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-loop-limit=@var{n} @gol
213 -fconstexpr-ops-limit=@var{n} -fno-elide-constructors @gol
214 -fno-enforce-eh-specs @gol
215 -fno-gnu-keywords @gol
216 -fno-implicit-templates @gol
217 -fno-implicit-inline-templates @gol
218 -fno-implement-inlines -fms-extensions @gol
219 -fnew-inheriting-ctors @gol
220 -fnew-ttp-matching @gol
221 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
222 -fno-optional-diags -fpermissive @gol
223 -fno-pretty-templates @gol
224 -frepo -fno-rtti -fsized-deallocation @gol
225 -ftemplate-backtrace-limit=@var{n} @gol
226 -ftemplate-depth=@var{n} @gol
227 -fno-threadsafe-statics -fuse-cxa-atexit @gol
228 -fno-weak -nostdinc++ @gol
229 -fvisibility-inlines-hidden @gol
230 -fvisibility-ms-compat @gol
231 -fext-numeric-literals @gol
232 -Wabi=@var{n} -Wabi-tag -Wconversion-null -Wctor-dtor-privacy @gol
233 -Wdelete-non-virtual-dtor -Wdeprecated-copy -Wdeprecated-copy-dtor @gol
234 -Wliteral-suffix @gol
235 -Wmultiple-inheritance -Wno-init-list-lifetime @gol
236 -Wnamespaces -Wnarrowing @gol
237 -Wpessimizing-move -Wredundant-move @gol
238 -Wnoexcept -Wnoexcept-type -Wclass-memaccess @gol
239 -Wnon-virtual-dtor -Wreorder -Wregister @gol
240 -Weffc++ -Wstrict-null-sentinel -Wtemplates @gol
241 -Wno-non-template-friend -Wold-style-cast @gol
242 -Woverloaded-virtual -Wno-pmf-conversions @gol
243 -Wno-class-conversion -Wno-terminate @gol
244 -Wsign-promo -Wvirtual-inheritance}
246 @item Objective-C and Objective-C++ Language Options
247 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
248 Objective-C and Objective-C++ Dialects}.
249 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
250 -fgnu-runtime -fnext-runtime @gol
251 -fno-nil-receivers @gol
252 -fobjc-abi-version=@var{n} @gol
253 -fobjc-call-cxx-cdtors @gol
254 -fobjc-direct-dispatch @gol
255 -fobjc-exceptions @gol
258 -fobjc-std=objc1 @gol
259 -fno-local-ivars @gol
260 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
261 -freplace-objc-classes @gol
264 -Wassign-intercept @gol
265 -Wno-protocol -Wselector @gol
266 -Wstrict-selector-match @gol
267 -Wundeclared-selector}
269 @item Diagnostic Message Formatting Options
270 @xref{Diagnostic Message Formatting Options,,Options to Control Diagnostic Messages Formatting}.
271 @gccoptlist{-fmessage-length=@var{n} @gol
272 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
273 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
274 -fdiagnostics-format=@r{[}text@r{|}json@r{]} @gol
275 -fno-diagnostics-show-option -fno-diagnostics-show-caret @gol
276 -fno-diagnostics-show-labels -fno-diagnostics-show-line-numbers @gol
277 -fdiagnostics-minimum-margin-width=@var{width} @gol
278 -fdiagnostics-parseable-fixits -fdiagnostics-generate-patch @gol
279 -fdiagnostics-show-template-tree -fno-elide-type @gol
282 @item Warning Options
283 @xref{Warning Options,,Options to Request or Suppress Warnings}.
284 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
285 -pedantic-errors @gol
286 -w -Wextra -Wall -Waddress -Waddress-of-packed-member @gol
287 -Waggregate-return -Waligned-new @gol
288 -Walloc-zero -Walloc-size-larger-than=@var{byte-size} @gol
289 -Walloca -Walloca-larger-than=@var{byte-size} @gol
290 -Wno-aggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
291 -Wno-attributes -Wattribute-alias=@var{n} @gol
292 -Wbool-compare -Wbool-operation @gol
293 -Wno-builtin-declaration-mismatch @gol
294 -Wno-builtin-macro-redefined -Wc90-c99-compat -Wc99-c11-compat @gol
295 -Wc++-compat -Wc++11-compat -Wc++14-compat -Wc++17-compat @gol
296 -Wcast-align -Wcast-align=strict -Wcast-function-type -Wcast-qual @gol
297 -Wchar-subscripts -Wcatch-value -Wcatch-value=@var{n} @gol
298 -Wclobbered -Wcomment -Wconditionally-supported @gol
299 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wdangling-else -Wdate-time @gol
300 -Wdelete-incomplete @gol
301 -Wno-attribute-warning @gol
302 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
303 -Wdisabled-optimization @gol
304 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
305 -Wno-div-by-zero -Wdouble-promotion @gol
306 -Wduplicated-branches -Wduplicated-cond @gol
307 -Wempty-body -Wenum-compare -Wno-endif-labels -Wexpansion-to-defined @gol
308 -Werror -Werror=* -Wextra-semi -Wfatal-errors @gol
309 -Wfloat-equal -Wformat -Wformat=2 @gol
310 -Wno-format-contains-nul -Wno-format-extra-args @gol
311 -Wformat-nonliteral -Wformat-overflow=@var{n} @gol
312 -Wformat-security -Wformat-signedness -Wformat-truncation=@var{n} @gol
313 -Wformat-y2k -Wframe-address @gol
314 -Wframe-larger-than=@var{byte-size} -Wno-free-nonheap-object @gol
315 -Wjump-misses-init @gol
316 -Whsa -Wif-not-aligned @gol
317 -Wignored-qualifiers -Wignored-attributes -Wincompatible-pointer-types @gol
318 -Wimplicit -Wimplicit-fallthrough -Wimplicit-fallthrough=@var{n} @gol
319 -Wimplicit-function-declaration -Wimplicit-int @gol
320 -Winit-self -Winline -Wno-int-conversion -Wint-in-bool-context @gol
321 -Wno-int-to-pointer-cast -Winvalid-memory-model -Wno-invalid-offsetof @gol
322 -Winvalid-pch -Wlarger-than=@var{byte-size} @gol
323 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
324 -Wmain -Wmaybe-uninitialized -Wmemset-elt-size -Wmemset-transposed-args @gol
325 -Wmisleading-indentation -Wmissing-attributes -Wmissing-braces @gol
326 -Wmissing-field-initializers -Wmissing-format-attribute @gol
327 -Wmissing-include-dirs -Wmissing-noreturn -Wmissing-profile @gol
328 -Wno-multichar -Wmultistatement-macros -Wnonnull -Wnonnull-compare @gol
329 -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
330 -Wnull-dereference -Wodr -Wno-overflow -Wopenmp-simd @gol
331 -Woverride-init-side-effects -Woverlength-strings @gol
332 -Wpacked -Wpacked-bitfield-compat -Wpacked-not-aligned -Wpadded @gol
333 -Wparentheses -Wno-pedantic-ms-format @gol
334 -Wplacement-new -Wplacement-new=@var{n} @gol
335 -Wpointer-arith -Wpointer-compare -Wno-pointer-to-int-cast @gol
336 -Wno-pragmas -Wno-prio-ctor-dtor -Wredundant-decls @gol
337 -Wrestrict -Wno-return-local-addr @gol
338 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
339 -Wshadow=global, -Wshadow=local, -Wshadow=compatible-local @gol
340 -Wshift-overflow -Wshift-overflow=@var{n} @gol
341 -Wshift-count-negative -Wshift-count-overflow -Wshift-negative-value @gol
342 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
343 -Wno-scalar-storage-order -Wsizeof-pointer-div @gol
344 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
345 -Wstack-protector -Wstack-usage=@var{byte-size} -Wstrict-aliasing @gol
346 -Wstrict-aliasing=n -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
347 -Wstringop-overflow=@var{n} -Wstringop-truncation -Wsubobject-linkage @gol
348 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{|}malloc@r{]} @gol
349 -Wsuggest-final-types @gol -Wsuggest-final-methods -Wsuggest-override @gol
350 -Wswitch -Wswitch-bool -Wswitch-default -Wswitch-enum @gol
351 -Wswitch-unreachable -Wsync-nand @gol
352 -Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs @gol
353 -Wtype-limits -Wundef @gol
354 -Wuninitialized -Wunknown-pragmas @gol
355 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
356 -Wunused-label -Wunused-local-typedefs -Wunused-macros @gol
357 -Wunused-parameter -Wno-unused-result @gol
358 -Wunused-value -Wunused-variable @gol
359 -Wunused-const-variable -Wunused-const-variable=@var{n} @gol
360 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
361 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
362 -Wvla -Wvla-larger-than=@var{byte-size} -Wvolatile-register-var @gol
364 -Wzero-as-null-pointer-constant}
366 @item C and Objective-C-only Warning Options
367 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
368 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
369 -Wold-style-declaration -Wold-style-definition @gol
370 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
371 -Wdeclaration-after-statement -Wpointer-sign}
373 @item Debugging Options
374 @xref{Debugging Options,,Options for Debugging Your Program}.
375 @gccoptlist{-g -g@var{level} -gdwarf -gdwarf-@var{version} @gol
376 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
377 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
378 -gas-loc-support -gno-as-loc-support @gol
379 -gas-locview-support -gno-as-locview-support @gol
380 -gcolumn-info -gno-column-info @gol
381 -gstatement-frontiers -gno-statement-frontiers @gol
382 -gvariable-location-views -gno-variable-location-views @gol
383 -ginternal-reset-location-views -gno-internal-reset-location-views @gol
384 -ginline-points -gno-inline-points @gol
385 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
386 -gsplit-dwarf -gdescribe-dies -gno-describe-dies @gol
387 -fdebug-prefix-map=@var{old}=@var{new} -fdebug-types-section @gol
388 -fno-eliminate-unused-debug-types @gol
389 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
390 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
391 -fno-eliminate-unused-debug-symbols -femit-class-debug-always @gol
392 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
393 -fvar-tracking -fvar-tracking-assignments}
395 @item Optimization Options
396 @xref{Optimize Options,,Options that Control Optimization}.
397 @gccoptlist{-faggressive-loop-optimizations @gol
398 -falign-functions[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
399 -falign-jumps[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
400 -falign-labels[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
401 -falign-loops[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
402 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
403 -fauto-inc-dec -fbranch-probabilities @gol
404 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
405 -fbtr-bb-exclusive -fcaller-saves @gol
406 -fcombine-stack-adjustments -fconserve-stack @gol
407 -fcompare-elim -fcprop-registers -fcrossjumping @gol
408 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
409 -fcx-limited-range @gol
410 -fdata-sections -fdce -fdelayed-branch @gol
411 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
412 -fdevirtualize-at-ltrans -fdse @gol
413 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
414 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
415 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
416 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
417 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
418 -fif-conversion2 -findirect-inlining @gol
419 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
420 -finline-small-functions -fipa-cp -fipa-cp-clone @gol
421 -fipa-bit-cp -fipa-vrp -fipa-pta -fipa-profile -fipa-pure-const @gol
422 -fipa-reference -fipa-reference-addressable @gol
423 -fipa-stack-alignment -fipa-icf -fira-algorithm=@var{algorithm} @gol
424 -flive-patching=@var{level} @gol
425 -fira-region=@var{region} -fira-hoist-pressure @gol
426 -fira-loop-pressure -fno-ira-share-save-slots @gol
427 -fno-ira-share-spill-slots @gol
428 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
429 -fivopts -fkeep-inline-functions -fkeep-static-functions @gol
430 -fkeep-static-consts -flimit-function-alignment -flive-range-shrinkage @gol
431 -floop-block -floop-interchange -floop-strip-mine @gol
432 -floop-unroll-and-jam -floop-nest-optimize @gol
433 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
434 -flto-partition=@var{alg} -fmerge-all-constants @gol
435 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
436 -fmove-loop-invariants -fno-branch-count-reg @gol
437 -fno-defer-pop -fno-fp-int-builtin-inexact -fno-function-cse @gol
438 -fno-guess-branch-probability -fno-inline -fno-math-errno -fno-peephole @gol
439 -fno-peephole2 -fno-printf-return-value -fno-sched-interblock @gol
440 -fno-sched-spec -fno-signed-zeros @gol
441 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
442 -fomit-frame-pointer -foptimize-sibling-calls @gol
443 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
444 -fprefetch-loop-arrays @gol
445 -fprofile-correction @gol
446 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
447 -fprofile-reorder-functions @gol
448 -freciprocal-math -free -frename-registers -freorder-blocks @gol
449 -freorder-blocks-algorithm=@var{algorithm} @gol
450 -freorder-blocks-and-partition -freorder-functions @gol
451 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
452 -frounding-math -fsave-optimization-record @gol
453 -fsched2-use-superblocks -fsched-pressure @gol
454 -fsched-spec-load -fsched-spec-load-dangerous @gol
455 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
456 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
457 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
458 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
459 -fschedule-fusion @gol
460 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
461 -fselective-scheduling -fselective-scheduling2 @gol
462 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
463 -fsemantic-interposition -fshrink-wrap -fshrink-wrap-separate @gol
464 -fsignaling-nans @gol
465 -fsingle-precision-constant -fsplit-ivs-in-unroller -fsplit-loops@gol
467 -fsplit-wide-types -fssa-backprop -fssa-phiopt @gol
468 -fstdarg-opt -fstore-merging -fstrict-aliasing @gol
469 -fthread-jumps -ftracer -ftree-bit-ccp @gol
470 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
471 -ftree-coalesce-vars -ftree-copy-prop -ftree-dce -ftree-dominator-opts @gol
472 -ftree-dse -ftree-forwprop -ftree-fre -fcode-hoisting @gol
473 -ftree-loop-if-convert -ftree-loop-im @gol
474 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
475 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
476 -ftree-loop-vectorize @gol
477 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
478 -ftree-reassoc -ftree-scev-cprop -ftree-sink -ftree-slsr -ftree-sra @gol
479 -ftree-switch-conversion -ftree-tail-merge @gol
480 -ftree-ter -ftree-vectorize -ftree-vrp -funconstrained-commons @gol
481 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
482 -funsafe-math-optimizations -funswitch-loops @gol
483 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
484 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
485 --param @var{name}=@var{value}
486 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
488 @item Program Instrumentation Options
489 @xref{Instrumentation Options,,Program Instrumentation Options}.
490 @gccoptlist{-p -pg -fprofile-arcs --coverage -ftest-coverage @gol
491 -fprofile-abs-path @gol
492 -fprofile-dir=@var{path} -fprofile-generate -fprofile-generate=@var{path} @gol
493 -fprofile-update=@var{method} -fprofile-filter-files=@var{regex} @gol
494 -fprofile-exclude-files=@var{regex} @gol
495 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
496 -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
497 -fsanitize-undefined-trap-on-error -fbounds-check @gol
498 -fcf-protection=@r{[}full@r{|}branch@r{|}return@r{|}none@r{]} @gol
499 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
500 -fstack-protector-explicit -fstack-check @gol
501 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
502 -fno-stack-limit -fsplit-stack @gol
503 -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} @gol
504 -fvtv-counts -fvtv-debug @gol
505 -finstrument-functions @gol
506 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
507 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}}
509 @item Preprocessor Options
510 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
511 @gccoptlist{-A@var{question}=@var{answer} @gol
512 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
513 -C -CC -D@var{macro}@r{[}=@var{defn}@r{]} @gol
514 -dD -dI -dM -dN -dU @gol
515 -fdebug-cpp -fdirectives-only -fdollars-in-identifiers @gol
516 -fexec-charset=@var{charset} -fextended-identifiers @gol
517 -finput-charset=@var{charset} -fmacro-prefix-map=@var{old}=@var{new} @gol
518 -fno-canonical-system-headers -fpch-deps -fpch-preprocess @gol
519 -fpreprocessed -ftabstop=@var{width} -ftrack-macro-expansion @gol
520 -fwide-exec-charset=@var{charset} -fworking-directory @gol
521 -H -imacros @var{file} -include @var{file} @gol
522 -M -MD -MF -MG -MM -MMD -MP -MQ -MT @gol
523 -no-integrated-cpp -P -pthread -remap @gol
524 -traditional -traditional-cpp -trigraphs @gol
525 -U@var{macro} -undef @gol
526 -Wp,@var{option} -Xpreprocessor @var{option}}
528 @item Assembler Options
529 @xref{Assembler Options,,Passing Options to the Assembler}.
530 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
533 @xref{Link Options,,Options for Linking}.
534 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
535 -nostartfiles -nodefaultlibs -nolibc -nostdlib @gol
536 -e @var{entry} --entry=@var{entry} @gol
537 -pie -pthread -r -rdynamic @gol
538 -s -static -static-pie -static-libgcc -static-libstdc++ @gol
539 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
540 -shared -shared-libgcc -symbolic @gol
541 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
542 -u @var{symbol} -z @var{keyword}}
544 @item Directory Options
545 @xref{Directory Options,,Options for Directory Search}.
546 @gccoptlist{-B@var{prefix} -I@var{dir} -I- @gol
547 -idirafter @var{dir} @gol
548 -imacros @var{file} -imultilib @var{dir} @gol
549 -iplugindir=@var{dir} -iprefix @var{file} @gol
550 -iquote @var{dir} -isysroot @var{dir} -isystem @var{dir} @gol
551 -iwithprefix @var{dir} -iwithprefixbefore @var{dir} @gol
552 -L@var{dir} -no-canonical-prefixes --no-sysroot-suffix @gol
553 -nostdinc -nostdinc++ --sysroot=@var{dir}}
555 @item Code Generation Options
556 @xref{Code Gen Options,,Options for Code Generation Conventions}.
557 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
558 -ffixed-@var{reg} -fexceptions @gol
559 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
560 -fasynchronous-unwind-tables @gol
562 -finhibit-size-directive -fno-common -fno-ident @gol
563 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol
564 -fno-jump-tables @gol
565 -frecord-gcc-switches @gol
566 -freg-struct-return -fshort-enums -fshort-wchar @gol
567 -fverbose-asm -fpack-struct[=@var{n}] @gol
568 -fleading-underscore -ftls-model=@var{model} @gol
569 -fstack-reuse=@var{reuse_level} @gol
570 -ftrampolines -ftrapv -fwrapv @gol
571 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
572 -fstrict-volatile-bitfields -fsync-libcalls}
574 @item Developer Options
575 @xref{Developer Options,,GCC Developer Options}.
576 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
577 -dumpfullversion -fchecking -fchecking=@var{n} -fdbg-cnt-list @gol
578 -fdbg-cnt=@var{counter-value-list} @gol
579 -fdisable-ipa-@var{pass_name} @gol
580 -fdisable-rtl-@var{pass_name} @gol
581 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
582 -fdisable-tree-@var{pass_name} @gol
583 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
584 -fdump-debug -fdump-earlydebug @gol
585 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
586 -fdump-final-insns@r{[}=@var{file}@r{]} @gol
587 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
589 -fdump-lang-@var{switch} @gol
590 -fdump-lang-@var{switch}-@var{options} @gol
591 -fdump-lang-@var{switch}-@var{options}=@var{filename} @gol
593 -fdump-rtl-@var{pass} -fdump-rtl-@var{pass}=@var{filename} @gol
594 -fdump-statistics @gol
596 -fdump-tree-@var{switch} @gol
597 -fdump-tree-@var{switch}-@var{options} @gol
598 -fdump-tree-@var{switch}-@var{options}=@var{filename} @gol
599 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
600 -fenable-@var{kind}-@var{pass} @gol
601 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
602 -fira-verbose=@var{n} @gol
603 -flto-report -flto-report-wpa -fmem-report-wpa @gol
604 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report @gol
605 -fopt-info -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
606 -fprofile-report @gol
607 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
608 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
609 -fstats -fstack-usage -ftime-report -ftime-report-details @gol
610 -fvar-tracking-assignments-toggle -gtoggle @gol
611 -print-file-name=@var{library} -print-libgcc-file-name @gol
612 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
613 -print-prog-name=@var{program} -print-search-dirs -Q @gol
614 -print-sysroot -print-sysroot-headers-suffix @gol
615 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
617 @item Machine-Dependent Options
618 @xref{Submodel Options,,Machine-Dependent Options}.
619 @c This list is ordered alphanumerically by subsection name.
620 @c Try and put the significant identifier (CPU or system) first,
621 @c so users have a clue at guessing where the ones they want will be.
623 @emph{AArch64 Options}
624 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
625 -mgeneral-regs-only @gol
626 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
627 -mstrict-align -mno-strict-align @gol
628 -momit-leaf-frame-pointer @gol
629 -mtls-dialect=desc -mtls-dialect=traditional @gol
630 -mtls-size=@var{size} @gol
631 -mfix-cortex-a53-835769 -mfix-cortex-a53-843419 @gol
632 -mlow-precision-recip-sqrt -mlow-precision-sqrt -mlow-precision-div @gol
633 -mpc-relative-literal-loads @gol
634 -msign-return-address=@var{scope} @gol
635 -mbranch-protection=@var{none}|@var{standard}|@var{pac-ret}[+@var{leaf}
636 +@var{b-key}]|@var{bti} @gol
637 -march=@var{name} -mcpu=@var{name} -mtune=@var{name} @gol
638 -moverride=@var{string} -mverbose-cost-dump @gol
639 -mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{sysreg} @gol
640 -mstack-protector-guard-offset=@var{offset} -mtrack-speculation }
642 @emph{Adapteva Epiphany Options}
643 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
644 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
645 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
646 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
647 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
648 -msplit-vecmove-early -m1reg-@var{reg}}
650 @emph{AMD GCN Options}
651 @gccoptlist{-march=@var{gpu} -mtune=@var{gpu} -mstack-size=@var{bytes}}
654 @gccoptlist{-mbarrel-shifter -mjli-always @gol
655 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
656 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
657 -mea -mno-mpy -mmul32x16 -mmul64 -matomic @gol
658 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
659 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
660 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
661 -mlong-calls -mmedium-calls -msdata -mirq-ctrl-saved @gol
662 -mrgf-banked-regs -mlpc-width=@var{width} -G @var{num} @gol
663 -mvolatile-cache -mtp-regno=@var{regno} @gol
664 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
665 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
666 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
667 -mlra-priority-compact mlra-priority-noncompact -mmillicode @gol
668 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
669 -mtune=@var{cpu} -mmultcost=@var{num} -mcode-density-frame @gol
670 -munalign-prob-threshold=@var{probability} -mmpy-option=@var{multo} @gol
671 -mdiv-rem -mcode-density -mll64 -mfpu=@var{fpu} -mrf16 -mbranch-index}
674 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
675 -mabi=@var{name} @gol
676 -mapcs-stack-check -mno-apcs-stack-check @gol
677 -mapcs-reentrant -mno-apcs-reentrant @gol
678 -mgeneral-regs-only @gol
679 -msched-prolog -mno-sched-prolog @gol
680 -mlittle-endian -mbig-endian @gol
682 -mfloat-abi=@var{name} @gol
683 -mfp16-format=@var{name}
684 -mthumb-interwork -mno-thumb-interwork @gol
685 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
686 -mtune=@var{name} -mprint-tune-info @gol
687 -mstructure-size-boundary=@var{n} @gol
688 -mabort-on-noreturn @gol
689 -mlong-calls -mno-long-calls @gol
690 -msingle-pic-base -mno-single-pic-base @gol
691 -mpic-register=@var{reg} @gol
692 -mnop-fun-dllimport @gol
693 -mpoke-function-name @gol
694 -mthumb -marm -mflip-thumb @gol
695 -mtpcs-frame -mtpcs-leaf-frame @gol
696 -mcaller-super-interworking -mcallee-super-interworking @gol
697 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
698 -mword-relocations @gol
699 -mfix-cortex-m3-ldrd @gol
700 -munaligned-access @gol
701 -mneon-for-64bits @gol
702 -mslow-flash-data @gol
703 -masm-syntax-unified @gol
705 -mverbose-cost-dump @gol
710 @gccoptlist{-mmcu=@var{mcu} -mabsdata -maccumulate-args @gol
711 -mbranch-cost=@var{cost} @gol
712 -mcall-prologues -mgas-isr-prologues -mint8 @gol
713 -mn_flash=@var{size} -mno-interrupts @gol
714 -mmain-is-OS_task -mrelax -mrmw -mstrict-X -mtiny-stack @gol
715 -mfract-convert-truncate @gol
716 -mshort-calls -nodevicelib @gol
717 -Waddr-space-convert -Wmisspelled-isr}
719 @emph{Blackfin Options}
720 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
721 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
722 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
723 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
724 -mno-id-shared-library -mshared-library-id=@var{n} @gol
725 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
726 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
727 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
731 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
732 -msim -msdata=@var{sdata-type}}
735 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
736 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
737 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
738 -mstack-align -mdata-align -mconst-align @gol
739 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
740 -melf -maout -melinux -mlinux -sim -sim2 @gol
741 -mmul-bug-workaround -mno-mul-bug-workaround}
744 @gccoptlist{-mmac @gol
745 -mcr16cplus -mcr16c @gol
746 -msim -mint32 -mbit-ops
747 -mdata-model=@var{model}}
750 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} @gol
751 -mbig-endian -EB -mlittle-endian -EL @gol
752 -mhard-float -msoft-float -mfpu=@var{fpu} -mdouble-float -mfdivdu @gol
753 -melrw -mistack -mmp -mcp -mcache -msecurity -mtrust @gol
754 -mdsp -medsp -mvdsp @gol
755 -mdiv -msmart -mhigh-registers -manchor @gol
756 -mpushpop -mmultiple-stld -mconstpool -mstack-size -mccrt @gol
757 -mbranch-cost=@var{n} -mcse-cc -msched-prolog}
759 @emph{Darwin Options}
760 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
761 -arch_only -bind_at_load -bundle -bundle_loader @gol
762 -client_name -compatibility_version -current_version @gol
764 -dependency-file -dylib_file -dylinker_install_name @gol
765 -dynamic -dynamiclib -exported_symbols_list @gol
766 -filelist -flat_namespace -force_cpusubtype_ALL @gol
767 -force_flat_namespace -headerpad_max_install_names @gol
769 -image_base -init -install_name -keep_private_externs @gol
770 -multi_module -multiply_defined -multiply_defined_unused @gol
771 -noall_load -no_dead_strip_inits_and_terms @gol
772 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
773 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
774 -private_bundle -read_only_relocs -sectalign @gol
775 -sectobjectsymbols -whyload -seg1addr @gol
776 -sectcreate -sectobjectsymbols -sectorder @gol
777 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
778 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
779 -segprot -segs_read_only_addr -segs_read_write_addr @gol
780 -single_module -static -sub_library -sub_umbrella @gol
781 -twolevel_namespace -umbrella -undefined @gol
782 -unexported_symbols_list -weak_reference_mismatches @gol
783 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
784 -mkernel -mone-byte-bool}
786 @emph{DEC Alpha Options}
787 @gccoptlist{-mno-fp-regs -msoft-float @gol
788 -mieee -mieee-with-inexact -mieee-conformant @gol
789 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
790 -mtrap-precision=@var{mode} -mbuild-constants @gol
791 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
792 -mbwx -mmax -mfix -mcix @gol
793 -mfloat-vax -mfloat-ieee @gol
794 -mexplicit-relocs -msmall-data -mlarge-data @gol
795 -msmall-text -mlarge-text @gol
796 -mmemory-latency=@var{time}}
799 @gccoptlist{-msmall-model -mno-lsim}
802 @gccoptlist{-msim -mlra -mnodiv -mft32b -mcompress -mnopm}
805 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
806 -mhard-float -msoft-float @gol
807 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
808 -mdouble -mno-double @gol
809 -mmedia -mno-media -mmuladd -mno-muladd @gol
810 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
811 -mlinked-fp -mlong-calls -malign-labels @gol
812 -mlibrary-pic -macc-4 -macc-8 @gol
813 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
814 -moptimize-membar -mno-optimize-membar @gol
815 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
816 -mvliw-branch -mno-vliw-branch @gol
817 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
818 -mno-nested-cond-exec -mtomcat-stats @gol
822 @emph{GNU/Linux Options}
823 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
824 -tno-android-cc -tno-android-ld}
826 @emph{H8/300 Options}
827 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
830 @gccoptlist{-march=@var{architecture-type} @gol
831 -mcaller-copies -mdisable-fpregs -mdisable-indexing @gol
832 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
833 -mfixed-range=@var{register-range} @gol
834 -mjump-in-delay -mlinker-opt -mlong-calls @gol
835 -mlong-load-store -mno-disable-fpregs @gol
836 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
837 -mno-jump-in-delay -mno-long-load-store @gol
838 -mno-portable-runtime -mno-soft-float @gol
839 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
840 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
841 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
842 -munix=@var{unix-std} -nolibdld -static -threads}
845 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
846 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
847 -mconstant-gp -mauto-pic -mfused-madd @gol
848 -minline-float-divide-min-latency @gol
849 -minline-float-divide-max-throughput @gol
850 -mno-inline-float-divide @gol
851 -minline-int-divide-min-latency @gol
852 -minline-int-divide-max-throughput @gol
853 -mno-inline-int-divide @gol
854 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
855 -mno-inline-sqrt @gol
856 -mdwarf2-asm -mearly-stop-bits @gol
857 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
858 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
859 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
860 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
861 -msched-spec-ldc -msched-spec-control-ldc @gol
862 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
863 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
864 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
865 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
868 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
869 -msign-extend-enabled -muser-enabled}
871 @emph{M32R/D Options}
872 @gccoptlist{-m32r2 -m32rx -m32r @gol
874 -malign-loops -mno-align-loops @gol
875 -missue-rate=@var{number} @gol
876 -mbranch-cost=@var{number} @gol
877 -mmodel=@var{code-size-model-type} @gol
878 -msdata=@var{sdata-type} @gol
879 -mno-flush-func -mflush-func=@var{name} @gol
880 -mno-flush-trap -mflush-trap=@var{number} @gol
884 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
886 @emph{M680x0 Options}
887 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
888 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
889 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
890 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
891 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
892 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
893 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
894 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
895 -mxgot -mno-xgot -mlong-jump-table-offsets}
898 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
899 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
900 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
901 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
902 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
905 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
906 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
907 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
908 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
911 @emph{MicroBlaze Options}
912 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
913 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
914 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
915 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
916 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model} @gol
917 -mpic-data-is-text-relative}
920 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
921 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
922 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
923 -mips16 -mno-mips16 -mflip-mips16 @gol
924 -minterlink-compressed -mno-interlink-compressed @gol
925 -minterlink-mips16 -mno-interlink-mips16 @gol
926 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
927 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
928 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
929 -mno-float -msingle-float -mdouble-float @gol
930 -modd-spreg -mno-odd-spreg @gol
931 -mabs=@var{mode} -mnan=@var{encoding} @gol
932 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
935 -mvirt -mno-virt @gol
938 -mginv -mno-ginv @gol
939 -mmicromips -mno-micromips @gol
941 -mloongson-mmi -mno-loongson-mmi @gol
942 -mloongson-ext -mno-loongson-ext @gol
943 -mloongson-ext2 -mno-loongson-ext2 @gol
944 -mfpu=@var{fpu-type} @gol
945 -msmartmips -mno-smartmips @gol
946 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
947 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
948 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
949 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
950 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
951 -membedded-data -mno-embedded-data @gol
952 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
953 -mcode-readable=@var{setting} @gol
954 -msplit-addresses -mno-split-addresses @gol
955 -mexplicit-relocs -mno-explicit-relocs @gol
956 -mcheck-zero-division -mno-check-zero-division @gol
957 -mdivide-traps -mdivide-breaks @gol
958 -mload-store-pairs -mno-load-store-pairs @gol
959 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
960 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
961 -mfix-24k -mno-fix-24k @gol
962 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
963 -mfix-r5900 -mno-fix-r5900 @gol
964 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
965 -mfix-vr4120 -mno-fix-vr4120 @gol
966 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
967 -mflush-func=@var{func} -mno-flush-func @gol
968 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
969 -mcompact-branches=@var{policy} @gol
970 -mfp-exceptions -mno-fp-exceptions @gol
971 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
972 -mlxc1-sxc1 -mno-lxc1-sxc1 -mmadd4 -mno-madd4 @gol
973 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address @gol
974 -mframe-header-opt -mno-frame-header-opt}
977 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
978 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
979 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
980 -mno-base-addresses -msingle-exit -mno-single-exit}
982 @emph{MN10300 Options}
983 @gccoptlist{-mmult-bug -mno-mult-bug @gol
984 -mno-am33 -mam33 -mam33-2 -mam34 @gol
985 -mtune=@var{cpu-type} @gol
986 -mreturn-pointer-on-d0 @gol
987 -mno-crt0 -mrelax -mliw -msetlb}
990 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
992 @emph{MSP430 Options}
993 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
995 -mcode-region= -mdata-region= @gol
996 -msilicon-errata= -msilicon-errata-warn= @gol
1000 @gccoptlist{-mbig-endian -mlittle-endian @gol
1001 -mreduced-regs -mfull-regs @gol
1002 -mcmov -mno-cmov @gol
1003 -mext-perf -mno-ext-perf @gol
1004 -mext-perf2 -mno-ext-perf2 @gol
1005 -mext-string -mno-ext-string @gol
1006 -mv3push -mno-v3push @gol
1007 -m16bit -mno-16bit @gol
1008 -misr-vector-size=@var{num} @gol
1009 -mcache-block-size=@var{num} @gol
1010 -march=@var{arch} @gol
1011 -mcmodel=@var{code-model} @gol
1012 -mctor-dtor -mrelax}
1014 @emph{Nios II Options}
1015 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
1016 -mgprel-sec=@var{regexp} -mr0rel-sec=@var{regexp} @gol
1018 -mno-bypass-cache -mbypass-cache @gol
1019 -mno-cache-volatile -mcache-volatile @gol
1020 -mno-fast-sw-div -mfast-sw-div @gol
1021 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
1022 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
1023 -mcustom-fpu-cfg=@var{name} @gol
1024 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
1025 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
1027 @emph{Nvidia PTX Options}
1028 @gccoptlist{-m32 -m64 -mmainkernel -moptimize}
1030 @emph{OpenRISC Options}
1031 @gccoptlist{-mboard=@var{name} -mnewlib -mhard-mul -mhard-div @gol
1032 -msoft-mul -msoft-div @gol
1033 -mcmov -mror -msext -msfimm -mshftimm}
1035 @emph{PDP-11 Options}
1036 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
1037 -mint32 -mno-int16 -mint16 -mno-int32 @gol
1038 -msplit -munix-asm -mdec-asm -mgnu-asm -mlra}
1040 @emph{picoChip Options}
1041 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
1042 -msymbol-as-address -mno-inefficient-warnings}
1044 @emph{PowerPC Options}
1045 See RS/6000 and PowerPC Options.
1047 @emph{RISC-V Options}
1048 @gccoptlist{-mbranch-cost=@var{N-instruction} @gol
1050 -mabi=@var{ABI-string} @gol
1051 -mfdiv -mno-fdiv @gol
1053 -march=@var{ISA-string} @gol
1054 -mtune=@var{processor-string} @gol
1055 -mpreferred-stack-boundary=@var{num} @gol
1056 -msmall-data-limit=@var{N-bytes} @gol
1057 -msave-restore -mno-save-restore @gol
1058 -mstrict-align -mno-strict-align @gol
1059 -mcmodel=medlow -mcmodel=medany @gol
1060 -mexplicit-relocs -mno-explicit-relocs @gol
1061 -mrelax -mno-relax @gol
1062 -mriscv-attribute -mmo-riscv-attribute}
1065 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
1066 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
1067 -m64bit-doubles -m32bit-doubles -msave-mduc-in-interrupts}
1069 @emph{RS/6000 and PowerPC Options}
1070 @gccoptlist{-mcpu=@var{cpu-type} @gol
1071 -mtune=@var{cpu-type} @gol
1072 -mcmodel=@var{code-model} @gol
1074 -maltivec -mno-altivec @gol
1075 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
1076 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
1077 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
1078 -mfprnd -mno-fprnd @gol
1079 -mcmpb -mno-cmpb -mhard-dfp -mno-hard-dfp @gol
1080 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
1081 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
1082 -malign-power -malign-natural @gol
1083 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
1084 -mupdate -mno-update @gol
1085 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
1086 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
1087 -mstrict-align -mno-strict-align -mrelocatable @gol
1088 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
1089 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
1090 -mdynamic-no-pic -mswdiv -msingle-pic-base @gol
1091 -mprioritize-restricted-insns=@var{priority} @gol
1092 -msched-costly-dep=@var{dependence_type} @gol
1093 -minsert-sched-nops=@var{scheme} @gol
1094 -mcall-aixdesc -mcall-eabi -mcall-freebsd @gol
1095 -mcall-linux -mcall-netbsd -mcall-openbsd @gol
1096 -mcall-sysv -mcall-sysv-eabi -mcall-sysv-noeabi @gol
1097 -mtraceback=@var{traceback_type} @gol
1098 -maix-struct-return -msvr4-struct-return @gol
1099 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
1100 -mlongcall -mno-longcall -mpltseq -mno-pltseq @gol
1101 -mblock-move-inline-limit=@var{num} @gol
1102 -mblock-compare-inline-limit=@var{num} @gol
1103 -mblock-compare-inline-loop-limit=@var{num} @gol
1104 -mstring-compare-inline-limit=@var{num} @gol
1105 -misel -mno-isel @gol
1106 -mvrsave -mno-vrsave @gol
1107 -mmulhw -mno-mulhw @gol
1108 -mdlmzb -mno-dlmzb @gol
1109 -mprototype -mno-prototype @gol
1110 -msim -mmvme -mads -myellowknife -memb -msdata @gol
1111 -msdata=@var{opt} -mreadonly-in-sdata -mvxworks -G @var{num} @gol
1112 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
1113 -mno-recip-precision @gol
1114 -mveclibabi=@var{type} -mfriz -mno-friz @gol
1115 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
1116 -msave-toc-indirect -mno-save-toc-indirect @gol
1117 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
1118 -mcrypto -mno-crypto -mhtm -mno-htm @gol
1119 -mquad-memory -mno-quad-memory @gol
1120 -mquad-memory-atomic -mno-quad-memory-atomic @gol
1121 -mcompat-align-parm -mno-compat-align-parm @gol
1122 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol
1123 -mgnu-attribute -mno-gnu-attribute @gol
1124 -mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{reg} @gol
1125 -mstack-protector-guard-offset=@var{offset} -mpcrel -mno-pcrel}
1128 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
1130 -mbig-endian-data -mlittle-endian-data @gol
1133 -mas100-syntax -mno-as100-syntax@gol
1135 -mmax-constant-size=@gol
1138 -mallow-string-insns -mno-allow-string-insns@gol
1140 -mno-warn-multiple-fast-interrupts@gol
1141 -msave-acc-in-interrupts}
1143 @emph{S/390 and zSeries Options}
1144 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1145 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
1146 -mlong-double-64 -mlong-double-128 @gol
1147 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
1148 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
1149 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
1150 -mhtm -mvx -mzvector @gol
1151 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
1152 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
1153 -mhotpatch=@var{halfwords},@var{halfwords}}
1155 @emph{Score Options}
1156 @gccoptlist{-meb -mel @gol
1160 -mscore5 -mscore5u -mscore7 -mscore7d}
1163 @gccoptlist{-m1 -m2 -m2e @gol
1164 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
1166 -m4-nofpu -m4-single-only -m4-single -m4 @gol
1167 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
1168 -mb -ml -mdalign -mrelax @gol
1169 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
1170 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
1171 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1172 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1173 -maccumulate-outgoing-args @gol
1174 -matomic-model=@var{atomic-model} @gol
1175 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1176 -mcbranch-force-delay-slot @gol
1177 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1178 -mpretend-cmove -mtas}
1180 @emph{Solaris 2 Options}
1181 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1184 @emph{SPARC Options}
1185 @gccoptlist{-mcpu=@var{cpu-type} @gol
1186 -mtune=@var{cpu-type} @gol
1187 -mcmodel=@var{code-model} @gol
1188 -mmemory-model=@var{mem-model} @gol
1189 -m32 -m64 -mapp-regs -mno-app-regs @gol
1190 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1191 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1192 -mhard-quad-float -msoft-quad-float @gol
1193 -mstack-bias -mno-stack-bias @gol
1194 -mstd-struct-return -mno-std-struct-return @gol
1195 -munaligned-doubles -mno-unaligned-doubles @gol
1196 -muser-mode -mno-user-mode @gol
1197 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1198 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1199 -mvis4 -mno-vis4 -mvis4b -mno-vis4b @gol
1200 -mcbcond -mno-cbcond -mfmaf -mno-fmaf -mfsmuld -mno-fsmuld @gol
1201 -mpopc -mno-popc -msubxc -mno-subxc @gol
1202 -mfix-at697f -mfix-ut699 -mfix-ut700 -mfix-gr712rc @gol
1206 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1207 -msafe-dma -munsafe-dma @gol
1209 -msmall-mem -mlarge-mem -mstdmain @gol
1210 -mfixed-range=@var{register-range} @gol
1212 -maddress-space-conversion -mno-address-space-conversion @gol
1213 -mcache-size=@var{cache-size} @gol
1214 -matomic-updates -mno-atomic-updates}
1216 @emph{System V Options}
1217 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1219 @emph{TILE-Gx Options}
1220 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1221 -mcmodel=@var{code-model}}
1223 @emph{TILEPro Options}
1224 @gccoptlist{-mcpu=@var{cpu} -m32}
1227 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1228 -mprolog-function -mno-prolog-function -mspace @gol
1229 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1230 -mapp-regs -mno-app-regs @gol
1231 -mdisable-callt -mno-disable-callt @gol
1232 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1233 -mv850e -mv850 -mv850e3v5 @gol
1244 @gccoptlist{-mg -mgnu -munix}
1246 @emph{Visium Options}
1247 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1248 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1251 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1252 -mpointer-size=@var{size}}
1254 @emph{VxWorks Options}
1255 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1256 -Xbind-lazy -Xbind-now}
1259 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1260 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1261 -mfpmath=@var{unit} @gol
1262 -masm=@var{dialect} -mno-fancy-math-387 @gol
1263 -mno-fp-ret-in-387 -m80387 -mhard-float -msoft-float @gol
1264 -mno-wide-multiply -mrtd -malign-double @gol
1265 -mpreferred-stack-boundary=@var{num} @gol
1266 -mincoming-stack-boundary=@var{num} @gol
1267 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1268 -mrecip -mrecip=@var{opt} @gol
1269 -mvzeroupper -mprefer-avx128 -mprefer-vector-width=@var{opt} @gol
1270 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1271 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1272 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1273 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma -mpconfig -mwbnoinvd @gol
1274 -mptwrite -mprefetchwt1 -mclflushopt -mclwb -mxsavec -mxsaves @gol
1275 -msse4a -m3dnow -m3dnowa -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop @gol
1276 -madx -mlzcnt -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mhle -mlwp @gol
1277 -mmwaitx -mclzero -mpku -mthreads -mgfni -mvaes -mwaitpkg @gol
1278 -mshstk -mmanual-endbr -mforce-indirect-call -mavx512vbmi2 -mavx512bf16 -menqcmd @gol
1279 -mvpclmulqdq -mavx512bitalg -mmovdiri -mmovdir64b -mavx512vpopcntdq @gol
1280 -mavx5124fmaps -mavx512vnni -mavx5124vnniw -mprfchw -mrdpid @gol
1282 -mcldemote -mms-bitfields -mno-align-stringops -minline-all-stringops @gol
1283 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1284 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1285 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1286 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1287 -mregparm=@var{num} -msseregparm @gol
1288 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1289 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1290 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1291 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1292 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1293 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1294 -minstrument-return=@var{type} -mfentry-name=@var{name} -mfentry-section=@var{name} @gol
1295 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1296 -malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol
1297 -mstack-protector-guard-reg=@var{reg} @gol
1298 -mstack-protector-guard-offset=@var{offset} @gol
1299 -mstack-protector-guard-symbol=@var{symbol} @gol
1300 -mgeneral-regs-only -mcall-ms2sysv-xlogues @gol
1301 -mindirect-branch=@var{choice} -mfunction-return=@var{choice} @gol
1302 -mindirect-branch-register}
1304 @emph{x86 Windows Options}
1305 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1306 -mnop-fun-dllimport -mthread @gol
1307 -municode -mwin32 -mwindows -fno-set-stack-executable}
1309 @emph{Xstormy16 Options}
1312 @emph{Xtensa Options}
1313 @gccoptlist{-mconst16 -mno-const16 @gol
1314 -mfused-madd -mno-fused-madd @gol
1316 -mserialize-volatile -mno-serialize-volatile @gol
1317 -mtext-section-literals -mno-text-section-literals @gol
1318 -mauto-litpools -mno-auto-litpools @gol
1319 -mtarget-align -mno-target-align @gol
1320 -mlongcalls -mno-longcalls}
1322 @emph{zSeries Options}
1323 See S/390 and zSeries Options.
1327 @node Overall Options
1328 @section Options Controlling the Kind of Output
1330 Compilation can involve up to four stages: preprocessing, compilation
1331 proper, assembly and linking, always in that order. GCC is capable of
1332 preprocessing and compiling several files either into several
1333 assembler input files, or into one assembler input file; then each
1334 assembler input file produces an object file, and linking combines all
1335 the object files (those newly compiled, and those specified as input)
1336 into an executable file.
1338 @cindex file name suffix
1339 For any given input file, the file name suffix determines what kind of
1340 compilation is done:
1344 C source code that must be preprocessed.
1347 C source code that should not be preprocessed.
1350 C++ source code that should not be preprocessed.
1353 Objective-C source code. Note that you must link with the @file{libobjc}
1354 library to make an Objective-C program work.
1357 Objective-C source code that should not be preprocessed.
1361 Objective-C++ source code. Note that you must link with the @file{libobjc}
1362 library to make an Objective-C++ program work. Note that @samp{.M} refers
1363 to a literal capital M@.
1365 @item @var{file}.mii
1366 Objective-C++ source code that should not be preprocessed.
1369 C, C++, Objective-C or Objective-C++ header file to be turned into a
1370 precompiled header (default), or C, C++ header file to be turned into an
1371 Ada spec (via the @option{-fdump-ada-spec} switch).
1374 @itemx @var{file}.cp
1375 @itemx @var{file}.cxx
1376 @itemx @var{file}.cpp
1377 @itemx @var{file}.CPP
1378 @itemx @var{file}.c++
1380 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1381 the last two letters must both be literally @samp{x}. Likewise,
1382 @samp{.C} refers to a literal capital C@.
1386 Objective-C++ source code that must be preprocessed.
1388 @item @var{file}.mii
1389 Objective-C++ source code that should not be preprocessed.
1393 @itemx @var{file}.hp
1394 @itemx @var{file}.hxx
1395 @itemx @var{file}.hpp
1396 @itemx @var{file}.HPP
1397 @itemx @var{file}.h++
1398 @itemx @var{file}.tcc
1399 C++ header file to be turned into a precompiled header or Ada spec.
1402 @itemx @var{file}.for
1403 @itemx @var{file}.ftn
1404 Fixed form Fortran source code that should not be preprocessed.
1407 @itemx @var{file}.FOR
1408 @itemx @var{file}.fpp
1409 @itemx @var{file}.FPP
1410 @itemx @var{file}.FTN
1411 Fixed form Fortran source code that must be preprocessed (with the traditional
1414 @item @var{file}.f90
1415 @itemx @var{file}.f95
1416 @itemx @var{file}.f03
1417 @itemx @var{file}.f08
1418 Free form Fortran source code that should not be preprocessed.
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 must be preprocessed (with the
1425 traditional preprocessor).
1430 @item @var{file}.brig
1431 BRIG files (binary representation of HSAIL).
1440 D documentation code (Ddoc).
1442 @item @var{file}.ads
1443 Ada source code file that contains a library unit declaration (a
1444 declaration of a package, subprogram, or generic, or a generic
1445 instantiation), or a library unit renaming declaration (a package,
1446 generic, or subprogram renaming declaration). Such files are also
1449 @item @var{file}.adb
1450 Ada source code file containing a library unit body (a subprogram or
1451 package body). Such files are also called @dfn{bodies}.
1453 @c GCC also knows about some suffixes for languages not yet included:
1461 @itemx @var{file}.sx
1462 Assembler code that must be preprocessed.
1465 An object file to be fed straight into linking.
1466 Any file name with no recognized suffix is treated this way.
1470 You can specify the input language explicitly with the @option{-x} option:
1473 @item -x @var{language}
1474 Specify explicitly the @var{language} for the following input files
1475 (rather than letting the compiler choose a default based on the file
1476 name suffix). This option applies to all following input files until
1477 the next @option{-x} option. Possible values for @var{language} are:
1479 c c-header cpp-output
1480 c++ c++-header c++-cpp-output
1481 objective-c objective-c-header objective-c-cpp-output
1482 objective-c++ objective-c++-header objective-c++-cpp-output
1483 assembler assembler-with-cpp
1486 f77 f77-cpp-input f95 f95-cpp-input
1492 Turn off any specification of a language, so that subsequent files are
1493 handled according to their file name suffixes (as they are if @option{-x}
1494 has not been used at all).
1497 If you only want some of the stages of compilation, you can use
1498 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1499 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1500 @command{gcc} is to stop. Note that some combinations (for example,
1501 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1506 Compile or assemble the source files, but do not link. The linking
1507 stage simply is not done. The ultimate output is in the form of an
1508 object file for each source file.
1510 By default, the object file name for a source file is made by replacing
1511 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1513 Unrecognized input files, not requiring compilation or assembly, are
1518 Stop after the stage of compilation proper; do not assemble. The output
1519 is in the form of an assembler code file for each non-assembler input
1522 By default, the assembler file name for a source file is made by
1523 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1525 Input files that don't require compilation are ignored.
1529 Stop after the preprocessing stage; do not run the compiler proper. The
1530 output is in the form of preprocessed source code, which is sent to the
1533 Input files that don't require preprocessing are ignored.
1535 @cindex output file option
1538 Place output in file @var{file}. This applies to whatever
1539 sort of output is being produced, whether it be an executable file,
1540 an object file, an assembler file or preprocessed C code.
1542 If @option{-o} is not specified, the default is to put an executable
1543 file in @file{a.out}, the object file for
1544 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1545 assembler file in @file{@var{source}.s}, a precompiled header file in
1546 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1551 Print (on standard error output) the commands executed to run the stages
1552 of compilation. Also print the version number of the compiler driver
1553 program and of the preprocessor and the compiler proper.
1557 Like @option{-v} except the commands are not executed and arguments
1558 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1559 This is useful for shell scripts to capture the driver-generated command lines.
1563 Print (on the standard output) a description of the command-line options
1564 understood by @command{gcc}. If the @option{-v} option is also specified
1565 then @option{--help} is also passed on to the various processes
1566 invoked by @command{gcc}, so that they can display the command-line options
1567 they accept. If the @option{-Wextra} option has also been specified
1568 (prior to the @option{--help} option), then command-line options that
1569 have no documentation associated with them are also displayed.
1572 @opindex target-help
1573 Print (on the standard output) a description of target-specific command-line
1574 options for each tool. For some targets extra target-specific
1575 information may also be printed.
1577 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1578 Print (on the standard output) a description of the command-line
1579 options understood by the compiler that fit into all specified classes
1580 and qualifiers. These are the supported classes:
1583 @item @samp{optimizers}
1584 Display all of the optimization options supported by the
1587 @item @samp{warnings}
1588 Display all of the options controlling warning messages
1589 produced by the compiler.
1592 Display target-specific options. Unlike the
1593 @option{--target-help} option however, target-specific options of the
1594 linker and assembler are not displayed. This is because those
1595 tools do not currently support the extended @option{--help=} syntax.
1598 Display the values recognized by the @option{--param}
1601 @item @var{language}
1602 Display the options supported for @var{language}, where
1603 @var{language} is the name of one of the languages supported in this
1607 Display the options that are common to all languages.
1610 These are the supported qualifiers:
1613 @item @samp{undocumented}
1614 Display only those options that are undocumented.
1617 Display options taking an argument that appears after an equal
1618 sign in the same continuous piece of text, such as:
1619 @samp{--help=target}.
1621 @item @samp{separate}
1622 Display options taking an argument that appears as a separate word
1623 following the original option, such as: @samp{-o output-file}.
1626 Thus for example to display all the undocumented target-specific
1627 switches supported by the compiler, use:
1630 --help=target,undocumented
1633 The sense of a qualifier can be inverted by prefixing it with the
1634 @samp{^} character, so for example to display all binary warning
1635 options (i.e., ones that are either on or off and that do not take an
1636 argument) that have a description, use:
1639 --help=warnings,^joined,^undocumented
1642 The argument to @option{--help=} should not consist solely of inverted
1645 Combining several classes is possible, although this usually
1646 restricts the output so much that there is nothing to display. One
1647 case where it does work, however, is when one of the classes is
1648 @var{target}. For example, to display all the target-specific
1649 optimization options, use:
1652 --help=target,optimizers
1655 The @option{--help=} option can be repeated on the command line. Each
1656 successive use displays its requested class of options, skipping
1657 those that have already been displayed. If @option{--help} is also
1658 specified anywhere on the command line then this takes precedence
1659 over any @option{--help=} option.
1661 If the @option{-Q} option appears on the command line before the
1662 @option{--help=} option, then the descriptive text displayed by
1663 @option{--help=} is changed. Instead of describing the displayed
1664 options, an indication is given as to whether the option is enabled,
1665 disabled or set to a specific value (assuming that the compiler
1666 knows this at the point where the @option{--help=} option is used).
1668 Here is a truncated example from the ARM port of @command{gcc}:
1671 % gcc -Q -mabi=2 --help=target -c
1672 The following options are target specific:
1674 -mabort-on-noreturn [disabled]
1678 The output is sensitive to the effects of previous command-line
1679 options, so for example it is possible to find out which optimizations
1680 are enabled at @option{-O2} by using:
1683 -Q -O2 --help=optimizers
1686 Alternatively you can discover which binary optimizations are enabled
1687 by @option{-O3} by using:
1690 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1691 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1692 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1697 Display the version number and copyrights of the invoked GCC@.
1699 @item -pass-exit-codes
1700 @opindex pass-exit-codes
1701 Normally the @command{gcc} program exits with the code of 1 if any
1702 phase of the compiler returns a non-success return code. If you specify
1703 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1704 the numerically highest error produced by any phase returning an error
1705 indication. The C, C++, and Fortran front ends return 4 if an internal
1706 compiler error is encountered.
1710 Use pipes rather than temporary files for communication between the
1711 various stages of compilation. This fails to work on some systems where
1712 the assembler is unable to read from a pipe; but the GNU assembler has
1715 @item -specs=@var{file}
1717 Process @var{file} after the compiler reads in the standard @file{specs}
1718 file, in order to override the defaults which the @command{gcc} driver
1719 program uses when determining what switches to pass to @command{cc1},
1720 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
1721 @option{-specs=@var{file}} can be specified on the command line, and they
1722 are processed in order, from left to right. @xref{Spec Files}, for
1723 information about the format of the @var{file}.
1727 Invoke all subcommands under a wrapper program. The name of the
1728 wrapper program and its parameters are passed as a comma separated
1732 gcc -c t.c -wrapper gdb,--args
1736 This invokes all subprograms of @command{gcc} under
1737 @samp{gdb --args}, thus the invocation of @command{cc1} is
1738 @samp{gdb --args cc1 @dots{}}.
1740 @item -ffile-prefix-map=@var{old}=@var{new}
1741 @opindex ffile-prefix-map
1742 When compiling files residing in directory @file{@var{old}}, record
1743 any references to them in the result of the compilation as if the
1744 files resided in directory @file{@var{new}} instead. Specifying this
1745 option is equivalent to specifying all the individual
1746 @option{-f*-prefix-map} options. This can be used to make reproducible
1747 builds that are location independent. See also
1748 @option{-fmacro-prefix-map} and @option{-fdebug-prefix-map}.
1750 @item -fplugin=@var{name}.so
1752 Load the plugin code in file @var{name}.so, assumed to be a
1753 shared object to be dlopen'd by the compiler. The base name of
1754 the shared object file is used to identify the plugin for the
1755 purposes of argument parsing (See
1756 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1757 Each plugin should define the callback functions specified in the
1760 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1761 @opindex fplugin-arg
1762 Define an argument called @var{key} with a value of @var{value}
1763 for the plugin called @var{name}.
1765 @item -fdump-ada-spec@r{[}-slim@r{]}
1766 @opindex fdump-ada-spec
1767 For C and C++ source and include files, generate corresponding Ada specs.
1768 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1769 GNAT User's Guide}, which provides detailed documentation on this feature.
1771 @item -fada-spec-parent=@var{unit}
1772 @opindex fada-spec-parent
1773 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1774 Ada specs as child units of parent @var{unit}.
1776 @item -fdump-go-spec=@var{file}
1777 @opindex fdump-go-spec
1778 For input files in any language, generate corresponding Go
1779 declarations in @var{file}. This generates Go @code{const},
1780 @code{type}, @code{var}, and @code{func} declarations which may be a
1781 useful way to start writing a Go interface to code written in some
1784 @include @value{srcdir}/../libiberty/at-file.texi
1788 @section Compiling C++ Programs
1790 @cindex suffixes for C++ source
1791 @cindex C++ source file suffixes
1792 C++ source files conventionally use one of the suffixes @samp{.C},
1793 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1794 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1795 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1796 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1797 files with these names and compiles them as C++ programs even if you
1798 call the compiler the same way as for compiling C programs (usually
1799 with the name @command{gcc}).
1803 However, the use of @command{gcc} does not add the C++ library.
1804 @command{g++} is a program that calls GCC and automatically specifies linking
1805 against the C++ library. It treats @samp{.c},
1806 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1807 files unless @option{-x} is used. This program is also useful when
1808 precompiling a C header file with a @samp{.h} extension for use in C++
1809 compilations. On many systems, @command{g++} is also installed with
1810 the name @command{c++}.
1812 @cindex invoking @command{g++}
1813 When you compile C++ programs, you may specify many of the same
1814 command-line options that you use for compiling programs in any
1815 language; or command-line options meaningful for C and related
1816 languages; or options that are meaningful only for C++ programs.
1817 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1818 explanations of options for languages related to C@.
1819 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1820 explanations of options that are meaningful only for C++ programs.
1822 @node C Dialect Options
1823 @section Options Controlling C Dialect
1824 @cindex dialect options
1825 @cindex language dialect options
1826 @cindex options, dialect
1828 The following options control the dialect of C (or languages derived
1829 from C, such as C++, Objective-C and Objective-C++) that the compiler
1833 @cindex ANSI support
1837 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1838 equivalent to @option{-std=c++98}.
1840 This turns off certain features of GCC that are incompatible with ISO
1841 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1842 such as the @code{asm} and @code{typeof} keywords, and
1843 predefined macros such as @code{unix} and @code{vax} that identify the
1844 type of system you are using. It also enables the undesirable and
1845 rarely used ISO trigraph feature. For the C compiler,
1846 it disables recognition of C++ style @samp{//} comments as well as
1847 the @code{inline} keyword.
1849 The alternate keywords @code{__asm__}, @code{__extension__},
1850 @code{__inline__} and @code{__typeof__} continue to work despite
1851 @option{-ansi}. You would not want to use them in an ISO C program, of
1852 course, but it is useful to put them in header files that might be included
1853 in compilations done with @option{-ansi}. Alternate predefined macros
1854 such as @code{__unix__} and @code{__vax__} are also available, with or
1855 without @option{-ansi}.
1857 The @option{-ansi} option does not cause non-ISO programs to be
1858 rejected gratuitously. For that, @option{-Wpedantic} is required in
1859 addition to @option{-ansi}. @xref{Warning Options}.
1861 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1862 option is used. Some header files may notice this macro and refrain
1863 from declaring certain functions or defining certain macros that the
1864 ISO standard doesn't call for; this is to avoid interfering with any
1865 programs that might use these names for other things.
1867 Functions that are normally built in but do not have semantics
1868 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1869 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1870 built-in functions provided by GCC}, for details of the functions
1875 Determine the language standard. @xref{Standards,,Language Standards
1876 Supported by GCC}, for details of these standard versions. This option
1877 is currently only supported when compiling C or C++.
1879 The compiler can accept several base standards, such as @samp{c90} or
1880 @samp{c++98}, and GNU dialects of those standards, such as
1881 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1882 compiler accepts all programs following that standard plus those
1883 using GNU extensions that do not contradict it. For example,
1884 @option{-std=c90} turns off certain features of GCC that are
1885 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1886 keywords, but not other GNU extensions that do not have a meaning in
1887 ISO C90, such as omitting the middle term of a @code{?:}
1888 expression. On the other hand, when a GNU dialect of a standard is
1889 specified, all features supported by the compiler are enabled, even when
1890 those features change the meaning of the base standard. As a result, some
1891 strict-conforming programs may be rejected. The particular standard
1892 is used by @option{-Wpedantic} to identify which features are GNU
1893 extensions given that version of the standard. For example
1894 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1895 comments, while @option{-std=gnu99 -Wpedantic} does not.
1897 A value for this option must be provided; possible values are
1903 Support all ISO C90 programs (certain GNU extensions that conflict
1904 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1906 @item iso9899:199409
1907 ISO C90 as modified in amendment 1.
1913 ISO C99. This standard is substantially completely supported, modulo
1914 bugs and floating-point issues
1915 (mainly but not entirely relating to optional C99 features from
1916 Annexes F and G). See
1917 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1918 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1923 ISO C11, the 2011 revision of the ISO C standard. This standard is
1924 substantially completely supported, modulo bugs, floating-point issues
1925 (mainly but not entirely relating to optional C11 features from
1926 Annexes F and G) and the optional Annexes K (Bounds-checking
1927 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1933 ISO C17, the 2017 revision of the ISO C standard
1934 (published in 2018). This standard is
1935 same as C11 except for corrections of defects (all of which are also
1936 applied with @option{-std=c11}) and a new value of
1937 @code{__STDC_VERSION__}, and so is supported to the same extent as C11.
1940 The next version of the ISO C standard, still under development. The
1941 support for this version is experimental and incomplete.
1945 GNU dialect of ISO C90 (including some C99 features).
1949 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1953 GNU dialect of ISO C11.
1954 The name @samp{gnu1x} is deprecated.
1958 GNU dialect of ISO C17. This is the default for C code.
1961 The next version of the ISO C standard, still under development, plus
1962 GNU extensions. The support for this version is experimental and
1967 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1968 additional defect reports. Same as @option{-ansi} for C++ code.
1972 GNU dialect of @option{-std=c++98}.
1976 The 2011 ISO C++ standard plus amendments.
1977 The name @samp{c++0x} is deprecated.
1981 GNU dialect of @option{-std=c++11}.
1982 The name @samp{gnu++0x} is deprecated.
1986 The 2014 ISO C++ standard plus amendments.
1987 The name @samp{c++1y} is deprecated.
1991 GNU dialect of @option{-std=c++14}.
1992 This is the default for C++ code.
1993 The name @samp{gnu++1y} is deprecated.
1997 The 2017 ISO C++ standard plus amendments.
1998 The name @samp{c++1z} is deprecated.
2002 GNU dialect of @option{-std=c++17}.
2003 The name @samp{gnu++1z} is deprecated.
2006 The next revision of the ISO C++ standard, tentatively planned for
2007 2020. Support is highly experimental, and will almost certainly
2008 change in incompatible ways in future releases.
2011 GNU dialect of @option{-std=c++2a}. Support is highly experimental,
2012 and will almost certainly change in incompatible ways in future
2016 @item -fgnu89-inline
2017 @opindex fgnu89-inline
2018 The option @option{-fgnu89-inline} tells GCC to use the traditional
2019 GNU semantics for @code{inline} functions when in C99 mode.
2020 @xref{Inline,,An Inline Function is As Fast As a Macro}.
2021 Using this option is roughly equivalent to adding the
2022 @code{gnu_inline} function attribute to all inline functions
2023 (@pxref{Function Attributes}).
2025 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
2026 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
2027 specifies the default behavior).
2028 This option is not supported in @option{-std=c90} or
2029 @option{-std=gnu90} mode.
2031 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
2032 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
2033 in effect for @code{inline} functions. @xref{Common Predefined
2034 Macros,,,cpp,The C Preprocessor}.
2036 @item -fpermitted-flt-eval-methods=@var{style}
2037 @opindex fpermitted-flt-eval-methods
2038 @opindex fpermitted-flt-eval-methods=c11
2039 @opindex fpermitted-flt-eval-methods=ts-18661-3
2040 ISO/IEC TS 18661-3 defines new permissible values for
2041 @code{FLT_EVAL_METHOD} that indicate that operations and constants with
2042 a semantic type that is an interchange or extended format should be
2043 evaluated to the precision and range of that type. These new values are
2044 a superset of those permitted under C99/C11, which does not specify the
2045 meaning of other positive values of @code{FLT_EVAL_METHOD}. As such, code
2046 conforming to C11 may not have been written expecting the possibility of
2049 @option{-fpermitted-flt-eval-methods} specifies whether the compiler
2050 should allow only the values of @code{FLT_EVAL_METHOD} specified in C99/C11,
2051 or the extended set of values specified in ISO/IEC TS 18661-3.
2053 @var{style} is either @code{c11} or @code{ts-18661-3} as appropriate.
2055 The default when in a standards compliant mode (@option{-std=c11} or similar)
2056 is @option{-fpermitted-flt-eval-methods=c11}. The default when in a GNU
2057 dialect (@option{-std=gnu11} or similar) is
2058 @option{-fpermitted-flt-eval-methods=ts-18661-3}.
2060 @item -aux-info @var{filename}
2062 Output to the given filename prototyped declarations for all functions
2063 declared and/or defined in a translation unit, including those in header
2064 files. This option is silently ignored in any language other than C@.
2066 Besides declarations, the file indicates, in comments, the origin of
2067 each declaration (source file and line), whether the declaration was
2068 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
2069 @samp{O} for old, respectively, in the first character after the line
2070 number and the colon), and whether it came from a declaration or a
2071 definition (@samp{C} or @samp{F}, respectively, in the following
2072 character). In the case of function definitions, a K&R-style list of
2073 arguments followed by their declarations is also provided, inside
2074 comments, after the declaration.
2076 @item -fallow-parameterless-variadic-functions
2077 @opindex fallow-parameterless-variadic-functions
2078 Accept variadic functions without named parameters.
2080 Although it is possible to define such a function, this is not very
2081 useful as it is not possible to read the arguments. This is only
2082 supported for C as this construct is allowed by C++.
2087 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
2088 keyword, so that code can use these words as identifiers. You can use
2089 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
2090 instead. @option{-ansi} implies @option{-fno-asm}.
2092 In C++, this switch only affects the @code{typeof} keyword, since
2093 @code{asm} and @code{inline} are standard keywords. You may want to
2094 use the @option{-fno-gnu-keywords} flag instead, which has the same
2095 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
2096 switch only affects the @code{asm} and @code{typeof} keywords, since
2097 @code{inline} is a standard keyword in ISO C99.
2100 @itemx -fno-builtin-@var{function}
2101 @opindex fno-builtin
2103 @cindex built-in functions
2104 Don't recognize built-in functions that do not begin with
2105 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
2106 functions provided by GCC}, for details of the functions affected,
2107 including those which are not built-in functions when @option{-ansi} or
2108 @option{-std} options for strict ISO C conformance are used because they
2109 do not have an ISO standard meaning.
2111 GCC normally generates special code to handle certain built-in functions
2112 more efficiently; for instance, calls to @code{alloca} may become single
2113 instructions which adjust the stack directly, and calls to @code{memcpy}
2114 may become inline copy loops. The resulting code is often both smaller
2115 and faster, but since the function calls no longer appear as such, you
2116 cannot set a breakpoint on those calls, nor can you change the behavior
2117 of the functions by linking with a different library. In addition,
2118 when a function is recognized as a built-in function, GCC may use
2119 information about that function to warn about problems with calls to
2120 that function, or to generate more efficient code, even if the
2121 resulting code still contains calls to that function. For example,
2122 warnings are given with @option{-Wformat} for bad calls to
2123 @code{printf} when @code{printf} is built in and @code{strlen} is
2124 known not to modify global memory.
2126 With the @option{-fno-builtin-@var{function}} option
2127 only the built-in function @var{function} is
2128 disabled. @var{function} must not begin with @samp{__builtin_}. If a
2129 function is named that is not built-in in this version of GCC, this
2130 option is ignored. There is no corresponding
2131 @option{-fbuiltin-@var{function}} option; if you wish to enable
2132 built-in functions selectively when using @option{-fno-builtin} or
2133 @option{-ffreestanding}, you may define macros such as:
2136 #define abs(n) __builtin_abs ((n))
2137 #define strcpy(d, s) __builtin_strcpy ((d), (s))
2143 Enable parsing of function definitions marked with @code{__GIMPLE}.
2144 This is an experimental feature that allows unit testing of GIMPLE
2149 @cindex hosted environment
2151 Assert that compilation targets a hosted environment. This implies
2152 @option{-fbuiltin}. A hosted environment is one in which the
2153 entire standard library is available, and in which @code{main} has a return
2154 type of @code{int}. Examples are nearly everything except a kernel.
2155 This is equivalent to @option{-fno-freestanding}.
2157 @item -ffreestanding
2158 @opindex ffreestanding
2159 @cindex hosted environment
2161 Assert that compilation targets a freestanding environment. This
2162 implies @option{-fno-builtin}. A freestanding environment
2163 is one in which the standard library may not exist, and program startup may
2164 not necessarily be at @code{main}. The most obvious example is an OS kernel.
2165 This is equivalent to @option{-fno-hosted}.
2167 @xref{Standards,,Language Standards Supported by GCC}, for details of
2168 freestanding and hosted environments.
2172 @cindex OpenACC accelerator programming
2173 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
2174 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
2175 compiler generates accelerated code according to the OpenACC Application
2176 Programming Interface v2.0 @w{@uref{https://www.openacc.org}}. This option
2177 implies @option{-pthread}, and thus is only supported on targets that
2178 have support for @option{-pthread}.
2180 @item -fopenacc-dim=@var{geom}
2181 @opindex fopenacc-dim
2182 @cindex OpenACC accelerator programming
2183 Specify default compute dimensions for parallel offload regions that do
2184 not explicitly specify. The @var{geom} value is a triple of
2185 ':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A size
2186 can be omitted, to use a target-specific default value.
2190 @cindex OpenMP parallel
2191 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
2192 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
2193 compiler generates parallel code according to the OpenMP Application
2194 Program Interface v4.5 @w{@uref{https://www.openmp.org}}. This option
2195 implies @option{-pthread}, and thus is only supported on targets that
2196 have support for @option{-pthread}. @option{-fopenmp} implies
2197 @option{-fopenmp-simd}.
2200 @opindex fopenmp-simd
2203 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
2204 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
2209 When the option @option{-fgnu-tm} is specified, the compiler
2210 generates code for the Linux variant of Intel's current Transactional
2211 Memory ABI specification document (Revision 1.1, May 6 2009). This is
2212 an experimental feature whose interface may change in future versions
2213 of GCC, as the official specification changes. Please note that not
2214 all architectures are supported for this feature.
2216 For more information on GCC's support for transactional memory,
2217 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
2218 Transactional Memory Library}.
2220 Note that the transactional memory feature is not supported with
2221 non-call exceptions (@option{-fnon-call-exceptions}).
2223 @item -fms-extensions
2224 @opindex fms-extensions
2225 Accept some non-standard constructs used in Microsoft header files.
2227 In C++ code, this allows member names in structures to be similar
2228 to previous types declarations.
2237 Some cases of unnamed fields in structures and unions are only
2238 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2239 fields within structs/unions}, for details.
2241 Note that this option is off for all targets but x86
2242 targets using ms-abi.
2244 @item -fplan9-extensions
2245 @opindex fplan9-extensions
2246 Accept some non-standard constructs used in Plan 9 code.
2248 This enables @option{-fms-extensions}, permits passing pointers to
2249 structures with anonymous fields to functions that expect pointers to
2250 elements of the type of the field, and permits referring to anonymous
2251 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2252 struct/union fields within structs/unions}, for details. This is only
2253 supported for C, not C++.
2255 @item -fcond-mismatch
2256 @opindex fcond-mismatch
2257 Allow conditional expressions with mismatched types in the second and
2258 third arguments. The value of such an expression is void. This option
2259 is not supported for C++.
2261 @item -flax-vector-conversions
2262 @opindex flax-vector-conversions
2263 Allow implicit conversions between vectors with differing numbers of
2264 elements and/or incompatible element types. This option should not be
2267 @item -funsigned-char
2268 @opindex funsigned-char
2269 Let the type @code{char} be unsigned, like @code{unsigned char}.
2271 Each kind of machine has a default for what @code{char} should
2272 be. It is either like @code{unsigned char} by default or like
2273 @code{signed char} by default.
2275 Ideally, a portable program should always use @code{signed char} or
2276 @code{unsigned char} when it depends on the signedness of an object.
2277 But many programs have been written to use plain @code{char} and
2278 expect it to be signed, or expect it to be unsigned, depending on the
2279 machines they were written for. This option, and its inverse, let you
2280 make such a program work with the opposite default.
2282 The type @code{char} is always a distinct type from each of
2283 @code{signed char} or @code{unsigned char}, even though its behavior
2284 is always just like one of those two.
2287 @opindex fsigned-char
2288 Let the type @code{char} be signed, like @code{signed char}.
2290 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2291 the negative form of @option{-funsigned-char}. Likewise, the option
2292 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2294 @item -fsigned-bitfields
2295 @itemx -funsigned-bitfields
2296 @itemx -fno-signed-bitfields
2297 @itemx -fno-unsigned-bitfields
2298 @opindex fsigned-bitfields
2299 @opindex funsigned-bitfields
2300 @opindex fno-signed-bitfields
2301 @opindex fno-unsigned-bitfields
2302 These options control whether a bit-field is signed or unsigned, when the
2303 declaration does not use either @code{signed} or @code{unsigned}. By
2304 default, such a bit-field is signed, because this is consistent: the
2305 basic integer types such as @code{int} are signed types.
2307 @item -fsso-struct=@var{endianness}
2308 @opindex fsso-struct
2309 Set the default scalar storage order of structures and unions to the
2310 specified endianness. The accepted values are @samp{big-endian},
2311 @samp{little-endian} and @samp{native} for the native endianness of
2312 the target (the default). This option is not supported for C++.
2314 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2315 code that is not binary compatible with code generated without it if the
2316 specified endianness is not the native endianness of the target.
2319 @node C++ Dialect Options
2320 @section Options Controlling C++ Dialect
2322 @cindex compiler options, C++
2323 @cindex C++ options, command-line
2324 @cindex options, C++
2325 This section describes the command-line options that are only meaningful
2326 for C++ programs. You can also use most of the GNU compiler options
2327 regardless of what language your program is in. For example, you
2328 might compile a file @file{firstClass.C} like this:
2331 g++ -g -fstrict-enums -O -c firstClass.C
2335 In this example, only @option{-fstrict-enums} is an option meant
2336 only for C++ programs; you can use the other options with any
2337 language supported by GCC@.
2339 Some options for compiling C programs, such as @option{-std}, are also
2340 relevant for C++ programs.
2341 @xref{C Dialect Options,,Options Controlling C Dialect}.
2343 Here is a list of options that are @emph{only} for compiling C++ programs:
2347 @item -fabi-version=@var{n}
2348 @opindex fabi-version
2349 Use version @var{n} of the C++ ABI@. The default is version 0.
2351 Version 0 refers to the version conforming most closely to
2352 the C++ ABI specification. Therefore, the ABI obtained using version 0
2353 will change in different versions of G++ as ABI bugs are fixed.
2355 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2357 Version 2 is the version of the C++ ABI that first appeared in G++
2358 3.4, and was the default through G++ 4.9.
2360 Version 3 corrects an error in mangling a constant address as a
2363 Version 4, which first appeared in G++ 4.5, implements a standard
2364 mangling for vector types.
2366 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2367 attribute const/volatile on function pointer types, decltype of a
2368 plain decl, and use of a function parameter in the declaration of
2371 Version 6, which first appeared in G++ 4.7, corrects the promotion
2372 behavior of C++11 scoped enums and the mangling of template argument
2373 packs, const/static_cast, prefix ++ and --, and a class scope function
2374 used as a template argument.
2376 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2377 builtin type and corrects the mangling of lambdas in default argument
2380 Version 8, which first appeared in G++ 4.9, corrects the substitution
2381 behavior of function types with function-cv-qualifiers.
2383 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2386 Version 10, which first appeared in G++ 6.1, adds mangling of
2387 attributes that affect type identity, such as ia32 calling convention
2388 attributes (e.g.@: @samp{stdcall}).
2390 Version 11, which first appeared in G++ 7, corrects the mangling of
2391 sizeof... expressions and operator names. For multiple entities with
2392 the same name within a function, that are declared in different scopes,
2393 the mangling now changes starting with the twelfth occurrence. It also
2394 implies @option{-fnew-inheriting-ctors}.
2396 Version 12, which first appeared in G++ 8, corrects the calling
2397 conventions for empty classes on the x86_64 target and for classes
2398 with only deleted copy/move constructors. It accidentally changes the
2399 calling convention for classes with a deleted copy constructor and a
2400 trivial move constructor.
2402 Version 13, which first appeared in G++ 8.2, fixes the accidental
2403 change in version 12.
2405 See also @option{-Wabi}.
2407 @item -fabi-compat-version=@var{n}
2408 @opindex fabi-compat-version
2409 On targets that support strong aliases, G++
2410 works around mangling changes by creating an alias with the correct
2411 mangled name when defining a symbol with an incorrect mangled name.
2412 This switch specifies which ABI version to use for the alias.
2414 With @option{-fabi-version=0} (the default), this defaults to 11 (GCC 7
2415 compatibility). If another ABI version is explicitly selected, this
2416 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2417 use @option{-fabi-compat-version=2}.
2419 If this option is not provided but @option{-Wabi=@var{n}} is, that
2420 version is used for compatibility aliases. If this option is provided
2421 along with @option{-Wabi} (without the version), the version from this
2422 option is used for the warning.
2424 @item -fno-access-control
2425 @opindex fno-access-control
2426 @opindex faccess-control
2427 Turn off all access checking. This switch is mainly useful for working
2428 around bugs in the access control code.
2431 @opindex faligned-new
2432 Enable support for C++17 @code{new} of types that require more
2433 alignment than @code{void* ::operator new(std::size_t)} provides. A
2434 numeric argument such as @code{-faligned-new=32} can be used to
2435 specify how much alignment (in bytes) is provided by that function,
2436 but few users will need to override the default of
2437 @code{alignof(std::max_align_t)}.
2439 This flag is enabled by default for @option{-std=c++17}.
2444 @opindex fno-char8_t
2445 Enable support for @code{char8_t} as adopted for C++2a. This includes
2446 the addition of a new @code{char8_t} fundamental type, changes to the
2447 types of UTF-8 string and character literals, new signatures for
2448 user-defined literals, associated standard library updates, and new
2449 @code{__cpp_char8_t} and @code{__cpp_lib_char8_t} feature test macros.
2451 This option enables functions to be overloaded for ordinary and UTF-8
2455 int f(const char *); // #1
2456 int f(const char8_t *); // #2
2457 int v1 = f("text"); // Calls #1
2458 int v2 = f(u8"text"); // Calls #2
2462 and introduces new signatures for user-defined literals:
2465 int operator""_udl1(char8_t);
2466 int v3 = u8'x'_udl1;
2467 int operator""_udl2(const char8_t*, std::size_t);
2468 int v4 = u8"text"_udl2;
2469 template<typename T, T...> int operator""_udl3();
2470 int v5 = u8"text"_udl3;
2474 The change to the types of UTF-8 string and character literals introduces
2475 incompatibilities with ISO C++11 and later standards. For example, the
2476 following code is well-formed under ISO C++11, but is ill-formed when
2477 @option{-fchar8_t} is specified.
2480 char ca[] = u8"xx"; // error: char-array initialized from wide
2482 const char *cp = u8"xx";// error: invalid conversion from
2483 // `const char8_t*' to `const char*'
2485 auto v = f(u8"xx"); // error: invalid conversion from
2486 // `const char8_t*' to `const char*'
2487 std::string s@{u8"xx"@}; // error: no matching function for call to
2488 // `std::basic_string<char>::basic_string()'
2489 using namespace std::literals;
2490 s = u8"xx"s; // error: conversion from
2491 // `basic_string<char8_t>' to non-scalar
2492 // type `basic_string<char>' requested
2497 Check that the pointer returned by @code{operator new} is non-null
2498 before attempting to modify the storage allocated. This check is
2499 normally unnecessary because the C++ standard specifies that
2500 @code{operator new} only returns @code{0} if it is declared
2501 @code{throw()}, in which case the compiler always checks the
2502 return value even without this option. In all other cases, when
2503 @code{operator new} has a non-empty exception specification, memory
2504 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2505 @samp{new (nothrow)}.
2509 Enable support for the C++ Extensions for Concepts Technical
2510 Specification, ISO 19217 (2015), which allows code like
2513 template <class T> concept bool Addable = requires (T t) @{ t + t; @};
2514 template <Addable T> T add (T a, T b) @{ return a + b; @}
2517 @item -fconstexpr-depth=@var{n}
2518 @opindex fconstexpr-depth
2519 Set the maximum nested evaluation depth for C++11 constexpr functions
2520 to @var{n}. A limit is needed to detect endless recursion during
2521 constant expression evaluation. The minimum specified by the standard
2524 @item -fconstexpr-loop-limit=@var{n}
2525 @opindex fconstexpr-loop-limit
2526 Set the maximum number of iterations for a loop in C++14 constexpr functions
2527 to @var{n}. A limit is needed to detect infinite loops during
2528 constant expression evaluation. The default is 262144 (1<<18).
2530 @item -fconstexpr-ops-limit=@var{n}
2531 @opindex fconstexpr-ops-limit
2532 Set the maximum number of operations during a single constexpr evaluation.
2533 Even when number of iterations of a single loop is limited with the above limit,
2534 if there are several nested loops and each of them has many iterations but still
2535 smaller than the above limit, or if in a body of some loop or even outside
2536 of a loop too many expressions need to be evaluated, the resulting constexpr
2537 evaluation might take too long.
2538 The default is 33554432 (1<<25).
2540 @item -fdeduce-init-list
2541 @opindex fdeduce-init-list
2542 Enable deduction of a template type parameter as
2543 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2546 template <class T> auto forward(T t) -> decltype (realfn (t))
2553 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2557 This deduction was implemented as a possible extension to the
2558 originally proposed semantics for the C++11 standard, but was not part
2559 of the final standard, so it is disabled by default. This option is
2560 deprecated, and may be removed in a future version of G++.
2562 @item -fno-elide-constructors
2563 @opindex fno-elide-constructors
2564 @opindex felide-constructors
2565 The C++ standard allows an implementation to omit creating a temporary
2566 that is only used to initialize another object of the same type.
2567 Specifying this option disables that optimization, and forces G++ to
2568 call the copy constructor in all cases. This option also causes G++
2569 to call trivial member functions which otherwise would be expanded inline.
2571 In C++17, the compiler is required to omit these temporaries, but this
2572 option still affects trivial member functions.
2574 @item -fno-enforce-eh-specs
2575 @opindex fno-enforce-eh-specs
2576 @opindex fenforce-eh-specs
2577 Don't generate code to check for violation of exception specifications
2578 at run time. This option violates the C++ standard, but may be useful
2579 for reducing code size in production builds, much like defining
2580 @code{NDEBUG}. This does not give user code permission to throw
2581 exceptions in violation of the exception specifications; the compiler
2582 still optimizes based on the specifications, so throwing an
2583 unexpected exception results in undefined behavior at run time.
2585 @item -fextern-tls-init
2586 @itemx -fno-extern-tls-init
2587 @opindex fextern-tls-init
2588 @opindex fno-extern-tls-init
2589 The C++11 and OpenMP standards allow @code{thread_local} and
2590 @code{threadprivate} variables to have dynamic (runtime)
2591 initialization. To support this, any use of such a variable goes
2592 through a wrapper function that performs any necessary initialization.
2593 When the use and definition of the variable are in the same
2594 translation unit, this overhead can be optimized away, but when the
2595 use is in a different translation unit there is significant overhead
2596 even if the variable doesn't actually need dynamic initialization. If
2597 the programmer can be sure that no use of the variable in a
2598 non-defining TU needs to trigger dynamic initialization (either
2599 because the variable is statically initialized, or a use of the
2600 variable in the defining TU will be executed before any uses in
2601 another TU), they can avoid this overhead with the
2602 @option{-fno-extern-tls-init} option.
2604 On targets that support symbol aliases, the default is
2605 @option{-fextern-tls-init}. On targets that do not support symbol
2606 aliases, the default is @option{-fno-extern-tls-init}.
2608 @item -fno-gnu-keywords
2609 @opindex fno-gnu-keywords
2610 @opindex fgnu-keywords
2611 Do not recognize @code{typeof} as a keyword, so that code can use this
2612 word as an identifier. You can use the keyword @code{__typeof__} instead.
2613 This option is implied by the strict ISO C++ dialects: @option{-ansi},
2614 @option{-std=c++98}, @option{-std=c++11}, etc.
2616 @item -fno-implicit-templates
2617 @opindex fno-implicit-templates
2618 @opindex fimplicit-templates
2619 Never emit code for non-inline templates that are instantiated
2620 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2621 If you use this option, you must take care to structure your code to
2622 include all the necessary explicit instantiations to avoid getting
2623 undefined symbols at link time.
2624 @xref{Template Instantiation}, for more information.
2626 @item -fno-implicit-inline-templates
2627 @opindex fno-implicit-inline-templates
2628 @opindex fimplicit-inline-templates
2629 Don't emit code for implicit instantiations of inline templates, either.
2630 The default is to handle inlines differently so that compiles with and
2631 without optimization need the same set of explicit instantiations.
2633 @item -fno-implement-inlines
2634 @opindex fno-implement-inlines
2635 @opindex fimplement-inlines
2636 To save space, do not emit out-of-line copies of inline functions
2637 controlled by @code{#pragma implementation}. This causes linker
2638 errors if these functions are not inlined everywhere they are called.
2640 @item -fms-extensions
2641 @opindex fms-extensions
2642 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2643 int and getting a pointer to member function via non-standard syntax.
2645 @item -fnew-inheriting-ctors
2646 @opindex fnew-inheriting-ctors
2647 Enable the P0136 adjustment to the semantics of C++11 constructor
2648 inheritance. This is part of C++17 but also considered to be a Defect
2649 Report against C++11 and C++14. This flag is enabled by default
2650 unless @option{-fabi-version=10} or lower is specified.
2652 @item -fnew-ttp-matching
2653 @opindex fnew-ttp-matching
2654 Enable the P0522 resolution to Core issue 150, template template
2655 parameters and default arguments: this allows a template with default
2656 template arguments as an argument for a template template parameter
2657 with fewer template parameters. This flag is enabled by default for
2658 @option{-std=c++17}.
2660 @item -fno-nonansi-builtins
2661 @opindex fno-nonansi-builtins
2662 @opindex fnonansi-builtins
2663 Disable built-in declarations of functions that are not mandated by
2664 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2665 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2668 @opindex fnothrow-opt
2669 Treat a @code{throw()} exception specification as if it were a
2670 @code{noexcept} specification to reduce or eliminate the text size
2671 overhead relative to a function with no exception specification. If
2672 the function has local variables of types with non-trivial
2673 destructors, the exception specification actually makes the
2674 function smaller because the EH cleanups for those variables can be
2675 optimized away. The semantic effect is that an exception thrown out of
2676 a function with such an exception specification results in a call
2677 to @code{terminate} rather than @code{unexpected}.
2679 @item -fno-operator-names
2680 @opindex fno-operator-names
2681 @opindex foperator-names
2682 Do not treat the operator name keywords @code{and}, @code{bitand},
2683 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2684 synonyms as keywords.
2686 @item -fno-optional-diags
2687 @opindex fno-optional-diags
2688 @opindex foptional-diags
2689 Disable diagnostics that the standard says a compiler does not need to
2690 issue. Currently, the only such diagnostic issued by G++ is the one for
2691 a name having multiple meanings within a class.
2694 @opindex fpermissive
2695 Downgrade some diagnostics about nonconformant code from errors to
2696 warnings. Thus, using @option{-fpermissive} allows some
2697 nonconforming code to compile.
2699 @item -fno-pretty-templates
2700 @opindex fno-pretty-templates
2701 @opindex fpretty-templates
2702 When an error message refers to a specialization of a function
2703 template, the compiler normally prints the signature of the
2704 template followed by the template arguments and any typedefs or
2705 typenames in the signature (e.g.@: @code{void f(T) [with T = int]}
2706 rather than @code{void f(int)}) so that it's clear which template is
2707 involved. When an error message refers to a specialization of a class
2708 template, the compiler omits any template arguments that match
2709 the default template arguments for that template. If either of these
2710 behaviors make it harder to understand the error message rather than
2711 easier, you can use @option{-fno-pretty-templates} to disable them.
2715 Enable automatic template instantiation at link time. This option also
2716 implies @option{-fno-implicit-templates}. @xref{Template
2717 Instantiation}, for more information.
2722 Disable generation of information about every class with virtual
2723 functions for use by the C++ run-time type identification features
2724 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2725 of the language, you can save some space by using this flag. Note that
2726 exception handling uses the same information, but G++ generates it as
2727 needed. The @code{dynamic_cast} operator can still be used for casts that
2728 do not require run-time type information, i.e.@: casts to @code{void *} or to
2729 unambiguous base classes.
2731 Mixing code compiled with @option{-frtti} with that compiled with
2732 @option{-fno-rtti} may not work. For example, programs may
2733 fail to link if a class compiled with @option{-fno-rtti} is used as a base
2734 for a class compiled with @option{-frtti}.
2736 @item -fsized-deallocation
2737 @opindex fsized-deallocation
2738 Enable the built-in global declarations
2740 void operator delete (void *, std::size_t) noexcept;
2741 void operator delete[] (void *, std::size_t) noexcept;
2743 as introduced in C++14. This is useful for user-defined replacement
2744 deallocation functions that, for example, use the size of the object
2745 to make deallocation faster. Enabled by default under
2746 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2747 warns about places that might want to add a definition.
2749 @item -fstrict-enums
2750 @opindex fstrict-enums
2751 Allow the compiler to optimize using the assumption that a value of
2752 enumerated type can only be one of the values of the enumeration (as
2753 defined in the C++ standard; basically, a value that can be
2754 represented in the minimum number of bits needed to represent all the
2755 enumerators). This assumption may not be valid if the program uses a
2756 cast to convert an arbitrary integer value to the enumerated type.
2758 @item -fstrong-eval-order
2759 @opindex fstrong-eval-order
2760 Evaluate member access, array subscripting, and shift expressions in
2761 left-to-right order, and evaluate assignment in right-to-left order,
2762 as adopted for C++17. Enabled by default with @option{-std=c++17}.
2763 @option{-fstrong-eval-order=some} enables just the ordering of member
2764 access and shift expressions, and is the default without
2765 @option{-std=c++17}.
2767 @item -ftemplate-backtrace-limit=@var{n}
2768 @opindex ftemplate-backtrace-limit
2769 Set the maximum number of template instantiation notes for a single
2770 warning or error to @var{n}. The default value is 10.
2772 @item -ftemplate-depth=@var{n}
2773 @opindex ftemplate-depth
2774 Set the maximum instantiation depth for template classes to @var{n}.
2775 A limit on the template instantiation depth is needed to detect
2776 endless recursions during template class instantiation. ANSI/ISO C++
2777 conforming programs must not rely on a maximum depth greater than 17
2778 (changed to 1024 in C++11). The default value is 900, as the compiler
2779 can run out of stack space before hitting 1024 in some situations.
2781 @item -fno-threadsafe-statics
2782 @opindex fno-threadsafe-statics
2783 @opindex fthreadsafe-statics
2784 Do not emit the extra code to use the routines specified in the C++
2785 ABI for thread-safe initialization of local statics. You can use this
2786 option to reduce code size slightly in code that doesn't need to be
2789 @item -fuse-cxa-atexit
2790 @opindex fuse-cxa-atexit
2791 Register destructors for objects with static storage duration with the
2792 @code{__cxa_atexit} function rather than the @code{atexit} function.
2793 This option is required for fully standards-compliant handling of static
2794 destructors, but only works if your C library supports
2795 @code{__cxa_atexit}.
2797 @item -fno-use-cxa-get-exception-ptr
2798 @opindex fno-use-cxa-get-exception-ptr
2799 @opindex fuse-cxa-get-exception-ptr
2800 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2801 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2802 if the runtime routine is not available.
2804 @item -fvisibility-inlines-hidden
2805 @opindex fvisibility-inlines-hidden
2806 This switch declares that the user does not attempt to compare
2807 pointers to inline functions or methods where the addresses of the two functions
2808 are taken in different shared objects.
2810 The effect of this is that GCC may, effectively, mark inline methods with
2811 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2812 appear in the export table of a DSO and do not require a PLT indirection
2813 when used within the DSO@. Enabling this option can have a dramatic effect
2814 on load and link times of a DSO as it massively reduces the size of the
2815 dynamic export table when the library makes heavy use of templates.
2817 The behavior of this switch is not quite the same as marking the
2818 methods as hidden directly, because it does not affect static variables
2819 local to the function or cause the compiler to deduce that
2820 the function is defined in only one shared object.
2822 You may mark a method as having a visibility explicitly to negate the
2823 effect of the switch for that method. For example, if you do want to
2824 compare pointers to a particular inline method, you might mark it as
2825 having default visibility. Marking the enclosing class with explicit
2826 visibility has no effect.
2828 Explicitly instantiated inline methods are unaffected by this option
2829 as their linkage might otherwise cross a shared library boundary.
2830 @xref{Template Instantiation}.
2832 @item -fvisibility-ms-compat
2833 @opindex fvisibility-ms-compat
2834 This flag attempts to use visibility settings to make GCC's C++
2835 linkage model compatible with that of Microsoft Visual Studio.
2837 The flag makes these changes to GCC's linkage model:
2841 It sets the default visibility to @code{hidden}, like
2842 @option{-fvisibility=hidden}.
2845 Types, but not their members, are not hidden by default.
2848 The One Definition Rule is relaxed for types without explicit
2849 visibility specifications that are defined in more than one
2850 shared object: those declarations are permitted if they are
2851 permitted when this option is not used.
2854 In new code it is better to use @option{-fvisibility=hidden} and
2855 export those classes that are intended to be externally visible.
2856 Unfortunately it is possible for code to rely, perhaps accidentally,
2857 on the Visual Studio behavior.
2859 Among the consequences of these changes are that static data members
2860 of the same type with the same name but defined in different shared
2861 objects are different, so changing one does not change the other;
2862 and that pointers to function members defined in different shared
2863 objects may not compare equal. When this flag is given, it is a
2864 violation of the ODR to define types with the same name differently.
2869 Do not use weak symbol support, even if it is provided by the linker.
2870 By default, G++ uses weak symbols if they are available. This
2871 option exists only for testing, and should not be used by end-users;
2872 it results in inferior code and has no benefits. This option may
2873 be removed in a future release of G++.
2877 Do not search for header files in the standard directories specific to
2878 C++, but do still search the other standard directories. (This option
2879 is used when building the C++ library.)
2882 In addition, these optimization, warning, and code generation options
2883 have meanings only for C++ programs:
2886 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2889 Warn when G++ it generates code that is probably not compatible with
2890 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2891 ABI with each major release, normally @option{-Wabi} will warn only if
2892 there is a check added later in a release series for an ABI issue
2893 discovered since the initial release. @option{-Wabi} will warn about
2894 more things if an older ABI version is selected (with
2895 @option{-fabi-version=@var{n}}).
2897 @option{-Wabi} can also be used with an explicit version number to
2898 warn about compatibility with a particular @option{-fabi-version}
2899 level, e.g.@: @option{-Wabi=2} to warn about changes relative to
2900 @option{-fabi-version=2}.
2902 If an explicit version number is provided and
2903 @option{-fabi-compat-version} is not specified, the version number
2904 from this option is used for compatibility aliases. If no explicit
2905 version number is provided with this option, but
2906 @option{-fabi-compat-version} is specified, that version number is
2907 used for ABI warnings.
2909 Although an effort has been made to warn about
2910 all such cases, there are probably some cases that are not warned about,
2911 even though G++ is generating incompatible code. There may also be
2912 cases where warnings are emitted even though the code that is generated
2915 You should rewrite your code to avoid these warnings if you are
2916 concerned about the fact that code generated by G++ may not be binary
2917 compatible with code generated by other compilers.
2919 Known incompatibilities in @option{-fabi-version=2} (which was the
2920 default from GCC 3.4 to 4.9) include:
2925 A template with a non-type template parameter of reference type was
2926 mangled incorrectly:
2929 template <int &> struct S @{@};
2933 This was fixed in @option{-fabi-version=3}.
2936 SIMD vector types declared using @code{__attribute ((vector_size))} were
2937 mangled in a non-standard way that does not allow for overloading of
2938 functions taking vectors of different sizes.
2940 The mangling was changed in @option{-fabi-version=4}.
2943 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2944 qualifiers, and @code{decltype} of a plain declaration was folded away.
2946 These mangling issues were fixed in @option{-fabi-version=5}.
2949 Scoped enumerators passed as arguments to a variadic function are
2950 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2951 On most targets this does not actually affect the parameter passing
2952 ABI, as there is no way to pass an argument smaller than @code{int}.
2954 Also, the ABI changed the mangling of template argument packs,
2955 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2956 a class scope function used as a template argument.
2958 These issues were corrected in @option{-fabi-version=6}.
2961 Lambdas in default argument scope were mangled incorrectly, and the
2962 ABI changed the mangling of @code{nullptr_t}.
2964 These issues were corrected in @option{-fabi-version=7}.
2967 When mangling a function type with function-cv-qualifiers, the
2968 un-qualified function type was incorrectly treated as a substitution
2971 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2974 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2975 unaligned accesses. Note that this did not affect the ABI of a
2976 function with a @code{nullptr_t} parameter, as parameters have a
2979 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2982 Target-specific attributes that affect the identity of a type, such as
2983 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2984 did not affect the mangled name, leading to name collisions when
2985 function pointers were used as template arguments.
2987 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2991 It also warns about psABI-related changes. The known psABI changes at this
2997 For SysV/x86-64, unions with @code{long double} members are
2998 passed in memory as specified in psABI. For example:
3008 @code{union U} is always passed in memory.
3012 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
3015 Warn when a type with an ABI tag is used in a context that does not
3016 have that ABI tag. See @ref{C++ Attributes} for more information
3019 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
3020 @opindex Wctor-dtor-privacy
3021 @opindex Wno-ctor-dtor-privacy
3022 Warn when a class seems unusable because all the constructors or
3023 destructors in that class are private, and it has neither friends nor
3024 public static member functions. Also warn if there are no non-private
3025 methods, and there's at least one private member function that isn't
3026 a constructor or destructor.
3028 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
3029 @opindex Wdelete-non-virtual-dtor
3030 @opindex Wno-delete-non-virtual-dtor
3031 Warn when @code{delete} is used to destroy an instance of a class that
3032 has virtual functions and non-virtual destructor. It is unsafe to delete
3033 an instance of a derived class through a pointer to a base class if the
3034 base class does not have a virtual destructor. This warning is enabled
3037 @item -Wdeprecated-copy @r{(C++ and Objective-C++ only)}
3038 @opindex Wdeprecated-copy
3039 @opindex Wno-deprecated-copy
3040 Warn that the implicit declaration of a copy constructor or copy
3041 assignment operator is deprecated if the class has a user-provided
3042 copy constructor or copy assignment operator, in C++11 and up. This
3043 warning is enabled by @option{-Wextra}. With
3044 @option{-Wdeprecated-copy-dtor}, also deprecate if the class has a
3045 user-provided destructor.
3047 @item -Wno-init-list-lifetime @r{(C++ and Objective-C++ only)}
3048 @opindex Winit-list-lifetime
3049 @opindex Wno-init-list-lifetime
3050 Do not warn about uses of @code{std::initializer_list} that are likely
3051 to result in dangling pointers. Since the underlying array for an
3052 @code{initializer_list} is handled like a normal C++ temporary object,
3053 it is easy to inadvertently keep a pointer to the array past the end
3054 of the array's lifetime. For example:
3058 If a function returns a temporary @code{initializer_list}, or a local
3059 @code{initializer_list} variable, the array's lifetime ends at the end
3060 of the return statement, so the value returned has a dangling pointer.
3063 If a new-expression creates an @code{initializer_list}, the array only
3064 lives until the end of the enclosing full-expression, so the
3065 @code{initializer_list} in the heap has a dangling pointer.
3068 When an @code{initializer_list} variable is assigned from a
3069 brace-enclosed initializer list, the temporary array created for the
3070 right side of the assignment only lives until the end of the
3071 full-expression, so at the next statement the @code{initializer_list}
3072 variable has a dangling pointer.
3075 // li's initial underlying array lives as long as li
3076 std::initializer_list<int> li = @{ 1,2,3 @};
3077 // assignment changes li to point to a temporary array
3079 // now the temporary is gone and li has a dangling pointer
3080 int i = li.begin()[0] // undefined behavior
3084 When a list constructor stores the @code{begin} pointer from the
3085 @code{initializer_list} argument, this doesn't extend the lifetime of
3086 the array, so if a class variable is constructed from a temporary
3087 @code{initializer_list}, the pointer is left dangling by the end of
3088 the variable declaration statement.
3092 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
3093 @opindex Wliteral-suffix
3094 @opindex Wno-literal-suffix
3095 Warn when a string or character literal is followed by a ud-suffix which does
3096 not begin with an underscore. As a conforming extension, GCC treats such
3097 suffixes as separate preprocessing tokens in order to maintain backwards
3098 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
3102 #define __STDC_FORMAT_MACROS
3103 #include <inttypes.h>
3108 printf("My int64: %" PRId64"\n", i64);
3112 In this case, @code{PRId64} is treated as a separate preprocessing token.
3114 Additionally, warn when a user-defined literal operator is declared with
3115 a literal suffix identifier that doesn't begin with an underscore. Literal
3116 suffix identifiers that don't begin with an underscore are reserved for
3117 future standardization.
3119 This warning is enabled by default.
3121 @item -Wlto-type-mismatch
3122 @opindex Wlto-type-mismatch
3123 @opindex Wno-lto-type-mismatch
3125 During the link-time optimization warn about type mismatches in
3126 global declarations from different compilation units.
3127 Requires @option{-flto} to be enabled. Enabled by default.
3129 @item -Wno-narrowing @r{(C++ and Objective-C++ only)}
3131 @opindex Wno-narrowing
3132 For C++11 and later standards, narrowing conversions are diagnosed by default,
3133 as required by the standard. A narrowing conversion from a constant produces
3134 an error, and a narrowing conversion from a non-constant produces a warning,
3135 but @option{-Wno-narrowing} suppresses the diagnostic.
3136 Note that this does not affect the meaning of well-formed code;
3137 narrowing conversions are still considered ill-formed in SFINAE contexts.
3139 With @option{-Wnarrowing} in C++98, warn when a narrowing
3140 conversion prohibited by C++11 occurs within
3144 int i = @{ 2.2 @}; // error: narrowing from double to int
3147 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
3149 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
3151 @opindex Wno-noexcept
3152 Warn when a noexcept-expression evaluates to false because of a call
3153 to a function that does not have a non-throwing exception
3154 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
3155 the compiler to never throw an exception.
3157 @item -Wnoexcept-type @r{(C++ and Objective-C++ only)}
3158 @opindex Wnoexcept-type
3159 @opindex Wno-noexcept-type
3160 Warn if the C++17 feature making @code{noexcept} part of a function
3161 type changes the mangled name of a symbol relative to C++14. Enabled
3162 by @option{-Wabi} and @option{-Wc++17-compat}.
3167 template <class T> void f(T t) @{ t(); @};
3169 void h() @{ f(g); @}
3173 In C++14, @code{f} calls @code{f<void(*)()>}, but in
3174 C++17 it calls @code{f<void(*)()noexcept>}.
3176 @item -Wclass-memaccess @r{(C++ and Objective-C++ only)}
3177 @opindex Wclass-memaccess
3178 @opindex Wno-class-memaccess
3179 Warn when the destination of a call to a raw memory function such as
3180 @code{memset} or @code{memcpy} is an object of class type, and when writing
3181 into such an object might bypass the class non-trivial or deleted constructor
3182 or copy assignment, violate const-correctness or encapsulation, or corrupt
3183 virtual table pointers. Modifying the representation of such objects may
3184 violate invariants maintained by member functions of the class. For example,
3185 the call to @code{memset} below is undefined because it modifies a non-trivial
3186 class object and is, therefore, diagnosed. The safe way to either initialize
3187 or clear the storage of objects of such types is by using the appropriate
3188 constructor or assignment operator, if one is available.
3190 std::string str = "abc";
3191 memset (&str, 0, sizeof str);
3193 The @option{-Wclass-memaccess} option is enabled by @option{-Wall}.
3194 Explicitly casting the pointer to the class object to @code{void *} or
3195 to a type that can be safely accessed by the raw memory function suppresses
3198 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
3199 @opindex Wnon-virtual-dtor
3200 @opindex Wno-non-virtual-dtor
3201 Warn when a class has virtual functions and an accessible non-virtual
3202 destructor itself or in an accessible polymorphic base class, in which
3203 case it is possible but unsafe to delete an instance of a derived
3204 class through a pointer to the class itself or base class. This
3205 warning is automatically enabled if @option{-Weffc++} is specified.
3207 @item -Wregister @r{(C++ and Objective-C++ only)}
3209 @opindex Wno-register
3210 Warn on uses of the @code{register} storage class specifier, except
3211 when it is part of the GNU @ref{Explicit Register Variables} extension.
3212 The use of the @code{register} keyword as storage class specifier has
3213 been deprecated in C++11 and removed in C++17.
3214 Enabled by default with @option{-std=c++17}.
3216 @item -Wreorder @r{(C++ and Objective-C++ only)}
3218 @opindex Wno-reorder
3219 @cindex reordering, warning
3220 @cindex warning for reordering of member initializers
3221 Warn when the order of member initializers given in the code does not
3222 match the order in which they must be executed. For instance:
3228 A(): j (0), i (1) @{ @}
3233 The compiler rearranges the member initializers for @code{i}
3234 and @code{j} to match the declaration order of the members, emitting
3235 a warning to that effect. This warning is enabled by @option{-Wall}.
3237 @item -Wno-pessimizing-move @r{(C++ and Objective-C++ only)}
3238 @opindex Wpessimizing-move
3239 @opindex Wno-pessimizing-move
3240 This warning warns when a call to @code{std::move} prevents copy
3241 elision. A typical scenario when copy elision can occur is when returning in
3242 a function with a class return type, when the expression being returned is the
3243 name of a non-volatile automatic object, and is not a function parameter, and
3244 has the same type as the function return type.
3254 return std::move (t);
3258 But in this example, the @code{std::move} call prevents copy elision.
3260 This warning is enabled by @option{-Wall}.
3262 @item -Wno-redundant-move @r{(C++ and Objective-C++ only)}
3263 @opindex Wredundant-move
3264 @opindex Wno-redundant-move
3265 This warning warns about redundant calls to @code{std::move}; that is, when
3266 a move operation would have been performed even without the @code{std::move}
3267 call. This happens because the compiler is forced to treat the object as if
3268 it were an rvalue in certain situations such as returning a local variable,
3269 where copy elision isn't applicable. Consider:
3278 return std::move (t);
3282 Here, the @code{std::move} call is redundant. Because G++ implements Core
3283 Issue 1579, another example is:
3286 struct T @{ // convertible to U
3296 return std::move (t);
3299 In this example, copy elision isn't applicable because the type of the
3300 expression being returned and the function return type differ, yet G++
3301 treats the return value as if it were designated by an rvalue.
3303 This warning is enabled by @option{-Wextra}.
3305 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
3306 @opindex fext-numeric-literals
3307 @opindex fno-ext-numeric-literals
3308 Accept imaginary, fixed-point, or machine-defined
3309 literal number suffixes as GNU extensions.
3310 When this option is turned off these suffixes are treated
3311 as C++11 user-defined literal numeric suffixes.
3312 This is on by default for all pre-C++11 dialects and all GNU dialects:
3313 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
3314 @option{-std=gnu++14}.
3315 This option is off by default
3316 for ISO C++11 onwards (@option{-std=c++11}, ...).
3319 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
3322 @item -Weffc++ @r{(C++ and Objective-C++ only)}
3325 Warn about violations of the following style guidelines from Scott Meyers'
3326 @cite{Effective C++} series of books:
3330 Define a copy constructor and an assignment operator for classes
3331 with dynamically-allocated memory.
3334 Prefer initialization to assignment in constructors.
3337 Have @code{operator=} return a reference to @code{*this}.
3340 Don't try to return a reference when you must return an object.
3343 Distinguish between prefix and postfix forms of increment and
3344 decrement operators.
3347 Never overload @code{&&}, @code{||}, or @code{,}.
3351 This option also enables @option{-Wnon-virtual-dtor}, which is also
3352 one of the effective C++ recommendations. However, the check is
3353 extended to warn about the lack of virtual destructor in accessible
3354 non-polymorphic bases classes too.
3356 When selecting this option, be aware that the standard library
3357 headers do not obey all of these guidelines; use @samp{grep -v}
3358 to filter out those warnings.
3360 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
3361 @opindex Wstrict-null-sentinel
3362 @opindex Wno-strict-null-sentinel
3363 Warn about the use of an uncasted @code{NULL} as sentinel. When
3364 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
3365 to @code{__null}. Although it is a null pointer constant rather than a
3366 null pointer, it is guaranteed to be of the same size as a pointer.
3367 But this use is not portable across different compilers.
3369 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
3370 @opindex Wno-non-template-friend
3371 @opindex Wnon-template-friend
3372 Disable warnings when non-template friend functions are declared
3373 within a template. In very old versions of GCC that predate implementation
3374 of the ISO standard, declarations such as
3375 @samp{friend int foo(int)}, where the name of the friend is an unqualified-id,
3376 could be interpreted as a particular specialization of a template
3377 function; the warning exists to diagnose compatibility problems,
3378 and is enabled by default.
3380 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
3381 @opindex Wold-style-cast
3382 @opindex Wno-old-style-cast
3383 Warn if an old-style (C-style) cast to a non-void type is used within
3384 a C++ program. The new-style casts (@code{dynamic_cast},
3385 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
3386 less vulnerable to unintended effects and much easier to search for.
3388 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
3389 @opindex Woverloaded-virtual
3390 @opindex Wno-overloaded-virtual
3391 @cindex overloaded virtual function, warning
3392 @cindex warning for overloaded virtual function
3393 Warn when a function declaration hides virtual functions from a
3394 base class. For example, in:
3401 struct B: public A @{
3406 the @code{A} class version of @code{f} is hidden in @code{B}, and code
3417 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
3418 @opindex Wno-pmf-conversions
3419 @opindex Wpmf-conversions
3420 Disable the diagnostic for converting a bound pointer to member function
3423 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
3424 @opindex Wsign-promo
3425 @opindex Wno-sign-promo
3426 Warn when overload resolution chooses a promotion from unsigned or
3427 enumerated type to a signed type, over a conversion to an unsigned type of
3428 the same size. Previous versions of G++ tried to preserve
3429 unsignedness, but the standard mandates the current behavior.
3431 @item -Wtemplates @r{(C++ and Objective-C++ only)}
3433 @opindex Wno-templates
3434 Warn when a primary template declaration is encountered. Some coding
3435 rules disallow templates, and this may be used to enforce that rule.
3436 The warning is inactive inside a system header file, such as the STL, so
3437 one can still use the STL. One may also instantiate or specialize
3440 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
3441 @opindex Wmultiple-inheritance
3442 @opindex Wno-multiple-inheritance
3443 Warn when a class is defined with multiple direct base classes. Some
3444 coding rules disallow multiple inheritance, and this may be used to
3445 enforce that rule. The warning is inactive inside a system header file,
3446 such as the STL, so one can still use the STL. One may also define
3447 classes that indirectly use multiple inheritance.
3449 @item -Wvirtual-inheritance
3450 @opindex Wvirtual-inheritance
3451 @opindex Wno-virtual-inheritance
3452 Warn when a class is defined with a virtual direct base class. Some
3453 coding rules disallow multiple inheritance, and this may be used to
3454 enforce that rule. The warning is inactive inside a system header file,
3455 such as the STL, so one can still use the STL. One may also define
3456 classes that indirectly use virtual inheritance.
3459 @opindex Wnamespaces
3460 @opindex Wno-namespaces
3461 Warn when a namespace definition is opened. Some coding rules disallow
3462 namespaces, and this may be used to enforce that rule. The warning is
3463 inactive inside a system header file, such as the STL, so one can still
3464 use the STL. One may also use using directives and qualified names.
3466 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
3468 @opindex Wno-terminate
3469 Disable the warning about a throw-expression that will immediately
3470 result in a call to @code{terminate}.
3472 @item -Wno-class-conversion @r{(C++ and Objective-C++ only)}
3473 @opindex Wno-class-conversion
3474 @opindex Wclass-conversion
3475 Disable the warning about the case when a conversion function converts an
3476 object to the same type, to a base class of that type, or to void; such
3477 a conversion function will never be called.
3480 @node Objective-C and Objective-C++ Dialect Options
3481 @section Options Controlling Objective-C and Objective-C++ Dialects
3483 @cindex compiler options, Objective-C and Objective-C++
3484 @cindex Objective-C and Objective-C++ options, command-line
3485 @cindex options, Objective-C and Objective-C++
3486 (NOTE: This manual does not describe the Objective-C and Objective-C++
3487 languages themselves. @xref{Standards,,Language Standards
3488 Supported by GCC}, for references.)
3490 This section describes the command-line options that are only meaningful
3491 for Objective-C and Objective-C++ programs. You can also use most of
3492 the language-independent GNU compiler options.
3493 For example, you might compile a file @file{some_class.m} like this:
3496 gcc -g -fgnu-runtime -O -c some_class.m
3500 In this example, @option{-fgnu-runtime} is an option meant only for
3501 Objective-C and Objective-C++ programs; you can use the other options with
3502 any language supported by GCC@.
3504 Note that since Objective-C is an extension of the C language, Objective-C
3505 compilations may also use options specific to the C front-end (e.g.,
3506 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3507 C++-specific options (e.g., @option{-Wabi}).
3509 Here is a list of options that are @emph{only} for compiling Objective-C
3510 and Objective-C++ programs:
3513 @item -fconstant-string-class=@var{class-name}
3514 @opindex fconstant-string-class
3515 Use @var{class-name} as the name of the class to instantiate for each
3516 literal string specified with the syntax @code{@@"@dots{}"}. The default
3517 class name is @code{NXConstantString} if the GNU runtime is being used, and
3518 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3519 @option{-fconstant-cfstrings} option, if also present, overrides the
3520 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3521 to be laid out as constant CoreFoundation strings.
3524 @opindex fgnu-runtime
3525 Generate object code compatible with the standard GNU Objective-C
3526 runtime. This is the default for most types of systems.
3528 @item -fnext-runtime
3529 @opindex fnext-runtime
3530 Generate output compatible with the NeXT runtime. This is the default
3531 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3532 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3535 @item -fno-nil-receivers
3536 @opindex fno-nil-receivers
3537 @opindex fnil-receivers
3538 Assume that all Objective-C message dispatches (@code{[receiver
3539 message:arg]}) in this translation unit ensure that the receiver is
3540 not @code{nil}. This allows for more efficient entry points in the
3541 runtime to be used. This option is only available in conjunction with
3542 the NeXT runtime and ABI version 0 or 1.
3544 @item -fobjc-abi-version=@var{n}
3545 @opindex fobjc-abi-version
3546 Use version @var{n} of the Objective-C ABI for the selected runtime.
3547 This option is currently supported only for the NeXT runtime. In that
3548 case, Version 0 is the traditional (32-bit) ABI without support for
3549 properties and other Objective-C 2.0 additions. Version 1 is the
3550 traditional (32-bit) ABI with support for properties and other
3551 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3552 nothing is specified, the default is Version 0 on 32-bit target
3553 machines, and Version 2 on 64-bit target machines.
3555 @item -fobjc-call-cxx-cdtors
3556 @opindex fobjc-call-cxx-cdtors
3557 For each Objective-C class, check if any of its instance variables is a
3558 C++ object with a non-trivial default constructor. If so, synthesize a
3559 special @code{- (id) .cxx_construct} instance method which runs
3560 non-trivial default constructors on any such instance variables, in order,
3561 and then return @code{self}. Similarly, check if any instance variable
3562 is a C++ object with a non-trivial destructor, and if so, synthesize a
3563 special @code{- (void) .cxx_destruct} method which runs
3564 all such default destructors, in reverse order.
3566 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3567 methods thusly generated only operate on instance variables
3568 declared in the current Objective-C class, and not those inherited
3569 from superclasses. It is the responsibility of the Objective-C
3570 runtime to invoke all such methods in an object's inheritance
3571 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3572 by the runtime immediately after a new object instance is allocated;
3573 the @code{- (void) .cxx_destruct} methods are invoked immediately
3574 before the runtime deallocates an object instance.
3576 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3577 support for invoking the @code{- (id) .cxx_construct} and
3578 @code{- (void) .cxx_destruct} methods.
3580 @item -fobjc-direct-dispatch
3581 @opindex fobjc-direct-dispatch
3582 Allow fast jumps to the message dispatcher. On Darwin this is
3583 accomplished via the comm page.
3585 @item -fobjc-exceptions
3586 @opindex fobjc-exceptions
3587 Enable syntactic support for structured exception handling in
3588 Objective-C, similar to what is offered by C++. This option
3589 is required to use the Objective-C keywords @code{@@try},
3590 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3591 @code{@@synchronized}. This option is available with both the GNU
3592 runtime and the NeXT runtime (but not available in conjunction with
3593 the NeXT runtime on Mac OS X 10.2 and earlier).
3597 Enable garbage collection (GC) in Objective-C and Objective-C++
3598 programs. This option is only available with the NeXT runtime; the
3599 GNU runtime has a different garbage collection implementation that
3600 does not require special compiler flags.
3602 @item -fobjc-nilcheck
3603 @opindex fobjc-nilcheck
3604 For the NeXT runtime with version 2 of the ABI, check for a nil
3605 receiver in method invocations before doing the actual method call.
3606 This is the default and can be disabled using
3607 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3608 checked for nil in this way no matter what this flag is set to.
3609 Currently this flag does nothing when the GNU runtime, or an older
3610 version of the NeXT runtime ABI, is used.
3612 @item -fobjc-std=objc1
3614 Conform to the language syntax of Objective-C 1.0, the language
3615 recognized by GCC 4.0. This only affects the Objective-C additions to
3616 the C/C++ language; it does not affect conformance to C/C++ standards,
3617 which is controlled by the separate C/C++ dialect option flags. When
3618 this option is used with the Objective-C or Objective-C++ compiler,
3619 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3620 This is useful if you need to make sure that your Objective-C code can
3621 be compiled with older versions of GCC@.
3623 @item -freplace-objc-classes
3624 @opindex freplace-objc-classes
3625 Emit a special marker instructing @command{ld(1)} not to statically link in
3626 the resulting object file, and allow @command{dyld(1)} to load it in at
3627 run time instead. This is used in conjunction with the Fix-and-Continue
3628 debugging mode, where the object file in question may be recompiled and
3629 dynamically reloaded in the course of program execution, without the need
3630 to restart the program itself. Currently, Fix-and-Continue functionality
3631 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3636 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3637 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3638 compile time) with static class references that get initialized at load time,
3639 which improves run-time performance. Specifying the @option{-fzero-link} flag
3640 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3641 to be retained. This is useful in Zero-Link debugging mode, since it allows
3642 for individual class implementations to be modified during program execution.
3643 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3644 regardless of command-line options.
3646 @item -fno-local-ivars
3647 @opindex fno-local-ivars
3648 @opindex flocal-ivars
3649 By default instance variables in Objective-C can be accessed as if
3650 they were local variables from within the methods of the class they're
3651 declared in. This can lead to shadowing between instance variables
3652 and other variables declared either locally inside a class method or
3653 globally with the same name. Specifying the @option{-fno-local-ivars}
3654 flag disables this behavior thus avoiding variable shadowing issues.
3656 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3657 @opindex fivar-visibility
3658 Set the default instance variable visibility to the specified option
3659 so that instance variables declared outside the scope of any access
3660 modifier directives default to the specified visibility.
3664 Dump interface declarations for all classes seen in the source file to a
3665 file named @file{@var{sourcename}.decl}.
3667 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3668 @opindex Wassign-intercept
3669 @opindex Wno-assign-intercept
3670 Warn whenever an Objective-C assignment is being intercepted by the
3673 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3674 @opindex Wno-protocol
3676 If a class is declared to implement a protocol, a warning is issued for
3677 every method in the protocol that is not implemented by the class. The
3678 default behavior is to issue a warning for every method not explicitly
3679 implemented in the class, even if a method implementation is inherited
3680 from the superclass. If you use the @option{-Wno-protocol} option, then
3681 methods inherited from the superclass are considered to be implemented,
3682 and no warning is issued for them.
3684 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3686 @opindex Wno-selector
3687 Warn if multiple methods of different types for the same selector are
3688 found during compilation. The check is performed on the list of methods
3689 in the final stage of compilation. Additionally, a check is performed
3690 for each selector appearing in a @code{@@selector(@dots{})}
3691 expression, and a corresponding method for that selector has been found
3692 during compilation. Because these checks scan the method table only at
3693 the end of compilation, these warnings are not produced if the final
3694 stage of compilation is not reached, for example because an error is
3695 found during compilation, or because the @option{-fsyntax-only} option is
3698 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3699 @opindex Wstrict-selector-match
3700 @opindex Wno-strict-selector-match
3701 Warn if multiple methods with differing argument and/or return types are
3702 found for a given selector when attempting to send a message using this
3703 selector to a receiver of type @code{id} or @code{Class}. When this flag
3704 is off (which is the default behavior), the compiler omits such warnings
3705 if any differences found are confined to types that share the same size
3708 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3709 @opindex Wundeclared-selector
3710 @opindex Wno-undeclared-selector
3711 Warn if a @code{@@selector(@dots{})} expression referring to an
3712 undeclared selector is found. A selector is considered undeclared if no
3713 method with that name has been declared before the
3714 @code{@@selector(@dots{})} expression, either explicitly in an
3715 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3716 an @code{@@implementation} section. This option always performs its
3717 checks as soon as a @code{@@selector(@dots{})} expression is found,
3718 while @option{-Wselector} only performs its checks in the final stage of
3719 compilation. This also enforces the coding style convention
3720 that methods and selectors must be declared before being used.
3722 @item -print-objc-runtime-info
3723 @opindex print-objc-runtime-info
3724 Generate C header describing the largest structure that is passed by
3729 @node Diagnostic Message Formatting Options
3730 @section Options to Control Diagnostic Messages Formatting
3731 @cindex options to control diagnostics formatting
3732 @cindex diagnostic messages
3733 @cindex message formatting
3735 Traditionally, diagnostic messages have been formatted irrespective of
3736 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3737 options described below
3738 to control the formatting algorithm for diagnostic messages,
3739 e.g.@: how many characters per line, how often source location
3740 information should be reported. Note that some language front ends may not
3741 honor these options.
3744 @item -fmessage-length=@var{n}
3745 @opindex fmessage-length
3746 Try to format error messages so that they fit on lines of about
3747 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3748 done; each error message appears on a single line. This is the
3749 default for all front ends.
3751 Note - this option also affects the display of the @samp{#error} and
3752 @samp{#warning} pre-processor directives, and the @samp{deprecated}
3753 function/type/variable attribute. It does not however affect the
3754 @samp{pragma GCC warning} and @samp{pragma GCC error} pragmas.
3756 @item -fdiagnostics-show-location=once
3757 @opindex fdiagnostics-show-location
3758 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3759 reporter to emit source location information @emph{once}; that is, in
3760 case the message is too long to fit on a single physical line and has to
3761 be wrapped, the source location won't be emitted (as prefix) again,
3762 over and over, in subsequent continuation lines. This is the default
3765 @item -fdiagnostics-show-location=every-line
3766 Only meaningful in line-wrapping mode. Instructs the diagnostic
3767 messages reporter to emit the same source location information (as
3768 prefix) for physical lines that result from the process of breaking
3769 a message which is too long to fit on a single line.
3771 @item -fdiagnostics-color[=@var{WHEN}]
3772 @itemx -fno-diagnostics-color
3773 @opindex fdiagnostics-color
3774 @cindex highlight, color
3775 @vindex GCC_COLORS @r{environment variable}
3776 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3777 or @samp{auto}. The default depends on how the compiler has been configured,
3778 it can be any of the above @var{WHEN} options or also @samp{never}
3779 if @env{GCC_COLORS} environment variable isn't present in the environment,
3780 and @samp{auto} otherwise.
3781 @samp{auto} means to use color only when the standard error is a terminal.
3782 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3783 aliases for @option{-fdiagnostics-color=always} and
3784 @option{-fdiagnostics-color=never}, respectively.
3786 The colors are defined by the environment variable @env{GCC_COLORS}.
3787 Its value is a colon-separated list of capabilities and Select Graphic
3788 Rendition (SGR) substrings. SGR commands are interpreted by the
3789 terminal or terminal emulator. (See the section in the documentation
3790 of your text terminal for permitted values and their meanings as
3791 character attributes.) These substring values are integers in decimal
3792 representation and can be concatenated with semicolons.
3793 Common values to concatenate include
3795 @samp{4} for underline,
3797 @samp{7} for inverse,
3798 @samp{39} for default foreground color,
3799 @samp{30} to @samp{37} for foreground colors,
3800 @samp{90} to @samp{97} for 16-color mode foreground colors,
3801 @samp{38;5;0} to @samp{38;5;255}
3802 for 88-color and 256-color modes foreground colors,
3803 @samp{49} for default background color,
3804 @samp{40} to @samp{47} for background colors,
3805 @samp{100} to @samp{107} for 16-color mode background colors,
3806 and @samp{48;5;0} to @samp{48;5;255}
3807 for 88-color and 256-color modes background colors.
3809 The default @env{GCC_COLORS} is
3811 error=01;31:warning=01;35:note=01;36:range1=32:range2=34:locus=01:\
3812 quote=01:fixit-insert=32:fixit-delete=31:\
3813 diff-filename=01:diff-hunk=32:diff-delete=31:diff-insert=32:\
3817 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3818 @samp{01;36} is bold cyan, @samp{32} is green, @samp{34} is blue,
3819 @samp{01} is bold, and @samp{31} is red.
3820 Setting @env{GCC_COLORS} to the empty string disables colors.
3821 Supported capabilities are as follows.
3825 @vindex error GCC_COLORS @r{capability}
3826 SGR substring for error: markers.
3829 @vindex warning GCC_COLORS @r{capability}
3830 SGR substring for warning: markers.
3833 @vindex note GCC_COLORS @r{capability}
3834 SGR substring for note: markers.
3837 @vindex range1 GCC_COLORS @r{capability}
3838 SGR substring for first additional range.
3841 @vindex range2 GCC_COLORS @r{capability}
3842 SGR substring for second additional range.
3845 @vindex locus GCC_COLORS @r{capability}
3846 SGR substring for location information, @samp{file:line} or
3847 @samp{file:line:column} etc.
3850 @vindex quote GCC_COLORS @r{capability}
3851 SGR substring for information printed within quotes.
3854 @vindex fixit-insert GCC_COLORS @r{capability}
3855 SGR substring for fix-it hints suggesting text to
3856 be inserted or replaced.
3859 @vindex fixit-delete GCC_COLORS @r{capability}
3860 SGR substring for fix-it hints suggesting text to
3863 @item diff-filename=
3864 @vindex diff-filename GCC_COLORS @r{capability}
3865 SGR substring for filename headers within generated patches.
3868 @vindex diff-hunk GCC_COLORS @r{capability}
3869 SGR substring for the starts of hunks within generated patches.
3872 @vindex diff-delete GCC_COLORS @r{capability}
3873 SGR substring for deleted lines within generated patches.
3876 @vindex diff-insert GCC_COLORS @r{capability}
3877 SGR substring for inserted lines within generated patches.
3880 @vindex type-diff GCC_COLORS @r{capability}
3881 SGR substring for highlighting mismatching types within template
3882 arguments in the C++ frontend.
3885 @item -fno-diagnostics-show-option
3886 @opindex fno-diagnostics-show-option
3887 @opindex fdiagnostics-show-option
3888 By default, each diagnostic emitted includes text indicating the
3889 command-line option that directly controls the diagnostic (if such an
3890 option is known to the diagnostic machinery). Specifying the
3891 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3893 @item -fno-diagnostics-show-caret
3894 @opindex fno-diagnostics-show-caret
3895 @opindex fdiagnostics-show-caret
3896 By default, each diagnostic emitted includes the original source line
3897 and a caret @samp{^} indicating the column. This option suppresses this
3898 information. The source line is truncated to @var{n} characters, if
3899 the @option{-fmessage-length=n} option is given. When the output is done
3900 to the terminal, the width is limited to the width given by the
3901 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3903 @item -fno-diagnostics-show-labels
3904 @opindex fno-diagnostics-show-labels
3905 @opindex fdiagnostics-show-labels
3906 By default, when printing source code (via @option{-fdiagnostics-show-caret}),
3907 diagnostics can label ranges of source code with pertinent information, such
3908 as the types of expressions:
3911 printf ("foo %s bar", long_i + long_j);
3917 This option suppresses the printing of these labels (in the example above,
3918 the vertical bars and the ``char *'' and ``long int'' text).
3920 @item -fno-diagnostics-show-line-numbers
3921 @opindex fno-diagnostics-show-line-numbers
3922 @opindex fdiagnostics-show-line-numbers
3923 By default, when printing source code (via @option{-fdiagnostics-show-caret}),
3924 a left margin is printed, showing line numbers. This option suppresses this
3927 @item -fdiagnostics-minimum-margin-width=@var{width}
3928 @opindex fdiagnostics-minimum-margin-width
3929 This option controls the minimum width of the left margin printed by
3930 @option{-fdiagnostics-show-line-numbers}. It defaults to 6.
3932 @item -fdiagnostics-parseable-fixits
3933 @opindex fdiagnostics-parseable-fixits
3934 Emit fix-it hints in a machine-parseable format, suitable for consumption
3935 by IDEs. For each fix-it, a line will be printed after the relevant
3936 diagnostic, starting with the string ``fix-it:''. For example:
3939 fix-it:"test.c":@{45:3-45:21@}:"gtk_widget_show_all"
3942 The location is expressed as a half-open range, expressed as a count of
3943 bytes, starting at byte 1 for the initial column. In the above example,
3944 bytes 3 through 20 of line 45 of ``test.c'' are to be replaced with the
3948 00000000011111111112222222222
3949 12345678901234567890123456789
3950 gtk_widget_showall (dlg);
3955 The filename and replacement string escape backslash as ``\\", tab as ``\t'',
3956 newline as ``\n'', double quotes as ``\"'', non-printable characters as octal
3957 (e.g. vertical tab as ``\013'').
3959 An empty replacement string indicates that the given range is to be removed.
3960 An empty range (e.g. ``45:3-45:3'') indicates that the string is to
3961 be inserted at the given position.
3963 @item -fdiagnostics-generate-patch
3964 @opindex fdiagnostics-generate-patch
3965 Print fix-it hints to stderr in unified diff format, after any diagnostics
3966 are printed. For example:
3973 void show_cb(GtkDialog *dlg)
3975 - gtk_widget_showall(dlg);
3976 + gtk_widget_show_all(dlg);
3981 The diff may or may not be colorized, following the same rules
3982 as for diagnostics (see @option{-fdiagnostics-color}).
3984 @item -fdiagnostics-show-template-tree
3985 @opindex fdiagnostics-show-template-tree
3987 In the C++ frontend, when printing diagnostics showing mismatching
3988 template types, such as:
3991 could not convert 'std::map<int, std::vector<double> >()'
3992 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
3995 the @option{-fdiagnostics-show-template-tree} flag enables printing a
3996 tree-like structure showing the common and differing parts of the types,
4006 The parts that differ are highlighted with color (``double'' and
4007 ``float'' in this case).
4009 @item -fno-elide-type
4010 @opindex fno-elide-type
4011 @opindex felide-type
4012 By default when the C++ frontend prints diagnostics showing mismatching
4013 template types, common parts of the types are printed as ``[...]'' to
4014 simplify the error message. For example:
4017 could not convert 'std::map<int, std::vector<double> >()'
4018 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
4021 Specifying the @option{-fno-elide-type} flag suppresses that behavior.
4022 This flag also affects the output of the
4023 @option{-fdiagnostics-show-template-tree} flag.
4025 @item -fno-show-column
4026 @opindex fno-show-column
4027 @opindex fshow-column
4028 Do not print column numbers in diagnostics. This may be necessary if
4029 diagnostics are being scanned by a program that does not understand the
4030 column numbers, such as @command{dejagnu}.
4032 @item -fdiagnostics-format=@var{FORMAT}
4033 @opindex fdiagnostics-format
4034 Select a different format for printing diagnostics.
4035 @var{FORMAT} is @samp{text} or @samp{json}.
4036 The default is @samp{text}.
4038 The @samp{json} format consists of a top-level JSON array containing JSON
4039 objects representing the diagnostics.
4041 The JSON is emitted as one line, without formatting; the examples below
4042 have been formatted for clarity.
4044 Diagnostics can have child diagnostics. For example, this error and note:
4047 misleading-indentation.c:15:3: warning: this 'if' clause does not
4048 guard... [-Wmisleading-indentation]
4051 misleading-indentation.c:17:5: note: ...this statement, but the latter
4052 is misleadingly indented as if it were guarded by the 'if'
4058 might be printed in JSON form (after formatting) like this:
4068 "file": "misleading-indentation.c",
4073 "file": "misleading-indentation.c",
4078 "message": "this \u2018if\u2019 clause does not guard...",
4079 "option": "-Wmisleading-indentation",
4087 "file": "misleading-indentation.c",
4092 "message": "...this statement, but the latter is @dots{}"
4101 where the @code{note} is a child of the @code{warning}.
4103 A diagnostic has a @code{kind}. If this is @code{warning}, then there is
4104 an @code{option} key describing the command-line option controlling the
4107 A diagnostic can contain zero or more locations. Each location has up
4108 to three positions within it: a @code{caret} position and optional
4109 @code{start} and @code{finish} positions. A location can also have
4110 an optional @code{label} string. For example, this error:
4113 bad-binary-ops.c:64:23: error: invalid operands to binary + (have 'S' @{aka
4114 'struct s'@} and 'T' @{aka 'struct t'@})
4115 64 | return callee_4a () + callee_4b ();
4116 | ~~~~~~~~~~~~ ^ ~~~~~~~~~~~~
4118 | | T @{aka struct t@}
4119 | S @{aka struct s@}
4123 has three locations. Its primary location is at the ``+'' token at column
4124 23. It has two secondary locations, describing the left and right-hand sides
4125 of the expression, which have labels. It might be printed in JSON form as:
4134 "column": 23, "file": "bad-binary-ops.c", "line": 64
4139 "column": 10, "file": "bad-binary-ops.c", "line": 64
4142 "column": 21, "file": "bad-binary-ops.c", "line": 64
4144 "label": "S @{aka struct s@}"
4148 "column": 25, "file": "bad-binary-ops.c", "line": 64
4151 "column": 36, "file": "bad-binary-ops.c", "line": 64
4153 "label": "T @{aka struct t@}"
4156 "message": "invalid operands to binary + @dots{}"
4160 If a diagnostic contains fix-it hints, it has a @code{fixits} array,
4161 consisting of half-open intervals, similar to the output of
4162 @option{-fdiagnostics-parseable-fixits}. For example, this diagnostic
4163 with a replacement fix-it hint:
4166 demo.c:8:15: error: 'struct s' has no member named 'colour'; did you
4168 8 | return ptr->colour;
4174 might be printed in JSON form as:
4209 "message": "\u2018struct s\u2019 has no member named @dots{}"
4214 where the fix-it hint suggests replacing the text from @code{start} up
4215 to but not including @code{next} with @code{string}'s value. Deletions
4216 are expressed via an empty value for @code{string}, insertions by
4217 having @code{start} equal @code{next}.
4221 @node Warning Options
4222 @section Options to Request or Suppress Warnings
4223 @cindex options to control warnings
4224 @cindex warning messages
4225 @cindex messages, warning
4226 @cindex suppressing warnings
4228 Warnings are diagnostic messages that report constructions that
4229 are not inherently erroneous but that are risky or suggest there
4230 may have been an error.
4232 The following language-independent options do not enable specific
4233 warnings but control the kinds of diagnostics produced by GCC@.
4236 @cindex syntax checking
4238 @opindex fsyntax-only
4239 Check the code for syntax errors, but don't do anything beyond that.
4241 @item -fmax-errors=@var{n}
4242 @opindex fmax-errors
4243 Limits the maximum number of error messages to @var{n}, at which point
4244 GCC bails out rather than attempting to continue processing the source
4245 code. If @var{n} is 0 (the default), there is no limit on the number
4246 of error messages produced. If @option{-Wfatal-errors} is also
4247 specified, then @option{-Wfatal-errors} takes precedence over this
4252 Inhibit all warning messages.
4257 Make all warnings into errors.
4262 Make the specified warning into an error. The specifier for a warning
4263 is appended; for example @option{-Werror=switch} turns the warnings
4264 controlled by @option{-Wswitch} into errors. This switch takes a
4265 negative form, to be used to negate @option{-Werror} for specific
4266 warnings; for example @option{-Wno-error=switch} makes
4267 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
4270 The warning message for each controllable warning includes the
4271 option that controls the warning. That option can then be used with
4272 @option{-Werror=} and @option{-Wno-error=} as described above.
4273 (Printing of the option in the warning message can be disabled using the
4274 @option{-fno-diagnostics-show-option} flag.)
4276 Note that specifying @option{-Werror=}@var{foo} automatically implies
4277 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
4280 @item -Wfatal-errors
4281 @opindex Wfatal-errors
4282 @opindex Wno-fatal-errors
4283 This option causes the compiler to abort compilation on the first error
4284 occurred rather than trying to keep going and printing further error
4289 You can request many specific warnings with options beginning with
4290 @samp{-W}, for example @option{-Wimplicit} to request warnings on
4291 implicit declarations. Each of these specific warning options also
4292 has a negative form beginning @samp{-Wno-} to turn off warnings; for
4293 example, @option{-Wno-implicit}. This manual lists only one of the
4294 two forms, whichever is not the default. For further
4295 language-specific options also refer to @ref{C++ Dialect Options} and
4296 @ref{Objective-C and Objective-C++ Dialect Options}.
4298 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
4299 options, such as @option{-Wunused}, which may turn on further options,
4300 such as @option{-Wunused-value}. The combined effect of positive and
4301 negative forms is that more specific options have priority over less
4302 specific ones, independently of their position in the command-line. For
4303 options of the same specificity, the last one takes effect. Options
4304 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
4305 as if they appeared at the end of the command-line.
4307 When an unrecognized warning option is requested (e.g.,
4308 @option{-Wunknown-warning}), GCC emits a diagnostic stating
4309 that the option is not recognized. However, if the @option{-Wno-} form
4310 is used, the behavior is slightly different: no diagnostic is
4311 produced for @option{-Wno-unknown-warning} unless other diagnostics
4312 are being produced. This allows the use of new @option{-Wno-} options
4313 with old compilers, but if something goes wrong, the compiler
4314 warns that an unrecognized option is present.
4321 @opindex Wno-pedantic
4322 Issue all the warnings demanded by strict ISO C and ISO C++;
4323 reject all programs that use forbidden extensions, and some other
4324 programs that do not follow ISO C and ISO C++. For ISO C, follows the
4325 version of the ISO C standard specified by any @option{-std} option used.
4327 Valid ISO C and ISO C++ programs should compile properly with or without
4328 this option (though a rare few require @option{-ansi} or a
4329 @option{-std} option specifying the required version of ISO C)@. However,
4330 without this option, certain GNU extensions and traditional C and C++
4331 features are supported as well. With this option, they are rejected.
4333 @option{-Wpedantic} does not cause warning messages for use of the
4334 alternate keywords whose names begin and end with @samp{__}. Pedantic
4335 warnings are also disabled in the expression that follows
4336 @code{__extension__}. However, only system header files should use
4337 these escape routes; application programs should avoid them.
4338 @xref{Alternate Keywords}.
4340 Some users try to use @option{-Wpedantic} to check programs for strict ISO
4341 C conformance. They soon find that it does not do quite what they want:
4342 it finds some non-ISO practices, but not all---only those for which
4343 ISO C @emph{requires} a diagnostic, and some others for which
4344 diagnostics have been added.
4346 A feature to report any failure to conform to ISO C might be useful in
4347 some instances, but would require considerable additional work and would
4348 be quite different from @option{-Wpedantic}. We don't have plans to
4349 support such a feature in the near future.
4351 Where the standard specified with @option{-std} represents a GNU
4352 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
4353 corresponding @dfn{base standard}, the version of ISO C on which the GNU
4354 extended dialect is based. Warnings from @option{-Wpedantic} are given
4355 where they are required by the base standard. (It does not make sense
4356 for such warnings to be given only for features not in the specified GNU
4357 C dialect, since by definition the GNU dialects of C include all
4358 features the compiler supports with the given option, and there would be
4359 nothing to warn about.)
4361 @item -pedantic-errors
4362 @opindex pedantic-errors
4363 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
4364 requires a diagnostic, in some cases where there is undefined behavior
4365 at compile-time and in some other cases that do not prevent compilation
4366 of programs that are valid according to the standard. This is not
4367 equivalent to @option{-Werror=pedantic}, since there are errors enabled
4368 by this option and not enabled by the latter and vice versa.
4373 This enables all the warnings about constructions that some users
4374 consider questionable, and that are easy to avoid (or modify to
4375 prevent the warning), even in conjunction with macros. This also
4376 enables some language-specific warnings described in @ref{C++ Dialect
4377 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
4379 @option{-Wall} turns on the following warning flags:
4381 @gccoptlist{-Waddress @gol
4382 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
4384 -Wbool-operation @gol
4385 -Wc++11-compat -Wc++14-compat @gol
4386 -Wcatch-value @r{(C++ and Objective-C++ only)} @gol
4387 -Wchar-subscripts @gol
4389 -Wduplicate-decl-specifier @r{(C and Objective-C only)} @gol
4390 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
4392 -Wint-in-bool-context @gol
4393 -Wimplicit @r{(C and Objective-C only)} @gol
4394 -Wimplicit-int @r{(C and Objective-C only)} @gol
4395 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
4396 -Winit-self @r{(only for C++)} @gol
4397 -Wlogical-not-parentheses @gol
4398 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
4399 -Wmaybe-uninitialized @gol
4400 -Wmemset-elt-size @gol
4401 -Wmemset-transposed-args @gol
4402 -Wmisleading-indentation @r{(only for C/C++)} @gol
4403 -Wmissing-attributes @gol
4404 -Wmissing-braces @r{(only for C/ObjC)} @gol
4405 -Wmultistatement-macros @gol
4406 -Wnarrowing @r{(only for C++)} @gol
4408 -Wnonnull-compare @gol
4411 -Wpessimizing-move @r{(only for C++)} @gol
4416 -Wsequence-point @gol
4417 -Wsign-compare @r{(only in C++)} @gol
4418 -Wsizeof-pointer-div @gol
4419 -Wsizeof-pointer-memaccess @gol
4420 -Wstrict-aliasing @gol
4421 -Wstrict-overflow=1 @gol
4423 -Wtautological-compare @gol
4425 -Wuninitialized @gol
4426 -Wunknown-pragmas @gol
4427 -Wunused-function @gol
4430 -Wunused-variable @gol
4431 -Wvolatile-register-var}
4433 Note that some warning flags are not implied by @option{-Wall}. Some of
4434 them warn about constructions that users generally do not consider
4435 questionable, but which occasionally you might wish to check for;
4436 others warn about constructions that are necessary or hard to avoid in
4437 some cases, and there is no simple way to modify the code to suppress
4438 the warning. Some of them are enabled by @option{-Wextra} but many of
4439 them must be enabled individually.
4445 This enables some extra warning flags that are not enabled by
4446 @option{-Wall}. (This option used to be called @option{-W}. The older
4447 name is still supported, but the newer name is more descriptive.)
4449 @gccoptlist{-Wclobbered @gol
4450 -Wcast-function-type @gol
4451 -Wdeprecated-copy @r{(C++ only)} @gol
4453 -Wignored-qualifiers @gol
4454 -Wimplicit-fallthrough=3 @gol
4455 -Wmissing-field-initializers @gol
4456 -Wmissing-parameter-type @r{(C only)} @gol
4457 -Wold-style-declaration @r{(C only)} @gol
4458 -Woverride-init @gol
4459 -Wsign-compare @r{(C only)} @gol
4460 -Wredundant-move @r{(only for C++)} @gol
4462 -Wuninitialized @gol
4463 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
4464 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
4465 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)}}
4468 The option @option{-Wextra} also prints warning messages for the
4474 A pointer is compared against integer zero with @code{<}, @code{<=},
4475 @code{>}, or @code{>=}.
4478 (C++ only) An enumerator and a non-enumerator both appear in a
4479 conditional expression.
4482 (C++ only) Ambiguous virtual bases.
4485 (C++ only) Subscripting an array that has been declared @code{register}.
4488 (C++ only) Taking the address of a variable that has been declared
4492 (C++ only) A base class is not initialized in the copy constructor
4497 @item -Wchar-subscripts
4498 @opindex Wchar-subscripts
4499 @opindex Wno-char-subscripts
4500 Warn if an array subscript has type @code{char}. This is a common cause
4501 of error, as programmers often forget that this type is signed on some
4503 This warning is enabled by @option{-Wall}.
4505 @item -Wno-coverage-mismatch
4506 @opindex Wno-coverage-mismatch
4507 @opindex Wcoverage-mismatch
4508 Warn if feedback profiles do not match when using the
4509 @option{-fprofile-use} option.
4510 If a source file is changed between compiling with @option{-fprofile-generate}
4511 and with @option{-fprofile-use}, the files with the profile feedback can fail
4512 to match the source file and GCC cannot use the profile feedback
4513 information. By default, this warning is enabled and is treated as an
4514 error. @option{-Wno-coverage-mismatch} can be used to disable the
4515 warning or @option{-Wno-error=coverage-mismatch} can be used to
4516 disable the error. Disabling the error for this warning can result in
4517 poorly optimized code and is useful only in the
4518 case of very minor changes such as bug fixes to an existing code-base.
4519 Completely disabling the warning is not recommended.
4522 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
4524 Suppress warning messages emitted by @code{#warning} directives.
4526 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
4527 @opindex Wdouble-promotion
4528 @opindex Wno-double-promotion
4529 Give a warning when a value of type @code{float} is implicitly
4530 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
4531 floating-point unit implement @code{float} in hardware, but emulate
4532 @code{double} in software. On such a machine, doing computations
4533 using @code{double} values is much more expensive because of the
4534 overhead required for software emulation.
4536 It is easy to accidentally do computations with @code{double} because
4537 floating-point literals are implicitly of type @code{double}. For
4541 float area(float radius)
4543 return 3.14159 * radius * radius;
4547 the compiler performs the entire computation with @code{double}
4548 because the floating-point literal is a @code{double}.
4550 @item -Wduplicate-decl-specifier @r{(C and Objective-C only)}
4551 @opindex Wduplicate-decl-specifier
4552 @opindex Wno-duplicate-decl-specifier
4553 Warn if a declaration has duplicate @code{const}, @code{volatile},
4554 @code{restrict} or @code{_Atomic} specifier. This warning is enabled by
4558 @itemx -Wformat=@var{n}
4561 @opindex ffreestanding
4562 @opindex fno-builtin
4564 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
4565 the arguments supplied have types appropriate to the format string
4566 specified, and that the conversions specified in the format string make
4567 sense. This includes standard functions, and others specified by format
4568 attributes (@pxref{Function Attributes}), in the @code{printf},
4569 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
4570 not in the C standard) families (or other target-specific families).
4571 Which functions are checked without format attributes having been
4572 specified depends on the standard version selected, and such checks of
4573 functions without the attribute specified are disabled by
4574 @option{-ffreestanding} or @option{-fno-builtin}.
4576 The formats are checked against the format features supported by GNU
4577 libc version 2.2. These include all ISO C90 and C99 features, as well
4578 as features from the Single Unix Specification and some BSD and GNU
4579 extensions. Other library implementations may not support all these
4580 features; GCC does not support warning about features that go beyond a
4581 particular library's limitations. However, if @option{-Wpedantic} is used
4582 with @option{-Wformat}, warnings are given about format features not
4583 in the selected standard version (but not for @code{strfmon} formats,
4584 since those are not in any version of the C standard). @xref{C Dialect
4585 Options,,Options Controlling C Dialect}.
4592 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
4593 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
4594 @option{-Wformat} also checks for null format arguments for several
4595 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
4596 aspects of this level of format checking can be disabled by the
4597 options: @option{-Wno-format-contains-nul},
4598 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
4599 @option{-Wformat} is enabled by @option{-Wall}.
4601 @item -Wno-format-contains-nul
4602 @opindex Wno-format-contains-nul
4603 @opindex Wformat-contains-nul
4604 If @option{-Wformat} is specified, do not warn about format strings that
4607 @item -Wno-format-extra-args
4608 @opindex Wno-format-extra-args
4609 @opindex Wformat-extra-args
4610 If @option{-Wformat} is specified, do not warn about excess arguments to a
4611 @code{printf} or @code{scanf} format function. The C standard specifies
4612 that such arguments are ignored.
4614 Where the unused arguments lie between used arguments that are
4615 specified with @samp{$} operand number specifications, normally
4616 warnings are still given, since the implementation could not know what
4617 type to pass to @code{va_arg} to skip the unused arguments. However,
4618 in the case of @code{scanf} formats, this option suppresses the
4619 warning if the unused arguments are all pointers, since the Single
4620 Unix Specification says that such unused arguments are allowed.
4622 @item -Wformat-overflow
4623 @itemx -Wformat-overflow=@var{level}
4624 @opindex Wformat-overflow
4625 @opindex Wno-format-overflow
4626 Warn about calls to formatted input/output functions such as @code{sprintf}
4627 and @code{vsprintf} that might overflow the destination buffer. When the
4628 exact number of bytes written by a format directive cannot be determined
4629 at compile-time it is estimated based on heuristics that depend on the
4630 @var{level} argument and on optimization. While enabling optimization
4631 will in most cases improve the accuracy of the warning, it may also
4632 result in false positives.
4635 @item -Wformat-overflow
4636 @itemx -Wformat-overflow=1
4637 @opindex Wformat-overflow
4638 @opindex Wno-format-overflow
4639 Level @var{1} of @option{-Wformat-overflow} enabled by @option{-Wformat}
4640 employs a conservative approach that warns only about calls that most
4641 likely overflow the buffer. At this level, numeric arguments to format
4642 directives with unknown values are assumed to have the value of one, and
4643 strings of unknown length to be empty. Numeric arguments that are known
4644 to be bounded to a subrange of their type, or string arguments whose output
4645 is bounded either by their directive's precision or by a finite set of
4646 string literals, are assumed to take on the value within the range that
4647 results in the most bytes on output. For example, the call to @code{sprintf}
4648 below is diagnosed because even with both @var{a} and @var{b} equal to zero,
4649 the terminating NUL character (@code{'\0'}) appended by the function
4650 to the destination buffer will be written past its end. Increasing
4651 the size of the buffer by a single byte is sufficient to avoid the
4652 warning, though it may not be sufficient to avoid the overflow.
4655 void f (int a, int b)
4658 sprintf (buf, "a = %i, b = %i\n", a, b);
4662 @item -Wformat-overflow=2
4663 Level @var{2} warns also about calls that might overflow the destination
4664 buffer given an argument of sufficient length or magnitude. At level
4665 @var{2}, unknown numeric arguments are assumed to have the minimum
4666 representable value for signed types with a precision greater than 1, and
4667 the maximum representable value otherwise. Unknown string arguments whose
4668 length cannot be assumed to be bounded either by the directive's precision,
4669 or by a finite set of string literals they may evaluate to, or the character
4670 array they may point to, are assumed to be 1 character long.
4672 At level @var{2}, the call in the example above is again diagnosed, but
4673 this time because with @var{a} equal to a 32-bit @code{INT_MIN} the first
4674 @code{%i} directive will write some of its digits beyond the end of
4675 the destination buffer. To make the call safe regardless of the values
4676 of the two variables, the size of the destination buffer must be increased
4677 to at least 34 bytes. GCC includes the minimum size of the buffer in
4678 an informational note following the warning.
4680 An alternative to increasing the size of the destination buffer is to
4681 constrain the range of formatted values. The maximum length of string
4682 arguments can be bounded by specifying the precision in the format
4683 directive. When numeric arguments of format directives can be assumed
4684 to be bounded by less than the precision of their type, choosing
4685 an appropriate length modifier to the format specifier will reduce
4686 the required buffer size. For example, if @var{a} and @var{b} in the
4687 example above can be assumed to be within the precision of
4688 the @code{short int} type then using either the @code{%hi} format
4689 directive or casting the argument to @code{short} reduces the maximum
4690 required size of the buffer to 24 bytes.
4693 void f (int a, int b)
4696 sprintf (buf, "a = %hi, b = %i\n", a, (short)b);
4701 @item -Wno-format-zero-length
4702 @opindex Wno-format-zero-length
4703 @opindex Wformat-zero-length
4704 If @option{-Wformat} is specified, do not warn about zero-length formats.
4705 The C standard specifies that zero-length formats are allowed.
4710 Enable @option{-Wformat} plus additional format checks. Currently
4711 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
4714 @item -Wformat-nonliteral
4715 @opindex Wformat-nonliteral
4716 @opindex Wno-format-nonliteral
4717 If @option{-Wformat} is specified, also warn if the format string is not a
4718 string literal and so cannot be checked, unless the format function
4719 takes its format arguments as a @code{va_list}.
4721 @item -Wformat-security
4722 @opindex Wformat-security
4723 @opindex Wno-format-security
4724 If @option{-Wformat} is specified, also warn about uses of format
4725 functions that represent possible security problems. At present, this
4726 warns about calls to @code{printf} and @code{scanf} functions where the
4727 format string is not a string literal and there are no format arguments,
4728 as in @code{printf (foo);}. This may be a security hole if the format
4729 string came from untrusted input and contains @samp{%n}. (This is
4730 currently a subset of what @option{-Wformat-nonliteral} warns about, but
4731 in future warnings may be added to @option{-Wformat-security} that are not
4732 included in @option{-Wformat-nonliteral}.)
4734 @item -Wformat-signedness
4735 @opindex Wformat-signedness
4736 @opindex Wno-format-signedness
4737 If @option{-Wformat} is specified, also warn if the format string
4738 requires an unsigned argument and the argument is signed and vice versa.
4740 @item -Wformat-truncation
4741 @itemx -Wformat-truncation=@var{level}
4742 @opindex Wformat-truncation
4743 @opindex Wno-format-truncation
4744 Warn about calls to formatted input/output functions such as @code{snprintf}
4745 and @code{vsnprintf} that might result in output truncation. When the exact
4746 number of bytes written by a format directive cannot be determined at
4747 compile-time it is estimated based on heuristics that depend on
4748 the @var{level} argument and on optimization. While enabling optimization
4749 will in most cases improve the accuracy of the warning, it may also result
4750 in false positives. Except as noted otherwise, the option uses the same
4751 logic @option{-Wformat-overflow}.
4754 @item -Wformat-truncation
4755 @itemx -Wformat-truncation=1
4756 @opindex Wformat-truncation
4757 @opindex Wno-format-truncation
4758 Level @var{1} of @option{-Wformat-truncation} enabled by @option{-Wformat}
4759 employs a conservative approach that warns only about calls to bounded
4760 functions whose return value is unused and that will most likely result
4761 in output truncation.
4763 @item -Wformat-truncation=2
4764 Level @var{2} warns also about calls to bounded functions whose return
4765 value is used and that might result in truncation given an argument of
4766 sufficient length or magnitude.
4770 @opindex Wformat-y2k
4771 @opindex Wno-format-y2k
4772 If @option{-Wformat} is specified, also warn about @code{strftime}
4773 formats that may yield only a two-digit year.
4778 @opindex Wno-nonnull
4779 Warn about passing a null pointer for arguments marked as
4780 requiring a non-null value by the @code{nonnull} function attribute.
4782 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
4783 can be disabled with the @option{-Wno-nonnull} option.
4785 @item -Wnonnull-compare
4786 @opindex Wnonnull-compare
4787 @opindex Wno-nonnull-compare
4788 Warn when comparing an argument marked with the @code{nonnull}
4789 function attribute against null inside the function.
4791 @option{-Wnonnull-compare} is included in @option{-Wall}. It
4792 can be disabled with the @option{-Wno-nonnull-compare} option.
4794 @item -Wnull-dereference
4795 @opindex Wnull-dereference
4796 @opindex Wno-null-dereference
4797 Warn if the compiler detects paths that trigger erroneous or
4798 undefined behavior due to dereferencing a null pointer. This option
4799 is only active when @option{-fdelete-null-pointer-checks} is active,
4800 which is enabled by optimizations in most targets. The precision of
4801 the warnings depends on the optimization options used.
4803 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
4805 @opindex Wno-init-self
4806 Warn about uninitialized variables that are initialized with themselves.
4807 Note this option can only be used with the @option{-Wuninitialized} option.
4809 For example, GCC warns about @code{i} being uninitialized in the
4810 following snippet only when @option{-Winit-self} has been specified:
4821 This warning is enabled by @option{-Wall} in C++.
4823 @item -Wimplicit-int @r{(C and Objective-C only)}
4824 @opindex Wimplicit-int
4825 @opindex Wno-implicit-int
4826 Warn when a declaration does not specify a type.
4827 This warning is enabled by @option{-Wall}.
4829 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
4830 @opindex Wimplicit-function-declaration
4831 @opindex Wno-implicit-function-declaration
4832 Give a warning whenever a function is used before being declared. In
4833 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
4834 enabled by default and it is made into an error by
4835 @option{-pedantic-errors}. This warning is also enabled by
4838 @item -Wimplicit @r{(C and Objective-C only)}
4840 @opindex Wno-implicit
4841 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
4842 This warning is enabled by @option{-Wall}.
4844 @item -Wimplicit-fallthrough
4845 @opindex Wimplicit-fallthrough
4846 @opindex Wno-implicit-fallthrough
4847 @option{-Wimplicit-fallthrough} is the same as @option{-Wimplicit-fallthrough=3}
4848 and @option{-Wno-implicit-fallthrough} is the same as
4849 @option{-Wimplicit-fallthrough=0}.
4851 @item -Wimplicit-fallthrough=@var{n}
4852 @opindex Wimplicit-fallthrough=
4853 Warn when a switch case falls through. For example:
4871 This warning does not warn when the last statement of a case cannot
4872 fall through, e.g. when there is a return statement or a call to function
4873 declared with the noreturn attribute. @option{-Wimplicit-fallthrough=}
4874 also takes into account control flow statements, such as ifs, and only
4875 warns when appropriate. E.g.@:
4885 @} else if (i < 1) @{
4895 Since there are occasions where a switch case fall through is desirable,
4896 GCC provides an attribute, @code{__attribute__ ((fallthrough))}, that is
4897 to be used along with a null statement to suppress this warning that
4898 would normally occur:
4906 __attribute__ ((fallthrough));
4913 C++17 provides a standard way to suppress the @option{-Wimplicit-fallthrough}
4914 warning using @code{[[fallthrough]];} instead of the GNU attribute. In C++11
4915 or C++14 users can use @code{[[gnu::fallthrough]];}, which is a GNU extension.
4916 Instead of these attributes, it is also possible to add a fallthrough comment
4917 to silence the warning. The whole body of the C or C++ style comment should
4918 match the given regular expressions listed below. The option argument @var{n}
4919 specifies what kind of comments are accepted:
4923 @item @option{-Wimplicit-fallthrough=0} disables the warning altogether.
4925 @item @option{-Wimplicit-fallthrough=1} matches @code{.*} regular
4926 expression, any comment is used as fallthrough comment.
4928 @item @option{-Wimplicit-fallthrough=2} case insensitively matches
4929 @code{.*falls?[ \t-]*thr(ough|u).*} regular expression.
4931 @item @option{-Wimplicit-fallthrough=3} case sensitively matches one of the
4932 following regular expressions:
4936 @item @code{-fallthrough}
4938 @item @code{@@fallthrough@@}
4940 @item @code{lint -fallthrough[ \t]*}
4942 @item @code{[ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?@*FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)?}
4944 @item @code{[ \t.!]*(Else,? |Intentional(ly)? )?@*Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4946 @item @code{[ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?@*fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4950 @item @option{-Wimplicit-fallthrough=4} case sensitively matches one of the
4951 following regular expressions:
4955 @item @code{-fallthrough}
4957 @item @code{@@fallthrough@@}
4959 @item @code{lint -fallthrough[ \t]*}
4961 @item @code{[ \t]*FALLTHR(OUGH|U)[ \t]*}
4965 @item @option{-Wimplicit-fallthrough=5} doesn't recognize any comments as
4966 fallthrough comments, only attributes disable the warning.
4970 The comment needs to be followed after optional whitespace and other comments
4971 by @code{case} or @code{default} keywords or by a user label that precedes some
4972 @code{case} or @code{default} label.
4987 The @option{-Wimplicit-fallthrough=3} warning is enabled by @option{-Wextra}.
4989 @item -Wif-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
4990 @opindex Wif-not-aligned
4991 @opindex Wno-if-not-aligned
4992 Control if warning triggered by the @code{warn_if_not_aligned} attribute
4993 should be issued. This is enabled by default.
4994 Use @option{-Wno-if-not-aligned} to disable it.
4996 @item -Wignored-qualifiers @r{(C and C++ only)}
4997 @opindex Wignored-qualifiers
4998 @opindex Wno-ignored-qualifiers
4999 Warn if the return type of a function has a type qualifier
5000 such as @code{const}. For ISO C such a type qualifier has no effect,
5001 since the value returned by a function is not an lvalue.
5002 For C++, the warning is only emitted for scalar types or @code{void}.
5003 ISO C prohibits qualified @code{void} return types on function
5004 definitions, so such return types always receive a warning
5005 even without this option.
5007 This warning is also enabled by @option{-Wextra}.
5009 @item -Wignored-attributes @r{(C and C++ only)}
5010 @opindex Wignored-attributes
5011 @opindex Wno-ignored-attributes
5012 Warn when an attribute is ignored. This is different from the
5013 @option{-Wattributes} option in that it warns whenever the compiler decides
5014 to drop an attribute, not that the attribute is either unknown, used in a
5015 wrong place, etc. This warning is enabled by default.
5020 Warn if the type of @code{main} is suspicious. @code{main} should be
5021 a function with external linkage, returning int, taking either zero
5022 arguments, two, or three arguments of appropriate types. This warning
5023 is enabled by default in C++ and is enabled by either @option{-Wall}
5024 or @option{-Wpedantic}.
5026 @item -Wmisleading-indentation @r{(C and C++ only)}
5027 @opindex Wmisleading-indentation
5028 @opindex Wno-misleading-indentation
5029 Warn when the indentation of the code does not reflect the block structure.
5030 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
5031 @code{for} clauses with a guarded statement that does not use braces,
5032 followed by an unguarded statement with the same indentation.
5034 In the following example, the call to ``bar'' is misleadingly indented as
5035 if it were guarded by the ``if'' conditional.
5038 if (some_condition ())
5040 bar (); /* Gotcha: this is not guarded by the "if". */
5043 In the case of mixed tabs and spaces, the warning uses the
5044 @option{-ftabstop=} option to determine if the statements line up
5047 The warning is not issued for code involving multiline preprocessor logic
5048 such as the following example.
5053 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
5059 The warning is not issued after a @code{#line} directive, since this
5060 typically indicates autogenerated code, and no assumptions can be made
5061 about the layout of the file that the directive references.
5063 This warning is enabled by @option{-Wall} in C and C++.
5065 @item -Wmissing-attributes
5066 @opindex Wmissing-attributes
5067 @opindex Wno-missing-attributes
5068 Warn when a declaration of a function is missing one or more attributes
5069 that a related function is declared with and whose absence may adversely
5070 affect the correctness or efficiency of generated code. For example,
5071 the warning is issued for declarations of aliases that use attributes
5072 to specify less restrictive requirements than those of their targets.
5073 This typically represents a potential optimization opportunity.
5074 By contrast, the @option{-Wattribute-alias=2} option controls warnings
5075 issued when the alias is more restrictive than the target, which could
5076 lead to incorrect code generation.
5077 Attributes considered include @code{alloc_align}, @code{alloc_size},
5078 @code{cold}, @code{const}, @code{hot}, @code{leaf}, @code{malloc},
5079 @code{nonnull}, @code{noreturn}, @code{nothrow}, @code{pure},
5080 @code{returns_nonnull}, and @code{returns_twice}.
5082 In C++, the warning is issued when an explicit specialization of a primary
5083 template declared with attribute @code{alloc_align}, @code{alloc_size},
5084 @code{assume_aligned}, @code{format}, @code{format_arg}, @code{malloc},
5085 or @code{nonnull} is declared without it. Attributes @code{deprecated},
5086 @code{error}, and @code{warning} suppress the warning.
5087 (@pxref{Function Attributes}).
5089 You can use the @code{copy} attribute to apply the same
5090 set of attributes to a declaration as that on another declaration without
5091 explicitly enumerating the attributes. This attribute can be applied
5092 to declarations of functions (@pxref{Common Function Attributes}),
5093 variables (@pxref{Common Variable Attributes}), or types
5094 (@pxref{Common Type Attributes}).
5096 @option{-Wmissing-attributes} is enabled by @option{-Wall}.
5098 For example, since the declaration of the primary function template
5099 below makes use of both attribute @code{malloc} and @code{alloc_size}
5100 the declaration of the explicit specialization of the template is
5101 diagnosed because it is missing one of the attributes.
5105 T* __attribute__ ((malloc, alloc_size (1)))
5109 void* __attribute__ ((malloc)) // missing alloc_size
5110 allocate<void> (size_t);
5113 @item -Wmissing-braces
5114 @opindex Wmissing-braces
5115 @opindex Wno-missing-braces
5116 Warn if an aggregate or union initializer is not fully bracketed. In
5117 the following example, the initializer for @code{a} is not fully
5118 bracketed, but that for @code{b} is fully bracketed. This warning is
5119 enabled by @option{-Wall} in C.
5122 int a[2][2] = @{ 0, 1, 2, 3 @};
5123 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
5126 This warning is enabled by @option{-Wall}.
5128 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
5129 @opindex Wmissing-include-dirs
5130 @opindex Wno-missing-include-dirs
5131 Warn if a user-supplied include directory does not exist.
5133 @item -Wmissing-profile
5134 @opindex Wmissing-profile
5135 @opindex Wno-missing-profile
5136 Warn if feedback profiles are missing when using the
5137 @option{-fprofile-use} option.
5138 This option diagnoses those cases where a new function or a new file is added
5139 to the user code between compiling with @option{-fprofile-generate} and with
5140 @option{-fprofile-use}, without regenerating the profiles. In these cases, the
5141 profile feedback data files do not contain any profile feedback information for
5142 the newly added function or file respectively. Also, in the case when profile
5143 count data (.gcda) files are removed, GCC cannot use any profile feedback
5144 information. In all these cases, warnings are issued to inform the user that a
5145 profile generation step is due. @option{-Wno-missing-profile} can be used to
5146 disable the warning. Ignoring the warning can result in poorly optimized code.
5147 Completely disabling the warning is not recommended and should be done only
5148 when non-existent profile data is justified.
5150 @item -Wmultistatement-macros
5151 @opindex Wmultistatement-macros
5152 @opindex Wno-multistatement-macros
5153 Warn about unsafe multiple statement macros that appear to be guarded
5154 by a clause such as @code{if}, @code{else}, @code{for}, @code{switch}, or
5155 @code{while}, in which only the first statement is actually guarded after
5156 the macro is expanded.
5161 #define DOIT x++; y++
5166 will increment @code{y} unconditionally, not just when @code{c} holds.
5167 The can usually be fixed by wrapping the macro in a do-while loop:
5169 #define DOIT do @{ x++; y++; @} while (0)
5174 This warning is enabled by @option{-Wall} in C and C++.
5177 @opindex Wparentheses
5178 @opindex Wno-parentheses
5179 Warn if parentheses are omitted in certain contexts, such
5180 as when there is an assignment in a context where a truth value
5181 is expected, or when operators are nested whose precedence people
5182 often get confused about.
5184 Also warn if a comparison like @code{x<=y<=z} appears; this is
5185 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
5186 interpretation from that of ordinary mathematical notation.
5188 Also warn for dangerous uses of the GNU extension to
5189 @code{?:} with omitted middle operand. When the condition
5190 in the @code{?}: operator is a boolean expression, the omitted value is
5191 always 1. Often programmers expect it to be a value computed
5192 inside the conditional expression instead.
5194 For C++ this also warns for some cases of unnecessary parentheses in
5195 declarations, which can indicate an attempt at a function call instead
5199 // Declares a local variable called mymutex.
5200 std::unique_lock<std::mutex> (mymutex);
5201 // User meant std::unique_lock<std::mutex> lock (mymutex);
5205 This warning is enabled by @option{-Wall}.
5207 @item -Wsequence-point
5208 @opindex Wsequence-point
5209 @opindex Wno-sequence-point
5210 Warn about code that may have undefined semantics because of violations
5211 of sequence point rules in the C and C++ standards.
5213 The C and C++ standards define the order in which expressions in a C/C++
5214 program are evaluated in terms of @dfn{sequence points}, which represent
5215 a partial ordering between the execution of parts of the program: those
5216 executed before the sequence point, and those executed after it. These
5217 occur after the evaluation of a full expression (one which is not part
5218 of a larger expression), after the evaluation of the first operand of a
5219 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
5220 function is called (but after the evaluation of its arguments and the
5221 expression denoting the called function), and in certain other places.
5222 Other than as expressed by the sequence point rules, the order of
5223 evaluation of subexpressions of an expression is not specified. All
5224 these rules describe only a partial order rather than a total order,
5225 since, for example, if two functions are called within one expression
5226 with no sequence point between them, the order in which the functions
5227 are called is not specified. However, the standards committee have
5228 ruled that function calls do not overlap.
5230 It is not specified when between sequence points modifications to the
5231 values of objects take effect. Programs whose behavior depends on this
5232 have undefined behavior; the C and C++ standards specify that ``Between
5233 the previous and next sequence point an object shall have its stored
5234 value modified at most once by the evaluation of an expression.
5235 Furthermore, the prior value shall be read only to determine the value
5236 to be stored.''. If a program breaks these rules, the results on any
5237 particular implementation are entirely unpredictable.
5239 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
5240 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
5241 diagnosed by this option, and it may give an occasional false positive
5242 result, but in general it has been found fairly effective at detecting
5243 this sort of problem in programs.
5245 The C++17 standard will define the order of evaluation of operands in
5246 more cases: in particular it requires that the right-hand side of an
5247 assignment be evaluated before the left-hand side, so the above
5248 examples are no longer undefined. But this warning will still warn
5249 about them, to help people avoid writing code that is undefined in C
5250 and earlier revisions of C++.
5252 The standard is worded confusingly, therefore there is some debate
5253 over the precise meaning of the sequence point rules in subtle cases.
5254 Links to discussions of the problem, including proposed formal
5255 definitions, may be found on the GCC readings page, at
5256 @uref{http://gcc.gnu.org/@/readings.html}.
5258 This warning is enabled by @option{-Wall} for C and C++.
5260 @item -Wno-return-local-addr
5261 @opindex Wno-return-local-addr
5262 @opindex Wreturn-local-addr
5263 Do not warn about returning a pointer (or in C++, a reference) to a
5264 variable that goes out of scope after the function returns.
5267 @opindex Wreturn-type
5268 @opindex Wno-return-type
5269 Warn whenever a function is defined with a return type that defaults
5270 to @code{int}. Also warn about any @code{return} statement with no
5271 return value in a function whose return type is not @code{void}
5272 (falling off the end of the function body is considered returning
5275 For C only, warn about a @code{return} statement with an expression in a
5276 function whose return type is @code{void}, unless the expression type is
5277 also @code{void}. As a GNU extension, the latter case is accepted
5278 without a warning unless @option{-Wpedantic} is used. Attempting
5279 to use the return value of a non-@code{void} function other than @code{main}
5280 that flows off the end by reaching the closing curly brace that terminates
5281 the function is undefined.
5283 Unlike in C, in C++, flowing off the end of a non-@code{void} function other
5284 than @code{main} results in undefined behavior even when the value of
5285 the function is not used.
5287 This warning is enabled by default in C++ and by @option{-Wall} otherwise.
5289 @item -Wshift-count-negative
5290 @opindex Wshift-count-negative
5291 @opindex Wno-shift-count-negative
5292 Warn if shift count is negative. This warning is enabled by default.
5294 @item -Wshift-count-overflow
5295 @opindex Wshift-count-overflow
5296 @opindex Wno-shift-count-overflow
5297 Warn if shift count >= width of type. This warning is enabled by default.
5299 @item -Wshift-negative-value
5300 @opindex Wshift-negative-value
5301 @opindex Wno-shift-negative-value
5302 Warn if left shifting a negative value. This warning is enabled by
5303 @option{-Wextra} in C99 and C++11 modes (and newer).
5305 @item -Wshift-overflow
5306 @itemx -Wshift-overflow=@var{n}
5307 @opindex Wshift-overflow
5308 @opindex Wno-shift-overflow
5309 Warn about left shift overflows. This warning is enabled by
5310 default in C99 and C++11 modes (and newer).
5313 @item -Wshift-overflow=1
5314 This is the warning level of @option{-Wshift-overflow} and is enabled
5315 by default in C99 and C++11 modes (and newer). This warning level does
5316 not warn about left-shifting 1 into the sign bit. (However, in C, such
5317 an overflow is still rejected in contexts where an integer constant expression
5318 is required.) No warning is emitted in C++2A mode (and newer), as signed left
5321 @item -Wshift-overflow=2
5322 This warning level also warns about left-shifting 1 into the sign bit,
5323 unless C++14 mode (or newer) is active.
5329 Warn whenever a @code{switch} statement has an index of enumerated type
5330 and lacks a @code{case} for one or more of the named codes of that
5331 enumeration. (The presence of a @code{default} label prevents this
5332 warning.) @code{case} labels outside the enumeration range also
5333 provoke warnings when this option is used (even if there is a
5334 @code{default} label).
5335 This warning is enabled by @option{-Wall}.
5337 @item -Wswitch-default
5338 @opindex Wswitch-default
5339 @opindex Wno-switch-default
5340 Warn whenever a @code{switch} statement does not have a @code{default}
5344 @opindex Wswitch-enum
5345 @opindex Wno-switch-enum
5346 Warn whenever a @code{switch} statement has an index of enumerated type
5347 and lacks a @code{case} for one or more of the named codes of that
5348 enumeration. @code{case} labels outside the enumeration range also
5349 provoke warnings when this option is used. The only difference
5350 between @option{-Wswitch} and this option is that this option gives a
5351 warning about an omitted enumeration code even if there is a
5352 @code{default} label.
5355 @opindex Wswitch-bool
5356 @opindex Wno-switch-bool
5357 Warn whenever a @code{switch} statement has an index of boolean type
5358 and the case values are outside the range of a boolean type.
5359 It is possible to suppress this warning by casting the controlling
5360 expression to a type other than @code{bool}. For example:
5363 switch ((int) (a == 4))
5369 This warning is enabled by default for C and C++ programs.
5371 @item -Wswitch-unreachable
5372 @opindex Wswitch-unreachable
5373 @opindex Wno-switch-unreachable
5374 Warn whenever a @code{switch} statement contains statements between the
5375 controlling expression and the first case label, which will never be
5376 executed. For example:
5388 @option{-Wswitch-unreachable} does not warn if the statement between the
5389 controlling expression and the first case label is just a declaration:
5402 This warning is enabled by default for C and C++ programs.
5404 @item -Wsync-nand @r{(C and C++ only)}
5406 @opindex Wno-sync-nand
5407 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
5408 built-in functions are used. These functions changed semantics in GCC 4.4.
5410 @item -Wunused-but-set-parameter
5411 @opindex Wunused-but-set-parameter
5412 @opindex Wno-unused-but-set-parameter
5413 Warn whenever a function parameter is assigned to, but otherwise unused
5414 (aside from its declaration).
5416 To suppress this warning use the @code{unused} attribute
5417 (@pxref{Variable Attributes}).
5419 This warning is also enabled by @option{-Wunused} together with
5422 @item -Wunused-but-set-variable
5423 @opindex Wunused-but-set-variable
5424 @opindex Wno-unused-but-set-variable
5425 Warn whenever a local variable is assigned to, but otherwise unused
5426 (aside from its declaration).
5427 This warning is enabled by @option{-Wall}.
5429 To suppress this warning use the @code{unused} attribute
5430 (@pxref{Variable Attributes}).
5432 This warning is also enabled by @option{-Wunused}, which is enabled
5435 @item -Wunused-function
5436 @opindex Wunused-function
5437 @opindex Wno-unused-function
5438 Warn whenever a static function is declared but not defined or a
5439 non-inline static function is unused.
5440 This warning is enabled by @option{-Wall}.
5442 @item -Wunused-label
5443 @opindex Wunused-label
5444 @opindex Wno-unused-label
5445 Warn whenever a label is declared but not used.
5446 This warning is enabled by @option{-Wall}.
5448 To suppress this warning use the @code{unused} attribute
5449 (@pxref{Variable Attributes}).
5451 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
5452 @opindex Wunused-local-typedefs
5453 @opindex Wno-unused-local-typedefs
5454 Warn when a typedef locally defined in a function is not used.
5455 This warning is enabled by @option{-Wall}.
5457 @item -Wunused-parameter
5458 @opindex Wunused-parameter
5459 @opindex Wno-unused-parameter
5460 Warn whenever a function parameter is unused aside from its declaration.
5462 To suppress this warning use the @code{unused} attribute
5463 (@pxref{Variable Attributes}).
5465 @item -Wno-unused-result
5466 @opindex Wunused-result
5467 @opindex Wno-unused-result
5468 Do not warn if a caller of a function marked with attribute
5469 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
5470 its return value. The default is @option{-Wunused-result}.
5472 @item -Wunused-variable
5473 @opindex Wunused-variable
5474 @opindex Wno-unused-variable
5475 Warn whenever a local or static variable is unused aside from its
5476 declaration. This option implies @option{-Wunused-const-variable=1} for C,
5477 but not for C++. This warning is enabled by @option{-Wall}.
5479 To suppress this warning use the @code{unused} attribute
5480 (@pxref{Variable Attributes}).
5482 @item -Wunused-const-variable
5483 @itemx -Wunused-const-variable=@var{n}
5484 @opindex Wunused-const-variable
5485 @opindex Wno-unused-const-variable
5486 Warn whenever a constant static variable is unused aside from its declaration.
5487 @option{-Wunused-const-variable=1} is enabled by @option{-Wunused-variable}
5488 for C, but not for C++. In C this declares variable storage, but in C++ this
5489 is not an error since const variables take the place of @code{#define}s.
5491 To suppress this warning use the @code{unused} attribute
5492 (@pxref{Variable Attributes}).
5495 @item -Wunused-const-variable=1
5496 This is the warning level that is enabled by @option{-Wunused-variable} for
5497 C. It warns only about unused static const variables defined in the main
5498 compilation unit, but not about static const variables declared in any
5501 @item -Wunused-const-variable=2
5502 This warning level also warns for unused constant static variables in
5503 headers (excluding system headers). This is the warning level of
5504 @option{-Wunused-const-variable} and must be explicitly requested since
5505 in C++ this isn't an error and in C it might be harder to clean up all
5509 @item -Wunused-value
5510 @opindex Wunused-value
5511 @opindex Wno-unused-value
5512 Warn whenever a statement computes a result that is explicitly not
5513 used. To suppress this warning cast the unused expression to
5514 @code{void}. This includes an expression-statement or the left-hand
5515 side of a comma expression that contains no side effects. For example,
5516 an expression such as @code{x[i,j]} causes a warning, while
5517 @code{x[(void)i,j]} does not.
5519 This warning is enabled by @option{-Wall}.
5524 All the above @option{-Wunused} options combined.
5526 In order to get a warning about an unused function parameter, you must
5527 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
5528 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
5530 @item -Wuninitialized
5531 @opindex Wuninitialized
5532 @opindex Wno-uninitialized
5533 Warn if an automatic variable is used without first being initialized
5534 or if a variable may be clobbered by a @code{setjmp} call. In C++,
5535 warn if a non-static reference or non-static @code{const} member
5536 appears in a class without constructors.
5538 If you want to warn about code that uses the uninitialized value of the
5539 variable in its own initializer, use the @option{-Winit-self} option.
5541 These warnings occur for individual uninitialized or clobbered
5542 elements of structure, union or array variables as well as for
5543 variables that are uninitialized or clobbered as a whole. They do
5544 not occur for variables or elements declared @code{volatile}. Because
5545 these warnings depend on optimization, the exact variables or elements
5546 for which there are warnings depends on the precise optimization
5547 options and version of GCC used.
5549 Note that there may be no warning about a variable that is used only
5550 to compute a value that itself is never used, because such
5551 computations may be deleted by data flow analysis before the warnings
5554 @item -Winvalid-memory-model
5555 @opindex Winvalid-memory-model
5556 @opindex Wno-invalid-memory-model
5557 Warn for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins},
5558 and the C11 atomic generic functions with a memory consistency argument
5559 that is either invalid for the operation or outside the range of values
5560 of the @code{memory_order} enumeration. For example, since the
5561 @code{__atomic_store} and @code{__atomic_store_n} built-ins are only
5562 defined for the relaxed, release, and sequentially consistent memory
5563 orders the following code is diagnosed:
5568 __atomic_store_n (i, 0, memory_order_consume);
5572 @option{-Winvalid-memory-model} is enabled by default.
5574 @item -Wmaybe-uninitialized
5575 @opindex Wmaybe-uninitialized
5576 @opindex Wno-maybe-uninitialized
5577 For an automatic (i.e.@: local) variable, if there exists a path from the
5578 function entry to a use of the variable that is initialized, but there exist
5579 some other paths for which the variable is not initialized, the compiler
5580 emits a warning if it cannot prove the uninitialized paths are not
5581 executed at run time.
5583 These warnings are only possible in optimizing compilation, because otherwise
5584 GCC does not keep track of the state of variables.
5586 These warnings are made optional because GCC may not be able to determine when
5587 the code is correct in spite of appearing to have an error. Here is one
5588 example of how this can happen:
5608 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
5609 always initialized, but GCC doesn't know this. To suppress the
5610 warning, you need to provide a default case with assert(0) or
5613 @cindex @code{longjmp} warnings
5614 This option also warns when a non-volatile automatic variable might be
5615 changed by a call to @code{longjmp}.
5616 The compiler sees only the calls to @code{setjmp}. It cannot know
5617 where @code{longjmp} will be called; in fact, a signal handler could
5618 call it at any point in the code. As a result, you may get a warning
5619 even when there is in fact no problem because @code{longjmp} cannot
5620 in fact be called at the place that would cause a problem.
5622 Some spurious warnings can be avoided if you declare all the functions
5623 you use that never return as @code{noreturn}. @xref{Function
5626 This warning is enabled by @option{-Wall} or @option{-Wextra}.
5628 @item -Wunknown-pragmas
5629 @opindex Wunknown-pragmas
5630 @opindex Wno-unknown-pragmas
5631 @cindex warning for unknown pragmas
5632 @cindex unknown pragmas, warning
5633 @cindex pragmas, warning of unknown
5634 Warn when a @code{#pragma} directive is encountered that is not understood by
5635 GCC@. If this command-line option is used, warnings are even issued
5636 for unknown pragmas in system header files. This is not the case if
5637 the warnings are only enabled by the @option{-Wall} command-line option.
5640 @opindex Wno-pragmas
5642 Do not warn about misuses of pragmas, such as incorrect parameters,
5643 invalid syntax, or conflicts between pragmas. See also
5644 @option{-Wunknown-pragmas}.
5646 @item -Wno-prio-ctor-dtor
5647 @opindex Wno-prio-ctor-dtor
5648 @opindex Wprio-ctor-dtor
5649 Do not warn if a priority from 0 to 100 is used for constructor or destructor.
5650 The use of constructor and destructor attributes allow you to assign a
5651 priority to the constructor/destructor to control its order of execution
5652 before @code{main} is called or after it returns. The priority values must be
5653 greater than 100 as the compiler reserves priority values between 0--100 for
5656 @item -Wstrict-aliasing
5657 @opindex Wstrict-aliasing
5658 @opindex Wno-strict-aliasing
5659 This option is only active when @option{-fstrict-aliasing} is active.
5660 It warns about code that might break the strict aliasing rules that the
5661 compiler is using for optimization. The warning does not catch all
5662 cases, but does attempt to catch the more common pitfalls. It is
5663 included in @option{-Wall}.
5664 It is equivalent to @option{-Wstrict-aliasing=3}
5666 @item -Wstrict-aliasing=n
5667 @opindex Wstrict-aliasing=n
5668 This option is only active when @option{-fstrict-aliasing} is active.
5669 It warns about code that might break the strict aliasing rules that the
5670 compiler is using for optimization.
5671 Higher levels correspond to higher accuracy (fewer false positives).
5672 Higher levels also correspond to more effort, similar to the way @option{-O}
5674 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
5676 Level 1: Most aggressive, quick, least accurate.
5677 Possibly useful when higher levels
5678 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
5679 false negatives. However, it has many false positives.
5680 Warns for all pointer conversions between possibly incompatible types,
5681 even if never dereferenced. Runs in the front end only.
5683 Level 2: Aggressive, quick, not too precise.
5684 May still have many false positives (not as many as level 1 though),
5685 and few false negatives (but possibly more than level 1).
5686 Unlike level 1, it only warns when an address is taken. Warns about
5687 incomplete types. Runs in the front end only.
5689 Level 3 (default for @option{-Wstrict-aliasing}):
5690 Should have very few false positives and few false
5691 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
5692 Takes care of the common pun+dereference pattern in the front end:
5693 @code{*(int*)&some_float}.
5694 If optimization is enabled, it also runs in the back end, where it deals
5695 with multiple statement cases using flow-sensitive points-to information.
5696 Only warns when the converted pointer is dereferenced.
5697 Does not warn about incomplete types.
5699 @item -Wstrict-overflow
5700 @itemx -Wstrict-overflow=@var{n}
5701 @opindex Wstrict-overflow
5702 @opindex Wno-strict-overflow
5703 This option is only active when signed overflow is undefined.
5704 It warns about cases where the compiler optimizes based on the
5705 assumption that signed overflow does not occur. Note that it does not
5706 warn about all cases where the code might overflow: it only warns
5707 about cases where the compiler implements some optimization. Thus
5708 this warning depends on the optimization level.
5710 An optimization that assumes that signed overflow does not occur is
5711 perfectly safe if the values of the variables involved are such that
5712 overflow never does, in fact, occur. Therefore this warning can
5713 easily give a false positive: a warning about code that is not
5714 actually a problem. To help focus on important issues, several
5715 warning levels are defined. No warnings are issued for the use of
5716 undefined signed overflow when estimating how many iterations a loop
5717 requires, in particular when determining whether a loop will be
5721 @item -Wstrict-overflow=1
5722 Warn about cases that are both questionable and easy to avoid. For
5723 example the compiler simplifies
5724 @code{x + 1 > x} to @code{1}. This level of
5725 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
5726 are not, and must be explicitly requested.
5728 @item -Wstrict-overflow=2
5729 Also warn about other cases where a comparison is simplified to a
5730 constant. For example: @code{abs (x) >= 0}. This can only be
5731 simplified when signed integer overflow is undefined, because
5732 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
5733 zero. @option{-Wstrict-overflow} (with no level) is the same as
5734 @option{-Wstrict-overflow=2}.
5736 @item -Wstrict-overflow=3
5737 Also warn about other cases where a comparison is simplified. For
5738 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
5740 @item -Wstrict-overflow=4
5741 Also warn about other simplifications not covered by the above cases.
5742 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
5744 @item -Wstrict-overflow=5
5745 Also warn about cases where the compiler reduces the magnitude of a
5746 constant involved in a comparison. For example: @code{x + 2 > y} is
5747 simplified to @code{x + 1 >= y}. This is reported only at the
5748 highest warning level because this simplification applies to many
5749 comparisons, so this warning level gives a very large number of
5753 @item -Wstringop-overflow
5754 @itemx -Wstringop-overflow=@var{type}
5755 @opindex Wstringop-overflow
5756 @opindex Wno-stringop-overflow
5757 Warn for calls to string manipulation functions such as @code{memcpy} and
5758 @code{strcpy} that are determined to overflow the destination buffer. The
5759 optional argument is one greater than the type of Object Size Checking to
5760 perform to determine the size of the destination. @xref{Object Size Checking}.
5761 The argument is meaningful only for functions that operate on character arrays
5762 but not for raw memory functions like @code{memcpy} which always make use
5763 of Object Size type-0. The option also warns for calls that specify a size
5764 in excess of the largest possible object or at most @code{SIZE_MAX / 2} bytes.
5765 The option produces the best results with optimization enabled but can detect
5766 a small subset of simple buffer overflows even without optimization in
5767 calls to the GCC built-in functions like @code{__builtin_memcpy} that
5768 correspond to the standard functions. In any case, the option warns about
5769 just a subset of buffer overflows detected by the corresponding overflow
5770 checking built-ins. For example, the option will issue a warning for
5771 the @code{strcpy} call below because it copies at least 5 characters
5772 (the string @code{"blue"} including the terminating NUL) into the buffer
5776 enum Color @{ blue, purple, yellow @};
5777 const char* f (enum Color clr)
5779 static char buf [4];
5783 case blue: str = "blue"; break;
5784 case purple: str = "purple"; break;
5785 case yellow: str = "yellow"; break;
5788 return strcpy (buf, str); // warning here
5792 Option @option{-Wstringop-overflow=2} is enabled by default.
5795 @item -Wstringop-overflow
5796 @itemx -Wstringop-overflow=1
5797 @opindex Wstringop-overflow
5798 @opindex Wno-stringop-overflow
5799 The @option{-Wstringop-overflow=1} option uses type-zero Object Size Checking
5800 to determine the sizes of destination objects. This is the default setting
5801 of the option. At this setting the option will not warn for writes past
5802 the end of subobjects of larger objects accessed by pointers unless the
5803 size of the largest surrounding object is known. When the destination may
5804 be one of several objects it is assumed to be the largest one of them. On
5805 Linux systems, when optimization is enabled at this setting the option warns
5806 for the same code as when the @code{_FORTIFY_SOURCE} macro is defined to
5809 @item -Wstringop-overflow=2
5810 The @option{-Wstringop-overflow=2} option uses type-one Object Size Checking
5811 to determine the sizes of destination objects. At this setting the option
5812 will warn about overflows when writing to members of the largest complete
5813 objects whose exact size is known. It will, however, not warn for excessive
5814 writes to the same members of unknown objects referenced by pointers since
5815 they may point to arrays containing unknown numbers of elements.
5817 @item -Wstringop-overflow=3
5818 The @option{-Wstringop-overflow=3} option uses type-two Object Size Checking
5819 to determine the sizes of destination objects. At this setting the option
5820 warns about overflowing the smallest object or data member. This is the
5821 most restrictive setting of the option that may result in warnings for safe
5824 @item -Wstringop-overflow=4
5825 The @option{-Wstringop-overflow=4} option uses type-three Object Size Checking
5826 to determine the sizes of destination objects. At this setting the option
5827 will warn about overflowing any data members, and when the destination is
5828 one of several objects it uses the size of the largest of them to decide
5829 whether to issue a warning. Similarly to @option{-Wstringop-overflow=3} this
5830 setting of the option may result in warnings for benign code.
5833 @item -Wstringop-truncation
5834 @opindex Wstringop-truncation
5835 @opindex Wno-stringop-truncation
5836 Warn for calls to bounded string manipulation functions such as @code{strncat},
5837 @code{strncpy}, and @code{stpncpy} that may either truncate the copied string
5838 or leave the destination unchanged.
5840 In the following example, the call to @code{strncat} specifies a bound that
5841 is less than the length of the source string. As a result, the copy of
5842 the source will be truncated and so the call is diagnosed. To avoid the
5843 warning use @code{bufsize - strlen (buf) - 1)} as the bound.
5846 void append (char *buf, size_t bufsize)
5848 strncat (buf, ".txt", 3);
5852 As another example, the following call to @code{strncpy} results in copying
5853 to @code{d} just the characters preceding the terminating NUL, without
5854 appending the NUL to the end. Assuming the result of @code{strncpy} is
5855 necessarily a NUL-terminated string is a common mistake, and so the call
5856 is diagnosed. To avoid the warning when the result is not expected to be
5857 NUL-terminated, call @code{memcpy} instead.
5860 void copy (char *d, const char *s)
5862 strncpy (d, s, strlen (s));
5866 In the following example, the call to @code{strncpy} specifies the size
5867 of the destination buffer as the bound. If the length of the source
5868 string is equal to or greater than this size the result of the copy will
5869 not be NUL-terminated. Therefore, the call is also diagnosed. To avoid
5870 the warning, specify @code{sizeof buf - 1} as the bound and set the last
5871 element of the buffer to @code{NUL}.
5874 void copy (const char *s)
5877 strncpy (buf, s, sizeof buf);
5882 In situations where a character array is intended to store a sequence
5883 of bytes with no terminating @code{NUL} such an array may be annotated
5884 with attribute @code{nonstring} to avoid this warning. Such arrays,
5885 however, are not suitable arguments to functions that expect
5886 @code{NUL}-terminated strings. To help detect accidental misuses of
5887 such arrays GCC issues warnings unless it can prove that the use is
5888 safe. @xref{Common Variable Attributes}.
5890 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{|}cold@r{|}malloc@r{]}
5891 @opindex Wsuggest-attribute=
5892 @opindex Wno-suggest-attribute=
5893 Warn for cases where adding an attribute may be beneficial. The
5894 attributes currently supported are listed below.
5897 @item -Wsuggest-attribute=pure
5898 @itemx -Wsuggest-attribute=const
5899 @itemx -Wsuggest-attribute=noreturn
5900 @itemx -Wmissing-noreturn
5901 @itemx -Wsuggest-attribute=malloc
5902 @opindex Wsuggest-attribute=pure
5903 @opindex Wno-suggest-attribute=pure
5904 @opindex Wsuggest-attribute=const
5905 @opindex Wno-suggest-attribute=const
5906 @opindex Wsuggest-attribute=noreturn
5907 @opindex Wno-suggest-attribute=noreturn
5908 @opindex Wmissing-noreturn
5909 @opindex Wno-missing-noreturn
5910 @opindex Wsuggest-attribute=malloc
5911 @opindex Wno-suggest-attribute=malloc
5913 Warn about functions that might be candidates for attributes
5914 @code{pure}, @code{const} or @code{noreturn} or @code{malloc}. The compiler
5915 only warns for functions visible in other compilation units or (in the case of
5916 @code{pure} and @code{const}) if it cannot prove that the function returns
5917 normally. A function returns normally if it doesn't contain an infinite loop or
5918 return abnormally by throwing, calling @code{abort} or trapping. This analysis
5919 requires option @option{-fipa-pure-const}, which is enabled by default at
5920 @option{-O} and higher. Higher optimization levels improve the accuracy
5923 @item -Wsuggest-attribute=format
5924 @itemx -Wmissing-format-attribute
5925 @opindex Wsuggest-attribute=format
5926 @opindex Wmissing-format-attribute
5927 @opindex Wno-suggest-attribute=format
5928 @opindex Wno-missing-format-attribute
5932 Warn about function pointers that might be candidates for @code{format}
5933 attributes. Note these are only possible candidates, not absolute ones.
5934 GCC guesses that function pointers with @code{format} attributes that
5935 are used in assignment, initialization, parameter passing or return
5936 statements should have a corresponding @code{format} attribute in the
5937 resulting type. I.e.@: the left-hand side of the assignment or
5938 initialization, the type of the parameter variable, or the return type
5939 of the containing function respectively should also have a @code{format}
5940 attribute to avoid the warning.
5942 GCC also warns about function definitions that might be
5943 candidates for @code{format} attributes. Again, these are only
5944 possible candidates. GCC guesses that @code{format} attributes
5945 might be appropriate for any function that calls a function like
5946 @code{vprintf} or @code{vscanf}, but this might not always be the
5947 case, and some functions for which @code{format} attributes are
5948 appropriate may not be detected.
5950 @item -Wsuggest-attribute=cold
5951 @opindex Wsuggest-attribute=cold
5952 @opindex Wno-suggest-attribute=cold
5954 Warn about functions that might be candidates for @code{cold} attribute. This
5955 is based on static detection and generally will only warn about functions which
5956 always leads to a call to another @code{cold} function such as wrappers of
5957 C++ @code{throw} or fatal error reporting functions leading to @code{abort}.
5960 @item -Wsuggest-final-types
5961 @opindex Wno-suggest-final-types
5962 @opindex Wsuggest-final-types
5963 Warn about types with virtual methods where code quality would be improved
5964 if the type were declared with the C++11 @code{final} specifier,
5966 declared in an anonymous namespace. This allows GCC to more aggressively
5967 devirtualize the polymorphic calls. This warning is more effective with link
5968 time optimization, where the information about the class hierarchy graph is
5971 @item -Wsuggest-final-methods
5972 @opindex Wno-suggest-final-methods
5973 @opindex Wsuggest-final-methods
5974 Warn about virtual methods where code quality would be improved if the method
5975 were declared with the C++11 @code{final} specifier,
5976 or, if possible, its type were
5977 declared in an anonymous namespace or with the @code{final} specifier.
5979 more effective with link-time optimization, where the information about the
5980 class hierarchy graph is more complete. It is recommended to first consider
5981 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
5984 @item -Wsuggest-override
5985 Warn about overriding virtual functions that are not marked with the override
5989 @opindex Wno-alloc-zero
5990 @opindex Walloc-zero
5991 Warn about calls to allocation functions decorated with attribute
5992 @code{alloc_size} that specify zero bytes, including those to the built-in
5993 forms of the functions @code{aligned_alloc}, @code{alloca}, @code{calloc},
5994 @code{malloc}, and @code{realloc}. Because the behavior of these functions
5995 when called with a zero size differs among implementations (and in the case
5996 of @code{realloc} has been deprecated) relying on it may result in subtle
5997 portability bugs and should be avoided.
5999 @item -Walloc-size-larger-than=@var{byte-size}
6000 @opindex Walloc-size-larger-than=
6001 @opindex Wno-alloc-size-larger-than
6002 Warn about calls to functions decorated with attribute @code{alloc_size}
6003 that attempt to allocate objects larger than the specified number of bytes,
6004 or where the result of the size computation in an integer type with infinite
6005 precision would exceed the value of @samp{PTRDIFF_MAX} on the target.
6006 @option{-Walloc-size-larger-than=}@samp{PTRDIFF_MAX} is enabled by default.
6007 Warnings controlled by the option can be disabled either by specifying
6008 @var{byte-size} of @samp{SIZE_MAX} or more or by
6009 @option{-Wno-alloc-size-larger-than}.
6010 @xref{Function Attributes}.
6012 @item -Wno-alloc-size-larger-than
6013 @opindex Wno-alloc-size-larger-than
6014 Disable @option{-Walloc-size-larger-than=} warnings. The option is
6015 equivalent to @option{-Walloc-size-larger-than=}@samp{SIZE_MAX} or
6021 This option warns on all uses of @code{alloca} in the source.
6023 @item -Walloca-larger-than=@var{byte-size}
6024 @opindex Walloca-larger-than=
6025 @opindex Wno-alloca-larger-than
6026 This option warns on calls to @code{alloca} with an integer argument whose
6027 value is either zero, or that is not bounded by a controlling predicate
6028 that limits its value to at most @var{byte-size}. It also warns for calls
6029 to @code{alloca} where the bound value is unknown. Arguments of non-integer
6030 types are considered unbounded even if they appear to be constrained to
6033 For example, a bounded case of @code{alloca} could be:
6036 void func (size_t n)
6047 In the above example, passing @code{-Walloca-larger-than=1000} would not
6048 issue a warning because the call to @code{alloca} is known to be at most
6049 1000 bytes. However, if @code{-Walloca-larger-than=500} were passed,
6050 the compiler would emit a warning.
6052 Unbounded uses, on the other hand, are uses of @code{alloca} with no
6053 controlling predicate constraining its integer argument. For example:
6058 void *p = alloca (n);
6063 If @code{-Walloca-larger-than=500} were passed, the above would trigger
6064 a warning, but this time because of the lack of bounds checking.
6066 Note, that even seemingly correct code involving signed integers could
6070 void func (signed int n)
6080 In the above example, @var{n} could be negative, causing a larger than
6081 expected argument to be implicitly cast into the @code{alloca} call.
6083 This option also warns when @code{alloca} is used in a loop.
6085 @option{-Walloca-larger-than=}@samp{PTRDIFF_MAX} is enabled by default
6086 but is usually only effective when @option{-ftree-vrp} is active (default
6087 for @option{-O2} and above).
6089 See also @option{-Wvla-larger-than=}@samp{byte-size}.
6091 @item -Wno-alloca-larger-than
6092 @opindex Wno-alloca-larger-than
6093 Disable @option{-Walloca-larger-than=} warnings. The option is
6094 equivalent to @option{-Walloca-larger-than=}@samp{SIZE_MAX} or larger.
6096 @item -Warray-bounds
6097 @itemx -Warray-bounds=@var{n}
6098 @opindex Wno-array-bounds
6099 @opindex Warray-bounds
6100 This option is only active when @option{-ftree-vrp} is active
6101 (default for @option{-O2} and above). It warns about subscripts to arrays
6102 that are always out of bounds. This warning is enabled by @option{-Wall}.
6105 @item -Warray-bounds=1
6106 This is the warning level of @option{-Warray-bounds} and is enabled
6107 by @option{-Wall}; higher levels are not, and must be explicitly requested.
6109 @item -Warray-bounds=2
6110 This warning level also warns about out of bounds access for
6111 arrays at the end of a struct and for arrays accessed through
6112 pointers. This warning level may give a larger number of
6113 false positives and is deactivated by default.
6116 @item -Wattribute-alias=@var{n}
6117 @itemx -Wno-attribute-alias
6118 @opindex -Wattribute-alias
6119 @opindex -Wno-attribute-alias
6120 Warn about declarations using the @code{alias} and similar attributes whose
6121 target is incompatible with the type of the alias.
6122 @xref{Function Attributes,,Declaring Attributes of Functions}.
6125 @item -Wattribute-alias=1
6126 The default warning level of the @option{-Wattribute-alias} option diagnoses
6127 incompatibilities between the type of the alias declaration and that of its
6128 target. Such incompatibilities are typically indicative of bugs.
6130 @item -Wattribute-alias=2
6132 At this level @option{-Wattribute-alias} also diagnoses cases where
6133 the attributes of the alias declaration are more restrictive than the
6134 attributes applied to its target. These mismatches can potentially
6135 result in incorrect code generation. In other cases they may be
6136 benign and could be resolved simply by adding the missing attribute to
6137 the target. For comparison, see the @option{-Wmissing-attributes}
6138 option, which controls diagnostics when the alias declaration is less
6139 restrictive than the target, rather than more restrictive.
6141 Attributes considered include @code{alloc_align}, @code{alloc_size},
6142 @code{cold}, @code{const}, @code{hot}, @code{leaf}, @code{malloc},
6143 @code{nonnull}, @code{noreturn}, @code{nothrow}, @code{pure},
6144 @code{returns_nonnull}, and @code{returns_twice}.
6147 @option{-Wattribute-alias} is equivalent to @option{-Wattribute-alias=1}.
6148 This is the default. You can disable these warnings with either
6149 @option{-Wno-attribute-alias} or @option{-Wattribute-alias=0}.
6151 @item -Wbool-compare
6152 @opindex Wno-bool-compare
6153 @opindex Wbool-compare
6154 Warn about boolean expression compared with an integer value different from
6155 @code{true}/@code{false}. For instance, the following comparison is
6160 if ((n > 1) == 2) @{ @dots{} @}
6162 This warning is enabled by @option{-Wall}.
6164 @item -Wbool-operation
6165 @opindex Wno-bool-operation
6166 @opindex Wbool-operation
6167 Warn about suspicious operations on expressions of a boolean type. For
6168 instance, bitwise negation of a boolean is very likely a bug in the program.
6169 For C, this warning also warns about incrementing or decrementing a boolean,
6170 which rarely makes sense. (In C++, decrementing a boolean is always invalid.
6171 Incrementing a boolean is invalid in C++17, and deprecated otherwise.)
6173 This warning is enabled by @option{-Wall}.
6175 @item -Wduplicated-branches
6176 @opindex Wno-duplicated-branches
6177 @opindex Wduplicated-branches
6178 Warn when an if-else has identical branches. This warning detects cases like
6185 It doesn't warn when both branches contain just a null statement. This warning
6186 also warn for conditional operators:
6188 int i = x ? *p : *p;
6191 @item -Wduplicated-cond
6192 @opindex Wno-duplicated-cond
6193 @opindex Wduplicated-cond
6194 Warn about duplicated conditions in an if-else-if chain. For instance,
6195 warn for the following code:
6197 if (p->q != NULL) @{ @dots{} @}
6198 else if (p->q != NULL) @{ @dots{} @}
6201 @item -Wframe-address
6202 @opindex Wno-frame-address
6203 @opindex Wframe-address
6204 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
6205 is called with an argument greater than 0. Such calls may return indeterminate
6206 values or crash the program. The warning is included in @option{-Wall}.
6208 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
6209 @opindex Wno-discarded-qualifiers
6210 @opindex Wdiscarded-qualifiers
6211 Do not warn if type qualifiers on pointers are being discarded.
6212 Typically, the compiler warns if a @code{const char *} variable is
6213 passed to a function that takes a @code{char *} parameter. This option
6214 can be used to suppress such a warning.
6216 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
6217 @opindex Wno-discarded-array-qualifiers
6218 @opindex Wdiscarded-array-qualifiers
6219 Do not warn if type qualifiers on arrays which are pointer targets
6220 are being discarded. Typically, the compiler warns if a
6221 @code{const int (*)[]} variable is passed to a function that
6222 takes a @code{int (*)[]} parameter. This option can be used to
6223 suppress such a warning.
6225 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
6226 @opindex Wno-incompatible-pointer-types
6227 @opindex Wincompatible-pointer-types
6228 Do not warn when there is a conversion between pointers that have incompatible
6229 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
6230 which warns for pointer argument passing or assignment with different
6233 @item -Wno-int-conversion @r{(C and Objective-C only)}
6234 @opindex Wno-int-conversion
6235 @opindex Wint-conversion
6236 Do not warn about incompatible integer to pointer and pointer to integer
6237 conversions. This warning is about implicit conversions; for explicit
6238 conversions the warnings @option{-Wno-int-to-pointer-cast} and
6239 @option{-Wno-pointer-to-int-cast} may be used.
6241 @item -Wno-div-by-zero
6242 @opindex Wno-div-by-zero
6243 @opindex Wdiv-by-zero
6244 Do not warn about compile-time integer division by zero. Floating-point
6245 division by zero is not warned about, as it can be a legitimate way of
6246 obtaining infinities and NaNs.
6248 @item -Wsystem-headers
6249 @opindex Wsystem-headers
6250 @opindex Wno-system-headers
6251 @cindex warnings from system headers
6252 @cindex system headers, warnings from
6253 Print warning messages for constructs found in system header files.
6254 Warnings from system headers are normally suppressed, on the assumption
6255 that they usually do not indicate real problems and would only make the
6256 compiler output harder to read. Using this command-line option tells
6257 GCC to emit warnings from system headers as if they occurred in user
6258 code. However, note that using @option{-Wall} in conjunction with this
6259 option does @emph{not} warn about unknown pragmas in system
6260 headers---for that, @option{-Wunknown-pragmas} must also be used.
6262 @item -Wtautological-compare
6263 @opindex Wtautological-compare
6264 @opindex Wno-tautological-compare
6265 Warn if a self-comparison always evaluates to true or false. This
6266 warning detects various mistakes such as:
6270 if (i > i) @{ @dots{} @}
6273 This warning also warns about bitwise comparisons that always evaluate
6274 to true or false, for instance:
6276 if ((a & 16) == 10) @{ @dots{} @}
6278 will always be false.
6280 This warning is enabled by @option{-Wall}.
6283 @opindex Wtrampolines
6284 @opindex Wno-trampolines
6285 Warn about trampolines generated for pointers to nested functions.
6286 A trampoline is a small piece of data or code that is created at run
6287 time on the stack when the address of a nested function is taken, and is
6288 used to call the nested function indirectly. For some targets, it is
6289 made up of data only and thus requires no special treatment. But, for
6290 most targets, it is made up of code and thus requires the stack to be
6291 made executable in order for the program to work properly.
6294 @opindex Wfloat-equal
6295 @opindex Wno-float-equal
6296 Warn if floating-point values are used in equality comparisons.
6298 The idea behind this is that sometimes it is convenient (for the
6299 programmer) to consider floating-point values as approximations to
6300 infinitely precise real numbers. If you are doing this, then you need
6301 to compute (by analyzing the code, or in some other way) the maximum or
6302 likely maximum error that the computation introduces, and allow for it
6303 when performing comparisons (and when producing output, but that's a
6304 different problem). In particular, instead of testing for equality, you
6305 should check to see whether the two values have ranges that overlap; and
6306 this is done with the relational operators, so equality comparisons are
6309 @item -Wtraditional @r{(C and Objective-C only)}
6310 @opindex Wtraditional
6311 @opindex Wno-traditional
6312 Warn about certain constructs that behave differently in traditional and
6313 ISO C@. Also warn about ISO C constructs that have no traditional C
6314 equivalent, and/or problematic constructs that should be avoided.
6318 Macro parameters that appear within string literals in the macro body.
6319 In traditional C macro replacement takes place within string literals,
6320 but in ISO C it does not.
6323 In traditional C, some preprocessor directives did not exist.
6324 Traditional preprocessors only considered a line to be a directive
6325 if the @samp{#} appeared in column 1 on the line. Therefore
6326 @option{-Wtraditional} warns about directives that traditional C
6327 understands but ignores because the @samp{#} does not appear as the
6328 first character on the line. It also suggests you hide directives like
6329 @code{#pragma} not understood by traditional C by indenting them. Some
6330 traditional implementations do not recognize @code{#elif}, so this option
6331 suggests avoiding it altogether.
6334 A function-like macro that appears without arguments.
6337 The unary plus operator.
6340 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
6341 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
6342 constants.) Note, these suffixes appear in macros defined in the system
6343 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
6344 Use of these macros in user code might normally lead to spurious
6345 warnings, however GCC's integrated preprocessor has enough context to
6346 avoid warning in these cases.
6349 A function declared external in one block and then used after the end of
6353 A @code{switch} statement has an operand of type @code{long}.
6356 A non-@code{static} function declaration follows a @code{static} one.
6357 This construct is not accepted by some traditional C compilers.
6360 The ISO type of an integer constant has a different width or
6361 signedness from its traditional type. This warning is only issued if
6362 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
6363 typically represent bit patterns, are not warned about.
6366 Usage of ISO string concatenation is detected.
6369 Initialization of automatic aggregates.
6372 Identifier conflicts with labels. Traditional C lacks a separate
6373 namespace for labels.
6376 Initialization of unions. If the initializer is zero, the warning is
6377 omitted. This is done under the assumption that the zero initializer in
6378 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
6379 initializer warnings and relies on default initialization to zero in the
6383 Conversions by prototypes between fixed/floating-point values and vice
6384 versa. The absence of these prototypes when compiling with traditional
6385 C causes serious problems. This is a subset of the possible
6386 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
6389 Use of ISO C style function definitions. This warning intentionally is
6390 @emph{not} issued for prototype declarations or variadic functions
6391 because these ISO C features appear in your code when using
6392 libiberty's traditional C compatibility macros, @code{PARAMS} and
6393 @code{VPARAMS}. This warning is also bypassed for nested functions
6394 because that feature is already a GCC extension and thus not relevant to
6395 traditional C compatibility.
6398 @item -Wtraditional-conversion @r{(C and Objective-C only)}
6399 @opindex Wtraditional-conversion
6400 @opindex Wno-traditional-conversion
6401 Warn if a prototype causes a type conversion that is different from what
6402 would happen to the same argument in the absence of a prototype. This
6403 includes conversions of fixed point to floating and vice versa, and
6404 conversions changing the width or signedness of a fixed-point argument
6405 except when the same as the default promotion.
6407 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
6408 @opindex Wdeclaration-after-statement
6409 @opindex Wno-declaration-after-statement
6410 Warn when a declaration is found after a statement in a block. This
6411 construct, known from C++, was introduced with ISO C99 and is by default
6412 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
6417 Warn whenever a local variable or type declaration shadows another
6418 variable, parameter, type, class member (in C++), or instance variable
6419 (in Objective-C) or whenever a built-in function is shadowed. Note
6420 that in C++, the compiler warns if a local variable shadows an
6421 explicit typedef, but not if it shadows a struct/class/enum.
6422 Same as @option{-Wshadow=global}.
6424 @item -Wno-shadow-ivar @r{(Objective-C only)}
6425 @opindex Wno-shadow-ivar
6426 @opindex Wshadow-ivar
6427 Do not warn whenever a local variable shadows an instance variable in an
6430 @item -Wshadow=global
6431 @opindex Wshadow=local
6432 The default for @option{-Wshadow}. Warns for any (global) shadowing.
6434 @item -Wshadow=local
6435 @opindex Wshadow=local
6436 Warn when a local variable shadows another local variable or parameter.
6437 This warning is enabled by @option{-Wshadow=global}.
6439 @item -Wshadow=compatible-local
6440 @opindex Wshadow=compatible-local
6441 Warn when a local variable shadows another local variable or parameter
6442 whose type is compatible with that of the shadowing variable. In C++,
6443 type compatibility here means the type of the shadowing variable can be
6444 converted to that of the shadowed variable. The creation of this flag
6445 (in addition to @option{-Wshadow=local}) is based on the idea that when
6446 a local variable shadows another one of incompatible type, it is most
6447 likely intentional, not a bug or typo, as shown in the following example:
6451 for (SomeIterator i = SomeObj.begin(); i != SomeObj.end(); ++i)
6453 for (int i = 0; i < N; ++i)
6462 Since the two variable @code{i} in the example above have incompatible types,
6463 enabling only @option{-Wshadow=compatible-local} will not emit a warning.
6464 Because their types are incompatible, if a programmer accidentally uses one
6465 in place of the other, type checking will catch that and emit an error or
6466 warning. So not warning (about shadowing) in this case will not lead to
6467 undetected bugs. Use of this flag instead of @option{-Wshadow=local} can
6468 possibly reduce the number of warnings triggered by intentional shadowing.
6470 This warning is enabled by @option{-Wshadow=local}.
6472 @item -Wlarger-than=@var{byte-size}
6473 @opindex Wlarger-than=
6474 @opindex Wlarger-than-@var{byte-size}
6475 Warn whenever an object is defined whose size exceeds @var{byte-size}.
6476 @option{-Wlarger-than=}@samp{PTRDIFF_MAX} is enabled by default.
6477 Warnings controlled by the option can be disabled either by specifying
6478 @var{byte-size} of @samp{SIZE_MAX} or more or by
6479 @option{-Wno-larger-than}.
6481 @item -Wno-larger-than
6482 @opindex Wno-larger-than
6483 Disable @option{-Wlarger-than=} warnings. The option is equivalent
6484 to @option{-Wlarger-than=}@samp{SIZE_MAX} or larger.
6486 @item -Wframe-larger-than=@var{byte-size}
6487 @opindex Wframe-larger-than=
6488 @opindex Wno-frame-larger-than
6489 Warn if the size of a function frame exceeds @var{byte-size}.
6490 The computation done to determine the stack frame size is approximate
6491 and not conservative.
6492 The actual requirements may be somewhat greater than @var{byte-size}
6493 even if you do not get a warning. In addition, any space allocated
6494 via @code{alloca}, variable-length arrays, or related constructs
6495 is not included by the compiler when determining
6496 whether or not to issue a warning.
6497 @option{-Wframe-larger-than=}@samp{PTRDIFF_MAX} is enabled by default.
6498 Warnings controlled by the option can be disabled either by specifying
6499 @var{byte-size} of @samp{SIZE_MAX} or more or by
6500 @option{-Wno-frame-larger-than}.
6502 @item -Wno-frame-larger-than
6503 @opindex Wno-frame-larger-than
6504 Disable @option{-Wframe-larger-than=} warnings. The option is equivalent
6505 to @option{-Wframe-larger-than=}@samp{SIZE_MAX} or larger.
6507 @item -Wno-free-nonheap-object
6508 @opindex Wno-free-nonheap-object
6509 @opindex Wfree-nonheap-object
6510 Do not warn when attempting to free an object that was not allocated
6513 @item -Wstack-usage=@var{byte-size}
6514 @opindex Wstack-usage
6515 @opindex Wno-stack-usage
6516 Warn if the stack usage of a function might exceed @var{byte-size}.
6517 The computation done to determine the stack usage is conservative.
6518 Any space allocated via @code{alloca}, variable-length arrays, or related
6519 constructs is included by the compiler when determining whether or not to
6522 The message is in keeping with the output of @option{-fstack-usage}.
6526 If the stack usage is fully static but exceeds the specified amount, it's:
6529 warning: stack usage is 1120 bytes
6532 If the stack usage is (partly) dynamic but bounded, it's:
6535 warning: stack usage might be 1648 bytes
6538 If the stack usage is (partly) dynamic and not bounded, it's:
6541 warning: stack usage might be unbounded
6545 @option{-Wstack-usage=}@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-stack-usage}.
6550 @item -Wno-stack-usage
6551 @opindex Wno-stack-usage
6552 Disable @option{-Wstack-usage=} warnings. The option is equivalent
6553 to @option{-Wstack-usage=}@samp{SIZE_MAX} or larger.
6555 @item -Wunsafe-loop-optimizations
6556 @opindex Wunsafe-loop-optimizations
6557 @opindex Wno-unsafe-loop-optimizations
6558 Warn if the loop cannot be optimized because the compiler cannot
6559 assume anything on the bounds of the loop indices. With
6560 @option{-funsafe-loop-optimizations} warn if the compiler makes
6563 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
6564 @opindex Wno-pedantic-ms-format
6565 @opindex Wpedantic-ms-format
6566 When used in combination with @option{-Wformat}
6567 and @option{-pedantic} without GNU extensions, this option
6568 disables the warnings about non-ISO @code{printf} / @code{scanf} format
6569 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
6570 which depend on the MS runtime.
6573 @opindex Waligned-new
6574 @opindex Wno-aligned-new
6575 Warn about a new-expression of a type that requires greater alignment
6576 than the @code{alignof(std::max_align_t)} but uses an allocation
6577 function without an explicit alignment parameter. This option is
6578 enabled by @option{-Wall}.
6580 Normally this only warns about global allocation functions, but
6581 @option{-Waligned-new=all} also warns about class member allocation
6584 @item -Wplacement-new
6585 @itemx -Wplacement-new=@var{n}
6586 @opindex Wplacement-new
6587 @opindex Wno-placement-new
6588 Warn about placement new expressions with undefined behavior, such as
6589 constructing an object in a buffer that is smaller than the type of
6590 the object. For example, the placement new expression below is diagnosed
6591 because it attempts to construct an array of 64 integers in a buffer only
6597 This warning is enabled by default.
6600 @item -Wplacement-new=1
6601 This is the default warning level of @option{-Wplacement-new}. At this
6602 level the warning is not issued for some strictly undefined constructs that
6603 GCC allows as extensions for compatibility with legacy code. For example,
6604 the following @code{new} expression is not diagnosed at this level even
6605 though it has undefined behavior according to the C++ standard because
6606 it writes past the end of the one-element array.
6608 struct S @{ int n, a[1]; @};
6609 S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]);
6610 new (s->a)int [32]();
6613 @item -Wplacement-new=2
6614 At this level, in addition to diagnosing all the same constructs as at level
6615 1, a diagnostic is also issued for placement new expressions that construct
6616 an object in the last member of structure whose type is an array of a single
6617 element and whose size is less than the size of the object being constructed.
6618 While the previous example would be diagnosed, the following construct makes
6619 use of the flexible member array extension to avoid the warning at level 2.
6621 struct S @{ int n, a[]; @};
6622 S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]);
6623 new (s->a)int [32]();
6628 @item -Wpointer-arith
6629 @opindex Wpointer-arith
6630 @opindex Wno-pointer-arith
6631 Warn about anything that depends on the ``size of'' a function type or
6632 of @code{void}. GNU C assigns these types a size of 1, for
6633 convenience in calculations with @code{void *} pointers and pointers
6634 to functions. In C++, warn also when an arithmetic operation involves
6635 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
6637 @item -Wpointer-compare
6638 @opindex Wpointer-compare
6639 @opindex Wno-pointer-compare
6640 Warn if a pointer is compared with a zero character constant. This usually
6641 means that the pointer was meant to be dereferenced. For example:
6644 const char *p = foo ();
6649 Note that the code above is invalid in C++11.
6651 This warning is enabled by default.
6654 @opindex Wtype-limits
6655 @opindex Wno-type-limits
6656 Warn if a comparison is always true or always false due to the limited
6657 range of the data type, but do not warn for constant expressions. For
6658 example, warn if an unsigned variable is compared against zero with
6659 @code{<} or @code{>=}. This warning is also enabled by
6662 @item -Wabsolute-value @r{(C and Objective-C only)}
6663 @opindex Wabsolute-value
6664 @opindex Wno-absolute-value
6665 Warn for calls to standard functions that compute the absolute value
6666 of an argument when a more appropriate standard function is available.
6667 For example, calling @code{abs(3.14)} triggers the warning because the
6668 appropriate function to call to compute the absolute value of a double
6669 argument is @code{fabs}. The option also triggers warnings when the
6670 argument in a call to such a function has an unsigned type. This
6671 warning can be suppressed with an explicit type cast and it is also
6672 enabled by @option{-Wextra}.
6674 @include cppwarnopts.texi
6676 @item -Wbad-function-cast @r{(C and Objective-C only)}
6677 @opindex Wbad-function-cast
6678 @opindex Wno-bad-function-cast
6679 Warn when a function call is cast to a non-matching type.
6680 For example, warn if a call to a function returning an integer type
6681 is cast to a pointer type.
6683 @item -Wc90-c99-compat @r{(C and Objective-C only)}
6684 @opindex Wc90-c99-compat
6685 @opindex Wno-c90-c99-compat
6686 Warn about features not present in ISO C90, but present in ISO C99.
6687 For instance, warn about use of variable length arrays, @code{long long}
6688 type, @code{bool} type, compound literals, designated initializers, and so
6689 on. This option is independent of the standards mode. Warnings are disabled
6690 in the expression that follows @code{__extension__}.
6692 @item -Wc99-c11-compat @r{(C and Objective-C only)}
6693 @opindex Wc99-c11-compat
6694 @opindex Wno-c99-c11-compat
6695 Warn about features not present in ISO C99, but present in ISO C11.
6696 For instance, warn about use of anonymous structures and unions,
6697 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
6698 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
6699 and so on. This option is independent of the standards mode. Warnings are
6700 disabled in the expression that follows @code{__extension__}.
6702 @item -Wc++-compat @r{(C and Objective-C only)}
6703 @opindex Wc++-compat
6704 @opindex Wno-c++-compat
6705 Warn about ISO C constructs that are outside of the common subset of
6706 ISO C and ISO C++, e.g.@: request for implicit conversion from
6707 @code{void *} to a pointer to non-@code{void} type.
6709 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
6710 @opindex Wc++11-compat
6711 @opindex Wno-c++11-compat
6712 Warn about C++ constructs whose meaning differs between ISO C++ 1998
6713 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
6714 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
6715 enabled by @option{-Wall}.
6717 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
6718 @opindex Wc++14-compat
6719 @opindex Wno-c++14-compat
6720 Warn about C++ constructs whose meaning differs between ISO C++ 2011
6721 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
6723 @item -Wc++17-compat @r{(C++ and Objective-C++ only)}
6724 @opindex Wc++17-compat
6725 @opindex Wno-c++17-compat
6726 Warn about C++ constructs whose meaning differs between ISO C++ 2014
6727 and ISO C++ 2017. This warning is enabled by @option{-Wall}.
6731 @opindex Wno-cast-qual
6732 Warn whenever a pointer is cast so as to remove a type qualifier from
6733 the target type. For example, warn if a @code{const char *} is cast
6734 to an ordinary @code{char *}.
6736 Also warn when making a cast that introduces a type qualifier in an
6737 unsafe way. For example, casting @code{char **} to @code{const char **}
6738 is unsafe, as in this example:
6741 /* p is char ** value. */
6742 const char **q = (const char **) p;
6743 /* Assignment of readonly string to const char * is OK. */
6745 /* Now char** pointer points to read-only memory. */
6750 @opindex Wcast-align
6751 @opindex Wno-cast-align
6752 Warn whenever a pointer is cast such that the required alignment of the
6753 target is increased. For example, warn if a @code{char *} is cast to
6754 an @code{int *} on machines where integers can only be accessed at
6755 two- or four-byte boundaries.
6757 @item -Wcast-align=strict
6758 @opindex Wcast-align=strict
6759 Warn whenever a pointer is cast such that the required alignment of the
6760 target is increased. For example, warn if a @code{char *} is cast to
6761 an @code{int *} regardless of the target machine.
6763 @item -Wcast-function-type
6764 @opindex Wcast-function-type
6765 @opindex Wno-cast-function-type
6766 Warn when a function pointer is cast to an incompatible function pointer.
6767 In a cast involving function types with a variable argument list only
6768 the types of initial arguments that are provided are considered.
6769 Any parameter of pointer-type matches any other pointer-type. Any benign
6770 differences in integral types are ignored, like @code{int} vs.@: @code{long}
6771 on ILP32 targets. Likewise type qualifiers are ignored. The function
6772 type @code{void (*) (void)} is special and matches everything, which can
6773 be used to suppress this warning.
6774 In a cast involving pointer to member types this warning warns whenever
6775 the type cast is changing the pointer to member type.
6776 This warning is enabled by @option{-Wextra}.
6778 @item -Wwrite-strings
6779 @opindex Wwrite-strings
6780 @opindex Wno-write-strings
6781 When compiling C, give string constants the type @code{const
6782 char[@var{length}]} so that copying the address of one into a
6783 non-@code{const} @code{char *} pointer produces a warning. These
6784 warnings help you find at compile time code that can try to write
6785 into a string constant, but only if you have been very careful about
6786 using @code{const} in declarations and prototypes. Otherwise, it is
6787 just a nuisance. This is why we did not make @option{-Wall} request
6790 When compiling C++, warn about the deprecated conversion from string
6791 literals to @code{char *}. This warning is enabled by default for C++
6795 @itemx -Wcatch-value=@var{n} @r{(C++ and Objective-C++ only)}
6796 @opindex Wcatch-value
6797 @opindex Wno-catch-value
6798 Warn about catch handlers that do not catch via reference.
6799 With @option{-Wcatch-value=1} (or @option{-Wcatch-value} for short)
6800 warn about polymorphic class types that are caught by value.
6801 With @option{-Wcatch-value=2} warn about all class types that are caught
6802 by value. With @option{-Wcatch-value=3} warn about all types that are
6803 not caught by reference. @option{-Wcatch-value} is enabled by @option{-Wall}.
6807 @opindex Wno-clobbered
6808 Warn for variables that might be changed by @code{longjmp} or
6809 @code{vfork}. This warning is also enabled by @option{-Wextra}.
6811 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
6812 @opindex Wconditionally-supported
6813 @opindex Wno-conditionally-supported
6814 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
6817 @opindex Wconversion
6818 @opindex Wno-conversion
6819 Warn for implicit conversions that may alter a value. This includes
6820 conversions between real and integer, like @code{abs (x)} when
6821 @code{x} is @code{double}; conversions between signed and unsigned,
6822 like @code{unsigned ui = -1}; and conversions to smaller types, like
6823 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
6824 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
6825 changed by the conversion like in @code{abs (2.0)}. Warnings about
6826 conversions between signed and unsigned integers can be disabled by
6827 using @option{-Wno-sign-conversion}.
6829 For C++, also warn for confusing overload resolution for user-defined
6830 conversions; and conversions that never use a type conversion
6831 operator: conversions to @code{void}, the same type, a base class or a
6832 reference to them. Warnings about conversions between signed and
6833 unsigned integers are disabled by default in C++ unless
6834 @option{-Wsign-conversion} is explicitly enabled.
6836 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
6837 @opindex Wconversion-null
6838 @opindex Wno-conversion-null
6839 Do not warn for conversions between @code{NULL} and non-pointer
6840 types. @option{-Wconversion-null} is enabled by default.
6842 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
6843 @opindex Wzero-as-null-pointer-constant
6844 @opindex Wno-zero-as-null-pointer-constant
6845 Warn when a literal @samp{0} is used as null pointer constant. This can
6846 be useful to facilitate the conversion to @code{nullptr} in C++11.
6848 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
6849 @opindex Wsubobject-linkage
6850 @opindex Wno-subobject-linkage
6851 Warn if a class type has a base or a field whose type uses the anonymous
6852 namespace or depends on a type with no linkage. If a type A depends on
6853 a type B with no or internal linkage, defining it in multiple
6854 translation units would be an ODR violation because the meaning of B
6855 is different in each translation unit. If A only appears in a single
6856 translation unit, the best way to silence the warning is to give it
6857 internal linkage by putting it in an anonymous namespace as well. The
6858 compiler doesn't give this warning for types defined in the main .C
6859 file, as those are unlikely to have multiple definitions.
6860 @option{-Wsubobject-linkage} is enabled by default.
6862 @item -Wdangling-else
6863 @opindex Wdangling-else
6864 @opindex Wno-dangling-else
6865 Warn about constructions where there may be confusion to which
6866 @code{if} statement an @code{else} branch belongs. Here is an example of
6881 In C/C++, every @code{else} branch belongs to the innermost possible
6882 @code{if} statement, which in this example is @code{if (b)}. This is
6883 often not what the programmer expected, as illustrated in the above
6884 example by indentation the programmer chose. When there is the
6885 potential for this confusion, GCC issues a warning when this flag
6886 is specified. To eliminate the warning, add explicit braces around
6887 the innermost @code{if} statement so there is no way the @code{else}
6888 can belong to the enclosing @code{if}. The resulting code
6905 This warning is enabled by @option{-Wparentheses}.
6909 @opindex Wno-date-time
6910 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
6911 are encountered as they might prevent bit-wise-identical reproducible
6914 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
6915 @opindex Wdelete-incomplete
6916 @opindex Wno-delete-incomplete
6917 Warn when deleting a pointer to incomplete type, which may cause
6918 undefined behavior at runtime. This warning is enabled by default.
6920 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
6921 @opindex Wuseless-cast
6922 @opindex Wno-useless-cast
6923 Warn when an expression is casted to its own type.
6926 @opindex Wempty-body
6927 @opindex Wno-empty-body
6928 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
6929 while} statement. This warning is also enabled by @option{-Wextra}.
6931 @item -Wenum-compare
6932 @opindex Wenum-compare
6933 @opindex Wno-enum-compare
6934 Warn about a comparison between values of different enumerated types.
6935 In C++ enumerated type mismatches in conditional expressions are also
6936 diagnosed and the warning is enabled by default. In C this warning is
6937 enabled by @option{-Wall}.
6939 @item -Wextra-semi @r{(C++, Objective-C++ only)}
6940 @opindex Wextra-semi
6941 @opindex Wno-extra-semi
6942 Warn about redundant semicolon after in-class function definition.
6944 @item -Wjump-misses-init @r{(C, Objective-C only)}
6945 @opindex Wjump-misses-init
6946 @opindex Wno-jump-misses-init
6947 Warn if a @code{goto} statement or a @code{switch} statement jumps
6948 forward across the initialization of a variable, or jumps backward to a
6949 label after the variable has been initialized. This only warns about
6950 variables that are initialized when they are declared. This warning is
6951 only supported for C and Objective-C; in C++ this sort of branch is an
6954 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
6955 can be disabled with the @option{-Wno-jump-misses-init} option.
6957 @item -Wsign-compare
6958 @opindex Wsign-compare
6959 @opindex Wno-sign-compare
6960 @cindex warning for comparison of signed and unsigned values
6961 @cindex comparison of signed and unsigned values, warning
6962 @cindex signed and unsigned values, comparison warning
6963 Warn when a comparison between signed and unsigned values could produce
6964 an incorrect result when the signed value is converted to unsigned.
6965 In C++, this warning is also enabled by @option{-Wall}. In C, it is
6966 also enabled by @option{-Wextra}.
6968 @item -Wsign-conversion
6969 @opindex Wsign-conversion
6970 @opindex Wno-sign-conversion
6971 Warn for implicit conversions that may change the sign of an integer
6972 value, like assigning a signed integer expression to an unsigned
6973 integer variable. An explicit cast silences the warning. In C, this
6974 option is enabled also by @option{-Wconversion}.
6976 @item -Wfloat-conversion
6977 @opindex Wfloat-conversion
6978 @opindex Wno-float-conversion
6979 Warn for implicit conversions that reduce the precision of a real value.
6980 This includes conversions from real to integer, and from higher precision
6981 real to lower precision real values. This option is also enabled by
6982 @option{-Wconversion}.
6984 @item -Wno-scalar-storage-order
6985 @opindex Wno-scalar-storage-order
6986 @opindex Wscalar-storage-order
6987 Do not warn on suspicious constructs involving reverse scalar storage order.
6989 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
6990 @opindex Wsized-deallocation
6991 @opindex Wno-sized-deallocation
6992 Warn about a definition of an unsized deallocation function
6994 void operator delete (void *) noexcept;
6995 void operator delete[] (void *) noexcept;
6997 without a definition of the corresponding sized deallocation function
6999 void operator delete (void *, std::size_t) noexcept;
7000 void operator delete[] (void *, std::size_t) noexcept;
7002 or vice versa. Enabled by @option{-Wextra} along with
7003 @option{-fsized-deallocation}.
7005 @item -Wsizeof-pointer-div
7006 @opindex Wsizeof-pointer-div
7007 @opindex Wno-sizeof-pointer-div
7008 Warn for suspicious divisions of two sizeof expressions that divide
7009 the pointer size by the element size, which is the usual way to compute
7010 the array size but won't work out correctly with pointers. This warning
7011 warns e.g.@: about @code{sizeof (ptr) / sizeof (ptr[0])} if @code{ptr} is
7012 not an array, but a pointer. This warning is enabled by @option{-Wall}.
7014 @item -Wsizeof-pointer-memaccess
7015 @opindex Wsizeof-pointer-memaccess
7016 @opindex Wno-sizeof-pointer-memaccess
7017 Warn for suspicious length parameters to certain string and memory built-in
7018 functions if the argument uses @code{sizeof}. This warning triggers for
7019 example for @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not
7020 an array, but a pointer, and suggests a possible fix, or about
7021 @code{memcpy (&foo, ptr, sizeof (&foo));}. @option{-Wsizeof-pointer-memaccess}
7022 also warns about calls to bounded string copy functions like @code{strncat}
7023 or @code{strncpy} that specify as the bound a @code{sizeof} expression of
7024 the source array. For example, in the following function the call to
7025 @code{strncat} specifies the size of the source string as the bound. That
7026 is almost certainly a mistake and so the call is diagnosed.
7028 void make_file (const char *name)
7030 char path[PATH_MAX];
7031 strncpy (path, name, sizeof path - 1);
7032 strncat (path, ".text", sizeof ".text");
7037 The @option{-Wsizeof-pointer-memaccess} option is enabled by @option{-Wall}.
7039 @item -Wsizeof-array-argument
7040 @opindex Wsizeof-array-argument
7041 @opindex Wno-sizeof-array-argument
7042 Warn when the @code{sizeof} operator is applied to a parameter that is
7043 declared as an array in a function definition. This warning is enabled by
7044 default for C and C++ programs.
7046 @item -Wmemset-elt-size
7047 @opindex Wmemset-elt-size
7048 @opindex Wno-memset-elt-size
7049 Warn for suspicious calls to the @code{memset} built-in function, if the
7050 first argument references an array, and the third argument is a number
7051 equal to the number of elements, but not equal to the size of the array
7052 in memory. This indicates that the user has omitted a multiplication by
7053 the element size. This warning is enabled by @option{-Wall}.
7055 @item -Wmemset-transposed-args
7056 @opindex Wmemset-transposed-args
7057 @opindex Wno-memset-transposed-args
7058 Warn for suspicious calls to the @code{memset} built-in function where
7059 the second argument is not zero and the third argument is zero. For
7060 example, the call @code{memset (buf, sizeof buf, 0)} is diagnosed because
7061 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostic
7062 is only emitted if the third argument is a literal zero. Otherwise, if
7063 it is an expression that is folded to zero, or a cast of zero to some
7064 type, it is far less likely that the arguments have been mistakenly
7065 transposed and no warning is emitted. This warning is enabled
7070 @opindex Wno-address
7071 Warn about suspicious uses of memory addresses. These include using
7072 the address of a function in a conditional expression, such as
7073 @code{void func(void); if (func)}, and comparisons against the memory
7074 address of a string literal, such as @code{if (x == "abc")}. Such
7075 uses typically indicate a programmer error: the address of a function
7076 always evaluates to true, so their use in a conditional usually
7077 indicate that the programmer forgot the parentheses in a function
7078 call; and comparisons against string literals result in unspecified
7079 behavior and are not portable in C, so they usually indicate that the
7080 programmer intended to use @code{strcmp}. This warning is enabled by
7083 @item -Waddress-of-packed-member
7084 @opindex Waddress-of-packed-member
7085 @opindex Wno-address-of-packed-member
7086 Warn when the address of packed member of struct or union is taken,
7087 which usually results in an unaligned pointer value. This is
7091 @opindex Wlogical-op
7092 @opindex Wno-logical-op
7093 Warn about suspicious uses of logical operators in expressions.
7094 This includes using logical operators in contexts where a
7095 bit-wise operator is likely to be expected. Also warns when
7096 the operands of a logical operator are the same:
7099 if (a < 0 && a < 0) @{ @dots{} @}
7102 @item -Wlogical-not-parentheses
7103 @opindex Wlogical-not-parentheses
7104 @opindex Wno-logical-not-parentheses
7105 Warn about logical not used on the left hand side operand of a comparison.
7106 This option does not warn if the right operand is considered to be a boolean
7107 expression. Its purpose is to detect suspicious code like the following:
7111 if (!a > 1) @{ @dots{} @}
7114 It is possible to suppress the warning by wrapping the LHS into
7117 if ((!a) > 1) @{ @dots{} @}
7120 This warning is enabled by @option{-Wall}.
7122 @item -Waggregate-return
7123 @opindex Waggregate-return
7124 @opindex Wno-aggregate-return
7125 Warn if any functions that return structures or unions are defined or
7126 called. (In languages where you can return an array, this also elicits
7129 @item -Wno-aggressive-loop-optimizations
7130 @opindex Wno-aggressive-loop-optimizations
7131 @opindex Waggressive-loop-optimizations
7132 Warn if in a loop with constant number of iterations the compiler detects
7133 undefined behavior in some statement during one or more of the iterations.
7135 @item -Wno-attributes
7136 @opindex Wno-attributes
7137 @opindex Wattributes
7138 Do not warn if an unexpected @code{__attribute__} is used, such as
7139 unrecognized attributes, function attributes applied to variables,
7140 etc. This does not stop errors for incorrect use of supported
7143 @item -Wno-builtin-declaration-mismatch
7144 @opindex Wno-builtin-declaration-mismatch
7145 @opindex Wbuiltin-declaration-mismatch
7146 Warn if a built-in function is declared with an incompatible signature
7147 or as a non-function, or when a built-in function declared with a type
7148 that does not include a prototype is called with arguments whose promoted
7149 types do not match those expected by the function. When @option{-Wextra}
7150 is specified, also warn when a built-in function that takes arguments is
7151 declared without a prototype. The @option{-Wno-builtin-declaration-mismatch}
7152 warning is enabled by default. To avoid the warning include the appropriate
7153 header to bring the prototypes of built-in functions into scope.
7155 For example, the call to @code{memset} below is diagnosed by the warning
7156 because the function expects a value of type @code{size_t} as its argument
7157 but the type of @code{32} is @code{int}. With @option{-Wextra},
7158 the declaration of the function is diagnosed as well.
7160 extern void* memset ();
7163 memset (d, '\0', 32);
7167 @item -Wno-builtin-macro-redefined
7168 @opindex Wno-builtin-macro-redefined
7169 @opindex Wbuiltin-macro-redefined
7170 Do not warn if certain built-in macros are redefined. This suppresses
7171 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
7172 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
7174 @item -Wstrict-prototypes @r{(C and Objective-C only)}
7175 @opindex Wstrict-prototypes
7176 @opindex Wno-strict-prototypes
7177 Warn if a function is declared or defined without specifying the
7178 argument types. (An old-style function definition is permitted without
7179 a warning if preceded by a declaration that specifies the argument
7182 @item -Wold-style-declaration @r{(C and Objective-C only)}
7183 @opindex Wold-style-declaration
7184 @opindex Wno-old-style-declaration
7185 Warn for obsolescent usages, according to the C Standard, in a
7186 declaration. For example, warn if storage-class specifiers like
7187 @code{static} are not the first things in a declaration. This warning
7188 is also enabled by @option{-Wextra}.
7190 @item -Wold-style-definition @r{(C and Objective-C only)}
7191 @opindex Wold-style-definition
7192 @opindex Wno-old-style-definition
7193 Warn if an old-style function definition is used. A warning is given
7194 even if there is a previous prototype.
7196 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
7197 @opindex Wmissing-parameter-type
7198 @opindex Wno-missing-parameter-type
7199 A function parameter is declared without a type specifier in K&R-style
7206 This warning is also enabled by @option{-Wextra}.
7208 @item -Wmissing-prototypes @r{(C and Objective-C only)}
7209 @opindex Wmissing-prototypes
7210 @opindex Wno-missing-prototypes
7211 Warn if a global function is defined without a previous prototype
7212 declaration. This warning is issued even if the definition itself
7213 provides a prototype. Use this option to detect global functions
7214 that do not have a matching prototype declaration in a header file.
7215 This option is not valid for C++ because all function declarations
7216 provide prototypes and a non-matching declaration declares an
7217 overload rather than conflict with an earlier declaration.
7218 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
7220 @item -Wmissing-declarations
7221 @opindex Wmissing-declarations
7222 @opindex Wno-missing-declarations
7223 Warn if a global function is defined without a previous declaration.
7224 Do so even if the definition itself provides a prototype.
7225 Use this option to detect global functions that are not declared in
7226 header files. In C, no warnings are issued for functions with previous
7227 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
7228 missing prototypes. In C++, no warnings are issued for function templates,
7229 or for inline functions, or for functions in anonymous namespaces.
7231 @item -Wmissing-field-initializers
7232 @opindex Wmissing-field-initializers
7233 @opindex Wno-missing-field-initializers
7237 Warn if a structure's initializer has some fields missing. For
7238 example, the following code causes such a warning, because
7239 @code{x.h} is implicitly zero:
7242 struct s @{ int f, g, h; @};
7243 struct s x = @{ 3, 4 @};
7246 This option does not warn about designated initializers, so the following
7247 modification does not trigger a warning:
7250 struct s @{ int f, g, h; @};
7251 struct s x = @{ .f = 3, .g = 4 @};
7254 In C this option does not warn about the universal zero initializer
7258 struct s @{ int f, g, h; @};
7259 struct s x = @{ 0 @};
7262 Likewise, in C++ this option does not warn about the empty @{ @}
7263 initializer, for example:
7266 struct s @{ int f, g, h; @};
7270 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
7271 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
7273 @item -Wno-multichar
7274 @opindex Wno-multichar
7276 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
7277 Usually they indicate a typo in the user's code, as they have
7278 implementation-defined values, and should not be used in portable code.
7280 @item -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]}
7281 @opindex Wnormalized=
7282 @opindex Wnormalized
7283 @opindex Wno-normalized
7286 @cindex character set, input normalization
7287 In ISO C and ISO C++, two identifiers are different if they are
7288 different sequences of characters. However, sometimes when characters
7289 outside the basic ASCII character set are used, you can have two
7290 different character sequences that look the same. To avoid confusion,
7291 the ISO 10646 standard sets out some @dfn{normalization rules} which
7292 when applied ensure that two sequences that look the same are turned into
7293 the same sequence. GCC can warn you if you are using identifiers that
7294 have not been normalized; this option controls that warning.
7296 There are four levels of warning supported by GCC@. The default is
7297 @option{-Wnormalized=nfc}, which warns about any identifier that is
7298 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
7299 recommended form for most uses. It is equivalent to
7300 @option{-Wnormalized}.
7302 Unfortunately, there are some characters allowed in identifiers by
7303 ISO C and ISO C++ that, when turned into NFC, are not allowed in
7304 identifiers. That is, there's no way to use these symbols in portable
7305 ISO C or C++ and have all your identifiers in NFC@.
7306 @option{-Wnormalized=id} suppresses the warning for these characters.
7307 It is hoped that future versions of the standards involved will correct
7308 this, which is why this option is not the default.
7310 You can switch the warning off for all characters by writing
7311 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
7312 only do this if you are using some other normalization scheme (like
7313 ``D''), because otherwise you can easily create bugs that are
7314 literally impossible to see.
7316 Some characters in ISO 10646 have distinct meanings but look identical
7317 in some fonts or display methodologies, especially once formatting has
7318 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
7319 LETTER N'', displays just like a regular @code{n} that has been
7320 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
7321 normalization scheme to convert all these into a standard form as
7322 well, and GCC warns if your code is not in NFKC if you use
7323 @option{-Wnormalized=nfkc}. This warning is comparable to warning
7324 about every identifier that contains the letter O because it might be
7325 confused with the digit 0, and so is not the default, but may be
7326 useful as a local coding convention if the programming environment
7327 cannot be fixed to display these characters distinctly.
7329 @item -Wno-attribute-warning
7330 @opindex Wno-attribute-warning
7331 @opindex Wattribute-warning
7332 Do not warn about usage of functions (@pxref{Function Attributes})
7333 declared with @code{warning} attribute. By default, this warning is
7334 enabled. @option{-Wno-attribute-warning} can be used to disable the
7335 warning or @option{-Wno-error=attribute-warning} can be used to
7336 disable the error when compiled with @option{-Werror} flag.
7338 @item -Wno-deprecated
7339 @opindex Wno-deprecated
7340 @opindex Wdeprecated
7341 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
7343 @item -Wno-deprecated-declarations
7344 @opindex Wno-deprecated-declarations
7345 @opindex Wdeprecated-declarations
7346 Do not warn about uses of functions (@pxref{Function Attributes}),
7347 variables (@pxref{Variable Attributes}), and types (@pxref{Type
7348 Attributes}) marked as deprecated by using the @code{deprecated}
7352 @opindex Wno-overflow
7354 Do not warn about compile-time overflow in constant expressions.
7359 Warn about One Definition Rule violations during link-time optimization.
7360 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
7363 @opindex Wopenmp-simd
7364 @opindex Wno-openmp-simd
7365 Warn if the vectorizer cost model overrides the OpenMP
7366 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
7367 option can be used to relax the cost model.
7369 @item -Woverride-init @r{(C and Objective-C only)}
7370 @opindex Woverride-init
7371 @opindex Wno-override-init
7375 Warn if an initialized field without side effects is overridden when
7376 using designated initializers (@pxref{Designated Inits, , Designated
7379 This warning is included in @option{-Wextra}. To get other
7380 @option{-Wextra} warnings without this one, use @option{-Wextra
7381 -Wno-override-init}.
7383 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
7384 @opindex Woverride-init-side-effects
7385 @opindex Wno-override-init-side-effects
7386 Warn if an initialized field with side effects is overridden when
7387 using designated initializers (@pxref{Designated Inits, , Designated
7388 Initializers}). This warning is enabled by default.
7393 Warn if a structure is given the packed attribute, but the packed
7394 attribute has no effect on the layout or size of the structure.
7395 Such structures may be mis-aligned for little benefit. For
7396 instance, in this code, the variable @code{f.x} in @code{struct bar}
7397 is misaligned even though @code{struct bar} does not itself
7398 have the packed attribute:
7405 @} __attribute__((packed));
7413 @item -Wpacked-bitfield-compat
7414 @opindex Wpacked-bitfield-compat
7415 @opindex Wno-packed-bitfield-compat
7416 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
7417 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
7418 the change can lead to differences in the structure layout. GCC
7419 informs you when the offset of such a field has changed in GCC 4.4.
7420 For example there is no longer a 4-bit padding between field @code{a}
7421 and @code{b} in this structure:
7428 @} __attribute__ ((packed));
7431 This warning is enabled by default. Use
7432 @option{-Wno-packed-bitfield-compat} to disable this warning.
7434 @item -Wpacked-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
7435 @opindex Wpacked-not-aligned
7436 @opindex Wno-packed-not-aligned
7437 Warn if a structure field with explicitly specified alignment in a
7438 packed struct or union is misaligned. For example, a warning will
7439 be issued on @code{struct S}, like, @code{warning: alignment 1 of
7440 'struct S' is less than 8}, in this code:
7444 struct __attribute__ ((aligned (8))) S8 @{ char a[8]; @};
7445 struct __attribute__ ((packed)) S @{
7451 This warning is enabled by @option{-Wall}.
7456 Warn if padding is included in a structure, either to align an element
7457 of the structure or to align the whole structure. Sometimes when this
7458 happens it is possible to rearrange the fields of the structure to
7459 reduce the padding and so make the structure smaller.
7461 @item -Wredundant-decls
7462 @opindex Wredundant-decls
7463 @opindex Wno-redundant-decls
7464 Warn if anything is declared more than once in the same scope, even in
7465 cases where multiple declaration is valid and changes nothing.
7469 @opindex Wno-restrict
7470 Warn when an object referenced by a @code{restrict}-qualified parameter
7471 (or, in C++, a @code{__restrict}-qualified parameter) is aliased by another
7472 argument, or when copies between such objects overlap. For example,
7473 the call to the @code{strcpy} function below attempts to truncate the string
7474 by replacing its initial characters with the last four. However, because
7475 the call writes the terminating NUL into @code{a[4]}, the copies overlap and
7476 the call is diagnosed.
7481 char a[] = "abcd1234";
7486 The @option{-Wrestrict} option detects some instances of simple overlap
7487 even without optimization but works best at @option{-O2} and above. It
7488 is included in @option{-Wall}.
7490 @item -Wnested-externs @r{(C and Objective-C only)}
7491 @opindex Wnested-externs
7492 @opindex Wno-nested-externs
7493 Warn if an @code{extern} declaration is encountered within a function.
7495 @item -Wno-inherited-variadic-ctor
7496 @opindex Winherited-variadic-ctor
7497 @opindex Wno-inherited-variadic-ctor
7498 Suppress warnings about use of C++11 inheriting constructors when the
7499 base class inherited from has a C variadic constructor; the warning is
7500 on by default because the ellipsis is not inherited.
7505 Warn if a function that is declared as inline cannot be inlined.
7506 Even with this option, the compiler does not warn about failures to
7507 inline functions declared in system headers.
7509 The compiler uses a variety of heuristics to determine whether or not
7510 to inline a function. For example, the compiler takes into account
7511 the size of the function being inlined and the amount of inlining
7512 that has already been done in the current function. Therefore,
7513 seemingly insignificant changes in the source program can cause the
7514 warnings produced by @option{-Winline} to appear or disappear.
7516 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
7517 @opindex Wno-invalid-offsetof
7518 @opindex Winvalid-offsetof
7519 Suppress warnings from applying the @code{offsetof} macro to a non-POD
7520 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
7521 to a non-standard-layout type is undefined. In existing C++ implementations,
7522 however, @code{offsetof} typically gives meaningful results.
7523 This flag is for users who are aware that they are
7524 writing nonportable code and who have deliberately chosen to ignore the
7527 The restrictions on @code{offsetof} may be relaxed in a future version
7528 of the C++ standard.
7530 @item -Wint-in-bool-context
7531 @opindex Wint-in-bool-context
7532 @opindex Wno-int-in-bool-context
7533 Warn for suspicious use of integer values where boolean values are expected,
7534 such as conditional expressions (?:) using non-boolean integer constants in
7535 boolean context, like @code{if (a <= b ? 2 : 3)}. Or left shifting of signed
7536 integers in boolean context, like @code{for (a = 0; 1 << a; a++);}. Likewise
7537 for all kinds of multiplications regardless of the data type.
7538 This warning is enabled by @option{-Wall}.
7540 @item -Wno-int-to-pointer-cast
7541 @opindex Wno-int-to-pointer-cast
7542 @opindex Wint-to-pointer-cast
7543 Suppress warnings from casts to pointer type of an integer of a
7544 different size. In C++, casting to a pointer type of smaller size is
7545 an error. @option{Wint-to-pointer-cast} is enabled by default.
7548 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
7549 @opindex Wno-pointer-to-int-cast
7550 @opindex Wpointer-to-int-cast
7551 Suppress warnings from casts from a pointer to an integer type of a
7555 @opindex Winvalid-pch
7556 @opindex Wno-invalid-pch
7557 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
7558 the search path but cannot be used.
7562 @opindex Wno-long-long
7563 Warn if @code{long long} type is used. This is enabled by either
7564 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
7565 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
7567 @item -Wvariadic-macros
7568 @opindex Wvariadic-macros
7569 @opindex Wno-variadic-macros
7570 Warn if variadic macros are used in ISO C90 mode, or if the GNU
7571 alternate syntax is used in ISO C99 mode. This is enabled by either
7572 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
7573 messages, use @option{-Wno-variadic-macros}.
7577 @opindex Wno-varargs
7578 Warn upon questionable usage of the macros used to handle variable
7579 arguments like @code{va_start}. This is default. To inhibit the
7580 warning messages, use @option{-Wno-varargs}.
7582 @item -Wvector-operation-performance
7583 @opindex Wvector-operation-performance
7584 @opindex Wno-vector-operation-performance
7585 Warn if vector operation is not implemented via SIMD capabilities of the
7586 architecture. Mainly useful for the performance tuning.
7587 Vector operation can be implemented @code{piecewise}, which means that the
7588 scalar operation is performed on every vector element;
7589 @code{in parallel}, which means that the vector operation is implemented
7590 using scalars of wider type, which normally is more performance efficient;
7591 and @code{as a single scalar}, which means that vector fits into a
7594 @item -Wno-virtual-move-assign
7595 @opindex Wvirtual-move-assign
7596 @opindex Wno-virtual-move-assign
7597 Suppress warnings about inheriting from a virtual base with a
7598 non-trivial C++11 move assignment operator. This is dangerous because
7599 if the virtual base is reachable along more than one path, it is
7600 moved multiple times, which can mean both objects end up in the
7601 moved-from state. If the move assignment operator is written to avoid
7602 moving from a moved-from object, this warning can be disabled.
7607 Warn if a variable-length array is used in the code.
7608 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
7609 the variable-length array.
7611 @item -Wvla-larger-than=@var{byte-size}
7612 @opindex Wvla-larger-than=
7613 @opindex Wno-vla-larger-than
7614 If this option is used, the compiler will warn for declarations of
7615 variable-length arrays whose size is either unbounded, or bounded
7616 by an argument that allows the array size to exceed @var{byte-size}
7617 bytes. This is similar to how @option{-Walloca-larger-than=}@var{byte-size}
7618 works, but with variable-length arrays.
7620 Note that GCC may optimize small variable-length arrays of a known
7621 value into plain arrays, so this warning may not get triggered for
7624 @option{-Wvla-larger-than=}@samp{PTRDIFF_MAX} is enabled by default but
7625 is typically only effective when @option{-ftree-vrp} is active (default
7626 for @option{-O2} and above).
7628 See also @option{-Walloca-larger-than=@var{byte-size}}.
7630 @item -Wno-vla-larger-than
7631 @opindex Wno-vla-larger-than
7632 Disable @option{-Wvla-larger-than=} warnings. The option is equivalent
7633 to @option{-Wvla-larger-than=}@samp{SIZE_MAX} or larger.
7635 @item -Wvolatile-register-var
7636 @opindex Wvolatile-register-var
7637 @opindex Wno-volatile-register-var
7638 Warn if a register variable is declared volatile. The volatile
7639 modifier does not inhibit all optimizations that may eliminate reads
7640 and/or writes to register variables. This warning is enabled by
7643 @item -Wdisabled-optimization
7644 @opindex Wdisabled-optimization
7645 @opindex Wno-disabled-optimization
7646 Warn if a requested optimization pass is disabled. This warning does
7647 not generally indicate that there is anything wrong with your code; it
7648 merely indicates that GCC's optimizers are unable to handle the code
7649 effectively. Often, the problem is that your code is too big or too
7650 complex; GCC refuses to optimize programs when the optimization
7651 itself is likely to take inordinate amounts of time.
7653 @item -Wpointer-sign @r{(C and Objective-C only)}
7654 @opindex Wpointer-sign
7655 @opindex Wno-pointer-sign
7656 Warn for pointer argument passing or assignment with different signedness.
7657 This option is only supported for C and Objective-C@. It is implied by
7658 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
7659 @option{-Wno-pointer-sign}.
7661 @item -Wstack-protector
7662 @opindex Wstack-protector
7663 @opindex Wno-stack-protector
7664 This option is only active when @option{-fstack-protector} is active. It
7665 warns about functions that are not protected against stack smashing.
7667 @item -Woverlength-strings
7668 @opindex Woverlength-strings
7669 @opindex Wno-overlength-strings
7670 Warn about string constants that are longer than the ``minimum
7671 maximum'' length specified in the C standard. Modern compilers
7672 generally allow string constants that are much longer than the
7673 standard's minimum limit, but very portable programs should avoid
7674 using longer strings.
7676 The limit applies @emph{after} string constant concatenation, and does
7677 not count the trailing NUL@. In C90, the limit was 509 characters; in
7678 C99, it was raised to 4095. C++98 does not specify a normative
7679 minimum maximum, so we do not diagnose overlength strings in C++@.
7681 This option is implied by @option{-Wpedantic}, and can be disabled with
7682 @option{-Wno-overlength-strings}.
7684 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
7685 @opindex Wunsuffixed-float-constants
7686 @opindex Wno-unsuffixed-float-constants
7688 Issue a warning for any floating constant that does not have
7689 a suffix. When used together with @option{-Wsystem-headers} it
7690 warns about such constants in system header files. This can be useful
7691 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
7692 from the decimal floating-point extension to C99.
7694 @item -Wno-designated-init @r{(C and Objective-C only)}
7695 Suppress warnings when a positional initializer is used to initialize
7696 a structure that has been marked with the @code{designated_init}
7700 Issue a warning when HSAIL cannot be emitted for the compiled function or
7705 @node Debugging Options
7706 @section Options for Debugging Your Program
7707 @cindex options, debugging
7708 @cindex debugging information options
7710 To tell GCC to emit extra information for use by a debugger, in almost
7711 all cases you need only to add @option{-g} to your other options.
7713 GCC allows you to use @option{-g} with
7714 @option{-O}. The shortcuts taken by optimized code may occasionally
7715 be surprising: some variables you declared may not exist
7716 at all; flow of control may briefly move where you did not expect it;
7717 some statements may not be executed because they compute constant
7718 results or their values are already at hand; some statements may
7719 execute in different places because they have been moved out of loops.
7720 Nevertheless it is possible to debug optimized output. This makes
7721 it reasonable to use the optimizer for programs that might have bugs.
7723 If you are not using some other optimization option, consider
7724 using @option{-Og} (@pxref{Optimize Options}) with @option{-g}.
7725 With no @option{-O} option at all, some compiler passes that collect
7726 information useful for debugging do not run at all, so that
7727 @option{-Og} may result in a better debugging experience.
7732 Produce debugging information in the operating system's native format
7733 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
7736 On most systems that use stabs format, @option{-g} enables use of extra
7737 debugging information that only GDB can use; this extra information
7738 makes debugging work better in GDB but probably makes other debuggers
7740 refuse to read the program. If you want to control for certain whether
7741 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
7742 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
7746 Produce debugging information for use by GDB@. This means to use the
7747 most expressive format available (DWARF, stabs, or the native format
7748 if neither of those are supported), including GDB extensions if at all
7752 @itemx -gdwarf-@var{version}
7754 Produce debugging information in DWARF format (if that is supported).
7755 The value of @var{version} may be either 2, 3, 4 or 5; the default version
7756 for most targets is 4. DWARF Version 5 is only experimental.
7758 Note that with DWARF Version 2, some ports require and always
7759 use some non-conflicting DWARF 3 extensions in the unwind tables.
7761 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
7762 for maximum benefit.
7764 GCC no longer supports DWARF Version 1, which is substantially
7765 different than Version 2 and later. For historical reasons, some
7766 other DWARF-related options such as
7767 @option{-fno-dwarf2-cfi-asm}) retain a reference to DWARF Version 2
7768 in their names, but apply to all currently-supported versions of DWARF.
7772 Produce debugging information in stabs format (if that is supported),
7773 without GDB extensions. This is the format used by DBX on most BSD
7774 systems. On MIPS, Alpha and System V Release 4 systems this option
7775 produces stabs debugging output that is not understood by DBX@.
7776 On System V Release 4 systems this option requires the GNU assembler.
7780 Produce debugging information in stabs format (if that is supported),
7781 using GNU extensions understood only by the GNU debugger (GDB)@. The
7782 use of these extensions is likely to make other debuggers crash or
7783 refuse to read the program.
7787 Produce debugging information in XCOFF format (if that is supported).
7788 This is the format used by the DBX debugger on IBM RS/6000 systems.
7792 Produce debugging information in XCOFF format (if that is supported),
7793 using GNU extensions understood only by the GNU debugger (GDB)@. The
7794 use of these extensions is likely to make other debuggers crash or
7795 refuse to read the program, and may cause assemblers other than the GNU
7796 assembler (GAS) to fail with an error.
7800 Produce debugging information in Alpha/VMS debug format (if that is
7801 supported). This is the format used by DEBUG on Alpha/VMS systems.
7804 @itemx -ggdb@var{level}
7805 @itemx -gstabs@var{level}
7806 @itemx -gxcoff@var{level}
7807 @itemx -gvms@var{level}
7808 Request debugging information and also use @var{level} to specify how
7809 much information. The default level is 2.
7811 Level 0 produces no debug information at all. Thus, @option{-g0} negates
7814 Level 1 produces minimal information, enough for making backtraces in
7815 parts of the program that you don't plan to debug. This includes
7816 descriptions of functions and external variables, and line number
7817 tables, but no information about local variables.
7819 Level 3 includes extra information, such as all the macro definitions
7820 present in the program. Some debuggers support macro expansion when
7821 you use @option{-g3}.
7823 If you use multiple @option{-g} options, with or without level numbers,
7824 the last such option is the one that is effective.
7826 @option{-gdwarf} does not accept a concatenated debug level, to avoid
7827 confusion with @option{-gdwarf-@var{level}}.
7828 Instead use an additional @option{-g@var{level}} option to change the
7829 debug level for DWARF.
7831 @item -fno-eliminate-unused-debug-symbols
7832 @opindex feliminate-unused-debug-symbols
7833 @opindex fno-eliminate-unused-debug-symbols
7834 By default, no debug information is produced for symbols that are not actually
7835 used. Use this option if you want debug information for all symbols.
7837 @item -femit-class-debug-always
7838 @opindex femit-class-debug-always
7839 Instead of emitting debugging information for a C++ class in only one
7840 object file, emit it in all object files using the class. This option
7841 should be used only with debuggers that are unable to handle the way GCC
7842 normally emits debugging information for classes because using this
7843 option increases the size of debugging information by as much as a
7846 @item -fno-merge-debug-strings
7847 @opindex fmerge-debug-strings
7848 @opindex fno-merge-debug-strings
7849 Direct the linker to not merge together strings in the debugging
7850 information that are identical in different object files. Merging is
7851 not supported by all assemblers or linkers. Merging decreases the size
7852 of the debug information in the output file at the cost of increasing
7853 link processing time. Merging is enabled by default.
7855 @item -fdebug-prefix-map=@var{old}=@var{new}
7856 @opindex fdebug-prefix-map
7857 When compiling files residing in directory @file{@var{old}}, record
7858 debugging information describing them as if the files resided in
7859 directory @file{@var{new}} instead. This can be used to replace a
7860 build-time path with an install-time path in the debug info. It can
7861 also be used to change an absolute path to a relative path by using
7862 @file{.} for @var{new}. This can give more reproducible builds, which
7863 are location independent, but may require an extra command to tell GDB
7864 where to find the source files. See also @option{-ffile-prefix-map}.
7866 @item -fvar-tracking
7867 @opindex fvar-tracking
7868 Run variable tracking pass. It computes where variables are stored at each
7869 position in code. Better debugging information is then generated
7870 (if the debugging information format supports this information).
7872 It is enabled by default when compiling with optimization (@option{-Os},
7873 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7874 the debug info format supports it.
7876 @item -fvar-tracking-assignments
7877 @opindex fvar-tracking-assignments
7878 @opindex fno-var-tracking-assignments
7879 Annotate assignments to user variables early in the compilation and
7880 attempt to carry the annotations over throughout the compilation all the
7881 way to the end, in an attempt to improve debug information while
7882 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7884 It can be enabled even if var-tracking is disabled, in which case
7885 annotations are created and maintained, but discarded at the end.
7886 By default, this flag is enabled together with @option{-fvar-tracking},
7887 except when selective scheduling is enabled.
7890 @opindex gsplit-dwarf
7891 Separate as much DWARF debugging information as possible into a
7892 separate output file with the extension @file{.dwo}. This option allows
7893 the build system to avoid linking files with debug information. To
7894 be useful, this option requires a debugger capable of reading @file{.dwo}
7897 @item -gdescribe-dies
7898 @opindex gdescribe-dies
7899 Add description attributes to some DWARF DIEs that have no name attribute,
7900 such as artificial variables, external references and call site
7905 Generate DWARF @code{.debug_pubnames} and @code{.debug_pubtypes} sections.
7907 @item -ggnu-pubnames
7908 @opindex ggnu-pubnames
7909 Generate @code{.debug_pubnames} and @code{.debug_pubtypes} sections in a format
7910 suitable for conversion into a GDB@ index. This option is only useful
7911 with a linker that can produce GDB@ index version 7.
7913 @item -fdebug-types-section
7914 @opindex fdebug-types-section
7915 @opindex fno-debug-types-section
7916 When using DWARF Version 4 or higher, type DIEs can be put into
7917 their own @code{.debug_types} section instead of making them part of the
7918 @code{.debug_info} section. It is more efficient to put them in a separate
7919 comdat section since the linker can then remove duplicates.
7920 But not all DWARF consumers support @code{.debug_types} sections yet
7921 and on some objects @code{.debug_types} produces larger instead of smaller
7922 debugging information.
7924 @item -grecord-gcc-switches
7925 @itemx -gno-record-gcc-switches
7926 @opindex grecord-gcc-switches
7927 @opindex gno-record-gcc-switches
7928 This switch causes the command-line options used to invoke the
7929 compiler that may affect code generation to be appended to the
7930 DW_AT_producer attribute in DWARF debugging information. The options
7931 are concatenated with spaces separating them from each other and from
7932 the compiler version.
7933 It is enabled by default.
7934 See also @option{-frecord-gcc-switches} for another
7935 way of storing compiler options into the object file.
7937 @item -gstrict-dwarf
7938 @opindex gstrict-dwarf
7939 Disallow using extensions of later DWARF standard version than selected
7940 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
7941 DWARF extensions from later standard versions is allowed.
7943 @item -gno-strict-dwarf
7944 @opindex gno-strict-dwarf
7945 Allow using extensions of later DWARF standard version than selected with
7946 @option{-gdwarf-@var{version}}.
7948 @item -gas-loc-support
7949 @opindex gas-loc-support
7950 Inform the compiler that the assembler supports @code{.loc} directives.
7951 It may then use them for the assembler to generate DWARF2+ line number
7954 This is generally desirable, because assembler-generated line-number
7955 tables are a lot more compact than those the compiler can generate
7958 This option will be enabled by default if, at GCC configure time, the
7959 assembler was found to support such directives.
7961 @item -gno-as-loc-support
7962 @opindex gno-as-loc-support
7963 Force GCC to generate DWARF2+ line number tables internally, if DWARF2+
7964 line number tables are to be generated.
7966 @item -gas-locview-support
7967 @opindex gas-locview-support
7968 Inform the compiler that the assembler supports @code{view} assignment
7969 and reset assertion checking in @code{.loc} directives.
7971 This option will be enabled by default if, at GCC configure time, the
7972 assembler was found to support them.
7974 @item -gno-as-locview-support
7975 Force GCC to assign view numbers internally, if
7976 @option{-gvariable-location-views} are explicitly requested.
7979 @itemx -gno-column-info
7980 @opindex gcolumn-info
7981 @opindex gno-column-info
7982 Emit location column information into DWARF debugging information, rather
7983 than just file and line.
7984 This option is enabled by default.
7986 @item -gstatement-frontiers
7987 @itemx -gno-statement-frontiers
7988 @opindex gstatement-frontiers
7989 @opindex gno-statement-frontiers
7990 This option causes GCC to create markers in the internal representation
7991 at the beginning of statements, and to keep them roughly in place
7992 throughout compilation, using them to guide the output of @code{is_stmt}
7993 markers in the line number table. This is enabled by default when
7994 compiling with optimization (@option{-Os}, @option{-O}, @option{-O2},
7995 @dots{}), and outputting DWARF 2 debug information at the normal level.
7997 @item -gvariable-location-views
7998 @itemx -gvariable-location-views=incompat5
7999 @itemx -gno-variable-location-views
8000 @opindex gvariable-location-views
8001 @opindex gvariable-location-views=incompat5
8002 @opindex gno-variable-location-views
8003 Augment variable location lists with progressive view numbers implied
8004 from the line number table. This enables debug information consumers to
8005 inspect state at certain points of the program, even if no instructions
8006 associated with the corresponding source locations are present at that
8007 point. If the assembler lacks support for view numbers in line number
8008 tables, this will cause the compiler to emit the line number table,
8009 which generally makes them somewhat less compact. The augmented line
8010 number tables and location lists are fully backward-compatible, so they
8011 can be consumed by debug information consumers that are not aware of
8012 these augmentations, but they won't derive any benefit from them either.
8014 This is enabled by default when outputting DWARF 2 debug information at
8015 the normal level, as long as there is assembler support,
8016 @option{-fvar-tracking-assignments} is enabled and
8017 @option{-gstrict-dwarf} is not. When assembler support is not
8018 available, this may still be enabled, but it will force GCC to output
8019 internal line number tables, and if
8020 @option{-ginternal-reset-location-views} is not enabled, that will most
8021 certainly lead to silently mismatching location views.
8023 There is a proposed representation for view numbers that is not backward
8024 compatible with the location list format introduced in DWARF 5, that can
8025 be enabled with @option{-gvariable-location-views=incompat5}. This
8026 option may be removed in the future, is only provided as a reference
8027 implementation of the proposed representation. Debug information
8028 consumers are not expected to support this extended format, and they
8029 would be rendered unable to decode location lists using it.
8031 @item -ginternal-reset-location-views
8032 @itemx -gnointernal-reset-location-views
8033 @opindex ginternal-reset-location-views
8034 @opindex gno-internal-reset-location-views
8035 Attempt to determine location views that can be omitted from location
8036 view lists. This requires the compiler to have very accurate insn
8037 length estimates, which isn't always the case, and it may cause
8038 incorrect view lists to be generated silently when using an assembler
8039 that does not support location view lists. The GNU assembler will flag
8040 any such error as a @code{view number mismatch}. This is only enabled
8041 on ports that define a reliable estimation function.
8043 @item -ginline-points
8044 @itemx -gno-inline-points
8045 @opindex ginline-points
8046 @opindex gno-inline-points
8047 Generate extended debug information for inlined functions. Location
8048 view tracking markers are inserted at inlined entry points, so that
8049 address and view numbers can be computed and output in debug
8050 information. This can be enabled independently of location views, in
8051 which case the view numbers won't be output, but it can only be enabled
8052 along with statement frontiers, and it is only enabled by default if
8053 location views are enabled.
8055 @item -gz@r{[}=@var{type}@r{]}
8057 Produce compressed debug sections in DWARF format, if that is supported.
8058 If @var{type} is not given, the default type depends on the capabilities
8059 of the assembler and linker used. @var{type} may be one of
8060 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
8061 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
8062 compression in traditional GNU format). If the linker doesn't support
8063 writing compressed debug sections, the option is rejected. Otherwise,
8064 if the assembler does not support them, @option{-gz} is silently ignored
8065 when producing object files.
8067 @item -femit-struct-debug-baseonly
8068 @opindex femit-struct-debug-baseonly
8069 Emit debug information for struct-like types
8070 only when the base name of the compilation source file
8071 matches the base name of file in which the struct is defined.
8073 This option substantially reduces the size of debugging information,
8074 but at significant potential loss in type information to the debugger.
8075 See @option{-femit-struct-debug-reduced} for a less aggressive option.
8076 See @option{-femit-struct-debug-detailed} for more detailed control.
8078 This option works only with DWARF debug output.
8080 @item -femit-struct-debug-reduced
8081 @opindex femit-struct-debug-reduced
8082 Emit debug information for struct-like types
8083 only when the base name of the compilation source file
8084 matches the base name of file in which the type is defined,
8085 unless the struct is a template or defined in a system header.
8087 This option significantly reduces the size of debugging information,
8088 with some potential loss in type information to the debugger.
8089 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
8090 See @option{-femit-struct-debug-detailed} for more detailed control.
8092 This option works only with DWARF debug output.
8094 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
8095 @opindex femit-struct-debug-detailed
8096 Specify the struct-like types
8097 for which the compiler generates debug information.
8098 The intent is to reduce duplicate struct debug information
8099 between different object files within the same program.
8101 This option is a detailed version of
8102 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
8103 which serves for most needs.
8105 A specification has the syntax@*
8106 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
8108 The optional first word limits the specification to
8109 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
8110 A struct type is used directly when it is the type of a variable, member.
8111 Indirect uses arise through pointers to structs.
8112 That is, when use of an incomplete struct is valid, the use is indirect.
8114 @samp{struct one direct; struct two * indirect;}.
8116 The optional second word limits the specification to
8117 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
8118 Generic structs are a bit complicated to explain.
8119 For C++, these are non-explicit specializations of template classes,
8120 or non-template classes within the above.
8121 Other programming languages have generics,
8122 but @option{-femit-struct-debug-detailed} does not yet implement them.
8124 The third word specifies the source files for those
8125 structs for which the compiler should emit debug information.
8126 The values @samp{none} and @samp{any} have the normal meaning.
8127 The value @samp{base} means that
8128 the base of name of the file in which the type declaration appears
8129 must match the base of the name of the main compilation file.
8130 In practice, this means that when compiling @file{foo.c}, debug information
8131 is generated for types declared in that file and @file{foo.h},
8132 but not other header files.
8133 The value @samp{sys} means those types satisfying @samp{base}
8134 or declared in system or compiler headers.
8136 You may need to experiment to determine the best settings for your application.
8138 The default is @option{-femit-struct-debug-detailed=all}.
8140 This option works only with DWARF debug output.
8142 @item -fno-dwarf2-cfi-asm
8143 @opindex fdwarf2-cfi-asm
8144 @opindex fno-dwarf2-cfi-asm
8145 Emit DWARF unwind info as compiler generated @code{.eh_frame} section
8146 instead of using GAS @code{.cfi_*} directives.
8148 @item -fno-eliminate-unused-debug-types
8149 @opindex feliminate-unused-debug-types
8150 @opindex fno-eliminate-unused-debug-types
8151 Normally, when producing DWARF output, GCC avoids producing debug symbol
8152 output for types that are nowhere used in the source file being compiled.
8153 Sometimes it is useful to have GCC emit debugging
8154 information for all types declared in a compilation
8155 unit, regardless of whether or not they are actually used
8156 in that compilation unit, for example
8157 if, in the debugger, you want to cast a value to a type that is
8158 not actually used in your program (but is declared). More often,
8159 however, this results in a significant amount of wasted space.
8162 @node Optimize Options
8163 @section Options That Control Optimization
8164 @cindex optimize options
8165 @cindex options, optimization
8167 These options control various sorts of optimizations.
8169 Without any optimization option, the compiler's goal is to reduce the
8170 cost of compilation and to make debugging produce the expected
8171 results. Statements are independent: if you stop the program with a
8172 breakpoint between statements, you can then assign a new value to any
8173 variable or change the program counter to any other statement in the
8174 function and get exactly the results you expect from the source
8177 Turning on optimization flags makes the compiler attempt to improve
8178 the performance and/or code size at the expense of compilation time
8179 and possibly the ability to debug the program.
8181 The compiler performs optimization based on the knowledge it has of the
8182 program. Compiling multiple files at once to a single output file mode allows
8183 the compiler to use information gained from all of the files when compiling
8186 Not all optimizations are controlled directly by a flag. Only
8187 optimizations that have a flag are listed in this section.
8189 Most optimizations are completely disabled at @option{-O0} or if an
8190 @option{-O} level is not set on the command line, even if individual
8191 optimization flags are specified. Similarly, @option{-Og} suppresses
8192 many optimization passes.
8194 Depending on the target and how GCC was configured, a slightly different
8195 set of optimizations may be enabled at each @option{-O} level than
8196 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
8197 to find out the exact set of optimizations that are enabled at each level.
8198 @xref{Overall Options}, for examples.
8205 Optimize. Optimizing compilation takes somewhat more time, and a lot
8206 more memory for a large function.
8208 With @option{-O}, the compiler tries to reduce code size and execution
8209 time, without performing any optimizations that take a great deal of
8212 @c Note that in addition to the default_options_table list in opts.c,
8213 @c several optimization flags default to true but control optimization
8214 @c passes that are explicitly disabled at -O0.
8216 @option{-O} turns on the following optimization flags:
8218 @c Please keep the following list alphabetized.
8219 @gccoptlist{-fauto-inc-dec @gol
8220 -fbranch-count-reg @gol
8221 -fcombine-stack-adjustments @gol
8223 -fcprop-registers @gol
8226 -fdelayed-branch @gol
8228 -fforward-propagate @gol
8229 -fguess-branch-probability @gol
8230 -fif-conversion @gol
8231 -fif-conversion2 @gol
8232 -finline-functions-called-once @gol
8234 -fipa-pure-const @gol
8235 -fipa-reference @gol
8236 -fipa-reference-addressable @gol
8237 -fmerge-constants @gol
8238 -fmove-loop-invariants @gol
8239 -fomit-frame-pointer @gol
8240 -freorder-blocks @gol
8242 -fshrink-wrap-separate @gol
8243 -fsplit-wide-types @gol
8249 -ftree-coalesce-vars @gol
8250 -ftree-copy-prop @gol
8252 -ftree-dominator-opts @gol
8254 -ftree-forwprop @gol
8258 -ftree-scev-cprop @gol
8267 Optimize even more. GCC performs nearly all supported optimizations
8268 that do not involve a space-speed tradeoff.
8269 As compared to @option{-O}, this option increases both compilation time
8270 and the performance of the generated code.
8272 @option{-O2} turns on all optimization flags specified by @option{-O}. It
8273 also turns on the following optimization flags:
8275 @c Please keep the following list alphabetized!
8276 @gccoptlist{-falign-functions -falign-jumps @gol
8277 -falign-labels -falign-loops @gol
8279 -fcode-hoisting @gol
8281 -fcse-follow-jumps -fcse-skip-blocks @gol
8282 -fdelete-null-pointer-checks @gol
8283 -fdevirtualize -fdevirtualize-speculatively @gol
8284 -fexpensive-optimizations @gol
8285 -fgcse -fgcse-lm @gol
8286 -fhoist-adjacent-loads @gol
8287 -finline-small-functions @gol
8288 -findirect-inlining @gol
8289 -fipa-bit-cp -fipa-cp -fipa-icf @gol
8290 -fipa-ra -fipa-sra -fipa-vrp @gol
8291 -fisolate-erroneous-paths-dereference @gol
8293 -foptimize-sibling-calls @gol
8294 -foptimize-strlen @gol
8295 -fpartial-inlining @gol
8297 -freorder-blocks-algorithm=stc @gol
8298 -freorder-blocks-and-partition -freorder-functions @gol
8299 -frerun-cse-after-loop @gol
8300 -fschedule-insns -fschedule-insns2 @gol
8301 -fsched-interblock -fsched-spec @gol
8302 -fstore-merging @gol
8303 -fstrict-aliasing @gol
8305 -ftree-builtin-call-dce @gol
8307 -ftree-switch-conversion -ftree-tail-merge @gol
8310 Please note the warning under @option{-fgcse} about
8311 invoking @option{-O2} on programs that use computed gotos.
8315 Optimize yet more. @option{-O3} turns on all optimizations specified
8316 by @option{-O2} and also turns on the following optimization flags:
8318 @c Please keep the following list alphabetized!
8319 @gccoptlist{-fgcse-after-reload @gol
8320 -finline-functions @gol
8322 -floop-interchange @gol
8323 -floop-unroll-and-jam @gol
8325 -fpredictive-commoning @gol
8327 -ftree-loop-distribute-patterns @gol
8328 -ftree-loop-distribution @gol
8329 -ftree-loop-vectorize @gol
8330 -ftree-partial-pre @gol
8331 -ftree-slp-vectorize @gol
8332 -funswitch-loops @gol
8333 -fvect-cost-model @gol
8334 -fversion-loops-for-strides}
8338 Reduce compilation time and make debugging produce the expected
8339 results. This is the default.
8343 Optimize for size. @option{-Os} enables all @option{-O2} optimizations
8344 except those that often increase code size:
8346 @gccoptlist{-falign-functions -falign-jumps @gol
8347 -falign-labels -falign-loops @gol
8348 -fprefetch-loop-arrays -freorder-blocks-algorithm=stc}
8350 It also enables @option{-finline-functions}, causes the compiler to tune for
8351 code size rather than execution speed, and performs further optimizations
8352 designed to reduce code size.
8356 Disregard strict standards compliance. @option{-Ofast} enables all
8357 @option{-O3} optimizations. It also enables optimizations that are not
8358 valid for all standard-compliant programs.
8359 It turns on @option{-ffast-math} and the Fortran-specific
8360 @option{-fstack-arrays}, unless @option{-fmax-stack-var-size} is
8361 specified, and @option{-fno-protect-parens}.
8365 Optimize debugging experience. @option{-Og} should be the optimization
8366 level of choice for the standard edit-compile-debug cycle, offering
8367 a reasonable level of optimization while maintaining fast compilation
8368 and a good debugging experience. It is a better choice than @option{-O0}
8369 for producing debuggable code because some compiler passes
8370 that collect debug information are disabled at @option{-O0}.
8372 Like @option{-O0}, @option{-Og} completely disables a number of
8373 optimization passes so that individual options controlling them have
8374 no effect. Otherwise @option{-Og} enables all @option{-O1}
8375 optimization flags except for those that may interfere with debugging:
8377 @gccoptlist{-fbranch-count-reg -fdelayed-branch @gol
8378 -fif-conversion -fif-conversion2 @gol
8379 -finline-functions-called-once @gol
8380 -fmove-loop-invariants -fssa-phiopt @gol
8381 -ftree-bit-ccp -ftree-pta -ftree-sra}
8385 If you use multiple @option{-O} options, with or without level numbers,
8386 the last such option is the one that is effective.
8388 Options of the form @option{-f@var{flag}} specify machine-independent
8389 flags. Most flags have both positive and negative forms; the negative
8390 form of @option{-ffoo} is @option{-fno-foo}. In the table
8391 below, only one of the forms is listed---the one you typically
8392 use. You can figure out the other form by either removing @samp{no-}
8395 The following options control specific optimizations. They are either
8396 activated by @option{-O} options or are related to ones that are. You
8397 can use the following flags in the rare cases when ``fine-tuning'' of
8398 optimizations to be performed is desired.
8401 @item -fno-defer-pop
8402 @opindex fno-defer-pop
8404 For machines that must pop arguments after a function call, always pop
8405 the arguments as soon as each function returns.
8406 At levels @option{-O1} and higher, @option{-fdefer-pop} is the default;
8407 this allows the compiler to let arguments accumulate on the stack for several
8408 function calls and pop them all at once.
8410 @item -fforward-propagate
8411 @opindex fforward-propagate
8412 Perform a forward propagation pass on RTL@. The pass tries to combine two
8413 instructions and checks if the result can be simplified. If loop unrolling
8414 is active, two passes are performed and the second is scheduled after
8417 This option is enabled by default at optimization levels @option{-O},
8418 @option{-O2}, @option{-O3}, @option{-Os}.
8420 @item -ffp-contract=@var{style}
8421 @opindex ffp-contract
8422 @option{-ffp-contract=off} disables floating-point expression contraction.
8423 @option{-ffp-contract=fast} enables floating-point expression contraction
8424 such as forming of fused multiply-add operations if the target has
8425 native support for them.
8426 @option{-ffp-contract=on} enables floating-point expression contraction
8427 if allowed by the language standard. This is currently not implemented
8428 and treated equal to @option{-ffp-contract=off}.
8430 The default is @option{-ffp-contract=fast}.
8432 @item -fomit-frame-pointer
8433 @opindex fomit-frame-pointer
8434 Omit the frame pointer in functions that don't need one. This avoids the
8435 instructions to save, set up and restore the frame pointer; on many targets
8436 it also makes an extra register available.
8438 On some targets this flag has no effect because the standard calling sequence
8439 always uses a frame pointer, so it cannot be omitted.
8441 Note that @option{-fno-omit-frame-pointer} doesn't guarantee the frame pointer
8442 is used in all functions. Several targets always omit the frame pointer in
8445 Enabled by default at @option{-O} and higher.
8447 @item -foptimize-sibling-calls
8448 @opindex foptimize-sibling-calls
8449 Optimize sibling and tail recursive calls.
8451 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8453 @item -foptimize-strlen
8454 @opindex foptimize-strlen
8455 Optimize various standard C string functions (e.g.@: @code{strlen},
8456 @code{strchr} or @code{strcpy}) and
8457 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
8459 Enabled at levels @option{-O2}, @option{-O3}.
8464 Do not expand any functions inline apart from those marked with
8465 the @code{always_inline} attribute. This is the default when not
8468 Single functions can be exempted from inlining by marking them
8469 with the @code{noinline} attribute.
8471 @item -finline-small-functions
8472 @opindex finline-small-functions
8473 Integrate functions into their callers when their body is smaller than expected
8474 function call code (so overall size of program gets smaller). The compiler
8475 heuristically decides which functions are simple enough to be worth integrating
8476 in this way. This inlining applies to all functions, even those not declared
8479 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8481 @item -findirect-inlining
8482 @opindex findirect-inlining
8483 Inline also indirect calls that are discovered to be known at compile
8484 time thanks to previous inlining. This option has any effect only
8485 when inlining itself is turned on by the @option{-finline-functions}
8486 or @option{-finline-small-functions} options.
8488 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8490 @item -finline-functions
8491 @opindex finline-functions
8492 Consider all functions for inlining, even if they are not declared inline.
8493 The compiler heuristically decides which functions are worth integrating
8496 If all calls to a given function are integrated, and the function is
8497 declared @code{static}, then the function is normally not output as
8498 assembler code in its own right.
8500 Enabled at levels @option{-O3}, @option{-Os}. Also enabled
8501 by @option{-fprofile-use} and @option{-fauto-profile}.
8503 @item -finline-functions-called-once
8504 @opindex finline-functions-called-once
8505 Consider all @code{static} functions called once for inlining into their
8506 caller even if they are not marked @code{inline}. If a call to a given
8507 function is integrated, then the function is not output as assembler code
8510 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os},
8511 but not @option{-Og}.
8513 @item -fearly-inlining
8514 @opindex fearly-inlining
8515 Inline functions marked by @code{always_inline} and functions whose body seems
8516 smaller than the function call overhead early before doing
8517 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
8518 makes profiling significantly cheaper and usually inlining faster on programs
8519 having large chains of nested wrapper functions.
8525 Perform interprocedural scalar replacement of aggregates, removal of
8526 unused parameters and replacement of parameters passed by reference
8527 by parameters passed by value.
8529 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
8531 @item -finline-limit=@var{n}
8532 @opindex finline-limit
8533 By default, GCC limits the size of functions that can be inlined. This flag
8534 allows coarse control of this limit. @var{n} is the size of functions that
8535 can be inlined in number of pseudo instructions.
8537 Inlining is actually controlled by a number of parameters, which may be
8538 specified individually by using @option{--param @var{name}=@var{value}}.
8539 The @option{-finline-limit=@var{n}} option sets some of these parameters
8543 @item max-inline-insns-single
8544 is set to @var{n}/2.
8545 @item max-inline-insns-auto
8546 is set to @var{n}/2.
8549 See below for a documentation of the individual
8550 parameters controlling inlining and for the defaults of these parameters.
8552 @emph{Note:} there may be no value to @option{-finline-limit} that results
8553 in default behavior.
8555 @emph{Note:} pseudo instruction represents, in this particular context, an
8556 abstract measurement of function's size. In no way does it represent a count
8557 of assembly instructions and as such its exact meaning might change from one
8558 release to an another.
8560 @item -fno-keep-inline-dllexport
8561 @opindex fno-keep-inline-dllexport
8562 @opindex fkeep-inline-dllexport
8563 This is a more fine-grained version of @option{-fkeep-inline-functions},
8564 which applies only to functions that are declared using the @code{dllexport}
8565 attribute or declspec. @xref{Function Attributes,,Declaring Attributes of
8568 @item -fkeep-inline-functions
8569 @opindex fkeep-inline-functions
8570 In C, emit @code{static} functions that are declared @code{inline}
8571 into the object file, even if the function has been inlined into all
8572 of its callers. This switch does not affect functions using the
8573 @code{extern inline} extension in GNU C90@. In C++, emit any and all
8574 inline functions into the object file.
8576 @item -fkeep-static-functions
8577 @opindex fkeep-static-functions
8578 Emit @code{static} functions into the object file, even if the function
8581 @item -fkeep-static-consts
8582 @opindex fkeep-static-consts
8583 Emit variables declared @code{static const} when optimization isn't turned
8584 on, even if the variables aren't referenced.
8586 GCC enables this option by default. If you want to force the compiler to
8587 check if a variable is referenced, regardless of whether or not
8588 optimization is turned on, use the @option{-fno-keep-static-consts} option.
8590 @item -fmerge-constants
8591 @opindex fmerge-constants
8592 Attempt to merge identical constants (string constants and floating-point
8593 constants) across compilation units.
8595 This option is the default for optimized compilation if the assembler and
8596 linker support it. Use @option{-fno-merge-constants} to inhibit this
8599 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8601 @item -fmerge-all-constants
8602 @opindex fmerge-all-constants
8603 Attempt to merge identical constants and identical variables.
8605 This option implies @option{-fmerge-constants}. In addition to
8606 @option{-fmerge-constants} this considers e.g.@: even constant initialized
8607 arrays or initialized constant variables with integral or floating-point
8608 types. Languages like C or C++ require each variable, including multiple
8609 instances of the same variable in recursive calls, to have distinct locations,
8610 so using this option results in non-conforming
8613 @item -fmodulo-sched
8614 @opindex fmodulo-sched
8615 Perform swing modulo scheduling immediately before the first scheduling
8616 pass. This pass looks at innermost loops and reorders their
8617 instructions by overlapping different iterations.
8619 @item -fmodulo-sched-allow-regmoves
8620 @opindex fmodulo-sched-allow-regmoves
8621 Perform more aggressive SMS-based modulo scheduling with register moves
8622 allowed. By setting this flag certain anti-dependences edges are
8623 deleted, which triggers the generation of reg-moves based on the
8624 life-range analysis. This option is effective only with
8625 @option{-fmodulo-sched} enabled.
8627 @item -fno-branch-count-reg
8628 @opindex fno-branch-count-reg
8629 @opindex fbranch-count-reg
8630 Disable the optimization pass that scans for opportunities to use
8631 ``decrement and branch'' instructions on a count register instead of
8632 instruction sequences that decrement a register, compare it against zero, and
8633 then branch based upon the result. This option is only meaningful on
8634 architectures that support such instructions, which include x86, PowerPC,
8635 IA-64 and S/390. Note that the @option{-fno-branch-count-reg} option
8636 doesn't remove the decrement and branch instructions from the generated
8637 instruction stream introduced by other optimization passes.
8639 The default is @option{-fbranch-count-reg} at @option{-O1} and higher,
8640 except for @option{-Og}.
8642 @item -fno-function-cse
8643 @opindex fno-function-cse
8644 @opindex ffunction-cse
8645 Do not put function addresses in registers; make each instruction that
8646 calls a constant function contain the function's address explicitly.
8648 This option results in less efficient code, but some strange hacks
8649 that alter the assembler output may be confused by the optimizations
8650 performed when this option is not used.
8652 The default is @option{-ffunction-cse}
8654 @item -fno-zero-initialized-in-bss
8655 @opindex fno-zero-initialized-in-bss
8656 @opindex fzero-initialized-in-bss
8657 If the target supports a BSS section, GCC by default puts variables that
8658 are initialized to zero into BSS@. This can save space in the resulting
8661 This option turns off this behavior because some programs explicitly
8662 rely on variables going to the data section---e.g., so that the
8663 resulting executable can find the beginning of that section and/or make
8664 assumptions based on that.
8666 The default is @option{-fzero-initialized-in-bss}.
8668 @item -fthread-jumps
8669 @opindex fthread-jumps
8670 Perform optimizations that check to see if a jump branches to a
8671 location where another comparison subsumed by the first is found. If
8672 so, the first branch is redirected to either the destination of the
8673 second branch or a point immediately following it, depending on whether
8674 the condition is known to be true or false.
8676 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8678 @item -fsplit-wide-types
8679 @opindex fsplit-wide-types
8680 When using a type that occupies multiple registers, such as @code{long
8681 long} on a 32-bit system, split the registers apart and allocate them
8682 independently. This normally generates better code for those types,
8683 but may make debugging more difficult.
8685 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
8688 @item -fcse-follow-jumps
8689 @opindex fcse-follow-jumps
8690 In common subexpression elimination (CSE), scan through jump instructions
8691 when the target of the jump is not reached by any other path. For
8692 example, when CSE encounters an @code{if} statement with an
8693 @code{else} clause, CSE follows the jump when the condition
8696 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8698 @item -fcse-skip-blocks
8699 @opindex fcse-skip-blocks
8700 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
8701 follow jumps that conditionally skip over blocks. When CSE
8702 encounters a simple @code{if} statement with no else clause,
8703 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
8704 body of the @code{if}.
8706 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8708 @item -frerun-cse-after-loop
8709 @opindex frerun-cse-after-loop
8710 Re-run common subexpression elimination after loop optimizations are
8713 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8717 Perform a global common subexpression elimination pass.
8718 This pass also performs global constant and copy propagation.
8720 @emph{Note:} When compiling a program using computed gotos, a GCC
8721 extension, you may get better run-time performance if you disable
8722 the global common subexpression elimination pass by adding
8723 @option{-fno-gcse} to the command line.
8725 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8729 When @option{-fgcse-lm} is enabled, global common subexpression elimination
8730 attempts to move loads that are only killed by stores into themselves. This
8731 allows a loop containing a load/store sequence to be changed to a load outside
8732 the loop, and a copy/store within the loop.
8734 Enabled by default when @option{-fgcse} is enabled.
8738 When @option{-fgcse-sm} is enabled, a store motion pass is run after
8739 global common subexpression elimination. This pass attempts to move
8740 stores out of loops. When used in conjunction with @option{-fgcse-lm},
8741 loops containing a load/store sequence can be changed to a load before
8742 the loop and a store after the loop.
8744 Not enabled at any optimization level.
8748 When @option{-fgcse-las} is enabled, the global common subexpression
8749 elimination pass eliminates redundant loads that come after stores to the
8750 same memory location (both partial and full redundancies).
8752 Not enabled at any optimization level.
8754 @item -fgcse-after-reload
8755 @opindex fgcse-after-reload
8756 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
8757 pass is performed after reload. The purpose of this pass is to clean up
8760 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
8762 @item -faggressive-loop-optimizations
8763 @opindex faggressive-loop-optimizations
8764 This option tells the loop optimizer to use language constraints to
8765 derive bounds for the number of iterations of a loop. This assumes that
8766 loop code does not invoke undefined behavior by for example causing signed
8767 integer overflows or out-of-bound array accesses. The bounds for the
8768 number of iterations of a loop are used to guide loop unrolling and peeling
8769 and loop exit test optimizations.
8770 This option is enabled by default.
8772 @item -funconstrained-commons
8773 @opindex funconstrained-commons
8774 This option tells the compiler that variables declared in common blocks
8775 (e.g.@: Fortran) may later be overridden with longer trailing arrays. This
8776 prevents certain optimizations that depend on knowing the array bounds.
8778 @item -fcrossjumping
8779 @opindex fcrossjumping
8780 Perform cross-jumping transformation.
8781 This transformation unifies equivalent code and saves code size. The
8782 resulting code may or may not perform better than without cross-jumping.
8784 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8786 @item -fauto-inc-dec
8787 @opindex fauto-inc-dec
8788 Combine increments or decrements of addresses with memory accesses.
8789 This pass is always skipped on architectures that do not have
8790 instructions to support this. Enabled by default at @option{-O} and
8791 higher on architectures that support this.
8795 Perform dead code elimination (DCE) on RTL@.
8796 Enabled by default at @option{-O} and higher.
8800 Perform dead store elimination (DSE) on RTL@.
8801 Enabled by default at @option{-O} and higher.
8803 @item -fif-conversion
8804 @opindex fif-conversion
8805 Attempt to transform conditional jumps into branch-less equivalents. This
8806 includes use of conditional moves, min, max, set flags and abs instructions, and
8807 some tricks doable by standard arithmetics. The use of conditional execution
8808 on chips where it is available is controlled by @option{-fif-conversion2}.
8810 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}, but
8811 not with @option{-Og}.
8813 @item -fif-conversion2
8814 @opindex fif-conversion2
8815 Use conditional execution (where available) to transform conditional jumps into
8816 branch-less equivalents.
8818 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}, but
8819 not with @option{-Og}.
8821 @item -fdeclone-ctor-dtor
8822 @opindex fdeclone-ctor-dtor
8823 The C++ ABI requires multiple entry points for constructors and
8824 destructors: one for a base subobject, one for a complete object, and
8825 one for a virtual destructor that calls operator delete afterwards.
8826 For a hierarchy with virtual bases, the base and complete variants are
8827 clones, which means two copies of the function. With this option, the
8828 base and complete variants are changed to be thunks that call a common
8831 Enabled by @option{-Os}.
8833 @item -fdelete-null-pointer-checks
8834 @opindex fdelete-null-pointer-checks
8835 Assume that programs cannot safely dereference null pointers, and that
8836 no code or data element resides at address zero.
8837 This option enables simple constant
8838 folding optimizations at all optimization levels. In addition, other
8839 optimization passes in GCC use this flag to control global dataflow
8840 analyses that eliminate useless checks for null pointers; these assume
8841 that a memory access to address zero always results in a trap, so
8842 that if a pointer is checked after it has already been dereferenced,
8845 Note however that in some environments this assumption is not true.
8846 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
8847 for programs that depend on that behavior.
8849 This option is enabled by default on most targets. On Nios II ELF, it
8850 defaults to off. On AVR, CR16, and MSP430, this option is completely disabled.
8852 Passes that use the dataflow information
8853 are enabled independently at different optimization levels.
8855 @item -fdevirtualize
8856 @opindex fdevirtualize
8857 Attempt to convert calls to virtual functions to direct calls. This
8858 is done both within a procedure and interprocedurally as part of
8859 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
8860 propagation (@option{-fipa-cp}).
8861 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8863 @item -fdevirtualize-speculatively
8864 @opindex fdevirtualize-speculatively
8865 Attempt to convert calls to virtual functions to speculative direct calls.
8866 Based on the analysis of the type inheritance graph, determine for a given call
8867 the set of likely targets. If the set is small, preferably of size 1, change
8868 the call into a conditional deciding between direct and indirect calls. The
8869 speculative calls enable more optimizations, such as inlining. When they seem
8870 useless after further optimization, they are converted back into original form.
8872 @item -fdevirtualize-at-ltrans
8873 @opindex fdevirtualize-at-ltrans
8874 Stream extra information needed for aggressive devirtualization when running
8875 the link-time optimizer in local transformation mode.
8876 This option enables more devirtualization but
8877 significantly increases the size of streamed data. For this reason it is
8878 disabled by default.
8880 @item -fexpensive-optimizations
8881 @opindex fexpensive-optimizations
8882 Perform a number of minor optimizations that are relatively expensive.
8884 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8888 Attempt to remove redundant extension instructions. This is especially
8889 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
8890 registers after writing to their lower 32-bit half.
8892 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
8893 @option{-O3}, @option{-Os}.
8895 @item -fno-lifetime-dse
8896 @opindex fno-lifetime-dse
8897 @opindex flifetime-dse
8898 In C++ the value of an object is only affected by changes within its
8899 lifetime: when the constructor begins, the object has an indeterminate
8900 value, and any changes during the lifetime of the object are dead when
8901 the object is destroyed. Normally dead store elimination will take
8902 advantage of this; if your code relies on the value of the object
8903 storage persisting beyond the lifetime of the object, you can use this
8904 flag to disable this optimization. To preserve stores before the
8905 constructor starts (e.g.@: because your operator new clears the object
8906 storage) but still treat the object as dead after the destructor you,
8907 can use @option{-flifetime-dse=1}. The default behavior can be
8908 explicitly selected with @option{-flifetime-dse=2}.
8909 @option{-flifetime-dse=0} is equivalent to @option{-fno-lifetime-dse}.
8911 @item -flive-range-shrinkage
8912 @opindex flive-range-shrinkage
8913 Attempt to decrease register pressure through register live range
8914 shrinkage. This is helpful for fast processors with small or moderate
8917 @item -fira-algorithm=@var{algorithm}
8918 @opindex fira-algorithm
8919 Use the specified coloring algorithm for the integrated register
8920 allocator. The @var{algorithm} argument can be @samp{priority}, which
8921 specifies Chow's priority coloring, or @samp{CB}, which specifies
8922 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
8923 for all architectures, but for those targets that do support it, it is
8924 the default because it generates better code.
8926 @item -fira-region=@var{region}
8927 @opindex fira-region
8928 Use specified regions for the integrated register allocator. The
8929 @var{region} argument should be one of the following:
8934 Use all loops as register allocation regions.
8935 This can give the best results for machines with a small and/or
8936 irregular register set.
8939 Use all loops except for loops with small register pressure
8940 as the regions. This value usually gives
8941 the best results in most cases and for most architectures,
8942 and is enabled by default when compiling with optimization for speed
8943 (@option{-O}, @option{-O2}, @dots{}).
8946 Use all functions as a single region.
8947 This typically results in the smallest code size, and is enabled by default for
8948 @option{-Os} or @option{-O0}.
8952 @item -fira-hoist-pressure
8953 @opindex fira-hoist-pressure
8954 Use IRA to evaluate register pressure in the code hoisting pass for
8955 decisions to hoist expressions. This option usually results in smaller
8956 code, but it can slow the compiler down.
8958 This option is enabled at level @option{-Os} for all targets.
8960 @item -fira-loop-pressure
8961 @opindex fira-loop-pressure
8962 Use IRA to evaluate register pressure in loops for decisions to move
8963 loop invariants. This option usually results in generation
8964 of faster and smaller code on machines with large register files (>= 32
8965 registers), but it can slow the compiler down.
8967 This option is enabled at level @option{-O3} for some targets.
8969 @item -fno-ira-share-save-slots
8970 @opindex fno-ira-share-save-slots
8971 @opindex fira-share-save-slots
8972 Disable sharing of stack slots used for saving call-used hard
8973 registers living through a call. Each hard register gets a
8974 separate stack slot, and as a result function stack frames are
8977 @item -fno-ira-share-spill-slots
8978 @opindex fno-ira-share-spill-slots
8979 @opindex fira-share-spill-slots
8980 Disable sharing of stack slots allocated for pseudo-registers. Each
8981 pseudo-register that does not get a hard register gets a separate
8982 stack slot, and as a result function stack frames are larger.
8986 Enable CFG-sensitive rematerialization in LRA. Instead of loading
8987 values of spilled pseudos, LRA tries to rematerialize (recalculate)
8988 values if it is profitable.
8990 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8992 @item -fdelayed-branch
8993 @opindex fdelayed-branch
8994 If supported for the target machine, attempt to reorder instructions
8995 to exploit instruction slots available after delayed branch
8998 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os},
8999 but not at @option{-Og}.
9001 @item -fschedule-insns
9002 @opindex fschedule-insns
9003 If supported for the target machine, attempt to reorder instructions to
9004 eliminate execution stalls due to required data being unavailable. This
9005 helps machines that have slow floating point or memory load instructions
9006 by allowing other instructions to be issued until the result of the load
9007 or floating-point instruction is required.
9009 Enabled at levels @option{-O2}, @option{-O3}.
9011 @item -fschedule-insns2
9012 @opindex fschedule-insns2
9013 Similar to @option{-fschedule-insns}, but requests an additional pass of
9014 instruction scheduling after register allocation has been done. This is
9015 especially useful on machines with a relatively small number of
9016 registers and where memory load instructions take more than one cycle.
9018 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9020 @item -fno-sched-interblock
9021 @opindex fno-sched-interblock
9022 @opindex fsched-interblock
9023 Disable instruction scheduling across basic blocks, which
9024 is normally enabled when scheduling before register allocation, i.e.@:
9025 with @option{-fschedule-insns} or at @option{-O2} or higher.
9027 @item -fno-sched-spec
9028 @opindex fno-sched-spec
9029 @opindex fsched-spec
9030 Disable speculative motion of non-load instructions, which
9031 is normally enabled when scheduling before register allocation, i.e.@:
9032 with @option{-fschedule-insns} or at @option{-O2} or higher.
9034 @item -fsched-pressure
9035 @opindex fsched-pressure
9036 Enable register pressure sensitive insn scheduling before register
9037 allocation. This only makes sense when scheduling before register
9038 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
9039 @option{-O2} or higher. Usage of this option can improve the
9040 generated code and decrease its size by preventing register pressure
9041 increase above the number of available hard registers and subsequent
9042 spills in register allocation.
9044 @item -fsched-spec-load
9045 @opindex fsched-spec-load
9046 Allow speculative motion of some load instructions. This only makes
9047 sense when scheduling before register allocation, i.e.@: with
9048 @option{-fschedule-insns} or at @option{-O2} or higher.
9050 @item -fsched-spec-load-dangerous
9051 @opindex fsched-spec-load-dangerous
9052 Allow speculative motion of more load instructions. This only makes
9053 sense when scheduling before register allocation, i.e.@: with
9054 @option{-fschedule-insns} or at @option{-O2} or higher.
9056 @item -fsched-stalled-insns
9057 @itemx -fsched-stalled-insns=@var{n}
9058 @opindex fsched-stalled-insns
9059 Define how many insns (if any) can be moved prematurely from the queue
9060 of stalled insns into the ready list during the second scheduling pass.
9061 @option{-fno-sched-stalled-insns} means that no insns are moved
9062 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
9063 on how many queued insns can be moved prematurely.
9064 @option{-fsched-stalled-insns} without a value is equivalent to
9065 @option{-fsched-stalled-insns=1}.
9067 @item -fsched-stalled-insns-dep
9068 @itemx -fsched-stalled-insns-dep=@var{n}
9069 @opindex fsched-stalled-insns-dep
9070 Define how many insn groups (cycles) are examined for a dependency
9071 on a stalled insn that is a candidate for premature removal from the queue
9072 of stalled insns. This has an effect only during the second scheduling pass,
9073 and only if @option{-fsched-stalled-insns} is used.
9074 @option{-fno-sched-stalled-insns-dep} is equivalent to
9075 @option{-fsched-stalled-insns-dep=0}.
9076 @option{-fsched-stalled-insns-dep} without a value is equivalent to
9077 @option{-fsched-stalled-insns-dep=1}.
9079 @item -fsched2-use-superblocks
9080 @opindex fsched2-use-superblocks
9081 When scheduling after register allocation, use superblock scheduling.
9082 This allows motion across basic block boundaries,
9083 resulting in faster schedules. This option is experimental, as not all machine
9084 descriptions used by GCC model the CPU closely enough to avoid unreliable
9085 results from the algorithm.
9087 This only makes sense when scheduling after register allocation, i.e.@: with
9088 @option{-fschedule-insns2} or at @option{-O2} or higher.
9090 @item -fsched-group-heuristic
9091 @opindex fsched-group-heuristic
9092 Enable the group heuristic in the scheduler. This heuristic favors
9093 the instruction that belongs to a schedule group. This is enabled
9094 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
9095 or @option{-fschedule-insns2} or at @option{-O2} or higher.
9097 @item -fsched-critical-path-heuristic
9098 @opindex fsched-critical-path-heuristic
9099 Enable the critical-path heuristic in the scheduler. This heuristic favors
9100 instructions on the critical path. This is enabled by default when
9101 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
9102 or @option{-fschedule-insns2} or at @option{-O2} or higher.
9104 @item -fsched-spec-insn-heuristic
9105 @opindex fsched-spec-insn-heuristic
9106 Enable the speculative instruction heuristic in the scheduler. This
9107 heuristic favors speculative instructions with greater dependency weakness.
9108 This is enabled by default when scheduling is enabled, i.e.@:
9109 with @option{-fschedule-insns} or @option{-fschedule-insns2}
9110 or at @option{-O2} or higher.
9112 @item -fsched-rank-heuristic
9113 @opindex fsched-rank-heuristic
9114 Enable the rank heuristic in the scheduler. This heuristic favors
9115 the instruction belonging to a basic block with greater size or frequency.
9116 This is enabled by default when scheduling is enabled, i.e.@:
9117 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
9118 at @option{-O2} or higher.
9120 @item -fsched-last-insn-heuristic
9121 @opindex fsched-last-insn-heuristic
9122 Enable the last-instruction heuristic in the scheduler. This heuristic
9123 favors the instruction that is less dependent on the last instruction
9124 scheduled. This is enabled by default when scheduling is enabled,
9125 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
9126 at @option{-O2} or higher.
9128 @item -fsched-dep-count-heuristic
9129 @opindex fsched-dep-count-heuristic
9130 Enable the dependent-count heuristic in the scheduler. This heuristic
9131 favors the instruction that has more instructions depending on it.
9132 This is enabled by default when scheduling is enabled, i.e.@:
9133 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
9134 at @option{-O2} or higher.
9136 @item -freschedule-modulo-scheduled-loops
9137 @opindex freschedule-modulo-scheduled-loops
9138 Modulo scheduling is performed before traditional scheduling. If a loop
9139 is modulo scheduled, later scheduling passes may change its schedule.
9140 Use this option to control that behavior.
9142 @item -fselective-scheduling
9143 @opindex fselective-scheduling
9144 Schedule instructions using selective scheduling algorithm. Selective
9145 scheduling runs instead of the first scheduler pass.
9147 @item -fselective-scheduling2
9148 @opindex fselective-scheduling2
9149 Schedule instructions using selective scheduling algorithm. Selective
9150 scheduling runs instead of the second scheduler pass.
9152 @item -fsel-sched-pipelining
9153 @opindex fsel-sched-pipelining
9154 Enable software pipelining of innermost loops during selective scheduling.
9155 This option has no effect unless one of @option{-fselective-scheduling} or
9156 @option{-fselective-scheduling2} is turned on.
9158 @item -fsel-sched-pipelining-outer-loops
9159 @opindex fsel-sched-pipelining-outer-loops
9160 When pipelining loops during selective scheduling, also pipeline outer loops.
9161 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
9163 @item -fsemantic-interposition
9164 @opindex fsemantic-interposition
9165 Some object formats, like ELF, allow interposing of symbols by the
9167 This means that for symbols exported from the DSO, the compiler cannot perform
9168 interprocedural propagation, inlining and other optimizations in anticipation
9169 that the function or variable in question may change. While this feature is
9170 useful, for example, to rewrite memory allocation functions by a debugging
9171 implementation, it is expensive in the terms of code quality.
9172 With @option{-fno-semantic-interposition} the compiler assumes that
9173 if interposition happens for functions the overwriting function will have
9174 precisely the same semantics (and side effects).
9175 Similarly if interposition happens
9176 for variables, the constructor of the variable will be the same. The flag
9177 has no effect for functions explicitly declared inline
9178 (where it is never allowed for interposition to change semantics)
9179 and for symbols explicitly declared weak.
9182 @opindex fshrink-wrap
9183 Emit function prologues only before parts of the function that need it,
9184 rather than at the top of the function. This flag is enabled by default at
9185 @option{-O} and higher.
9187 @item -fshrink-wrap-separate
9188 @opindex fshrink-wrap-separate
9189 Shrink-wrap separate parts of the prologue and epilogue separately, so that
9190 those parts are only executed when needed.
9191 This option is on by default, but has no effect unless @option{-fshrink-wrap}
9192 is also turned on and the target supports this.
9194 @item -fcaller-saves
9195 @opindex fcaller-saves
9196 Enable allocation of values to registers that are clobbered by
9197 function calls, by emitting extra instructions to save and restore the
9198 registers around such calls. Such allocation is done only when it
9199 seems to result in better code.
9201 This option is always enabled by default on certain machines, usually
9202 those which have no call-preserved registers to use instead.
9204 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9206 @item -fcombine-stack-adjustments
9207 @opindex fcombine-stack-adjustments
9208 Tracks stack adjustments (pushes and pops) and stack memory references
9209 and then tries to find ways to combine them.
9211 Enabled by default at @option{-O1} and higher.
9215 Use caller save registers for allocation if those registers are not used by
9216 any called function. In that case it is not necessary to save and restore
9217 them around calls. This is only possible if called functions are part of
9218 same compilation unit as current function and they are compiled before it.
9220 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}, however the option
9221 is disabled if generated code will be instrumented for profiling
9222 (@option{-p}, or @option{-pg}) or if callee's register usage cannot be known
9223 exactly (this happens on targets that do not expose prologues
9224 and epilogues in RTL).
9226 @item -fconserve-stack
9227 @opindex fconserve-stack
9228 Attempt to minimize stack usage. The compiler attempts to use less
9229 stack space, even if that makes the program slower. This option
9230 implies setting the @option{large-stack-frame} parameter to 100
9231 and the @option{large-stack-frame-growth} parameter to 400.
9233 @item -ftree-reassoc
9234 @opindex ftree-reassoc
9235 Perform reassociation on trees. This flag is enabled by default
9236 at @option{-O} and higher.
9238 @item -fcode-hoisting
9239 @opindex fcode-hoisting
9240 Perform code hoisting. Code hoisting tries to move the
9241 evaluation of expressions executed on all paths to the function exit
9242 as early as possible. This is especially useful as a code size
9243 optimization, but it often helps for code speed as well.
9244 This flag is enabled by default at @option{-O2} and higher.
9248 Perform partial redundancy elimination (PRE) on trees. This flag is
9249 enabled by default at @option{-O2} and @option{-O3}.
9251 @item -ftree-partial-pre
9252 @opindex ftree-partial-pre
9253 Make partial redundancy elimination (PRE) more aggressive. This flag is
9254 enabled by default at @option{-O3}.
9256 @item -ftree-forwprop
9257 @opindex ftree-forwprop
9258 Perform forward propagation on trees. This flag is enabled by default
9259 at @option{-O} and higher.
9263 Perform full redundancy elimination (FRE) on trees. The difference
9264 between FRE and PRE is that FRE only considers expressions
9265 that are computed on all paths leading to the redundant computation.
9266 This analysis is faster than PRE, though it exposes fewer redundancies.
9267 This flag is enabled by default at @option{-O} and higher.
9269 @item -ftree-phiprop
9270 @opindex ftree-phiprop
9271 Perform hoisting of loads from conditional pointers on trees. This
9272 pass is enabled by default at @option{-O} and higher.
9274 @item -fhoist-adjacent-loads
9275 @opindex fhoist-adjacent-loads
9276 Speculatively hoist loads from both branches of an if-then-else if the
9277 loads are from adjacent locations in the same structure and the target
9278 architecture has a conditional move instruction. This flag is enabled
9279 by default at @option{-O2} and higher.
9281 @item -ftree-copy-prop
9282 @opindex ftree-copy-prop
9283 Perform copy propagation on trees. This pass eliminates unnecessary
9284 copy operations. This flag is enabled by default at @option{-O} and
9287 @item -fipa-pure-const
9288 @opindex fipa-pure-const
9289 Discover which functions are pure or constant.
9290 Enabled by default at @option{-O} and higher.
9292 @item -fipa-reference
9293 @opindex fipa-reference
9294 Discover which static variables do not escape the
9296 Enabled by default at @option{-O} and higher.
9298 @item -fipa-reference-addressable
9299 @opindex fipa-reference-addressable
9300 Discover read-only, write-only and non-addressable static variables.
9301 Enabled by default at @option{-O} and higher.
9303 @item -fipa-stack-alignment
9304 @opindex fipa-stack-alignment
9305 Reduce stack alignment on call sites if possible.
9310 Perform interprocedural pointer analysis and interprocedural modification
9311 and reference analysis. This option can cause excessive memory and
9312 compile-time usage on large compilation units. It is not enabled by
9313 default at any optimization level.
9316 @opindex fipa-profile
9317 Perform interprocedural profile propagation. The functions called only from
9318 cold functions are marked as cold. Also functions executed once (such as
9319 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
9320 functions and loop less parts of functions executed once are then optimized for
9322 Enabled by default at @option{-O} and higher.
9326 Perform interprocedural constant propagation.
9327 This optimization analyzes the program to determine when values passed
9328 to functions are constants and then optimizes accordingly.
9329 This optimization can substantially increase performance
9330 if the application has constants passed to functions.
9331 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
9332 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9334 @item -fipa-cp-clone
9335 @opindex fipa-cp-clone
9336 Perform function cloning to make interprocedural constant propagation stronger.
9337 When enabled, interprocedural constant propagation performs function cloning
9338 when externally visible function can be called with constant arguments.
9339 Because this optimization can create multiple copies of functions,
9340 it may significantly increase code size
9341 (see @option{--param ipcp-unit-growth=@var{value}}).
9342 This flag is enabled by default at @option{-O3}.
9343 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9346 @opindex fipa-bit-cp
9347 When enabled, perform interprocedural bitwise constant
9348 propagation. This flag is enabled by default at @option{-O2} and
9349 by @option{-fprofile-use} and @option{-fauto-profile}.
9350 It requires that @option{-fipa-cp} is enabled.
9354 When enabled, perform interprocedural propagation of value
9355 ranges. This flag is enabled by default at @option{-O2}. It requires
9356 that @option{-fipa-cp} is enabled.
9360 Perform Identical Code Folding for functions and read-only variables.
9361 The optimization reduces code size and may disturb unwind stacks by replacing
9362 a function by equivalent one with a different name. The optimization works
9363 more effectively with link-time optimization enabled.
9365 Although the behavior is similar to the Gold Linker's ICF optimization, GCC ICF
9366 works on different levels and thus the optimizations are not same - there are
9367 equivalences that are found only by GCC and equivalences found only by Gold.
9369 This flag is enabled by default at @option{-O2} and @option{-Os}.
9371 @item -flive-patching=@var{level}
9372 @opindex flive-patching
9373 Control GCC's optimizations to produce output suitable for live-patching.
9375 If the compiler's optimization uses a function's body or information extracted
9376 from its body to optimize/change another function, the latter is called an
9377 impacted function of the former. If a function is patched, its impacted
9378 functions should be patched too.
9380 The impacted functions are determined by the compiler's interprocedural
9381 optimizations. For example, a caller is impacted when inlining a function
9383 cloning a function and changing its caller to call this new clone,
9384 or extracting a function's pureness/constness information to optimize
9385 its direct or indirect callers, etc.
9387 Usually, the more IPA optimizations enabled, the larger the number of
9388 impacted functions for each function. In order to control the number of
9389 impacted functions and more easily compute the list of impacted function,
9390 IPA optimizations can be partially enabled at two different levels.
9392 The @var{level} argument should be one of the following:
9398 Only enable inlining and cloning optimizations, which includes inlining,
9399 cloning, interprocedural scalar replacement of aggregates and partial inlining.
9400 As a result, when patching a function, all its callers and its clones'
9401 callers are impacted, therefore need to be patched as well.
9403 @option{-flive-patching=inline-clone} disables the following optimization flags:
9404 @gccoptlist{-fwhole-program -fipa-pta -fipa-reference -fipa-ra @gol
9405 -fipa-icf -fipa-icf-functions -fipa-icf-variables @gol
9406 -fipa-bit-cp -fipa-vrp -fipa-pure-const -fipa-reference-addressable @gol
9407 -fipa-stack-alignment}
9409 @item inline-only-static
9411 Only enable inlining of static functions.
9412 As a result, when patching a static function, all its callers are impacted
9413 and so need to be patched as well.
9415 In addition to all the flags that @option{-flive-patching=inline-clone}
9417 @option{-flive-patching=inline-only-static} disables the following additional
9419 @gccoptlist{-fipa-cp-clone -fipa-sra -fpartial-inlining -fipa-cp}
9423 When @option{-flive-patching} is specified without any value, the default value
9424 is @var{inline-clone}.
9426 This flag is disabled by default.
9428 Note that @option{-flive-patching} is not supported with link-time optimization
9431 @item -fisolate-erroneous-paths-dereference
9432 @opindex fisolate-erroneous-paths-dereference
9433 Detect paths that trigger erroneous or undefined behavior due to
9434 dereferencing a null pointer. Isolate those paths from the main control
9435 flow and turn the statement with erroneous or undefined behavior into a trap.
9436 This flag is enabled by default at @option{-O2} and higher and depends on
9437 @option{-fdelete-null-pointer-checks} also being enabled.
9439 @item -fisolate-erroneous-paths-attribute
9440 @opindex fisolate-erroneous-paths-attribute
9441 Detect paths that trigger erroneous or undefined behavior due to a null value
9442 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
9443 attribute. Isolate those paths from the main control flow and turn the
9444 statement with erroneous or undefined behavior into a trap. This is not
9445 currently enabled, but may be enabled by @option{-O2} in the future.
9449 Perform forward store motion on trees. This flag is
9450 enabled by default at @option{-O} and higher.
9452 @item -ftree-bit-ccp
9453 @opindex ftree-bit-ccp
9454 Perform sparse conditional bit constant propagation on trees and propagate
9455 pointer alignment information.
9456 This pass only operates on local scalar variables and is enabled by default
9457 at @option{-O1} and higher, except for @option{-Og}.
9458 It requires that @option{-ftree-ccp} is enabled.
9462 Perform sparse conditional constant propagation (CCP) on trees. This
9463 pass only operates on local scalar variables and is enabled by default
9464 at @option{-O} and higher.
9466 @item -fssa-backprop
9467 @opindex fssa-backprop
9468 Propagate information about uses of a value up the definition chain
9469 in order to simplify the definitions. For example, this pass strips
9470 sign operations if the sign of a value never matters. The flag is
9471 enabled by default at @option{-O} and higher.
9474 @opindex fssa-phiopt
9475 Perform pattern matching on SSA PHI nodes to optimize conditional
9476 code. This pass is enabled by default at @option{-O1} and higher,
9477 except for @option{-Og}.
9479 @item -ftree-switch-conversion
9480 @opindex ftree-switch-conversion
9481 Perform conversion of simple initializations in a switch to
9482 initializations from a scalar array. This flag is enabled by default
9483 at @option{-O2} and higher.
9485 @item -ftree-tail-merge
9486 @opindex ftree-tail-merge
9487 Look for identical code sequences. When found, replace one with a jump to the
9488 other. This optimization is known as tail merging or cross jumping. This flag
9489 is enabled by default at @option{-O2} and higher. The compilation time
9491 be limited using @option{max-tail-merge-comparisons} parameter and
9492 @option{max-tail-merge-iterations} parameter.
9496 Perform dead code elimination (DCE) on trees. This flag is enabled by
9497 default at @option{-O} and higher.
9499 @item -ftree-builtin-call-dce
9500 @opindex ftree-builtin-call-dce
9501 Perform conditional dead code elimination (DCE) for calls to built-in functions
9502 that may set @code{errno} but are otherwise free of side effects. This flag is
9503 enabled by default at @option{-O2} and higher if @option{-Os} is not also
9506 @item -ftree-dominator-opts
9507 @opindex ftree-dominator-opts
9508 Perform a variety of simple scalar cleanups (constant/copy
9509 propagation, redundancy elimination, range propagation and expression
9510 simplification) based on a dominator tree traversal. This also
9511 performs jump threading (to reduce jumps to jumps). This flag is
9512 enabled by default at @option{-O} and higher.
9516 Perform dead store elimination (DSE) on trees. A dead store is a store into
9517 a memory location that is later overwritten by another store without
9518 any intervening loads. In this case the earlier store can be deleted. This
9519 flag is enabled by default at @option{-O} and higher.
9523 Perform loop header copying on trees. This is beneficial since it increases
9524 effectiveness of code motion optimizations. It also saves one jump. This flag
9525 is enabled by default at @option{-O} and higher. It is not enabled
9526 for @option{-Os}, since it usually increases code size.
9528 @item -ftree-loop-optimize
9529 @opindex ftree-loop-optimize
9530 Perform loop optimizations on trees. This flag is enabled by default
9531 at @option{-O} and higher.
9533 @item -ftree-loop-linear
9534 @itemx -floop-strip-mine
9536 @opindex ftree-loop-linear
9537 @opindex floop-strip-mine
9538 @opindex floop-block
9539 Perform loop nest optimizations. Same as
9540 @option{-floop-nest-optimize}. To use this code transformation, GCC has
9541 to be configured with @option{--with-isl} to enable the Graphite loop
9542 transformation infrastructure.
9544 @item -fgraphite-identity
9545 @opindex fgraphite-identity
9546 Enable the identity transformation for graphite. For every SCoP we generate
9547 the polyhedral representation and transform it back to gimple. Using
9548 @option{-fgraphite-identity} we can check the costs or benefits of the
9549 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
9550 are also performed by the code generator isl, like index splitting and
9551 dead code elimination in loops.
9553 @item -floop-nest-optimize
9554 @opindex floop-nest-optimize
9555 Enable the isl based loop nest optimizer. This is a generic loop nest
9556 optimizer based on the Pluto optimization algorithms. It calculates a loop
9557 structure optimized for data-locality and parallelism. This option
9560 @item -floop-parallelize-all
9561 @opindex floop-parallelize-all
9562 Use the Graphite data dependence analysis to identify loops that can
9563 be parallelized. Parallelize all the loops that can be analyzed to
9564 not contain loop carried dependences without checking that it is
9565 profitable to parallelize the loops.
9567 @item -ftree-coalesce-vars
9568 @opindex ftree-coalesce-vars
9569 While transforming the program out of the SSA representation, attempt to
9570 reduce copying by coalescing versions of different user-defined
9571 variables, instead of just compiler temporaries. This may severely
9572 limit the ability to debug an optimized program compiled with
9573 @option{-fno-var-tracking-assignments}. In the negated form, this flag
9574 prevents SSA coalescing of user variables. This option is enabled by
9575 default if optimization is enabled, and it does very little otherwise.
9577 @item -ftree-loop-if-convert
9578 @opindex ftree-loop-if-convert
9579 Attempt to transform conditional jumps in the innermost loops to
9580 branch-less equivalents. The intent is to remove control-flow from
9581 the innermost loops in order to improve the ability of the
9582 vectorization pass to handle these loops. This is enabled by default
9583 if vectorization is enabled.
9585 @item -ftree-loop-distribution
9586 @opindex ftree-loop-distribution
9587 Perform loop distribution. This flag can improve cache performance on
9588 big loop bodies and allow further loop optimizations, like
9589 parallelization or vectorization, to take place. For example, the loop
9605 This flag is enabled by default at @option{-O3}.
9606 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9608 @item -ftree-loop-distribute-patterns
9609 @opindex ftree-loop-distribute-patterns
9610 Perform loop distribution of patterns that can be code generated with
9611 calls to a library. This flag is enabled by default at @option{-O3}, and
9612 by @option{-fprofile-use} and @option{-fauto-profile}.
9614 This pass distributes the initialization loops and generates a call to
9615 memset zero. For example, the loop
9631 and the initialization loop is transformed into a call to memset zero.
9632 This flag is enabled by default at @option{-O3}.
9633 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9635 @item -floop-interchange
9636 @opindex floop-interchange
9637 Perform loop interchange outside of graphite. This flag can improve cache
9638 performance on loop nest and allow further loop optimizations, like
9639 vectorization, to take place. For example, the loop
9641 for (int i = 0; i < N; i++)
9642 for (int j = 0; j < N; j++)
9643 for (int k = 0; k < N; k++)
9644 c[i][j] = c[i][j] + a[i][k]*b[k][j];
9648 for (int i = 0; i < N; i++)
9649 for (int k = 0; k < N; k++)
9650 for (int j = 0; j < N; j++)
9651 c[i][j] = c[i][j] + a[i][k]*b[k][j];
9653 This flag is enabled by default at @option{-O3}.
9654 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9656 @item -floop-unroll-and-jam
9657 @opindex floop-unroll-and-jam
9658 Apply unroll and jam transformations on feasible loops. In a loop
9659 nest this unrolls the outer loop by some factor and fuses the resulting
9660 multiple inner loops. This flag is enabled by default at @option{-O3}.
9661 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9663 @item -ftree-loop-im
9664 @opindex ftree-loop-im
9665 Perform loop invariant motion on trees. This pass moves only invariants that
9666 are hard to handle at RTL level (function calls, operations that expand to
9667 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
9668 operands of conditions that are invariant out of the loop, so that we can use
9669 just trivial invariantness analysis in loop unswitching. The pass also includes
9672 @item -ftree-loop-ivcanon
9673 @opindex ftree-loop-ivcanon
9674 Create a canonical counter for number of iterations in loops for which
9675 determining number of iterations requires complicated analysis. Later
9676 optimizations then may determine the number easily. Useful especially
9677 in connection with unrolling.
9679 @item -ftree-scev-cprop
9680 @opindex ftree-scev-cprop
9681 Perform final value replacement. If a variable is modified in a loop
9682 in such a way that its value when exiting the loop can be determined using
9683 only its initial value and the number of loop iterations, replace uses of
9684 the final value by such a computation, provided it is sufficiently cheap.
9685 This reduces data dependencies and may allow further simplifications.
9686 Enabled by default at @option{-O} and higher.
9690 Perform induction variable optimizations (strength reduction, induction
9691 variable merging and induction variable elimination) on trees.
9693 @item -ftree-parallelize-loops=n
9694 @opindex ftree-parallelize-loops
9695 Parallelize loops, i.e., split their iteration space to run in n threads.
9696 This is only possible for loops whose iterations are independent
9697 and can be arbitrarily reordered. The optimization is only
9698 profitable on multiprocessor machines, for loops that are CPU-intensive,
9699 rather than constrained e.g.@: by memory bandwidth. This option
9700 implies @option{-pthread}, and thus is only supported on targets
9701 that have support for @option{-pthread}.
9705 Perform function-local points-to analysis on trees. This flag is
9706 enabled by default at @option{-O1} and higher, except for @option{-Og}.
9710 Perform scalar replacement of aggregates. This pass replaces structure
9711 references with scalars to prevent committing structures to memory too
9712 early. This flag is enabled by default at @option{-O1} and higher,
9713 except for @option{-Og}.
9715 @item -fstore-merging
9716 @opindex fstore-merging
9717 Perform merging of narrow stores to consecutive memory addresses. This pass
9718 merges contiguous stores of immediate values narrower than a word into fewer
9719 wider stores to reduce the number of instructions. This is enabled by default
9720 at @option{-O2} and higher as well as @option{-Os}.
9724 Perform temporary expression replacement during the SSA->normal phase. Single
9725 use/single def temporaries are replaced at their use location with their
9726 defining expression. This results in non-GIMPLE code, but gives the expanders
9727 much more complex trees to work on resulting in better RTL generation. This is
9728 enabled by default at @option{-O} and higher.
9732 Perform straight-line strength reduction on trees. This recognizes related
9733 expressions involving multiplications and replaces them by less expensive
9734 calculations when possible. This is enabled by default at @option{-O} and
9737 @item -ftree-vectorize
9738 @opindex ftree-vectorize
9739 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
9740 and @option{-ftree-slp-vectorize} if not explicitly specified.
9742 @item -ftree-loop-vectorize
9743 @opindex ftree-loop-vectorize
9744 Perform loop vectorization on trees. This flag is enabled by default at
9745 @option{-O3} and by @option{-ftree-vectorize}, @option{-fprofile-use},
9746 and @option{-fauto-profile}.
9748 @item -ftree-slp-vectorize
9749 @opindex ftree-slp-vectorize
9750 Perform basic block vectorization on trees. This flag is enabled by default at
9751 @option{-O3} and by @option{-ftree-vectorize}, @option{-fprofile-use},
9752 and @option{-fauto-profile}.
9754 @item -fvect-cost-model=@var{model}
9755 @opindex fvect-cost-model
9756 Alter the cost model used for vectorization. The @var{model} argument
9757 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
9758 With the @samp{unlimited} model the vectorized code-path is assumed
9759 to be profitable while with the @samp{dynamic} model a runtime check
9760 guards the vectorized code-path to enable it only for iteration
9761 counts that will likely execute faster than when executing the original
9762 scalar loop. The @samp{cheap} model disables vectorization of
9763 loops where doing so would be cost prohibitive for example due to
9764 required runtime checks for data dependence or alignment but otherwise
9765 is equal to the @samp{dynamic} model.
9766 The default cost model depends on other optimization flags and is
9767 either @samp{dynamic} or @samp{cheap}.
9769 @item -fsimd-cost-model=@var{model}
9770 @opindex fsimd-cost-model
9771 Alter the cost model used for vectorization of loops marked with the OpenMP
9772 simd directive. The @var{model} argument should be one of
9773 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
9774 have the same meaning as described in @option{-fvect-cost-model} and by
9775 default a cost model defined with @option{-fvect-cost-model} is used.
9779 Perform Value Range Propagation on trees. This is similar to the
9780 constant propagation pass, but instead of values, ranges of values are
9781 propagated. This allows the optimizers to remove unnecessary range
9782 checks like array bound checks and null pointer checks. This is
9783 enabled by default at @option{-O2} and higher. Null pointer check
9784 elimination is only done if @option{-fdelete-null-pointer-checks} is
9788 @opindex fsplit-paths
9789 Split paths leading to loop backedges. This can improve dead code
9790 elimination and common subexpression elimination. This is enabled by
9791 default at @option{-O3} and above.
9793 @item -fsplit-ivs-in-unroller
9794 @opindex fsplit-ivs-in-unroller
9795 Enables expression of values of induction variables in later iterations
9796 of the unrolled loop using the value in the first iteration. This breaks
9797 long dependency chains, thus improving efficiency of the scheduling passes.
9799 A combination of @option{-fweb} and CSE is often sufficient to obtain the
9800 same effect. However, that is not reliable in cases where the loop body
9801 is more complicated than a single basic block. It also does not work at all
9802 on some architectures due to restrictions in the CSE pass.
9804 This optimization is enabled by default.
9806 @item -fvariable-expansion-in-unroller
9807 @opindex fvariable-expansion-in-unroller
9808 With this option, the compiler creates multiple copies of some
9809 local variables when unrolling a loop, which can result in superior code.
9811 @item -fpartial-inlining
9812 @opindex fpartial-inlining
9813 Inline parts of functions. This option has any effect only
9814 when inlining itself is turned on by the @option{-finline-functions}
9815 or @option{-finline-small-functions} options.
9817 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9819 @item -fpredictive-commoning
9820 @opindex fpredictive-commoning
9821 Perform predictive commoning optimization, i.e., reusing computations
9822 (especially memory loads and stores) performed in previous
9823 iterations of loops.
9825 This option is enabled at level @option{-O3}.
9826 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9828 @item -fprefetch-loop-arrays
9829 @opindex fprefetch-loop-arrays
9830 If supported by the target machine, generate instructions to prefetch
9831 memory to improve the performance of loops that access large arrays.
9833 This option may generate better or worse code; results are highly
9834 dependent on the structure of loops within the source code.
9836 Disabled at level @option{-Os}.
9838 @item -fno-printf-return-value
9839 @opindex fno-printf-return-value
9840 @opindex fprintf-return-value
9841 Do not substitute constants for known return value of formatted output
9842 functions such as @code{sprintf}, @code{snprintf}, @code{vsprintf}, and
9843 @code{vsnprintf} (but not @code{printf} of @code{fprintf}). This
9844 transformation allows GCC to optimize or even eliminate branches based
9845 on the known return value of these functions called with arguments that
9846 are either constant, or whose values are known to be in a range that
9847 makes determining the exact return value possible. For example, when
9848 @option{-fprintf-return-value} is in effect, both the branch and the
9849 body of the @code{if} statement (but not the call to @code{snprint})
9850 can be optimized away when @code{i} is a 32-bit or smaller integer
9851 because the return value is guaranteed to be at most 8.
9855 if (snprintf (buf, "%08x", i) >= sizeof buf)
9859 The @option{-fprintf-return-value} option relies on other optimizations
9860 and yields best results with @option{-O2} and above. It works in tandem
9861 with the @option{-Wformat-overflow} and @option{-Wformat-truncation}
9862 options. The @option{-fprintf-return-value} option is enabled by default.
9865 @itemx -fno-peephole2
9866 @opindex fno-peephole
9868 @opindex fno-peephole2
9870 Disable any machine-specific peephole optimizations. The difference
9871 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
9872 are implemented in the compiler; some targets use one, some use the
9873 other, a few use both.
9875 @option{-fpeephole} is enabled by default.
9876 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9878 @item -fno-guess-branch-probability
9879 @opindex fno-guess-branch-probability
9880 @opindex fguess-branch-probability
9881 Do not guess branch probabilities using heuristics.
9883 GCC uses heuristics to guess branch probabilities if they are
9884 not provided by profiling feedback (@option{-fprofile-arcs}). These
9885 heuristics are based on the control flow graph. If some branch probabilities
9886 are specified by @code{__builtin_expect}, then the heuristics are
9887 used to guess branch probabilities for the rest of the control flow graph,
9888 taking the @code{__builtin_expect} info into account. The interactions
9889 between the heuristics and @code{__builtin_expect} can be complex, and in
9890 some cases, it may be useful to disable the heuristics so that the effects
9891 of @code{__builtin_expect} are easier to understand.
9893 It is also possible to specify expected probability of the expression
9894 with @code{__builtin_expect_with_probability} built-in function.
9896 The default is @option{-fguess-branch-probability} at levels
9897 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9899 @item -freorder-blocks
9900 @opindex freorder-blocks
9901 Reorder basic blocks in the compiled function in order to reduce number of
9902 taken branches and improve code locality.
9904 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9906 @item -freorder-blocks-algorithm=@var{algorithm}
9907 @opindex freorder-blocks-algorithm
9908 Use the specified algorithm for basic block reordering. The
9909 @var{algorithm} argument can be @samp{simple}, which does not increase
9910 code size (except sometimes due to secondary effects like alignment),
9911 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
9912 put all often executed code together, minimizing the number of branches
9913 executed by making extra copies of code.
9915 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
9916 @samp{stc} at levels @option{-O2}, @option{-O3}.
9918 @item -freorder-blocks-and-partition
9919 @opindex freorder-blocks-and-partition
9920 In addition to reordering basic blocks in the compiled function, in order
9921 to reduce number of taken branches, partitions hot and cold basic blocks
9922 into separate sections of the assembly and @file{.o} files, to improve
9923 paging and cache locality performance.
9925 This optimization is automatically turned off in the presence of
9926 exception handling or unwind tables (on targets using setjump/longjump or target specific scheme), for linkonce sections, for functions with a user-defined
9927 section attribute and on any architecture that does not support named
9928 sections. When @option{-fsplit-stack} is used this option is not
9929 enabled by default (to avoid linker errors), but may be enabled
9930 explicitly (if using a working linker).
9932 Enabled for x86 at levels @option{-O2}, @option{-O3}, @option{-Os}.
9934 @item -freorder-functions
9935 @opindex freorder-functions
9936 Reorder functions in the object file in order to
9937 improve code locality. This is implemented by using special
9938 subsections @code{.text.hot} for most frequently executed functions and
9939 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
9940 the linker so object file format must support named sections and linker must
9941 place them in a reasonable way.
9943 This option isn't effective unless you either provide profile feedback
9944 (see @option{-fprofile-arcs} for details) or manually annotate functions with
9945 @code{hot} or @code{cold} attributes (@pxref{Common Function Attributes}).
9947 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9949 @item -fstrict-aliasing
9950 @opindex fstrict-aliasing
9951 Allow the compiler to assume the strictest aliasing rules applicable to
9952 the language being compiled. For C (and C++), this activates
9953 optimizations based on the type of expressions. In particular, an
9954 object of one type is assumed never to reside at the same address as an
9955 object of a different type, unless the types are almost the same. For
9956 example, an @code{unsigned int} can alias an @code{int}, but not a
9957 @code{void*} or a @code{double}. A character type may alias any other
9960 @anchor{Type-punning}Pay special attention to code like this:
9973 The practice of reading from a different union member than the one most
9974 recently written to (called ``type-punning'') is common. Even with
9975 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
9976 is accessed through the union type. So, the code above works as
9977 expected. @xref{Structures unions enumerations and bit-fields
9978 implementation}. However, this code might not:
9989 Similarly, access by taking the address, casting the resulting pointer
9990 and dereferencing the result has undefined behavior, even if the cast
9991 uses a union type, e.g.:
9995 return ((union a_union *) &d)->i;
9999 The @option{-fstrict-aliasing} option is enabled at levels
10000 @option{-O2}, @option{-O3}, @option{-Os}.
10002 @item -falign-functions
10003 @itemx -falign-functions=@var{n}
10004 @itemx -falign-functions=@var{n}:@var{m}
10005 @itemx -falign-functions=@var{n}:@var{m}:@var{n2}
10006 @itemx -falign-functions=@var{n}:@var{m}:@var{n2}:@var{m2}
10007 @opindex falign-functions
10008 Align the start of functions to the next power-of-two greater than
10009 @var{n}, skipping up to @var{m}-1 bytes. This ensures that at least
10010 the first @var{m} bytes of the function can be fetched by the CPU
10011 without crossing an @var{n}-byte alignment boundary.
10013 If @var{m} is not specified, it defaults to @var{n}.
10015 Examples: @option{-falign-functions=32} aligns functions to the next
10016 32-byte boundary, @option{-falign-functions=24} aligns to the next
10017 32-byte boundary only if this can be done by skipping 23 bytes or less,
10018 @option{-falign-functions=32:7} aligns to the next
10019 32-byte boundary only if this can be done by skipping 6 bytes or less.
10021 The second pair of @var{n2}:@var{m2} values allows you to specify
10022 a secondary alignment: @option{-falign-functions=64:7:32:3} aligns to
10023 the next 64-byte boundary if this can be done by skipping 6 bytes or less,
10024 otherwise aligns to the next 32-byte boundary if this can be done
10025 by skipping 2 bytes or less.
10026 If @var{m2} is not specified, it defaults to @var{n2}.
10028 Some assemblers only support this flag when @var{n} is a power of two;
10029 in that case, it is rounded up.
10031 @option{-fno-align-functions} and @option{-falign-functions=1} are
10032 equivalent and mean that functions are not aligned.
10034 If @var{n} is not specified or is zero, use a machine-dependent default.
10035 The maximum allowed @var{n} option value is 65536.
10037 Enabled at levels @option{-O2}, @option{-O3}.
10039 @item -flimit-function-alignment
10040 If this option is enabled, the compiler tries to avoid unnecessarily
10041 overaligning functions. It attempts to instruct the assembler to align
10042 by the amount specified by @option{-falign-functions}, but not to
10043 skip more bytes than the size of the function.
10045 @item -falign-labels
10046 @itemx -falign-labels=@var{n}
10047 @itemx -falign-labels=@var{n}:@var{m}
10048 @itemx -falign-labels=@var{n}:@var{m}:@var{n2}
10049 @itemx -falign-labels=@var{n}:@var{m}:@var{n2}:@var{m2}
10050 @opindex falign-labels
10051 Align all branch targets to a power-of-two boundary.
10053 Parameters of this option are analogous to the @option{-falign-functions} option.
10054 @option{-fno-align-labels} and @option{-falign-labels=1} are
10055 equivalent and mean that labels are not aligned.
10057 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
10058 are greater than this value, then their values are used instead.
10060 If @var{n} is not specified or is zero, use a machine-dependent default
10061 which is very likely to be @samp{1}, meaning no alignment.
10062 The maximum allowed @var{n} option value is 65536.
10064 Enabled at levels @option{-O2}, @option{-O3}.
10066 @item -falign-loops
10067 @itemx -falign-loops=@var{n}
10068 @itemx -falign-loops=@var{n}:@var{m}
10069 @itemx -falign-loops=@var{n}:@var{m}:@var{n2}
10070 @itemx -falign-loops=@var{n}:@var{m}:@var{n2}:@var{m2}
10071 @opindex falign-loops
10072 Align loops to a power-of-two boundary. If the loops are executed
10073 many times, this makes up for any execution of the dummy padding
10076 Parameters of this option are analogous to the @option{-falign-functions} option.
10077 @option{-fno-align-loops} and @option{-falign-loops=1} are
10078 equivalent and mean that loops are not aligned.
10079 The maximum allowed @var{n} option value is 65536.
10081 If @var{n} is not specified or is zero, use a machine-dependent default.
10083 Enabled at levels @option{-O2}, @option{-O3}.
10085 @item -falign-jumps
10086 @itemx -falign-jumps=@var{n}
10087 @itemx -falign-jumps=@var{n}:@var{m}
10088 @itemx -falign-jumps=@var{n}:@var{m}:@var{n2}
10089 @itemx -falign-jumps=@var{n}:@var{m}:@var{n2}:@var{m2}
10090 @opindex falign-jumps
10091 Align branch targets to a power-of-two boundary, for branch targets
10092 where the targets can only be reached by jumping. In this case,
10093 no dummy operations need be executed.
10095 Parameters of this option are analogous to the @option{-falign-functions} option.
10096 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
10097 equivalent and mean that loops are not aligned.
10099 If @var{n} is not specified or is zero, use a machine-dependent default.
10100 The maximum allowed @var{n} option value is 65536.
10102 Enabled at levels @option{-O2}, @option{-O3}.
10104 @item -funit-at-a-time
10105 @opindex funit-at-a-time
10106 This option is left for compatibility reasons. @option{-funit-at-a-time}
10107 has no effect, while @option{-fno-unit-at-a-time} implies
10108 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
10110 Enabled by default.
10112 @item -fno-toplevel-reorder
10113 @opindex fno-toplevel-reorder
10114 @opindex ftoplevel-reorder
10115 Do not reorder top-level functions, variables, and @code{asm}
10116 statements. Output them in the same order that they appear in the
10117 input file. When this option is used, unreferenced static variables
10118 are not removed. This option is intended to support existing code
10119 that relies on a particular ordering. For new code, it is better to
10120 use attributes when possible.
10122 @option{-ftoplevel-reorder} is the default at @option{-O1} and higher, and
10123 also at @option{-O0} if @option{-fsection-anchors} is explicitly requested.
10124 Additionally @option{-fno-toplevel-reorder} implies
10125 @option{-fno-section-anchors}.
10129 Constructs webs as commonly used for register allocation purposes and assign
10130 each web individual pseudo register. This allows the register allocation pass
10131 to operate on pseudos directly, but also strengthens several other optimization
10132 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
10133 however, make debugging impossible, since variables no longer stay in a
10136 Enabled by default with @option{-funroll-loops}.
10138 @item -fwhole-program
10139 @opindex fwhole-program
10140 Assume that the current compilation unit represents the whole program being
10141 compiled. All public functions and variables with the exception of @code{main}
10142 and those merged by attribute @code{externally_visible} become static functions
10143 and in effect are optimized more aggressively by interprocedural optimizers.
10145 This option should not be used in combination with @option{-flto}.
10146 Instead relying on a linker plugin should provide safer and more precise
10149 @item -flto[=@var{n}]
10151 This option runs the standard link-time optimizer. When invoked
10152 with source code, it generates GIMPLE (one of GCC's internal
10153 representations) and writes it to special ELF sections in the object
10154 file. When the object files are linked together, all the function
10155 bodies are read from these ELF sections and instantiated as if they
10156 had been part of the same translation unit.
10158 To use the link-time optimizer, @option{-flto} and optimization
10159 options should be specified at compile time and during the final link.
10160 It is recommended that you compile all the files participating in the
10161 same link with the same options and also specify those options at
10166 gcc -c -O2 -flto foo.c
10167 gcc -c -O2 -flto bar.c
10168 gcc -o myprog -flto -O2 foo.o bar.o
10171 The first two invocations to GCC save a bytecode representation
10172 of GIMPLE into special ELF sections inside @file{foo.o} and
10173 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
10174 @file{foo.o} and @file{bar.o}, merges the two files into a single
10175 internal image, and compiles the result as usual. Since both
10176 @file{foo.o} and @file{bar.o} are merged into a single image, this
10177 causes all the interprocedural analyses and optimizations in GCC to
10178 work across the two files as if they were a single one. This means,
10179 for example, that the inliner is able to inline functions in
10180 @file{bar.o} into functions in @file{foo.o} and vice-versa.
10182 Another (simpler) way to enable link-time optimization is:
10185 gcc -o myprog -flto -O2 foo.c bar.c
10188 The above generates bytecode for @file{foo.c} and @file{bar.c},
10189 merges them together into a single GIMPLE representation and optimizes
10190 them as usual to produce @file{myprog}.
10192 The important thing to keep in mind is that to enable link-time
10193 optimizations you need to use the GCC driver to perform the link step.
10194 GCC automatically performs link-time optimization if any of the
10195 objects involved were compiled with the @option{-flto} command-line option.
10196 You can always override
10197 the automatic decision to do link-time optimization
10198 by passing @option{-fno-lto} to the link command.
10200 To make whole program optimization effective, it is necessary to make
10201 certain whole program assumptions. The compiler needs to know
10202 what functions and variables can be accessed by libraries and runtime
10203 outside of the link-time optimized unit. When supported by the linker,
10204 the linker plugin (see @option{-fuse-linker-plugin}) passes information
10205 to the compiler about used and externally visible symbols. When
10206 the linker plugin is not available, @option{-fwhole-program} should be
10207 used to allow the compiler to make these assumptions, which leads
10208 to more aggressive optimization decisions.
10210 When a file is compiled with @option{-flto} without
10211 @option{-fuse-linker-plugin}, the generated object file is larger than
10212 a regular object file because it contains GIMPLE bytecodes and the usual
10213 final code (see @option{-ffat-lto-objects}. This means that
10214 object files with LTO information can be linked as normal object
10215 files; if @option{-fno-lto} is passed to the linker, no
10216 interprocedural optimizations are applied. Note that when
10217 @option{-fno-fat-lto-objects} is enabled the compile stage is faster
10218 but you cannot perform a regular, non-LTO link on them.
10220 When producing the final binary, GCC only
10221 applies link-time optimizations to those files that contain bytecode.
10222 Therefore, you can mix and match object files and libraries with
10223 GIMPLE bytecodes and final object code. GCC automatically selects
10224 which files to optimize in LTO mode and which files to link without
10225 further processing.
10227 Generally, options specified at link time override those
10228 specified at compile time, although in some cases GCC attempts to infer
10229 link-time options from the settings used to compile the input files.
10231 If you do not specify an optimization level option @option{-O} at
10232 link time, then GCC uses the highest optimization level
10233 used when compiling the object files. Note that it is generally
10234 ineffective to specify an optimization level option only at link time and
10235 not at compile time, for two reasons. First, compiling without
10236 optimization suppresses compiler passes that gather information
10237 needed for effective optimization at link time. Second, some early
10238 optimization passes can be performed only at compile time and
10241 There are some code generation flags preserved by GCC when
10242 generating bytecodes, as they need to be used during the final link.
10243 Currently, the following options and their settings are taken from
10244 the first object file that explicitly specifies them:
10245 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
10246 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
10247 and all the @option{-m} target flags.
10249 Certain ABI-changing flags are required to match in all compilation units,
10250 and trying to override this at link time with a conflicting value
10251 is ignored. This includes options such as @option{-freg-struct-return}
10252 and @option{-fpcc-struct-return}.
10254 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
10255 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
10256 are passed through to the link stage and merged conservatively for
10257 conflicting translation units. Specifically
10258 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
10259 precedence; and for example @option{-ffp-contract=off} takes precedence
10260 over @option{-ffp-contract=fast}. You can override them at link time.
10262 If LTO encounters objects with C linkage declared with incompatible
10263 types in separate translation units to be linked together (undefined
10264 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
10265 issued. The behavior is still undefined at run time. Similar
10266 diagnostics may be raised for other languages.
10268 Another feature of LTO is that it is possible to apply interprocedural
10269 optimizations on files written in different languages:
10273 g++ -c -flto bar.cc
10274 gfortran -c -flto baz.f90
10275 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
10278 Notice that the final link is done with @command{g++} to get the C++
10279 runtime libraries and @option{-lgfortran} is added to get the Fortran
10280 runtime libraries. In general, when mixing languages in LTO mode, you
10281 should use the same link command options as when mixing languages in a
10282 regular (non-LTO) compilation.
10284 If object files containing GIMPLE bytecode are stored in a library archive, say
10285 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
10286 are using a linker with plugin support. To create static libraries suitable
10287 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
10288 and @command{ranlib};
10289 to show the symbols of object files with GIMPLE bytecode, use
10290 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
10291 and @command{nm} have been compiled with plugin support. At link time, use the
10292 flag @option{-fuse-linker-plugin} to ensure that the library participates in
10293 the LTO optimization process:
10296 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
10299 With the linker plugin enabled, the linker extracts the needed
10300 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
10301 to make them part of the aggregated GIMPLE image to be optimized.
10303 If you are not using a linker with plugin support and/or do not
10304 enable the linker plugin, then the objects inside @file{libfoo.a}
10305 are extracted and linked as usual, but they do not participate
10306 in the LTO optimization process. In order to make a static library suitable
10307 for both LTO optimization and usual linkage, compile its object files with
10308 @option{-flto} @option{-ffat-lto-objects}.
10310 Link-time optimizations do not require the presence of the whole program to
10311 operate. If the program does not require any symbols to be exported, it is
10312 possible to combine @option{-flto} and @option{-fwhole-program} to allow
10313 the interprocedural optimizers to use more aggressive assumptions which may
10314 lead to improved optimization opportunities.
10315 Use of @option{-fwhole-program} is not needed when linker plugin is
10316 active (see @option{-fuse-linker-plugin}).
10318 The current implementation of LTO makes no
10319 attempt to generate bytecode that is portable between different
10320 types of hosts. The bytecode files are versioned and there is a
10321 strict version check, so bytecode files generated in one version of
10322 GCC do not work with an older or newer version of GCC.
10324 Link-time optimization does not work well with generation of debugging
10325 information on systems other than those using a combination of ELF and
10328 If you specify the optional @var{n}, the optimization and code
10329 generation done at link time is executed in parallel using @var{n}
10330 parallel jobs by utilizing an installed @command{make} program. The
10331 environment variable @env{MAKE} may be used to override the program
10332 used. The default value for @var{n} is 1.
10334 You can also specify @option{-flto=jobserver} to use GNU make's
10335 job server mode to determine the number of parallel jobs. This
10336 is useful when the Makefile calling GCC is already executing in parallel.
10337 You must prepend a @samp{+} to the command recipe in the parent Makefile
10338 for this to work. This option likely only works if @env{MAKE} is
10341 @item -flto-partition=@var{alg}
10342 @opindex flto-partition
10343 Specify the partitioning algorithm used by the link-time optimizer.
10344 The value is either @samp{1to1} to specify a partitioning mirroring
10345 the original source files or @samp{balanced} to specify partitioning
10346 into equally sized chunks (whenever possible) or @samp{max} to create
10347 new partition for every symbol where possible. Specifying @samp{none}
10348 as an algorithm disables partitioning and streaming completely.
10349 The default value is @samp{balanced}. While @samp{1to1} can be used
10350 as an workaround for various code ordering issues, the @samp{max}
10351 partitioning is intended for internal testing only.
10352 The value @samp{one} specifies that exactly one partition should be
10353 used while the value @samp{none} bypasses partitioning and executes
10354 the link-time optimization step directly from the WPA phase.
10356 @item -flto-odr-type-merging
10357 @opindex flto-odr-type-merging
10358 Enable streaming of mangled types names of C++ types and their unification
10359 at link time. This increases size of LTO object files, but enables
10360 diagnostics about One Definition Rule violations.
10362 @item -flto-compression-level=@var{n}
10363 @opindex flto-compression-level
10364 This option specifies the level of compression used for intermediate
10365 language written to LTO object files, and is only meaningful in
10366 conjunction with LTO mode (@option{-flto}). Valid
10367 values are 0 (no compression) to 9 (maximum compression). Values
10368 outside this range are clamped to either 0 or 9. If the option is not
10369 given, a default balanced compression setting is used.
10371 @item -fuse-linker-plugin
10372 @opindex fuse-linker-plugin
10373 Enables the use of a linker plugin during link-time optimization. This
10374 option relies on plugin support in the linker, which is available in gold
10375 or in GNU ld 2.21 or newer.
10377 This option enables the extraction of object files with GIMPLE bytecode out
10378 of library archives. This improves the quality of optimization by exposing
10379 more code to the link-time optimizer. This information specifies what
10380 symbols can be accessed externally (by non-LTO object or during dynamic
10381 linking). Resulting code quality improvements on binaries (and shared
10382 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
10383 See @option{-flto} for a description of the effect of this flag and how to
10386 This option is enabled by default when LTO support in GCC is enabled
10387 and GCC was configured for use with
10388 a linker supporting plugins (GNU ld 2.21 or newer or gold).
10390 @item -ffat-lto-objects
10391 @opindex ffat-lto-objects
10392 Fat LTO objects are object files that contain both the intermediate language
10393 and the object code. This makes them usable for both LTO linking and normal
10394 linking. This option is effective only when compiling with @option{-flto}
10395 and is ignored at link time.
10397 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
10398 requires the complete toolchain to be aware of LTO. It requires a linker with
10399 linker plugin support for basic functionality. Additionally,
10400 @command{nm}, @command{ar} and @command{ranlib}
10401 need to support linker plugins to allow a full-featured build environment
10402 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
10403 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
10404 to these tools. With non fat LTO makefiles need to be modified to use them.
10406 Note that modern binutils provide plugin auto-load mechanism.
10407 Installing the linker plugin into @file{$libdir/bfd-plugins} has the same
10408 effect as usage of the command wrappers (@command{gcc-ar}, @command{gcc-nm} and
10409 @command{gcc-ranlib}).
10411 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
10414 @item -fcompare-elim
10415 @opindex fcompare-elim
10416 After register allocation and post-register allocation instruction splitting,
10417 identify arithmetic instructions that compute processor flags similar to a
10418 comparison operation based on that arithmetic. If possible, eliminate the
10419 explicit comparison operation.
10421 This pass only applies to certain targets that cannot explicitly represent
10422 the comparison operation before register allocation is complete.
10424 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
10426 @item -fcprop-registers
10427 @opindex fcprop-registers
10428 After register allocation and post-register allocation instruction splitting,
10429 perform a copy-propagation pass to try to reduce scheduling dependencies
10430 and occasionally eliminate the copy.
10432 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
10434 @item -fprofile-correction
10435 @opindex fprofile-correction
10436 Profiles collected using an instrumented binary for multi-threaded programs may
10437 be inconsistent due to missed counter updates. When this option is specified,
10438 GCC uses heuristics to correct or smooth out such inconsistencies. By
10439 default, GCC emits an error message when an inconsistent profile is detected.
10441 This option is enabled by @option{-fauto-profile}.
10443 @item -fprofile-use
10444 @itemx -fprofile-use=@var{path}
10445 @opindex fprofile-use
10446 Enable profile feedback-directed optimizations,
10447 and the following optimizations, many of which
10448 are generally profitable only with profile feedback available:
10450 @gccoptlist{-fbranch-probabilities -fprofile-values @gol
10451 -funroll-loops -fpeel-loops -ftracer -fvpt @gol
10452 -finline-functions -fipa-cp -fipa-cp-clone -fipa-bit-cp @gol
10453 -fpredictive-commoning -fsplit-loops -funswitch-loops @gol
10454 -fgcse-after-reload -ftree-loop-vectorize -ftree-slp-vectorize @gol
10455 -fvect-cost-model=dynamic -ftree-loop-distribute-patterns @gol
10456 -fprofile-reorder-functions}
10458 Before you can use this option, you must first generate profiling information.
10459 @xref{Instrumentation Options}, for information about the
10460 @option{-fprofile-generate} option.
10462 By default, GCC emits an error message if the feedback profiles do not
10463 match the source code. This error can be turned into a warning by using
10464 @option{-Wno-error=coverage-mismatch}. Note this may result in poorly
10465 optimized code. Additionally, by default, GCC also emits a warning message if
10466 the feedback profiles do not exist (see @option{-Wmissing-profile}).
10468 If @var{path} is specified, GCC looks at the @var{path} to find
10469 the profile feedback data files. See @option{-fprofile-dir}.
10471 @item -fauto-profile
10472 @itemx -fauto-profile=@var{path}
10473 @opindex fauto-profile
10474 Enable sampling-based feedback-directed optimizations,
10475 and the following optimizations,
10476 many of which are generally profitable only with profile feedback available:
10478 @gccoptlist{-fbranch-probabilities -fprofile-values @gol
10479 -funroll-loops -fpeel-loops -ftracer -fvpt @gol
10480 -finline-functions -fipa-cp -fipa-cp-clone -fipa-bit-cp @gol
10481 -fpredictive-commoning -fsplit-loops -funswitch-loops @gol
10482 -fgcse-after-reload -ftree-loop-vectorize -ftree-slp-vectorize @gol
10483 -fvect-cost-model=dynamic -ftree-loop-distribute-patterns @gol
10484 -fprofile-correction}
10486 @var{path} is the name of a file containing AutoFDO profile information.
10487 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
10489 Producing an AutoFDO profile data file requires running your program
10490 with the @command{perf} utility on a supported GNU/Linux target system.
10491 For more information, see @uref{https://perf.wiki.kernel.org/}.
10495 perf record -e br_inst_retired:near_taken -b -o perf.data \
10499 Then use the @command{create_gcov} tool to convert the raw profile data
10500 to a format that can be used by GCC.@ You must also supply the
10501 unstripped binary for your program to this tool.
10502 See @uref{https://github.com/google/autofdo}.
10506 create_gcov --binary=your_program.unstripped --profile=perf.data \
10507 --gcov=profile.afdo
10511 The following options control compiler behavior regarding floating-point
10512 arithmetic. These options trade off between speed and
10513 correctness. All must be specifically enabled.
10516 @item -ffloat-store
10517 @opindex ffloat-store
10518 Do not store floating-point variables in registers, and inhibit other
10519 options that might change whether a floating-point value is taken from a
10520 register or memory.
10522 @cindex floating-point precision
10523 This option prevents undesirable excess precision on machines such as
10524 the 68000 where the floating registers (of the 68881) keep more
10525 precision than a @code{double} is supposed to have. Similarly for the
10526 x86 architecture. For most programs, the excess precision does only
10527 good, but a few programs rely on the precise definition of IEEE floating
10528 point. Use @option{-ffloat-store} for such programs, after modifying
10529 them to store all pertinent intermediate computations into variables.
10531 @item -fexcess-precision=@var{style}
10532 @opindex fexcess-precision
10533 This option allows further control over excess precision on machines
10534 where floating-point operations occur in a format with more precision or
10535 range than the IEEE standard and interchange floating-point types. By
10536 default, @option{-fexcess-precision=fast} is in effect; this means that
10537 operations may be carried out in a wider precision than the types specified
10538 in the source if that would result in faster code, and it is unpredictable
10539 when rounding to the types specified in the source code takes place.
10540 When compiling C, if @option{-fexcess-precision=standard} is specified then
10541 excess precision follows the rules specified in ISO C99; in particular,
10542 both casts and assignments cause values to be rounded to their
10543 semantic types (whereas @option{-ffloat-store} only affects
10544 assignments). This option is enabled by default for C if a strict
10545 conformance option such as @option{-std=c99} is used.
10546 @option{-ffast-math} enables @option{-fexcess-precision=fast} by default
10547 regardless of whether a strict conformance option is used.
10550 @option{-fexcess-precision=standard} is not implemented for languages
10551 other than C. On the x86, it has no effect if @option{-mfpmath=sse}
10552 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
10553 semantics apply without excess precision, and in the latter, rounding
10557 @opindex ffast-math
10558 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
10559 @option{-ffinite-math-only}, @option{-fno-rounding-math},
10560 @option{-fno-signaling-nans}, @option{-fcx-limited-range} and
10561 @option{-fexcess-precision=fast}.
10563 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
10565 This option is not turned on by any @option{-O} option besides
10566 @option{-Ofast} since it can result in incorrect output for programs
10567 that depend on an exact implementation of IEEE or ISO rules/specifications
10568 for math functions. It may, however, yield faster code for programs
10569 that do not require the guarantees of these specifications.
10571 @item -fno-math-errno
10572 @opindex fno-math-errno
10573 @opindex fmath-errno
10574 Do not set @code{errno} after calling math functions that are executed
10575 with a single instruction, e.g., @code{sqrt}. A program that relies on
10576 IEEE exceptions for math error handling may want to use this flag
10577 for speed while maintaining IEEE arithmetic compatibility.
10579 This option is not turned on by any @option{-O} option since
10580 it can result in incorrect output for programs that depend on
10581 an exact implementation of IEEE or ISO rules/specifications for
10582 math functions. It may, however, yield faster code for programs
10583 that do not require the guarantees of these specifications.
10585 The default is @option{-fmath-errno}.
10587 On Darwin systems, the math library never sets @code{errno}. There is
10588 therefore no reason for the compiler to consider the possibility that
10589 it might, and @option{-fno-math-errno} is the default.
10591 @item -funsafe-math-optimizations
10592 @opindex funsafe-math-optimizations
10594 Allow optimizations for floating-point arithmetic that (a) assume
10595 that arguments and results are valid and (b) may violate IEEE or
10596 ANSI standards. When used at link time, it may include libraries
10597 or startup files that change the default FPU control word or other
10598 similar optimizations.
10600 This option is not turned on by any @option{-O} option since
10601 it can result in incorrect output for programs that depend on
10602 an exact implementation of IEEE or ISO rules/specifications for
10603 math functions. It may, however, yield faster code for programs
10604 that do not require the guarantees of these specifications.
10605 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
10606 @option{-fassociative-math} and @option{-freciprocal-math}.
10608 The default is @option{-fno-unsafe-math-optimizations}.
10610 @item -fassociative-math
10611 @opindex fassociative-math
10613 Allow re-association of operands in series of floating-point operations.
10614 This violates the ISO C and C++ language standard by possibly changing
10615 computation result. NOTE: re-ordering may change the sign of zero as
10616 well as ignore NaNs and inhibit or create underflow or overflow (and
10617 thus cannot be used on code that relies on rounding behavior like
10618 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
10619 and thus may not be used when ordered comparisons are required.
10620 This option requires that both @option{-fno-signed-zeros} and
10621 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
10622 much sense with @option{-frounding-math}. For Fortran the option
10623 is automatically enabled when both @option{-fno-signed-zeros} and
10624 @option{-fno-trapping-math} are in effect.
10626 The default is @option{-fno-associative-math}.
10628 @item -freciprocal-math
10629 @opindex freciprocal-math
10631 Allow the reciprocal of a value to be used instead of dividing by
10632 the value if this enables optimizations. For example @code{x / y}
10633 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
10634 is subject to common subexpression elimination. Note that this loses
10635 precision and increases the number of flops operating on the value.
10637 The default is @option{-fno-reciprocal-math}.
10639 @item -ffinite-math-only
10640 @opindex ffinite-math-only
10641 Allow optimizations for floating-point arithmetic that assume
10642 that arguments and results are not NaNs or +-Infs.
10644 This option is not turned on by any @option{-O} option since
10645 it can result in incorrect output for programs that depend on
10646 an exact implementation of IEEE or ISO rules/specifications for
10647 math functions. It may, however, yield faster code for programs
10648 that do not require the guarantees of these specifications.
10650 The default is @option{-fno-finite-math-only}.
10652 @item -fno-signed-zeros
10653 @opindex fno-signed-zeros
10654 @opindex fsigned-zeros
10655 Allow optimizations for floating-point arithmetic that ignore the
10656 signedness of zero. IEEE arithmetic specifies the behavior of
10657 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
10658 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
10659 This option implies that the sign of a zero result isn't significant.
10661 The default is @option{-fsigned-zeros}.
10663 @item -fno-trapping-math
10664 @opindex fno-trapping-math
10665 @opindex ftrapping-math
10666 Compile code assuming that floating-point operations cannot generate
10667 user-visible traps. These traps include division by zero, overflow,
10668 underflow, inexact result and invalid operation. This option requires
10669 that @option{-fno-signaling-nans} be in effect. Setting this option may
10670 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
10672 This option should never be turned on by any @option{-O} option since
10673 it can result in incorrect output for programs that depend on
10674 an exact implementation of IEEE or ISO rules/specifications for
10677 The default is @option{-ftrapping-math}.
10679 @item -frounding-math
10680 @opindex frounding-math
10681 Disable transformations and optimizations that assume default floating-point
10682 rounding behavior. This is round-to-zero for all floating point
10683 to integer conversions, and round-to-nearest for all other arithmetic
10684 truncations. This option should be specified for programs that change
10685 the FP rounding mode dynamically, or that may be executed with a
10686 non-default rounding mode. This option disables constant folding of
10687 floating-point expressions at compile time (which may be affected by
10688 rounding mode) and arithmetic transformations that are unsafe in the
10689 presence of sign-dependent rounding modes.
10691 The default is @option{-fno-rounding-math}.
10693 This option is experimental and does not currently guarantee to
10694 disable all GCC optimizations that are affected by rounding mode.
10695 Future versions of GCC may provide finer control of this setting
10696 using C99's @code{FENV_ACCESS} pragma. This command-line option
10697 will be used to specify the default state for @code{FENV_ACCESS}.
10699 @item -fsignaling-nans
10700 @opindex fsignaling-nans
10701 Compile code assuming that IEEE signaling NaNs may generate user-visible
10702 traps during floating-point operations. Setting this option disables
10703 optimizations that may change the number of exceptions visible with
10704 signaling NaNs. This option implies @option{-ftrapping-math}.
10706 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
10709 The default is @option{-fno-signaling-nans}.
10711 This option is experimental and does not currently guarantee to
10712 disable all GCC optimizations that affect signaling NaN behavior.
10714 @item -fno-fp-int-builtin-inexact
10715 @opindex fno-fp-int-builtin-inexact
10716 @opindex ffp-int-builtin-inexact
10717 Do not allow the built-in functions @code{ceil}, @code{floor},
10718 @code{round} and @code{trunc}, and their @code{float} and @code{long
10719 double} variants, to generate code that raises the ``inexact''
10720 floating-point exception for noninteger arguments. ISO C99 and C11
10721 allow these functions to raise the ``inexact'' exception, but ISO/IEC
10722 TS 18661-1:2014, the C bindings to IEEE 754-2008, does not allow these
10723 functions to do so.
10725 The default is @option{-ffp-int-builtin-inexact}, allowing the
10726 exception to be raised. This option does nothing unless
10727 @option{-ftrapping-math} is in effect.
10729 Even if @option{-fno-fp-int-builtin-inexact} is used, if the functions
10730 generate a call to a library function then the ``inexact'' exception
10731 may be raised if the library implementation does not follow TS 18661.
10733 @item -fsingle-precision-constant
10734 @opindex fsingle-precision-constant
10735 Treat floating-point constants as single precision instead of
10736 implicitly converting them to double-precision constants.
10738 @item -fcx-limited-range
10739 @opindex fcx-limited-range
10740 When enabled, this option states that a range reduction step is not
10741 needed when performing complex division. Also, there is no checking
10742 whether the result of a complex multiplication or division is @code{NaN
10743 + I*NaN}, with an attempt to rescue the situation in that case. The
10744 default is @option{-fno-cx-limited-range}, but is enabled by
10745 @option{-ffast-math}.
10747 This option controls the default setting of the ISO C99
10748 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
10751 @item -fcx-fortran-rules
10752 @opindex fcx-fortran-rules
10753 Complex multiplication and division follow Fortran rules. Range
10754 reduction is done as part of complex division, but there is no checking
10755 whether the result of a complex multiplication or division is @code{NaN
10756 + I*NaN}, with an attempt to rescue the situation in that case.
10758 The default is @option{-fno-cx-fortran-rules}.
10762 The following options control optimizations that may improve
10763 performance, but are not enabled by any @option{-O} options. This
10764 section includes experimental options that may produce broken code.
10767 @item -fbranch-probabilities
10768 @opindex fbranch-probabilities
10769 After running a program compiled with @option{-fprofile-arcs}
10770 (@pxref{Instrumentation Options}),
10771 you can compile it a second time using
10772 @option{-fbranch-probabilities}, to improve optimizations based on
10773 the number of times each branch was taken. When a program
10774 compiled with @option{-fprofile-arcs} exits, it saves arc execution
10775 counts to a file called @file{@var{sourcename}.gcda} for each source
10776 file. The information in this data file is very dependent on the
10777 structure of the generated code, so you must use the same source code
10778 and the same optimization options for both compilations.
10780 With @option{-fbranch-probabilities}, GCC puts a
10781 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
10782 These can be used to improve optimization. Currently, they are only
10783 used in one place: in @file{reorg.c}, instead of guessing which path a
10784 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
10785 exactly determine which path is taken more often.
10787 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10789 @item -fprofile-values
10790 @opindex fprofile-values
10791 If combined with @option{-fprofile-arcs}, it adds code so that some
10792 data about values of expressions in the program is gathered.
10794 With @option{-fbranch-probabilities}, it reads back the data gathered
10795 from profiling values of expressions for usage in optimizations.
10797 Enabled by @option{-fprofile-generate}, @option{-fprofile-use}, and
10798 @option{-fauto-profile}.
10800 @item -fprofile-reorder-functions
10801 @opindex fprofile-reorder-functions
10802 Function reordering based on profile instrumentation collects
10803 first time of execution of a function and orders these functions
10804 in ascending order.
10806 Enabled with @option{-fprofile-use}.
10810 If combined with @option{-fprofile-arcs}, this option instructs the compiler
10811 to add code to gather information about values of expressions.
10813 With @option{-fbranch-probabilities}, it reads back the data gathered
10814 and actually performs the optimizations based on them.
10815 Currently the optimizations include specialization of division operations
10816 using the knowledge about the value of the denominator.
10818 Enabled with @option{-fprofile-use} and @option{-fauto-profile}.
10820 @item -frename-registers
10821 @opindex frename-registers
10822 Attempt to avoid false dependencies in scheduled code by making use
10823 of registers left over after register allocation. This optimization
10824 most benefits processors with lots of registers. Depending on the
10825 debug information format adopted by the target, however, it can
10826 make debugging impossible, since variables no longer stay in
10827 a ``home register''.
10829 Enabled by default with @option{-funroll-loops}.
10831 @item -fschedule-fusion
10832 @opindex fschedule-fusion
10833 Performs a target dependent pass over the instruction stream to schedule
10834 instructions of same type together because target machine can execute them
10835 more efficiently if they are adjacent to each other in the instruction flow.
10837 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
10841 Perform tail duplication to enlarge superblock size. This transformation
10842 simplifies the control flow of the function allowing other optimizations to do
10845 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10847 @item -funroll-loops
10848 @opindex funroll-loops
10849 Unroll loops whose number of iterations can be determined at compile time or
10850 upon entry to the loop. @option{-funroll-loops} implies
10851 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
10852 It also turns on complete loop peeling (i.e.@: complete removal of loops with
10853 a small constant number of iterations). This option makes code larger, and may
10854 or may not make it run faster.
10856 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10858 @item -funroll-all-loops
10859 @opindex funroll-all-loops
10860 Unroll all loops, even if their number of iterations is uncertain when
10861 the loop is entered. This usually makes programs run more slowly.
10862 @option{-funroll-all-loops} implies the same options as
10863 @option{-funroll-loops}.
10866 @opindex fpeel-loops
10867 Peels loops for which there is enough information that they do not
10868 roll much (from profile feedback or static analysis). It also turns on
10869 complete loop peeling (i.e.@: complete removal of loops with small constant
10870 number of iterations).
10872 Enabled by @option{-O3}, @option{-fprofile-use}, and @option{-fauto-profile}.
10874 @item -fmove-loop-invariants
10875 @opindex fmove-loop-invariants
10876 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
10877 at level @option{-O1} and higher, except for @option{-Og}.
10879 @item -fsplit-loops
10880 @opindex fsplit-loops
10881 Split a loop into two if it contains a condition that's always true
10882 for one side of the iteration space and false for the other.
10884 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10886 @item -funswitch-loops
10887 @opindex funswitch-loops
10888 Move branches with loop invariant conditions out of the loop, with duplicates
10889 of the loop on both branches (modified according to result of the condition).
10891 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10893 @item -fversion-loops-for-strides
10894 @opindex fversion-loops-for-strides
10895 If a loop iterates over an array with a variable stride, create another
10896 version of the loop that assumes the stride is always one. For example:
10899 for (int i = 0; i < n; ++i)
10900 x[i * stride] = @dots{};
10907 for (int i = 0; i < n; ++i)
10910 for (int i = 0; i < n; ++i)
10911 x[i * stride] = @dots{};
10914 This is particularly useful for assumed-shape arrays in Fortran where
10915 (for example) it allows better vectorization assuming contiguous accesses.
10916 This flag is enabled by default at @option{-O3}.
10917 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10919 @item -ffunction-sections
10920 @itemx -fdata-sections
10921 @opindex ffunction-sections
10922 @opindex fdata-sections
10923 Place each function or data item into its own section in the output
10924 file if the target supports arbitrary sections. The name of the
10925 function or the name of the data item determines the section's name
10926 in the output file.
10928 Use these options on systems where the linker can perform optimizations to
10929 improve locality of reference in the instruction space. Most systems using the
10930 ELF object format have linkers with such optimizations. On AIX, the linker
10931 rearranges sections (CSECTs) based on the call graph. The performance impact
10934 Together with a linker garbage collection (linker @option{--gc-sections}
10935 option) these options may lead to smaller statically-linked executables (after
10938 On ELF/DWARF systems these options do not degenerate the quality of the debug
10939 information. There could be issues with other object files/debug info formats.
10941 Only use these options when there are significant benefits from doing so. When
10942 you specify these options, the assembler and linker create larger object and
10943 executable files and are also slower. These options affect code generation.
10944 They prevent optimizations by the compiler and assembler using relative
10945 locations inside a translation unit since the locations are unknown until
10946 link time. An example of such an optimization is relaxing calls to short call
10949 @item -fbranch-target-load-optimize
10950 @opindex fbranch-target-load-optimize
10951 Perform branch target register load optimization before prologue / epilogue
10953 The use of target registers can typically be exposed only during reload,
10954 thus hoisting loads out of loops and doing inter-block scheduling needs
10955 a separate optimization pass.
10957 @item -fbranch-target-load-optimize2
10958 @opindex fbranch-target-load-optimize2
10959 Perform branch target register load optimization after prologue / epilogue
10962 @item -fbtr-bb-exclusive
10963 @opindex fbtr-bb-exclusive
10964 When performing branch target register load optimization, don't reuse
10965 branch target registers within any basic block.
10968 @opindex fstdarg-opt
10969 Optimize the prologue of variadic argument functions with respect to usage of
10972 @item -fsection-anchors
10973 @opindex fsection-anchors
10974 Try to reduce the number of symbolic address calculations by using
10975 shared ``anchor'' symbols to address nearby objects. This transformation
10976 can help to reduce the number of GOT entries and GOT accesses on some
10979 For example, the implementation of the following function @code{foo}:
10982 static int a, b, c;
10983 int foo (void) @{ return a + b + c; @}
10987 usually calculates the addresses of all three variables, but if you
10988 compile it with @option{-fsection-anchors}, it accesses the variables
10989 from a common anchor point instead. The effect is similar to the
10990 following pseudocode (which isn't valid C):
10995 register int *xr = &x;
10996 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
11000 Not all targets support this option.
11002 @item --param @var{name}=@var{value}
11004 In some places, GCC uses various constants to control the amount of
11005 optimization that is done. For example, GCC does not inline functions
11006 that contain more than a certain number of instructions. You can
11007 control some of these constants on the command line using the
11008 @option{--param} option.
11010 The names of specific parameters, and the meaning of the values, are
11011 tied to the internals of the compiler, and are subject to change
11012 without notice in future releases.
11014 In order to get minimal, maximal and default value of a parameter,
11015 one can use @option{--help=param -Q} options.
11017 In each case, the @var{value} is an integer. The allowable choices for
11021 @item predictable-branch-outcome
11022 When branch is predicted to be taken with probability lower than this threshold
11023 (in percent), then it is considered well predictable.
11025 @item max-rtl-if-conversion-insns
11026 RTL if-conversion tries to remove conditional branches around a block and
11027 replace them with conditionally executed instructions. This parameter
11028 gives the maximum number of instructions in a block which should be
11029 considered for if-conversion. The compiler will
11030 also use other heuristics to decide whether if-conversion is likely to be
11033 @item max-rtl-if-conversion-predictable-cost
11034 @itemx max-rtl-if-conversion-unpredictable-cost
11035 RTL if-conversion will try to remove conditional branches around a block
11036 and replace them with conditionally executed instructions. These parameters
11037 give the maximum permissible cost for the sequence that would be generated
11038 by if-conversion depending on whether the branch is statically determined
11039 to be predictable or not. The units for this parameter are the same as
11040 those for the GCC internal seq_cost metric. The compiler will try to
11041 provide a reasonable default for this parameter using the BRANCH_COST
11044 @item max-crossjump-edges
11045 The maximum number of incoming edges to consider for cross-jumping.
11046 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
11047 the number of edges incoming to each block. Increasing values mean
11048 more aggressive optimization, making the compilation time increase with
11049 probably small improvement in executable size.
11051 @item min-crossjump-insns
11052 The minimum number of instructions that must be matched at the end
11053 of two blocks before cross-jumping is performed on them. This
11054 value is ignored in the case where all instructions in the block being
11055 cross-jumped from are matched.
11057 @item max-grow-copy-bb-insns
11058 The maximum code size expansion factor when copying basic blocks
11059 instead of jumping. The expansion is relative to a jump instruction.
11061 @item max-goto-duplication-insns
11062 The maximum number of instructions to duplicate to a block that jumps
11063 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
11064 passes, GCC factors computed gotos early in the compilation process,
11065 and unfactors them as late as possible. Only computed jumps at the
11066 end of a basic blocks with no more than max-goto-duplication-insns are
11069 @item max-delay-slot-insn-search
11070 The maximum number of instructions to consider when looking for an
11071 instruction to fill a delay slot. If more than this arbitrary number of
11072 instructions are searched, the time savings from filling the delay slot
11073 are minimal, so stop searching. Increasing values mean more
11074 aggressive optimization, making the compilation time increase with probably
11075 small improvement in execution time.
11077 @item max-delay-slot-live-search
11078 When trying to fill delay slots, the maximum number of instructions to
11079 consider when searching for a block with valid live register
11080 information. Increasing this arbitrarily chosen value means more
11081 aggressive optimization, increasing the compilation time. This parameter
11082 should be removed when the delay slot code is rewritten to maintain the
11083 control-flow graph.
11085 @item max-gcse-memory
11086 The approximate maximum amount of memory that can be allocated in
11087 order to perform the global common subexpression elimination
11088 optimization. If more memory than specified is required, the
11089 optimization is not done.
11091 @item max-gcse-insertion-ratio
11092 If the ratio of expression insertions to deletions is larger than this value
11093 for any expression, then RTL PRE inserts or removes the expression and thus
11094 leaves partially redundant computations in the instruction stream.
11096 @item max-pending-list-length
11097 The maximum number of pending dependencies scheduling allows
11098 before flushing the current state and starting over. Large functions
11099 with few branches or calls can create excessively large lists which
11100 needlessly consume memory and resources.
11102 @item max-modulo-backtrack-attempts
11103 The maximum number of backtrack attempts the scheduler should make
11104 when modulo scheduling a loop. Larger values can exponentially increase
11107 @item max-inline-insns-single
11108 Several parameters control the tree inliner used in GCC@.
11109 This number sets the maximum number of instructions (counted in GCC's
11110 internal representation) in a single function that the tree inliner
11111 considers for inlining. This only affects functions declared
11112 inline and methods implemented in a class declaration (C++).
11114 @item max-inline-insns-auto
11115 When you use @option{-finline-functions} (included in @option{-O3}),
11116 a lot of functions that would otherwise not be considered for inlining
11117 by the compiler are investigated. To those functions, a different
11118 (more restrictive) limit compared to functions declared inline can
11121 @item max-inline-insns-small
11122 This is bound applied to calls which are considered relevant with
11123 @option{-finline-small-functions}.
11125 @item max-inline-insns-size
11126 This is bound applied to calls which are optimized for size. Small growth
11127 may be desirable to anticipate optimization oppurtunities exposed by inlining.
11129 @item uninlined-function-insns
11130 Number of instructions accounted by inliner for function overhead such as
11131 function prologue and epilogue.
11133 @item uninlined-function-time
11134 Extra time accounted by inliner for function overhead such as time needed to
11135 execute function prologue and epilogue
11137 @item uninlined-thunk-insns
11138 @item uninlined-thunk-time
11139 Same as @option{--param uninlined-function-insns} and
11140 @option{--param uninlined-function-time} but applied to function thunks
11142 @item inline-min-speedup
11143 When estimated performance improvement of caller + callee runtime exceeds this
11144 threshold (in percent), the function can be inlined regardless of the limit on
11145 @option{--param max-inline-insns-single} and @option{--param
11146 max-inline-insns-auto}.
11148 @item large-function-insns
11149 The limit specifying really large functions. For functions larger than this
11150 limit after inlining, inlining is constrained by
11151 @option{--param large-function-growth}. This parameter is useful primarily
11152 to avoid extreme compilation time caused by non-linear algorithms used by the
11155 @item large-function-growth
11156 Specifies maximal growth of large function caused by inlining in percents.
11157 For example, parameter value 100 limits large function growth to 2.0 times
11160 @item large-unit-insns
11161 The limit specifying large translation unit. Growth caused by inlining of
11162 units larger than this limit is limited by @option{--param inline-unit-growth}.
11163 For small units this might be too tight.
11164 For example, consider a unit consisting of function A
11165 that is inline and B that just calls A three times. If B is small relative to
11166 A, the growth of unit is 300\% and yet such inlining is very sane. For very
11167 large units consisting of small inlineable functions, however, the overall unit
11168 growth limit is needed to avoid exponential explosion of code size. Thus for
11169 smaller units, the size is increased to @option{--param large-unit-insns}
11170 before applying @option{--param inline-unit-growth}.
11172 @item inline-unit-growth
11173 Specifies maximal overall growth of the compilation unit caused by inlining.
11174 For example, parameter value 20 limits unit growth to 1.2 times the original
11175 size. Cold functions (either marked cold via an attribute or by profile
11176 feedback) are not accounted into the unit size.
11178 @item ipcp-unit-growth
11179 Specifies maximal overall growth of the compilation unit caused by
11180 interprocedural constant propagation. For example, parameter value 10 limits
11181 unit growth to 1.1 times the original size.
11183 @item large-stack-frame
11184 The limit specifying large stack frames. While inlining the algorithm is trying
11185 to not grow past this limit too much.
11187 @item large-stack-frame-growth
11188 Specifies maximal growth of large stack frames caused by inlining in percents.
11189 For example, parameter value 1000 limits large stack frame growth to 11 times
11192 @item max-inline-insns-recursive
11193 @itemx max-inline-insns-recursive-auto
11194 Specifies the maximum number of instructions an out-of-line copy of a
11195 self-recursive inline
11196 function can grow into by performing recursive inlining.
11198 @option{--param max-inline-insns-recursive} applies to functions
11200 For functions not declared inline, recursive inlining
11201 happens only when @option{-finline-functions} (included in @option{-O3}) is
11202 enabled; @option{--param max-inline-insns-recursive-auto} applies instead.
11204 @item max-inline-recursive-depth
11205 @itemx max-inline-recursive-depth-auto
11206 Specifies the maximum recursion depth used for recursive inlining.
11208 @option{--param max-inline-recursive-depth} applies to functions
11209 declared inline. For functions not declared inline, recursive inlining
11210 happens only when @option{-finline-functions} (included in @option{-O3}) is
11211 enabled; @option{--param max-inline-recursive-depth-auto} applies instead.
11213 @item min-inline-recursive-probability
11214 Recursive inlining is profitable only for function having deep recursion
11215 in average and can hurt for function having little recursion depth by
11216 increasing the prologue size or complexity of function body to other
11219 When profile feedback is available (see @option{-fprofile-generate}) the actual
11220 recursion depth can be guessed from the probability that function recurses
11221 via a given call expression. This parameter limits inlining only to call
11222 expressions whose probability exceeds the given threshold (in percents).
11224 @item early-inlining-insns
11225 Specify growth that the early inliner can make. In effect it increases
11226 the amount of inlining for code having a large abstraction penalty.
11228 @item max-early-inliner-iterations
11229 Limit of iterations of the early inliner. This basically bounds
11230 the number of nested indirect calls the early inliner can resolve.
11231 Deeper chains are still handled by late inlining.
11233 @item comdat-sharing-probability
11234 Probability (in percent) that C++ inline function with comdat visibility
11235 are shared across multiple compilation units.
11237 @item profile-func-internal-id
11238 A parameter to control whether to use function internal id in profile
11239 database lookup. If the value is 0, the compiler uses an id that
11240 is based on function assembler name and filename, which makes old profile
11241 data more tolerant to source changes such as function reordering etc.
11243 @item min-vect-loop-bound
11244 The minimum number of iterations under which loops are not vectorized
11245 when @option{-ftree-vectorize} is used. The number of iterations after
11246 vectorization needs to be greater than the value specified by this option
11247 to allow vectorization.
11249 @item gcse-cost-distance-ratio
11250 Scaling factor in calculation of maximum distance an expression
11251 can be moved by GCSE optimizations. This is currently supported only in the
11252 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
11253 is with simple expressions, i.e., the expressions that have cost
11254 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
11255 hoisting of simple expressions.
11257 @item gcse-unrestricted-cost
11258 Cost, roughly measured as the cost of a single typical machine
11259 instruction, at which GCSE optimizations do not constrain
11260 the distance an expression can travel. This is currently
11261 supported only in the code hoisting pass. The lesser the cost,
11262 the more aggressive code hoisting is. Specifying 0
11263 allows all expressions to travel unrestricted distances.
11265 @item max-hoist-depth
11266 The depth of search in the dominator tree for expressions to hoist.
11267 This is used to avoid quadratic behavior in hoisting algorithm.
11268 The value of 0 does not limit on the search, but may slow down compilation
11271 @item max-tail-merge-comparisons
11272 The maximum amount of similar bbs to compare a bb with. This is used to
11273 avoid quadratic behavior in tree tail merging.
11275 @item max-tail-merge-iterations
11276 The maximum amount of iterations of the pass over the function. This is used to
11277 limit compilation time in tree tail merging.
11279 @item store-merging-allow-unaligned
11280 Allow the store merging pass to introduce unaligned stores if it is legal to
11283 @item max-stores-to-merge
11284 The maximum number of stores to attempt to merge into wider stores in the store
11287 @item max-unrolled-insns
11288 The maximum number of instructions that a loop may have to be unrolled.
11289 If a loop is unrolled, this parameter also determines how many times
11290 the loop code is unrolled.
11292 @item max-average-unrolled-insns
11293 The maximum number of instructions biased by probabilities of their execution
11294 that a loop may have to be unrolled. If a loop is unrolled,
11295 this parameter also determines how many times the loop code is unrolled.
11297 @item max-unroll-times
11298 The maximum number of unrollings of a single loop.
11300 @item max-peeled-insns
11301 The maximum number of instructions that a loop may have to be peeled.
11302 If a loop is peeled, this parameter also determines how many times
11303 the loop code is peeled.
11305 @item max-peel-times
11306 The maximum number of peelings of a single loop.
11308 @item max-peel-branches
11309 The maximum number of branches on the hot path through the peeled sequence.
11311 @item max-completely-peeled-insns
11312 The maximum number of insns of a completely peeled loop.
11314 @item max-completely-peel-times
11315 The maximum number of iterations of a loop to be suitable for complete peeling.
11317 @item max-completely-peel-loop-nest-depth
11318 The maximum depth of a loop nest suitable for complete peeling.
11320 @item max-unswitch-insns
11321 The maximum number of insns of an unswitched loop.
11323 @item max-unswitch-level
11324 The maximum number of branches unswitched in a single loop.
11326 @item lim-expensive
11327 The minimum cost of an expensive expression in the loop invariant motion.
11329 @item iv-consider-all-candidates-bound
11330 Bound on number of candidates for induction variables, below which
11331 all candidates are considered for each use in induction variable
11332 optimizations. If there are more candidates than this,
11333 only the most relevant ones are considered to avoid quadratic time complexity.
11335 @item iv-max-considered-uses
11336 The induction variable optimizations give up on loops that contain more
11337 induction variable uses.
11339 @item iv-always-prune-cand-set-bound
11340 If the number of candidates in the set is smaller than this value,
11341 always try to remove unnecessary ivs from the set
11342 when adding a new one.
11344 @item avg-loop-niter
11345 Average number of iterations of a loop.
11347 @item dse-max-object-size
11348 Maximum size (in bytes) of objects tracked bytewise by dead store elimination.
11349 Larger values may result in larger compilation times.
11351 @item dse-max-alias-queries-per-store
11352 Maximum number of queries into the alias oracle per store.
11353 Larger values result in larger compilation times and may result in more
11354 removed dead stores.
11356 @item scev-max-expr-size
11357 Bound on size of expressions used in the scalar evolutions analyzer.
11358 Large expressions slow the analyzer.
11360 @item scev-max-expr-complexity
11361 Bound on the complexity of the expressions in the scalar evolutions analyzer.
11362 Complex expressions slow the analyzer.
11364 @item max-tree-if-conversion-phi-args
11365 Maximum number of arguments in a PHI supported by TREE if conversion
11366 unless the loop is marked with simd pragma.
11368 @item vect-max-version-for-alignment-checks
11369 The maximum number of run-time checks that can be performed when
11370 doing loop versioning for alignment in the vectorizer.
11372 @item vect-max-version-for-alias-checks
11373 The maximum number of run-time checks that can be performed when
11374 doing loop versioning for alias in the vectorizer.
11376 @item vect-max-peeling-for-alignment
11377 The maximum number of loop peels to enhance access alignment
11378 for vectorizer. Value -1 means no limit.
11380 @item max-iterations-to-track
11381 The maximum number of iterations of a loop the brute-force algorithm
11382 for analysis of the number of iterations of the loop tries to evaluate.
11384 @item hot-bb-count-ws-permille
11385 A basic block profile count is considered hot if it contributes to
11386 the given permillage (i.e.@: 0...1000) of the entire profiled execution.
11388 @item hot-bb-frequency-fraction
11389 Select fraction of the entry block frequency of executions of basic block in
11390 function given basic block needs to have to be considered hot.
11392 @item max-predicted-iterations
11393 The maximum number of loop iterations we predict statically. This is useful
11394 in cases where a function contains a single loop with known bound and
11395 another loop with unknown bound.
11396 The known number of iterations is predicted correctly, while
11397 the unknown number of iterations average to roughly 10. This means that the
11398 loop without bounds appears artificially cold relative to the other one.
11400 @item builtin-expect-probability
11401 Control the probability of the expression having the specified value. This
11402 parameter takes a percentage (i.e.@: 0 ... 100) as input.
11404 @item builtin-string-cmp-inline-length
11405 The maximum length of a constant string for a builtin string cmp call
11406 eligible for inlining.
11408 @item align-threshold
11410 Select fraction of the maximal frequency of executions of a basic block in
11411 a function to align the basic block.
11413 @item align-loop-iterations
11415 A loop expected to iterate at least the selected number of iterations is
11418 @item tracer-dynamic-coverage
11419 @itemx tracer-dynamic-coverage-feedback
11421 This value is used to limit superblock formation once the given percentage of
11422 executed instructions is covered. This limits unnecessary code size
11425 The @option{tracer-dynamic-coverage-feedback} parameter
11426 is used only when profile
11427 feedback is available. The real profiles (as opposed to statically estimated
11428 ones) are much less balanced allowing the threshold to be larger value.
11430 @item tracer-max-code-growth
11431 Stop tail duplication once code growth has reached given percentage. This is
11432 a rather artificial limit, as most of the duplicates are eliminated later in
11433 cross jumping, so it may be set to much higher values than is the desired code
11436 @item tracer-min-branch-ratio
11438 Stop reverse growth when the reverse probability of best edge is less than this
11439 threshold (in percent).
11441 @item tracer-min-branch-probability
11442 @itemx tracer-min-branch-probability-feedback
11444 Stop forward growth if the best edge has probability lower than this
11447 Similarly to @option{tracer-dynamic-coverage} two parameters are
11448 provided. @option{tracer-min-branch-probability-feedback} is used for
11449 compilation with profile feedback and @option{tracer-min-branch-probability}
11450 compilation without. The value for compilation with profile feedback
11451 needs to be more conservative (higher) in order to make tracer
11454 @item stack-clash-protection-guard-size
11455 Specify the size of the operating system provided stack guard as
11456 2 raised to @var{num} bytes. Higher values may reduce the
11457 number of explicit probes, but a value larger than the operating system
11458 provided guard will leave code vulnerable to stack clash style attacks.
11460 @item stack-clash-protection-probe-interval
11461 Stack clash protection involves probing stack space as it is allocated. This
11462 param controls the maximum distance between probes into the stack as 2 raised
11463 to @var{num} bytes. Higher values may reduce the number of explicit probes, but a value
11464 larger than the operating system provided guard will leave code vulnerable to
11465 stack clash style attacks.
11467 @item max-cse-path-length
11469 The maximum number of basic blocks on path that CSE considers.
11471 @item max-cse-insns
11472 The maximum number of instructions CSE processes before flushing.
11474 @item ggc-min-expand
11476 GCC uses a garbage collector to manage its own memory allocation. This
11477 parameter specifies the minimum percentage by which the garbage
11478 collector's heap should be allowed to expand between collections.
11479 Tuning this may improve compilation speed; it has no effect on code
11482 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
11483 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
11484 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
11485 GCC is not able to calculate RAM on a particular platform, the lower
11486 bound of 30% is used. Setting this parameter and
11487 @option{ggc-min-heapsize} to zero causes a full collection to occur at
11488 every opportunity. This is extremely slow, but can be useful for
11491 @item ggc-min-heapsize
11493 Minimum size of the garbage collector's heap before it begins bothering
11494 to collect garbage. The first collection occurs after the heap expands
11495 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
11496 tuning this may improve compilation speed, and has no effect on code
11499 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
11500 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
11501 with a lower bound of 4096 (four megabytes) and an upper bound of
11502 131072 (128 megabytes). If GCC is not able to calculate RAM on a
11503 particular platform, the lower bound is used. Setting this parameter
11504 very large effectively disables garbage collection. Setting this
11505 parameter and @option{ggc-min-expand} to zero causes a full collection
11506 to occur at every opportunity.
11508 @item max-reload-search-insns
11509 The maximum number of instruction reload should look backward for equivalent
11510 register. Increasing values mean more aggressive optimization, making the
11511 compilation time increase with probably slightly better performance.
11513 @item max-cselib-memory-locations
11514 The maximum number of memory locations cselib should take into account.
11515 Increasing values mean more aggressive optimization, making the compilation time
11516 increase with probably slightly better performance.
11518 @item max-sched-ready-insns
11519 The maximum number of instructions ready to be issued the scheduler should
11520 consider at any given time during the first scheduling pass. Increasing
11521 values mean more thorough searches, making the compilation time increase
11522 with probably little benefit.
11524 @item max-sched-region-blocks
11525 The maximum number of blocks in a region to be considered for
11526 interblock scheduling.
11528 @item max-pipeline-region-blocks
11529 The maximum number of blocks in a region to be considered for
11530 pipelining in the selective scheduler.
11532 @item max-sched-region-insns
11533 The maximum number of insns in a region to be considered for
11534 interblock scheduling.
11536 @item max-pipeline-region-insns
11537 The maximum number of insns in a region to be considered for
11538 pipelining in the selective scheduler.
11540 @item min-spec-prob
11541 The minimum probability (in percents) of reaching a source block
11542 for interblock speculative scheduling.
11544 @item max-sched-extend-regions-iters
11545 The maximum number of iterations through CFG to extend regions.
11546 A value of 0 disables region extensions.
11548 @item max-sched-insn-conflict-delay
11549 The maximum conflict delay for an insn to be considered for speculative motion.
11551 @item sched-spec-prob-cutoff
11552 The minimal probability of speculation success (in percents), so that
11553 speculative insns are scheduled.
11555 @item sched-state-edge-prob-cutoff
11556 The minimum probability an edge must have for the scheduler to save its
11559 @item sched-mem-true-dep-cost
11560 Minimal distance (in CPU cycles) between store and load targeting same
11563 @item selsched-max-lookahead
11564 The maximum size of the lookahead window of selective scheduling. It is a
11565 depth of search for available instructions.
11567 @item selsched-max-sched-times
11568 The maximum number of times that an instruction is scheduled during
11569 selective scheduling. This is the limit on the number of iterations
11570 through which the instruction may be pipelined.
11572 @item selsched-insns-to-rename
11573 The maximum number of best instructions in the ready list that are considered
11574 for renaming in the selective scheduler.
11577 The minimum value of stage count that swing modulo scheduler
11580 @item max-last-value-rtl
11581 The maximum size measured as number of RTLs that can be recorded in an expression
11582 in combiner for a pseudo register as last known value of that register.
11584 @item max-combine-insns
11585 The maximum number of instructions the RTL combiner tries to combine.
11587 @item integer-share-limit
11588 Small integer constants can use a shared data structure, reducing the
11589 compiler's memory usage and increasing its speed. This sets the maximum
11590 value of a shared integer constant.
11592 @item ssp-buffer-size
11593 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
11594 protection when @option{-fstack-protection} is used.
11596 @item min-size-for-stack-sharing
11597 The minimum size of variables taking part in stack slot sharing when not
11600 @item max-jump-thread-duplication-stmts
11601 Maximum number of statements allowed in a block that needs to be
11602 duplicated when threading jumps.
11604 @item max-fields-for-field-sensitive
11605 Maximum number of fields in a structure treated in
11606 a field sensitive manner during pointer analysis.
11608 @item prefetch-latency
11609 Estimate on average number of instructions that are executed before
11610 prefetch finishes. The distance prefetched ahead is proportional
11611 to this constant. Increasing this number may also lead to less
11612 streams being prefetched (see @option{simultaneous-prefetches}).
11614 @item simultaneous-prefetches
11615 Maximum number of prefetches that can run at the same time.
11617 @item l1-cache-line-size
11618 The size of cache line in L1 data cache, in bytes.
11620 @item l1-cache-size
11621 The size of L1 data cache, in kilobytes.
11623 @item l2-cache-size
11624 The size of L2 data cache, in kilobytes.
11626 @item prefetch-dynamic-strides
11627 Whether the loop array prefetch pass should issue software prefetch hints
11628 for strides that are non-constant. In some cases this may be
11629 beneficial, though the fact the stride is non-constant may make it
11630 hard to predict when there is clear benefit to issuing these hints.
11632 Set to 1 if the prefetch hints should be issued for non-constant
11633 strides. Set to 0 if prefetch hints should be issued only for strides that
11634 are known to be constant and below @option{prefetch-minimum-stride}.
11636 @item prefetch-minimum-stride
11637 Minimum constant stride, in bytes, to start using prefetch hints for. If
11638 the stride is less than this threshold, prefetch hints will not be issued.
11640 This setting is useful for processors that have hardware prefetchers, in
11641 which case there may be conflicts between the hardware prefetchers and
11642 the software prefetchers. If the hardware prefetchers have a maximum
11643 stride they can handle, it should be used here to improve the use of
11644 software prefetchers.
11646 A value of -1 means we don't have a threshold and therefore
11647 prefetch hints can be issued for any constant stride.
11649 This setting is only useful for strides that are known and constant.
11651 @item loop-interchange-max-num-stmts
11652 The maximum number of stmts in a loop to be interchanged.
11654 @item loop-interchange-stride-ratio
11655 The minimum ratio between stride of two loops for interchange to be profitable.
11657 @item min-insn-to-prefetch-ratio
11658 The minimum ratio between the number of instructions and the
11659 number of prefetches to enable prefetching in a loop.
11661 @item prefetch-min-insn-to-mem-ratio
11662 The minimum ratio between the number of instructions and the
11663 number of memory references to enable prefetching in a loop.
11665 @item use-canonical-types
11666 Whether the compiler should use the ``canonical'' type system.
11667 Should always be 1, which uses a more efficient internal
11668 mechanism for comparing types in C++ and Objective-C++. However, if
11669 bugs in the canonical type system are causing compilation failures,
11670 set this value to 0 to disable canonical types.
11672 @item switch-conversion-max-branch-ratio
11673 Switch initialization conversion refuses to create arrays that are
11674 bigger than @option{switch-conversion-max-branch-ratio} times the number of
11675 branches in the switch.
11677 @item max-partial-antic-length
11678 Maximum length of the partial antic set computed during the tree
11679 partial redundancy elimination optimization (@option{-ftree-pre}) when
11680 optimizing at @option{-O3} and above. For some sorts of source code
11681 the enhanced partial redundancy elimination optimization can run away,
11682 consuming all of the memory available on the host machine. This
11683 parameter sets a limit on the length of the sets that are computed,
11684 which prevents the runaway behavior. Setting a value of 0 for
11685 this parameter allows an unlimited set length.
11687 @item rpo-vn-max-loop-depth
11688 Maximum loop depth that is value-numbered optimistically.
11689 When the limit hits the innermost
11690 @var{rpo-vn-max-loop-depth} loops and the outermost loop in the
11691 loop nest are value-numbered optimistically and the remaining ones not.
11693 @item sccvn-max-alias-queries-per-access
11694 Maximum number of alias-oracle queries we perform when looking for
11695 redundancies for loads and stores. If this limit is hit the search
11696 is aborted and the load or store is not considered redundant. The
11697 number of queries is algorithmically limited to the number of
11698 stores on all paths from the load to the function entry.
11700 @item ira-max-loops-num
11701 IRA uses regional register allocation by default. If a function
11702 contains more loops than the number given by this parameter, only at most
11703 the given number of the most frequently-executed loops form regions
11704 for regional register allocation.
11706 @item ira-max-conflict-table-size
11707 Although IRA uses a sophisticated algorithm to compress the conflict
11708 table, the table can still require excessive amounts of memory for
11709 huge functions. If the conflict table for a function could be more
11710 than the size in MB given by this parameter, the register allocator
11711 instead uses a faster, simpler, and lower-quality
11712 algorithm that does not require building a pseudo-register conflict table.
11714 @item ira-loop-reserved-regs
11715 IRA can be used to evaluate more accurate register pressure in loops
11716 for decisions to move loop invariants (see @option{-O3}). The number
11717 of available registers reserved for some other purposes is given
11718 by this parameter. Default of the parameter
11719 is the best found from numerous experiments.
11721 @item lra-inheritance-ebb-probability-cutoff
11722 LRA tries to reuse values reloaded in registers in subsequent insns.
11723 This optimization is called inheritance. EBB is used as a region to
11724 do this optimization. The parameter defines a minimal fall-through
11725 edge probability in percentage used to add BB to inheritance EBB in
11726 LRA. The default value was chosen
11727 from numerous runs of SPEC2000 on x86-64.
11729 @item loop-invariant-max-bbs-in-loop
11730 Loop invariant motion can be very expensive, both in compilation time and
11731 in amount of needed compile-time memory, with very large loops. Loops
11732 with more basic blocks than this parameter won't have loop invariant
11733 motion optimization performed on them.
11735 @item loop-max-datarefs-for-datadeps
11736 Building data dependencies is expensive for very large loops. This
11737 parameter limits the number of data references in loops that are
11738 considered for data dependence analysis. These large loops are no
11739 handled by the optimizations using loop data dependencies.
11741 @item max-vartrack-size
11742 Sets a maximum number of hash table slots to use during variable
11743 tracking dataflow analysis of any function. If this limit is exceeded
11744 with variable tracking at assignments enabled, analysis for that
11745 function is retried without it, after removing all debug insns from
11746 the function. If the limit is exceeded even without debug insns, var
11747 tracking analysis is completely disabled for the function. Setting
11748 the parameter to zero makes it unlimited.
11750 @item max-vartrack-expr-depth
11751 Sets a maximum number of recursion levels when attempting to map
11752 variable names or debug temporaries to value expressions. This trades
11753 compilation time for more complete debug information. If this is set too
11754 low, value expressions that are available and could be represented in
11755 debug information may end up not being used; setting this higher may
11756 enable the compiler to find more complex debug expressions, but compile
11757 time and memory use may grow.
11759 @item max-debug-marker-count
11760 Sets a threshold on the number of debug markers (e.g.@: begin stmt
11761 markers) to avoid complexity explosion at inlining or expanding to RTL.
11762 If a function has more such gimple stmts than the set limit, such stmts
11763 will be dropped from the inlined copy of a function, and from its RTL
11766 @item min-nondebug-insn-uid
11767 Use uids starting at this parameter for nondebug insns. The range below
11768 the parameter is reserved exclusively for debug insns created by
11769 @option{-fvar-tracking-assignments}, but debug insns may get
11770 (non-overlapping) uids above it if the reserved range is exhausted.
11772 @item ipa-sra-ptr-growth-factor
11773 IPA-SRA replaces a pointer to an aggregate with one or more new
11774 parameters only when their cumulative size is less or equal to
11775 @option{ipa-sra-ptr-growth-factor} times the size of the original
11778 @item sra-max-scalarization-size-Ospeed
11779 @itemx sra-max-scalarization-size-Osize
11780 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
11781 replace scalar parts of aggregates with uses of independent scalar
11782 variables. These parameters control the maximum size, in storage units,
11783 of aggregate which is considered for replacement when compiling for
11785 (@option{sra-max-scalarization-size-Ospeed}) or size
11786 (@option{sra-max-scalarization-size-Osize}) respectively.
11788 @item tm-max-aggregate-size
11789 When making copies of thread-local variables in a transaction, this
11790 parameter specifies the size in bytes after which variables are
11791 saved with the logging functions as opposed to save/restore code
11792 sequence pairs. This option only applies when using
11795 @item graphite-max-nb-scop-params
11796 To avoid exponential effects in the Graphite loop transforms, the
11797 number of parameters in a Static Control Part (SCoP) is bounded.
11798 A value of zero can be used to lift
11799 the bound. A variable whose value is unknown at compilation time and
11800 defined outside a SCoP is a parameter of the SCoP.
11802 @item loop-block-tile-size
11803 Loop blocking or strip mining transforms, enabled with
11804 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
11805 loop in the loop nest by a given number of iterations. The strip
11806 length can be changed using the @option{loop-block-tile-size}
11809 @item ipa-cp-value-list-size
11810 IPA-CP attempts to track all possible values and types passed to a function's
11811 parameter in order to propagate them and perform devirtualization.
11812 @option{ipa-cp-value-list-size} is the maximum number of values and types it
11813 stores per one formal parameter of a function.
11815 @item ipa-cp-eval-threshold
11816 IPA-CP calculates its own score of cloning profitability heuristics
11817 and performs those cloning opportunities with scores that exceed
11818 @option{ipa-cp-eval-threshold}.
11820 @item ipa-cp-recursion-penalty
11821 Percentage penalty the recursive functions will receive when they
11822 are evaluated for cloning.
11824 @item ipa-cp-single-call-penalty
11825 Percentage penalty functions containing a single call to another
11826 function will receive when they are evaluated for cloning.
11828 @item ipa-max-agg-items
11829 IPA-CP is also capable to propagate a number of scalar values passed
11830 in an aggregate. @option{ipa-max-agg-items} controls the maximum
11831 number of such values per one parameter.
11833 @item ipa-cp-loop-hint-bonus
11834 When IPA-CP determines that a cloning candidate would make the number
11835 of iterations of a loop known, it adds a bonus of
11836 @option{ipa-cp-loop-hint-bonus} to the profitability score of
11839 @item ipa-cp-array-index-hint-bonus
11840 When IPA-CP determines that a cloning candidate would make the index of
11841 an array access known, it adds a bonus of
11842 @option{ipa-cp-array-index-hint-bonus} to the profitability
11843 score of the candidate.
11845 @item ipa-max-aa-steps
11846 During its analysis of function bodies, IPA-CP employs alias analysis
11847 in order to track values pointed to by function parameters. In order
11848 not spend too much time analyzing huge functions, it gives up and
11849 consider all memory clobbered after examining
11850 @option{ipa-max-aa-steps} statements modifying memory.
11852 @item lto-partitions
11853 Specify desired number of partitions produced during WHOPR compilation.
11854 The number of partitions should exceed the number of CPUs used for compilation.
11856 @item lto-min-partition
11857 Size of minimal partition for WHOPR (in estimated instructions).
11858 This prevents expenses of splitting very small programs into too many
11861 @item lto-max-partition
11862 Size of max partition for WHOPR (in estimated instructions).
11863 to provide an upper bound for individual size of partition.
11864 Meant to be used only with balanced partitioning.
11866 @item lto-max-streaming-parallelism
11867 Maximal number of parallel processes used for LTO streaming.
11869 @item cxx-max-namespaces-for-diagnostic-help
11870 The maximum number of namespaces to consult for suggestions when C++
11871 name lookup fails for an identifier.
11873 @item sink-frequency-threshold
11874 The maximum relative execution frequency (in percents) of the target block
11875 relative to a statement's original block to allow statement sinking of a
11876 statement. Larger numbers result in more aggressive statement sinking.
11877 A small positive adjustment is applied for
11878 statements with memory operands as those are even more profitable so sink.
11880 @item max-stores-to-sink
11881 The maximum number of conditional store pairs that can be sunk. Set to 0
11882 if either vectorization (@option{-ftree-vectorize}) or if-conversion
11883 (@option{-ftree-loop-if-convert}) is disabled.
11885 @item allow-store-data-races
11886 Allow optimizers to introduce new data races on stores.
11887 Set to 1 to allow, otherwise to 0.
11889 @item case-values-threshold
11890 The smallest number of different values for which it is best to use a
11891 jump-table instead of a tree of conditional branches. If the value is
11892 0, use the default for the machine.
11894 @item jump-table-max-growth-ratio-for-size
11895 The maximum code size growth ratio when expanding
11896 into a jump table (in percent). The parameter is used when
11897 optimizing for size.
11899 @item jump-table-max-growth-ratio-for-speed
11900 The maximum code size growth ratio when expanding
11901 into a jump table (in percent). The parameter is used when
11902 optimizing for speed.
11904 @item tree-reassoc-width
11905 Set the maximum number of instructions executed in parallel in
11906 reassociated tree. This parameter overrides target dependent
11907 heuristics used by default if has non zero value.
11909 @item sched-pressure-algorithm
11910 Choose between the two available implementations of
11911 @option{-fsched-pressure}. Algorithm 1 is the original implementation
11912 and is the more likely to prevent instructions from being reordered.
11913 Algorithm 2 was designed to be a compromise between the relatively
11914 conservative approach taken by algorithm 1 and the rather aggressive
11915 approach taken by the default scheduler. It relies more heavily on
11916 having a regular register file and accurate register pressure classes.
11917 See @file{haifa-sched.c} in the GCC sources for more details.
11919 The default choice depends on the target.
11921 @item max-slsr-cand-scan
11922 Set the maximum number of existing candidates that are considered when
11923 seeking a basis for a new straight-line strength reduction candidate.
11926 Enable buffer overflow detection for global objects. This kind
11927 of protection is enabled by default if you are using
11928 @option{-fsanitize=address} option.
11929 To disable global objects protection use @option{--param asan-globals=0}.
11932 Enable buffer overflow detection for stack objects. This kind of
11933 protection is enabled by default when using @option{-fsanitize=address}.
11934 To disable stack protection use @option{--param asan-stack=0} option.
11936 @item asan-instrument-reads
11937 Enable buffer overflow detection for memory reads. This kind of
11938 protection is enabled by default when using @option{-fsanitize=address}.
11939 To disable memory reads protection use
11940 @option{--param asan-instrument-reads=0}.
11942 @item asan-instrument-writes
11943 Enable buffer overflow detection for memory writes. This kind of
11944 protection is enabled by default when using @option{-fsanitize=address}.
11945 To disable memory writes protection use
11946 @option{--param asan-instrument-writes=0} option.
11948 @item asan-memintrin
11949 Enable detection for built-in functions. This kind of protection
11950 is enabled by default when using @option{-fsanitize=address}.
11951 To disable built-in functions protection use
11952 @option{--param asan-memintrin=0}.
11954 @item asan-use-after-return
11955 Enable detection of use-after-return. This kind of protection
11956 is enabled by default when using the @option{-fsanitize=address} option.
11957 To disable it use @option{--param asan-use-after-return=0}.
11959 Note: By default the check is disabled at run time. To enable it,
11960 add @code{detect_stack_use_after_return=1} to the environment variable
11961 @env{ASAN_OPTIONS}.
11963 @item asan-instrumentation-with-call-threshold
11964 If number of memory accesses in function being instrumented
11965 is greater or equal to this number, use callbacks instead of inline checks.
11966 E.g. to disable inline code use
11967 @option{--param asan-instrumentation-with-call-threshold=0}.
11969 @item use-after-scope-direct-emission-threshold
11970 If the size of a local variable in bytes is smaller or equal to this
11971 number, directly poison (or unpoison) shadow memory instead of using
11972 run-time callbacks.
11974 @item max-fsm-thread-path-insns
11975 Maximum number of instructions to copy when duplicating blocks on a
11976 finite state automaton jump thread path.
11978 @item max-fsm-thread-length
11979 Maximum number of basic blocks on a finite state automaton jump thread
11982 @item max-fsm-thread-paths
11983 Maximum number of new jump thread paths to create for a finite state
11986 @item parloops-chunk-size
11987 Chunk size of omp schedule for loops parallelized by parloops.
11989 @item parloops-schedule
11990 Schedule type of omp schedule for loops parallelized by parloops (static,
11991 dynamic, guided, auto, runtime).
11993 @item parloops-min-per-thread
11994 The minimum number of iterations per thread of an innermost parallelized
11995 loop for which the parallelized variant is preferred over the single threaded
11996 one. Note that for a parallelized loop nest the
11997 minimum number of iterations of the outermost loop per thread is two.
11999 @item max-ssa-name-query-depth
12000 Maximum depth of recursion when querying properties of SSA names in things
12001 like fold routines. One level of recursion corresponds to following a
12004 @item hsa-gen-debug-stores
12005 Enable emission of special debug stores within HSA kernels which are
12006 then read and reported by libgomp plugin. Generation of these stores
12007 is disabled by default, use @option{--param hsa-gen-debug-stores=1} to
12010 @item max-speculative-devirt-maydefs
12011 The maximum number of may-defs we analyze when looking for a must-def
12012 specifying the dynamic type of an object that invokes a virtual call
12013 we may be able to devirtualize speculatively.
12015 @item max-vrp-switch-assertions
12016 The maximum number of assertions to add along the default edge of a switch
12017 statement during VRP.
12019 @item unroll-jam-min-percent
12020 The minimum percentage of memory references that must be optimized
12021 away for the unroll-and-jam transformation to be considered profitable.
12023 @item unroll-jam-max-unroll
12024 The maximum number of times the outer loop should be unrolled by
12025 the unroll-and-jam transformation.
12027 @item max-rtl-if-conversion-unpredictable-cost
12028 Maximum permissible cost for the sequence that would be generated
12029 by the RTL if-conversion pass for a branch that is considered unpredictable.
12031 @item max-variable-expansions-in-unroller
12032 If @option{-fvariable-expansion-in-unroller} is used, the maximum number
12033 of times that an individual variable will be expanded during loop unrolling.
12035 @item tracer-min-branch-probability-feedback
12036 Stop forward growth if the probability of best edge is less than
12037 this threshold (in percent). Used when profile feedback is available.
12039 @item partial-inlining-entry-probability
12040 Maximum probability of the entry BB of split region
12041 (in percent relative to entry BB of the function)
12042 to make partial inlining happen.
12044 @item max-tracked-strlens
12045 Maximum number of strings for which strlen optimization pass will
12046 track string lengths.
12048 @item gcse-after-reload-partial-fraction
12049 The threshold ratio for performing partial redundancy
12050 elimination after reload.
12052 @item gcse-after-reload-critical-fraction
12053 The threshold ratio of critical edges execution count that
12054 permit performing redundancy elimination after reload.
12056 @item max-loop-header-insns
12057 The maximum number of insns in loop header duplicated
12058 by the copy loop headers pass.
12060 @item vect-epilogues-nomask
12061 Enable loop epilogue vectorization using smaller vector size.
12063 @item slp-max-insns-in-bb
12064 Maximum number of instructions in basic block to be
12065 considered for SLP vectorization.
12067 @item avoid-fma-max-bits
12068 Maximum number of bits for which we avoid creating FMAs.
12070 @item sms-loop-average-count-threshold
12071 A threshold on the average loop count considered by the swing modulo scheduler.
12073 @item sms-dfa-history
12074 The number of cycles the swing modulo scheduler considers when checking
12075 conflicts using DFA.
12077 @item hot-bb-count-fraction
12078 Select fraction of the maximal count of repetitions of basic block
12079 in program given basic block needs
12080 to have to be considered hot (used in non-LTO mode)
12082 @item max-inline-insns-recursive-auto
12083 The maximum number of instructions non-inline function
12084 can grow to via recursive inlining.
12086 @item graphite-allow-codegen-errors
12087 Whether codegen errors should be ICEs when @option{-fchecking}.
12089 @item sms-max-ii-factor
12090 A factor for tuning the upper bound that swing modulo scheduler
12091 uses for scheduling a loop.
12093 @item lra-max-considered-reload-pseudos
12094 The max number of reload pseudos which are considered during
12095 spilling a non-reload pseudo.
12097 @item max-pow-sqrt-depth
12098 Maximum depth of sqrt chains to use when synthesizing exponentiation
12099 by a real constant.
12101 @item max-dse-active-local-stores
12102 Maximum number of active local stores in RTL dead store elimination.
12104 @item asan-instrument-allocas
12105 Enable asan allocas/VLAs protection.
12107 @item max-iterations-computation-cost
12108 Bound on the cost of an expression to compute the number of iterations.
12110 @item max-isl-operations
12111 Maximum number of isl operations, 0 means unlimited.
12113 @item graphite-max-arrays-per-scop
12114 Maximum number of arrays per scop.
12116 @item max-vartrack-reverse-op-size
12117 Max. size of loc list for which reverse ops should be added.
12119 @item unlikely-bb-count-fraction
12120 The minimum fraction of profile runs a given basic block execution count
12121 must be not to be considered unlikely.
12123 @item tracer-dynamic-coverage-feedback
12124 The percentage of function, weighted by execution frequency,
12125 that must be covered by trace formation.
12126 Used when profile feedback is available.
12128 @item max-inline-recursive-depth-auto
12129 The maximum depth of recursive inlining for non-inline functions.
12131 @item fsm-scale-path-stmts
12132 Scale factor to apply to the number of statements in a threading path
12133 when comparing to the number of (scaled) blocks.
12135 @item fsm-maximum-phi-arguments
12136 Maximum number of arguments a PHI may have before the FSM threader
12137 will not try to thread through its block.
12139 @item uninit-control-dep-attempts
12140 Maximum number of nested calls to search for control dependencies
12141 during uninitialized variable analysis.
12143 @item max-once-peeled-insns
12144 The maximum number of insns of a peeled loop that rolls only once.
12146 @item sra-max-scalarization-size-Osize
12147 Maximum size, in storage units, of an aggregate
12148 which should be considered for scalarization when compiling for size.
12150 @item fsm-scale-path-blocks
12151 Scale factor to apply to the number of blocks in a threading path
12152 when comparing to the number of (scaled) statements.
12154 @item sched-autopref-queue-depth
12155 Hardware autoprefetcher scheduler model control flag.
12156 Number of lookahead cycles the model looks into; at '
12157 ' only enable instruction sorting heuristic.
12159 @item loop-versioning-max-inner-insns
12160 The maximum number of instructions that an inner loop can have
12161 before the loop versioning pass considers it too big to copy.
12163 @item loop-versioning-max-outer-insns
12164 The maximum number of instructions that an outer loop can have
12165 before the loop versioning pass considers it too big to copy,
12166 discounting any instructions in inner loops that directly benefit
12172 @node Instrumentation Options
12173 @section Program Instrumentation Options
12174 @cindex instrumentation options
12175 @cindex program instrumentation options
12176 @cindex run-time error checking options
12177 @cindex profiling options
12178 @cindex options, program instrumentation
12179 @cindex options, run-time error checking
12180 @cindex options, profiling
12182 GCC supports a number of command-line options that control adding
12183 run-time instrumentation to the code it normally generates.
12184 For example, one purpose of instrumentation is collect profiling
12185 statistics for use in finding program hot spots, code coverage
12186 analysis, or profile-guided optimizations.
12187 Another class of program instrumentation is adding run-time checking
12188 to detect programming errors like invalid pointer
12189 dereferences or out-of-bounds array accesses, as well as deliberately
12190 hostile attacks such as stack smashing or C++ vtable hijacking.
12191 There is also a general hook which can be used to implement other
12192 forms of tracing or function-level instrumentation for debug or
12193 program analysis purposes.
12196 @cindex @command{prof}
12197 @cindex @command{gprof}
12202 Generate extra code to write profile information suitable for the
12203 analysis program @command{prof} (for @option{-p}) or @command{gprof}
12204 (for @option{-pg}). You must use this option when compiling
12205 the source files you want data about, and you must also use it when
12208 You can use the function attribute @code{no_instrument_function} to
12209 suppress profiling of individual functions when compiling with these options.
12210 @xref{Common Function Attributes}.
12212 @item -fprofile-arcs
12213 @opindex fprofile-arcs
12214 Add code so that program flow @dfn{arcs} are instrumented. During
12215 execution the program records how many times each branch and call is
12216 executed and how many times it is taken or returns. On targets that support
12217 constructors with priority support, profiling properly handles constructors,
12218 destructors and C++ constructors (and destructors) of classes which are used
12219 as a type of a global variable.
12222 program exits it saves this data to a file called
12223 @file{@var{auxname}.gcda} for each source file. The data may be used for
12224 profile-directed optimizations (@option{-fbranch-probabilities}), or for
12225 test coverage analysis (@option{-ftest-coverage}). Each object file's
12226 @var{auxname} is generated from the name of the output file, if
12227 explicitly specified and it is not the final executable, otherwise it is
12228 the basename of the source file. In both cases any suffix is removed
12229 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
12230 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
12231 @xref{Cross-profiling}.
12233 @cindex @command{gcov}
12237 This option is used to compile and link code instrumented for coverage
12238 analysis. The option is a synonym for @option{-fprofile-arcs}
12239 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
12240 linking). See the documentation for those options for more details.
12245 Compile the source files with @option{-fprofile-arcs} plus optimization
12246 and code generation options. For test coverage analysis, use the
12247 additional @option{-ftest-coverage} option. You do not need to profile
12248 every source file in a program.
12251 Compile the source files additionally with @option{-fprofile-abs-path}
12252 to create absolute path names in the @file{.gcno} files. This allows
12253 @command{gcov} to find the correct sources in projects where compilations
12254 occur with different working directories.
12257 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
12258 (the latter implies the former).
12261 Run the program on a representative workload to generate the arc profile
12262 information. This may be repeated any number of times. You can run
12263 concurrent instances of your program, and provided that the file system
12264 supports locking, the data files will be correctly updated. Unless
12265 a strict ISO C dialect option is in effect, @code{fork} calls are
12266 detected and correctly handled without double counting.
12269 For profile-directed optimizations, compile the source files again with
12270 the same optimization and code generation options plus
12271 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
12272 Control Optimization}).
12275 For test coverage analysis, use @command{gcov} to produce human readable
12276 information from the @file{.gcno} and @file{.gcda} files. Refer to the
12277 @command{gcov} documentation for further information.
12281 With @option{-fprofile-arcs}, for each function of your program GCC
12282 creates a program flow graph, then finds a spanning tree for the graph.
12283 Only arcs that are not on the spanning tree have to be instrumented: the
12284 compiler adds code to count the number of times that these arcs are
12285 executed. When an arc is the only exit or only entrance to a block, the
12286 instrumentation code can be added to the block; otherwise, a new basic
12287 block must be created to hold the instrumentation code.
12290 @item -ftest-coverage
12291 @opindex ftest-coverage
12292 Produce a notes file that the @command{gcov} code-coverage utility
12293 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
12294 show program coverage. Each source file's note file is called
12295 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
12296 above for a description of @var{auxname} and instructions on how to
12297 generate test coverage data. Coverage data matches the source files
12298 more closely if you do not optimize.
12300 @item -fprofile-abs-path
12301 @opindex fprofile-abs-path
12302 Automatically convert relative source file names to absolute path names
12303 in the @file{.gcno} files. This allows @command{gcov} to find the correct
12304 sources in projects where compilations occur with different working
12307 @item -fprofile-dir=@var{path}
12308 @opindex fprofile-dir
12310 Set the directory to search for the profile data files in to @var{path}.
12311 This option affects only the profile data generated by
12312 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
12313 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
12314 and its related options. Both absolute and relative paths can be used.
12315 By default, GCC uses the current directory as @var{path}, thus the
12316 profile data file appears in the same directory as the object file.
12317 In order to prevent the file name clashing, if the object file name is
12318 not an absolute path, we mangle the absolute path of the
12319 @file{@var{sourcename}.gcda} file and use it as the file name of a
12322 When an executable is run in a massive parallel environment, it is recommended
12323 to save profile to different folders. That can be done with variables
12324 in @var{path} that are exported during run-time:
12332 value of environment variable @var{VAR}
12336 @item -fprofile-generate
12337 @itemx -fprofile-generate=@var{path}
12338 @opindex fprofile-generate
12340 Enable options usually used for instrumenting application to produce
12341 profile useful for later recompilation with profile feedback based
12342 optimization. You must use @option{-fprofile-generate} both when
12343 compiling and when linking your program.
12345 The following options are enabled:
12346 @option{-fprofile-arcs}, @option{-fprofile-values},
12347 @option{-finline-functions}, and @option{-fipa-bit-cp}.
12349 If @var{path} is specified, GCC looks at the @var{path} to find
12350 the profile feedback data files. See @option{-fprofile-dir}.
12352 To optimize the program based on the collected profile information, use
12353 @option{-fprofile-use}. @xref{Optimize Options}, for more information.
12355 @item -fprofile-update=@var{method}
12356 @opindex fprofile-update
12358 Alter the update method for an application instrumented for profile
12359 feedback based optimization. The @var{method} argument should be one of
12360 @samp{single}, @samp{atomic} or @samp{prefer-atomic}.
12361 The first one is useful for single-threaded applications,
12362 while the second one prevents profile corruption by emitting thread-safe code.
12364 @strong{Warning:} When an application does not properly join all threads
12365 (or creates an detached thread), a profile file can be still corrupted.
12367 Using @samp{prefer-atomic} would be transformed either to @samp{atomic},
12368 when supported by a target, or to @samp{single} otherwise. The GCC driver
12369 automatically selects @samp{prefer-atomic} when @option{-pthread}
12370 is present in the command line.
12372 @item -fprofile-filter-files=@var{regex}
12373 @opindex fprofile-filter-files
12375 Instrument only functions from files where names match
12376 any regular expression (separated by a semi-colon).
12378 For example, @option{-fprofile-filter-files=main.c;module.*.c} will instrument
12379 only @file{main.c} and all C files starting with 'module'.
12381 @item -fprofile-exclude-files=@var{regex}
12382 @opindex fprofile-exclude-files
12384 Instrument only functions from files where names do not match
12385 all the regular expressions (separated by a semi-colon).
12387 For example, @option{-fprofile-exclude-files=/usr/*} will prevent instrumentation
12388 of all files that are located in @file{/usr/} folder.
12390 @item -fsanitize=address
12391 @opindex fsanitize=address
12392 Enable AddressSanitizer, a fast memory error detector.
12393 Memory access instructions are instrumented to detect
12394 out-of-bounds and use-after-free bugs.
12395 The option enables @option{-fsanitize-address-use-after-scope}.
12396 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
12397 more details. The run-time behavior can be influenced using the
12398 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
12399 the available options are shown at startup of the instrumented program. See
12400 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
12401 for a list of supported options.
12402 The option cannot be combined with @option{-fsanitize=thread}.
12404 @item -fsanitize=kernel-address
12405 @opindex fsanitize=kernel-address
12406 Enable AddressSanitizer for Linux kernel.
12407 See @uref{https://github.com/google/kasan/wiki} for more details.
12409 @item -fsanitize=pointer-compare
12410 @opindex fsanitize=pointer-compare
12411 Instrument comparison operation (<, <=, >, >=) with pointer operands.
12412 The option must be combined with either @option{-fsanitize=kernel-address} or
12413 @option{-fsanitize=address}
12414 The option cannot be combined with @option{-fsanitize=thread}.
12415 Note: By default the check is disabled at run time. To enable it,
12416 add @code{detect_invalid_pointer_pairs=2} to the environment variable
12417 @env{ASAN_OPTIONS}. Using @code{detect_invalid_pointer_pairs=1} detects
12418 invalid operation only when both pointers are non-null.
12420 @item -fsanitize=pointer-subtract
12421 @opindex fsanitize=pointer-subtract
12422 Instrument subtraction with pointer operands.
12423 The option must be combined with either @option{-fsanitize=kernel-address} or
12424 @option{-fsanitize=address}
12425 The option cannot be combined with @option{-fsanitize=thread}.
12426 Note: By default the check is disabled at run time. To enable it,
12427 add @code{detect_invalid_pointer_pairs=2} to the environment variable
12428 @env{ASAN_OPTIONS}. Using @code{detect_invalid_pointer_pairs=1} detects
12429 invalid operation only when both pointers are non-null.
12431 @item -fsanitize=thread
12432 @opindex fsanitize=thread
12433 Enable ThreadSanitizer, a fast data race detector.
12434 Memory access instructions are instrumented to detect
12435 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
12436 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
12437 environment variable; see
12438 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
12440 The option cannot be combined with @option{-fsanitize=address},
12441 @option{-fsanitize=leak}.
12443 Note that sanitized atomic builtins cannot throw exceptions when
12444 operating on invalid memory addresses with non-call exceptions
12445 (@option{-fnon-call-exceptions}).
12447 @item -fsanitize=leak
12448 @opindex fsanitize=leak
12449 Enable LeakSanitizer, a memory leak detector.
12450 This option only matters for linking of executables and
12451 the executable is linked against a library that overrides @code{malloc}
12452 and other allocator functions. See
12453 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
12454 details. The run-time behavior can be influenced using the
12455 @env{LSAN_OPTIONS} environment variable.
12456 The option cannot be combined with @option{-fsanitize=thread}.
12458 @item -fsanitize=undefined
12459 @opindex fsanitize=undefined
12460 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
12461 Various computations are instrumented to detect undefined behavior
12462 at runtime. Current suboptions are:
12466 @item -fsanitize=shift
12467 @opindex fsanitize=shift
12468 This option enables checking that the result of a shift operation is
12469 not undefined. Note that what exactly is considered undefined differs
12470 slightly between C and C++, as well as between ISO C90 and C99, etc.
12471 This option has two suboptions, @option{-fsanitize=shift-base} and
12472 @option{-fsanitize=shift-exponent}.
12474 @item -fsanitize=shift-exponent
12475 @opindex fsanitize=shift-exponent
12476 This option enables checking that the second argument of a shift operation
12477 is not negative and is smaller than the precision of the promoted first
12480 @item -fsanitize=shift-base
12481 @opindex fsanitize=shift-base
12482 If the second argument of a shift operation is within range, check that the
12483 result of a shift operation is not undefined. Note that what exactly is
12484 considered undefined differs slightly between C and C++, as well as between
12485 ISO C90 and C99, etc.
12487 @item -fsanitize=integer-divide-by-zero
12488 @opindex fsanitize=integer-divide-by-zero
12489 Detect integer division by zero as well as @code{INT_MIN / -1} division.
12491 @item -fsanitize=unreachable
12492 @opindex fsanitize=unreachable
12493 With this option, the compiler turns the @code{__builtin_unreachable}
12494 call into a diagnostics message call instead. When reaching the
12495 @code{__builtin_unreachable} call, the behavior is undefined.
12497 @item -fsanitize=vla-bound
12498 @opindex fsanitize=vla-bound
12499 This option instructs the compiler to check that the size of a variable
12500 length array is positive.
12502 @item -fsanitize=null
12503 @opindex fsanitize=null
12504 This option enables pointer checking. Particularly, the application
12505 built with this option turned on will issue an error message when it
12506 tries to dereference a NULL pointer, or if a reference (possibly an
12507 rvalue reference) is bound to a NULL pointer, or if a method is invoked
12508 on an object pointed by a NULL pointer.
12510 @item -fsanitize=return
12511 @opindex fsanitize=return
12512 This option enables return statement checking. Programs
12513 built with this option turned on will issue an error message
12514 when the end of a non-void function is reached without actually
12515 returning a value. This option works in C++ only.
12517 @item -fsanitize=signed-integer-overflow
12518 @opindex fsanitize=signed-integer-overflow
12519 This option enables signed integer overflow checking. We check that
12520 the result of @code{+}, @code{*}, and both unary and binary @code{-}
12521 does not overflow in the signed arithmetics. Note, integer promotion
12522 rules must be taken into account. That is, the following is not an
12525 signed char a = SCHAR_MAX;
12529 @item -fsanitize=bounds
12530 @opindex fsanitize=bounds
12531 This option enables instrumentation of array bounds. Various out of bounds
12532 accesses are detected. Flexible array members, flexible array member-like
12533 arrays, and initializers of variables with static storage are not instrumented.
12535 @item -fsanitize=bounds-strict
12536 @opindex fsanitize=bounds-strict
12537 This option enables strict instrumentation of array bounds. Most out of bounds
12538 accesses are detected, including flexible array members and flexible array
12539 member-like arrays. Initializers of variables with static storage are not
12542 @item -fsanitize=alignment
12543 @opindex fsanitize=alignment
12545 This option enables checking of alignment of pointers when they are
12546 dereferenced, or when a reference is bound to insufficiently aligned target,
12547 or when a method or constructor is invoked on insufficiently aligned object.
12549 @item -fsanitize=object-size
12550 @opindex fsanitize=object-size
12551 This option enables instrumentation of memory references using the
12552 @code{__builtin_object_size} function. Various out of bounds pointer
12553 accesses are detected.
12555 @item -fsanitize=float-divide-by-zero
12556 @opindex fsanitize=float-divide-by-zero
12557 Detect floating-point division by zero. Unlike other similar options,
12558 @option{-fsanitize=float-divide-by-zero} is not enabled by
12559 @option{-fsanitize=undefined}, since floating-point division by zero can
12560 be a legitimate way of obtaining infinities and NaNs.
12562 @item -fsanitize=float-cast-overflow
12563 @opindex fsanitize=float-cast-overflow
12564 This option enables floating-point type to integer conversion checking.
12565 We check that the result of the conversion does not overflow.
12566 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
12567 not enabled by @option{-fsanitize=undefined}.
12568 This option does not work well with @code{FE_INVALID} exceptions enabled.
12570 @item -fsanitize=nonnull-attribute
12571 @opindex fsanitize=nonnull-attribute
12573 This option enables instrumentation of calls, checking whether null values
12574 are not passed to arguments marked as requiring a non-null value by the
12575 @code{nonnull} function attribute.
12577 @item -fsanitize=returns-nonnull-attribute
12578 @opindex fsanitize=returns-nonnull-attribute
12580 This option enables instrumentation of return statements in functions
12581 marked with @code{returns_nonnull} function attribute, to detect returning
12582 of null values from such functions.
12584 @item -fsanitize=bool
12585 @opindex fsanitize=bool
12587 This option enables instrumentation of loads from bool. If a value other
12588 than 0/1 is loaded, a run-time error is issued.
12590 @item -fsanitize=enum
12591 @opindex fsanitize=enum
12593 This option enables instrumentation of loads from an enum type. If
12594 a value outside the range of values for the enum type is loaded,
12595 a run-time error is issued.
12597 @item -fsanitize=vptr
12598 @opindex fsanitize=vptr
12600 This option enables instrumentation of C++ member function calls, member
12601 accesses and some conversions between pointers to base and derived classes,
12602 to verify the referenced object has the correct dynamic type.
12604 @item -fsanitize=pointer-overflow
12605 @opindex fsanitize=pointer-overflow
12607 This option enables instrumentation of pointer arithmetics. If the pointer
12608 arithmetics overflows, a run-time error is issued.
12610 @item -fsanitize=builtin
12611 @opindex fsanitize=builtin
12613 This option enables instrumentation of arguments to selected builtin
12614 functions. If an invalid value is passed to such arguments, a run-time
12615 error is issued. E.g.@ passing 0 as the argument to @code{__builtin_ctz}
12616 or @code{__builtin_clz} invokes undefined behavior and is diagnosed
12621 While @option{-ftrapv} causes traps for signed overflows to be emitted,
12622 @option{-fsanitize=undefined} gives a diagnostic message.
12623 This currently works only for the C family of languages.
12625 @item -fno-sanitize=all
12626 @opindex fno-sanitize=all
12628 This option disables all previously enabled sanitizers.
12629 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
12632 @item -fasan-shadow-offset=@var{number}
12633 @opindex fasan-shadow-offset
12634 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
12635 It is useful for experimenting with different shadow memory layouts in
12636 Kernel AddressSanitizer.
12638 @item -fsanitize-sections=@var{s1},@var{s2},...
12639 @opindex fsanitize-sections
12640 Sanitize global variables in selected user-defined sections. @var{si} may
12643 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
12644 @opindex fsanitize-recover
12645 @opindex fno-sanitize-recover
12646 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
12647 mentioned in comma-separated list of @var{opts}. Enabling this option
12648 for a sanitizer component causes it to attempt to continue
12649 running the program as if no error happened. This means multiple
12650 runtime errors can be reported in a single program run, and the exit
12651 code of the program may indicate success even when errors
12652 have been reported. The @option{-fno-sanitize-recover=} option
12653 can be used to alter
12654 this behavior: only the first detected error is reported
12655 and program then exits with a non-zero exit code.
12657 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
12658 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
12659 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
12660 @option{-fsanitize=bounds-strict},
12661 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
12662 For these sanitizers error recovery is turned on by default,
12663 except @option{-fsanitize=address}, for which this feature is experimental.
12664 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
12665 accepted, the former enables recovery for all sanitizers that support it,
12666 the latter disables recovery for all sanitizers that support it.
12668 Even if a recovery mode is turned on the compiler side, it needs to be also
12669 enabled on the runtime library side, otherwise the failures are still fatal.
12670 The runtime library defaults to @code{halt_on_error=0} for
12671 ThreadSanitizer and UndefinedBehaviorSanitizer, while default value for
12672 AddressSanitizer is @code{halt_on_error=1}. This can be overridden through
12673 setting the @code{halt_on_error} flag in the corresponding environment variable.
12675 Syntax without an explicit @var{opts} parameter is deprecated. It is
12676 equivalent to specifying an @var{opts} list of:
12679 undefined,float-cast-overflow,float-divide-by-zero,bounds-strict
12682 @item -fsanitize-address-use-after-scope
12683 @opindex fsanitize-address-use-after-scope
12684 Enable sanitization of local variables to detect use-after-scope bugs.
12685 The option sets @option{-fstack-reuse} to @samp{none}.
12687 @item -fsanitize-undefined-trap-on-error
12688 @opindex fsanitize-undefined-trap-on-error
12689 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
12690 report undefined behavior using @code{__builtin_trap} rather than
12691 a @code{libubsan} library routine. The advantage of this is that the
12692 @code{libubsan} library is not needed and is not linked in, so this
12693 is usable even in freestanding environments.
12695 @item -fsanitize-coverage=trace-pc
12696 @opindex fsanitize-coverage=trace-pc
12697 Enable coverage-guided fuzzing code instrumentation.
12698 Inserts a call to @code{__sanitizer_cov_trace_pc} into every basic block.
12700 @item -fsanitize-coverage=trace-cmp
12701 @opindex fsanitize-coverage=trace-cmp
12702 Enable dataflow guided fuzzing code instrumentation.
12703 Inserts a call to @code{__sanitizer_cov_trace_cmp1},
12704 @code{__sanitizer_cov_trace_cmp2}, @code{__sanitizer_cov_trace_cmp4} or
12705 @code{__sanitizer_cov_trace_cmp8} for integral comparison with both operands
12706 variable or @code{__sanitizer_cov_trace_const_cmp1},
12707 @code{__sanitizer_cov_trace_const_cmp2},
12708 @code{__sanitizer_cov_trace_const_cmp4} or
12709 @code{__sanitizer_cov_trace_const_cmp8} for integral comparison with one
12710 operand constant, @code{__sanitizer_cov_trace_cmpf} or
12711 @code{__sanitizer_cov_trace_cmpd} for float or double comparisons and
12712 @code{__sanitizer_cov_trace_switch} for switch statements.
12714 @item -fcf-protection=@r{[}full@r{|}branch@r{|}return@r{|}none@r{]}
12715 @opindex fcf-protection
12716 Enable code instrumentation of control-flow transfers to increase
12717 program security by checking that target addresses of control-flow
12718 transfer instructions (such as indirect function call, function return,
12719 indirect jump) are valid. This prevents diverting the flow of control
12720 to an unexpected target. This is intended to protect against such
12721 threats as Return-oriented Programming (ROP), and similarly
12722 call/jmp-oriented programming (COP/JOP).
12724 The value @code{branch} tells the compiler to implement checking of
12725 validity of control-flow transfer at the point of indirect branch
12726 instructions, i.e.@: call/jmp instructions. The value @code{return}
12727 implements checking of validity at the point of returning from a
12728 function. The value @code{full} is an alias for specifying both
12729 @code{branch} and @code{return}. The value @code{none} turns off
12732 The macro @code{__CET__} is defined when @option{-fcf-protection} is
12733 used. The first bit of @code{__CET__} is set to 1 for the value
12734 @code{branch} and the second bit of @code{__CET__} is set to 1 for
12737 You can also use the @code{nocf_check} attribute to identify
12738 which functions and calls should be skipped from instrumentation
12739 (@pxref{Function Attributes}).
12741 Currently the x86 GNU/Linux target provides an implementation based
12742 on Intel Control-flow Enforcement Technology (CET).
12744 @item -fstack-protector
12745 @opindex fstack-protector
12746 Emit extra code to check for buffer overflows, such as stack smashing
12747 attacks. This is done by adding a guard variable to functions with
12748 vulnerable objects. This includes functions that call @code{alloca}, and
12749 functions with buffers larger than 8 bytes. The guards are initialized
12750 when a function is entered and then checked when the function exits.
12751 If a guard check fails, an error message is printed and the program exits.
12753 @item -fstack-protector-all
12754 @opindex fstack-protector-all
12755 Like @option{-fstack-protector} except that all functions are protected.
12757 @item -fstack-protector-strong
12758 @opindex fstack-protector-strong
12759 Like @option{-fstack-protector} but includes additional functions to
12760 be protected --- those that have local array definitions, or have
12761 references to local frame addresses.
12763 @item -fstack-protector-explicit
12764 @opindex fstack-protector-explicit
12765 Like @option{-fstack-protector} but only protects those functions which
12766 have the @code{stack_protect} attribute.
12768 @item -fstack-check
12769 @opindex fstack-check
12770 Generate code to verify that you do not go beyond the boundary of the
12771 stack. You should specify this flag if you are running in an
12772 environment with multiple threads, but you only rarely need to specify it in
12773 a single-threaded environment since stack overflow is automatically
12774 detected on nearly all systems if there is only one stack.
12776 Note that this switch does not actually cause checking to be done; the
12777 operating system or the language runtime must do that. The switch causes
12778 generation of code to ensure that they see the stack being extended.
12780 You can additionally specify a string parameter: @samp{no} means no
12781 checking, @samp{generic} means force the use of old-style checking,
12782 @samp{specific} means use the best checking method and is equivalent
12783 to bare @option{-fstack-check}.
12785 Old-style checking is a generic mechanism that requires no specific
12786 target support in the compiler but comes with the following drawbacks:
12790 Modified allocation strategy for large objects: they are always
12791 allocated dynamically if their size exceeds a fixed threshold. Note this
12792 may change the semantics of some code.
12795 Fixed limit on the size of the static frame of functions: when it is
12796 topped by a particular function, stack checking is not reliable and
12797 a warning is issued by the compiler.
12800 Inefficiency: because of both the modified allocation strategy and the
12801 generic implementation, code performance is hampered.
12804 Note that old-style stack checking is also the fallback method for
12805 @samp{specific} if no target support has been added in the compiler.
12807 @samp{-fstack-check=} is designed for Ada's needs to detect infinite recursion
12808 and stack overflows. @samp{specific} is an excellent choice when compiling
12809 Ada code. It is not generally sufficient to protect against stack-clash
12810 attacks. To protect against those you want @samp{-fstack-clash-protection}.
12812 @item -fstack-clash-protection
12813 @opindex fstack-clash-protection
12814 Generate code to prevent stack clash style attacks. When this option is
12815 enabled, the compiler will only allocate one page of stack space at a time
12816 and each page is accessed immediately after allocation. Thus, it prevents
12817 allocations from jumping over any stack guard page provided by the
12820 Most targets do not fully support stack clash protection. However, on
12821 those targets @option{-fstack-clash-protection} will protect dynamic stack
12822 allocations. @option{-fstack-clash-protection} may also provide limited
12823 protection for static stack allocations if the target supports
12824 @option{-fstack-check=specific}.
12826 @item -fstack-limit-register=@var{reg}
12827 @itemx -fstack-limit-symbol=@var{sym}
12828 @itemx -fno-stack-limit
12829 @opindex fstack-limit-register
12830 @opindex fstack-limit-symbol
12831 @opindex fno-stack-limit
12832 Generate code to ensure that the stack does not grow beyond a certain value,
12833 either the value of a register or the address of a symbol. If a larger
12834 stack is required, a signal is raised at run time. For most targets,
12835 the signal is raised before the stack overruns the boundary, so
12836 it is possible to catch the signal without taking special precautions.
12838 For instance, if the stack starts at absolute address @samp{0x80000000}
12839 and grows downwards, you can use the flags
12840 @option{-fstack-limit-symbol=__stack_limit} and
12841 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
12842 of 128KB@. Note that this may only work with the GNU linker.
12844 You can locally override stack limit checking by using the
12845 @code{no_stack_limit} function attribute (@pxref{Function Attributes}).
12847 @item -fsplit-stack
12848 @opindex fsplit-stack
12849 Generate code to automatically split the stack before it overflows.
12850 The resulting program has a discontiguous stack which can only
12851 overflow if the program is unable to allocate any more memory. This
12852 is most useful when running threaded programs, as it is no longer
12853 necessary to calculate a good stack size to use for each thread. This
12854 is currently only implemented for the x86 targets running
12857 When code compiled with @option{-fsplit-stack} calls code compiled
12858 without @option{-fsplit-stack}, there may not be much stack space
12859 available for the latter code to run. If compiling all code,
12860 including library code, with @option{-fsplit-stack} is not an option,
12861 then the linker can fix up these calls so that the code compiled
12862 without @option{-fsplit-stack} always has a large stack. Support for
12863 this is implemented in the gold linker in GNU binutils release 2.21
12866 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
12867 @opindex fvtable-verify
12868 This option is only available when compiling C++ code.
12869 It turns on (or off, if using @option{-fvtable-verify=none}) the security
12870 feature that verifies at run time, for every virtual call, that
12871 the vtable pointer through which the call is made is valid for the type of
12872 the object, and has not been corrupted or overwritten. If an invalid vtable
12873 pointer is detected at run time, an error is reported and execution of the
12874 program is immediately halted.
12876 This option causes run-time data structures to be built at program startup,
12877 which are used for verifying the vtable pointers.
12878 The options @samp{std} and @samp{preinit}
12879 control the timing of when these data structures are built. In both cases the
12880 data structures are built before execution reaches @code{main}. Using
12881 @option{-fvtable-verify=std} causes the data structures to be built after
12882 shared libraries have been loaded and initialized.
12883 @option{-fvtable-verify=preinit} causes them to be built before shared
12884 libraries have been loaded and initialized.
12886 If this option appears multiple times in the command line with different
12887 values specified, @samp{none} takes highest priority over both @samp{std} and
12888 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
12891 @opindex fvtv-debug
12892 When used in conjunction with @option{-fvtable-verify=std} or
12893 @option{-fvtable-verify=preinit}, causes debug versions of the
12894 runtime functions for the vtable verification feature to be called.
12895 This flag also causes the compiler to log information about which
12896 vtable pointers it finds for each class.
12897 This information is written to a file named @file{vtv_set_ptr_data.log}
12898 in the directory named by the environment variable @env{VTV_LOGS_DIR}
12899 if that is defined or the current working directory otherwise.
12901 Note: This feature @emph{appends} data to the log file. If you want a fresh log
12902 file, be sure to delete any existing one.
12905 @opindex fvtv-counts
12906 This is a debugging flag. When used in conjunction with
12907 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
12908 causes the compiler to keep track of the total number of virtual calls
12909 it encounters and the number of verifications it inserts. It also
12910 counts the number of calls to certain run-time library functions
12911 that it inserts and logs this information for each compilation unit.
12912 The compiler writes this information to a file named
12913 @file{vtv_count_data.log} in the directory named by the environment
12914 variable @env{VTV_LOGS_DIR} if that is defined or the current working
12915 directory otherwise. It also counts the size of the vtable pointer sets
12916 for each class, and writes this information to @file{vtv_class_set_sizes.log}
12917 in the same directory.
12919 Note: This feature @emph{appends} data to the log files. To get fresh log
12920 files, be sure to delete any existing ones.
12922 @item -finstrument-functions
12923 @opindex finstrument-functions
12924 Generate instrumentation calls for entry and exit to functions. Just
12925 after function entry and just before function exit, the following
12926 profiling functions are called with the address of the current
12927 function and its call site. (On some platforms,
12928 @code{__builtin_return_address} does not work beyond the current
12929 function, so the call site information may not be available to the
12930 profiling functions otherwise.)
12933 void __cyg_profile_func_enter (void *this_fn,
12935 void __cyg_profile_func_exit (void *this_fn,
12939 The first argument is the address of the start of the current function,
12940 which may be looked up exactly in the symbol table.
12942 This instrumentation is also done for functions expanded inline in other
12943 functions. The profiling calls indicate where, conceptually, the
12944 inline function is entered and exited. This means that addressable
12945 versions of such functions must be available. If all your uses of a
12946 function are expanded inline, this may mean an additional expansion of
12947 code size. If you use @code{extern inline} in your C code, an
12948 addressable version of such functions must be provided. (This is
12949 normally the case anyway, but if you get lucky and the optimizer always
12950 expands the functions inline, you might have gotten away without
12951 providing static copies.)
12953 A function may be given the attribute @code{no_instrument_function}, in
12954 which case this instrumentation is not done. This can be used, for
12955 example, for the profiling functions listed above, high-priority
12956 interrupt routines, and any functions from which the profiling functions
12957 cannot safely be called (perhaps signal handlers, if the profiling
12958 routines generate output or allocate memory).
12959 @xref{Common Function Attributes}.
12961 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
12962 @opindex finstrument-functions-exclude-file-list
12964 Set the list of functions that are excluded from instrumentation (see
12965 the description of @option{-finstrument-functions}). If the file that
12966 contains a function definition matches with one of @var{file}, then
12967 that function is not instrumented. The match is done on substrings:
12968 if the @var{file} parameter is a substring of the file name, it is
12969 considered to be a match.
12974 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
12978 excludes any inline function defined in files whose pathnames
12979 contain @file{/bits/stl} or @file{include/sys}.
12981 If, for some reason, you want to include letter @samp{,} in one of
12982 @var{sym}, write @samp{\,}. For example,
12983 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
12984 (note the single quote surrounding the option).
12986 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
12987 @opindex finstrument-functions-exclude-function-list
12989 This is similar to @option{-finstrument-functions-exclude-file-list},
12990 but this option sets the list of function names to be excluded from
12991 instrumentation. The function name to be matched is its user-visible
12992 name, such as @code{vector<int> blah(const vector<int> &)}, not the
12993 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
12994 match is done on substrings: if the @var{sym} parameter is a substring
12995 of the function name, it is considered to be a match. For C99 and C++
12996 extended identifiers, the function name must be given in UTF-8, not
12997 using universal character names.
12999 @item -fpatchable-function-entry=@var{N}[,@var{M}]
13000 @opindex fpatchable-function-entry
13001 Generate @var{N} NOPs right at the beginning
13002 of each function, with the function entry point before the @var{M}th NOP.
13003 If @var{M} is omitted, it defaults to @code{0} so the
13004 function entry points to the address just at the first NOP.
13005 The NOP instructions reserve extra space which can be used to patch in
13006 any desired instrumentation at run time, provided that the code segment
13007 is writable. The amount of space is controllable indirectly via
13008 the number of NOPs; the NOP instruction used corresponds to the instruction
13009 emitted by the internal GCC back-end interface @code{gen_nop}. This behavior
13010 is target-specific and may also depend on the architecture variant and/or
13011 other compilation options.
13013 For run-time identification, the starting addresses of these areas,
13014 which correspond to their respective function entries minus @var{M},
13015 are additionally collected in the @code{__patchable_function_entries}
13016 section of the resulting binary.
13018 Note that the value of @code{__attribute__ ((patchable_function_entry
13019 (N,M)))} takes precedence over command-line option
13020 @option{-fpatchable-function-entry=N,M}. This can be used to increase
13021 the area size or to remove it completely on a single function.
13022 If @code{N=0}, no pad location is recorded.
13024 The NOP instructions are inserted at---and maybe before, depending on
13025 @var{M}---the function entry address, even before the prologue.
13030 @node Preprocessor Options
13031 @section Options Controlling the Preprocessor
13032 @cindex preprocessor options
13033 @cindex options, preprocessor
13035 These options control the C preprocessor, which is run on each C source
13036 file before actual compilation.
13038 If you use the @option{-E} option, nothing is done except preprocessing.
13039 Some of these options make sense only together with @option{-E} because
13040 they cause the preprocessor output to be unsuitable for actual
13043 In addition to the options listed here, there are a number of options
13044 to control search paths for include files documented in
13045 @ref{Directory Options}.
13046 Options to control preprocessor diagnostics are listed in
13047 @ref{Warning Options}.
13050 @include cppopts.texi
13052 @item -Wp,@var{option}
13054 You can use @option{-Wp,@var{option}} to bypass the compiler driver
13055 and pass @var{option} directly through to the preprocessor. If
13056 @var{option} contains commas, it is split into multiple options at the
13057 commas. However, many options are modified, translated or interpreted
13058 by the compiler driver before being passed to the preprocessor, and
13059 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
13060 interface is undocumented and subject to change, so whenever possible
13061 you should avoid using @option{-Wp} and let the driver handle the
13064 @item -Xpreprocessor @var{option}
13065 @opindex Xpreprocessor
13066 Pass @var{option} as an option to the preprocessor. You can use this to
13067 supply system-specific preprocessor options that GCC does not
13070 If you want to pass an option that takes an argument, you must use
13071 @option{-Xpreprocessor} twice, once for the option and once for the argument.
13073 @item -no-integrated-cpp
13074 @opindex no-integrated-cpp
13075 Perform preprocessing as a separate pass before compilation.
13076 By default, GCC performs preprocessing as an integrated part of
13077 input tokenization and parsing.
13078 If this option is provided, the appropriate language front end
13079 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
13080 and Objective-C, respectively) is instead invoked twice,
13081 once for preprocessing only and once for actual compilation
13082 of the preprocessed input.
13083 This option may be useful in conjunction with the @option{-B} or
13084 @option{-wrapper} options to specify an alternate preprocessor or
13085 perform additional processing of the program source between
13086 normal preprocessing and compilation.
13090 @node Assembler Options
13091 @section Passing Options to the Assembler
13093 @c prevent bad page break with this line
13094 You can pass options to the assembler.
13097 @item -Wa,@var{option}
13099 Pass @var{option} as an option to the assembler. If @var{option}
13100 contains commas, it is split into multiple options at the commas.
13102 @item -Xassembler @var{option}
13103 @opindex Xassembler
13104 Pass @var{option} as an option to the assembler. You can use this to
13105 supply system-specific assembler options that GCC does not
13108 If you want to pass an option that takes an argument, you must use
13109 @option{-Xassembler} twice, once for the option and once for the argument.
13114 @section Options for Linking
13115 @cindex link options
13116 @cindex options, linking
13118 These options come into play when the compiler links object files into
13119 an executable output file. They are meaningless if the compiler is
13120 not doing a link step.
13124 @item @var{object-file-name}
13125 A file name that does not end in a special recognized suffix is
13126 considered to name an object file or library. (Object files are
13127 distinguished from libraries by the linker according to the file
13128 contents.) If linking is done, these object files are used as input
13137 If any of these options is used, then the linker is not run, and
13138 object file names should not be used as arguments. @xref{Overall
13141 @item -flinker-output=@var{type}
13142 @opindex flinker-output
13143 This option controls code generation of the link time optimizer. By
13144 default the linker output is automatically determined by the linker
13145 plugin. For debugging the compiler and if incremental linking with a
13146 non-LTO object file is desired, it may be useful to control the type
13149 If @var{type} is @samp{exec} code generation produces a static
13150 binary. In this case @option{-fpic} and @option{-fpie} are both
13153 If @var{type} is @samp{dyn} code generation produces a shared
13154 library. In this case @option{-fpic} or @option{-fPIC} is preserved,
13155 but not enabled automatically. This allows to build shared libraries
13156 without position independent code on architectures where this is
13157 possible, i.e.@: on x86.
13159 If @var{type} is @samp{pie} code generation produces an @option{-fpie}
13160 executable. This results in similar optimizations as @samp{exec}
13161 except that @option{-fpie} is not disabled if specified at compilation
13164 If @var{type} is @samp{rel} the compiler assumes that incremental linking is
13165 done. The sections containing intermediate code for link-time optimization are
13166 merged, pre-optimized, and output to the resulting object file. In addition, if
13167 @option{-ffat-lto-objects} is specified the binary code is produced for future
13168 non-LTO linking. The object file produced by incremental linking will be smaller
13169 than a static library produced from the same object files. At link time the
13170 result of incremental linking will also load faster to compiler than a static
13171 library assuming that the majority of objects in the library are used.
13173 Finally @samp{nolto-rel} configures the compiler for incremental linking where
13174 code generation is forced, a final binary is produced and the intermediate
13175 code for later link-time optimization is stripped. When multiple object files
13176 are linked together the resulting code will be optimized better than with
13177 link-time optimizations disabled (for example, cross-module inlining will
13178 happen), most of benefits of whole program optimizations are however lost.
13180 During the incremental link (by @option{-r}) the linker plugin will default to
13181 @option{rel}. With current interfaces to GNU Binutils it is however not
13182 possible to incrementally link LTO objects and non-LTO objects into a single
13183 mixed object file. In the case any of object files in incremental link cannot
13184 be used for link-time optimization the linker plugin will issue a warning and
13185 use @samp{nolto-rel}. To maintain the whole program optimization it is
13186 recommended to link such objects into static library instead. Alternatively it
13187 is possible to use H.J. Lu's binutils with support for mixed objects.
13190 @opindex fuse-ld=bfd
13191 Use the @command{bfd} linker instead of the default linker.
13193 @item -fuse-ld=gold
13194 @opindex fuse-ld=gold
13195 Use the @command{gold} linker instead of the default linker.
13198 @opindex fuse-ld=lld
13199 Use the LLVM @command{lld} linker instead of the default linker.
13202 @item -l@var{library}
13203 @itemx -l @var{library}
13205 Search the library named @var{library} when linking. (The second
13206 alternative with the library as a separate argument is only for
13207 POSIX compliance and is not recommended.)
13209 The @option{-l} option is passed directly to the linker by GCC. Refer
13210 to your linker documentation for exact details. The general
13211 description below applies to the GNU linker.
13213 The linker searches a standard list of directories for the library.
13214 The directories searched include several standard system directories
13215 plus any that you specify with @option{-L}.
13217 Static libraries are archives of object files, and have file names
13218 like @file{lib@var{library}.a}. Some targets also support shared
13219 libraries, which typically have names like @file{lib@var{library}.so}.
13220 If both static and shared libraries are found, the linker gives
13221 preference to linking with the shared library unless the
13222 @option{-static} option is used.
13224 It makes a difference where in the command you write this option; the
13225 linker searches and processes libraries and object files in the order they
13226 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
13227 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
13228 to functions in @samp{z}, those functions may not be loaded.
13232 You need this special case of the @option{-l} option in order to
13233 link an Objective-C or Objective-C++ program.
13235 @item -nostartfiles
13236 @opindex nostartfiles
13237 Do not use the standard system startup files when linking.
13238 The standard system libraries are used normally, unless @option{-nostdlib},
13239 @option{-nolibc}, or @option{-nodefaultlibs} is used.
13241 @item -nodefaultlibs
13242 @opindex nodefaultlibs
13243 Do not use the standard system libraries when linking.
13244 Only the libraries you specify are passed to the linker, and options
13245 specifying linkage of the system libraries, such as @option{-static-libgcc}
13246 or @option{-shared-libgcc}, are ignored.
13247 The standard startup files are used normally, unless @option{-nostartfiles}
13250 The compiler may generate calls to @code{memcmp},
13251 @code{memset}, @code{memcpy} and @code{memmove}.
13252 These entries are usually resolved by entries in
13253 libc. These entry points should be supplied through some other
13254 mechanism when this option is specified.
13258 Do not use the C library or system libraries tightly coupled with it when
13259 linking. Still link with the startup files, @file{libgcc} or toolchain
13260 provided language support libraries such as @file{libgnat}, @file{libgfortran}
13261 or @file{libstdc++} unless options preventing their inclusion are used as
13262 well. This typically removes @option{-lc} from the link command line, as well
13263 as system libraries that normally go with it and become meaningless when
13264 absence of a C library is assumed, for example @option{-lpthread} or
13265 @option{-lm} in some configurations. This is intended for bare-board
13266 targets when there is indeed no C library available.
13270 Do not use the standard system startup files or libraries when linking.
13271 No startup files and only the libraries you specify are passed to
13272 the linker, and options specifying linkage of the system libraries, such as
13273 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
13275 The compiler may generate calls to @code{memcmp}, @code{memset},
13276 @code{memcpy} and @code{memmove}.
13277 These entries are usually resolved by entries in
13278 libc. These entry points should be supplied through some other
13279 mechanism when this option is specified.
13281 @cindex @option{-lgcc}, use with @option{-nostdlib}
13282 @cindex @option{-nostdlib} and unresolved references
13283 @cindex unresolved references and @option{-nostdlib}
13284 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
13285 @cindex @option{-nodefaultlibs} and unresolved references
13286 @cindex unresolved references and @option{-nodefaultlibs}
13287 One of the standard libraries bypassed by @option{-nostdlib} and
13288 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
13289 which GCC uses to overcome shortcomings of particular machines, or special
13290 needs for some languages.
13291 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
13292 Collection (GCC) Internals},
13293 for more discussion of @file{libgcc.a}.)
13294 In most cases, you need @file{libgcc.a} even when you want to avoid
13295 other standard libraries. In other words, when you specify @option{-nostdlib}
13296 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
13297 This ensures that you have no unresolved references to internal GCC
13298 library subroutines.
13299 (An example of such an internal subroutine is @code{__main}, used to ensure C++
13300 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
13301 GNU Compiler Collection (GCC) Internals}.)
13303 @item -e @var{entry}
13304 @itemx --entry=@var{entry}
13308 Specify that the program entry point is @var{entry}. The argument is
13309 interpreted by the linker; the GNU linker accepts either a symbol name
13314 Produce a dynamically linked position independent executable on targets
13315 that support it. For predictable results, you must also specify the same
13316 set of options used for compilation (@option{-fpie}, @option{-fPIE},
13317 or model suboptions) when you specify this linker option.
13321 Don't produce a dynamically linked position independent executable.
13324 @opindex static-pie
13325 Produce a static position independent executable on targets that support
13326 it. A static position independent executable is similar to a static
13327 executable, but can be loaded at any address without a dynamic linker.
13328 For predictable results, you must also specify the same set of options
13329 used for compilation (@option{-fpie}, @option{-fPIE}, or model
13330 suboptions) when you specify this linker option.
13334 Link with the POSIX threads library. This option is supported on
13335 GNU/Linux targets, most other Unix derivatives, and also on
13336 x86 Cygwin and MinGW targets. On some targets this option also sets
13337 flags for the preprocessor, so it should be used consistently for both
13338 compilation and linking.
13342 Produce a relocatable object as output. This is also known as partial
13347 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
13348 that support it. This instructs the linker to add all symbols, not
13349 only used ones, to the dynamic symbol table. This option is needed
13350 for some uses of @code{dlopen} or to allow obtaining backtraces
13351 from within a program.
13355 Remove all symbol table and relocation information from the executable.
13359 On systems that support dynamic linking, this overrides @option{-pie}
13360 and prevents linking with the shared libraries. On other systems, this
13361 option has no effect.
13365 Produce a shared object which can then be linked with other objects to
13366 form an executable. Not all systems support this option. For predictable
13367 results, you must also specify the same set of options used for compilation
13368 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
13369 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
13370 needs to build supplementary stub code for constructors to work. On
13371 multi-libbed systems, @samp{gcc -shared} must select the correct support
13372 libraries to link against. Failing to supply the correct flags may lead
13373 to subtle defects. Supplying them in cases where they are not necessary
13376 @item -shared-libgcc
13377 @itemx -static-libgcc
13378 @opindex shared-libgcc
13379 @opindex static-libgcc
13380 On systems that provide @file{libgcc} as a shared library, these options
13381 force the use of either the shared or static version, respectively.
13382 If no shared version of @file{libgcc} was built when the compiler was
13383 configured, these options have no effect.
13385 There are several situations in which an application should use the
13386 shared @file{libgcc} instead of the static version. The most common
13387 of these is when the application wishes to throw and catch exceptions
13388 across different shared libraries. In that case, each of the libraries
13389 as well as the application itself should use the shared @file{libgcc}.
13391 Therefore, the G++ driver automatically adds @option{-shared-libgcc}
13392 whenever you build a shared library or a main executable, because C++
13393 programs typically use exceptions, so this is the right thing to do.
13395 If, instead, you use the GCC driver to create shared libraries, you may
13396 find that they are not always linked with the shared @file{libgcc}.
13397 If GCC finds, at its configuration time, that you have a non-GNU linker
13398 or a GNU linker that does not support option @option{--eh-frame-hdr},
13399 it links the shared version of @file{libgcc} into shared libraries
13400 by default. Otherwise, it takes advantage of the linker and optimizes
13401 away the linking with the shared version of @file{libgcc}, linking with
13402 the static version of libgcc by default. This allows exceptions to
13403 propagate through such shared libraries, without incurring relocation
13404 costs at library load time.
13406 However, if a library or main executable is supposed to throw or catch
13407 exceptions, you must link it using the G++ driver, or using the option
13408 @option{-shared-libgcc}, such that it is linked with the shared
13411 @item -static-libasan
13412 @opindex static-libasan
13413 When the @option{-fsanitize=address} option is used to link a program,
13414 the GCC driver automatically links against @option{libasan}. If
13415 @file{libasan} is available as a shared library, and the @option{-static}
13416 option is not used, then this links against the shared version of
13417 @file{libasan}. The @option{-static-libasan} option directs the GCC
13418 driver to link @file{libasan} statically, without necessarily linking
13419 other libraries statically.
13421 @item -static-libtsan
13422 @opindex static-libtsan
13423 When the @option{-fsanitize=thread} option is used to link a program,
13424 the GCC driver automatically links against @option{libtsan}. If
13425 @file{libtsan} is available as a shared library, and the @option{-static}
13426 option is not used, then this links against the shared version of
13427 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
13428 driver to link @file{libtsan} statically, without necessarily linking
13429 other libraries statically.
13431 @item -static-liblsan
13432 @opindex static-liblsan
13433 When the @option{-fsanitize=leak} option is used to link a program,
13434 the GCC driver automatically links against @option{liblsan}. If
13435 @file{liblsan} is available as a shared library, and the @option{-static}
13436 option is not used, then this links against the shared version of
13437 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
13438 driver to link @file{liblsan} statically, without necessarily linking
13439 other libraries statically.
13441 @item -static-libubsan
13442 @opindex static-libubsan
13443 When the @option{-fsanitize=undefined} option is used to link a program,
13444 the GCC driver automatically links against @option{libubsan}. If
13445 @file{libubsan} is available as a shared library, and the @option{-static}
13446 option is not used, then this links against the shared version of
13447 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
13448 driver to link @file{libubsan} statically, without necessarily linking
13449 other libraries statically.
13451 @item -static-libstdc++
13452 @opindex static-libstdc++
13453 When the @command{g++} program is used to link a C++ program, it
13454 normally automatically links against @option{libstdc++}. If
13455 @file{libstdc++} is available as a shared library, and the
13456 @option{-static} option is not used, then this links against the
13457 shared version of @file{libstdc++}. That is normally fine. However, it
13458 is sometimes useful to freeze the version of @file{libstdc++} used by
13459 the program without going all the way to a fully static link. The
13460 @option{-static-libstdc++} option directs the @command{g++} driver to
13461 link @file{libstdc++} statically, without necessarily linking other
13462 libraries statically.
13466 Bind references to global symbols when building a shared object. Warn
13467 about any unresolved references (unless overridden by the link editor
13468 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
13471 @item -T @var{script}
13473 @cindex linker script
13474 Use @var{script} as the linker script. This option is supported by most
13475 systems using the GNU linker. On some targets, such as bare-board
13476 targets without an operating system, the @option{-T} option may be required
13477 when linking to avoid references to undefined symbols.
13479 @item -Xlinker @var{option}
13481 Pass @var{option} as an option to the linker. You can use this to
13482 supply system-specific linker options that GCC does not recognize.
13484 If you want to pass an option that takes a separate argument, you must use
13485 @option{-Xlinker} twice, once for the option and once for the argument.
13486 For example, to pass @option{-assert definitions}, you must write
13487 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
13488 @option{-Xlinker "-assert definitions"}, because this passes the entire
13489 string as a single argument, which is not what the linker expects.
13491 When using the GNU linker, it is usually more convenient to pass
13492 arguments to linker options using the @option{@var{option}=@var{value}}
13493 syntax than as separate arguments. For example, you can specify
13494 @option{-Xlinker -Map=output.map} rather than
13495 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
13496 this syntax for command-line options.
13498 @item -Wl,@var{option}
13500 Pass @var{option} as an option to the linker. If @var{option} contains
13501 commas, it is split into multiple options at the commas. You can use this
13502 syntax to pass an argument to the option.
13503 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
13504 linker. When using the GNU linker, you can also get the same effect with
13505 @option{-Wl,-Map=output.map}.
13507 @item -u @var{symbol}
13509 Pretend the symbol @var{symbol} is undefined, to force linking of
13510 library modules to define it. You can use @option{-u} multiple times with
13511 different symbols to force loading of additional library modules.
13513 @item -z @var{keyword}
13515 @option{-z} is passed directly on to the linker along with the keyword
13516 @var{keyword}. See the section in the documentation of your linker for
13517 permitted values and their meanings.
13520 @node Directory Options
13521 @section Options for Directory Search
13522 @cindex directory options
13523 @cindex options, directory search
13524 @cindex search path
13526 These options specify directories to search for header files, for
13527 libraries and for parts of the compiler:
13530 @include cppdiropts.texi
13532 @item -iplugindir=@var{dir}
13533 @opindex iplugindir=
13534 Set the directory to search for plugins that are passed
13535 by @option{-fplugin=@var{name}} instead of
13536 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
13537 to be used by the user, but only passed by the driver.
13541 Add directory @var{dir} to the list of directories to be searched
13544 @item -B@var{prefix}
13546 This option specifies where to find the executables, libraries,
13547 include files, and data files of the compiler itself.
13549 The compiler driver program runs one or more of the subprograms
13550 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
13551 @var{prefix} as a prefix for each program it tries to run, both with and
13552 without @samp{@var{machine}/@var{version}/} for the corresponding target
13553 machine and compiler version.
13555 For each subprogram to be run, the compiler driver first tries the
13556 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
13557 is not specified, the driver tries two standard prefixes,
13558 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
13559 those results in a file name that is found, the unmodified program
13560 name is searched for using the directories specified in your
13561 @env{PATH} environment variable.
13563 The compiler checks to see if the path provided by @option{-B}
13564 refers to a directory, and if necessary it adds a directory
13565 separator character at the end of the path.
13567 @option{-B} prefixes that effectively specify directory names also apply
13568 to libraries in the linker, because the compiler translates these
13569 options into @option{-L} options for the linker. They also apply to
13570 include files in the preprocessor, because the compiler translates these
13571 options into @option{-isystem} options for the preprocessor. In this case,
13572 the compiler appends @samp{include} to the prefix.
13574 The runtime support file @file{libgcc.a} can also be searched for using
13575 the @option{-B} prefix, if needed. If it is not found there, the two
13576 standard prefixes above are tried, and that is all. The file is left
13577 out of the link if it is not found by those means.
13579 Another way to specify a prefix much like the @option{-B} prefix is to use
13580 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
13583 As a special kludge, if the path provided by @option{-B} is
13584 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
13585 9, then it is replaced by @file{[dir/]include}. This is to help
13586 with boot-strapping the compiler.
13588 @item -no-canonical-prefixes
13589 @opindex no-canonical-prefixes
13590 Do not expand any symbolic links, resolve references to @samp{/../}
13591 or @samp{/./}, or make the path absolute when generating a relative
13594 @item --sysroot=@var{dir}
13596 Use @var{dir} as the logical root directory for headers and libraries.
13597 For example, if the compiler normally searches for headers in
13598 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
13599 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
13601 If you use both this option and the @option{-isysroot} option, then
13602 the @option{--sysroot} option applies to libraries, but the
13603 @option{-isysroot} option applies to header files.
13605 The GNU linker (beginning with version 2.16) has the necessary support
13606 for this option. If your linker does not support this option, the
13607 header file aspect of @option{--sysroot} still works, but the
13608 library aspect does not.
13610 @item --no-sysroot-suffix
13611 @opindex no-sysroot-suffix
13612 For some targets, a suffix is added to the root directory specified
13613 with @option{--sysroot}, depending on the other options used, so that
13614 headers may for example be found in
13615 @file{@var{dir}/@var{suffix}/usr/include} instead of
13616 @file{@var{dir}/usr/include}. This option disables the addition of
13621 @node Code Gen Options
13622 @section Options for Code Generation Conventions
13623 @cindex code generation conventions
13624 @cindex options, code generation
13625 @cindex run-time options
13627 These machine-independent options control the interface conventions
13628 used in code generation.
13630 Most of them have both positive and negative forms; the negative form
13631 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
13632 one of the forms is listed---the one that is not the default. You
13633 can figure out the other form by either removing @samp{no-} or adding
13637 @item -fstack-reuse=@var{reuse-level}
13638 @opindex fstack_reuse
13639 This option controls stack space reuse for user declared local/auto variables
13640 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
13641 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
13642 local variables and temporaries, @samp{named_vars} enables the reuse only for
13643 user defined local variables with names, and @samp{none} disables stack reuse
13644 completely. The default value is @samp{all}. The option is needed when the
13645 program extends the lifetime of a scoped local variable or a compiler generated
13646 temporary beyond the end point defined by the language. When a lifetime of
13647 a variable ends, and if the variable lives in memory, the optimizing compiler
13648 has the freedom to reuse its stack space with other temporaries or scoped
13649 local variables whose live range does not overlap with it. Legacy code extending
13650 local lifetime is likely to break with the stack reuse optimization.
13669 if (*p == 10) // out of scope use of local1
13680 A(int k) : i(k), j(k) @{ @}
13687 void foo(const A& ar)
13694 foo(A(10)); // temp object's lifetime ends when foo returns
13700 ap->i+= 10; // ap references out of scope temp whose space
13701 // is reused with a. What is the value of ap->i?
13706 The lifetime of a compiler generated temporary is well defined by the C++
13707 standard. When a lifetime of a temporary ends, and if the temporary lives
13708 in memory, the optimizing compiler has the freedom to reuse its stack
13709 space with other temporaries or scoped local variables whose live range
13710 does not overlap with it. However some of the legacy code relies on
13711 the behavior of older compilers in which temporaries' stack space is
13712 not reused, the aggressive stack reuse can lead to runtime errors. This
13713 option is used to control the temporary stack reuse optimization.
13717 This option generates traps for signed overflow on addition, subtraction,
13718 multiplication operations.
13719 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
13720 @option{-ftrapv} @option{-fwrapv} on the command-line results in
13721 @option{-fwrapv} being effective. Note that only active options override, so
13722 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
13723 results in @option{-ftrapv} being effective.
13727 This option instructs the compiler to assume that signed arithmetic
13728 overflow of addition, subtraction and multiplication wraps around
13729 using twos-complement representation. This flag enables some optimizations
13730 and disables others.
13731 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
13732 @option{-ftrapv} @option{-fwrapv} on the command-line results in
13733 @option{-fwrapv} being effective. Note that only active options override, so
13734 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
13735 results in @option{-ftrapv} being effective.
13737 @item -fwrapv-pointer
13738 @opindex fwrapv-pointer
13739 This option instructs the compiler to assume that pointer arithmetic
13740 overflow on addition and subtraction wraps around using twos-complement
13741 representation. This flag disables some optimizations which assume
13742 pointer overflow is invalid.
13744 @item -fstrict-overflow
13745 @opindex fstrict-overflow
13746 This option implies @option{-fno-wrapv} @option{-fno-wrapv-pointer} and when
13747 negated implies @option{-fwrapv} @option{-fwrapv-pointer}.
13750 @opindex fexceptions
13751 Enable exception handling. Generates extra code needed to propagate
13752 exceptions. For some targets, this implies GCC generates frame
13753 unwind information for all functions, which can produce significant data
13754 size overhead, although it does not affect execution. If you do not
13755 specify this option, GCC enables it by default for languages like
13756 C++ that normally require exception handling, and disables it for
13757 languages like C that do not normally require it. However, you may need
13758 to enable this option when compiling C code that needs to interoperate
13759 properly with exception handlers written in C++. You may also wish to
13760 disable this option if you are compiling older C++ programs that don't
13761 use exception handling.
13763 @item -fnon-call-exceptions
13764 @opindex fnon-call-exceptions
13765 Generate code that allows trapping instructions to throw exceptions.
13766 Note that this requires platform-specific runtime support that does
13767 not exist everywhere. Moreover, it only allows @emph{trapping}
13768 instructions to throw exceptions, i.e.@: memory references or floating-point
13769 instructions. It does not allow exceptions to be thrown from
13770 arbitrary signal handlers such as @code{SIGALRM}.
13772 @item -fdelete-dead-exceptions
13773 @opindex fdelete-dead-exceptions
13774 Consider that instructions that may throw exceptions but don't otherwise
13775 contribute to the execution of the program can be optimized away.
13776 This option is enabled by default for the Ada front end, as permitted by
13777 the Ada language specification.
13778 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
13780 @item -funwind-tables
13781 @opindex funwind-tables
13782 Similar to @option{-fexceptions}, except that it just generates any needed
13783 static data, but does not affect the generated code in any other way.
13784 You normally do not need to enable this option; instead, a language processor
13785 that needs this handling enables it on your behalf.
13787 @item -fasynchronous-unwind-tables
13788 @opindex fasynchronous-unwind-tables
13789 Generate unwind table in DWARF format, if supported by target machine. The
13790 table is exact at each instruction boundary, so it can be used for stack
13791 unwinding from asynchronous events (such as debugger or garbage collector).
13793 @item -fno-gnu-unique
13794 @opindex fno-gnu-unique
13795 @opindex fgnu-unique
13796 On systems with recent GNU assembler and C library, the C++ compiler
13797 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
13798 of template static data members and static local variables in inline
13799 functions are unique even in the presence of @code{RTLD_LOCAL}; this
13800 is necessary to avoid problems with a library used by two different
13801 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
13802 therefore disagreeing with the other one about the binding of the
13803 symbol. But this causes @code{dlclose} to be ignored for affected
13804 DSOs; if your program relies on reinitialization of a DSO via
13805 @code{dlclose} and @code{dlopen}, you can use
13806 @option{-fno-gnu-unique}.
13808 @item -fpcc-struct-return
13809 @opindex fpcc-struct-return
13810 Return ``short'' @code{struct} and @code{union} values in memory like
13811 longer ones, rather than in registers. This convention is less
13812 efficient, but it has the advantage of allowing intercallability between
13813 GCC-compiled files and files compiled with other compilers, particularly
13814 the Portable C Compiler (pcc).
13816 The precise convention for returning structures in memory depends
13817 on the target configuration macros.
13819 Short structures and unions are those whose size and alignment match
13820 that of some integer type.
13822 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
13823 switch is not binary compatible with code compiled with the
13824 @option{-freg-struct-return} switch.
13825 Use it to conform to a non-default application binary interface.
13827 @item -freg-struct-return
13828 @opindex freg-struct-return
13829 Return @code{struct} and @code{union} values in registers when possible.
13830 This is more efficient for small structures than
13831 @option{-fpcc-struct-return}.
13833 If you specify neither @option{-fpcc-struct-return} nor
13834 @option{-freg-struct-return}, GCC defaults to whichever convention is
13835 standard for the target. If there is no standard convention, GCC
13836 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
13837 the principal compiler. In those cases, we can choose the standard, and
13838 we chose the more efficient register return alternative.
13840 @strong{Warning:} code compiled with the @option{-freg-struct-return}
13841 switch is not binary compatible with code compiled with the
13842 @option{-fpcc-struct-return} switch.
13843 Use it to conform to a non-default application binary interface.
13845 @item -fshort-enums
13846 @opindex fshort-enums
13847 Allocate to an @code{enum} type only as many bytes as it needs for the
13848 declared range of possible values. Specifically, the @code{enum} type
13849 is equivalent to the smallest integer type that has enough room.
13851 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
13852 code that is not binary compatible with code generated without that switch.
13853 Use it to conform to a non-default application binary interface.
13855 @item -fshort-wchar
13856 @opindex fshort-wchar
13857 Override the underlying type for @code{wchar_t} to be @code{short
13858 unsigned int} instead of the default for the target. This option is
13859 useful for building programs to run under WINE@.
13861 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
13862 code that is not binary compatible with code generated without that switch.
13863 Use it to conform to a non-default application binary interface.
13866 @opindex fno-common
13868 @cindex tentative definitions
13869 In C code, this option controls the placement of global variables
13870 defined without an initializer, known as @dfn{tentative definitions}
13871 in the C standard. Tentative definitions are distinct from declarations
13872 of a variable with the @code{extern} keyword, which do not allocate storage.
13874 Unix C compilers have traditionally allocated storage for
13875 uninitialized global variables in a common block. This allows the
13876 linker to resolve all tentative definitions of the same variable
13877 in different compilation units to the same object, or to a non-tentative
13879 This is the behavior specified by @option{-fcommon}, and is the default for
13880 GCC on most targets.
13881 On the other hand, this behavior is not required by ISO
13882 C, and on some targets may carry a speed or code size penalty on
13883 variable references.
13885 The @option{-fno-common} option specifies that the compiler should instead
13886 place uninitialized global variables in the BSS section of the object file.
13887 This inhibits the merging of tentative definitions by the linker so
13888 you get a multiple-definition error if the same
13889 variable is defined in more than one compilation unit.
13890 Compiling with @option{-fno-common} is useful on targets for which
13891 it provides better performance, or if you wish to verify that the
13892 program will work on other systems that always treat uninitialized
13893 variable definitions this way.
13898 Ignore the @code{#ident} directive.
13900 @item -finhibit-size-directive
13901 @opindex finhibit-size-directive
13902 Don't output a @code{.size} assembler directive, or anything else that
13903 would cause trouble if the function is split in the middle, and the
13904 two halves are placed at locations far apart in memory. This option is
13905 used when compiling @file{crtstuff.c}; you should not need to use it
13908 @item -fverbose-asm
13909 @opindex fverbose-asm
13910 Put extra commentary information in the generated assembly code to
13911 make it more readable. This option is generally only of use to those
13912 who actually need to read the generated assembly code (perhaps while
13913 debugging the compiler itself).
13915 @option{-fno-verbose-asm}, the default, causes the
13916 extra information to be omitted and is useful when comparing two assembler
13919 The added comments include:
13924 information on the compiler version and command-line options,
13927 the source code lines associated with the assembly instructions,
13928 in the form FILENAME:LINENUMBER:CONTENT OF LINE,
13931 hints on which high-level expressions correspond to
13932 the various assembly instruction operands.
13936 For example, given this C source file:
13944 for (i = 0; i < n; i++)
13951 compiling to (x86_64) assembly via @option{-S} and emitting the result
13952 direct to stdout via @option{-o} @option{-}
13955 gcc -S test.c -fverbose-asm -Os -o -
13958 gives output similar to this:
13962 # GNU C11 (GCC) version 7.0.0 20160809 (experimental) (x86_64-pc-linux-gnu)
13969 .type test, @@function
13973 # test.c:4: int total = 0;
13974 xorl %eax, %eax # <retval>
13975 # test.c:6: for (i = 0; i < n; i++)
13976 xorl %edx, %edx # i
13978 # test.c:6: for (i = 0; i < n; i++)
13979 cmpl %edi, %edx # n, i
13981 # test.c:7: total += i * i;
13982 movl %edx, %ecx # i, tmp92
13983 imull %edx, %ecx # i, tmp92
13984 # test.c:6: for (i = 0; i < n; i++)
13986 # test.c:7: total += i * i;
13987 addl %ecx, %eax # tmp92, <retval>
13995 .ident "GCC: (GNU) 7.0.0 20160809 (experimental)"
13996 .section .note.GNU-stack,"",@@progbits
13999 The comments are intended for humans rather than machines and hence the
14000 precise format of the comments is subject to change.
14002 @item -frecord-gcc-switches
14003 @opindex frecord-gcc-switches
14004 This switch causes the command line used to invoke the
14005 compiler to be recorded into the object file that is being created.
14006 This switch is only implemented on some targets and the exact format
14007 of the recording is target and binary file format dependent, but it
14008 usually takes the form of a section containing ASCII text. This
14009 switch is related to the @option{-fverbose-asm} switch, but that
14010 switch only records information in the assembler output file as
14011 comments, so it never reaches the object file.
14012 See also @option{-grecord-gcc-switches} for another
14013 way of storing compiler options into the object file.
14017 @cindex global offset table
14019 Generate position-independent code (PIC) suitable for use in a shared
14020 library, if supported for the target machine. Such code accesses all
14021 constant addresses through a global offset table (GOT)@. The dynamic
14022 loader resolves the GOT entries when the program starts (the dynamic
14023 loader is not part of GCC; it is part of the operating system). If
14024 the GOT size for the linked executable exceeds a machine-specific
14025 maximum size, you get an error message from the linker indicating that
14026 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
14027 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
14028 on the m68k and RS/6000. The x86 has no such limit.)
14030 Position-independent code requires special support, and therefore works
14031 only on certain machines. For the x86, GCC supports PIC for System V
14032 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
14033 position-independent.
14035 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
14040 If supported for the target machine, emit position-independent code,
14041 suitable for dynamic linking and avoiding any limit on the size of the
14042 global offset table. This option makes a difference on AArch64, m68k,
14043 PowerPC and SPARC@.
14045 Position-independent code requires special support, and therefore works
14046 only on certain machines.
14048 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
14055 These options are similar to @option{-fpic} and @option{-fPIC}, but the
14056 generated position-independent code can be only linked into executables.
14057 Usually these options are used to compile code that will be linked using
14058 the @option{-pie} GCC option.
14060 @option{-fpie} and @option{-fPIE} both define the macros
14061 @code{__pie__} and @code{__PIE__}. The macros have the value 1
14062 for @option{-fpie} and 2 for @option{-fPIE}.
14067 Do not use the PLT for external function calls in position-independent code.
14068 Instead, load the callee address at call sites from the GOT and branch to it.
14069 This leads to more efficient code by eliminating PLT stubs and exposing
14070 GOT loads to optimizations. On architectures such as 32-bit x86 where
14071 PLT stubs expect the GOT pointer in a specific register, this gives more
14072 register allocation freedom to the compiler.
14073 Lazy binding requires use of the PLT;
14074 with @option{-fno-plt} all external symbols are resolved at load time.
14076 Alternatively, the function attribute @code{noplt} can be used to avoid calls
14077 through the PLT for specific external functions.
14079 In position-dependent code, a few targets also convert calls to
14080 functions that are marked to not use the PLT to use the GOT instead.
14082 @item -fno-jump-tables
14083 @opindex fno-jump-tables
14084 @opindex fjump-tables
14085 Do not use jump tables for switch statements even where it would be
14086 more efficient than other code generation strategies. This option is
14087 of use in conjunction with @option{-fpic} or @option{-fPIC} for
14088 building code that forms part of a dynamic linker and cannot
14089 reference the address of a jump table. On some targets, jump tables
14090 do not require a GOT and this option is not needed.
14092 @item -ffixed-@var{reg}
14094 Treat the register named @var{reg} as a fixed register; generated code
14095 should never refer to it (except perhaps as a stack pointer, frame
14096 pointer or in some other fixed role).
14098 @var{reg} must be the name of a register. The register names accepted
14099 are machine-specific and are defined in the @code{REGISTER_NAMES}
14100 macro in the machine description macro file.
14102 This flag does not have a negative form, because it specifies a
14105 @item -fcall-used-@var{reg}
14106 @opindex fcall-used
14107 Treat the register named @var{reg} as an allocable register that is
14108 clobbered by function calls. It may be allocated for temporaries or
14109 variables that do not live across a call. Functions compiled this way
14110 do not save and restore the register @var{reg}.
14112 It is an error to use this flag with the frame pointer or stack pointer.
14113 Use of this flag for other registers that have fixed pervasive roles in
14114 the machine's execution model produces disastrous results.
14116 This flag does not have a negative form, because it specifies a
14119 @item -fcall-saved-@var{reg}
14120 @opindex fcall-saved
14121 Treat the register named @var{reg} as an allocable register saved by
14122 functions. It may be allocated even for temporaries or variables that
14123 live across a call. Functions compiled this way save and restore
14124 the register @var{reg} if they use it.
14126 It is an error to use this flag with the frame pointer or stack pointer.
14127 Use of this flag for other registers that have fixed pervasive roles in
14128 the machine's execution model produces disastrous results.
14130 A different sort of disaster results from the use of this flag for
14131 a register in which function values may be returned.
14133 This flag does not have a negative form, because it specifies a
14136 @item -fpack-struct[=@var{n}]
14137 @opindex fpack-struct
14138 Without a value specified, pack all structure members together without
14139 holes. When a value is specified (which must be a small power of two), pack
14140 structure members according to this value, representing the maximum
14141 alignment (that is, objects with default alignment requirements larger than
14142 this are output potentially unaligned at the next fitting location.
14144 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
14145 code that is not binary compatible with code generated without that switch.
14146 Additionally, it makes the code suboptimal.
14147 Use it to conform to a non-default application binary interface.
14149 @item -fleading-underscore
14150 @opindex fleading-underscore
14151 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
14152 change the way C symbols are represented in the object file. One use
14153 is to help link with legacy assembly code.
14155 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
14156 generate code that is not binary compatible with code generated without that
14157 switch. Use it to conform to a non-default application binary interface.
14158 Not all targets provide complete support for this switch.
14160 @item -ftls-model=@var{model}
14161 @opindex ftls-model
14162 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
14163 The @var{model} argument should be one of @samp{global-dynamic},
14164 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
14165 Note that the choice is subject to optimization: the compiler may use
14166 a more efficient model for symbols not visible outside of the translation
14167 unit, or if @option{-fpic} is not given on the command line.
14169 The default without @option{-fpic} is @samp{initial-exec}; with
14170 @option{-fpic} the default is @samp{global-dynamic}.
14172 @item -ftrampolines
14173 @opindex ftrampolines
14174 For targets that normally need trampolines for nested functions, always
14175 generate them instead of using descriptors. Otherwise, for targets that
14176 do not need them, like for example HP-PA or IA-64, do nothing.
14178 A trampoline is a small piece of code that is created at run time on the
14179 stack when the address of a nested function is taken, and is used to call
14180 the nested function indirectly. Therefore, it requires the stack to be
14181 made executable in order for the program to work properly.
14183 @option{-fno-trampolines} is enabled by default on a language by language
14184 basis to let the compiler avoid generating them, if it computes that this
14185 is safe, and replace them with descriptors. Descriptors are made up of data
14186 only, but the generated code must be prepared to deal with them. As of this
14187 writing, @option{-fno-trampolines} is enabled by default only for Ada.
14189 Moreover, code compiled with @option{-ftrampolines} and code compiled with
14190 @option{-fno-trampolines} are not binary compatible if nested functions are
14191 present. This option must therefore be used on a program-wide basis and be
14192 manipulated with extreme care.
14194 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
14195 @opindex fvisibility
14196 Set the default ELF image symbol visibility to the specified option---all
14197 symbols are marked with this unless overridden within the code.
14198 Using this feature can very substantially improve linking and
14199 load times of shared object libraries, produce more optimized
14200 code, provide near-perfect API export and prevent symbol clashes.
14201 It is @strong{strongly} recommended that you use this in any shared objects
14204 Despite the nomenclature, @samp{default} always means public; i.e.,
14205 available to be linked against from outside the shared object.
14206 @samp{protected} and @samp{internal} are pretty useless in real-world
14207 usage so the only other commonly used option is @samp{hidden}.
14208 The default if @option{-fvisibility} isn't specified is
14209 @samp{default}, i.e., make every symbol public.
14211 A good explanation of the benefits offered by ensuring ELF
14212 symbols have the correct visibility is given by ``How To Write
14213 Shared Libraries'' by Ulrich Drepper (which can be found at
14214 @w{@uref{https://www.akkadia.org/drepper/}})---however a superior
14215 solution made possible by this option to marking things hidden when
14216 the default is public is to make the default hidden and mark things
14217 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
14218 and @code{__attribute__ ((visibility("default")))} instead of
14219 @code{__declspec(dllexport)} you get almost identical semantics with
14220 identical syntax. This is a great boon to those working with
14221 cross-platform projects.
14223 For those adding visibility support to existing code, you may find
14224 @code{#pragma GCC visibility} of use. This works by you enclosing
14225 the declarations you wish to set visibility for with (for example)
14226 @code{#pragma GCC visibility push(hidden)} and
14227 @code{#pragma GCC visibility pop}.
14228 Bear in mind that symbol visibility should be viewed @strong{as
14229 part of the API interface contract} and thus all new code should
14230 always specify visibility when it is not the default; i.e., declarations
14231 only for use within the local DSO should @strong{always} be marked explicitly
14232 as hidden as so to avoid PLT indirection overheads---making this
14233 abundantly clear also aids readability and self-documentation of the code.
14234 Note that due to ISO C++ specification requirements, @code{operator new} and
14235 @code{operator delete} must always be of default visibility.
14237 Be aware that headers from outside your project, in particular system
14238 headers and headers from any other library you use, may not be
14239 expecting to be compiled with visibility other than the default. You
14240 may need to explicitly say @code{#pragma GCC visibility push(default)}
14241 before including any such headers.
14243 @code{extern} declarations are not affected by @option{-fvisibility}, so
14244 a lot of code can be recompiled with @option{-fvisibility=hidden} with
14245 no modifications. However, this means that calls to @code{extern}
14246 functions with no explicit visibility use the PLT, so it is more
14247 effective to use @code{__attribute ((visibility))} and/or
14248 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
14249 declarations should be treated as hidden.
14251 Note that @option{-fvisibility} does affect C++ vague linkage
14252 entities. This means that, for instance, an exception class that is
14253 be thrown between DSOs must be explicitly marked with default
14254 visibility so that the @samp{type_info} nodes are unified between
14257 An overview of these techniques, their benefits and how to use them
14258 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
14260 @item -fstrict-volatile-bitfields
14261 @opindex fstrict-volatile-bitfields
14262 This option should be used if accesses to volatile bit-fields (or other
14263 structure fields, although the compiler usually honors those types
14264 anyway) should use a single access of the width of the
14265 field's type, aligned to a natural alignment if possible. For
14266 example, targets with memory-mapped peripheral registers might require
14267 all such accesses to be 16 bits wide; with this flag you can
14268 declare all peripheral bit-fields as @code{unsigned short} (assuming short
14269 is 16 bits on these targets) to force GCC to use 16-bit accesses
14270 instead of, perhaps, a more efficient 32-bit access.
14272 If this option is disabled, the compiler uses the most efficient
14273 instruction. In the previous example, that might be a 32-bit load
14274 instruction, even though that accesses bytes that do not contain
14275 any portion of the bit-field, or memory-mapped registers unrelated to
14276 the one being updated.
14278 In some cases, such as when the @code{packed} attribute is applied to a
14279 structure field, it may not be possible to access the field with a single
14280 read or write that is correctly aligned for the target machine. In this
14281 case GCC falls back to generating multiple accesses rather than code that
14282 will fault or truncate the result at run time.
14284 Note: Due to restrictions of the C/C++11 memory model, write accesses are
14285 not allowed to touch non bit-field members. It is therefore recommended
14286 to define all bits of the field's type as bit-field members.
14288 The default value of this option is determined by the application binary
14289 interface for the target processor.
14291 @item -fsync-libcalls
14292 @opindex fsync-libcalls
14293 This option controls whether any out-of-line instance of the @code{__sync}
14294 family of functions may be used to implement the C++11 @code{__atomic}
14295 family of functions.
14297 The default value of this option is enabled, thus the only useful form
14298 of the option is @option{-fno-sync-libcalls}. This option is used in
14299 the implementation of the @file{libatomic} runtime library.
14303 @node Developer Options
14304 @section GCC Developer Options
14305 @cindex developer options
14306 @cindex debugging GCC
14307 @cindex debug dump options
14308 @cindex dump options
14309 @cindex compilation statistics
14311 This section describes command-line options that are primarily of
14312 interest to GCC developers, including options to support compiler
14313 testing and investigation of compiler bugs and compile-time
14314 performance problems. This includes options that produce debug dumps
14315 at various points in the compilation; that print statistics such as
14316 memory use and execution time; and that print information about GCC's
14317 configuration, such as where it searches for libraries. You should
14318 rarely need to use any of these options for ordinary compilation and
14321 Many developer options that cause GCC to dump output to a file take an
14322 optional @samp{=@var{filename}} suffix. You can specify @samp{stdout}
14323 or @samp{-} to dump to standard output, and @samp{stderr} for standard
14326 If @samp{=@var{filename}} is omitted, a default dump file name is
14327 constructed by concatenating the base dump file name, a pass number,
14328 phase letter, and pass name. The base dump file name is the name of
14329 output file produced by the compiler if explicitly specified and not
14330 an executable; otherwise it is the source file name.
14331 The pass number is determined by the order passes are registered with
14332 the compiler's pass manager.
14333 This is generally the same as the order of execution, but passes
14334 registered by plugins, target-specific passes, or passes that are
14335 otherwise registered late are numbered higher than the pass named
14336 @samp{final}, even if they are executed earlier. The phase letter is
14337 one of @samp{i} (inter-procedural analysis), @samp{l}
14338 (language-specific), @samp{r} (RTL), or @samp{t} (tree).
14339 The files are created in the directory of the output file.
14343 @item -d@var{letters}
14344 @itemx -fdump-rtl-@var{pass}
14345 @itemx -fdump-rtl-@var{pass}=@var{filename}
14347 @opindex fdump-rtl-@var{pass}
14348 Says to make debugging dumps during compilation at times specified by
14349 @var{letters}. This is used for debugging the RTL-based passes of the
14352 Some @option{-d@var{letters}} switches have different meaning when
14353 @option{-E} is used for preprocessing. @xref{Preprocessor Options},
14354 for information about preprocessor-specific dump options.
14356 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
14357 @option{-d} option @var{letters}. Here are the possible
14358 letters for use in @var{pass} and @var{letters}, and their meanings:
14362 @item -fdump-rtl-alignments
14363 @opindex fdump-rtl-alignments
14364 Dump after branch alignments have been computed.
14366 @item -fdump-rtl-asmcons
14367 @opindex fdump-rtl-asmcons
14368 Dump after fixing rtl statements that have unsatisfied in/out constraints.
14370 @item -fdump-rtl-auto_inc_dec
14371 @opindex fdump-rtl-auto_inc_dec
14372 Dump after auto-inc-dec discovery. This pass is only run on
14373 architectures that have auto inc or auto dec instructions.
14375 @item -fdump-rtl-barriers
14376 @opindex fdump-rtl-barriers
14377 Dump after cleaning up the barrier instructions.
14379 @item -fdump-rtl-bbpart
14380 @opindex fdump-rtl-bbpart
14381 Dump after partitioning hot and cold basic blocks.
14383 @item -fdump-rtl-bbro
14384 @opindex fdump-rtl-bbro
14385 Dump after block reordering.
14387 @item -fdump-rtl-btl1
14388 @itemx -fdump-rtl-btl2
14389 @opindex fdump-rtl-btl2
14390 @opindex fdump-rtl-btl2
14391 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
14392 after the two branch
14393 target load optimization passes.
14395 @item -fdump-rtl-bypass
14396 @opindex fdump-rtl-bypass
14397 Dump after jump bypassing and control flow optimizations.
14399 @item -fdump-rtl-combine
14400 @opindex fdump-rtl-combine
14401 Dump after the RTL instruction combination pass.
14403 @item -fdump-rtl-compgotos
14404 @opindex fdump-rtl-compgotos
14405 Dump after duplicating the computed gotos.
14407 @item -fdump-rtl-ce1
14408 @itemx -fdump-rtl-ce2
14409 @itemx -fdump-rtl-ce3
14410 @opindex fdump-rtl-ce1
14411 @opindex fdump-rtl-ce2
14412 @opindex fdump-rtl-ce3
14413 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
14414 @option{-fdump-rtl-ce3} enable dumping after the three
14415 if conversion passes.
14417 @item -fdump-rtl-cprop_hardreg
14418 @opindex fdump-rtl-cprop_hardreg
14419 Dump after hard register copy propagation.
14421 @item -fdump-rtl-csa
14422 @opindex fdump-rtl-csa
14423 Dump after combining stack adjustments.
14425 @item -fdump-rtl-cse1
14426 @itemx -fdump-rtl-cse2
14427 @opindex fdump-rtl-cse1
14428 @opindex fdump-rtl-cse2
14429 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
14430 the two common subexpression elimination passes.
14432 @item -fdump-rtl-dce
14433 @opindex fdump-rtl-dce
14434 Dump after the standalone dead code elimination passes.
14436 @item -fdump-rtl-dbr
14437 @opindex fdump-rtl-dbr
14438 Dump after delayed branch scheduling.
14440 @item -fdump-rtl-dce1
14441 @itemx -fdump-rtl-dce2
14442 @opindex fdump-rtl-dce1
14443 @opindex fdump-rtl-dce2
14444 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
14445 the two dead store elimination passes.
14447 @item -fdump-rtl-eh
14448 @opindex fdump-rtl-eh
14449 Dump after finalization of EH handling code.
14451 @item -fdump-rtl-eh_ranges
14452 @opindex fdump-rtl-eh_ranges
14453 Dump after conversion of EH handling range regions.
14455 @item -fdump-rtl-expand
14456 @opindex fdump-rtl-expand
14457 Dump after RTL generation.
14459 @item -fdump-rtl-fwprop1
14460 @itemx -fdump-rtl-fwprop2
14461 @opindex fdump-rtl-fwprop1
14462 @opindex fdump-rtl-fwprop2
14463 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
14464 dumping after the two forward propagation passes.
14466 @item -fdump-rtl-gcse1
14467 @itemx -fdump-rtl-gcse2
14468 @opindex fdump-rtl-gcse1
14469 @opindex fdump-rtl-gcse2
14470 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
14471 after global common subexpression elimination.
14473 @item -fdump-rtl-init-regs
14474 @opindex fdump-rtl-init-regs
14475 Dump after the initialization of the registers.
14477 @item -fdump-rtl-initvals
14478 @opindex fdump-rtl-initvals
14479 Dump after the computation of the initial value sets.
14481 @item -fdump-rtl-into_cfglayout
14482 @opindex fdump-rtl-into_cfglayout
14483 Dump after converting to cfglayout mode.
14485 @item -fdump-rtl-ira
14486 @opindex fdump-rtl-ira
14487 Dump after iterated register allocation.
14489 @item -fdump-rtl-jump
14490 @opindex fdump-rtl-jump
14491 Dump after the second jump optimization.
14493 @item -fdump-rtl-loop2
14494 @opindex fdump-rtl-loop2
14495 @option{-fdump-rtl-loop2} enables dumping after the rtl
14496 loop optimization passes.
14498 @item -fdump-rtl-mach
14499 @opindex fdump-rtl-mach
14500 Dump after performing the machine dependent reorganization pass, if that
14503 @item -fdump-rtl-mode_sw
14504 @opindex fdump-rtl-mode_sw
14505 Dump after removing redundant mode switches.
14507 @item -fdump-rtl-rnreg
14508 @opindex fdump-rtl-rnreg
14509 Dump after register renumbering.
14511 @item -fdump-rtl-outof_cfglayout
14512 @opindex fdump-rtl-outof_cfglayout
14513 Dump after converting from cfglayout mode.
14515 @item -fdump-rtl-peephole2
14516 @opindex fdump-rtl-peephole2
14517 Dump after the peephole pass.
14519 @item -fdump-rtl-postreload
14520 @opindex fdump-rtl-postreload
14521 Dump after post-reload optimizations.
14523 @item -fdump-rtl-pro_and_epilogue
14524 @opindex fdump-rtl-pro_and_epilogue
14525 Dump after generating the function prologues and epilogues.
14527 @item -fdump-rtl-sched1
14528 @itemx -fdump-rtl-sched2
14529 @opindex fdump-rtl-sched1
14530 @opindex fdump-rtl-sched2
14531 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
14532 after the basic block scheduling passes.
14534 @item -fdump-rtl-ree
14535 @opindex fdump-rtl-ree
14536 Dump after sign/zero extension elimination.
14538 @item -fdump-rtl-seqabstr
14539 @opindex fdump-rtl-seqabstr
14540 Dump after common sequence discovery.
14542 @item -fdump-rtl-shorten
14543 @opindex fdump-rtl-shorten
14544 Dump after shortening branches.
14546 @item -fdump-rtl-sibling
14547 @opindex fdump-rtl-sibling
14548 Dump after sibling call optimizations.
14550 @item -fdump-rtl-split1
14551 @itemx -fdump-rtl-split2
14552 @itemx -fdump-rtl-split3
14553 @itemx -fdump-rtl-split4
14554 @itemx -fdump-rtl-split5
14555 @opindex fdump-rtl-split1
14556 @opindex fdump-rtl-split2
14557 @opindex fdump-rtl-split3
14558 @opindex fdump-rtl-split4
14559 @opindex fdump-rtl-split5
14560 These options enable dumping after five rounds of
14561 instruction splitting.
14563 @item -fdump-rtl-sms
14564 @opindex fdump-rtl-sms
14565 Dump after modulo scheduling. This pass is only run on some
14568 @item -fdump-rtl-stack
14569 @opindex fdump-rtl-stack
14570 Dump after conversion from GCC's ``flat register file'' registers to the
14571 x87's stack-like registers. This pass is only run on x86 variants.
14573 @item -fdump-rtl-subreg1
14574 @itemx -fdump-rtl-subreg2
14575 @opindex fdump-rtl-subreg1
14576 @opindex fdump-rtl-subreg2
14577 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
14578 the two subreg expansion passes.
14580 @item -fdump-rtl-unshare
14581 @opindex fdump-rtl-unshare
14582 Dump after all rtl has been unshared.
14584 @item -fdump-rtl-vartrack
14585 @opindex fdump-rtl-vartrack
14586 Dump after variable tracking.
14588 @item -fdump-rtl-vregs
14589 @opindex fdump-rtl-vregs
14590 Dump after converting virtual registers to hard registers.
14592 @item -fdump-rtl-web
14593 @opindex fdump-rtl-web
14594 Dump after live range splitting.
14596 @item -fdump-rtl-regclass
14597 @itemx -fdump-rtl-subregs_of_mode_init
14598 @itemx -fdump-rtl-subregs_of_mode_finish
14599 @itemx -fdump-rtl-dfinit
14600 @itemx -fdump-rtl-dfinish
14601 @opindex fdump-rtl-regclass
14602 @opindex fdump-rtl-subregs_of_mode_init
14603 @opindex fdump-rtl-subregs_of_mode_finish
14604 @opindex fdump-rtl-dfinit
14605 @opindex fdump-rtl-dfinish
14606 These dumps are defined but always produce empty files.
14609 @itemx -fdump-rtl-all
14611 @opindex fdump-rtl-all
14612 Produce all the dumps listed above.
14616 Annotate the assembler output with miscellaneous debugging information.
14620 Dump all macro definitions, at the end of preprocessing, in addition to
14625 Produce a core dump whenever an error occurs.
14629 Annotate the assembler output with a comment indicating which
14630 pattern and alternative is used. The length and cost of each instruction are
14635 Dump the RTL in the assembler output as a comment before each instruction.
14636 Also turns on @option{-dp} annotation.
14640 Just generate RTL for a function instead of compiling it. Usually used
14641 with @option{-fdump-rtl-expand}.
14645 @opindex fdump-debug
14646 Dump debugging information generated during the debug
14649 @item -fdump-earlydebug
14650 @opindex fdump-earlydebug
14651 Dump debugging information generated during the early debug
14654 @item -fdump-noaddr
14655 @opindex fdump-noaddr
14656 When doing debugging dumps, suppress address output. This makes it more
14657 feasible to use diff on debugging dumps for compiler invocations with
14658 different compiler binaries and/or different
14659 text / bss / data / heap / stack / dso start locations.
14662 @opindex freport-bug
14663 Collect and dump debug information into a temporary file if an
14664 internal compiler error (ICE) occurs.
14666 @item -fdump-unnumbered
14667 @opindex fdump-unnumbered
14668 When doing debugging dumps, suppress instruction numbers and address output.
14669 This makes it more feasible to use diff on debugging dumps for compiler
14670 invocations with different options, in particular with and without
14673 @item -fdump-unnumbered-links
14674 @opindex fdump-unnumbered-links
14675 When doing debugging dumps (see @option{-d} option above), suppress
14676 instruction numbers for the links to the previous and next instructions
14679 @item -fdump-ipa-@var{switch}
14680 @itemx -fdump-ipa-@var{switch}-@var{options}
14682 Control the dumping at various stages of inter-procedural analysis
14683 language tree to a file. The file name is generated by appending a
14684 switch specific suffix to the source file name, and the file is created
14685 in the same directory as the output file. The following dumps are
14690 Enables all inter-procedural analysis dumps.
14693 Dumps information about call-graph optimization, unused function removal,
14694 and inlining decisions.
14697 Dump after function inlining.
14701 Additionally, the options @option{-optimized}, @option{-missed},
14702 @option{-note}, and @option{-all} can be provided, with the same meaning
14703 as for @option{-fopt-info}, defaulting to @option{-optimized}.
14705 For example, @option{-fdump-ipa-inline-optimized-missed} will emit
14706 information on callsites that were inlined, along with callsites
14707 that were not inlined.
14709 By default, the dump will contain messages about successful
14710 optimizations (equivalent to @option{-optimized}) together with
14711 low-level details about the analysis.
14713 @item -fdump-lang-all
14714 @itemx -fdump-lang-@var{switch}
14715 @itemx -fdump-lang-@var{switch}-@var{options}
14716 @itemx -fdump-lang-@var{switch}-@var{options}=@var{filename}
14717 @opindex fdump-lang-all
14718 @opindex fdump-lang
14719 Control the dumping of language-specific information. The @var{options}
14720 and @var{filename} portions behave as described in the
14721 @option{-fdump-tree} option. The following @var{switch} values are
14727 Enable all language-specific dumps.
14730 Dump class hierarchy information. Virtual table information is emitted
14731 unless '@option{slim}' is specified. This option is applicable to C++ only.
14734 Dump the raw internal tree data. This option is applicable to C++ only.
14738 @item -fdump-passes
14739 @opindex fdump-passes
14740 Print on @file{stderr} the list of optimization passes that are turned
14741 on and off by the current command-line options.
14743 @item -fdump-statistics-@var{option}
14744 @opindex fdump-statistics
14745 Enable and control dumping of pass statistics in a separate file. The
14746 file name is generated by appending a suffix ending in
14747 @samp{.statistics} to the source file name, and the file is created in
14748 the same directory as the output file. If the @samp{-@var{option}}
14749 form is used, @samp{-stats} causes counters to be summed over the
14750 whole compilation unit while @samp{-details} dumps every event as
14751 the passes generate them. The default with no option is to sum
14752 counters for each function compiled.
14754 @item -fdump-tree-all
14755 @itemx -fdump-tree-@var{switch}
14756 @itemx -fdump-tree-@var{switch}-@var{options}
14757 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
14758 @opindex fdump-tree-all
14759 @opindex fdump-tree
14760 Control the dumping at various stages of processing the intermediate
14761 language tree to a file. If the @samp{-@var{options}}
14762 form is used, @var{options} is a list of @samp{-} separated options
14763 which control the details of the dump. Not all options are applicable
14764 to all dumps; those that are not meaningful are ignored. The
14765 following options are available
14769 Print the address of each node. Usually this is not meaningful as it
14770 changes according to the environment and source file. Its primary use
14771 is for tying up a dump file with a debug environment.
14773 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
14774 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
14775 use working backward from mangled names in the assembly file.
14777 When dumping front-end intermediate representations, inhibit dumping
14778 of members of a scope or body of a function merely because that scope
14779 has been reached. Only dump such items when they are directly reachable
14780 by some other path.
14782 When dumping pretty-printed trees, this option inhibits dumping the
14783 bodies of control structures.
14785 When dumping RTL, print the RTL in slim (condensed) form instead of
14786 the default LISP-like representation.
14788 Print a raw representation of the tree. By default, trees are
14789 pretty-printed into a C-like representation.
14791 Enable more detailed dumps (not honored by every dump option). Also
14792 include information from the optimization passes.
14794 Enable dumping various statistics about the pass (not honored by every dump
14797 Enable showing basic block boundaries (disabled in raw dumps).
14799 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
14800 dump a representation of the control flow graph suitable for viewing with
14801 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
14802 the file is pretty-printed as a subgraph, so that GraphViz can render them
14803 all in a single plot.
14805 This option currently only works for RTL dumps, and the RTL is always
14806 dumped in slim form.
14808 Enable showing virtual operands for every statement.
14810 Enable showing line numbers for statements.
14812 Enable showing the unique ID (@code{DECL_UID}) for each variable.
14814 Enable showing the tree dump for each statement.
14816 Enable showing the EH region number holding each statement.
14818 Enable showing scalar evolution analysis details.
14820 Enable showing optimization information (only available in certain
14823 Enable showing missed optimization information (only available in certain
14826 Enable other detailed optimization information (only available in
14829 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
14830 and @option{lineno}.
14832 Turn on all optimization options, i.e., @option{optimized},
14833 @option{missed}, and @option{note}.
14836 To determine what tree dumps are available or find the dump for a pass
14837 of interest follow the steps below.
14841 Invoke GCC with @option{-fdump-passes} and in the @file{stderr} output
14842 look for a code that corresponds to the pass you are interested in.
14843 For example, the codes @code{tree-evrp}, @code{tree-vrp1}, and
14844 @code{tree-vrp2} correspond to the three Value Range Propagation passes.
14845 The number at the end distinguishes distinct invocations of the same pass.
14847 To enable the creation of the dump file, append the pass code to
14848 the @option{-fdump-} option prefix and invoke GCC with it. For example,
14849 to enable the dump from the Early Value Range Propagation pass, invoke
14850 GCC with the @option{-fdump-tree-evrp} option. Optionally, you may
14851 specify the name of the dump file. If you don't specify one, GCC
14852 creates as described below.
14854 Find the pass dump in a file whose name is composed of three components
14855 separated by a period: the name of the source file GCC was invoked to
14856 compile, a numeric suffix indicating the pass number followed by the
14857 letter @samp{t} for tree passes (and the letter @samp{r} for RTL passes),
14858 and finally the pass code. For example, the Early VRP pass dump might
14859 be in a file named @file{myfile.c.038t.evrp} in the current working
14860 directory. Note that the numeric codes are not stable and may change
14861 from one version of GCC to another.
14865 @itemx -fopt-info-@var{options}
14866 @itemx -fopt-info-@var{options}=@var{filename}
14868 Controls optimization dumps from various optimization passes. If the
14869 @samp{-@var{options}} form is used, @var{options} is a list of
14870 @samp{-} separated option keywords to select the dump details and
14873 The @var{options} can be divided into three groups:
14876 options describing what kinds of messages should be emitted,
14878 options describing the verbosity of the dump, and
14880 options describing which optimizations should be included.
14882 The options from each group can be freely mixed as they are
14883 non-overlapping. However, in case of any conflicts,
14884 the later options override the earlier options on the command
14887 The following options control which kinds of messages should be emitted:
14891 Print information when an optimization is successfully applied. It is
14892 up to a pass to decide which information is relevant. For example, the
14893 vectorizer passes print the source location of loops which are
14894 successfully vectorized.
14896 Print information about missed optimizations. Individual passes
14897 control which information to include in the output.
14899 Print verbose information about optimizations, such as certain
14900 transformations, more detailed messages about decisions etc.
14902 Print detailed optimization information. This includes
14903 @samp{optimized}, @samp{missed}, and @samp{note}.
14906 The following option controls the dump verbosity:
14910 By default, only ``high-level'' messages are emitted. This option enables
14911 additional, more detailed, messages, which are likely to only be of interest
14915 One or more of the following option keywords can be used to describe a
14916 group of optimizations:
14920 Enable dumps from all interprocedural optimizations.
14922 Enable dumps from all loop optimizations.
14924 Enable dumps from all inlining optimizations.
14926 Enable dumps from all OMP (Offloading and Multi Processing) optimizations.
14928 Enable dumps from all vectorization optimizations.
14930 Enable dumps from all optimizations. This is a superset of
14931 the optimization groups listed above.
14934 If @var{options} is
14935 omitted, it defaults to @samp{optimized-optall}, which means to dump messages
14936 about successful optimizations from all the passes, omitting messages
14937 that are treated as ``internals''.
14939 If the @var{filename} is provided, then the dumps from all the
14940 applicable optimizations are concatenated into the @var{filename}.
14941 Otherwise the dump is output onto @file{stderr}. Though multiple
14942 @option{-fopt-info} options are accepted, only one of them can include
14943 a @var{filename}. If other filenames are provided then all but the
14944 first such option are ignored.
14946 Note that the output @var{filename} is overwritten
14947 in case of multiple translation units. If a combined output from
14948 multiple translation units is desired, @file{stderr} should be used
14951 In the following example, the optimization info is output to
14960 gcc -O3 -fopt-info-missed=missed.all
14964 outputs missed optimization report from all the passes into
14965 @file{missed.all}, and this one:
14968 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
14972 prints information about missed optimization opportunities from
14973 vectorization passes on @file{stderr}.
14974 Note that @option{-fopt-info-vec-missed} is equivalent to
14975 @option{-fopt-info-missed-vec}. The order of the optimization group
14976 names and message types listed after @option{-fopt-info} does not matter.
14978 As another example,
14980 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
14984 outputs information about missed optimizations as well as
14985 optimized locations from all the inlining passes into
14991 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
14995 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
14996 in conflict since only one output file is allowed. In this case, only
14997 the first option takes effect and the subsequent options are
14998 ignored. Thus only @file{vec.miss} is produced which contains
14999 dumps from the vectorizer about missed opportunities.
15001 @item -fsave-optimization-record
15002 @opindex fsave-optimization-record
15003 Write a SRCFILE.opt-record.json.gz file detailing what optimizations
15004 were performed, for those optimizations that support @option{-fopt-info}.
15006 This option is experimental and the format of the data within the
15007 compressed JSON file is subject to change.
15009 It is roughly equivalent to a machine-readable version of
15010 @option{-fopt-info-all}, as a collection of messages with source file,
15011 line number and column number, with the following additional data for
15017 the execution count of the code being optimized, along with metadata about
15018 whether this was from actual profile data, or just an estimate, allowing
15019 consumers to prioritize messages by code hotness,
15022 the function name of the code being optimized, where applicable,
15025 the ``inlining chain'' for the code being optimized, so that when
15026 a function is inlined into several different places (which might
15027 themselves be inlined), the reader can distinguish between the copies,
15030 objects identifying those parts of the message that refer to expressions,
15031 statements or symbol-table nodes, which of these categories they are, and,
15032 when available, their source code location,
15035 the GCC pass that emitted the message, and
15038 the location in GCC's own code from which the message was emitted
15042 Additionally, some messages are logically nested within other
15043 messages, reflecting implementation details of the optimization
15046 @item -fsched-verbose=@var{n}
15047 @opindex fsched-verbose
15048 On targets that use instruction scheduling, this option controls the
15049 amount of debugging output the scheduler prints to the dump files.
15051 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
15052 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
15053 For @var{n} greater than one, it also output basic block probabilities,
15054 detailed ready list information and unit/insn info. For @var{n} greater
15055 than two, it includes RTL at abort point, control-flow and regions info.
15056 And for @var{n} over four, @option{-fsched-verbose} also includes
15061 @item -fenable-@var{kind}-@var{pass}
15062 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
15066 This is a set of options that are used to explicitly disable/enable
15067 optimization passes. These options are intended for use for debugging GCC.
15068 Compiler users should use regular options for enabling/disabling
15073 @item -fdisable-ipa-@var{pass}
15074 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
15075 statically invoked in the compiler multiple times, the pass name should be
15076 appended with a sequential number starting from 1.
15078 @item -fdisable-rtl-@var{pass}
15079 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
15080 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
15081 statically invoked in the compiler multiple times, the pass name should be
15082 appended with a sequential number starting from 1. @var{range-list} is a
15083 comma-separated list of function ranges or assembler names. Each range is a number
15084 pair separated by a colon. The range is inclusive in both ends. If the range
15085 is trivial, the number pair can be simplified as a single number. If the
15086 function's call graph node's @var{uid} falls within one of the specified ranges,
15087 the @var{pass} is disabled for that function. The @var{uid} is shown in the
15088 function header of a dump file, and the pass names can be dumped by using
15089 option @option{-fdump-passes}.
15091 @item -fdisable-tree-@var{pass}
15092 @itemx -fdisable-tree-@var{pass}=@var{range-list}
15093 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
15096 @item -fenable-ipa-@var{pass}
15097 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
15098 statically invoked in the compiler multiple times, the pass name should be
15099 appended with a sequential number starting from 1.
15101 @item -fenable-rtl-@var{pass}
15102 @itemx -fenable-rtl-@var{pass}=@var{range-list}
15103 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
15104 description and examples.
15106 @item -fenable-tree-@var{pass}
15107 @itemx -fenable-tree-@var{pass}=@var{range-list}
15108 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
15109 of option arguments.
15113 Here are some examples showing uses of these options.
15117 # disable ccp1 for all functions
15118 -fdisable-tree-ccp1
15119 # disable complete unroll for function whose cgraph node uid is 1
15120 -fenable-tree-cunroll=1
15121 # disable gcse2 for functions at the following ranges [1,1],
15122 # [300,400], and [400,1000]
15123 # disable gcse2 for functions foo and foo2
15124 -fdisable-rtl-gcse2=foo,foo2
15125 # disable early inlining
15126 -fdisable-tree-einline
15127 # disable ipa inlining
15128 -fdisable-ipa-inline
15129 # enable tree full unroll
15130 -fenable-tree-unroll
15135 @itemx -fchecking=@var{n}
15137 @opindex fno-checking
15138 Enable internal consistency checking. The default depends on
15139 the compiler configuration. @option{-fchecking=2} enables further
15140 internal consistency checking that might affect code generation.
15142 @item -frandom-seed=@var{string}
15143 @opindex frandom-seed
15144 This option provides a seed that GCC uses in place of
15145 random numbers in generating certain symbol names
15146 that have to be different in every compiled file. It is also used to
15147 place unique stamps in coverage data files and the object files that
15148 produce them. You can use the @option{-frandom-seed} option to produce
15149 reproducibly identical object files.
15151 The @var{string} can either be a number (decimal, octal or hex) or an
15152 arbitrary string (in which case it's converted to a number by
15155 The @var{string} should be different for every file you compile.
15158 @itemx -save-temps=cwd
15159 @opindex save-temps
15160 Store the usual ``temporary'' intermediate files permanently; place them
15161 in the current directory and name them based on the source file. Thus,
15162 compiling @file{foo.c} with @option{-c -save-temps} produces files
15163 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
15164 preprocessed @file{foo.i} output file even though the compiler now
15165 normally uses an integrated preprocessor.
15167 When used in combination with the @option{-x} command-line option,
15168 @option{-save-temps} is sensible enough to avoid over writing an
15169 input source file with the same extension as an intermediate file.
15170 The corresponding intermediate file may be obtained by renaming the
15171 source file before using @option{-save-temps}.
15173 If you invoke GCC in parallel, compiling several different source
15174 files that share a common base name in different subdirectories or the
15175 same source file compiled for multiple output destinations, it is
15176 likely that the different parallel compilers will interfere with each
15177 other, and overwrite the temporary files. For instance:
15180 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
15181 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
15184 may result in @file{foo.i} and @file{foo.o} being written to
15185 simultaneously by both compilers.
15187 @item -save-temps=obj
15188 @opindex save-temps=obj
15189 Store the usual ``temporary'' intermediate files permanently. If the
15190 @option{-o} option is used, the temporary files are based on the
15191 object file. If the @option{-o} option is not used, the
15192 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
15197 gcc -save-temps=obj -c foo.c
15198 gcc -save-temps=obj -c bar.c -o dir/xbar.o
15199 gcc -save-temps=obj foobar.c -o dir2/yfoobar
15203 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
15204 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
15205 @file{dir2/yfoobar.o}.
15207 @item -time@r{[}=@var{file}@r{]}
15209 Report the CPU time taken by each subprocess in the compilation
15210 sequence. For C source files, this is the compiler proper and assembler
15211 (plus the linker if linking is done).
15213 Without the specification of an output file, the output looks like this:
15220 The first number on each line is the ``user time'', that is time spent
15221 executing the program itself. The second number is ``system time'',
15222 time spent executing operating system routines on behalf of the program.
15223 Both numbers are in seconds.
15225 With the specification of an output file, the output is appended to the
15226 named file, and it looks like this:
15229 0.12 0.01 cc1 @var{options}
15230 0.00 0.01 as @var{options}
15233 The ``user time'' and the ``system time'' are moved before the program
15234 name, and the options passed to the program are displayed, so that one
15235 can later tell what file was being compiled, and with which options.
15237 @item -fdump-final-insns@r{[}=@var{file}@r{]}
15238 @opindex fdump-final-insns
15239 Dump the final internal representation (RTL) to @var{file}. If the
15240 optional argument is omitted (or if @var{file} is @code{.}), the name
15241 of the dump file is determined by appending @code{.gkd} to the
15242 compilation output file name.
15244 @item -fcompare-debug@r{[}=@var{opts}@r{]}
15245 @opindex fcompare-debug
15246 @opindex fno-compare-debug
15247 If no error occurs during compilation, run the compiler a second time,
15248 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
15249 passed to the second compilation. Dump the final internal
15250 representation in both compilations, and print an error if they differ.
15252 If the equal sign is omitted, the default @option{-gtoggle} is used.
15254 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
15255 and nonzero, implicitly enables @option{-fcompare-debug}. If
15256 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
15257 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
15260 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
15261 is equivalent to @option{-fno-compare-debug}, which disables the dumping
15262 of the final representation and the second compilation, preventing even
15263 @env{GCC_COMPARE_DEBUG} from taking effect.
15265 To verify full coverage during @option{-fcompare-debug} testing, set
15266 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
15267 which GCC rejects as an invalid option in any actual compilation
15268 (rather than preprocessing, assembly or linking). To get just a
15269 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
15270 not overridden} will do.
15272 @item -fcompare-debug-second
15273 @opindex fcompare-debug-second
15274 This option is implicitly passed to the compiler for the second
15275 compilation requested by @option{-fcompare-debug}, along with options to
15276 silence warnings, and omitting other options that would cause the compiler
15277 to produce output to files or to standard output as a side effect. Dump
15278 files and preserved temporary files are renamed so as to contain the
15279 @code{.gk} additional extension during the second compilation, to avoid
15280 overwriting those generated by the first.
15282 When this option is passed to the compiler driver, it causes the
15283 @emph{first} compilation to be skipped, which makes it useful for little
15284 other than debugging the compiler proper.
15288 Turn off generation of debug info, if leaving out this option
15289 generates it, or turn it on at level 2 otherwise. The position of this
15290 argument in the command line does not matter; it takes effect after all
15291 other options are processed, and it does so only once, no matter how
15292 many times it is given. This is mainly intended to be used with
15293 @option{-fcompare-debug}.
15295 @item -fvar-tracking-assignments-toggle
15296 @opindex fvar-tracking-assignments-toggle
15297 @opindex fno-var-tracking-assignments-toggle
15298 Toggle @option{-fvar-tracking-assignments}, in the same way that
15299 @option{-gtoggle} toggles @option{-g}.
15303 Makes the compiler print out each function name as it is compiled, and
15304 print some statistics about each pass when it finishes.
15306 @item -ftime-report
15307 @opindex ftime-report
15308 Makes the compiler print some statistics about the time consumed by each
15309 pass when it finishes.
15311 @item -ftime-report-details
15312 @opindex ftime-report-details
15313 Record the time consumed by infrastructure parts separately for each pass.
15315 @item -fira-verbose=@var{n}
15316 @opindex fira-verbose
15317 Control the verbosity of the dump file for the integrated register allocator.
15318 The default value is 5. If the value @var{n} is greater or equal to 10,
15319 the dump output is sent to stderr using the same format as @var{n} minus 10.
15322 @opindex flto-report
15323 Prints a report with internal details on the workings of the link-time
15324 optimizer. The contents of this report vary from version to version.
15325 It is meant to be useful to GCC developers when processing object
15326 files in LTO mode (via @option{-flto}).
15328 Disabled by default.
15330 @item -flto-report-wpa
15331 @opindex flto-report-wpa
15332 Like @option{-flto-report}, but only print for the WPA phase of Link
15336 @opindex fmem-report
15337 Makes the compiler print some statistics about permanent memory
15338 allocation when it finishes.
15340 @item -fmem-report-wpa
15341 @opindex fmem-report-wpa
15342 Makes the compiler print some statistics about permanent memory
15343 allocation for the WPA phase only.
15345 @item -fpre-ipa-mem-report
15346 @opindex fpre-ipa-mem-report
15347 @item -fpost-ipa-mem-report
15348 @opindex fpost-ipa-mem-report
15349 Makes the compiler print some statistics about permanent memory
15350 allocation before or after interprocedural optimization.
15352 @item -fprofile-report
15353 @opindex fprofile-report
15354 Makes the compiler print some statistics about consistency of the
15355 (estimated) profile and effect of individual passes.
15357 @item -fstack-usage
15358 @opindex fstack-usage
15359 Makes the compiler output stack usage information for the program, on a
15360 per-function basis. The filename for the dump is made by appending
15361 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
15362 the output file, if explicitly specified and it is not an executable,
15363 otherwise it is the basename of the source file. An entry is made up
15368 The name of the function.
15372 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
15375 The qualifier @code{static} means that the function manipulates the stack
15376 statically: a fixed number of bytes are allocated for the frame on function
15377 entry and released on function exit; no stack adjustments are otherwise made
15378 in the function. The second field is this fixed number of bytes.
15380 The qualifier @code{dynamic} means that the function manipulates the stack
15381 dynamically: in addition to the static allocation described above, stack
15382 adjustments are made in the body of the function, for example to push/pop
15383 arguments around function calls. If the qualifier @code{bounded} is also
15384 present, the amount of these adjustments is bounded at compile time and
15385 the second field is an upper bound of the total amount of stack used by
15386 the function. If it is not present, the amount of these adjustments is
15387 not bounded at compile time and the second field only represents the
15392 Emit statistics about front-end processing at the end of the compilation.
15393 This option is supported only by the C++ front end, and
15394 the information is generally only useful to the G++ development team.
15396 @item -fdbg-cnt-list
15397 @opindex fdbg-cnt-list
15398 Print the name and the counter upper bound for all debug counters.
15401 @item -fdbg-cnt=@var{counter-value-list}
15403 Set the internal debug counter lower and upper bound. @var{counter-value-list}
15404 is a comma-separated list of @var{name}:@var{lower_bound}:@var{upper_bound}
15405 tuples which sets the lower and the upper bound of each debug
15406 counter @var{name}. The @var{lower_bound} is optional and is zero
15407 initialized if not set.
15408 All debug counters have the initial upper bound of @code{UINT_MAX};
15409 thus @code{dbg_cnt} returns true always unless the upper bound
15410 is set by this option.
15411 For example, with @option{-fdbg-cnt=dce:2:4,tail_call:10},
15412 @code{dbg_cnt(dce)} returns true only for third and fourth invocation.
15413 For @code{dbg_cnt(tail_call)} true is returned for first 10 invocations.
15415 @item -print-file-name=@var{library}
15416 @opindex print-file-name
15417 Print the full absolute name of the library file @var{library} that
15418 would be used when linking---and don't do anything else. With this
15419 option, GCC does not compile or link anything; it just prints the
15422 @item -print-multi-directory
15423 @opindex print-multi-directory
15424 Print the directory name corresponding to the multilib selected by any
15425 other switches present in the command line. This directory is supposed
15426 to exist in @env{GCC_EXEC_PREFIX}.
15428 @item -print-multi-lib
15429 @opindex print-multi-lib
15430 Print the mapping from multilib directory names to compiler switches
15431 that enable them. The directory name is separated from the switches by
15432 @samp{;}, and each switch starts with an @samp{@@} instead of the
15433 @samp{-}, without spaces between multiple switches. This is supposed to
15434 ease shell processing.
15436 @item -print-multi-os-directory
15437 @opindex print-multi-os-directory
15438 Print the path to OS libraries for the selected
15439 multilib, relative to some @file{lib} subdirectory. If OS libraries are
15440 present in the @file{lib} subdirectory and no multilibs are used, this is
15441 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
15442 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
15443 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
15444 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
15446 @item -print-multiarch
15447 @opindex print-multiarch
15448 Print the path to OS libraries for the selected multiarch,
15449 relative to some @file{lib} subdirectory.
15451 @item -print-prog-name=@var{program}
15452 @opindex print-prog-name
15453 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
15455 @item -print-libgcc-file-name
15456 @opindex print-libgcc-file-name
15457 Same as @option{-print-file-name=libgcc.a}.
15459 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
15460 but you do want to link with @file{libgcc.a}. You can do:
15463 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
15466 @item -print-search-dirs
15467 @opindex print-search-dirs
15468 Print the name of the configured installation directory and a list of
15469 program and library directories @command{gcc} searches---and don't do anything else.
15471 This is useful when @command{gcc} prints the error message
15472 @samp{installation problem, cannot exec cpp0: No such file or directory}.
15473 To resolve this you either need to put @file{cpp0} and the other compiler
15474 components where @command{gcc} expects to find them, or you can set the environment
15475 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
15476 Don't forget the trailing @samp{/}.
15477 @xref{Environment Variables}.
15479 @item -print-sysroot
15480 @opindex print-sysroot
15481 Print the target sysroot directory that is used during
15482 compilation. This is the target sysroot specified either at configure
15483 time or using the @option{--sysroot} option, possibly with an extra
15484 suffix that depends on compilation options. If no target sysroot is
15485 specified, the option prints nothing.
15487 @item -print-sysroot-headers-suffix
15488 @opindex print-sysroot-headers-suffix
15489 Print the suffix added to the target sysroot when searching for
15490 headers, or give an error if the compiler is not configured with such
15491 a suffix---and don't do anything else.
15494 @opindex dumpmachine
15495 Print the compiler's target machine (for example,
15496 @samp{i686-pc-linux-gnu})---and don't do anything else.
15499 @opindex dumpversion
15500 Print the compiler version (for example, @code{3.0}, @code{6.3.0} or @code{7})---and don't do
15501 anything else. This is the compiler version used in filesystem paths and
15502 specs. Depending on how the compiler has been configured it can be just
15503 a single number (major version), two numbers separated by a dot (major and
15504 minor version) or three numbers separated by dots (major, minor and patchlevel
15507 @item -dumpfullversion
15508 @opindex dumpfullversion
15509 Print the full compiler version---and don't do anything else. The output is
15510 always three numbers separated by dots, major, minor and patchlevel version.
15514 Print the compiler's built-in specs---and don't do anything else. (This
15515 is used when GCC itself is being built.) @xref{Spec Files}.
15518 @node Submodel Options
15519 @section Machine-Dependent Options
15520 @cindex submodel options
15521 @cindex specifying hardware config
15522 @cindex hardware models and configurations, specifying
15523 @cindex target-dependent options
15524 @cindex machine-dependent options
15526 Each target machine supported by GCC can have its own options---for
15527 example, to allow you to compile for a particular processor variant or
15528 ABI, or to control optimizations specific to that machine. By
15529 convention, the names of machine-specific options start with
15532 Some configurations of the compiler also support additional target-specific
15533 options, usually for compatibility with other compilers on the same
15536 @c This list is ordered alphanumerically by subsection name.
15537 @c It should be the same order and spelling as these options are listed
15538 @c in Machine Dependent Options
15541 * AArch64 Options::
15542 * Adapteva Epiphany Options::
15543 * AMD GCN Options::
15547 * Blackfin Options::
15553 * DEC Alpha Options::
15557 * GNU/Linux Options::
15567 * MicroBlaze Options::
15570 * MN10300 Options::
15574 * Nios II Options::
15575 * Nvidia PTX Options::
15576 * OpenRISC Options::
15578 * picoChip Options::
15579 * PowerPC Options::
15582 * RS/6000 and PowerPC Options::
15584 * S/390 and zSeries Options::
15587 * Solaris 2 Options::
15590 * System V Options::
15591 * TILE-Gx Options::
15592 * TILEPro Options::
15597 * VxWorks Options::
15599 * x86 Windows Options::
15600 * Xstormy16 Options::
15602 * zSeries Options::
15605 @node AArch64 Options
15606 @subsection AArch64 Options
15607 @cindex AArch64 Options
15609 These options are defined for AArch64 implementations:
15613 @item -mabi=@var{name}
15615 Generate code for the specified data model. Permissible values
15616 are @samp{ilp32} for SysV-like data model where int, long int and pointers
15617 are 32 bits, and @samp{lp64} for SysV-like data model where int is 32 bits,
15618 but long int and pointers are 64 bits.
15620 The default depends on the specific target configuration. Note that
15621 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
15622 entire program with the same ABI, and link with a compatible set of libraries.
15625 @opindex mbig-endian
15626 Generate big-endian code. This is the default when GCC is configured for an
15627 @samp{aarch64_be-*-*} target.
15629 @item -mgeneral-regs-only
15630 @opindex mgeneral-regs-only
15631 Generate code which uses only the general-purpose registers. This will prevent
15632 the compiler from using floating-point and Advanced SIMD registers but will not
15633 impose any restrictions on the assembler.
15635 @item -mlittle-endian
15636 @opindex mlittle-endian
15637 Generate little-endian code. This is the default when GCC is configured for an
15638 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
15640 @item -mcmodel=tiny
15641 @opindex mcmodel=tiny
15642 Generate code for the tiny code model. The program and its statically defined
15643 symbols must be within 1MB of each other. Programs can be statically or
15644 dynamically linked.
15646 @item -mcmodel=small
15647 @opindex mcmodel=small
15648 Generate code for the small code model. The program and its statically defined
15649 symbols must be within 4GB of each other. Programs can be statically or
15650 dynamically linked. This is the default code model.
15652 @item -mcmodel=large
15653 @opindex mcmodel=large
15654 Generate code for the large code model. This makes no assumptions about
15655 addresses and sizes of sections. Programs can be statically linked only.
15657 @item -mstrict-align
15658 @itemx -mno-strict-align
15659 @opindex mstrict-align
15660 @opindex mno-strict-align
15661 Avoid or allow generating memory accesses that may not be aligned on a natural
15662 object boundary as described in the architecture specification.
15664 @item -momit-leaf-frame-pointer
15665 @itemx -mno-omit-leaf-frame-pointer
15666 @opindex momit-leaf-frame-pointer
15667 @opindex mno-omit-leaf-frame-pointer
15668 Omit or keep the frame pointer in leaf functions. The former behavior is the
15671 @item -mstack-protector-guard=@var{guard}
15672 @itemx -mstack-protector-guard-reg=@var{reg}
15673 @itemx -mstack-protector-guard-offset=@var{offset}
15674 @opindex mstack-protector-guard
15675 @opindex mstack-protector-guard-reg
15676 @opindex mstack-protector-guard-offset
15677 Generate stack protection code using canary at @var{guard}. Supported
15678 locations are @samp{global} for a global canary or @samp{sysreg} for a
15679 canary in an appropriate system register.
15681 With the latter choice the options
15682 @option{-mstack-protector-guard-reg=@var{reg}} and
15683 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
15684 which system register to use as base register for reading the canary,
15685 and from what offset from that base register. There is no default
15686 register or offset as this is entirely for use within the Linux
15689 @item -mstack-protector-guard=@var{guard}
15690 @itemx -mstack-protector-guard-reg=@var{reg}
15691 @itemx -mstack-protector-guard-offset=@var{offset}
15692 @opindex mstack-protector-guard
15693 @opindex mstack-protector-guard-reg
15694 @opindex mstack-protector-guard-offset
15695 Generate stack protection code using canary at @var{guard}. Supported
15696 locations are @samp{global} for a global canary or @samp{sysreg} for a
15697 canary in an appropriate system register.
15699 With the latter choice the options
15700 @option{-mstack-protector-guard-reg=@var{reg}} and
15701 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
15702 which system register to use as base register for reading the canary,
15703 and from what offset from that base register. There is no default
15704 register or offset as this is entirely for use within the Linux
15707 @item -mtls-dialect=desc
15708 @opindex mtls-dialect=desc
15709 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
15710 of TLS variables. This is the default.
15712 @item -mtls-dialect=traditional
15713 @opindex mtls-dialect=traditional
15714 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
15717 @item -mtls-size=@var{size}
15719 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
15720 This option requires binutils 2.26 or newer.
15722 @item -mfix-cortex-a53-835769
15723 @itemx -mno-fix-cortex-a53-835769
15724 @opindex mfix-cortex-a53-835769
15725 @opindex mno-fix-cortex-a53-835769
15726 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
15727 This involves inserting a NOP instruction between memory instructions and
15728 64-bit integer multiply-accumulate instructions.
15730 @item -mfix-cortex-a53-843419
15731 @itemx -mno-fix-cortex-a53-843419
15732 @opindex mfix-cortex-a53-843419
15733 @opindex mno-fix-cortex-a53-843419
15734 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
15735 This erratum workaround is made at link time and this will only pass the
15736 corresponding flag to the linker.
15738 @item -mlow-precision-recip-sqrt
15739 @itemx -mno-low-precision-recip-sqrt
15740 @opindex mlow-precision-recip-sqrt
15741 @opindex mno-low-precision-recip-sqrt
15742 Enable or disable the reciprocal square root approximation.
15743 This option only has an effect if @option{-ffast-math} or
15744 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
15745 precision of reciprocal square root results to about 16 bits for
15746 single precision and to 32 bits for double precision.
15748 @item -mlow-precision-sqrt
15749 @itemx -mno-low-precision-sqrt
15750 @opindex mlow-precision-sqrt
15751 @opindex mno-low-precision-sqrt
15752 Enable or disable the square root approximation.
15753 This option only has an effect if @option{-ffast-math} or
15754 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
15755 precision of square root results to about 16 bits for
15756 single precision and to 32 bits for double precision.
15757 If enabled, it implies @option{-mlow-precision-recip-sqrt}.
15759 @item -mlow-precision-div
15760 @itemx -mno-low-precision-div
15761 @opindex mlow-precision-div
15762 @opindex mno-low-precision-div
15763 Enable or disable the division approximation.
15764 This option only has an effect if @option{-ffast-math} or
15765 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
15766 precision of division results to about 16 bits for
15767 single precision and to 32 bits for double precision.
15769 @item -mtrack-speculation
15770 @itemx -mno-track-speculation
15771 Enable or disable generation of additional code to track speculative
15772 execution through conditional branches. The tracking state can then
15773 be used by the compiler when expanding calls to
15774 @code{__builtin_speculation_safe_copy} to permit a more efficient code
15775 sequence to be generated.
15777 @item -march=@var{name}
15779 Specify the name of the target architecture and, optionally, one or
15780 more feature modifiers. This option has the form
15781 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
15783 The permissible values for @var{arch} are @samp{armv8-a},
15784 @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a}, @samp{armv8.4-a},
15785 @samp{armv8.5-a} or @var{native}.
15787 The value @samp{armv8.5-a} implies @samp{armv8.4-a} and enables compiler
15788 support for the ARMv8.5-A architecture extensions.
15790 The value @samp{armv8.4-a} implies @samp{armv8.3-a} and enables compiler
15791 support for the ARMv8.4-A architecture extensions.
15793 The value @samp{armv8.3-a} implies @samp{armv8.2-a} and enables compiler
15794 support for the ARMv8.3-A architecture extensions.
15796 The value @samp{armv8.2-a} implies @samp{armv8.1-a} and enables compiler
15797 support for the ARMv8.2-A architecture extensions.
15799 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
15800 support for the ARMv8.1-A architecture extension. In particular, it
15801 enables the @samp{+crc}, @samp{+lse}, and @samp{+rdma} features.
15803 The value @samp{native} is available on native AArch64 GNU/Linux and
15804 causes the compiler to pick the architecture of the host system. This
15805 option has no effect if the compiler is unable to recognize the
15806 architecture of the host system,
15808 The permissible values for @var{feature} are listed in the sub-section
15809 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
15810 Feature Modifiers}. Where conflicting feature modifiers are
15811 specified, the right-most feature is used.
15813 GCC uses @var{name} to determine what kind of instructions it can emit
15814 when generating assembly code. If @option{-march} is specified
15815 without either of @option{-mtune} or @option{-mcpu} also being
15816 specified, the code is tuned to perform well across a range of target
15817 processors implementing the target architecture.
15819 @item -mtune=@var{name}
15821 Specify the name of the target processor for which GCC should tune the
15822 performance of the code. Permissible values for this option are:
15823 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
15824 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
15825 @samp{cortex-a76}, @samp{ares}, @samp{exynos-m1}, @samp{emag}, @samp{falkor},
15826 @samp{neoverse-e1},@samp{neoverse-n1},@samp{qdf24xx}, @samp{saphira},
15827 @samp{phecda}, @samp{xgene1}, @samp{vulcan}, @samp{octeontx},
15828 @samp{octeontx81}, @samp{octeontx83}, @samp{thunderx}, @samp{thunderxt88},
15829 @samp{thunderxt88p1}, @samp{thunderxt81}, @samp{tsv110},
15830 @samp{thunderxt83}, @samp{thunderx2t99},
15831 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15832 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53},
15833 @samp{cortex-a75.cortex-a55}, @samp{cortex-a76.cortex-a55}
15836 The values @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15837 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53},
15838 @samp{cortex-a75.cortex-a55}, @samp{cortex-a76.cortex-a55} specify that GCC
15839 should tune for a big.LITTLE system.
15841 Additionally on native AArch64 GNU/Linux systems the value
15842 @samp{native} tunes performance to the host system. This option has no effect
15843 if the compiler is unable to recognize the processor of the host system.
15845 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
15846 are specified, the code is tuned to perform well across a range
15847 of target processors.
15849 This option cannot be suffixed by feature modifiers.
15851 @item -mcpu=@var{name}
15853 Specify the name of the target processor, optionally suffixed by one
15854 or more feature modifiers. This option has the form
15855 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
15856 the permissible values for @var{cpu} are the same as those available
15857 for @option{-mtune}. The permissible values for @var{feature} are
15858 documented in the sub-section on
15859 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
15860 Feature Modifiers}. Where conflicting feature modifiers are
15861 specified, the right-most feature is used.
15863 GCC uses @var{name} to determine what kind of instructions it can emit when
15864 generating assembly code (as if by @option{-march}) and to determine
15865 the target processor for which to tune for performance (as if
15866 by @option{-mtune}). Where this option is used in conjunction
15867 with @option{-march} or @option{-mtune}, those options take precedence
15868 over the appropriate part of this option.
15870 @item -moverride=@var{string}
15872 Override tuning decisions made by the back-end in response to a
15873 @option{-mtune=} switch. The syntax, semantics, and accepted values
15874 for @var{string} in this option are not guaranteed to be consistent
15877 This option is only intended to be useful when developing GCC.
15879 @item -mverbose-cost-dump
15880 @opindex mverbose-cost-dump
15881 Enable verbose cost model dumping in the debug dump files. This option is
15882 provided for use in debugging the compiler.
15884 @item -mpc-relative-literal-loads
15885 @itemx -mno-pc-relative-literal-loads
15886 @opindex mpc-relative-literal-loads
15887 @opindex mno-pc-relative-literal-loads
15888 Enable or disable PC-relative literal loads. With this option literal pools are
15889 accessed using a single instruction and emitted after each function. This
15890 limits the maximum size of functions to 1MB. This is enabled by default for
15891 @option{-mcmodel=tiny}.
15893 @item -msign-return-address=@var{scope}
15894 @opindex msign-return-address
15895 Select the function scope on which return address signing will be applied.
15896 Permissible values are @samp{none}, which disables return address signing,
15897 @samp{non-leaf}, which enables pointer signing for functions which are not leaf
15898 functions, and @samp{all}, which enables pointer signing for all functions. The
15899 default value is @samp{none}. This option has been deprecated by
15900 -mbranch-protection.
15902 @item -mbranch-protection=@var{none}|@var{standard}|@var{pac-ret}[+@var{leaf}+@var{b-key}]|@var{bti}
15903 @opindex mbranch-protection
15904 Select the branch protection features to use.
15905 @samp{none} is the default and turns off all types of branch protection.
15906 @samp{standard} turns on all types of branch protection features. If a feature
15907 has additional tuning options, then @samp{standard} sets it to its standard
15909 @samp{pac-ret[+@var{leaf}]} turns on return address signing to its standard
15910 level: signing functions that save the return address to memory (non-leaf
15911 functions will practically always do this) using the a-key. The optional
15912 argument @samp{leaf} can be used to extend the signing to include leaf
15913 functions. The optional argument @samp{b-key} can be used to sign the functions
15914 with the B-key instead of the A-key.
15915 @samp{bti} turns on branch target identification mechanism.
15917 @item -msve-vector-bits=@var{bits}
15918 @opindex msve-vector-bits
15919 Specify the number of bits in an SVE vector register. This option only has
15920 an effect when SVE is enabled.
15922 GCC supports two forms of SVE code generation: ``vector-length
15923 agnostic'' output that works with any size of vector register and
15924 ``vector-length specific'' output that allows GCC to make assumptions
15925 about the vector length when it is useful for optimization reasons.
15926 The possible values of @samp{bits} are: @samp{scalable}, @samp{128},
15927 @samp{256}, @samp{512}, @samp{1024} and @samp{2048}.
15928 Specifying @samp{scalable} selects vector-length agnostic
15929 output. At present @samp{-msve-vector-bits=128} also generates vector-length
15930 agnostic output. All other values generate vector-length specific code.
15931 The behavior of these values may change in future releases and no value except
15932 @samp{scalable} should be relied on for producing code that is portable across
15933 different hardware SVE vector lengths.
15935 The default is @samp{-msve-vector-bits=scalable}, which produces
15936 vector-length agnostic code.
15939 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
15940 @anchor{aarch64-feature-modifiers}
15941 @cindex @option{-march} feature modifiers
15942 @cindex @option{-mcpu} feature modifiers
15943 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
15944 the following and their inverses @option{no@var{feature}}:
15948 Enable CRC extension. This is on by default for
15949 @option{-march=armv8.1-a}.
15951 Enable Crypto extension. This also enables Advanced SIMD and floating-point
15954 Enable floating-point instructions. This is on by default for all possible
15955 values for options @option{-march} and @option{-mcpu}.
15957 Enable Advanced SIMD instructions. This also enables floating-point
15958 instructions. This is on by default for all possible values for options
15959 @option{-march} and @option{-mcpu}.
15961 Enable Scalable Vector Extension instructions. This also enables Advanced
15962 SIMD and floating-point instructions.
15964 Enable Large System Extension instructions. This is on by default for
15965 @option{-march=armv8.1-a}.
15967 Enable Round Double Multiply Accumulate instructions. This is on by default
15968 for @option{-march=armv8.1-a}.
15970 Enable FP16 extension. This also enables floating-point instructions.
15972 Enable FP16 fmla extension. This also enables FP16 extensions and
15973 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.
15976 Enable the RcPc extension. This does not change code generation from GCC,
15977 but is passed on to the assembler, enabling inline asm statements to use
15978 instructions from the RcPc extension.
15980 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
15982 Enable the Armv8-a aes and pmull crypto extension. This also enables Advanced
15985 Enable the Armv8-a sha2 crypto extension. This also enables Advanced SIMD instructions.
15987 Enable the sha512 and sha3 crypto extension. This also enables Advanced SIMD
15988 instructions. Use of this option with architectures prior to Armv8.2-A is not supported.
15990 Enable the sm3 and sm4 crypto extension. This also enables Advanced SIMD instructions.
15991 Use of this option with architectures prior to Armv8.2-A is not supported.
15993 Enable the Statistical Profiling extension. This option is only to enable the
15994 extension at the assembler level and does not affect code generation.
15996 Enable the Armv8.5-a Random Number instructions. This option is only to
15997 enable the extension at the assembler level and does not affect code
16000 Enable the Armv8.5-a Memory Tagging Extensions. This option is only to
16001 enable the extension at the assembler level and does not affect code
16004 Enable the Armv8-a Speculation Barrier instruction. This option is only to
16005 enable the extension at the assembler level and does not affect code
16006 generation. This option is enabled by default for @option{-march=armv8.5-a}.
16008 Enable the Armv8-a Speculative Store Bypass Safe instruction. This option
16009 is only to enable the extension at the assembler level and does not affect code
16010 generation. This option is enabled by default for @option{-march=armv8.5-a}.
16012 Enable the Armv8-a Execution and Data Prediction Restriction instructions.
16013 This option is only to enable the extension at the assembler level and does
16014 not affect code generation. This option is enabled by default for
16016 Enable the Armv8-a Scalable Vector Extension 2. This also enables SVE
16019 Enable SVE2 bitperm instructions. This also enables SVE2 instructions.
16021 Enable SVE2 sm4 instructions. This also enables SVE2 instructions.
16023 Enable SVE2 aes instructions. This also enables SVE2 instructions.
16025 Enable SVE2 sha3 instructions. This also enables SVE2 instructions.
16026 @option{-march=armv8.5-a}.
16030 Feature @option{crypto} implies @option{aes}, @option{sha2}, and @option{simd},
16031 which implies @option{fp}.
16032 Conversely, @option{nofp} implies @option{nosimd}, which implies
16033 @option{nocrypto}, @option{noaes} and @option{nosha2}.
16035 @node Adapteva Epiphany Options
16036 @subsection Adapteva Epiphany Options
16038 These @samp{-m} options are defined for Adapteva Epiphany:
16041 @item -mhalf-reg-file
16042 @opindex mhalf-reg-file
16043 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
16044 That allows code to run on hardware variants that lack these registers.
16046 @item -mprefer-short-insn-regs
16047 @opindex mprefer-short-insn-regs
16048 Preferentially allocate registers that allow short instruction generation.
16049 This can result in increased instruction count, so this may either reduce or
16050 increase overall code size.
16052 @item -mbranch-cost=@var{num}
16053 @opindex mbranch-cost
16054 Set the cost of branches to roughly @var{num} ``simple'' instructions.
16055 This cost is only a heuristic and is not guaranteed to produce
16056 consistent results across releases.
16060 Enable the generation of conditional moves.
16062 @item -mnops=@var{num}
16064 Emit @var{num} NOPs before every other generated instruction.
16066 @item -mno-soft-cmpsf
16067 @opindex mno-soft-cmpsf
16068 @opindex msoft-cmpsf
16069 For single-precision floating-point comparisons, emit an @code{fsub} instruction
16070 and test the flags. This is faster than a software comparison, but can
16071 get incorrect results in the presence of NaNs, or when two different small
16072 numbers are compared such that their difference is calculated as zero.
16073 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
16074 software comparisons.
16076 @item -mstack-offset=@var{num}
16077 @opindex mstack-offset
16078 Set the offset between the top of the stack and the stack pointer.
16079 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
16080 can be used by leaf functions without stack allocation.
16081 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
16082 Note also that this option changes the ABI; compiling a program with a
16083 different stack offset than the libraries have been compiled with
16084 generally does not work.
16085 This option can be useful if you want to evaluate if a different stack
16086 offset would give you better code, but to actually use a different stack
16087 offset to build working programs, it is recommended to configure the
16088 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
16090 @item -mno-round-nearest
16091 @opindex mno-round-nearest
16092 @opindex mround-nearest
16093 Make the scheduler assume that the rounding mode has been set to
16094 truncating. The default is @option{-mround-nearest}.
16097 @opindex mlong-calls
16098 If not otherwise specified by an attribute, assume all calls might be beyond
16099 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
16100 function address into a register before performing a (otherwise direct) call.
16101 This is the default.
16103 @item -mshort-calls
16104 @opindex short-calls
16105 If not otherwise specified by an attribute, assume all direct calls are
16106 in the range of the @code{b} / @code{bl} instructions, so use these instructions
16107 for direct calls. The default is @option{-mlong-calls}.
16111 Assume addresses can be loaded as 16-bit unsigned values. This does not
16112 apply to function addresses for which @option{-mlong-calls} semantics
16115 @item -mfp-mode=@var{mode}
16117 Set the prevailing mode of the floating-point unit.
16118 This determines the floating-point mode that is provided and expected
16119 at function call and return time. Making this mode match the mode you
16120 predominantly need at function start can make your programs smaller and
16121 faster by avoiding unnecessary mode switches.
16123 @var{mode} can be set to one the following values:
16127 Any mode at function entry is valid, and retained or restored when
16128 the function returns, and when it calls other functions.
16129 This mode is useful for compiling libraries or other compilation units
16130 you might want to incorporate into different programs with different
16131 prevailing FPU modes, and the convenience of being able to use a single
16132 object file outweighs the size and speed overhead for any extra
16133 mode switching that might be needed, compared with what would be needed
16134 with a more specific choice of prevailing FPU mode.
16137 This is the mode used for floating-point calculations with
16138 truncating (i.e.@: round towards zero) rounding mode. That includes
16139 conversion from floating point to integer.
16141 @item round-nearest
16142 This is the mode used for floating-point calculations with
16143 round-to-nearest-or-even rounding mode.
16146 This is the mode used to perform integer calculations in the FPU, e.g.@:
16147 integer multiply, or integer multiply-and-accumulate.
16150 The default is @option{-mfp-mode=caller}
16152 @item -mno-split-lohi
16153 @itemx -mno-postinc
16154 @itemx -mno-postmodify
16155 @opindex mno-split-lohi
16156 @opindex msplit-lohi
16157 @opindex mno-postinc
16159 @opindex mno-postmodify
16160 @opindex mpostmodify
16161 Code generation tweaks that disable, respectively, splitting of 32-bit
16162 loads, generation of post-increment addresses, and generation of
16163 post-modify addresses. The defaults are @option{msplit-lohi},
16164 @option{-mpost-inc}, and @option{-mpost-modify}.
16166 @item -mnovect-double
16167 @opindex mno-vect-double
16168 @opindex mvect-double
16169 Change the preferred SIMD mode to SImode. The default is
16170 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
16172 @item -max-vect-align=@var{num}
16173 @opindex max-vect-align
16174 The maximum alignment for SIMD vector mode types.
16175 @var{num} may be 4 or 8. The default is 8.
16176 Note that this is an ABI change, even though many library function
16177 interfaces are unaffected if they don't use SIMD vector modes
16178 in places that affect size and/or alignment of relevant types.
16180 @item -msplit-vecmove-early
16181 @opindex msplit-vecmove-early
16182 Split vector moves into single word moves before reload. In theory this
16183 can give better register allocation, but so far the reverse seems to be
16184 generally the case.
16186 @item -m1reg-@var{reg}
16188 Specify a register to hold the constant @minus{}1, which makes loading small negative
16189 constants and certain bitmasks faster.
16190 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
16191 which specify use of that register as a fixed register,
16192 and @samp{none}, which means that no register is used for this
16193 purpose. The default is @option{-m1reg-none}.
16197 @node AMD GCN Options
16198 @subsection AMD GCN Options
16199 @cindex AMD GCN Options
16201 These options are defined specifically for the AMD GCN port.
16205 @item -march=@var{gpu}
16207 @itemx -mtune=@var{gpu}
16209 Set architecture type or tuning for @var{gpu}. Supported values for @var{gpu}
16215 Compile for GCN3 Fiji devices (gfx803).
16218 Compile for GCN5 Vega 10 devices (gfx900).
16222 @item -mstack-size=@var{bytes}
16223 @opindex mstack-size
16224 Specify how many @var{bytes} of stack space will be requested for each GPU
16225 thread (wave-front). Beware that there may be many threads and limited memory
16226 available. The size of the stack allocation may also have an impact on
16227 run-time performance. The default is 32KB when using OpenACC or OpenMP, and
16233 @subsection ARC Options
16234 @cindex ARC options
16236 The following options control the architecture variant for which code
16239 @c architecture variants
16242 @item -mbarrel-shifter
16243 @opindex mbarrel-shifter
16244 Generate instructions supported by barrel shifter. This is the default
16245 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
16248 @opindex mjli-alawys
16249 Force to call a function using jli_s instruction. This option is
16250 valid only for ARCv2 architecture.
16252 @item -mcpu=@var{cpu}
16254 Set architecture type, register usage, and instruction scheduling
16255 parameters for @var{cpu}. There are also shortcut alias options
16256 available for backward compatibility and convenience. Supported
16257 values for @var{cpu} are
16263 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
16267 Compile for ARC601. Alias: @option{-mARC601}.
16272 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
16273 This is the default when configured with @option{--with-cpu=arc700}@.
16276 Compile for ARC EM.
16279 Compile for ARC HS.
16282 Compile for ARC EM CPU with no hardware extensions.
16285 Compile for ARC EM4 CPU.
16288 Compile for ARC EM4 DMIPS CPU.
16291 Compile for ARC EM4 DMIPS CPU with the single-precision floating-point
16295 Compile for ARC EM4 DMIPS CPU with single-precision floating-point and
16296 double assist instructions.
16299 Compile for ARC HS CPU with no hardware extensions except the atomic
16303 Compile for ARC HS34 CPU.
16306 Compile for ARC HS38 CPU.
16309 Compile for ARC HS38 CPU with all hardware extensions on.
16312 Compile for ARC 600 CPU with @code{norm} instructions enabled.
16314 @item arc600_mul32x16
16315 Compile for ARC 600 CPU with @code{norm} and 32x16-bit multiply
16316 instructions enabled.
16319 Compile for ARC 600 CPU with @code{norm} and @code{mul64}-family
16320 instructions enabled.
16323 Compile for ARC 601 CPU with @code{norm} instructions enabled.
16325 @item arc601_mul32x16
16326 Compile for ARC 601 CPU with @code{norm} and 32x16-bit multiply
16327 instructions enabled.
16330 Compile for ARC 601 CPU with @code{norm} and @code{mul64}-family
16331 instructions enabled.
16334 Compile for ARC 700 on NPS400 chip.
16337 Compile for ARC EM minimalist configuration featuring reduced register
16344 @itemx -mdpfp-compact
16345 @opindex mdpfp-compact
16346 Generate double-precision FPX instructions, tuned for the compact
16350 @opindex mdpfp-fast
16351 Generate double-precision FPX instructions, tuned for the fast
16354 @item -mno-dpfp-lrsr
16355 @opindex mno-dpfp-lrsr
16356 Disable @code{lr} and @code{sr} instructions from using FPX extension
16361 Generate extended arithmetic instructions. Currently only
16362 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
16363 supported. This is always enabled for @option{-mcpu=ARC700}.
16368 Do not generate @code{mpy}-family instructions for ARC700. This option is
16373 Generate 32x16-bit multiply and multiply-accumulate instructions.
16377 Generate @code{mul64} and @code{mulu64} instructions.
16378 Only valid for @option{-mcpu=ARC600}.
16382 Generate @code{norm} instructions. This is the default if @option{-mcpu=ARC700}
16387 @itemx -mspfp-compact
16388 @opindex mspfp-compact
16389 Generate single-precision FPX instructions, tuned for the compact
16393 @opindex mspfp-fast
16394 Generate single-precision FPX instructions, tuned for the fast
16399 Enable generation of ARC SIMD instructions via target-specific
16400 builtins. Only valid for @option{-mcpu=ARC700}.
16403 @opindex msoft-float
16404 This option ignored; it is provided for compatibility purposes only.
16405 Software floating-point code is emitted by default, and this default
16406 can overridden by FPX options; @option{-mspfp}, @option{-mspfp-compact}, or
16407 @option{-mspfp-fast} for single precision, and @option{-mdpfp},
16408 @option{-mdpfp-compact}, or @option{-mdpfp-fast} for double precision.
16412 Generate @code{swap} instructions.
16416 This enables use of the locked load/store conditional extension to implement
16417 atomic memory built-in functions. Not available for ARC 6xx or ARC
16422 Enable @code{div} and @code{rem} instructions for ARCv2 cores.
16424 @item -mcode-density
16425 @opindex mcode-density
16426 Enable code density instructions for ARC EM.
16427 This option is on by default for ARC HS.
16431 Enable double load/store operations for ARC HS cores.
16433 @item -mtp-regno=@var{regno}
16435 Specify thread pointer register number.
16437 @item -mmpy-option=@var{multo}
16438 @opindex mmpy-option
16439 Compile ARCv2 code with a multiplier design option. You can specify
16440 the option using either a string or numeric value for @var{multo}.
16441 @samp{wlh1} is the default value. The recognized values are:
16446 No multiplier available.
16450 16x16 multiplier, fully pipelined.
16451 The following instructions are enabled: @code{mpyw} and @code{mpyuw}.
16455 32x32 multiplier, fully
16456 pipelined (1 stage). The following instructions are additionally
16457 enabled: @code{mpy}, @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16461 32x32 multiplier, fully pipelined
16462 (2 stages). The following instructions are additionally enabled: @code{mpy},
16463 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16467 Two 16x16 multipliers, blocking,
16468 sequential. The following instructions are additionally enabled: @code{mpy},
16469 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16473 One 16x16 multiplier, blocking,
16474 sequential. The following instructions are additionally enabled: @code{mpy},
16475 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16479 One 32x4 multiplier, blocking,
16480 sequential. The following instructions are additionally enabled: @code{mpy},
16481 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16485 ARC HS SIMD support.
16489 ARC HS SIMD support.
16493 ARC HS SIMD support.
16497 This option is only available for ARCv2 cores@.
16499 @item -mfpu=@var{fpu}
16501 Enables support for specific floating-point hardware extensions for ARCv2
16502 cores. Supported values for @var{fpu} are:
16507 Enables support for single-precision floating-point hardware
16511 Enables support for double-precision floating-point hardware
16512 extensions. The single-precision floating-point extension is also
16513 enabled. Not available for ARC EM@.
16516 Enables support for double-precision floating-point hardware
16517 extensions using double-precision assist instructions. The single-precision
16518 floating-point extension is also enabled. This option is
16519 only available for ARC EM@.
16522 Enables support for double-precision floating-point hardware
16523 extensions using double-precision assist instructions.
16524 The single-precision floating-point, square-root, and divide
16525 extensions are also enabled. This option is
16526 only available for ARC EM@.
16529 Enables support for double-precision floating-point hardware
16530 extensions using double-precision assist instructions.
16531 The single-precision floating-point and fused multiply and add
16532 hardware extensions are also enabled. This option is
16533 only available for ARC EM@.
16536 Enables support for double-precision floating-point hardware
16537 extensions using double-precision assist instructions.
16538 All single-precision floating-point hardware extensions are also
16539 enabled. This option is only available for ARC EM@.
16542 Enables support for single-precision floating-point, square-root and divide
16543 hardware extensions@.
16546 Enables support for double-precision floating-point, square-root and divide
16547 hardware extensions. This option
16548 includes option @samp{fpus_div}. Not available for ARC EM@.
16551 Enables support for single-precision floating-point and
16552 fused multiply and add hardware extensions@.
16555 Enables support for double-precision floating-point and
16556 fused multiply and add hardware extensions. This option
16557 includes option @samp{fpus_fma}. Not available for ARC EM@.
16560 Enables support for all single-precision floating-point hardware
16564 Enables support for all single- and double-precision floating-point
16565 hardware extensions. Not available for ARC EM@.
16569 @item -mirq-ctrl-saved=@var{register-range}, @var{blink}, @var{lp_count}
16570 @opindex mirq-ctrl-saved
16571 Specifies general-purposes registers that the processor automatically
16572 saves/restores on interrupt entry and exit. @var{register-range} is
16573 specified as two registers separated by a dash. The register range
16574 always starts with @code{r0}, the upper limit is @code{fp} register.
16575 @var{blink} and @var{lp_count} are optional. This option is only
16576 valid for ARC EM and ARC HS cores.
16578 @item -mrgf-banked-regs=@var{number}
16579 @opindex mrgf-banked-regs
16580 Specifies the number of registers replicated in second register bank
16581 on entry to fast interrupt. Fast interrupts are interrupts with the
16582 highest priority level P0. These interrupts save only PC and STATUS32
16583 registers to avoid memory transactions during interrupt entry and exit
16584 sequences. Use this option when you are using fast interrupts in an
16585 ARC V2 family processor. Permitted values are 4, 8, 16, and 32.
16587 @item -mlpc-width=@var{width}
16588 @opindex mlpc-width
16589 Specify the width of the @code{lp_count} register. Valid values for
16590 @var{width} are 8, 16, 20, 24, 28 and 32 bits. The default width is
16591 fixed to 32 bits. If the width is less than 32, the compiler does not
16592 attempt to transform loops in your program to use the zero-delay loop
16593 mechanism unless it is known that the @code{lp_count} register can
16594 hold the required loop-counter value. Depending on the width
16595 specified, the compiler and run-time library might continue to use the
16596 loop mechanism for various needs. This option defines macro
16597 @code{__ARC_LPC_WIDTH__} with the value of @var{width}.
16601 This option instructs the compiler to generate code for a 16-entry
16602 register file. This option defines the @code{__ARC_RF16__}
16603 preprocessor macro.
16605 @item -mbranch-index
16606 @opindex mbranch-index
16607 Enable use of @code{bi} or @code{bih} instructions to implement jump
16612 The following options are passed through to the assembler, and also
16613 define preprocessor macro symbols.
16615 @c Flags used by the assembler, but for which we define preprocessor
16616 @c macro symbols as well.
16619 @opindex mdsp-packa
16620 Passed down to the assembler to enable the DSP Pack A extensions.
16621 Also sets the preprocessor symbol @code{__Xdsp_packa}. This option is
16626 Passed down to the assembler to enable the dual Viterbi butterfly
16627 extension. Also sets the preprocessor symbol @code{__Xdvbf}. This
16628 option is deprecated.
16630 @c ARC700 4.10 extension instruction
16633 Passed down to the assembler to enable the locked load/store
16634 conditional extension. Also sets the preprocessor symbol
16639 Passed down to the assembler. Also sets the preprocessor symbol
16640 @code{__Xxmac_d16}. This option is deprecated.
16644 Passed down to the assembler. Also sets the preprocessor symbol
16645 @code{__Xxmac_24}. This option is deprecated.
16647 @c ARC700 4.10 extension instruction
16650 Passed down to the assembler to enable the 64-bit time-stamp counter
16651 extension instruction. Also sets the preprocessor symbol
16652 @code{__Xrtsc}. This option is deprecated.
16654 @c ARC700 4.10 extension instruction
16657 Passed down to the assembler to enable the swap byte ordering
16658 extension instruction. Also sets the preprocessor symbol
16662 @opindex mtelephony
16663 Passed down to the assembler to enable dual- and single-operand
16664 instructions for telephony. Also sets the preprocessor symbol
16665 @code{__Xtelephony}. This option is deprecated.
16669 Passed down to the assembler to enable the XY memory extension. Also
16670 sets the preprocessor symbol @code{__Xxy}.
16674 The following options control how the assembly code is annotated:
16676 @c Assembly annotation options
16680 Annotate assembler instructions with estimated addresses.
16682 @item -mannotate-align
16683 @opindex mannotate-align
16684 Explain what alignment considerations lead to the decision to make an
16685 instruction short or long.
16689 The following options are passed through to the linker:
16691 @c options passed through to the linker
16695 Passed through to the linker, to specify use of the @code{arclinux} emulation.
16696 This option is enabled by default in tool chains built for
16697 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
16698 when profiling is not requested.
16700 @item -marclinux_prof
16701 @opindex marclinux_prof
16702 Passed through to the linker, to specify use of the
16703 @code{arclinux_prof} emulation. This option is enabled by default in
16704 tool chains built for @w{@code{arc-linux-uclibc}} and
16705 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
16709 The following options control the semantics of generated code:
16711 @c semantically relevant code generation options
16714 @opindex mlong-calls
16715 Generate calls as register indirect calls, thus providing access
16716 to the full 32-bit address range.
16718 @item -mmedium-calls
16719 @opindex mmedium-calls
16720 Don't use less than 25-bit addressing range for calls, which is the
16721 offset available for an unconditional branch-and-link
16722 instruction. Conditional execution of function calls is suppressed, to
16723 allow use of the 25-bit range, rather than the 21-bit range with
16724 conditional branch-and-link. This is the default for tool chains built
16725 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
16729 Put definitions of externally-visible data in a small data section if
16730 that data is no bigger than @var{num} bytes. The default value of
16731 @var{num} is 4 for any ARC configuration, or 8 when we have double
16732 load/store operations.
16737 Do not generate sdata references. This is the default for tool chains
16738 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
16741 @item -mvolatile-cache
16742 @opindex mvolatile-cache
16743 Use ordinarily cached memory accesses for volatile references. This is the
16746 @item -mno-volatile-cache
16747 @opindex mno-volatile-cache
16748 @opindex mvolatile-cache
16749 Enable cache bypass for volatile references.
16753 The following options fine tune code generation:
16754 @c code generation tuning options
16757 @opindex malign-call
16758 Do alignment optimizations for call instructions.
16760 @item -mauto-modify-reg
16761 @opindex mauto-modify-reg
16762 Enable the use of pre/post modify with register displacement.
16764 @item -mbbit-peephole
16765 @opindex mbbit-peephole
16766 Enable bbit peephole2.
16770 This option disables a target-specific pass in @file{arc_reorg} to
16771 generate compare-and-branch (@code{br@var{cc}}) instructions.
16772 It has no effect on
16773 generation of these instructions driven by the combiner pass.
16775 @item -mcase-vector-pcrel
16776 @opindex mcase-vector-pcrel
16777 Use PC-relative switch case tables to enable case table shortening.
16778 This is the default for @option{-Os}.
16780 @item -mcompact-casesi
16781 @opindex mcompact-casesi
16782 Enable compact @code{casesi} pattern. This is the default for @option{-Os},
16783 and only available for ARCv1 cores. This option is deprecated.
16785 @item -mno-cond-exec
16786 @opindex mno-cond-exec
16787 Disable the ARCompact-specific pass to generate conditional
16788 execution instructions.
16790 Due to delay slot scheduling and interactions between operand numbers,
16791 literal sizes, instruction lengths, and the support for conditional execution,
16792 the target-independent pass to generate conditional execution is often lacking,
16793 so the ARC port has kept a special pass around that tries to find more
16794 conditional execution generation opportunities after register allocation,
16795 branch shortening, and delay slot scheduling have been done. This pass
16796 generally, but not always, improves performance and code size, at the cost of
16797 extra compilation time, which is why there is an option to switch it off.
16798 If you have a problem with call instructions exceeding their allowable
16799 offset range because they are conditionalized, you should consider using
16800 @option{-mmedium-calls} instead.
16802 @item -mearly-cbranchsi
16803 @opindex mearly-cbranchsi
16804 Enable pre-reload use of the @code{cbranchsi} pattern.
16806 @item -mexpand-adddi
16807 @opindex mexpand-adddi
16808 Expand @code{adddi3} and @code{subdi3} at RTL generation time into
16809 @code{add.f}, @code{adc} etc. This option is deprecated.
16811 @item -mindexed-loads
16812 @opindex mindexed-loads
16813 Enable the use of indexed loads. This can be problematic because some
16814 optimizers then assume that indexed stores exist, which is not
16819 Enable Local Register Allocation. This is still experimental for ARC,
16820 so by default the compiler uses standard reload
16821 (i.e.@: @option{-mno-lra}).
16823 @item -mlra-priority-none
16824 @opindex mlra-priority-none
16825 Don't indicate any priority for target registers.
16827 @item -mlra-priority-compact
16828 @opindex mlra-priority-compact
16829 Indicate target register priority for r0..r3 / r12..r15.
16831 @item -mlra-priority-noncompact
16832 @opindex mlra-priority-noncompact
16833 Reduce target register priority for r0..r3 / r12..r15.
16836 @opindex mmillicode
16837 When optimizing for size (using @option{-Os}), prologues and epilogues
16838 that have to save or restore a large number of registers are often
16839 shortened by using call to a special function in libgcc; this is
16840 referred to as a @emph{millicode} call. As these calls can pose
16841 performance issues, and/or cause linking issues when linking in a
16842 nonstandard way, this option is provided to turn on or off millicode
16845 @item -mcode-density-frame
16846 @opindex mcode-density-frame
16847 This option enable the compiler to emit @code{enter} and @code{leave}
16848 instructions. These instructions are only valid for CPUs with
16849 code-density feature.
16852 @opindex mmixed-code
16853 Tweak register allocation to help 16-bit instruction generation.
16854 This generally has the effect of decreasing the average instruction size
16855 while increasing the instruction count.
16859 Enable @samp{q} instruction alternatives.
16860 This is the default for @option{-Os}.
16864 Enable @samp{Rcq} constraint handling.
16865 Most short code generation depends on this.
16866 This is the default.
16870 Enable @samp{Rcw} constraint handling.
16871 Most ccfsm condexec mostly depends on this.
16872 This is the default.
16874 @item -msize-level=@var{level}
16875 @opindex msize-level
16876 Fine-tune size optimization with regards to instruction lengths and alignment.
16877 The recognized values for @var{level} are:
16880 No size optimization. This level is deprecated and treated like @samp{1}.
16883 Short instructions are used opportunistically.
16886 In addition, alignment of loops and of code after barriers are dropped.
16889 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
16893 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
16894 the behavior when this is not set is equivalent to level @samp{1}.
16896 @item -mtune=@var{cpu}
16898 Set instruction scheduling parameters for @var{cpu}, overriding any implied
16899 by @option{-mcpu=}.
16901 Supported values for @var{cpu} are
16905 Tune for ARC600 CPU.
16908 Tune for ARC601 CPU.
16911 Tune for ARC700 CPU with standard multiplier block.
16914 Tune for ARC700 CPU with XMAC block.
16917 Tune for ARC725D CPU.
16920 Tune for ARC750D CPU.
16924 @item -mmultcost=@var{num}
16926 Cost to assume for a multiply instruction, with @samp{4} being equal to a
16927 normal instruction.
16929 @item -munalign-prob-threshold=@var{probability}
16930 @opindex munalign-prob-threshold
16931 Set probability threshold for unaligning branches.
16932 When tuning for @samp{ARC700} and optimizing for speed, branches without
16933 filled delay slot are preferably emitted unaligned and long, unless
16934 profiling indicates that the probability for the branch to be taken
16935 is below @var{probability}. @xref{Cross-profiling}.
16936 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
16940 The following options are maintained for backward compatibility, but
16941 are now deprecated and will be removed in a future release:
16943 @c Deprecated options
16951 @opindex mbig-endian
16954 Compile code for big-endian targets. Use of these options is now
16955 deprecated. Big-endian code is supported by configuring GCC to build
16956 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets,
16957 for which big endian is the default.
16959 @item -mlittle-endian
16960 @opindex mlittle-endian
16963 Compile code for little-endian targets. Use of these options is now
16964 deprecated. Little-endian code is supported by configuring GCC to build
16965 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets,
16966 for which little endian is the default.
16968 @item -mbarrel_shifter
16969 @opindex mbarrel_shifter
16970 Replaced by @option{-mbarrel-shifter}.
16972 @item -mdpfp_compact
16973 @opindex mdpfp_compact
16974 Replaced by @option{-mdpfp-compact}.
16977 @opindex mdpfp_fast
16978 Replaced by @option{-mdpfp-fast}.
16981 @opindex mdsp_packa
16982 Replaced by @option{-mdsp-packa}.
16986 Replaced by @option{-mea}.
16990 Replaced by @option{-mmac-24}.
16994 Replaced by @option{-mmac-d16}.
16996 @item -mspfp_compact
16997 @opindex mspfp_compact
16998 Replaced by @option{-mspfp-compact}.
17001 @opindex mspfp_fast
17002 Replaced by @option{-mspfp-fast}.
17004 @item -mtune=@var{cpu}
17006 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
17007 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
17008 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively.
17010 @item -multcost=@var{num}
17012 Replaced by @option{-mmultcost}.
17017 @subsection ARM Options
17018 @cindex ARM options
17020 These @samp{-m} options are defined for the ARM port:
17023 @item -mabi=@var{name}
17025 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
17026 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
17029 @opindex mapcs-frame
17030 Generate a stack frame that is compliant with the ARM Procedure Call
17031 Standard for all functions, even if this is not strictly necessary for
17032 correct execution of the code. Specifying @option{-fomit-frame-pointer}
17033 with this option causes the stack frames not to be generated for
17034 leaf functions. The default is @option{-mno-apcs-frame}.
17035 This option is deprecated.
17039 This is a synonym for @option{-mapcs-frame} and is deprecated.
17042 @c not currently implemented
17043 @item -mapcs-stack-check
17044 @opindex mapcs-stack-check
17045 Generate code to check the amount of stack space available upon entry to
17046 every function (that actually uses some stack space). If there is
17047 insufficient space available then either the function
17048 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
17049 called, depending upon the amount of stack space required. The runtime
17050 system is required to provide these functions. The default is
17051 @option{-mno-apcs-stack-check}, since this produces smaller code.
17053 @c not currently implemented
17054 @item -mapcs-reentrant
17055 @opindex mapcs-reentrant
17056 Generate reentrant, position-independent code. The default is
17057 @option{-mno-apcs-reentrant}.
17060 @item -mthumb-interwork
17061 @opindex mthumb-interwork
17062 Generate code that supports calling between the ARM and Thumb
17063 instruction sets. Without this option, on pre-v5 architectures, the
17064 two instruction sets cannot be reliably used inside one program. The
17065 default is @option{-mno-thumb-interwork}, since slightly larger code
17066 is generated when @option{-mthumb-interwork} is specified. In AAPCS
17067 configurations this option is meaningless.
17069 @item -mno-sched-prolog
17070 @opindex mno-sched-prolog
17071 @opindex msched-prolog
17072 Prevent the reordering of instructions in the function prologue, or the
17073 merging of those instruction with the instructions in the function's
17074 body. This means that all functions start with a recognizable set
17075 of instructions (or in fact one of a choice from a small set of
17076 different function prologues), and this information can be used to
17077 locate the start of functions inside an executable piece of code. The
17078 default is @option{-msched-prolog}.
17080 @item -mfloat-abi=@var{name}
17081 @opindex mfloat-abi
17082 Specifies which floating-point ABI to use. Permissible values
17083 are: @samp{soft}, @samp{softfp} and @samp{hard}.
17085 Specifying @samp{soft} causes GCC to generate output containing
17086 library calls for floating-point operations.
17087 @samp{softfp} allows the generation of code using hardware floating-point
17088 instructions, but still uses the soft-float calling conventions.
17089 @samp{hard} allows generation of floating-point instructions
17090 and uses FPU-specific calling conventions.
17092 The default depends on the specific target configuration. Note that
17093 the hard-float and soft-float ABIs are not link-compatible; you must
17094 compile your entire program with the same ABI, and link with a
17095 compatible set of libraries.
17097 @item -mgeneral-regs-only
17098 @opindex mgeneral-regs-only
17099 Generate code which uses only the general-purpose registers. This will prevent
17100 the compiler from using floating-point and Advanced SIMD registers but will not
17101 impose any restrictions on the assembler.
17103 @item -mlittle-endian
17104 @opindex mlittle-endian
17105 Generate code for a processor running in little-endian mode. This is
17106 the default for all standard configurations.
17109 @opindex mbig-endian
17110 Generate code for a processor running in big-endian mode; the default is
17111 to compile code for a little-endian processor.
17116 When linking a big-endian image select between BE8 and BE32 formats.
17117 The option has no effect for little-endian images and is ignored. The
17118 default is dependent on the selected target architecture. For ARMv6
17119 and later architectures the default is BE8, for older architectures
17120 the default is BE32. BE32 format has been deprecated by ARM.
17122 @item -march=@var{name}@r{[}+extension@dots{}@r{]}
17124 This specifies the name of the target ARM architecture. GCC uses this
17125 name to determine what kind of instructions it can emit when generating
17126 assembly code. This option can be used in conjunction with or instead
17127 of the @option{-mcpu=} option.
17129 Permissible names are:
17131 @samp{armv5t}, @samp{armv5te},
17132 @samp{armv6}, @samp{armv6j}, @samp{armv6k}, @samp{armv6kz}, @samp{armv6t2},
17133 @samp{armv6z}, @samp{armv6zk},
17134 @samp{armv7}, @samp{armv7-a}, @samp{armv7ve},
17135 @samp{armv8-a}, @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a},
17140 @samp{armv6-m}, @samp{armv6s-m},
17141 @samp{armv7-m}, @samp{armv7e-m},
17142 @samp{armv8-m.base}, @samp{armv8-m.main},
17143 @samp{iwmmxt} and @samp{iwmmxt2}.
17145 Additionally, the following architectures, which lack support for the
17146 Thumb execution state, are recognized but support is deprecated: @samp{armv4}.
17148 Many of the architectures support extensions. These can be added by
17149 appending @samp{+@var{extension}} to the architecture name. Extension
17150 options are processed in order and capabilities accumulate. An extension
17151 will also enable any necessary base extensions
17152 upon which it depends. For example, the @samp{+crypto} extension
17153 will always enable the @samp{+simd} extension. The exception to the
17154 additive construction is for extensions that are prefixed with
17155 @samp{+no@dots{}}: these extensions disable the specified option and
17156 any other extensions that may depend on the presence of that
17159 For example, @samp{-march=armv7-a+simd+nofp+vfpv4} is equivalent to
17160 writing @samp{-march=armv7-a+vfpv4} since the @samp{+simd} option is
17161 entirely disabled by the @samp{+nofp} option that follows it.
17163 Most extension names are generically named, but have an effect that is
17164 dependent upon the architecture to which it is applied. For example,
17165 the @samp{+simd} option can be applied to both @samp{armv7-a} and
17166 @samp{armv8-a} architectures, but will enable the original ARMv7-A
17167 Advanced SIMD (Neon) extensions for @samp{armv7-a} and the ARMv8-A
17168 variant for @samp{armv8-a}.
17170 The table below lists the supported extensions for each architecture.
17171 Architectures not mentioned do not support any extensions.
17184 The VFPv2 floating-point instructions. The extension @samp{+vfpv2} can be
17185 used as an alias for this extension.
17188 Disable the floating-point instructions.
17192 The common subset of the ARMv7-A, ARMv7-R and ARMv7-M architectures.
17195 The VFPv3 floating-point instructions, with 16 double-precision
17196 registers. The extension @samp{+vfpv3-d16} can be used as an alias
17197 for this extension. Note that floating-point is not supported by the
17198 base ARMv7-M architecture, but is compatible with both the ARMv7-A and
17199 ARMv7-R architectures.
17202 Disable the floating-point instructions.
17208 The multiprocessing extension.
17211 The security extension.
17214 The VFPv3 floating-point instructions, with 16 double-precision
17215 registers. The extension @samp{+vfpv3-d16} can be used as an alias
17216 for this extension.
17219 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
17220 The extensions @samp{+neon} and @samp{+neon-vfpv3} can be used as aliases
17221 for this extension.
17224 The VFPv3 floating-point instructions, with 32 double-precision
17227 @item +vfpv3-d16-fp16
17228 The VFPv3 floating-point instructions, with 16 double-precision
17229 registers and the half-precision floating-point conversion operations.
17232 The VFPv3 floating-point instructions, with 32 double-precision
17233 registers and the half-precision floating-point conversion operations.
17236 The VFPv4 floating-point instructions, with 16 double-precision
17240 The VFPv4 floating-point instructions, with 32 double-precision
17244 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
17245 the half-precision floating-point conversion operations.
17248 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions.
17251 Disable the Advanced SIMD instructions (does not disable floating point).
17254 Disable the floating-point and Advanced SIMD instructions.
17258 The extended version of the ARMv7-A architecture with support for
17262 The VFPv4 floating-point instructions, with 16 double-precision registers.
17263 The extension @samp{+vfpv4-d16} can be used as an alias for this extension.
17266 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions. The
17267 extension @samp{+neon-vfpv4} can be used as an alias for this extension.
17270 The VFPv3 floating-point instructions, with 16 double-precision
17274 The VFPv3 floating-point instructions, with 32 double-precision
17277 @item +vfpv3-d16-fp16
17278 The VFPv3 floating-point instructions, with 16 double-precision
17279 registers and the half-precision floating-point conversion operations.
17282 The VFPv3 floating-point instructions, with 32 double-precision
17283 registers and the half-precision floating-point conversion operations.
17286 The VFPv4 floating-point instructions, with 16 double-precision
17290 The VFPv4 floating-point instructions, with 32 double-precision
17294 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
17295 The extension @samp{+neon-vfpv3} can be used as an alias for this extension.
17298 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
17299 the half-precision floating-point conversion operations.
17302 Disable the Advanced SIMD instructions (does not disable floating point).
17305 Disable the floating-point and Advanced SIMD instructions.
17311 The Cyclic Redundancy Check (CRC) instructions.
17313 The ARMv8-A Advanced SIMD and floating-point instructions.
17315 The cryptographic instructions.
17317 Disable the cryptographic instructions.
17319 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17321 Speculation Barrier Instruction.
17323 Execution and Data Prediction Restriction Instructions.
17329 The ARMv8.1-A Advanced SIMD and floating-point instructions.
17332 The cryptographic instructions. This also enables the Advanced SIMD and
17333 floating-point instructions.
17336 Disable the cryptographic instructions.
17339 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17342 Speculation Barrier Instruction.
17345 Execution and Data Prediction Restriction Instructions.
17352 The half-precision floating-point data processing instructions.
17353 This also enables the Advanced SIMD and floating-point instructions.
17356 The half-precision floating-point fmla extension. This also enables
17357 the half-precision floating-point extension and Advanced SIMD and
17358 floating-point instructions.
17361 The ARMv8.1-A Advanced SIMD and floating-point instructions.
17364 The cryptographic instructions. This also enables the Advanced SIMD and
17365 floating-point instructions.
17368 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
17371 Disable the cryptographic extension.
17374 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17377 Speculation Barrier Instruction.
17380 Execution and Data Prediction Restriction Instructions.
17386 The half-precision floating-point data processing instructions.
17387 This also enables the Advanced SIMD and floating-point instructions as well
17388 as the Dot Product extension and the half-precision floating-point fmla
17392 The ARMv8.3-A Advanced SIMD and floating-point instructions as well as the
17393 Dot Product extension.
17396 The cryptographic instructions. This also enables the Advanced SIMD and
17397 floating-point instructions as well as the Dot Product extension.
17400 Disable the cryptographic extension.
17403 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17406 Speculation Barrier Instruction.
17409 Execution and Data Prediction Restriction Instructions.
17415 The half-precision floating-point data processing instructions.
17416 This also enables the Advanced SIMD and floating-point instructions as well
17417 as the Dot Product extension and the half-precision floating-point fmla
17421 The ARMv8.3-A Advanced SIMD and floating-point instructions as well as the
17422 Dot Product extension.
17425 The cryptographic instructions. This also enables the Advanced SIMD and
17426 floating-point instructions as well as the Dot Product extension.
17429 Disable the cryptographic extension.
17432 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17438 The single-precision VFPv3 floating-point instructions. The extension
17439 @samp{+vfpv3xd} can be used as an alias for this extension.
17442 The VFPv3 floating-point instructions with 16 double-precision registers.
17443 The extension +vfpv3-d16 can be used as an alias for this extension.
17445 @item +vfpv3xd-d16-fp16
17446 The single-precision VFPv3 floating-point instructions with 16 double-precision
17447 registers and the half-precision floating-point conversion operations.
17449 @item +vfpv3-d16-fp16
17450 The VFPv3 floating-point instructions with 16 double-precision
17451 registers and the half-precision floating-point conversion operations.
17454 Disable the floating-point extension.
17457 The ARM-state integer division instructions.
17460 Disable the ARM-state integer division extension.
17466 The single-precision VFPv4 floating-point instructions.
17469 The single-precision FPv5 floating-point instructions.
17472 The single- and double-precision FPv5 floating-point instructions.
17475 Disable the floating-point extensions.
17481 The DSP instructions.
17484 Disable the DSP extension.
17487 The single-precision floating-point instructions.
17490 The single- and double-precision floating-point instructions.
17493 Disable the floating-point extension.
17499 The Cyclic Redundancy Check (CRC) instructions.
17501 The single-precision FPv5 floating-point instructions.
17503 The ARMv8-A Advanced SIMD and floating-point instructions.
17505 The cryptographic instructions.
17507 Disable the cryptographic instructions.
17509 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17514 @option{-march=native} causes the compiler to auto-detect the architecture
17515 of the build computer. At present, this feature is only supported on
17516 GNU/Linux, and not all architectures are recognized. If the auto-detect
17517 is unsuccessful the option has no effect.
17519 @item -mtune=@var{name}
17521 This option specifies the name of the target ARM processor for
17522 which GCC should tune the performance of the code.
17523 For some ARM implementations better performance can be obtained by using
17525 Permissible names are: @samp{arm7tdmi}, @samp{arm7tdmi-s}, @samp{arm710t},
17526 @samp{arm720t}, @samp{arm740t}, @samp{strongarm}, @samp{strongarm110},
17527 @samp{strongarm1100}, 0@samp{strongarm1110}, @samp{arm8}, @samp{arm810},
17528 @samp{arm9}, @samp{arm9e}, @samp{arm920}, @samp{arm920t}, @samp{arm922t},
17529 @samp{arm946e-s}, @samp{arm966e-s}, @samp{arm968e-s}, @samp{arm926ej-s},
17530 @samp{arm940t}, @samp{arm9tdmi}, @samp{arm10tdmi}, @samp{arm1020t},
17531 @samp{arm1026ej-s}, @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
17532 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
17533 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
17534 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
17535 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
17536 @samp{cortex-a32}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
17537 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
17538 @samp{cortex-a76}, @samp{ares}, @samp{cortex-r4}, @samp{cortex-r4f},
17539 @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8}, @samp{cortex-r52},
17540 @samp{cortex-m0}, @samp{cortex-m0plus}, @samp{cortex-m1}, @samp{cortex-m3},
17541 @samp{cortex-m4}, @samp{cortex-m7}, @samp{cortex-m23}, @samp{cortex-m33},
17542 @samp{cortex-m1.small-multiply}, @samp{cortex-m0.small-multiply},
17543 @samp{cortex-m0plus.small-multiply}, @samp{exynos-m1}, @samp{marvell-pj4},
17544 @samp{neoverse-n1}, @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2},
17545 @samp{ep9312}, @samp{fa526}, @samp{fa626}, @samp{fa606te}, @samp{fa626te},
17546 @samp{fmp626}, @samp{fa726te}, @samp{xgene1}.
17548 Additionally, this option can specify that GCC should tune the performance
17549 of the code for a big.LITTLE system. Permissible names are:
17550 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
17551 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
17552 @samp{cortex-a72.cortex-a35}, @samp{cortex-a73.cortex-a53},
17553 @samp{cortex-a75.cortex-a55}, @samp{cortex-a76.cortex-a55}.
17555 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
17556 performance for a blend of processors within architecture @var{arch}.
17557 The aim is to generate code that run well on the current most popular
17558 processors, balancing between optimizations that benefit some CPUs in the
17559 range, and avoiding performance pitfalls of other CPUs. The effects of
17560 this option may change in future GCC versions as CPU models come and go.
17562 @option{-mtune} permits the same extension options as @option{-mcpu}, but
17563 the extension options do not affect the tuning of the generated code.
17565 @option{-mtune=native} causes the compiler to auto-detect the CPU
17566 of the build computer. At present, this feature is only supported on
17567 GNU/Linux, and not all architectures are recognized. If the auto-detect is
17568 unsuccessful the option has no effect.
17570 @item -mcpu=@var{name}@r{[}+extension@dots{}@r{]}
17572 This specifies the name of the target ARM processor. GCC uses this name
17573 to derive the name of the target ARM architecture (as if specified
17574 by @option{-march}) and the ARM processor type for which to tune for
17575 performance (as if specified by @option{-mtune}). Where this option
17576 is used in conjunction with @option{-march} or @option{-mtune},
17577 those options take precedence over the appropriate part of this option.
17579 Many of the supported CPUs implement optional architectural
17580 extensions. Where this is so the architectural extensions are
17581 normally enabled by default. If implementations that lack the
17582 extension exist, then the extension syntax can be used to disable
17583 those extensions that have been omitted. For floating-point and
17584 Advanced SIMD (Neon) instructions, the settings of the options
17585 @option{-mfloat-abi} and @option{-mfpu} must also be considered:
17586 floating-point and Advanced SIMD instructions will only be used if
17587 @option{-mfloat-abi} is not set to @samp{soft}; and any setting of
17588 @option{-mfpu} other than @samp{auto} will override the available
17589 floating-point and SIMD extension instructions.
17591 For example, @samp{cortex-a9} can be found in three major
17592 configurations: integer only, with just a floating-point unit or with
17593 floating-point and Advanced SIMD. The default is to enable all the
17594 instructions, but the extensions @samp{+nosimd} and @samp{+nofp} can
17595 be used to disable just the SIMD or both the SIMD and floating-point
17596 instructions respectively.
17598 Permissible names for this option are the same as those for
17601 The following extension options are common to the listed CPUs:
17605 Disable the DSP instructions on @samp{cortex-m33}.
17608 Disables the floating-point instructions on @samp{arm9e},
17609 @samp{arm946e-s}, @samp{arm966e-s}, @samp{arm968e-s}, @samp{arm10e},
17610 @samp{arm1020e}, @samp{arm1022e}, @samp{arm926ej-s},
17611 @samp{arm1026ej-s}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8},
17612 @samp{cortex-m4}, @samp{cortex-m7} and @samp{cortex-m33}.
17613 Disables the floating-point and SIMD instructions on
17614 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7},
17615 @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a12},
17616 @samp{cortex-a15}, @samp{cortex-a17}, @samp{cortex-a15.cortex-a7},
17617 @samp{cortex-a17.cortex-a7}, @samp{cortex-a32}, @samp{cortex-a35},
17618 @samp{cortex-a53} and @samp{cortex-a55}.
17621 Disables the double-precision component of the floating-point instructions
17622 on @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8}, @samp{cortex-r52} and
17626 Disables the SIMD (but not floating-point) instructions on
17627 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}
17628 and @samp{cortex-a9}.
17631 Enables the cryptographic instructions on @samp{cortex-a32},
17632 @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55}, @samp{cortex-a57},
17633 @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75}, @samp{exynos-m1},
17634 @samp{xgene1}, @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
17635 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53} and
17636 @samp{cortex-a75.cortex-a55}.
17639 Additionally the @samp{generic-armv7-a} pseudo target defaults to
17640 VFPv3 with 16 double-precision registers. It supports the following
17641 extension options: @samp{mp}, @samp{sec}, @samp{vfpv3-d16},
17642 @samp{vfpv3}, @samp{vfpv3-d16-fp16}, @samp{vfpv3-fp16},
17643 @samp{vfpv4-d16}, @samp{vfpv4}, @samp{neon}, @samp{neon-vfpv3},
17644 @samp{neon-fp16}, @samp{neon-vfpv4}. The meanings are the same as for
17645 the extensions to @option{-march=armv7-a}.
17647 @option{-mcpu=generic-@var{arch}} is also permissible, and is
17648 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
17649 See @option{-mtune} for more information.
17651 @option{-mcpu=native} causes the compiler to auto-detect the CPU
17652 of the build computer. At present, this feature is only supported on
17653 GNU/Linux, and not all architectures are recognized. If the auto-detect
17654 is unsuccessful the option has no effect.
17656 @item -mfpu=@var{name}
17658 This specifies what floating-point hardware (or hardware emulation) is
17659 available on the target. Permissible names are: @samp{auto}, @samp{vfpv2},
17661 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
17662 @samp{vfpv3xd-fp16}, @samp{neon-vfpv3}, @samp{neon-fp16}, @samp{vfpv4},
17663 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
17664 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
17665 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
17666 Note that @samp{neon} is an alias for @samp{neon-vfpv3} and @samp{vfp}
17667 is an alias for @samp{vfpv2}.
17669 The setting @samp{auto} is the default and is special. It causes the
17670 compiler to select the floating-point and Advanced SIMD instructions
17671 based on the settings of @option{-mcpu} and @option{-march}.
17673 If the selected floating-point hardware includes the NEON extension
17674 (e.g.@: @option{-mfpu=neon}), note that floating-point
17675 operations are not generated by GCC's auto-vectorization pass unless
17676 @option{-funsafe-math-optimizations} is also specified. This is
17677 because NEON hardware does not fully implement the IEEE 754 standard for
17678 floating-point arithmetic (in particular denormal values are treated as
17679 zero), so the use of NEON instructions may lead to a loss of precision.
17681 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}).
17683 @item -mfp16-format=@var{name}
17684 @opindex mfp16-format
17685 Specify the format of the @code{__fp16} half-precision floating-point type.
17686 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
17687 the default is @samp{none}, in which case the @code{__fp16} type is not
17688 defined. @xref{Half-Precision}, for more information.
17690 @item -mstructure-size-boundary=@var{n}
17691 @opindex mstructure-size-boundary
17692 The sizes of all structures and unions are rounded up to a multiple
17693 of the number of bits set by this option. Permissible values are 8, 32
17694 and 64. The default value varies for different toolchains. For the COFF
17695 targeted toolchain the default value is 8. A value of 64 is only allowed
17696 if the underlying ABI supports it.
17698 Specifying a larger number can produce faster, more efficient code, but
17699 can also increase the size of the program. Different values are potentially
17700 incompatible. Code compiled with one value cannot necessarily expect to
17701 work with code or libraries compiled with another value, if they exchange
17702 information using structures or unions.
17704 This option is deprecated.
17706 @item -mabort-on-noreturn
17707 @opindex mabort-on-noreturn
17708 Generate a call to the function @code{abort} at the end of a
17709 @code{noreturn} function. It is executed if the function tries to
17713 @itemx -mno-long-calls
17714 @opindex mlong-calls
17715 @opindex mno-long-calls
17716 Tells the compiler to perform function calls by first loading the
17717 address of the function into a register and then performing a subroutine
17718 call on this register. This switch is needed if the target function
17719 lies outside of the 64-megabyte addressing range of the offset-based
17720 version of subroutine call instruction.
17722 Even if this switch is enabled, not all function calls are turned
17723 into long calls. The heuristic is that static functions, functions
17724 that have the @code{short_call} attribute, functions that are inside
17725 the scope of a @code{#pragma no_long_calls} directive, and functions whose
17726 definitions have already been compiled within the current compilation
17727 unit are not turned into long calls. The exceptions to this rule are
17728 that weak function definitions, functions with the @code{long_call}
17729 attribute or the @code{section} attribute, and functions that are within
17730 the scope of a @code{#pragma long_calls} directive are always
17731 turned into long calls.
17733 This feature is not enabled by default. Specifying
17734 @option{-mno-long-calls} restores the default behavior, as does
17735 placing the function calls within the scope of a @code{#pragma
17736 long_calls_off} directive. Note these switches have no effect on how
17737 the compiler generates code to handle function calls via function
17740 @item -msingle-pic-base
17741 @opindex msingle-pic-base
17742 Treat the register used for PIC addressing as read-only, rather than
17743 loading it in the prologue for each function. The runtime system is
17744 responsible for initializing this register with an appropriate value
17745 before execution begins.
17747 @item -mpic-register=@var{reg}
17748 @opindex mpic-register
17749 Specify the register to be used for PIC addressing.
17750 For standard PIC base case, the default is any suitable register
17751 determined by compiler. For single PIC base case, the default is
17752 @samp{R9} if target is EABI based or stack-checking is enabled,
17753 otherwise the default is @samp{R10}.
17755 @item -mpic-data-is-text-relative
17756 @opindex mpic-data-is-text-relative
17757 Assume that the displacement between the text and data segments is fixed
17758 at static link time. This permits using PC-relative addressing
17759 operations to access data known to be in the data segment. For
17760 non-VxWorks RTP targets, this option is enabled by default. When
17761 disabled on such targets, it will enable @option{-msingle-pic-base} by
17764 @item -mpoke-function-name
17765 @opindex mpoke-function-name
17766 Write the name of each function into the text section, directly
17767 preceding the function prologue. The generated code is similar to this:
17771 .ascii "arm_poke_function_name", 0
17774 .word 0xff000000 + (t1 - t0)
17775 arm_poke_function_name
17777 stmfd sp!, @{fp, ip, lr, pc@}
17781 When performing a stack backtrace, code can inspect the value of
17782 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
17783 location @code{pc - 12} and the top 8 bits are set, then we know that
17784 there is a function name embedded immediately preceding this location
17785 and has length @code{((pc[-3]) & 0xff000000)}.
17792 Select between generating code that executes in ARM and Thumb
17793 states. The default for most configurations is to generate code
17794 that executes in ARM state, but the default can be changed by
17795 configuring GCC with the @option{--with-mode=}@var{state}
17798 You can also override the ARM and Thumb mode for each function
17799 by using the @code{target("thumb")} and @code{target("arm")} function attributes
17800 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
17803 @opindex mflip-thumb
17804 Switch ARM/Thumb modes on alternating functions.
17805 This option is provided for regression testing of mixed Thumb/ARM code
17806 generation, and is not intended for ordinary use in compiling code.
17809 @opindex mtpcs-frame
17810 Generate a stack frame that is compliant with the Thumb Procedure Call
17811 Standard for all non-leaf functions. (A leaf function is one that does
17812 not call any other functions.) The default is @option{-mno-tpcs-frame}.
17814 @item -mtpcs-leaf-frame
17815 @opindex mtpcs-leaf-frame
17816 Generate a stack frame that is compliant with the Thumb Procedure Call
17817 Standard for all leaf functions. (A leaf function is one that does
17818 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
17820 @item -mcallee-super-interworking
17821 @opindex mcallee-super-interworking
17822 Gives all externally visible functions in the file being compiled an ARM
17823 instruction set header which switches to Thumb mode before executing the
17824 rest of the function. This allows these functions to be called from
17825 non-interworking code. This option is not valid in AAPCS configurations
17826 because interworking is enabled by default.
17828 @item -mcaller-super-interworking
17829 @opindex mcaller-super-interworking
17830 Allows calls via function pointers (including virtual functions) to
17831 execute correctly regardless of whether the target code has been
17832 compiled for interworking or not. There is a small overhead in the cost
17833 of executing a function pointer if this option is enabled. This option
17834 is not valid in AAPCS configurations because interworking is enabled
17837 @item -mtp=@var{name}
17839 Specify the access model for the thread local storage pointer. The valid
17840 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
17841 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
17842 (supported in the arm6k architecture), and @samp{auto}, which uses the
17843 best available method for the selected processor. The default setting is
17846 @item -mtls-dialect=@var{dialect}
17847 @opindex mtls-dialect
17848 Specify the dialect to use for accessing thread local storage. Two
17849 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
17850 @samp{gnu} dialect selects the original GNU scheme for supporting
17851 local and global dynamic TLS models. The @samp{gnu2} dialect
17852 selects the GNU descriptor scheme, which provides better performance
17853 for shared libraries. The GNU descriptor scheme is compatible with
17854 the original scheme, but does require new assembler, linker and
17855 library support. Initial and local exec TLS models are unaffected by
17856 this option and always use the original scheme.
17858 @item -mword-relocations
17859 @opindex mword-relocations
17860 Only generate absolute relocations on word-sized values (i.e.@: R_ARM_ABS32).
17861 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
17862 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
17863 is specified. This option conflicts with @option{-mslow-flash-data}.
17865 @item -mfix-cortex-m3-ldrd
17866 @opindex mfix-cortex-m3-ldrd
17867 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
17868 with overlapping destination and base registers are used. This option avoids
17869 generating these instructions. This option is enabled by default when
17870 @option{-mcpu=cortex-m3} is specified.
17872 @item -munaligned-access
17873 @itemx -mno-unaligned-access
17874 @opindex munaligned-access
17875 @opindex mno-unaligned-access
17876 Enables (or disables) reading and writing of 16- and 32- bit values
17877 from addresses that are not 16- or 32- bit aligned. By default
17878 unaligned access is disabled for all pre-ARMv6, all ARMv6-M and for
17879 ARMv8-M Baseline architectures, and enabled for all other
17880 architectures. If unaligned access is not enabled then words in packed
17881 data structures are accessed a byte at a time.
17883 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
17884 generated object file to either true or false, depending upon the
17885 setting of this option. If unaligned access is enabled then the
17886 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
17889 @item -mneon-for-64bits
17890 @opindex mneon-for-64bits
17891 Enables using Neon to handle scalar 64-bits operations. This is
17892 disabled by default since the cost of moving data from core registers
17895 @item -mslow-flash-data
17896 @opindex mslow-flash-data
17897 Assume loading data from flash is slower than fetching instruction.
17898 Therefore literal load is minimized for better performance.
17899 This option is only supported when compiling for ARMv7 M-profile and
17900 off by default. It conflicts with @option{-mword-relocations}.
17902 @item -masm-syntax-unified
17903 @opindex masm-syntax-unified
17904 Assume inline assembler is using unified asm syntax. The default is
17905 currently off which implies divided syntax. This option has no impact
17906 on Thumb2. However, this may change in future releases of GCC.
17907 Divided syntax should be considered deprecated.
17909 @item -mrestrict-it
17910 @opindex mrestrict-it
17911 Restricts generation of IT blocks to conform to the rules of ARMv8-A.
17912 IT blocks can only contain a single 16-bit instruction from a select
17913 set of instructions. This option is on by default for ARMv8-A Thumb mode.
17915 @item -mprint-tune-info
17916 @opindex mprint-tune-info
17917 Print CPU tuning information as comment in assembler file. This is
17918 an option used only for regression testing of the compiler and not
17919 intended for ordinary use in compiling code. This option is disabled
17922 @item -mverbose-cost-dump
17923 @opindex mverbose-cost-dump
17924 Enable verbose cost model dumping in the debug dump files. This option is
17925 provided for use in debugging the compiler.
17928 @opindex mpure-code
17929 Do not allow constant data to be placed in code sections.
17930 Additionally, when compiling for ELF object format give all text sections the
17931 ELF processor-specific section attribute @code{SHF_ARM_PURECODE}. This option
17932 is only available when generating non-pic code for M-profile targets with the
17937 Generate secure code as per the "ARMv8-M Security Extensions: Requirements on
17938 Development Tools Engineering Specification", which can be found on
17939 @url{http://infocenter.arm.com/help/topic/com.arm.doc.ecm0359818/ECM0359818_armv8m_security_extensions_reqs_on_dev_tools_1_0.pdf}.
17943 @subsection AVR Options
17944 @cindex AVR Options
17946 These options are defined for AVR implementations:
17949 @item -mmcu=@var{mcu}
17951 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
17953 The default for this option is@tie{}@samp{avr2}.
17955 GCC supports the following AVR devices and ISAs:
17957 @include avr-mmcu.texi
17962 Assume that all data in static storage can be accessed by LDS / STS
17963 instructions. This option has only an effect on reduced Tiny devices like
17964 ATtiny40. See also the @code{absdata}
17965 @ref{AVR Variable Attributes,variable attribute}.
17967 @item -maccumulate-args
17968 @opindex maccumulate-args
17969 Accumulate outgoing function arguments and acquire/release the needed
17970 stack space for outgoing function arguments once in function
17971 prologue/epilogue. Without this option, outgoing arguments are pushed
17972 before calling a function and popped afterwards.
17974 Popping the arguments after the function call can be expensive on
17975 AVR so that accumulating the stack space might lead to smaller
17976 executables because arguments need not be removed from the
17977 stack after such a function call.
17979 This option can lead to reduced code size for functions that perform
17980 several calls to functions that get their arguments on the stack like
17981 calls to printf-like functions.
17983 @item -mbranch-cost=@var{cost}
17984 @opindex mbranch-cost
17985 Set the branch costs for conditional branch instructions to
17986 @var{cost}. Reasonable values for @var{cost} are small, non-negative
17987 integers. The default branch cost is 0.
17989 @item -mcall-prologues
17990 @opindex mcall-prologues
17991 Functions prologues/epilogues are expanded as calls to appropriate
17992 subroutines. Code size is smaller.
17994 @item -mgas-isr-prologues
17995 @opindex mgas-isr-prologues
17996 Interrupt service routines (ISRs) may use the @code{__gcc_isr} pseudo
17997 instruction supported by GNU Binutils.
17998 If this option is on, the feature can still be disabled for individual
17999 ISRs by means of the @ref{AVR Function Attributes,,@code{no_gccisr}}
18000 function attribute. This feature is activated per default
18001 if optimization is on (but not with @option{-Og}, @pxref{Optimize Options}),
18002 and if GNU Binutils support @w{@uref{https://sourceware.org/PR21683,PR21683}}.
18006 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
18007 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
18008 and @code{long long} is 4 bytes. Please note that this option does not
18009 conform to the C standards, but it results in smaller code
18012 @item -mmain-is-OS_task
18013 @opindex mmain-is-OS_task
18014 Do not save registers in @code{main}. The effect is the same like
18015 attaching attribute @ref{AVR Function Attributes,,@code{OS_task}}
18016 to @code{main}. It is activated per default if optimization is on.
18018 @item -mn-flash=@var{num}
18020 Assume that the flash memory has a size of
18021 @var{num} times 64@tie{}KiB.
18023 @item -mno-interrupts
18024 @opindex mno-interrupts
18025 Generated code is not compatible with hardware interrupts.
18026 Code size is smaller.
18030 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
18031 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
18032 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
18033 the assembler's command line and the @option{--relax} option to the
18034 linker's command line.
18036 Jump relaxing is performed by the linker because jump offsets are not
18037 known before code is located. Therefore, the assembler code generated by the
18038 compiler is the same, but the instructions in the executable may
18039 differ from instructions in the assembler code.
18041 Relaxing must be turned on if linker stubs are needed, see the
18042 section on @code{EIND} and linker stubs below.
18046 Assume that the device supports the Read-Modify-Write
18047 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
18049 @item -mshort-calls
18050 @opindex mshort-calls
18052 Assume that @code{RJMP} and @code{RCALL} can target the whole
18055 This option is used internally for multilib selection. It is
18056 not an optimization option, and you don't need to set it by hand.
18060 Treat the stack pointer register as an 8-bit register,
18061 i.e.@: assume the high byte of the stack pointer is zero.
18062 In general, you don't need to set this option by hand.
18064 This option is used internally by the compiler to select and
18065 build multilibs for architectures @code{avr2} and @code{avr25}.
18066 These architectures mix devices with and without @code{SPH}.
18067 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
18068 the compiler driver adds or removes this option from the compiler
18069 proper's command line, because the compiler then knows if the device
18070 or architecture has an 8-bit stack pointer and thus no @code{SPH}
18075 Use address register @code{X} in a way proposed by the hardware. This means
18076 that @code{X} is only used in indirect, post-increment or
18077 pre-decrement addressing.
18079 Without this option, the @code{X} register may be used in the same way
18080 as @code{Y} or @code{Z} which then is emulated by additional
18082 For example, loading a value with @code{X+const} addressing with a
18083 small non-negative @code{const < 64} to a register @var{Rn} is
18087 adiw r26, const ; X += const
18088 ld @var{Rn}, X ; @var{Rn} = *X
18089 sbiw r26, const ; X -= const
18093 @opindex mtiny-stack
18094 Only change the lower 8@tie{}bits of the stack pointer.
18096 @item -mfract-convert-truncate
18097 @opindex mfract-convert-truncate
18098 Allow to use truncation instead of rounding towards zero for fractional fixed-point types.
18101 @opindex nodevicelib
18102 Don't link against AVR-LibC's device specific library @code{lib<mcu>.a}.
18104 @item -Waddr-space-convert
18105 @opindex Waddr-space-convert
18106 @opindex Wno-addr-space-convert
18107 Warn about conversions between address spaces in the case where the
18108 resulting address space is not contained in the incoming address space.
18110 @item -Wmisspelled-isr
18111 @opindex Wmisspelled-isr
18112 @opindex Wno-misspelled-isr
18113 Warn if the ISR is misspelled, i.e.@: without __vector prefix.
18114 Enabled by default.
18117 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
18118 @cindex @code{EIND}
18119 Pointers in the implementation are 16@tie{}bits wide.
18120 The address of a function or label is represented as word address so
18121 that indirect jumps and calls can target any code address in the
18122 range of 64@tie{}Ki words.
18124 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
18125 bytes of program memory space, there is a special function register called
18126 @code{EIND} that serves as most significant part of the target address
18127 when @code{EICALL} or @code{EIJMP} instructions are used.
18129 Indirect jumps and calls on these devices are handled as follows by
18130 the compiler and are subject to some limitations:
18135 The compiler never sets @code{EIND}.
18138 The compiler uses @code{EIND} implicitly in @code{EICALL}/@code{EIJMP}
18139 instructions or might read @code{EIND} directly in order to emulate an
18140 indirect call/jump by means of a @code{RET} instruction.
18143 The compiler assumes that @code{EIND} never changes during the startup
18144 code or during the application. In particular, @code{EIND} is not
18145 saved/restored in function or interrupt service routine
18149 For indirect calls to functions and computed goto, the linker
18150 generates @emph{stubs}. Stubs are jump pads sometimes also called
18151 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
18152 The stub contains a direct jump to the desired address.
18155 Linker relaxation must be turned on so that the linker generates
18156 the stubs correctly in all situations. See the compiler option
18157 @option{-mrelax} and the linker option @option{--relax}.
18158 There are corner cases where the linker is supposed to generate stubs
18159 but aborts without relaxation and without a helpful error message.
18162 The default linker script is arranged for code with @code{EIND = 0}.
18163 If code is supposed to work for a setup with @code{EIND != 0}, a custom
18164 linker script has to be used in order to place the sections whose
18165 name start with @code{.trampolines} into the segment where @code{EIND}
18169 The startup code from libgcc never sets @code{EIND}.
18170 Notice that startup code is a blend of code from libgcc and AVR-LibC.
18171 For the impact of AVR-LibC on @code{EIND}, see the
18172 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
18175 It is legitimate for user-specific startup code to set up @code{EIND}
18176 early, for example by means of initialization code located in
18177 section @code{.init3}. Such code runs prior to general startup code
18178 that initializes RAM and calls constructors, but after the bit
18179 of startup code from AVR-LibC that sets @code{EIND} to the segment
18180 where the vector table is located.
18182 #include <avr/io.h>
18185 __attribute__((section(".init3"),naked,used,no_instrument_function))
18186 init3_set_eind (void)
18188 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
18189 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
18194 The @code{__trampolines_start} symbol is defined in the linker script.
18197 Stubs are generated automatically by the linker if
18198 the following two conditions are met:
18201 @item The address of a label is taken by means of the @code{gs} modifier
18202 (short for @emph{generate stubs}) like so:
18204 LDI r24, lo8(gs(@var{func}))
18205 LDI r25, hi8(gs(@var{func}))
18207 @item The final location of that label is in a code segment
18208 @emph{outside} the segment where the stubs are located.
18212 The compiler emits such @code{gs} modifiers for code labels in the
18213 following situations:
18215 @item Taking address of a function or code label.
18216 @item Computed goto.
18217 @item If prologue-save function is used, see @option{-mcall-prologues}
18218 command-line option.
18219 @item Switch/case dispatch tables. If you do not want such dispatch
18220 tables you can specify the @option{-fno-jump-tables} command-line option.
18221 @item C and C++ constructors/destructors called during startup/shutdown.
18222 @item If the tools hit a @code{gs()} modifier explained above.
18226 Jumping to non-symbolic addresses like so is @emph{not} supported:
18231 /* Call function at word address 0x2 */
18232 return ((int(*)(void)) 0x2)();
18236 Instead, a stub has to be set up, i.e.@: the function has to be called
18237 through a symbol (@code{func_4} in the example):
18242 extern int func_4 (void);
18244 /* Call function at byte address 0x4 */
18249 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
18250 Alternatively, @code{func_4} can be defined in the linker script.
18253 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
18254 @cindex @code{RAMPD}
18255 @cindex @code{RAMPX}
18256 @cindex @code{RAMPY}
18257 @cindex @code{RAMPZ}
18258 Some AVR devices support memories larger than the 64@tie{}KiB range
18259 that can be accessed with 16-bit pointers. To access memory locations
18260 outside this 64@tie{}KiB range, the content of a @code{RAMP}
18261 register is used as high part of the address:
18262 The @code{X}, @code{Y}, @code{Z} address register is concatenated
18263 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
18264 register, respectively, to get a wide address. Similarly,
18265 @code{RAMPD} is used together with direct addressing.
18269 The startup code initializes the @code{RAMP} special function
18270 registers with zero.
18273 If a @ref{AVR Named Address Spaces,named address space} other than
18274 generic or @code{__flash} is used, then @code{RAMPZ} is set
18275 as needed before the operation.
18278 If the device supports RAM larger than 64@tie{}KiB and the compiler
18279 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
18280 is reset to zero after the operation.
18283 If the device comes with a specific @code{RAMP} register, the ISR
18284 prologue/epilogue saves/restores that SFR and initializes it with
18285 zero in case the ISR code might (implicitly) use it.
18288 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
18289 If you use inline assembler to read from locations outside the
18290 16-bit address range and change one of the @code{RAMP} registers,
18291 you must reset it to zero after the access.
18295 @subsubsection AVR Built-in Macros
18297 GCC defines several built-in macros so that the user code can test
18298 for the presence or absence of features. Almost any of the following
18299 built-in macros are deduced from device capabilities and thus
18300 triggered by the @option{-mmcu=} command-line option.
18302 For even more AVR-specific built-in macros see
18303 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
18308 Build-in macro that resolves to a decimal number that identifies the
18309 architecture and depends on the @option{-mmcu=@var{mcu}} option.
18310 Possible values are:
18312 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
18313 @code{4}, @code{5}, @code{51}, @code{6}
18315 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
18316 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
18321 @code{102}, @code{103}, @code{104},
18322 @code{105}, @code{106}, @code{107}
18324 for @var{mcu}=@code{avrtiny},
18325 @code{avrxmega2}, @code{avrxmega3}, @code{avrxmega4},
18326 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
18327 If @var{mcu} specifies a device, this built-in macro is set
18328 accordingly. For example, with @option{-mmcu=atmega8} the macro is
18329 defined to @code{4}.
18331 @item __AVR_@var{Device}__
18332 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
18333 the device's name. For example, @option{-mmcu=atmega8} defines the
18334 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
18335 @code{__AVR_ATtiny261A__}, etc.
18337 The built-in macros' names follow
18338 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
18339 the device name as from the AVR user manual. The difference between
18340 @var{Device} in the built-in macro and @var{device} in
18341 @option{-mmcu=@var{device}} is that the latter is always lowercase.
18343 If @var{device} is not a device but only a core architecture like
18344 @samp{avr51}, this macro is not defined.
18346 @item __AVR_DEVICE_NAME__
18347 Setting @option{-mmcu=@var{device}} defines this built-in macro to
18348 the device's name. For example, with @option{-mmcu=atmega8} the macro
18349 is defined to @code{atmega8}.
18351 If @var{device} is not a device but only a core architecture like
18352 @samp{avr51}, this macro is not defined.
18354 @item __AVR_XMEGA__
18355 The device / architecture belongs to the XMEGA family of devices.
18357 @item __AVR_HAVE_ELPM__
18358 The device has the @code{ELPM} instruction.
18360 @item __AVR_HAVE_ELPMX__
18361 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
18362 R@var{n},Z+} instructions.
18364 @item __AVR_HAVE_MOVW__
18365 The device has the @code{MOVW} instruction to perform 16-bit
18366 register-register moves.
18368 @item __AVR_HAVE_LPMX__
18369 The device has the @code{LPM R@var{n},Z} and
18370 @code{LPM R@var{n},Z+} instructions.
18372 @item __AVR_HAVE_MUL__
18373 The device has a hardware multiplier.
18375 @item __AVR_HAVE_JMP_CALL__
18376 The device has the @code{JMP} and @code{CALL} instructions.
18377 This is the case for devices with more than 8@tie{}KiB of program
18380 @item __AVR_HAVE_EIJMP_EICALL__
18381 @itemx __AVR_3_BYTE_PC__
18382 The device has the @code{EIJMP} and @code{EICALL} instructions.
18383 This is the case for devices with more than 128@tie{}KiB of program memory.
18384 This also means that the program counter
18385 (PC) is 3@tie{}bytes wide.
18387 @item __AVR_2_BYTE_PC__
18388 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
18389 with up to 128@tie{}KiB of program memory.
18391 @item __AVR_HAVE_8BIT_SP__
18392 @itemx __AVR_HAVE_16BIT_SP__
18393 The stack pointer (SP) register is treated as 8-bit respectively
18394 16-bit register by the compiler.
18395 The definition of these macros is affected by @option{-mtiny-stack}.
18397 @item __AVR_HAVE_SPH__
18399 The device has the SPH (high part of stack pointer) special function
18400 register or has an 8-bit stack pointer, respectively.
18401 The definition of these macros is affected by @option{-mmcu=} and
18402 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
18405 @item __AVR_HAVE_RAMPD__
18406 @itemx __AVR_HAVE_RAMPX__
18407 @itemx __AVR_HAVE_RAMPY__
18408 @itemx __AVR_HAVE_RAMPZ__
18409 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
18410 @code{RAMPZ} special function register, respectively.
18412 @item __NO_INTERRUPTS__
18413 This macro reflects the @option{-mno-interrupts} command-line option.
18415 @item __AVR_ERRATA_SKIP__
18416 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
18417 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
18418 instructions because of a hardware erratum. Skip instructions are
18419 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
18420 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
18423 @item __AVR_ISA_RMW__
18424 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
18426 @item __AVR_SFR_OFFSET__=@var{offset}
18427 Instructions that can address I/O special function registers directly
18428 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
18429 address as if addressed by an instruction to access RAM like @code{LD}
18430 or @code{STS}. This offset depends on the device architecture and has
18431 to be subtracted from the RAM address in order to get the
18432 respective I/O@tie{}address.
18434 @item __AVR_SHORT_CALLS__
18435 The @option{-mshort-calls} command line option is set.
18437 @item __AVR_PM_BASE_ADDRESS__=@var{addr}
18438 Some devices support reading from flash memory by means of @code{LD*}
18439 instructions. The flash memory is seen in the data address space
18440 at an offset of @code{__AVR_PM_BASE_ADDRESS__}. If this macro
18441 is not defined, this feature is not available. If defined,
18442 the address space is linear and there is no need to put
18443 @code{.rodata} into RAM. This is handled by the default linker
18444 description file, and is currently available for
18445 @code{avrtiny} and @code{avrxmega3}. Even more convenient,
18446 there is no need to use address spaces like @code{__flash} or
18447 features like attribute @code{progmem} and @code{pgm_read_*}.
18449 @item __WITH_AVRLIBC__
18450 The compiler is configured to be used together with AVR-Libc.
18451 See the @option{--with-avrlibc} configure option.
18455 @node Blackfin Options
18456 @subsection Blackfin Options
18457 @cindex Blackfin Options
18460 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
18462 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
18463 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
18464 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
18465 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
18466 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
18467 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
18468 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
18469 @samp{bf561}, @samp{bf592}.
18471 The optional @var{sirevision} specifies the silicon revision of the target
18472 Blackfin processor. Any workarounds available for the targeted silicon revision
18473 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
18474 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
18475 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
18476 hexadecimal digits representing the major and minor numbers in the silicon
18477 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
18478 is not defined. If @var{sirevision} is @samp{any}, the
18479 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
18480 If this optional @var{sirevision} is not used, GCC assumes the latest known
18481 silicon revision of the targeted Blackfin processor.
18483 GCC defines a preprocessor macro for the specified @var{cpu}.
18484 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
18485 provided by libgloss to be linked in if @option{-msim} is not given.
18487 Without this option, @samp{bf532} is used as the processor by default.
18489 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
18490 only the preprocessor macro is defined.
18494 Specifies that the program will be run on the simulator. This causes
18495 the simulator BSP provided by libgloss to be linked in. This option
18496 has effect only for @samp{bfin-elf} toolchain.
18497 Certain other options, such as @option{-mid-shared-library} and
18498 @option{-mfdpic}, imply @option{-msim}.
18500 @item -momit-leaf-frame-pointer
18501 @opindex momit-leaf-frame-pointer
18502 Don't keep the frame pointer in a register for leaf functions. This
18503 avoids the instructions to save, set up and restore frame pointers and
18504 makes an extra register available in leaf functions.
18506 @item -mspecld-anomaly
18507 @opindex mspecld-anomaly
18508 When enabled, the compiler ensures that the generated code does not
18509 contain speculative loads after jump instructions. If this option is used,
18510 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
18512 @item -mno-specld-anomaly
18513 @opindex mno-specld-anomaly
18514 @opindex mspecld-anomaly
18515 Don't generate extra code to prevent speculative loads from occurring.
18517 @item -mcsync-anomaly
18518 @opindex mcsync-anomaly
18519 When enabled, the compiler ensures that the generated code does not
18520 contain CSYNC or SSYNC instructions too soon after conditional branches.
18521 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
18523 @item -mno-csync-anomaly
18524 @opindex mno-csync-anomaly
18525 @opindex mcsync-anomaly
18526 Don't generate extra code to prevent CSYNC or SSYNC instructions from
18527 occurring too soon after a conditional branch.
18531 When enabled, the compiler is free to take advantage of the knowledge that
18532 the entire program fits into the low 64k of memory.
18535 @opindex mno-low64k
18536 Assume that the program is arbitrarily large. This is the default.
18538 @item -mstack-check-l1
18539 @opindex mstack-check-l1
18540 Do stack checking using information placed into L1 scratchpad memory by the
18543 @item -mid-shared-library
18544 @opindex mid-shared-library
18545 Generate code that supports shared libraries via the library ID method.
18546 This allows for execute in place and shared libraries in an environment
18547 without virtual memory management. This option implies @option{-fPIC}.
18548 With a @samp{bfin-elf} target, this option implies @option{-msim}.
18550 @item -mno-id-shared-library
18551 @opindex mno-id-shared-library
18552 @opindex mid-shared-library
18553 Generate code that doesn't assume ID-based shared libraries are being used.
18554 This is the default.
18556 @item -mleaf-id-shared-library
18557 @opindex mleaf-id-shared-library
18558 Generate code that supports shared libraries via the library ID method,
18559 but assumes that this library or executable won't link against any other
18560 ID shared libraries. That allows the compiler to use faster code for jumps
18563 @item -mno-leaf-id-shared-library
18564 @opindex mno-leaf-id-shared-library
18565 @opindex mleaf-id-shared-library
18566 Do not assume that the code being compiled won't link against any ID shared
18567 libraries. Slower code is generated for jump and call insns.
18569 @item -mshared-library-id=n
18570 @opindex mshared-library-id
18571 Specifies the identification number of the ID-based shared library being
18572 compiled. Specifying a value of 0 generates more compact code; specifying
18573 other values forces the allocation of that number to the current
18574 library but is no more space- or time-efficient than omitting this option.
18578 Generate code that allows the data segment to be located in a different
18579 area of memory from the text segment. This allows for execute in place in
18580 an environment without virtual memory management by eliminating relocations
18581 against the text section.
18583 @item -mno-sep-data
18584 @opindex mno-sep-data
18586 Generate code that assumes that the data segment follows the text segment.
18587 This is the default.
18590 @itemx -mno-long-calls
18591 @opindex mlong-calls
18592 @opindex mno-long-calls
18593 Tells the compiler to perform function calls by first loading the
18594 address of the function into a register and then performing a subroutine
18595 call on this register. This switch is needed if the target function
18596 lies outside of the 24-bit addressing range of the offset-based
18597 version of subroutine call instruction.
18599 This feature is not enabled by default. Specifying
18600 @option{-mno-long-calls} restores the default behavior. Note these
18601 switches have no effect on how the compiler generates code to handle
18602 function calls via function pointers.
18606 Link with the fast floating-point library. This library relaxes some of
18607 the IEEE floating-point standard's rules for checking inputs against
18608 Not-a-Number (NAN), in the interest of performance.
18611 @opindex minline-plt
18612 Enable inlining of PLT entries in function calls to functions that are
18613 not known to bind locally. It has no effect without @option{-mfdpic}.
18616 @opindex mmulticore
18617 Build a standalone application for multicore Blackfin processors.
18618 This option causes proper start files and link scripts supporting
18619 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
18620 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
18622 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
18623 selects the one-application-per-core programming model. Without
18624 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
18625 programming model is used. In this model, the main function of Core B
18626 should be named as @code{coreb_main}.
18628 If this option is not used, the single-core application programming
18633 Build a standalone application for Core A of BF561 when using
18634 the one-application-per-core programming model. Proper start files
18635 and link scripts are used to support Core A, and the macro
18636 @code{__BFIN_COREA} is defined.
18637 This option can only be used in conjunction with @option{-mmulticore}.
18641 Build a standalone application for Core B of BF561 when using
18642 the one-application-per-core programming model. Proper start files
18643 and link scripts are used to support Core B, and the macro
18644 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
18645 should be used instead of @code{main}.
18646 This option can only be used in conjunction with @option{-mmulticore}.
18650 Build a standalone application for SDRAM. Proper start files and
18651 link scripts are used to put the application into SDRAM, and the macro
18652 @code{__BFIN_SDRAM} is defined.
18653 The loader should initialize SDRAM before loading the application.
18657 Assume that ICPLBs are enabled at run time. This has an effect on certain
18658 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
18659 are enabled; for standalone applications the default is off.
18663 @subsection C6X Options
18664 @cindex C6X Options
18667 @item -march=@var{name}
18669 This specifies the name of the target architecture. GCC uses this
18670 name to determine what kind of instructions it can emit when generating
18671 assembly code. Permissible names are: @samp{c62x},
18672 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
18675 @opindex mbig-endian
18676 Generate code for a big-endian target.
18678 @item -mlittle-endian
18679 @opindex mlittle-endian
18680 Generate code for a little-endian target. This is the default.
18684 Choose startup files and linker script suitable for the simulator.
18686 @item -msdata=default
18687 @opindex msdata=default
18688 Put small global and static data in the @code{.neardata} section,
18689 which is pointed to by register @code{B14}. Put small uninitialized
18690 global and static data in the @code{.bss} section, which is adjacent
18691 to the @code{.neardata} section. Put small read-only data into the
18692 @code{.rodata} section. The corresponding sections used for large
18693 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
18696 @opindex msdata=all
18697 Put all data, not just small objects, into the sections reserved for
18698 small data, and use addressing relative to the @code{B14} register to
18702 @opindex msdata=none
18703 Make no use of the sections reserved for small data, and use absolute
18704 addresses to access all data. Put all initialized global and static
18705 data in the @code{.fardata} section, and all uninitialized data in the
18706 @code{.far} section. Put all constant data into the @code{.const}
18711 @subsection CRIS Options
18712 @cindex CRIS Options
18714 These options are defined specifically for the CRIS ports.
18717 @item -march=@var{architecture-type}
18718 @itemx -mcpu=@var{architecture-type}
18721 Generate code for the specified architecture. The choices for
18722 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
18723 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
18724 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
18727 @item -mtune=@var{architecture-type}
18729 Tune to @var{architecture-type} everything applicable about the generated
18730 code, except for the ABI and the set of available instructions. The
18731 choices for @var{architecture-type} are the same as for
18732 @option{-march=@var{architecture-type}}.
18734 @item -mmax-stack-frame=@var{n}
18735 @opindex mmax-stack-frame
18736 Warn when the stack frame of a function exceeds @var{n} bytes.
18742 The options @option{-metrax4} and @option{-metrax100} are synonyms for
18743 @option{-march=v3} and @option{-march=v8} respectively.
18745 @item -mmul-bug-workaround
18746 @itemx -mno-mul-bug-workaround
18747 @opindex mmul-bug-workaround
18748 @opindex mno-mul-bug-workaround
18749 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
18750 models where it applies. This option is active by default.
18754 Enable CRIS-specific verbose debug-related information in the assembly
18755 code. This option also has the effect of turning off the @samp{#NO_APP}
18756 formatted-code indicator to the assembler at the beginning of the
18761 Do not use condition-code results from previous instruction; always emit
18762 compare and test instructions before use of condition codes.
18764 @item -mno-side-effects
18765 @opindex mno-side-effects
18766 @opindex mside-effects
18767 Do not emit instructions with side effects in addressing modes other than
18770 @item -mstack-align
18771 @itemx -mno-stack-align
18772 @itemx -mdata-align
18773 @itemx -mno-data-align
18774 @itemx -mconst-align
18775 @itemx -mno-const-align
18776 @opindex mstack-align
18777 @opindex mno-stack-align
18778 @opindex mdata-align
18779 @opindex mno-data-align
18780 @opindex mconst-align
18781 @opindex mno-const-align
18782 These options (@samp{no-} options) arrange (eliminate arrangements) for the
18783 stack frame, individual data and constants to be aligned for the maximum
18784 single data access size for the chosen CPU model. The default is to
18785 arrange for 32-bit alignment. ABI details such as structure layout are
18786 not affected by these options.
18794 Similar to the stack- data- and const-align options above, these options
18795 arrange for stack frame, writable data and constants to all be 32-bit,
18796 16-bit or 8-bit aligned. The default is 32-bit alignment.
18798 @item -mno-prologue-epilogue
18799 @itemx -mprologue-epilogue
18800 @opindex mno-prologue-epilogue
18801 @opindex mprologue-epilogue
18802 With @option{-mno-prologue-epilogue}, the normal function prologue and
18803 epilogue which set up the stack frame are omitted and no return
18804 instructions or return sequences are generated in the code. Use this
18805 option only together with visual inspection of the compiled code: no
18806 warnings or errors are generated when call-saved registers must be saved,
18807 or storage for local variables needs to be allocated.
18811 @opindex mno-gotplt
18813 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
18814 instruction sequences that load addresses for functions from the PLT part
18815 of the GOT rather than (traditional on other architectures) calls to the
18816 PLT@. The default is @option{-mgotplt}.
18820 Legacy no-op option only recognized with the cris-axis-elf and
18821 cris-axis-linux-gnu targets.
18825 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
18829 This option, recognized for the cris-axis-elf, arranges
18830 to link with input-output functions from a simulator library. Code,
18831 initialized data and zero-initialized data are allocated consecutively.
18835 Like @option{-sim}, but pass linker options to locate initialized data at
18836 0x40000000 and zero-initialized data at 0x80000000.
18840 @subsection CR16 Options
18841 @cindex CR16 Options
18843 These options are defined specifically for the CR16 ports.
18849 Enable the use of multiply-accumulate instructions. Disabled by default.
18853 @opindex mcr16cplus
18855 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
18860 Links the library libsim.a which is in compatible with simulator. Applicable
18861 to ELF compiler only.
18865 Choose integer type as 32-bit wide.
18869 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
18871 @item -mdata-model=@var{model}
18872 @opindex mdata-model
18873 Choose a data model. The choices for @var{model} are @samp{near},
18874 @samp{far} or @samp{medium}. @samp{medium} is default.
18875 However, @samp{far} is not valid with @option{-mcr16c}, as the
18876 CR16C architecture does not support the far data model.
18879 @node C-SKY Options
18880 @subsection C-SKY Options
18881 @cindex C-SKY Options
18883 GCC supports these options when compiling for C-SKY V2 processors.
18887 @item -march=@var{arch}
18889 Specify the C-SKY target architecture. Valid values for @var{arch} are:
18890 @samp{ck801}, @samp{ck802}, @samp{ck803}, @samp{ck807}, and @samp{ck810}.
18891 The default is @samp{ck810}.
18893 @item -mcpu=@var{cpu}
18895 Specify the C-SKY target processor. Valid values for @var{cpu} are:
18896 @samp{ck801}, @samp{ck801t},
18897 @samp{ck802}, @samp{ck802t}, @samp{ck802j},
18898 @samp{ck803}, @samp{ck803h}, @samp{ck803t}, @samp{ck803ht},
18899 @samp{ck803f}, @samp{ck803fh}, @samp{ck803e}, @samp{ck803eh},
18900 @samp{ck803et}, @samp{ck803eht}, @samp{ck803ef}, @samp{ck803efh},
18901 @samp{ck803ft}, @samp{ck803eft}, @samp{ck803efht}, @samp{ck803r1},
18902 @samp{ck803hr1}, @samp{ck803tr1}, @samp{ck803htr1}, @samp{ck803fr1},
18903 @samp{ck803fhr1}, @samp{ck803er1}, @samp{ck803ehr1}, @samp{ck803etr1},
18904 @samp{ck803ehtr1}, @samp{ck803efr1}, @samp{ck803efhr1}, @samp{ck803ftr1},
18905 @samp{ck803eftr1}, @samp{ck803efhtr1},
18906 @samp{ck803s}, @samp{ck803st}, @samp{ck803se}, @samp{ck803sf},
18907 @samp{ck803sef}, @samp{ck803seft},
18908 @samp{ck807e}, @samp{ck807ef}, @samp{ck807}, @samp{ck807f},
18909 @samp{ck810e}, @samp{ck810et}, @samp{ck810ef}, @samp{ck810eft},
18910 @samp{ck810}, @samp{ck810v}, @samp{ck810f}, @samp{ck810t}, @samp{ck810fv},
18911 @samp{ck810tv}, @samp{ck810ft}, and @samp{ck810ftv}.
18914 @opindex mbig-endian
18917 @itemx -mlittle-endian
18918 @opindex mlittle-endian
18922 Select big- or little-endian code. The default is little-endian.
18925 @opindex mhard-float
18926 @itemx -msoft-float
18927 @opindex msoft-float
18929 Select hardware or software floating-point implementations.
18930 The default is soft float.
18932 @item -mdouble-float
18933 @itemx -mno-double-float
18934 @opindex mdouble-float
18935 When @option{-mhard-float} is in effect, enable generation of
18936 double-precision float instructions. This is the default except
18937 when compiling for CK803.
18942 When @option{-mhard-float} is in effect, enable generation of
18943 @code{frecipd}, @code{fsqrtd}, and @code{fdivd} instructions.
18944 This is the default except when compiling for CK803.
18946 @item -mfpu=@var{fpu}
18948 Select the floating-point processor. This option can only be used with
18949 @option{-mhard-float}.
18950 Values for @var{fpu} are
18951 @samp{fpv2_sf} (equivalent to @samp{-mno-double-float -mno-fdivdu}),
18952 @samp{fpv2} (@samp{-mdouble-float -mno-divdu}), and
18953 @samp{fpv2_divd} (@samp{-mdouble-float -mdivdu}).
18958 Enable the extended @code{lrw} instruction. This option defaults to on
18959 for CK801 and off otherwise.
18964 Enable interrupt stack instructions; the default is off.
18966 The @option{-mistack} option is required to handle the
18967 @code{interrupt} and @code{isr} function attributes
18968 (@pxref{C-SKY Function Attributes}).
18972 Enable multiprocessor instructions; the default is off.
18976 Enable coprocessor instructions; the default is off.
18980 Enable coprocessor instructions; the default is off.
18984 Enable C-SKY security instructions; the default is off.
18988 Enable C-SKY trust instructions; the default is off.
18996 Enable C-SKY DSP, Enhanced DSP, or Vector DSP instructions, respectively.
18997 All of these options default to off.
19002 Generate divide instructions. Default is off.
19007 Generate code for Smart Mode, using only registers numbered 0-7 to allow
19008 use of 16-bit instructions. This option is ignored for CK801 where this
19009 is the required behavior, and it defaults to on for CK802.
19010 For other targets, the default is off.
19012 @item -mhigh-registers
19013 @itemx -mno-high-registers
19014 @opindex mhigh-registers
19015 Generate code using the high registers numbered 16-31. This option
19016 is not supported on CK801, CK802, or CK803, and is enabled by default
19017 for other processors.
19022 Generate code using global anchor symbol addresses.
19025 @itemx -mno-pushpop
19027 Generate code using @code{push} and @code{pop} instructions. This option
19030 @item -mmultiple-stld
19032 @itemx -mno-multiple-stld
19034 @opindex mmultiple-stld
19035 Generate code using @code{stm} and @code{ldm} instructions. This option
19036 isn't supported on CK801 but is enabled by default on other processors.
19039 @itemx -mno-constpool
19040 @opindex mconstpool
19041 Create constant pools in the compiler instead of deferring it to the
19042 assembler. This option is the default and required for correct code
19043 generation on CK801 and CK802, and is optional on other processors.
19046 @item -mno-stack-size
19047 @opindex mstack-size
19048 Emit @code{.stack_size} directives for each function in the assembly
19049 output. This option defaults to off.
19054 Generate code for the C-SKY compiler runtime instead of libgcc. This
19055 option defaults to off.
19057 @item -mbranch-cost=@var{n}
19058 @opindex mbranch-cost=
19059 Set the branch costs to roughly @code{n} instructions. The default is 1.
19061 @item -msched-prolog
19062 @itemx -mno-sched-prolog
19063 @opindex msched-prolog
19064 Permit scheduling of function prologue and epilogue sequences. Using
19065 this option can result in code that is not compliant with the C-SKY V2 ABI
19066 prologue requirements and that cannot be debugged or backtraced.
19067 It is disabled by default.
19071 @node Darwin Options
19072 @subsection Darwin Options
19073 @cindex Darwin options
19075 These options are defined for all architectures running the Darwin operating
19078 FSF GCC on Darwin does not create ``fat'' object files; it creates
19079 an object file for the single architecture that GCC was built to
19080 target. Apple's GCC on Darwin does create ``fat'' files if multiple
19081 @option{-arch} options are used; it does so by running the compiler or
19082 linker multiple times and joining the results together with
19085 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
19086 @samp{i686}) is determined by the flags that specify the ISA
19087 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
19088 @option{-force_cpusubtype_ALL} option can be used to override this.
19090 The Darwin tools vary in their behavior when presented with an ISA
19091 mismatch. The assembler, @file{as}, only permits instructions to
19092 be used that are valid for the subtype of the file it is generating,
19093 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
19094 The linker for shared libraries, @file{/usr/bin/libtool}, fails
19095 and prints an error if asked to create a shared library with a less
19096 restrictive subtype than its input files (for instance, trying to put
19097 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
19098 for executables, @command{ld}, quietly gives the executable the most
19099 restrictive subtype of any of its input files.
19104 Add the framework directory @var{dir} to the head of the list of
19105 directories to be searched for header files. These directories are
19106 interleaved with those specified by @option{-I} options and are
19107 scanned in a left-to-right order.
19109 A framework directory is a directory with frameworks in it. A
19110 framework is a directory with a @file{Headers} and/or
19111 @file{PrivateHeaders} directory contained directly in it that ends
19112 in @file{.framework}. The name of a framework is the name of this
19113 directory excluding the @file{.framework}. Headers associated with
19114 the framework are found in one of those two directories, with
19115 @file{Headers} being searched first. A subframework is a framework
19116 directory that is in a framework's @file{Frameworks} directory.
19117 Includes of subframework headers can only appear in a header of a
19118 framework that contains the subframework, or in a sibling subframework
19119 header. Two subframeworks are siblings if they occur in the same
19120 framework. A subframework should not have the same name as a
19121 framework; a warning is issued if this is violated. Currently a
19122 subframework cannot have subframeworks; in the future, the mechanism
19123 may be extended to support this. The standard frameworks can be found
19124 in @file{/System/Library/Frameworks} and
19125 @file{/Library/Frameworks}. An example include looks like
19126 @code{#include <Framework/header.h>}, where @file{Framework} denotes
19127 the name of the framework and @file{header.h} is found in the
19128 @file{PrivateHeaders} or @file{Headers} directory.
19130 @item -iframework@var{dir}
19131 @opindex iframework
19132 Like @option{-F} except the directory is a treated as a system
19133 directory. The main difference between this @option{-iframework} and
19134 @option{-F} is that with @option{-iframework} the compiler does not
19135 warn about constructs contained within header files found via
19136 @var{dir}. This option is valid only for the C family of languages.
19140 Emit debugging information for symbols that are used. For stabs
19141 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
19142 This is by default ON@.
19146 Emit debugging information for all symbols and types.
19148 @item -mmacosx-version-min=@var{version}
19149 The earliest version of MacOS X that this executable will run on
19150 is @var{version}. Typical values of @var{version} include @code{10.1},
19151 @code{10.2}, and @code{10.3.9}.
19153 If the compiler was built to use the system's headers by default,
19154 then the default for this option is the system version on which the
19155 compiler is running, otherwise the default is to make choices that
19156 are compatible with as many systems and code bases as possible.
19160 Enable kernel development mode. The @option{-mkernel} option sets
19161 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
19162 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
19163 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
19164 applicable. This mode also sets @option{-mno-altivec},
19165 @option{-msoft-float}, @option{-fno-builtin} and
19166 @option{-mlong-branch} for PowerPC targets.
19168 @item -mone-byte-bool
19169 @opindex mone-byte-bool
19170 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
19171 By default @code{sizeof(bool)} is @code{4} when compiling for
19172 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
19173 option has no effect on x86.
19175 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
19176 to generate code that is not binary compatible with code generated
19177 without that switch. Using this switch may require recompiling all
19178 other modules in a program, including system libraries. Use this
19179 switch to conform to a non-default data model.
19181 @item -mfix-and-continue
19182 @itemx -ffix-and-continue
19183 @itemx -findirect-data
19184 @opindex mfix-and-continue
19185 @opindex ffix-and-continue
19186 @opindex findirect-data
19187 Generate code suitable for fast turnaround development, such as to
19188 allow GDB to dynamically load @file{.o} files into already-running
19189 programs. @option{-findirect-data} and @option{-ffix-and-continue}
19190 are provided for backwards compatibility.
19194 Loads all members of static archive libraries.
19195 See man ld(1) for more information.
19197 @item -arch_errors_fatal
19198 @opindex arch_errors_fatal
19199 Cause the errors having to do with files that have the wrong architecture
19202 @item -bind_at_load
19203 @opindex bind_at_load
19204 Causes the output file to be marked such that the dynamic linker will
19205 bind all undefined references when the file is loaded or launched.
19209 Produce a Mach-o bundle format file.
19210 See man ld(1) for more information.
19212 @item -bundle_loader @var{executable}
19213 @opindex bundle_loader
19214 This option specifies the @var{executable} that will load the build
19215 output file being linked. See man ld(1) for more information.
19218 @opindex dynamiclib
19219 When passed this option, GCC produces a dynamic library instead of
19220 an executable when linking, using the Darwin @file{libtool} command.
19222 @item -force_cpusubtype_ALL
19223 @opindex force_cpusubtype_ALL
19224 This causes GCC's output file to have the @samp{ALL} subtype, instead of
19225 one controlled by the @option{-mcpu} or @option{-march} option.
19227 @item -allowable_client @var{client_name}
19228 @itemx -client_name
19229 @itemx -compatibility_version
19230 @itemx -current_version
19232 @itemx -dependency-file
19234 @itemx -dylinker_install_name
19236 @itemx -exported_symbols_list
19239 @itemx -flat_namespace
19240 @itemx -force_flat_namespace
19241 @itemx -headerpad_max_install_names
19244 @itemx -install_name
19245 @itemx -keep_private_externs
19246 @itemx -multi_module
19247 @itemx -multiply_defined
19248 @itemx -multiply_defined_unused
19251 @itemx -no_dead_strip_inits_and_terms
19252 @itemx -nofixprebinding
19253 @itemx -nomultidefs
19255 @itemx -noseglinkedit
19256 @itemx -pagezero_size
19258 @itemx -prebind_all_twolevel_modules
19259 @itemx -private_bundle
19261 @itemx -read_only_relocs
19263 @itemx -sectobjectsymbols
19267 @itemx -sectobjectsymbols
19270 @itemx -segs_read_only_addr
19272 @itemx -segs_read_write_addr
19273 @itemx -seg_addr_table
19274 @itemx -seg_addr_table_filename
19275 @itemx -seglinkedit
19277 @itemx -segs_read_only_addr
19278 @itemx -segs_read_write_addr
19279 @itemx -single_module
19281 @itemx -sub_library
19283 @itemx -sub_umbrella
19284 @itemx -twolevel_namespace
19287 @itemx -unexported_symbols_list
19288 @itemx -weak_reference_mismatches
19289 @itemx -whatsloaded
19290 @opindex allowable_client
19291 @opindex client_name
19292 @opindex compatibility_version
19293 @opindex current_version
19294 @opindex dead_strip
19295 @opindex dependency-file
19296 @opindex dylib_file
19297 @opindex dylinker_install_name
19299 @opindex exported_symbols_list
19301 @opindex flat_namespace
19302 @opindex force_flat_namespace
19303 @opindex headerpad_max_install_names
19304 @opindex image_base
19306 @opindex install_name
19307 @opindex keep_private_externs
19308 @opindex multi_module
19309 @opindex multiply_defined
19310 @opindex multiply_defined_unused
19311 @opindex noall_load
19312 @opindex no_dead_strip_inits_and_terms
19313 @opindex nofixprebinding
19314 @opindex nomultidefs
19316 @opindex noseglinkedit
19317 @opindex pagezero_size
19319 @opindex prebind_all_twolevel_modules
19320 @opindex private_bundle
19321 @opindex read_only_relocs
19323 @opindex sectobjectsymbols
19326 @opindex sectcreate
19327 @opindex sectobjectsymbols
19330 @opindex segs_read_only_addr
19331 @opindex segs_read_write_addr
19332 @opindex seg_addr_table
19333 @opindex seg_addr_table_filename
19334 @opindex seglinkedit
19336 @opindex segs_read_only_addr
19337 @opindex segs_read_write_addr
19338 @opindex single_module
19340 @opindex sub_library
19341 @opindex sub_umbrella
19342 @opindex twolevel_namespace
19345 @opindex unexported_symbols_list
19346 @opindex weak_reference_mismatches
19347 @opindex whatsloaded
19348 These options are passed to the Darwin linker. The Darwin linker man page
19349 describes them in detail.
19352 @node DEC Alpha Options
19353 @subsection DEC Alpha Options
19355 These @samp{-m} options are defined for the DEC Alpha implementations:
19358 @item -mno-soft-float
19359 @itemx -msoft-float
19360 @opindex mno-soft-float
19361 @opindex msoft-float
19362 Use (do not use) the hardware floating-point instructions for
19363 floating-point operations. When @option{-msoft-float} is specified,
19364 functions in @file{libgcc.a} are used to perform floating-point
19365 operations. Unless they are replaced by routines that emulate the
19366 floating-point operations, or compiled in such a way as to call such
19367 emulations routines, these routines issue floating-point
19368 operations. If you are compiling for an Alpha without floating-point
19369 operations, you must ensure that the library is built so as not to call
19372 Note that Alpha implementations without floating-point operations are
19373 required to have floating-point registers.
19376 @itemx -mno-fp-regs
19378 @opindex mno-fp-regs
19379 Generate code that uses (does not use) the floating-point register set.
19380 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
19381 register set is not used, floating-point operands are passed in integer
19382 registers as if they were integers and floating-point results are passed
19383 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
19384 so any function with a floating-point argument or return value called by code
19385 compiled with @option{-mno-fp-regs} must also be compiled with that
19388 A typical use of this option is building a kernel that does not use,
19389 and hence need not save and restore, any floating-point registers.
19393 The Alpha architecture implements floating-point hardware optimized for
19394 maximum performance. It is mostly compliant with the IEEE floating-point
19395 standard. However, for full compliance, software assistance is
19396 required. This option generates code fully IEEE-compliant code
19397 @emph{except} that the @var{inexact-flag} is not maintained (see below).
19398 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
19399 defined during compilation. The resulting code is less efficient but is
19400 able to correctly support denormalized numbers and exceptional IEEE
19401 values such as not-a-number and plus/minus infinity. Other Alpha
19402 compilers call this option @option{-ieee_with_no_inexact}.
19404 @item -mieee-with-inexact
19405 @opindex mieee-with-inexact
19406 This is like @option{-mieee} except the generated code also maintains
19407 the IEEE @var{inexact-flag}. Turning on this option causes the
19408 generated code to implement fully-compliant IEEE math. In addition to
19409 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
19410 macro. On some Alpha implementations the resulting code may execute
19411 significantly slower than the code generated by default. Since there is
19412 very little code that depends on the @var{inexact-flag}, you should
19413 normally not specify this option. Other Alpha compilers call this
19414 option @option{-ieee_with_inexact}.
19416 @item -mfp-trap-mode=@var{trap-mode}
19417 @opindex mfp-trap-mode
19418 This option controls what floating-point related traps are enabled.
19419 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
19420 The trap mode can be set to one of four values:
19424 This is the default (normal) setting. The only traps that are enabled
19425 are the ones that cannot be disabled in software (e.g., division by zero
19429 In addition to the traps enabled by @samp{n}, underflow traps are enabled
19433 Like @samp{u}, but the instructions are marked to be safe for software
19434 completion (see Alpha architecture manual for details).
19437 Like @samp{su}, but inexact traps are enabled as well.
19440 @item -mfp-rounding-mode=@var{rounding-mode}
19441 @opindex mfp-rounding-mode
19442 Selects the IEEE rounding mode. Other Alpha compilers call this option
19443 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
19448 Normal IEEE rounding mode. Floating-point numbers are rounded towards
19449 the nearest machine number or towards the even machine number in case
19453 Round towards minus infinity.
19456 Chopped rounding mode. Floating-point numbers are rounded towards zero.
19459 Dynamic rounding mode. A field in the floating-point control register
19460 (@var{fpcr}, see Alpha architecture reference manual) controls the
19461 rounding mode in effect. The C library initializes this register for
19462 rounding towards plus infinity. Thus, unless your program modifies the
19463 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
19466 @item -mtrap-precision=@var{trap-precision}
19467 @opindex mtrap-precision
19468 In the Alpha architecture, floating-point traps are imprecise. This
19469 means without software assistance it is impossible to recover from a
19470 floating trap and program execution normally needs to be terminated.
19471 GCC can generate code that can assist operating system trap handlers
19472 in determining the exact location that caused a floating-point trap.
19473 Depending on the requirements of an application, different levels of
19474 precisions can be selected:
19478 Program precision. This option is the default and means a trap handler
19479 can only identify which program caused a floating-point exception.
19482 Function precision. The trap handler can determine the function that
19483 caused a floating-point exception.
19486 Instruction precision. The trap handler can determine the exact
19487 instruction that caused a floating-point exception.
19490 Other Alpha compilers provide the equivalent options called
19491 @option{-scope_safe} and @option{-resumption_safe}.
19493 @item -mieee-conformant
19494 @opindex mieee-conformant
19495 This option marks the generated code as IEEE conformant. You must not
19496 use this option unless you also specify @option{-mtrap-precision=i} and either
19497 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
19498 is to emit the line @samp{.eflag 48} in the function prologue of the
19499 generated assembly file.
19501 @item -mbuild-constants
19502 @opindex mbuild-constants
19503 Normally GCC examines a 32- or 64-bit integer constant to
19504 see if it can construct it from smaller constants in two or three
19505 instructions. If it cannot, it outputs the constant as a literal and
19506 generates code to load it from the data segment at run time.
19508 Use this option to require GCC to construct @emph{all} integer constants
19509 using code, even if it takes more instructions (the maximum is six).
19511 You typically use this option to build a shared library dynamic
19512 loader. Itself a shared library, it must relocate itself in memory
19513 before it can find the variables and constants in its own data segment.
19531 Indicate whether GCC should generate code to use the optional BWX,
19532 CIX, FIX and MAX instruction sets. The default is to use the instruction
19533 sets supported by the CPU type specified via @option{-mcpu=} option or that
19534 of the CPU on which GCC was built if none is specified.
19537 @itemx -mfloat-ieee
19538 @opindex mfloat-vax
19539 @opindex mfloat-ieee
19540 Generate code that uses (does not use) VAX F and G floating-point
19541 arithmetic instead of IEEE single and double precision.
19543 @item -mexplicit-relocs
19544 @itemx -mno-explicit-relocs
19545 @opindex mexplicit-relocs
19546 @opindex mno-explicit-relocs
19547 Older Alpha assemblers provided no way to generate symbol relocations
19548 except via assembler macros. Use of these macros does not allow
19549 optimal instruction scheduling. GNU binutils as of version 2.12
19550 supports a new syntax that allows the compiler to explicitly mark
19551 which relocations should apply to which instructions. This option
19552 is mostly useful for debugging, as GCC detects the capabilities of
19553 the assembler when it is built and sets the default accordingly.
19556 @itemx -mlarge-data
19557 @opindex msmall-data
19558 @opindex mlarge-data
19559 When @option{-mexplicit-relocs} is in effect, static data is
19560 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
19561 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
19562 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
19563 16-bit relocations off of the @code{$gp} register. This limits the
19564 size of the small data area to 64KB, but allows the variables to be
19565 directly accessed via a single instruction.
19567 The default is @option{-mlarge-data}. With this option the data area
19568 is limited to just below 2GB@. Programs that require more than 2GB of
19569 data must use @code{malloc} or @code{mmap} to allocate the data in the
19570 heap instead of in the program's data segment.
19572 When generating code for shared libraries, @option{-fpic} implies
19573 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
19576 @itemx -mlarge-text
19577 @opindex msmall-text
19578 @opindex mlarge-text
19579 When @option{-msmall-text} is used, the compiler assumes that the
19580 code of the entire program (or shared library) fits in 4MB, and is
19581 thus reachable with a branch instruction. When @option{-msmall-data}
19582 is used, the compiler can assume that all local symbols share the
19583 same @code{$gp} value, and thus reduce the number of instructions
19584 required for a function call from 4 to 1.
19586 The default is @option{-mlarge-text}.
19588 @item -mcpu=@var{cpu_type}
19590 Set the instruction set and instruction scheduling parameters for
19591 machine type @var{cpu_type}. You can specify either the @samp{EV}
19592 style name or the corresponding chip number. GCC supports scheduling
19593 parameters for the EV4, EV5 and EV6 family of processors and
19594 chooses the default values for the instruction set from the processor
19595 you specify. If you do not specify a processor type, GCC defaults
19596 to the processor on which the compiler was built.
19598 Supported values for @var{cpu_type} are
19604 Schedules as an EV4 and has no instruction set extensions.
19608 Schedules as an EV5 and has no instruction set extensions.
19612 Schedules as an EV5 and supports the BWX extension.
19617 Schedules as an EV5 and supports the BWX and MAX extensions.
19621 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
19625 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
19628 Native toolchains also support the value @samp{native},
19629 which selects the best architecture option for the host processor.
19630 @option{-mcpu=native} has no effect if GCC does not recognize
19633 @item -mtune=@var{cpu_type}
19635 Set only the instruction scheduling parameters for machine type
19636 @var{cpu_type}. The instruction set is not changed.
19638 Native toolchains also support the value @samp{native},
19639 which selects the best architecture option for the host processor.
19640 @option{-mtune=native} has no effect if GCC does not recognize
19643 @item -mmemory-latency=@var{time}
19644 @opindex mmemory-latency
19645 Sets the latency the scheduler should assume for typical memory
19646 references as seen by the application. This number is highly
19647 dependent on the memory access patterns used by the application
19648 and the size of the external cache on the machine.
19650 Valid options for @var{time} are
19654 A decimal number representing clock cycles.
19660 The compiler contains estimates of the number of clock cycles for
19661 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
19662 (also called Dcache, Scache, and Bcache), as well as to main memory.
19663 Note that L3 is only valid for EV5.
19669 @subsection FR30 Options
19670 @cindex FR30 Options
19672 These options are defined specifically for the FR30 port.
19676 @item -msmall-model
19677 @opindex msmall-model
19678 Use the small address space model. This can produce smaller code, but
19679 it does assume that all symbolic values and addresses fit into a
19684 Assume that runtime support has been provided and so there is no need
19685 to include the simulator library (@file{libsim.a}) on the linker
19691 @subsection FT32 Options
19692 @cindex FT32 Options
19694 These options are defined specifically for the FT32 port.
19700 Specifies that the program will be run on the simulator. This causes
19701 an alternate runtime startup and library to be linked.
19702 You must not use this option when generating programs that will run on
19703 real hardware; you must provide your own runtime library for whatever
19704 I/O functions are needed.
19708 Enable Local Register Allocation. This is still experimental for FT32,
19709 so by default the compiler uses standard reload.
19713 Do not use div and mod instructions.
19717 Enable use of the extended instructions of the FT32B processor.
19721 Compress all code using the Ft32B code compression scheme.
19725 Do not generate code that reads program memory.
19730 @subsection FRV Options
19731 @cindex FRV Options
19737 Only use the first 32 general-purpose registers.
19742 Use all 64 general-purpose registers.
19747 Use only the first 32 floating-point registers.
19752 Use all 64 floating-point registers.
19755 @opindex mhard-float
19757 Use hardware instructions for floating-point operations.
19760 @opindex msoft-float
19762 Use library routines for floating-point operations.
19767 Dynamically allocate condition code registers.
19772 Do not try to dynamically allocate condition code registers, only
19773 use @code{icc0} and @code{fcc0}.
19778 Change ABI to use double word insns.
19784 Do not use double word instructions.
19789 Use floating-point double instructions.
19792 @opindex mno-double
19794 Do not use floating-point double instructions.
19799 Use media instructions.
19804 Do not use media instructions.
19809 Use multiply and add/subtract instructions.
19812 @opindex mno-muladd
19814 Do not use multiply and add/subtract instructions.
19819 Select the FDPIC ABI, which uses function descriptors to represent
19820 pointers to functions. Without any PIC/PIE-related options, it
19821 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
19822 assumes GOT entries and small data are within a 12-bit range from the
19823 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
19824 are computed with 32 bits.
19825 With a @samp{bfin-elf} target, this option implies @option{-msim}.
19828 @opindex minline-plt
19830 Enable inlining of PLT entries in function calls to functions that are
19831 not known to bind locally. It has no effect without @option{-mfdpic}.
19832 It's enabled by default if optimizing for speed and compiling for
19833 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
19834 optimization option such as @option{-O3} or above is present in the
19840 Assume a large TLS segment when generating thread-local code.
19845 Do not assume a large TLS segment when generating thread-local code.
19850 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
19851 that is known to be in read-only sections. It's enabled by default,
19852 except for @option{-fpic} or @option{-fpie}: even though it may help
19853 make the global offset table smaller, it trades 1 instruction for 4.
19854 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
19855 one of which may be shared by multiple symbols, and it avoids the need
19856 for a GOT entry for the referenced symbol, so it's more likely to be a
19857 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
19859 @item -multilib-library-pic
19860 @opindex multilib-library-pic
19862 Link with the (library, not FD) pic libraries. It's implied by
19863 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
19864 @option{-fpic} without @option{-mfdpic}. You should never have to use
19868 @opindex mlinked-fp
19870 Follow the EABI requirement of always creating a frame pointer whenever
19871 a stack frame is allocated. This option is enabled by default and can
19872 be disabled with @option{-mno-linked-fp}.
19875 @opindex mlong-calls
19877 Use indirect addressing to call functions outside the current
19878 compilation unit. This allows the functions to be placed anywhere
19879 within the 32-bit address space.
19881 @item -malign-labels
19882 @opindex malign-labels
19884 Try to align labels to an 8-byte boundary by inserting NOPs into the
19885 previous packet. This option only has an effect when VLIW packing
19886 is enabled. It doesn't create new packets; it merely adds NOPs to
19889 @item -mlibrary-pic
19890 @opindex mlibrary-pic
19892 Generate position-independent EABI code.
19897 Use only the first four media accumulator registers.
19902 Use all eight media accumulator registers.
19907 Pack VLIW instructions.
19912 Do not pack VLIW instructions.
19915 @opindex mno-eflags
19917 Do not mark ABI switches in e_flags.
19920 @opindex mcond-move
19922 Enable the use of conditional-move instructions (default).
19924 This switch is mainly for debugging the compiler and will likely be removed
19925 in a future version.
19927 @item -mno-cond-move
19928 @opindex mno-cond-move
19930 Disable the use of conditional-move instructions.
19932 This switch is mainly for debugging the compiler and will likely be removed
19933 in a future version.
19938 Enable the use of conditional set instructions (default).
19940 This switch is mainly for debugging the compiler and will likely be removed
19941 in a future version.
19946 Disable the use of conditional set instructions.
19948 This switch is mainly for debugging the compiler and will likely be removed
19949 in a future version.
19952 @opindex mcond-exec
19954 Enable the use of conditional execution (default).
19956 This switch is mainly for debugging the compiler and will likely be removed
19957 in a future version.
19959 @item -mno-cond-exec
19960 @opindex mno-cond-exec
19962 Disable the use of conditional execution.
19964 This switch is mainly for debugging the compiler and will likely be removed
19965 in a future version.
19967 @item -mvliw-branch
19968 @opindex mvliw-branch
19970 Run a pass to pack branches into VLIW instructions (default).
19972 This switch is mainly for debugging the compiler and will likely be removed
19973 in a future version.
19975 @item -mno-vliw-branch
19976 @opindex mno-vliw-branch
19978 Do not run a pass to pack branches into VLIW instructions.
19980 This switch is mainly for debugging the compiler and will likely be removed
19981 in a future version.
19983 @item -mmulti-cond-exec
19984 @opindex mmulti-cond-exec
19986 Enable optimization of @code{&&} and @code{||} in conditional execution
19989 This switch is mainly for debugging the compiler and will likely be removed
19990 in a future version.
19992 @item -mno-multi-cond-exec
19993 @opindex mno-multi-cond-exec
19995 Disable optimization of @code{&&} and @code{||} in conditional execution.
19997 This switch is mainly for debugging the compiler and will likely be removed
19998 in a future version.
20000 @item -mnested-cond-exec
20001 @opindex mnested-cond-exec
20003 Enable nested conditional execution optimizations (default).
20005 This switch is mainly for debugging the compiler and will likely be removed
20006 in a future version.
20008 @item -mno-nested-cond-exec
20009 @opindex mno-nested-cond-exec
20011 Disable nested conditional execution optimizations.
20013 This switch is mainly for debugging the compiler and will likely be removed
20014 in a future version.
20016 @item -moptimize-membar
20017 @opindex moptimize-membar
20019 This switch removes redundant @code{membar} instructions from the
20020 compiler-generated code. It is enabled by default.
20022 @item -mno-optimize-membar
20023 @opindex mno-optimize-membar
20024 @opindex moptimize-membar
20026 This switch disables the automatic removal of redundant @code{membar}
20027 instructions from the generated code.
20029 @item -mtomcat-stats
20030 @opindex mtomcat-stats
20032 Cause gas to print out tomcat statistics.
20034 @item -mcpu=@var{cpu}
20037 Select the processor type for which to generate code. Possible values are
20038 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
20039 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
20043 @node GNU/Linux Options
20044 @subsection GNU/Linux Options
20046 These @samp{-m} options are defined for GNU/Linux targets:
20051 Use the GNU C library. This is the default except
20052 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
20053 @samp{*-*-linux-*android*} targets.
20057 Use uClibc C library. This is the default on
20058 @samp{*-*-linux-*uclibc*} targets.
20062 Use the musl C library. This is the default on
20063 @samp{*-*-linux-*musl*} targets.
20067 Use Bionic C library. This is the default on
20068 @samp{*-*-linux-*android*} targets.
20072 Compile code compatible with Android platform. This is the default on
20073 @samp{*-*-linux-*android*} targets.
20075 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
20076 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
20077 this option makes the GCC driver pass Android-specific options to the linker.
20078 Finally, this option causes the preprocessor macro @code{__ANDROID__}
20081 @item -tno-android-cc
20082 @opindex tno-android-cc
20083 Disable compilation effects of @option{-mandroid}, i.e., do not enable
20084 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
20085 @option{-fno-rtti} by default.
20087 @item -tno-android-ld
20088 @opindex tno-android-ld
20089 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
20090 linking options to the linker.
20094 @node H8/300 Options
20095 @subsection H8/300 Options
20097 These @samp{-m} options are defined for the H8/300 implementations:
20102 Shorten some address references at link time, when possible; uses the
20103 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
20104 ld, Using ld}, for a fuller description.
20108 Generate code for the H8/300H@.
20112 Generate code for the H8S@.
20116 Generate code for the H8S and H8/300H in the normal mode. This switch
20117 must be used either with @option{-mh} or @option{-ms}.
20121 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
20125 Extended registers are stored on stack before execution of function
20126 with monitor attribute. Default option is @option{-mexr}.
20127 This option is valid only for H8S targets.
20132 Extended registers are not stored on stack before execution of function
20133 with monitor attribute. Default option is @option{-mno-exr}.
20134 This option is valid only for H8S targets.
20138 Make @code{int} data 32 bits by default.
20141 @opindex malign-300
20142 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
20143 The default for the H8/300H and H8S is to align longs and floats on
20145 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
20146 This option has no effect on the H8/300.
20150 @subsection HPPA Options
20151 @cindex HPPA Options
20153 These @samp{-m} options are defined for the HPPA family of computers:
20156 @item -march=@var{architecture-type}
20158 Generate code for the specified architecture. The choices for
20159 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
20160 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
20161 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
20162 architecture option for your machine. Code compiled for lower numbered
20163 architectures runs on higher numbered architectures, but not the
20166 @item -mpa-risc-1-0
20167 @itemx -mpa-risc-1-1
20168 @itemx -mpa-risc-2-0
20169 @opindex mpa-risc-1-0
20170 @opindex mpa-risc-1-1
20171 @opindex mpa-risc-2-0
20172 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
20174 @item -mcaller-copies
20175 @opindex mcaller-copies
20176 The caller copies function arguments passed by hidden reference. This
20177 option should be used with care as it is not compatible with the default
20178 32-bit runtime. However, only aggregates larger than eight bytes are
20179 passed by hidden reference and the option provides better compatibility
20182 @item -mjump-in-delay
20183 @opindex mjump-in-delay
20184 This option is ignored and provided for compatibility purposes only.
20186 @item -mdisable-fpregs
20187 @opindex mdisable-fpregs
20188 Prevent floating-point registers from being used in any manner. This is
20189 necessary for compiling kernels that perform lazy context switching of
20190 floating-point registers. If you use this option and attempt to perform
20191 floating-point operations, the compiler aborts.
20193 @item -mdisable-indexing
20194 @opindex mdisable-indexing
20195 Prevent the compiler from using indexing address modes. This avoids some
20196 rather obscure problems when compiling MIG generated code under MACH@.
20198 @item -mno-space-regs
20199 @opindex mno-space-regs
20200 @opindex mspace-regs
20201 Generate code that assumes the target has no space registers. This allows
20202 GCC to generate faster indirect calls and use unscaled index address modes.
20204 Such code is suitable for level 0 PA systems and kernels.
20206 @item -mfast-indirect-calls
20207 @opindex mfast-indirect-calls
20208 Generate code that assumes calls never cross space boundaries. This
20209 allows GCC to emit code that performs faster indirect calls.
20211 This option does not work in the presence of shared libraries or nested
20214 @item -mfixed-range=@var{register-range}
20215 @opindex mfixed-range
20216 Generate code treating the given register range as fixed registers.
20217 A fixed register is one that the register allocator cannot use. This is
20218 useful when compiling kernel code. A register range is specified as
20219 two registers separated by a dash. Multiple register ranges can be
20220 specified separated by a comma.
20222 @item -mlong-load-store
20223 @opindex mlong-load-store
20224 Generate 3-instruction load and store sequences as sometimes required by
20225 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
20228 @item -mportable-runtime
20229 @opindex mportable-runtime
20230 Use the portable calling conventions proposed by HP for ELF systems.
20234 Enable the use of assembler directives only GAS understands.
20236 @item -mschedule=@var{cpu-type}
20238 Schedule code according to the constraints for the machine type
20239 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
20240 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
20241 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
20242 proper scheduling option for your machine. The default scheduling is
20246 @opindex mlinker-opt
20247 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
20248 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
20249 linkers in which they give bogus error messages when linking some programs.
20252 @opindex msoft-float
20253 Generate output containing library calls for floating point.
20254 @strong{Warning:} the requisite libraries are not available for all HPPA
20255 targets. Normally the facilities of the machine's usual C compiler are
20256 used, but this cannot be done directly in cross-compilation. You must make
20257 your own arrangements to provide suitable library functions for
20260 @option{-msoft-float} changes the calling convention in the output file;
20261 therefore, it is only useful if you compile @emph{all} of a program with
20262 this option. In particular, you need to compile @file{libgcc.a}, the
20263 library that comes with GCC, with @option{-msoft-float} in order for
20268 Generate the predefine, @code{_SIO}, for server IO@. The default is
20269 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
20270 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
20271 options are available under HP-UX and HI-UX@.
20275 Use options specific to GNU @command{ld}.
20276 This passes @option{-shared} to @command{ld} when
20277 building a shared library. It is the default when GCC is configured,
20278 explicitly or implicitly, with the GNU linker. This option does not
20279 affect which @command{ld} is called; it only changes what parameters
20280 are passed to that @command{ld}.
20281 The @command{ld} that is called is determined by the
20282 @option{--with-ld} configure option, GCC's program search path, and
20283 finally by the user's @env{PATH}. The linker used by GCC can be printed
20284 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
20285 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
20289 Use options specific to HP @command{ld}.
20290 This passes @option{-b} to @command{ld} when building
20291 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
20292 links. It is the default when GCC is configured, explicitly or
20293 implicitly, with the HP linker. This option does not affect
20294 which @command{ld} is called; it only changes what parameters are passed to that
20296 The @command{ld} that is called is determined by the @option{--with-ld}
20297 configure option, GCC's program search path, and finally by the user's
20298 @env{PATH}. The linker used by GCC can be printed using @samp{which
20299 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
20300 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
20303 @opindex mno-long-calls
20304 @opindex mlong-calls
20305 Generate code that uses long call sequences. This ensures that a call
20306 is always able to reach linker generated stubs. The default is to generate
20307 long calls only when the distance from the call site to the beginning
20308 of the function or translation unit, as the case may be, exceeds a
20309 predefined limit set by the branch type being used. The limits for
20310 normal calls are 7,600,000 and 240,000 bytes, respectively for the
20311 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
20314 Distances are measured from the beginning of functions when using the
20315 @option{-ffunction-sections} option, or when using the @option{-mgas}
20316 and @option{-mno-portable-runtime} options together under HP-UX with
20319 It is normally not desirable to use this option as it degrades
20320 performance. However, it may be useful in large applications,
20321 particularly when partial linking is used to build the application.
20323 The types of long calls used depends on the capabilities of the
20324 assembler and linker, and the type of code being generated. The
20325 impact on systems that support long absolute calls, and long pic
20326 symbol-difference or pc-relative calls should be relatively small.
20327 However, an indirect call is used on 32-bit ELF systems in pic code
20328 and it is quite long.
20330 @item -munix=@var{unix-std}
20332 Generate compiler predefines and select a startfile for the specified
20333 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
20334 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
20335 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
20336 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
20337 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
20340 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
20341 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
20342 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
20343 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
20344 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
20345 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
20347 It is @emph{important} to note that this option changes the interfaces
20348 for various library routines. It also affects the operational behavior
20349 of the C library. Thus, @emph{extreme} care is needed in using this
20352 Library code that is intended to operate with more than one UNIX
20353 standard must test, set and restore the variable @code{__xpg4_extended_mask}
20354 as appropriate. Most GNU software doesn't provide this capability.
20358 Suppress the generation of link options to search libdld.sl when the
20359 @option{-static} option is specified on HP-UX 10 and later.
20363 The HP-UX implementation of setlocale in libc has a dependency on
20364 libdld.sl. There isn't an archive version of libdld.sl. Thus,
20365 when the @option{-static} option is specified, special link options
20366 are needed to resolve this dependency.
20368 On HP-UX 10 and later, the GCC driver adds the necessary options to
20369 link with libdld.sl when the @option{-static} option is specified.
20370 This causes the resulting binary to be dynamic. On the 64-bit port,
20371 the linkers generate dynamic binaries by default in any case. The
20372 @option{-nolibdld} option can be used to prevent the GCC driver from
20373 adding these link options.
20377 Add support for multithreading with the @dfn{dce thread} library
20378 under HP-UX@. This option sets flags for both the preprocessor and
20382 @node IA-64 Options
20383 @subsection IA-64 Options
20384 @cindex IA-64 Options
20386 These are the @samp{-m} options defined for the Intel IA-64 architecture.
20390 @opindex mbig-endian
20391 Generate code for a big-endian target. This is the default for HP-UX@.
20393 @item -mlittle-endian
20394 @opindex mlittle-endian
20395 Generate code for a little-endian target. This is the default for AIX5
20401 @opindex mno-gnu-as
20402 Generate (or don't) code for the GNU assembler. This is the default.
20403 @c Also, this is the default if the configure option @option{--with-gnu-as}
20409 @opindex mno-gnu-ld
20410 Generate (or don't) code for the GNU linker. This is the default.
20411 @c Also, this is the default if the configure option @option{--with-gnu-ld}
20416 Generate code that does not use a global pointer register. The result
20417 is not position independent code, and violates the IA-64 ABI@.
20419 @item -mvolatile-asm-stop
20420 @itemx -mno-volatile-asm-stop
20421 @opindex mvolatile-asm-stop
20422 @opindex mno-volatile-asm-stop
20423 Generate (or don't) a stop bit immediately before and after volatile asm
20426 @item -mregister-names
20427 @itemx -mno-register-names
20428 @opindex mregister-names
20429 @opindex mno-register-names
20430 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
20431 the stacked registers. This may make assembler output more readable.
20437 Disable (or enable) optimizations that use the small data section. This may
20438 be useful for working around optimizer bugs.
20440 @item -mconstant-gp
20441 @opindex mconstant-gp
20442 Generate code that uses a single constant global pointer value. This is
20443 useful when compiling kernel code.
20447 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
20448 This is useful when compiling firmware code.
20450 @item -minline-float-divide-min-latency
20451 @opindex minline-float-divide-min-latency
20452 Generate code for inline divides of floating-point values
20453 using the minimum latency algorithm.
20455 @item -minline-float-divide-max-throughput
20456 @opindex minline-float-divide-max-throughput
20457 Generate code for inline divides of floating-point values
20458 using the maximum throughput algorithm.
20460 @item -mno-inline-float-divide
20461 @opindex mno-inline-float-divide
20462 Do not generate inline code for divides of floating-point values.
20464 @item -minline-int-divide-min-latency
20465 @opindex minline-int-divide-min-latency
20466 Generate code for inline divides of integer values
20467 using the minimum latency algorithm.
20469 @item -minline-int-divide-max-throughput
20470 @opindex minline-int-divide-max-throughput
20471 Generate code for inline divides of integer values
20472 using the maximum throughput algorithm.
20474 @item -mno-inline-int-divide
20475 @opindex mno-inline-int-divide
20476 @opindex minline-int-divide
20477 Do not generate inline code for divides of integer values.
20479 @item -minline-sqrt-min-latency
20480 @opindex minline-sqrt-min-latency
20481 Generate code for inline square roots
20482 using the minimum latency algorithm.
20484 @item -minline-sqrt-max-throughput
20485 @opindex minline-sqrt-max-throughput
20486 Generate code for inline square roots
20487 using the maximum throughput algorithm.
20489 @item -mno-inline-sqrt
20490 @opindex mno-inline-sqrt
20491 Do not generate inline code for @code{sqrt}.
20494 @itemx -mno-fused-madd
20495 @opindex mfused-madd
20496 @opindex mno-fused-madd
20497 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
20498 instructions. The default is to use these instructions.
20500 @item -mno-dwarf2-asm
20501 @itemx -mdwarf2-asm
20502 @opindex mno-dwarf2-asm
20503 @opindex mdwarf2-asm
20504 Don't (or do) generate assembler code for the DWARF line number debugging
20505 info. This may be useful when not using the GNU assembler.
20507 @item -mearly-stop-bits
20508 @itemx -mno-early-stop-bits
20509 @opindex mearly-stop-bits
20510 @opindex mno-early-stop-bits
20511 Allow stop bits to be placed earlier than immediately preceding the
20512 instruction that triggered the stop bit. This can improve instruction
20513 scheduling, but does not always do so.
20515 @item -mfixed-range=@var{register-range}
20516 @opindex mfixed-range
20517 Generate code treating the given register range as fixed registers.
20518 A fixed register is one that the register allocator cannot use. This is
20519 useful when compiling kernel code. A register range is specified as
20520 two registers separated by a dash. Multiple register ranges can be
20521 specified separated by a comma.
20523 @item -mtls-size=@var{tls-size}
20525 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
20528 @item -mtune=@var{cpu-type}
20530 Tune the instruction scheduling for a particular CPU, Valid values are
20531 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
20532 and @samp{mckinley}.
20538 Generate code for a 32-bit or 64-bit environment.
20539 The 32-bit environment sets int, long and pointer to 32 bits.
20540 The 64-bit environment sets int to 32 bits and long and pointer
20541 to 64 bits. These are HP-UX specific flags.
20543 @item -mno-sched-br-data-spec
20544 @itemx -msched-br-data-spec
20545 @opindex mno-sched-br-data-spec
20546 @opindex msched-br-data-spec
20547 (Dis/En)able data speculative scheduling before reload.
20548 This results in generation of @code{ld.a} instructions and
20549 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
20550 The default setting is disabled.
20552 @item -msched-ar-data-spec
20553 @itemx -mno-sched-ar-data-spec
20554 @opindex msched-ar-data-spec
20555 @opindex mno-sched-ar-data-spec
20556 (En/Dis)able data speculative scheduling after reload.
20557 This results in generation of @code{ld.a} instructions and
20558 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
20559 The default setting is enabled.
20561 @item -mno-sched-control-spec
20562 @itemx -msched-control-spec
20563 @opindex mno-sched-control-spec
20564 @opindex msched-control-spec
20565 (Dis/En)able control speculative scheduling. This feature is
20566 available only during region scheduling (i.e.@: before reload).
20567 This results in generation of the @code{ld.s} instructions and
20568 the corresponding check instructions @code{chk.s}.
20569 The default setting is disabled.
20571 @item -msched-br-in-data-spec
20572 @itemx -mno-sched-br-in-data-spec
20573 @opindex msched-br-in-data-spec
20574 @opindex mno-sched-br-in-data-spec
20575 (En/Dis)able speculative scheduling of the instructions that
20576 are dependent on the data speculative loads before reload.
20577 This is effective only with @option{-msched-br-data-spec} enabled.
20578 The default setting is enabled.
20580 @item -msched-ar-in-data-spec
20581 @itemx -mno-sched-ar-in-data-spec
20582 @opindex msched-ar-in-data-spec
20583 @opindex mno-sched-ar-in-data-spec
20584 (En/Dis)able speculative scheduling of the instructions that
20585 are dependent on the data speculative loads after reload.
20586 This is effective only with @option{-msched-ar-data-spec} enabled.
20587 The default setting is enabled.
20589 @item -msched-in-control-spec
20590 @itemx -mno-sched-in-control-spec
20591 @opindex msched-in-control-spec
20592 @opindex mno-sched-in-control-spec
20593 (En/Dis)able speculative scheduling of the instructions that
20594 are dependent on the control speculative loads.
20595 This is effective only with @option{-msched-control-spec} enabled.
20596 The default setting is enabled.
20598 @item -mno-sched-prefer-non-data-spec-insns
20599 @itemx -msched-prefer-non-data-spec-insns
20600 @opindex mno-sched-prefer-non-data-spec-insns
20601 @opindex msched-prefer-non-data-spec-insns
20602 If enabled, data-speculative instructions are chosen for schedule
20603 only if there are no other choices at the moment. This makes
20604 the use of the data speculation much more conservative.
20605 The default setting is disabled.
20607 @item -mno-sched-prefer-non-control-spec-insns
20608 @itemx -msched-prefer-non-control-spec-insns
20609 @opindex mno-sched-prefer-non-control-spec-insns
20610 @opindex msched-prefer-non-control-spec-insns
20611 If enabled, control-speculative instructions are chosen for schedule
20612 only if there are no other choices at the moment. This makes
20613 the use of the control speculation much more conservative.
20614 The default setting is disabled.
20616 @item -mno-sched-count-spec-in-critical-path
20617 @itemx -msched-count-spec-in-critical-path
20618 @opindex mno-sched-count-spec-in-critical-path
20619 @opindex msched-count-spec-in-critical-path
20620 If enabled, speculative dependencies are considered during
20621 computation of the instructions priorities. This makes the use of the
20622 speculation a bit more conservative.
20623 The default setting is disabled.
20625 @item -msched-spec-ldc
20626 @opindex msched-spec-ldc
20627 Use a simple data speculation check. This option is on by default.
20629 @item -msched-control-spec-ldc
20630 @opindex msched-spec-ldc
20631 Use a simple check for control speculation. This option is on by default.
20633 @item -msched-stop-bits-after-every-cycle
20634 @opindex msched-stop-bits-after-every-cycle
20635 Place a stop bit after every cycle when scheduling. This option is on
20638 @item -msched-fp-mem-deps-zero-cost
20639 @opindex msched-fp-mem-deps-zero-cost
20640 Assume that floating-point stores and loads are not likely to cause a conflict
20641 when placed into the same instruction group. This option is disabled by
20644 @item -msel-sched-dont-check-control-spec
20645 @opindex msel-sched-dont-check-control-spec
20646 Generate checks for control speculation in selective scheduling.
20647 This flag is disabled by default.
20649 @item -msched-max-memory-insns=@var{max-insns}
20650 @opindex msched-max-memory-insns
20651 Limit on the number of memory insns per instruction group, giving lower
20652 priority to subsequent memory insns attempting to schedule in the same
20653 instruction group. Frequently useful to prevent cache bank conflicts.
20654 The default value is 1.
20656 @item -msched-max-memory-insns-hard-limit
20657 @opindex msched-max-memory-insns-hard-limit
20658 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
20659 disallowing more than that number in an instruction group.
20660 Otherwise, the limit is ``soft'', meaning that non-memory operations
20661 are preferred when the limit is reached, but memory operations may still
20667 @subsection LM32 Options
20668 @cindex LM32 options
20670 These @option{-m} options are defined for the LatticeMico32 architecture:
20673 @item -mbarrel-shift-enabled
20674 @opindex mbarrel-shift-enabled
20675 Enable barrel-shift instructions.
20677 @item -mdivide-enabled
20678 @opindex mdivide-enabled
20679 Enable divide and modulus instructions.
20681 @item -mmultiply-enabled
20682 @opindex multiply-enabled
20683 Enable multiply instructions.
20685 @item -msign-extend-enabled
20686 @opindex msign-extend-enabled
20687 Enable sign extend instructions.
20689 @item -muser-enabled
20690 @opindex muser-enabled
20691 Enable user-defined instructions.
20696 @subsection M32C Options
20697 @cindex M32C options
20700 @item -mcpu=@var{name}
20702 Select the CPU for which code is generated. @var{name} may be one of
20703 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
20704 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
20705 the M32C/80 series.
20709 Specifies that the program will be run on the simulator. This causes
20710 an alternate runtime library to be linked in which supports, for
20711 example, file I/O@. You must not use this option when generating
20712 programs that will run on real hardware; you must provide your own
20713 runtime library for whatever I/O functions are needed.
20715 @item -memregs=@var{number}
20717 Specifies the number of memory-based pseudo-registers GCC uses
20718 during code generation. These pseudo-registers are used like real
20719 registers, so there is a tradeoff between GCC's ability to fit the
20720 code into available registers, and the performance penalty of using
20721 memory instead of registers. Note that all modules in a program must
20722 be compiled with the same value for this option. Because of that, you
20723 must not use this option with GCC's default runtime libraries.
20727 @node M32R/D Options
20728 @subsection M32R/D Options
20729 @cindex M32R/D options
20731 These @option{-m} options are defined for Renesas M32R/D architectures:
20736 Generate code for the M32R/2@.
20740 Generate code for the M32R/X@.
20744 Generate code for the M32R@. This is the default.
20746 @item -mmodel=small
20747 @opindex mmodel=small
20748 Assume all objects live in the lower 16MB of memory (so that their addresses
20749 can be loaded with the @code{ld24} instruction), and assume all subroutines
20750 are reachable with the @code{bl} instruction.
20751 This is the default.
20753 The addressability of a particular object can be set with the
20754 @code{model} attribute.
20756 @item -mmodel=medium
20757 @opindex mmodel=medium
20758 Assume objects may be anywhere in the 32-bit address space (the compiler
20759 generates @code{seth/add3} instructions to load their addresses), and
20760 assume all subroutines are reachable with the @code{bl} instruction.
20762 @item -mmodel=large
20763 @opindex mmodel=large
20764 Assume objects may be anywhere in the 32-bit address space (the compiler
20765 generates @code{seth/add3} instructions to load their addresses), and
20766 assume subroutines may not be reachable with the @code{bl} instruction
20767 (the compiler generates the much slower @code{seth/add3/jl}
20768 instruction sequence).
20771 @opindex msdata=none
20772 Disable use of the small data area. Variables are put into
20773 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
20774 @code{section} attribute has been specified).
20775 This is the default.
20777 The small data area consists of sections @code{.sdata} and @code{.sbss}.
20778 Objects may be explicitly put in the small data area with the
20779 @code{section} attribute using one of these sections.
20781 @item -msdata=sdata
20782 @opindex msdata=sdata
20783 Put small global and static data in the small data area, but do not
20784 generate special code to reference them.
20787 @opindex msdata=use
20788 Put small global and static data in the small data area, and generate
20789 special instructions to reference them.
20793 @cindex smaller data references
20794 Put global and static objects less than or equal to @var{num} bytes
20795 into the small data or BSS sections instead of the normal data or BSS
20796 sections. The default value of @var{num} is 8.
20797 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
20798 for this option to have any effect.
20800 All modules should be compiled with the same @option{-G @var{num}} value.
20801 Compiling with different values of @var{num} may or may not work; if it
20802 doesn't the linker gives an error message---incorrect code is not
20807 Makes the M32R-specific code in the compiler display some statistics
20808 that might help in debugging programs.
20810 @item -malign-loops
20811 @opindex malign-loops
20812 Align all loops to a 32-byte boundary.
20814 @item -mno-align-loops
20815 @opindex mno-align-loops
20816 Do not enforce a 32-byte alignment for loops. This is the default.
20818 @item -missue-rate=@var{number}
20819 @opindex missue-rate=@var{number}
20820 Issue @var{number} instructions per cycle. @var{number} can only be 1
20823 @item -mbranch-cost=@var{number}
20824 @opindex mbranch-cost=@var{number}
20825 @var{number} can only be 1 or 2. If it is 1 then branches are
20826 preferred over conditional code, if it is 2, then the opposite applies.
20828 @item -mflush-trap=@var{number}
20829 @opindex mflush-trap=@var{number}
20830 Specifies the trap number to use to flush the cache. The default is
20831 12. Valid numbers are between 0 and 15 inclusive.
20833 @item -mno-flush-trap
20834 @opindex mno-flush-trap
20835 Specifies that the cache cannot be flushed by using a trap.
20837 @item -mflush-func=@var{name}
20838 @opindex mflush-func=@var{name}
20839 Specifies the name of the operating system function to call to flush
20840 the cache. The default is @samp{_flush_cache}, but a function call
20841 is only used if a trap is not available.
20843 @item -mno-flush-func
20844 @opindex mno-flush-func
20845 Indicates that there is no OS function for flushing the cache.
20849 @node M680x0 Options
20850 @subsection M680x0 Options
20851 @cindex M680x0 options
20853 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
20854 The default settings depend on which architecture was selected when
20855 the compiler was configured; the defaults for the most common choices
20859 @item -march=@var{arch}
20861 Generate code for a specific M680x0 or ColdFire instruction set
20862 architecture. Permissible values of @var{arch} for M680x0
20863 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
20864 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
20865 architectures are selected according to Freescale's ISA classification
20866 and the permissible values are: @samp{isaa}, @samp{isaaplus},
20867 @samp{isab} and @samp{isac}.
20869 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
20870 code for a ColdFire target. The @var{arch} in this macro is one of the
20871 @option{-march} arguments given above.
20873 When used together, @option{-march} and @option{-mtune} select code
20874 that runs on a family of similar processors but that is optimized
20875 for a particular microarchitecture.
20877 @item -mcpu=@var{cpu}
20879 Generate code for a specific M680x0 or ColdFire processor.
20880 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
20881 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
20882 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
20883 below, which also classifies the CPUs into families:
20885 @multitable @columnfractions 0.20 0.80
20886 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
20887 @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}
20888 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
20889 @item @samp{5206e} @tab @samp{5206e}
20890 @item @samp{5208} @tab @samp{5207} @samp{5208}
20891 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
20892 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
20893 @item @samp{5216} @tab @samp{5214} @samp{5216}
20894 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
20895 @item @samp{5225} @tab @samp{5224} @samp{5225}
20896 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
20897 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
20898 @item @samp{5249} @tab @samp{5249}
20899 @item @samp{5250} @tab @samp{5250}
20900 @item @samp{5271} @tab @samp{5270} @samp{5271}
20901 @item @samp{5272} @tab @samp{5272}
20902 @item @samp{5275} @tab @samp{5274} @samp{5275}
20903 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
20904 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
20905 @item @samp{5307} @tab @samp{5307}
20906 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
20907 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
20908 @item @samp{5407} @tab @samp{5407}
20909 @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}
20912 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
20913 @var{arch} is compatible with @var{cpu}. Other combinations of
20914 @option{-mcpu} and @option{-march} are rejected.
20916 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
20917 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
20918 where the value of @var{family} is given by the table above.
20920 @item -mtune=@var{tune}
20922 Tune the code for a particular microarchitecture within the
20923 constraints set by @option{-march} and @option{-mcpu}.
20924 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
20925 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
20926 and @samp{cpu32}. The ColdFire microarchitectures
20927 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
20929 You can also use @option{-mtune=68020-40} for code that needs
20930 to run relatively well on 68020, 68030 and 68040 targets.
20931 @option{-mtune=68020-60} is similar but includes 68060 targets
20932 as well. These two options select the same tuning decisions as
20933 @option{-m68020-40} and @option{-m68020-60} respectively.
20935 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
20936 when tuning for 680x0 architecture @var{arch}. It also defines
20937 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
20938 option is used. If GCC is tuning for a range of architectures,
20939 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
20940 it defines the macros for every architecture in the range.
20942 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
20943 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
20944 of the arguments given above.
20950 Generate output for a 68000. This is the default
20951 when the compiler is configured for 68000-based systems.
20952 It is equivalent to @option{-march=68000}.
20954 Use this option for microcontrollers with a 68000 or EC000 core,
20955 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
20959 Generate output for a 68010. This is the default
20960 when the compiler is configured for 68010-based systems.
20961 It is equivalent to @option{-march=68010}.
20967 Generate output for a 68020. This is the default
20968 when the compiler is configured for 68020-based systems.
20969 It is equivalent to @option{-march=68020}.
20973 Generate output for a 68030. This is the default when the compiler is
20974 configured for 68030-based systems. It is equivalent to
20975 @option{-march=68030}.
20979 Generate output for a 68040. This is the default when the compiler is
20980 configured for 68040-based systems. It is equivalent to
20981 @option{-march=68040}.
20983 This option inhibits the use of 68881/68882 instructions that have to be
20984 emulated by software on the 68040. Use this option if your 68040 does not
20985 have code to emulate those instructions.
20989 Generate output for a 68060. This is the default when the compiler is
20990 configured for 68060-based systems. It is equivalent to
20991 @option{-march=68060}.
20993 This option inhibits the use of 68020 and 68881/68882 instructions that
20994 have to be emulated by software on the 68060. Use this option if your 68060
20995 does not have code to emulate those instructions.
20999 Generate output for a CPU32. This is the default
21000 when the compiler is configured for CPU32-based systems.
21001 It is equivalent to @option{-march=cpu32}.
21003 Use this option for microcontrollers with a
21004 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
21005 68336, 68340, 68341, 68349 and 68360.
21009 Generate output for a 520X ColdFire CPU@. This is the default
21010 when the compiler is configured for 520X-based systems.
21011 It is equivalent to @option{-mcpu=5206}, and is now deprecated
21012 in favor of that option.
21014 Use this option for microcontroller with a 5200 core, including
21015 the MCF5202, MCF5203, MCF5204 and MCF5206.
21019 Generate output for a 5206e ColdFire CPU@. The option is now
21020 deprecated in favor of the equivalent @option{-mcpu=5206e}.
21024 Generate output for a member of the ColdFire 528X family.
21025 The option is now deprecated in favor of the equivalent
21026 @option{-mcpu=528x}.
21030 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
21031 in favor of the equivalent @option{-mcpu=5307}.
21035 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
21036 in favor of the equivalent @option{-mcpu=5407}.
21040 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
21041 This includes use of hardware floating-point instructions.
21042 The option is equivalent to @option{-mcpu=547x}, and is now
21043 deprecated in favor of that option.
21047 Generate output for a 68040, without using any of the new instructions.
21048 This results in code that can run relatively efficiently on either a
21049 68020/68881 or a 68030 or a 68040. The generated code does use the
21050 68881 instructions that are emulated on the 68040.
21052 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
21056 Generate output for a 68060, without using any of the new instructions.
21057 This results in code that can run relatively efficiently on either a
21058 68020/68881 or a 68030 or a 68040. The generated code does use the
21059 68881 instructions that are emulated on the 68060.
21061 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
21065 @opindex mhard-float
21067 Generate floating-point instructions. This is the default for 68020
21068 and above, and for ColdFire devices that have an FPU@. It defines the
21069 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
21070 on ColdFire targets.
21073 @opindex msoft-float
21074 Do not generate floating-point instructions; use library calls instead.
21075 This is the default for 68000, 68010, and 68832 targets. It is also
21076 the default for ColdFire devices that have no FPU.
21082 Generate (do not generate) ColdFire hardware divide and remainder
21083 instructions. If @option{-march} is used without @option{-mcpu},
21084 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
21085 architectures. Otherwise, the default is taken from the target CPU
21086 (either the default CPU, or the one specified by @option{-mcpu}). For
21087 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
21088 @option{-mcpu=5206e}.
21090 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
21094 Consider type @code{int} to be 16 bits wide, like @code{short int}.
21095 Additionally, parameters passed on the stack are also aligned to a
21096 16-bit boundary even on targets whose API mandates promotion to 32-bit.
21100 Do not consider type @code{int} to be 16 bits wide. This is the default.
21103 @itemx -mno-bitfield
21104 @opindex mnobitfield
21105 @opindex mno-bitfield
21106 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
21107 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
21111 Do use the bit-field instructions. The @option{-m68020} option implies
21112 @option{-mbitfield}. This is the default if you use a configuration
21113 designed for a 68020.
21117 Use a different function-calling convention, in which functions
21118 that take a fixed number of arguments return with the @code{rtd}
21119 instruction, which pops their arguments while returning. This
21120 saves one instruction in the caller since there is no need to pop
21121 the arguments there.
21123 This calling convention is incompatible with the one normally
21124 used on Unix, so you cannot use it if you need to call libraries
21125 compiled with the Unix compiler.
21127 Also, you must provide function prototypes for all functions that
21128 take variable numbers of arguments (including @code{printf});
21129 otherwise incorrect code is generated for calls to those
21132 In addition, seriously incorrect code results if you call a
21133 function with too many arguments. (Normally, extra arguments are
21134 harmlessly ignored.)
21136 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
21137 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
21139 The default is @option{-mno-rtd}.
21142 @itemx -mno-align-int
21143 @opindex malign-int
21144 @opindex mno-align-int
21145 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
21146 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
21147 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
21148 Aligning variables on 32-bit boundaries produces code that runs somewhat
21149 faster on processors with 32-bit busses at the expense of more memory.
21151 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
21152 aligns structures containing the above types differently than
21153 most published application binary interface specifications for the m68k.
21157 Use the pc-relative addressing mode of the 68000 directly, instead of
21158 using a global offset table. At present, this option implies @option{-fpic},
21159 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
21160 not presently supported with @option{-mpcrel}, though this could be supported for
21161 68020 and higher processors.
21163 @item -mno-strict-align
21164 @itemx -mstrict-align
21165 @opindex mno-strict-align
21166 @opindex mstrict-align
21167 Do not (do) assume that unaligned memory references are handled by
21171 Generate code that allows the data segment to be located in a different
21172 area of memory from the text segment. This allows for execute-in-place in
21173 an environment without virtual memory management. This option implies
21176 @item -mno-sep-data
21177 Generate code that assumes that the data segment follows the text segment.
21178 This is the default.
21180 @item -mid-shared-library
21181 Generate code that supports shared libraries via the library ID method.
21182 This allows for execute-in-place and shared libraries in an environment
21183 without virtual memory management. This option implies @option{-fPIC}.
21185 @item -mno-id-shared-library
21186 Generate code that doesn't assume ID-based shared libraries are being used.
21187 This is the default.
21189 @item -mshared-library-id=n
21190 Specifies the identification number of the ID-based shared library being
21191 compiled. Specifying a value of 0 generates more compact code; specifying
21192 other values forces the allocation of that number to the current
21193 library, but is no more space- or time-efficient than omitting this option.
21199 When generating position-independent code for ColdFire, generate code
21200 that works if the GOT has more than 8192 entries. This code is
21201 larger and slower than code generated without this option. On M680x0
21202 processors, this option is not needed; @option{-fPIC} suffices.
21204 GCC normally uses a single instruction to load values from the GOT@.
21205 While this is relatively efficient, it only works if the GOT
21206 is smaller than about 64k. Anything larger causes the linker
21207 to report an error such as:
21209 @cindex relocation truncated to fit (ColdFire)
21211 relocation truncated to fit: R_68K_GOT16O foobar
21214 If this happens, you should recompile your code with @option{-mxgot}.
21215 It should then work with very large GOTs. However, code generated with
21216 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
21217 the value of a global symbol.
21219 Note that some linkers, including newer versions of the GNU linker,
21220 can create multiple GOTs and sort GOT entries. If you have such a linker,
21221 you should only need to use @option{-mxgot} when compiling a single
21222 object file that accesses more than 8192 GOT entries. Very few do.
21224 These options have no effect unless GCC is generating
21225 position-independent code.
21227 @item -mlong-jump-table-offsets
21228 @opindex mlong-jump-table-offsets
21229 Use 32-bit offsets in @code{switch} tables. The default is to use
21234 @node MCore Options
21235 @subsection MCore Options
21236 @cindex MCore options
21238 These are the @samp{-m} options defined for the Motorola M*Core
21244 @itemx -mno-hardlit
21246 @opindex mno-hardlit
21247 Inline constants into the code stream if it can be done in two
21248 instructions or less.
21254 Use the divide instruction. (Enabled by default).
21256 @item -mrelax-immediate
21257 @itemx -mno-relax-immediate
21258 @opindex mrelax-immediate
21259 @opindex mno-relax-immediate
21260 Allow arbitrary-sized immediates in bit operations.
21262 @item -mwide-bitfields
21263 @itemx -mno-wide-bitfields
21264 @opindex mwide-bitfields
21265 @opindex mno-wide-bitfields
21266 Always treat bit-fields as @code{int}-sized.
21268 @item -m4byte-functions
21269 @itemx -mno-4byte-functions
21270 @opindex m4byte-functions
21271 @opindex mno-4byte-functions
21272 Force all functions to be aligned to a 4-byte boundary.
21274 @item -mcallgraph-data
21275 @itemx -mno-callgraph-data
21276 @opindex mcallgraph-data
21277 @opindex mno-callgraph-data
21278 Emit callgraph information.
21281 @itemx -mno-slow-bytes
21282 @opindex mslow-bytes
21283 @opindex mno-slow-bytes
21284 Prefer word access when reading byte quantities.
21286 @item -mlittle-endian
21287 @itemx -mbig-endian
21288 @opindex mlittle-endian
21289 @opindex mbig-endian
21290 Generate code for a little-endian target.
21296 Generate code for the 210 processor.
21300 Assume that runtime support has been provided and so omit the
21301 simulator library (@file{libsim.a)} from the linker command line.
21303 @item -mstack-increment=@var{size}
21304 @opindex mstack-increment
21305 Set the maximum amount for a single stack increment operation. Large
21306 values can increase the speed of programs that contain functions
21307 that need a large amount of stack space, but they can also trigger a
21308 segmentation fault if the stack is extended too much. The default
21314 @subsection MeP Options
21315 @cindex MeP options
21321 Enables the @code{abs} instruction, which is the absolute difference
21322 between two registers.
21326 Enables all the optional instructions---average, multiply, divide, bit
21327 operations, leading zero, absolute difference, min/max, clip, and
21333 Enables the @code{ave} instruction, which computes the average of two
21336 @item -mbased=@var{n}
21338 Variables of size @var{n} bytes or smaller are placed in the
21339 @code{.based} section by default. Based variables use the @code{$tp}
21340 register as a base register, and there is a 128-byte limit to the
21341 @code{.based} section.
21345 Enables the bit operation instructions---bit test (@code{btstm}), set
21346 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
21347 test-and-set (@code{tas}).
21349 @item -mc=@var{name}
21351 Selects which section constant data is placed in. @var{name} may
21352 be @samp{tiny}, @samp{near}, or @samp{far}.
21356 Enables the @code{clip} instruction. Note that @option{-mclip} is not
21357 useful unless you also provide @option{-mminmax}.
21359 @item -mconfig=@var{name}
21361 Selects one of the built-in core configurations. Each MeP chip has
21362 one or more modules in it; each module has a core CPU and a variety of
21363 coprocessors, optional instructions, and peripherals. The
21364 @code{MeP-Integrator} tool, not part of GCC, provides these
21365 configurations through this option; using this option is the same as
21366 using all the corresponding command-line options. The default
21367 configuration is @samp{default}.
21371 Enables the coprocessor instructions. By default, this is a 32-bit
21372 coprocessor. Note that the coprocessor is normally enabled via the
21373 @option{-mconfig=} option.
21377 Enables the 32-bit coprocessor's instructions.
21381 Enables the 64-bit coprocessor's instructions.
21385 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
21389 Causes constant variables to be placed in the @code{.near} section.
21393 Enables the @code{div} and @code{divu} instructions.
21397 Generate big-endian code.
21401 Generate little-endian code.
21403 @item -mio-volatile
21404 @opindex mio-volatile
21405 Tells the compiler that any variable marked with the @code{io}
21406 attribute is to be considered volatile.
21410 Causes variables to be assigned to the @code{.far} section by default.
21414 Enables the @code{leadz} (leading zero) instruction.
21418 Causes variables to be assigned to the @code{.near} section by default.
21422 Enables the @code{min} and @code{max} instructions.
21426 Enables the multiplication and multiply-accumulate instructions.
21430 Disables all the optional instructions enabled by @option{-mall-opts}.
21434 Enables the @code{repeat} and @code{erepeat} instructions, used for
21435 low-overhead looping.
21439 Causes all variables to default to the @code{.tiny} section. Note
21440 that there is a 65536-byte limit to this section. Accesses to these
21441 variables use the @code{%gp} base register.
21445 Enables the saturation instructions. Note that the compiler does not
21446 currently generate these itself, but this option is included for
21447 compatibility with other tools, like @code{as}.
21451 Link the SDRAM-based runtime instead of the default ROM-based runtime.
21455 Link the simulator run-time libraries.
21459 Link the simulator runtime libraries, excluding built-in support
21460 for reset and exception vectors and tables.
21464 Causes all functions to default to the @code{.far} section. Without
21465 this option, functions default to the @code{.near} section.
21467 @item -mtiny=@var{n}
21469 Variables that are @var{n} bytes or smaller are allocated to the
21470 @code{.tiny} section. These variables use the @code{$gp} base
21471 register. The default for this option is 4, but note that there's a
21472 65536-byte limit to the @code{.tiny} section.
21476 @node MicroBlaze Options
21477 @subsection MicroBlaze Options
21478 @cindex MicroBlaze Options
21483 @opindex msoft-float
21484 Use software emulation for floating point (default).
21487 @opindex mhard-float
21488 Use hardware floating-point instructions.
21492 Do not optimize block moves, use @code{memcpy}.
21494 @item -mno-clearbss
21495 @opindex mno-clearbss
21496 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
21498 @item -mcpu=@var{cpu-type}
21500 Use features of, and schedule code for, the given CPU.
21501 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
21502 where @var{X} is a major version, @var{YY} is the minor version, and
21503 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
21504 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v6.00.a}.
21506 @item -mxl-soft-mul
21507 @opindex mxl-soft-mul
21508 Use software multiply emulation (default).
21510 @item -mxl-soft-div
21511 @opindex mxl-soft-div
21512 Use software emulation for divides (default).
21514 @item -mxl-barrel-shift
21515 @opindex mxl-barrel-shift
21516 Use the hardware barrel shifter.
21518 @item -mxl-pattern-compare
21519 @opindex mxl-pattern-compare
21520 Use pattern compare instructions.
21522 @item -msmall-divides
21523 @opindex msmall-divides
21524 Use table lookup optimization for small signed integer divisions.
21526 @item -mxl-stack-check
21527 @opindex mxl-stack-check
21528 This option is deprecated. Use @option{-fstack-check} instead.
21531 @opindex mxl-gp-opt
21532 Use GP-relative @code{.sdata}/@code{.sbss} sections.
21534 @item -mxl-multiply-high
21535 @opindex mxl-multiply-high
21536 Use multiply high instructions for high part of 32x32 multiply.
21538 @item -mxl-float-convert
21539 @opindex mxl-float-convert
21540 Use hardware floating-point conversion instructions.
21542 @item -mxl-float-sqrt
21543 @opindex mxl-float-sqrt
21544 Use hardware floating-point square root instruction.
21547 @opindex mbig-endian
21548 Generate code for a big-endian target.
21550 @item -mlittle-endian
21551 @opindex mlittle-endian
21552 Generate code for a little-endian target.
21555 @opindex mxl-reorder
21556 Use reorder instructions (swap and byte reversed load/store).
21558 @item -mxl-mode-@var{app-model}
21559 Select application model @var{app-model}. Valid models are
21562 normal executable (default), uses startup code @file{crt0.o}.
21564 @item -mpic-data-is-text-relative
21565 @opindex mpic-data-is-text-relative
21566 Assume that the displacement between the text and data segments is fixed
21567 at static link time. This allows data to be referenced by offset from start of
21568 text address instead of GOT since PC-relative addressing is not supported.
21571 for use with Xilinx Microprocessor Debugger (XMD) based
21572 software intrusive debug agent called xmdstub. This uses startup file
21573 @file{crt1.o} and sets the start address of the program to 0x800.
21576 for applications that are loaded using a bootloader.
21577 This model uses startup file @file{crt2.o} which does not contain a processor
21578 reset vector handler. This is suitable for transferring control on a
21579 processor reset to the bootloader rather than the application.
21582 for applications that do not require any of the
21583 MicroBlaze vectors. This option may be useful for applications running
21584 within a monitoring application. This model uses @file{crt3.o} as a startup file.
21587 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
21588 @option{-mxl-mode-@var{app-model}}.
21593 @subsection MIPS Options
21594 @cindex MIPS options
21600 Generate big-endian code.
21604 Generate little-endian code. This is the default for @samp{mips*el-*-*}
21607 @item -march=@var{arch}
21609 Generate code that runs on @var{arch}, which can be the name of a
21610 generic MIPS ISA, or the name of a particular processor.
21612 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
21613 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
21614 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
21615 @samp{mips64r5} and @samp{mips64r6}.
21616 The processor names are:
21617 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
21618 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
21619 @samp{5kc}, @samp{5kf},
21621 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
21622 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
21623 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
21624 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
21625 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
21626 @samp{i6400}, @samp{i6500},
21628 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a}, @samp{gs464},
21629 @samp{gs464e}, @samp{gs264e},
21631 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
21632 @samp{m5100}, @samp{m5101},
21633 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
21635 @samp{p5600}, @samp{p6600},
21636 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
21637 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r5900},
21638 @samp{r6000}, @samp{r8000},
21639 @samp{rm7000}, @samp{rm9000},
21640 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
21643 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
21644 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
21645 @samp{xlr} and @samp{xlp}.
21646 The special value @samp{from-abi} selects the
21647 most compatible architecture for the selected ABI (that is,
21648 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
21650 The native Linux/GNU toolchain also supports the value @samp{native},
21651 which selects the best architecture option for the host processor.
21652 @option{-march=native} has no effect if GCC does not recognize
21655 In processor names, a final @samp{000} can be abbreviated as @samp{k}
21656 (for example, @option{-march=r2k}). Prefixes are optional, and
21657 @samp{vr} may be written @samp{r}.
21659 Names of the form @samp{@var{n}f2_1} refer to processors with
21660 FPUs clocked at half the rate of the core, names of the form
21661 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
21662 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
21663 processors with FPUs clocked a ratio of 3:2 with respect to the core.
21664 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
21665 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
21666 accepted as synonyms for @samp{@var{n}f1_1}.
21668 GCC defines two macros based on the value of this option. The first
21669 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
21670 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
21671 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
21672 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
21673 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
21675 Note that the @code{_MIPS_ARCH} macro uses the processor names given
21676 above. In other words, it has the full prefix and does not
21677 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
21678 the macro names the resolved architecture (either @code{"mips1"} or
21679 @code{"mips3"}). It names the default architecture when no
21680 @option{-march} option is given.
21682 @item -mtune=@var{arch}
21684 Optimize for @var{arch}. Among other things, this option controls
21685 the way instructions are scheduled, and the perceived cost of arithmetic
21686 operations. The list of @var{arch} values is the same as for
21689 When this option is not used, GCC optimizes for the processor
21690 specified by @option{-march}. By using @option{-march} and
21691 @option{-mtune} together, it is possible to generate code that
21692 runs on a family of processors, but optimize the code for one
21693 particular member of that family.
21695 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
21696 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
21697 @option{-march} ones described above.
21701 Equivalent to @option{-march=mips1}.
21705 Equivalent to @option{-march=mips2}.
21709 Equivalent to @option{-march=mips3}.
21713 Equivalent to @option{-march=mips4}.
21717 Equivalent to @option{-march=mips32}.
21721 Equivalent to @option{-march=mips32r3}.
21725 Equivalent to @option{-march=mips32r5}.
21729 Equivalent to @option{-march=mips32r6}.
21733 Equivalent to @option{-march=mips64}.
21737 Equivalent to @option{-march=mips64r2}.
21741 Equivalent to @option{-march=mips64r3}.
21745 Equivalent to @option{-march=mips64r5}.
21749 Equivalent to @option{-march=mips64r6}.
21754 @opindex mno-mips16
21755 Generate (do not generate) MIPS16 code. If GCC is targeting a
21756 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
21758 MIPS16 code generation can also be controlled on a per-function basis
21759 by means of @code{mips16} and @code{nomips16} attributes.
21760 @xref{Function Attributes}, for more information.
21762 @item -mflip-mips16
21763 @opindex mflip-mips16
21764 Generate MIPS16 code on alternating functions. This option is provided
21765 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
21766 not intended for ordinary use in compiling user code.
21768 @item -minterlink-compressed
21769 @itemx -mno-interlink-compressed
21770 @opindex minterlink-compressed
21771 @opindex mno-interlink-compressed
21772 Require (do not require) that code using the standard (uncompressed) MIPS ISA
21773 be link-compatible with MIPS16 and microMIPS code, and vice versa.
21775 For example, code using the standard ISA encoding cannot jump directly
21776 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
21777 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
21778 knows that the target of the jump is not compressed.
21780 @item -minterlink-mips16
21781 @itemx -mno-interlink-mips16
21782 @opindex minterlink-mips16
21783 @opindex mno-interlink-mips16
21784 Aliases of @option{-minterlink-compressed} and
21785 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
21786 and are retained for backwards compatibility.
21798 Generate code for the given ABI@.
21800 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
21801 generates 64-bit code when you select a 64-bit architecture, but you
21802 can use @option{-mgp32} to get 32-bit code instead.
21804 For information about the O64 ABI, see
21805 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
21807 GCC supports a variant of the o32 ABI in which floating-point registers
21808 are 64 rather than 32 bits wide. You can select this combination with
21809 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
21810 and @code{mfhc1} instructions and is therefore only supported for
21811 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
21813 The register assignments for arguments and return values remain the
21814 same, but each scalar value is passed in a single 64-bit register
21815 rather than a pair of 32-bit registers. For example, scalar
21816 floating-point values are returned in @samp{$f0} only, not a
21817 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
21818 remains the same in that the even-numbered double-precision registers
21821 Two additional variants of the o32 ABI are supported to enable
21822 a transition from 32-bit to 64-bit registers. These are FPXX
21823 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
21824 The FPXX extension mandates that all code must execute correctly
21825 when run using 32-bit or 64-bit registers. The code can be interlinked
21826 with either FP32 or FP64, but not both.
21827 The FP64A extension is similar to the FP64 extension but forbids the
21828 use of odd-numbered single-precision registers. This can be used
21829 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
21830 processors and allows both FP32 and FP64A code to interlink and
21831 run in the same process without changing FPU modes.
21834 @itemx -mno-abicalls
21836 @opindex mno-abicalls
21837 Generate (do not generate) code that is suitable for SVR4-style
21838 dynamic objects. @option{-mabicalls} is the default for SVR4-based
21843 Generate (do not generate) code that is fully position-independent,
21844 and that can therefore be linked into shared libraries. This option
21845 only affects @option{-mabicalls}.
21847 All @option{-mabicalls} code has traditionally been position-independent,
21848 regardless of options like @option{-fPIC} and @option{-fpic}. However,
21849 as an extension, the GNU toolchain allows executables to use absolute
21850 accesses for locally-binding symbols. It can also use shorter GP
21851 initialization sequences and generate direct calls to locally-defined
21852 functions. This mode is selected by @option{-mno-shared}.
21854 @option{-mno-shared} depends on binutils 2.16 or higher and generates
21855 objects that can only be linked by the GNU linker. However, the option
21856 does not affect the ABI of the final executable; it only affects the ABI
21857 of relocatable objects. Using @option{-mno-shared} generally makes
21858 executables both smaller and quicker.
21860 @option{-mshared} is the default.
21866 Assume (do not assume) that the static and dynamic linkers
21867 support PLTs and copy relocations. This option only affects
21868 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
21869 has no effect without @option{-msym32}.
21871 You can make @option{-mplt} the default by configuring
21872 GCC with @option{--with-mips-plt}. The default is
21873 @option{-mno-plt} otherwise.
21879 Lift (do not lift) the usual restrictions on the size of the global
21882 GCC normally uses a single instruction to load values from the GOT@.
21883 While this is relatively efficient, it only works if the GOT
21884 is smaller than about 64k. Anything larger causes the linker
21885 to report an error such as:
21887 @cindex relocation truncated to fit (MIPS)
21889 relocation truncated to fit: R_MIPS_GOT16 foobar
21892 If this happens, you should recompile your code with @option{-mxgot}.
21893 This works with very large GOTs, although the code is also
21894 less efficient, since it takes three instructions to fetch the
21895 value of a global symbol.
21897 Note that some linkers can create multiple GOTs. If you have such a
21898 linker, you should only need to use @option{-mxgot} when a single object
21899 file accesses more than 64k's worth of GOT entries. Very few do.
21901 These options have no effect unless GCC is generating position
21906 Assume that general-purpose registers are 32 bits wide.
21910 Assume that general-purpose registers are 64 bits wide.
21914 Assume that floating-point registers are 32 bits wide.
21918 Assume that floating-point registers are 64 bits wide.
21922 Do not assume the width of floating-point registers.
21925 @opindex mhard-float
21926 Use floating-point coprocessor instructions.
21929 @opindex msoft-float
21930 Do not use floating-point coprocessor instructions. Implement
21931 floating-point calculations using library calls instead.
21935 Equivalent to @option{-msoft-float}, but additionally asserts that the
21936 program being compiled does not perform any floating-point operations.
21937 This option is presently supported only by some bare-metal MIPS
21938 configurations, where it may select a special set of libraries
21939 that lack all floating-point support (including, for example, the
21940 floating-point @code{printf} formats).
21941 If code compiled with @option{-mno-float} accidentally contains
21942 floating-point operations, it is likely to suffer a link-time
21943 or run-time failure.
21945 @item -msingle-float
21946 @opindex msingle-float
21947 Assume that the floating-point coprocessor only supports single-precision
21950 @item -mdouble-float
21951 @opindex mdouble-float
21952 Assume that the floating-point coprocessor supports double-precision
21953 operations. This is the default.
21956 @itemx -mno-odd-spreg
21957 @opindex modd-spreg
21958 @opindex mno-odd-spreg
21959 Enable the use of odd-numbered single-precision floating-point registers
21960 for the o32 ABI. This is the default for processors that are known to
21961 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
21965 @itemx -mabs=legacy
21967 @opindex mabs=legacy
21968 These options control the treatment of the special not-a-number (NaN)
21969 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
21970 @code{neg.@i{fmt}} machine instructions.
21972 By default or when @option{-mabs=legacy} is used the legacy
21973 treatment is selected. In this case these instructions are considered
21974 arithmetic and avoided where correct operation is required and the
21975 input operand might be a NaN. A longer sequence of instructions that
21976 manipulate the sign bit of floating-point datum manually is used
21977 instead unless the @option{-ffinite-math-only} option has also been
21980 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
21981 this case these instructions are considered non-arithmetic and therefore
21982 operating correctly in all cases, including in particular where the
21983 input operand is a NaN. These instructions are therefore always used
21984 for the respective operations.
21987 @itemx -mnan=legacy
21989 @opindex mnan=legacy
21990 These options control the encoding of the special not-a-number (NaN)
21991 IEEE 754 floating-point data.
21993 The @option{-mnan=legacy} option selects the legacy encoding. In this
21994 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
21995 significand field being 0, whereas signaling NaNs (sNaNs) are denoted
21996 by the first bit of their trailing significand field being 1.
21998 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
21999 this case qNaNs are denoted by the first bit of their trailing
22000 significand field being 1, whereas sNaNs are denoted by the first bit of
22001 their trailing significand field being 0.
22003 The default is @option{-mnan=legacy} unless GCC has been configured with
22004 @option{--with-nan=2008}.
22010 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
22011 implement atomic memory built-in functions. When neither option is
22012 specified, GCC uses the instructions if the target architecture
22015 @option{-mllsc} is useful if the runtime environment can emulate the
22016 instructions and @option{-mno-llsc} can be useful when compiling for
22017 nonstandard ISAs. You can make either option the default by
22018 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
22019 respectively. @option{--with-llsc} is the default for some
22020 configurations; see the installation documentation for details.
22026 Use (do not use) revision 1 of the MIPS DSP ASE@.
22027 @xref{MIPS DSP Built-in Functions}. This option defines the
22028 preprocessor macro @code{__mips_dsp}. It also defines
22029 @code{__mips_dsp_rev} to 1.
22035 Use (do not use) revision 2 of the MIPS DSP ASE@.
22036 @xref{MIPS DSP Built-in Functions}. This option defines the
22037 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
22038 It also defines @code{__mips_dsp_rev} to 2.
22041 @itemx -mno-smartmips
22042 @opindex msmartmips
22043 @opindex mno-smartmips
22044 Use (do not use) the MIPS SmartMIPS ASE.
22046 @item -mpaired-single
22047 @itemx -mno-paired-single
22048 @opindex mpaired-single
22049 @opindex mno-paired-single
22050 Use (do not use) paired-single floating-point instructions.
22051 @xref{MIPS Paired-Single Support}. This option requires
22052 hardware floating-point support to be enabled.
22058 Use (do not use) MIPS Digital Media Extension instructions.
22059 This option can only be used when generating 64-bit code and requires
22060 hardware floating-point support to be enabled.
22065 @opindex mno-mips3d
22066 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
22067 The option @option{-mips3d} implies @option{-mpaired-single}.
22070 @itemx -mno-micromips
22071 @opindex mmicromips
22072 @opindex mno-mmicromips
22073 Generate (do not generate) microMIPS code.
22075 MicroMIPS code generation can also be controlled on a per-function basis
22076 by means of @code{micromips} and @code{nomicromips} attributes.
22077 @xref{Function Attributes}, for more information.
22083 Use (do not use) MT Multithreading instructions.
22089 Use (do not use) the MIPS MCU ASE instructions.
22095 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
22101 Use (do not use) the MIPS Virtualization (VZ) instructions.
22107 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
22113 Use (do not use) the MIPS Cyclic Redundancy Check (CRC) instructions.
22119 Use (do not use) the MIPS Global INValidate (GINV) instructions.
22121 @item -mloongson-mmi
22122 @itemx -mno-loongson-mmi
22123 @opindex mloongson-mmi
22124 @opindex mno-loongson-mmi
22125 Use (do not use) the MIPS Loongson MultiMedia extensions Instructions (MMI).
22127 @item -mloongson-ext
22128 @itemx -mno-loongson-ext
22129 @opindex mloongson-ext
22130 @opindex mno-loongson-ext
22131 Use (do not use) the MIPS Loongson EXTensions (EXT) instructions.
22133 @item -mloongson-ext2
22134 @itemx -mno-loongson-ext2
22135 @opindex mloongson-ext2
22136 @opindex mno-loongson-ext2
22137 Use (do not use) the MIPS Loongson EXTensions r2 (EXT2) instructions.
22141 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
22142 an explanation of the default and the way that the pointer size is
22147 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
22149 The default size of @code{int}s, @code{long}s and pointers depends on
22150 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
22151 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
22152 32-bit @code{long}s. Pointers are the same size as @code{long}s,
22153 or the same size as integer registers, whichever is smaller.
22159 Assume (do not assume) that all symbols have 32-bit values, regardless
22160 of the selected ABI@. This option is useful in combination with
22161 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
22162 to generate shorter and faster references to symbolic addresses.
22166 Put definitions of externally-visible data in a small data section
22167 if that data is no bigger than @var{num} bytes. GCC can then generate
22168 more efficient accesses to the data; see @option{-mgpopt} for details.
22170 The default @option{-G} option depends on the configuration.
22172 @item -mlocal-sdata
22173 @itemx -mno-local-sdata
22174 @opindex mlocal-sdata
22175 @opindex mno-local-sdata
22176 Extend (do not extend) the @option{-G} behavior to local data too,
22177 such as to static variables in C@. @option{-mlocal-sdata} is the
22178 default for all configurations.
22180 If the linker complains that an application is using too much small data,
22181 you might want to try rebuilding the less performance-critical parts with
22182 @option{-mno-local-sdata}. You might also want to build large
22183 libraries with @option{-mno-local-sdata}, so that the libraries leave
22184 more room for the main program.
22186 @item -mextern-sdata
22187 @itemx -mno-extern-sdata
22188 @opindex mextern-sdata
22189 @opindex mno-extern-sdata
22190 Assume (do not assume) that externally-defined data is in
22191 a small data section if the size of that data is within the @option{-G} limit.
22192 @option{-mextern-sdata} is the default for all configurations.
22194 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
22195 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
22196 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
22197 is placed in a small data section. If @var{Var} is defined by another
22198 module, you must either compile that module with a high-enough
22199 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
22200 definition. If @var{Var} is common, you must link the application
22201 with a high-enough @option{-G} setting.
22203 The easiest way of satisfying these restrictions is to compile
22204 and link every module with the same @option{-G} option. However,
22205 you may wish to build a library that supports several different
22206 small data limits. You can do this by compiling the library with
22207 the highest supported @option{-G} setting and additionally using
22208 @option{-mno-extern-sdata} to stop the library from making assumptions
22209 about externally-defined data.
22215 Use (do not use) GP-relative accesses for symbols that are known to be
22216 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
22217 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
22220 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
22221 might not hold the value of @code{_gp}. For example, if the code is
22222 part of a library that might be used in a boot monitor, programs that
22223 call boot monitor routines pass an unknown value in @code{$gp}.
22224 (In such situations, the boot monitor itself is usually compiled
22225 with @option{-G0}.)
22227 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
22228 @option{-mno-extern-sdata}.
22230 @item -membedded-data
22231 @itemx -mno-embedded-data
22232 @opindex membedded-data
22233 @opindex mno-embedded-data
22234 Allocate variables to the read-only data section first if possible, then
22235 next in the small data section if possible, otherwise in data. This gives
22236 slightly slower code than the default, but reduces the amount of RAM required
22237 when executing, and thus may be preferred for some embedded systems.
22239 @item -muninit-const-in-rodata
22240 @itemx -mno-uninit-const-in-rodata
22241 @opindex muninit-const-in-rodata
22242 @opindex mno-uninit-const-in-rodata
22243 Put uninitialized @code{const} variables in the read-only data section.
22244 This option is only meaningful in conjunction with @option{-membedded-data}.
22246 @item -mcode-readable=@var{setting}
22247 @opindex mcode-readable
22248 Specify whether GCC may generate code that reads from executable sections.
22249 There are three possible settings:
22252 @item -mcode-readable=yes
22253 Instructions may freely access executable sections. This is the
22256 @item -mcode-readable=pcrel
22257 MIPS16 PC-relative load instructions can access executable sections,
22258 but other instructions must not do so. This option is useful on 4KSc
22259 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
22260 It is also useful on processors that can be configured to have a dual
22261 instruction/data SRAM interface and that, like the M4K, automatically
22262 redirect PC-relative loads to the instruction RAM.
22264 @item -mcode-readable=no
22265 Instructions must not access executable sections. This option can be
22266 useful on targets that are configured to have a dual instruction/data
22267 SRAM interface but that (unlike the M4K) do not automatically redirect
22268 PC-relative loads to the instruction RAM.
22271 @item -msplit-addresses
22272 @itemx -mno-split-addresses
22273 @opindex msplit-addresses
22274 @opindex mno-split-addresses
22275 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
22276 relocation operators. This option has been superseded by
22277 @option{-mexplicit-relocs} but is retained for backwards compatibility.
22279 @item -mexplicit-relocs
22280 @itemx -mno-explicit-relocs
22281 @opindex mexplicit-relocs
22282 @opindex mno-explicit-relocs
22283 Use (do not use) assembler relocation operators when dealing with symbolic
22284 addresses. The alternative, selected by @option{-mno-explicit-relocs},
22285 is to use assembler macros instead.
22287 @option{-mexplicit-relocs} is the default if GCC was configured
22288 to use an assembler that supports relocation operators.
22290 @item -mcheck-zero-division
22291 @itemx -mno-check-zero-division
22292 @opindex mcheck-zero-division
22293 @opindex mno-check-zero-division
22294 Trap (do not trap) on integer division by zero.
22296 The default is @option{-mcheck-zero-division}.
22298 @item -mdivide-traps
22299 @itemx -mdivide-breaks
22300 @opindex mdivide-traps
22301 @opindex mdivide-breaks
22302 MIPS systems check for division by zero by generating either a
22303 conditional trap or a break instruction. Using traps results in
22304 smaller code, but is only supported on MIPS II and later. Also, some
22305 versions of the Linux kernel have a bug that prevents trap from
22306 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
22307 allow conditional traps on architectures that support them and
22308 @option{-mdivide-breaks} to force the use of breaks.
22310 The default is usually @option{-mdivide-traps}, but this can be
22311 overridden at configure time using @option{--with-divide=breaks}.
22312 Divide-by-zero checks can be completely disabled using
22313 @option{-mno-check-zero-division}.
22315 @item -mload-store-pairs
22316 @itemx -mno-load-store-pairs
22317 @opindex mload-store-pairs
22318 @opindex mno-load-store-pairs
22319 Enable (disable) an optimization that pairs consecutive load or store
22320 instructions to enable load/store bonding. This option is enabled by
22321 default but only takes effect when the selected architecture is known
22322 to support bonding.
22327 @opindex mno-memcpy
22328 Force (do not force) the use of @code{memcpy} for non-trivial block
22329 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
22330 most constant-sized copies.
22333 @itemx -mno-long-calls
22334 @opindex mlong-calls
22335 @opindex mno-long-calls
22336 Disable (do not disable) use of the @code{jal} instruction. Calling
22337 functions using @code{jal} is more efficient but requires the caller
22338 and callee to be in the same 256 megabyte segment.
22340 This option has no effect on abicalls code. The default is
22341 @option{-mno-long-calls}.
22347 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
22348 instructions, as provided by the R4650 ISA@.
22354 Enable (disable) use of the @code{madd} and @code{msub} integer
22355 instructions. The default is @option{-mimadd} on architectures
22356 that support @code{madd} and @code{msub} except for the 74k
22357 architecture where it was found to generate slower code.
22360 @itemx -mno-fused-madd
22361 @opindex mfused-madd
22362 @opindex mno-fused-madd
22363 Enable (disable) use of the floating-point multiply-accumulate
22364 instructions, when they are available. The default is
22365 @option{-mfused-madd}.
22367 On the R8000 CPU when multiply-accumulate instructions are used,
22368 the intermediate product is calculated to infinite precision
22369 and is not subject to the FCSR Flush to Zero bit. This may be
22370 undesirable in some circumstances. On other processors the result
22371 is numerically identical to the equivalent computation using
22372 separate multiply, add, subtract and negate instructions.
22376 Tell the MIPS assembler to not run its preprocessor over user
22377 assembler files (with a @samp{.s} suffix) when assembling them.
22380 @itemx -mno-fix-24k
22382 @opindex mno-fix-24k
22383 Work around the 24K E48 (lost data on stores during refill) errata.
22384 The workarounds are implemented by the assembler rather than by GCC@.
22387 @itemx -mno-fix-r4000
22388 @opindex mfix-r4000
22389 @opindex mno-fix-r4000
22390 Work around certain R4000 CPU errata:
22393 A double-word or a variable shift may give an incorrect result if executed
22394 immediately after starting an integer division.
22396 A double-word or a variable shift may give an incorrect result if executed
22397 while an integer multiplication is in progress.
22399 An integer division may give an incorrect result if started in a delay slot
22400 of a taken branch or a jump.
22404 @itemx -mno-fix-r4400
22405 @opindex mfix-r4400
22406 @opindex mno-fix-r4400
22407 Work around certain R4400 CPU errata:
22410 A double-word or a variable shift may give an incorrect result if executed
22411 immediately after starting an integer division.
22415 @itemx -mno-fix-r10000
22416 @opindex mfix-r10000
22417 @opindex mno-fix-r10000
22418 Work around certain R10000 errata:
22421 @code{ll}/@code{sc} sequences may not behave atomically on revisions
22422 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
22425 This option can only be used if the target architecture supports
22426 branch-likely instructions. @option{-mfix-r10000} is the default when
22427 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
22431 @itemx -mno-fix-r5900
22432 @opindex mfix-r5900
22433 Do not attempt to schedule the preceding instruction into the delay slot
22434 of a branch instruction placed at the end of a short loop of six
22435 instructions or fewer and always schedule a @code{nop} instruction there
22436 instead. The short loop bug under certain conditions causes loops to
22437 execute only once or twice, due to a hardware bug in the R5900 chip. The
22438 workaround is implemented by the assembler rather than by GCC@.
22441 @itemx -mno-fix-rm7000
22442 @opindex mfix-rm7000
22443 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
22444 workarounds are implemented by the assembler rather than by GCC@.
22447 @itemx -mno-fix-vr4120
22448 @opindex mfix-vr4120
22449 Work around certain VR4120 errata:
22452 @code{dmultu} does not always produce the correct result.
22454 @code{div} and @code{ddiv} do not always produce the correct result if one
22455 of the operands is negative.
22457 The workarounds for the division errata rely on special functions in
22458 @file{libgcc.a}. At present, these functions are only provided by
22459 the @code{mips64vr*-elf} configurations.
22461 Other VR4120 errata require a NOP to be inserted between certain pairs of
22462 instructions. These errata are handled by the assembler, not by GCC itself.
22465 @opindex mfix-vr4130
22466 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
22467 workarounds are implemented by the assembler rather than by GCC,
22468 although GCC avoids using @code{mflo} and @code{mfhi} if the
22469 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
22470 instructions are available instead.
22473 @itemx -mno-fix-sb1
22475 Work around certain SB-1 CPU core errata.
22476 (This flag currently works around the SB-1 revision 2
22477 ``F1'' and ``F2'' floating-point errata.)
22479 @item -mr10k-cache-barrier=@var{setting}
22480 @opindex mr10k-cache-barrier
22481 Specify whether GCC should insert cache barriers to avoid the
22482 side effects of speculation on R10K processors.
22484 In common with many processors, the R10K tries to predict the outcome
22485 of a conditional branch and speculatively executes instructions from
22486 the ``taken'' branch. It later aborts these instructions if the
22487 predicted outcome is wrong. However, on the R10K, even aborted
22488 instructions can have side effects.
22490 This problem only affects kernel stores and, depending on the system,
22491 kernel loads. As an example, a speculatively-executed store may load
22492 the target memory into cache and mark the cache line as dirty, even if
22493 the store itself is later aborted. If a DMA operation writes to the
22494 same area of memory before the ``dirty'' line is flushed, the cached
22495 data overwrites the DMA-ed data. See the R10K processor manual
22496 for a full description, including other potential problems.
22498 One workaround is to insert cache barrier instructions before every memory
22499 access that might be speculatively executed and that might have side
22500 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
22501 controls GCC's implementation of this workaround. It assumes that
22502 aborted accesses to any byte in the following regions does not have
22507 the memory occupied by the current function's stack frame;
22510 the memory occupied by an incoming stack argument;
22513 the memory occupied by an object with a link-time-constant address.
22516 It is the kernel's responsibility to ensure that speculative
22517 accesses to these regions are indeed safe.
22519 If the input program contains a function declaration such as:
22525 then the implementation of @code{foo} must allow @code{j foo} and
22526 @code{jal foo} to be executed speculatively. GCC honors this
22527 restriction for functions it compiles itself. It expects non-GCC
22528 functions (such as hand-written assembly code) to do the same.
22530 The option has three forms:
22533 @item -mr10k-cache-barrier=load-store
22534 Insert a cache barrier before a load or store that might be
22535 speculatively executed and that might have side effects even
22538 @item -mr10k-cache-barrier=store
22539 Insert a cache barrier before a store that might be speculatively
22540 executed and that might have side effects even if aborted.
22542 @item -mr10k-cache-barrier=none
22543 Disable the insertion of cache barriers. This is the default setting.
22546 @item -mflush-func=@var{func}
22547 @itemx -mno-flush-func
22548 @opindex mflush-func
22549 Specifies the function to call to flush the I and D caches, or to not
22550 call any such function. If called, the function must take the same
22551 arguments as the common @code{_flush_func}, that is, the address of the
22552 memory range for which the cache is being flushed, the size of the
22553 memory range, and the number 3 (to flush both caches). The default
22554 depends on the target GCC was configured for, but commonly is either
22555 @code{_flush_func} or @code{__cpu_flush}.
22557 @item mbranch-cost=@var{num}
22558 @opindex mbranch-cost
22559 Set the cost of branches to roughly @var{num} ``simple'' instructions.
22560 This cost is only a heuristic and is not guaranteed to produce
22561 consistent results across releases. A zero cost redundantly selects
22562 the default, which is based on the @option{-mtune} setting.
22564 @item -mbranch-likely
22565 @itemx -mno-branch-likely
22566 @opindex mbranch-likely
22567 @opindex mno-branch-likely
22568 Enable or disable use of Branch Likely instructions, regardless of the
22569 default for the selected architecture. By default, Branch Likely
22570 instructions may be generated if they are supported by the selected
22571 architecture. An exception is for the MIPS32 and MIPS64 architectures
22572 and processors that implement those architectures; for those, Branch
22573 Likely instructions are not be generated by default because the MIPS32
22574 and MIPS64 architectures specifically deprecate their use.
22576 @item -mcompact-branches=never
22577 @itemx -mcompact-branches=optimal
22578 @itemx -mcompact-branches=always
22579 @opindex mcompact-branches=never
22580 @opindex mcompact-branches=optimal
22581 @opindex mcompact-branches=always
22582 These options control which form of branches will be generated. The
22583 default is @option{-mcompact-branches=optimal}.
22585 The @option{-mcompact-branches=never} option ensures that compact branch
22586 instructions will never be generated.
22588 The @option{-mcompact-branches=always} option ensures that a compact
22589 branch instruction will be generated if available. If a compact branch
22590 instruction is not available, a delay slot form of the branch will be
22593 This option is supported from MIPS Release 6 onwards.
22595 The @option{-mcompact-branches=optimal} option will cause a delay slot
22596 branch to be used if one is available in the current ISA and the delay
22597 slot is successfully filled. If the delay slot is not filled, a compact
22598 branch will be chosen if one is available.
22600 @item -mfp-exceptions
22601 @itemx -mno-fp-exceptions
22602 @opindex mfp-exceptions
22603 Specifies whether FP exceptions are enabled. This affects how
22604 FP instructions are scheduled for some processors.
22605 The default is that FP exceptions are
22608 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
22609 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
22612 @item -mvr4130-align
22613 @itemx -mno-vr4130-align
22614 @opindex mvr4130-align
22615 The VR4130 pipeline is two-way superscalar, but can only issue two
22616 instructions together if the first one is 8-byte aligned. When this
22617 option is enabled, GCC aligns pairs of instructions that it
22618 thinks should execute in parallel.
22620 This option only has an effect when optimizing for the VR4130.
22621 It normally makes code faster, but at the expense of making it bigger.
22622 It is enabled by default at optimization level @option{-O3}.
22627 Enable (disable) generation of @code{synci} instructions on
22628 architectures that support it. The @code{synci} instructions (if
22629 enabled) are generated when @code{__builtin___clear_cache} is
22632 This option defaults to @option{-mno-synci}, but the default can be
22633 overridden by configuring GCC with @option{--with-synci}.
22635 When compiling code for single processor systems, it is generally safe
22636 to use @code{synci}. However, on many multi-core (SMP) systems, it
22637 does not invalidate the instruction caches on all cores and may lead
22638 to undefined behavior.
22640 @item -mrelax-pic-calls
22641 @itemx -mno-relax-pic-calls
22642 @opindex mrelax-pic-calls
22643 Try to turn PIC calls that are normally dispatched via register
22644 @code{$25} into direct calls. This is only possible if the linker can
22645 resolve the destination at link time and if the destination is within
22646 range for a direct call.
22648 @option{-mrelax-pic-calls} is the default if GCC was configured to use
22649 an assembler and a linker that support the @code{.reloc} assembly
22650 directive and @option{-mexplicit-relocs} is in effect. With
22651 @option{-mno-explicit-relocs}, this optimization can be performed by the
22652 assembler and the linker alone without help from the compiler.
22654 @item -mmcount-ra-address
22655 @itemx -mno-mcount-ra-address
22656 @opindex mmcount-ra-address
22657 @opindex mno-mcount-ra-address
22658 Emit (do not emit) code that allows @code{_mcount} to modify the
22659 calling function's return address. When enabled, this option extends
22660 the usual @code{_mcount} interface with a new @var{ra-address}
22661 parameter, which has type @code{intptr_t *} and is passed in register
22662 @code{$12}. @code{_mcount} can then modify the return address by
22663 doing both of the following:
22666 Returning the new address in register @code{$31}.
22668 Storing the new address in @code{*@var{ra-address}},
22669 if @var{ra-address} is nonnull.
22672 The default is @option{-mno-mcount-ra-address}.
22674 @item -mframe-header-opt
22675 @itemx -mno-frame-header-opt
22676 @opindex mframe-header-opt
22677 Enable (disable) frame header optimization in the o32 ABI. When using the
22678 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
22679 function to write out register arguments. When enabled, this optimization
22680 will suppress the allocation of the frame header if it can be determined that
22683 This optimization is off by default at all optimization levels.
22686 @itemx -mno-lxc1-sxc1
22687 @opindex mlxc1-sxc1
22688 When applicable, enable (disable) the generation of @code{lwxc1},
22689 @code{swxc1}, @code{ldxc1}, @code{sdxc1} instructions. Enabled by default.
22694 When applicable, enable (disable) the generation of 4-operand @code{madd.s},
22695 @code{madd.d} and related instructions. Enabled by default.
22700 @subsection MMIX Options
22701 @cindex MMIX Options
22703 These options are defined for the MMIX:
22707 @itemx -mno-libfuncs
22709 @opindex mno-libfuncs
22710 Specify that intrinsic library functions are being compiled, passing all
22711 values in registers, no matter the size.
22714 @itemx -mno-epsilon
22716 @opindex mno-epsilon
22717 Generate floating-point comparison instructions that compare with respect
22718 to the @code{rE} epsilon register.
22720 @item -mabi=mmixware
22722 @opindex mabi=mmixware
22724 Generate code that passes function parameters and return values that (in
22725 the called function) are seen as registers @code{$0} and up, as opposed to
22726 the GNU ABI which uses global registers @code{$231} and up.
22728 @item -mzero-extend
22729 @itemx -mno-zero-extend
22730 @opindex mzero-extend
22731 @opindex mno-zero-extend
22732 When reading data from memory in sizes shorter than 64 bits, use (do not
22733 use) zero-extending load instructions by default, rather than
22734 sign-extending ones.
22737 @itemx -mno-knuthdiv
22739 @opindex mno-knuthdiv
22740 Make the result of a division yielding a remainder have the same sign as
22741 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
22742 remainder follows the sign of the dividend. Both methods are
22743 arithmetically valid, the latter being almost exclusively used.
22745 @item -mtoplevel-symbols
22746 @itemx -mno-toplevel-symbols
22747 @opindex mtoplevel-symbols
22748 @opindex mno-toplevel-symbols
22749 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
22750 code can be used with the @code{PREFIX} assembly directive.
22754 Generate an executable in the ELF format, rather than the default
22755 @samp{mmo} format used by the @command{mmix} simulator.
22757 @item -mbranch-predict
22758 @itemx -mno-branch-predict
22759 @opindex mbranch-predict
22760 @opindex mno-branch-predict
22761 Use (do not use) the probable-branch instructions, when static branch
22762 prediction indicates a probable branch.
22764 @item -mbase-addresses
22765 @itemx -mno-base-addresses
22766 @opindex mbase-addresses
22767 @opindex mno-base-addresses
22768 Generate (do not generate) code that uses @emph{base addresses}. Using a
22769 base address automatically generates a request (handled by the assembler
22770 and the linker) for a constant to be set up in a global register. The
22771 register is used for one or more base address requests within the range 0
22772 to 255 from the value held in the register. The generally leads to short
22773 and fast code, but the number of different data items that can be
22774 addressed is limited. This means that a program that uses lots of static
22775 data may require @option{-mno-base-addresses}.
22777 @item -msingle-exit
22778 @itemx -mno-single-exit
22779 @opindex msingle-exit
22780 @opindex mno-single-exit
22781 Force (do not force) generated code to have a single exit point in each
22785 @node MN10300 Options
22786 @subsection MN10300 Options
22787 @cindex MN10300 options
22789 These @option{-m} options are defined for Matsushita MN10300 architectures:
22794 Generate code to avoid bugs in the multiply instructions for the MN10300
22795 processors. This is the default.
22797 @item -mno-mult-bug
22798 @opindex mno-mult-bug
22799 Do not generate code to avoid bugs in the multiply instructions for the
22800 MN10300 processors.
22804 Generate code using features specific to the AM33 processor.
22808 Do not generate code using features specific to the AM33 processor. This
22813 Generate code using features specific to the AM33/2.0 processor.
22817 Generate code using features specific to the AM34 processor.
22819 @item -mtune=@var{cpu-type}
22821 Use the timing characteristics of the indicated CPU type when
22822 scheduling instructions. This does not change the targeted processor
22823 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
22824 @samp{am33-2} or @samp{am34}.
22826 @item -mreturn-pointer-on-d0
22827 @opindex mreturn-pointer-on-d0
22828 When generating a function that returns a pointer, return the pointer
22829 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
22830 only in @code{a0}, and attempts to call such functions without a prototype
22831 result in errors. Note that this option is on by default; use
22832 @option{-mno-return-pointer-on-d0} to disable it.
22836 Do not link in the C run-time initialization object file.
22840 Indicate to the linker that it should perform a relaxation optimization pass
22841 to shorten branches, calls and absolute memory addresses. This option only
22842 has an effect when used on the command line for the final link step.
22844 This option makes symbolic debugging impossible.
22848 Allow the compiler to generate @emph{Long Instruction Word}
22849 instructions if the target is the @samp{AM33} or later. This is the
22850 default. This option defines the preprocessor macro @code{__LIW__}.
22854 Do not allow the compiler to generate @emph{Long Instruction Word}
22855 instructions. This option defines the preprocessor macro
22860 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
22861 instructions if the target is the @samp{AM33} or later. This is the
22862 default. This option defines the preprocessor macro @code{__SETLB__}.
22866 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
22867 instructions. This option defines the preprocessor macro
22868 @code{__NO_SETLB__}.
22872 @node Moxie Options
22873 @subsection Moxie Options
22874 @cindex Moxie Options
22880 Generate big-endian code. This is the default for @samp{moxie-*-*}
22885 Generate little-endian code.
22889 Generate mul.x and umul.x instructions. This is the default for
22890 @samp{moxiebox-*-*} configurations.
22894 Do not link in the C run-time initialization object file.
22898 @node MSP430 Options
22899 @subsection MSP430 Options
22900 @cindex MSP430 Options
22902 These options are defined for the MSP430:
22908 Force assembly output to always use hex constants. Normally such
22909 constants are signed decimals, but this option is available for
22910 testsuite and/or aesthetic purposes.
22914 Select the MCU to target. This is used to create a C preprocessor
22915 symbol based upon the MCU name, converted to upper case and pre- and
22916 post-fixed with @samp{__}. This in turn is used by the
22917 @file{msp430.h} header file to select an MCU-specific supplementary
22920 The option also sets the ISA to use. If the MCU name is one that is
22921 known to only support the 430 ISA then that is selected, otherwise the
22922 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
22923 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
22924 name selects the 430X ISA.
22926 In addition an MCU-specific linker script is added to the linker
22927 command line. The script's name is the name of the MCU with
22928 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
22929 command line defines the C preprocessor symbol @code{__XXX__} and
22930 cause the linker to search for a script called @file{xxx.ld}.
22932 This option is also passed on to the assembler.
22935 @itemx -mno-warn-mcu
22937 @opindex mno-warn-mcu
22938 This option enables or disables warnings about conflicts between the
22939 MCU name specified by the @option{-mmcu} option and the ISA set by the
22940 @option{-mcpu} option and/or the hardware multiply support set by the
22941 @option{-mhwmult} option. It also toggles warnings about unrecognized
22942 MCU names. This option is on by default.
22946 Specifies the ISA to use. Accepted values are @samp{msp430},
22947 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
22948 @option{-mmcu=} option should be used to select the ISA.
22952 Link to the simulator runtime libraries and linker script. Overrides
22953 any scripts that would be selected by the @option{-mmcu=} option.
22957 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
22961 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
22965 This option is passed to the assembler and linker, and allows the
22966 linker to perform certain optimizations that cannot be done until
22971 Describes the type of hardware multiply supported by the target.
22972 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
22973 for the original 16-bit-only multiply supported by early MCUs.
22974 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
22975 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
22976 A value of @samp{auto} can also be given. This tells GCC to deduce
22977 the hardware multiply support based upon the MCU name provided by the
22978 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
22979 the MCU name is not recognized then no hardware multiply support is
22980 assumed. @code{auto} is the default setting.
22982 Hardware multiplies are normally performed by calling a library
22983 routine. This saves space in the generated code. When compiling at
22984 @option{-O3} or higher however the hardware multiplier is invoked
22985 inline. This makes for bigger, but faster code.
22987 The hardware multiply routines disable interrupts whilst running and
22988 restore the previous interrupt state when they finish. This makes
22989 them safe to use inside interrupt handlers as well as in normal code.
22993 Enable the use of a minimum runtime environment - no static
22994 initializers or constructors. This is intended for memory-constrained
22995 devices. The compiler includes special symbols in some objects
22996 that tell the linker and runtime which code fragments are required.
22998 @item -mcode-region=
22999 @itemx -mdata-region=
23000 @opindex mcode-region
23001 @opindex mdata-region
23002 These options tell the compiler where to place functions and data that
23003 do not have one of the @code{lower}, @code{upper}, @code{either} or
23004 @code{section} attributes. Possible values are @code{lower},
23005 @code{upper}, @code{either} or @code{any}. The first three behave
23006 like the corresponding attribute. The fourth possible value -
23007 @code{any} - is the default. It leaves placement entirely up to the
23008 linker script and how it assigns the standard sections
23009 (@code{.text}, @code{.data}, etc) to the memory regions.
23011 @item -msilicon-errata=
23012 @opindex msilicon-errata
23013 This option passes on a request to assembler to enable the fixes for
23014 the named silicon errata.
23016 @item -msilicon-errata-warn=
23017 @opindex msilicon-errata-warn
23018 This option passes on a request to the assembler to enable warning
23019 messages when a silicon errata might need to be applied.
23023 @node NDS32 Options
23024 @subsection NDS32 Options
23025 @cindex NDS32 Options
23027 These options are defined for NDS32 implementations:
23032 @opindex mbig-endian
23033 Generate code in big-endian mode.
23035 @item -mlittle-endian
23036 @opindex mlittle-endian
23037 Generate code in little-endian mode.
23039 @item -mreduced-regs
23040 @opindex mreduced-regs
23041 Use reduced-set registers for register allocation.
23044 @opindex mfull-regs
23045 Use full-set registers for register allocation.
23049 Generate conditional move instructions.
23053 Do not generate conditional move instructions.
23057 Generate performance extension instructions.
23059 @item -mno-ext-perf
23060 @opindex mno-ext-perf
23061 Do not generate performance extension instructions.
23064 @opindex mext-perf2
23065 Generate performance extension 2 instructions.
23067 @item -mno-ext-perf2
23068 @opindex mno-ext-perf2
23069 Do not generate performance extension 2 instructions.
23072 @opindex mext-string
23073 Generate string extension instructions.
23075 @item -mno-ext-string
23076 @opindex mno-ext-string
23077 Do not generate string extension instructions.
23081 Generate v3 push25/pop25 instructions.
23084 @opindex mno-v3push
23085 Do not generate v3 push25/pop25 instructions.
23089 Generate 16-bit instructions.
23092 @opindex mno-16-bit
23093 Do not generate 16-bit instructions.
23095 @item -misr-vector-size=@var{num}
23096 @opindex misr-vector-size
23097 Specify the size of each interrupt vector, which must be 4 or 16.
23099 @item -mcache-block-size=@var{num}
23100 @opindex mcache-block-size
23101 Specify the size of each cache block,
23102 which must be a power of 2 between 4 and 512.
23104 @item -march=@var{arch}
23106 Specify the name of the target architecture.
23108 @item -mcmodel=@var{code-model}
23110 Set the code model to one of
23113 All the data and read-only data segments must be within 512KB addressing space.
23114 The text segment must be within 16MB addressing space.
23115 @item @samp{medium}
23116 The data segment must be within 512KB while the read-only data segment can be
23117 within 4GB addressing space. The text segment should be still within 16MB
23120 All the text and data segments can be within 4GB addressing space.
23124 @opindex mctor-dtor
23125 Enable constructor/destructor feature.
23129 Guide linker to relax instructions.
23133 @node Nios II Options
23134 @subsection Nios II Options
23135 @cindex Nios II options
23136 @cindex Altera Nios II options
23138 These are the options defined for the Altera Nios II processor.
23144 @cindex smaller data references
23145 Put global and static objects less than or equal to @var{num} bytes
23146 into the small data or BSS sections instead of the normal data or BSS
23147 sections. The default value of @var{num} is 8.
23149 @item -mgpopt=@var{option}
23154 Generate (do not generate) GP-relative accesses. The following
23155 @var{option} names are recognized:
23160 Do not generate GP-relative accesses.
23163 Generate GP-relative accesses for small data objects that are not
23164 external, weak, or uninitialized common symbols.
23165 Also use GP-relative addressing for objects that
23166 have been explicitly placed in a small data section via a @code{section}
23170 As for @samp{local}, but also generate GP-relative accesses for
23171 small data objects that are external, weak, or common. If you use this option,
23172 you must ensure that all parts of your program (including libraries) are
23173 compiled with the same @option{-G} setting.
23176 Generate GP-relative accesses for all data objects in the program. If you
23177 use this option, the entire data and BSS segments
23178 of your program must fit in 64K of memory and you must use an appropriate
23179 linker script to allocate them within the addressable range of the
23183 Generate GP-relative addresses for function pointers as well as data
23184 pointers. If you use this option, the entire text, data, and BSS segments
23185 of your program must fit in 64K of memory and you must use an appropriate
23186 linker script to allocate them within the addressable range of the
23191 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
23192 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
23194 The default is @option{-mgpopt} except when @option{-fpic} or
23195 @option{-fPIC} is specified to generate position-independent code.
23196 Note that the Nios II ABI does not permit GP-relative accesses from
23199 You may need to specify @option{-mno-gpopt} explicitly when building
23200 programs that include large amounts of small data, including large
23201 GOT data sections. In this case, the 16-bit offset for GP-relative
23202 addressing may not be large enough to allow access to the entire
23203 small data section.
23205 @item -mgprel-sec=@var{regexp}
23206 @opindex mgprel-sec
23207 This option specifies additional section names that can be accessed via
23208 GP-relative addressing. It is most useful in conjunction with
23209 @code{section} attributes on variable declarations
23210 (@pxref{Common Variable Attributes}) and a custom linker script.
23211 The @var{regexp} is a POSIX Extended Regular Expression.
23213 This option does not affect the behavior of the @option{-G} option, and
23214 the specified sections are in addition to the standard @code{.sdata}
23215 and @code{.sbss} small-data sections that are recognized by @option{-mgpopt}.
23217 @item -mr0rel-sec=@var{regexp}
23218 @opindex mr0rel-sec
23219 This option specifies names of sections that can be accessed via a
23220 16-bit offset from @code{r0}; that is, in the low 32K or high 32K
23221 of the 32-bit address space. It is most useful in conjunction with
23222 @code{section} attributes on variable declarations
23223 (@pxref{Common Variable Attributes}) and a custom linker script.
23224 The @var{regexp} is a POSIX Extended Regular Expression.
23226 In contrast to the use of GP-relative addressing for small data,
23227 zero-based addressing is never generated by default and there are no
23228 conventional section names used in standard linker scripts for sections
23229 in the low or high areas of memory.
23235 Generate little-endian (default) or big-endian (experimental) code,
23238 @item -march=@var{arch}
23240 This specifies the name of the target Nios II architecture. GCC uses this
23241 name to determine what kind of instructions it can emit when generating
23242 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
23244 The preprocessor macro @code{__nios2_arch__} is available to programs,
23245 with value 1 or 2, indicating the targeted ISA level.
23247 @item -mbypass-cache
23248 @itemx -mno-bypass-cache
23249 @opindex mno-bypass-cache
23250 @opindex mbypass-cache
23251 Force all load and store instructions to always bypass cache by
23252 using I/O variants of the instructions. The default is not to
23255 @item -mno-cache-volatile
23256 @itemx -mcache-volatile
23257 @opindex mcache-volatile
23258 @opindex mno-cache-volatile
23259 Volatile memory access bypass the cache using the I/O variants of
23260 the load and store instructions. The default is not to bypass the cache.
23262 @item -mno-fast-sw-div
23263 @itemx -mfast-sw-div
23264 @opindex mno-fast-sw-div
23265 @opindex mfast-sw-div
23266 Do not use table-based fast divide for small numbers. The default
23267 is to use the fast divide at @option{-O3} and above.
23271 @itemx -mno-hw-mulx
23275 @opindex mno-hw-mul
23277 @opindex mno-hw-mulx
23279 @opindex mno-hw-div
23281 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
23282 instructions by the compiler. The default is to emit @code{mul}
23283 and not emit @code{div} and @code{mulx}.
23289 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
23290 CDX (code density) instructions. Enabling these instructions also
23291 requires @option{-march=r2}. Since these instructions are optional
23292 extensions to the R2 architecture, the default is not to emit them.
23294 @item -mcustom-@var{insn}=@var{N}
23295 @itemx -mno-custom-@var{insn}
23296 @opindex mcustom-@var{insn}
23297 @opindex mno-custom-@var{insn}
23298 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
23299 custom instruction with encoding @var{N} when generating code that uses
23300 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
23301 instruction 253 for single-precision floating-point add operations instead
23302 of the default behavior of using a library call.
23304 The following values of @var{insn} are supported. Except as otherwise
23305 noted, floating-point operations are expected to be implemented with
23306 normal IEEE 754 semantics and correspond directly to the C operators or the
23307 equivalent GCC built-in functions (@pxref{Other Builtins}).
23309 Single-precision floating point:
23312 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
23313 Binary arithmetic operations.
23319 Unary absolute value.
23321 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
23322 Comparison operations.
23324 @item @samp{fmins}, @samp{fmaxs}
23325 Floating-point minimum and maximum. These instructions are only
23326 generated if @option{-ffinite-math-only} is specified.
23328 @item @samp{fsqrts}
23329 Unary square root operation.
23331 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
23332 Floating-point trigonometric and exponential functions. These instructions
23333 are only generated if @option{-funsafe-math-optimizations} is also specified.
23337 Double-precision floating point:
23340 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
23341 Binary arithmetic operations.
23347 Unary absolute value.
23349 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
23350 Comparison operations.
23352 @item @samp{fmind}, @samp{fmaxd}
23353 Double-precision minimum and maximum. These instructions are only
23354 generated if @option{-ffinite-math-only} is specified.
23356 @item @samp{fsqrtd}
23357 Unary square root operation.
23359 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
23360 Double-precision trigonometric and exponential functions. These instructions
23361 are only generated if @option{-funsafe-math-optimizations} is also specified.
23367 @item @samp{fextsd}
23368 Conversion from single precision to double precision.
23370 @item @samp{ftruncds}
23371 Conversion from double precision to single precision.
23373 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
23374 Conversion from floating point to signed or unsigned integer types, with
23375 truncation towards zero.
23378 Conversion from single-precision floating point to signed integer,
23379 rounding to the nearest integer and ties away from zero.
23380 This corresponds to the @code{__builtin_lroundf} function when
23381 @option{-fno-math-errno} is used.
23383 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
23384 Conversion from signed or unsigned integer types to floating-point types.
23388 In addition, all of the following transfer instructions for internal
23389 registers X and Y must be provided to use any of the double-precision
23390 floating-point instructions. Custom instructions taking two
23391 double-precision source operands expect the first operand in the
23392 64-bit register X. The other operand (or only operand of a unary
23393 operation) is given to the custom arithmetic instruction with the
23394 least significant half in source register @var{src1} and the most
23395 significant half in @var{src2}. A custom instruction that returns a
23396 double-precision result returns the most significant 32 bits in the
23397 destination register and the other half in 32-bit register Y.
23398 GCC automatically generates the necessary code sequences to write
23399 register X and/or read register Y when double-precision floating-point
23400 instructions are used.
23405 Write @var{src1} into the least significant half of X and @var{src2} into
23406 the most significant half of X.
23409 Write @var{src1} into Y.
23411 @item @samp{frdxhi}, @samp{frdxlo}
23412 Read the most or least (respectively) significant half of X and store it in
23416 Read the value of Y and store it into @var{dest}.
23419 Note that you can gain more local control over generation of Nios II custom
23420 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
23421 and @code{target("no-custom-@var{insn}")} function attributes
23422 (@pxref{Function Attributes})
23423 or pragmas (@pxref{Function Specific Option Pragmas}).
23425 @item -mcustom-fpu-cfg=@var{name}
23426 @opindex mcustom-fpu-cfg
23428 This option enables a predefined, named set of custom instruction encodings
23429 (see @option{-mcustom-@var{insn}} above).
23430 Currently, the following sets are defined:
23432 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
23433 @gccoptlist{-mcustom-fmuls=252 @gol
23434 -mcustom-fadds=253 @gol
23435 -mcustom-fsubs=254 @gol
23436 -fsingle-precision-constant}
23438 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
23439 @gccoptlist{-mcustom-fmuls=252 @gol
23440 -mcustom-fadds=253 @gol
23441 -mcustom-fsubs=254 @gol
23442 -mcustom-fdivs=255 @gol
23443 -fsingle-precision-constant}
23445 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
23446 @gccoptlist{-mcustom-floatus=243 @gol
23447 -mcustom-fixsi=244 @gol
23448 -mcustom-floatis=245 @gol
23449 -mcustom-fcmpgts=246 @gol
23450 -mcustom-fcmples=249 @gol
23451 -mcustom-fcmpeqs=250 @gol
23452 -mcustom-fcmpnes=251 @gol
23453 -mcustom-fmuls=252 @gol
23454 -mcustom-fadds=253 @gol
23455 -mcustom-fsubs=254 @gol
23456 -mcustom-fdivs=255 @gol
23457 -fsingle-precision-constant}
23459 Custom instruction assignments given by individual
23460 @option{-mcustom-@var{insn}=} options override those given by
23461 @option{-mcustom-fpu-cfg=}, regardless of the
23462 order of the options on the command line.
23464 Note that you can gain more local control over selection of a FPU
23465 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
23466 function attribute (@pxref{Function Attributes})
23467 or pragma (@pxref{Function Specific Option Pragmas}).
23471 These additional @samp{-m} options are available for the Altera Nios II
23472 ELF (bare-metal) target:
23478 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
23479 startup and termination code, and is typically used in conjunction with
23480 @option{-msys-crt0=} to specify the location of the alternate startup code
23481 provided by the HAL BSP.
23485 Link with a limited version of the C library, @option{-lsmallc}, rather than
23488 @item -msys-crt0=@var{startfile}
23490 @var{startfile} is the file name of the startfile (crt0) to use
23491 when linking. This option is only useful in conjunction with @option{-mhal}.
23493 @item -msys-lib=@var{systemlib}
23495 @var{systemlib} is the library name of the library that provides
23496 low-level system calls required by the C library,
23497 e.g.@: @code{read} and @code{write}.
23498 This option is typically used to link with a library provided by a HAL BSP.
23502 @node Nvidia PTX Options
23503 @subsection Nvidia PTX Options
23504 @cindex Nvidia PTX options
23505 @cindex nvptx options
23507 These options are defined for Nvidia PTX:
23515 Generate code for 32-bit or 64-bit ABI.
23517 @item -misa=@var{ISA-string}
23519 Generate code for given the specified PTX ISA (e.g.@: @samp{sm_35}). ISA
23520 strings must be lower-case. Valid ISA strings include @samp{sm_30} and
23521 @samp{sm_35}. The default ISA is sm_30.
23524 @opindex mmainkernel
23525 Link in code for a __main kernel. This is for stand-alone instead of
23526 offloading execution.
23530 Apply partitioned execution optimizations. This is the default when any
23531 level of optimization is selected.
23534 @opindex msoft-stack
23535 Generate code that does not use @code{.local} memory
23536 directly for stack storage. Instead, a per-warp stack pointer is
23537 maintained explicitly. This enables variable-length stack allocation (with
23538 variable-length arrays or @code{alloca}), and when global memory is used for
23539 underlying storage, makes it possible to access automatic variables from other
23540 threads, or with atomic instructions. This code generation variant is used
23541 for OpenMP offloading, but the option is exposed on its own for the purpose
23542 of testing the compiler; to generate code suitable for linking into programs
23543 using OpenMP offloading, use option @option{-mgomp}.
23545 @item -muniform-simt
23546 @opindex muniform-simt
23547 Switch to code generation variant that allows to execute all threads in each
23548 warp, while maintaining memory state and side effects as if only one thread
23549 in each warp was active outside of OpenMP SIMD regions. All atomic operations
23550 and calls to runtime (malloc, free, vprintf) are conditionally executed (iff
23551 current lane index equals the master lane index), and the register being
23552 assigned is copied via a shuffle instruction from the master lane. Outside of
23553 SIMD regions lane 0 is the master; inside, each thread sees itself as the
23554 master. Shared memory array @code{int __nvptx_uni[]} stores all-zeros or
23555 all-ones bitmasks for each warp, indicating current mode (0 outside of SIMD
23556 regions). Each thread can bitwise-and the bitmask at position @code{tid.y}
23557 with current lane index to compute the master lane index.
23561 Generate code for use in OpenMP offloading: enables @option{-msoft-stack} and
23562 @option{-muniform-simt} options, and selects corresponding multilib variant.
23566 @node OpenRISC Options
23567 @subsection OpenRISC Options
23568 @cindex OpenRISC Options
23570 These options are defined for OpenRISC:
23574 @item -mboard=@var{name}
23576 Configure a board specific runtime. This will be passed to the linker for
23577 newlib board library linking. The default is @code{or1ksim}.
23581 For compatibility, it's always newlib for elf now.
23585 Generate code for hardware which supports divide instructions. This is the
23590 Generate code for hardware which supports multiply instructions. This is the
23595 Generate code for hardware which supports the conditional move (@code{l.cmov})
23600 Generate code for hardware which supports rotate right instructions.
23604 Generate code for hardware which supports sign-extension instructions.
23608 Generate code for hardware which supports set flag immediate (@code{l.sf*i})
23613 Generate code for hardware which supports shift immediate related instructions
23614 (i.e. @code{l.srai}, @code{l.srli}, @code{l.slli}, @code{1.rori}). Note, to
23615 enable generation of the @code{l.rori} instruction the @option{-mror} flag must
23620 Generate code for hardware which requires divide instruction emulation.
23624 Generate code for hardware which requires multiply instruction emulation.
23628 @node PDP-11 Options
23629 @subsection PDP-11 Options
23630 @cindex PDP-11 Options
23632 These options are defined for the PDP-11:
23637 Use hardware FPP floating point. This is the default. (FIS floating
23638 point on the PDP-11/40 is not supported.) Implies -m45.
23641 @opindex msoft-float
23642 Do not use hardware floating point.
23646 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
23650 Return floating-point results in memory. This is the default.
23654 Generate code for a PDP-11/40. Implies -msoft-float -mno-split.
23658 Generate code for a PDP-11/45. This is the default.
23662 Generate code for a PDP-11/10. Implies -msoft-float -mno-split.
23668 Use 16-bit @code{int}. This is the default.
23674 Use 32-bit @code{int}.
23678 Target has split instruction and data space. Implies -m45.
23682 Use Unix assembler syntax.
23686 Use DEC assembler syntax.
23690 Use GNU assembler syntax. This is the default.
23694 Use the new LRA register allocator. By default, the old ``reload''
23698 @node picoChip Options
23699 @subsection picoChip Options
23700 @cindex picoChip options
23702 These @samp{-m} options are defined for picoChip implementations:
23706 @item -mae=@var{ae_type}
23708 Set the instruction set, register set, and instruction scheduling
23709 parameters for array element type @var{ae_type}. Supported values
23710 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
23712 @option{-mae=ANY} selects a completely generic AE type. Code
23713 generated with this option runs on any of the other AE types. The
23714 code is not as efficient as it would be if compiled for a specific
23715 AE type, and some types of operation (e.g., multiplication) do not
23716 work properly on all types of AE.
23718 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
23719 for compiled code, and is the default.
23721 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
23722 option may suffer from poor performance of byte (char) manipulation,
23723 since the DSP AE does not provide hardware support for byte load/stores.
23725 @item -msymbol-as-address
23726 Enable the compiler to directly use a symbol name as an address in a
23727 load/store instruction, without first loading it into a
23728 register. Typically, the use of this option generates larger
23729 programs, which run faster than when the option isn't used. However, the
23730 results vary from program to program, so it is left as a user option,
23731 rather than being permanently enabled.
23733 @item -mno-inefficient-warnings
23734 Disables warnings about the generation of inefficient code. These
23735 warnings can be generated, for example, when compiling code that
23736 performs byte-level memory operations on the MAC AE type. The MAC AE has
23737 no hardware support for byte-level memory operations, so all byte
23738 load/stores must be synthesized from word load/store operations. This is
23739 inefficient and a warning is generated to indicate
23740 that you should rewrite the code to avoid byte operations, or to target
23741 an AE type that has the necessary hardware support. This option disables
23746 @node PowerPC Options
23747 @subsection PowerPC Options
23748 @cindex PowerPC options
23750 These are listed under @xref{RS/6000 and PowerPC Options}.
23752 @node RISC-V Options
23753 @subsection RISC-V Options
23754 @cindex RISC-V Options
23756 These command-line options are defined for RISC-V targets:
23759 @item -mbranch-cost=@var{n}
23760 @opindex mbranch-cost
23761 Set the cost of branches to roughly @var{n} instructions.
23766 When generating PIC code, do or don't allow the use of PLTs. Ignored for
23767 non-PIC. The default is @option{-mplt}.
23769 @item -mabi=@var{ABI-string}
23771 Specify integer and floating-point calling convention. @var{ABI-string}
23772 contains two parts: the size of integer types and the registers used for
23773 floating-point types. For example @samp{-march=rv64ifd -mabi=lp64d} means that
23774 @samp{long} and pointers are 64-bit (implicitly defining @samp{int} to be
23775 32-bit), and that floating-point values up to 64 bits wide are passed in F
23776 registers. Contrast this with @samp{-march=rv64ifd -mabi=lp64f}, which still
23777 allows the compiler to generate code that uses the F and D extensions but only
23778 allows floating-point values up to 32 bits long to be passed in registers; or
23779 @samp{-march=rv64ifd -mabi=lp64}, in which no floating-point arguments will be
23780 passed in registers.
23782 The default for this argument is system dependent, users who want a specific
23783 calling convention should specify one explicitly. The valid calling
23784 conventions are: @samp{ilp32}, @samp{ilp32f}, @samp{ilp32d}, @samp{lp64},
23785 @samp{lp64f}, and @samp{lp64d}. Some calling conventions are impossible to
23786 implement on some ISAs: for example, @samp{-march=rv32if -mabi=ilp32d} is
23787 invalid because the ABI requires 64-bit values be passed in F registers, but F
23788 registers are only 32 bits wide. There is also the @samp{ilp32e} ABI that can
23789 only be used with the @samp{rv32e} architecture. This ABI is not well
23790 specified at present, and is subject to change.
23795 Do or don't use hardware floating-point divide and square root instructions.
23796 This requires the F or D extensions for floating-point registers. The default
23797 is to use them if the specified architecture has these instructions.
23802 Do or don't use hardware instructions for integer division. This requires the
23803 M extension. The default is to use them if the specified architecture has
23804 these instructions.
23806 @item -march=@var{ISA-string}
23808 Generate code for given RISC-V ISA (e.g.@: @samp{rv64im}). ISA strings must be
23809 lower-case. Examples include @samp{rv64i}, @samp{rv32g}, @samp{rv32e}, and
23812 @item -mtune=@var{processor-string}
23814 Optimize the output for the given processor, specified by microarchitecture
23815 name. Permissible values for this option are: @samp{rocket},
23816 @samp{sifive-3-series}, @samp{sifive-5-series}, @samp{sifive-7-series},
23819 When @option{-mtune=} is not specified, the default is @samp{rocket}.
23821 The @samp{size} choice is not intended for use by end-users. This is used
23822 when @option{-Os} is specified. It overrides the instruction cost info
23823 provided by @option{-mtune=}, but does not override the pipeline info. This
23824 helps reduce code size while still giving good performance.
23826 @item -mpreferred-stack-boundary=@var{num}
23827 @opindex mpreferred-stack-boundary
23828 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
23829 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
23830 the default is 4 (16 bytes or 128-bits).
23832 @strong{Warning:} If you use this switch, then you must build all modules with
23833 the same value, including any libraries. This includes the system libraries
23834 and startup modules.
23836 @item -msmall-data-limit=@var{n}
23837 @opindex msmall-data-limit
23838 Put global and static data smaller than @var{n} bytes into a special section
23841 @item -msave-restore
23842 @itemx -mno-save-restore
23843 @opindex msave-restore
23844 Do or don't use smaller but slower prologue and epilogue code that uses
23845 library function calls. The default is to use fast inline prologues and
23848 @item -mstrict-align
23849 @itemx -mno-strict-align
23850 @opindex mstrict-align
23851 Do not or do generate unaligned memory accesses. The default is set depending
23852 on whether the processor we are optimizing for supports fast unaligned access
23855 @item -mcmodel=medlow
23856 @opindex mcmodel=medlow
23857 Generate code for the medium-low code model. The program and its statically
23858 defined symbols must lie within a single 2 GiB address range and must lie
23859 between absolute addresses @minus{}2 GiB and +2 GiB. Programs can be
23860 statically or dynamically linked. This is the default code model.
23862 @item -mcmodel=medany
23863 @opindex mcmodel=medany
23864 Generate code for the medium-any code model. The program and its statically
23865 defined symbols must be within any single 2 GiB address range. Programs can be
23866 statically or dynamically linked.
23868 @item -mexplicit-relocs
23869 @itemx -mno-exlicit-relocs
23870 Use or do not use assembler relocation operators when dealing with symbolic
23871 addresses. The alternative is to use assembler macros instead, which may
23872 limit optimization.
23876 Take advantage of linker relaxations to reduce the number of instructions
23877 required to materialize symbol addresses. The default is to take advantage of
23878 linker relaxations.
23880 @item -memit-attribute
23881 @itemx -mno-emit-attribute
23882 Emit (do not emit) RISC-V attribute to record extra information into ELF
23883 objects. This feature requires at least binutils 2.32.
23887 @subsection RL78 Options
23888 @cindex RL78 Options
23894 Links in additional target libraries to support operation within a
23903 Specifies the type of hardware multiplication and division support to
23904 be used. The simplest is @code{none}, which uses software for both
23905 multiplication and division. This is the default. The @code{g13}
23906 value is for the hardware multiply/divide peripheral found on the
23907 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
23908 the multiplication and division instructions supported by the RL78/G14
23909 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
23910 the value @code{mg10} is an alias for @code{none}.
23912 In addition a C preprocessor macro is defined, based upon the setting
23913 of this option. Possible values are: @code{__RL78_MUL_NONE__},
23914 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
23921 Specifies the RL78 core to target. The default is the G14 core, also
23922 known as an S3 core or just RL78. The G13 or S2 core does not have
23923 multiply or divide instructions, instead it uses a hardware peripheral
23924 for these operations. The G10 or S1 core does not have register
23925 banks, so it uses a different calling convention.
23927 If this option is set it also selects the type of hardware multiply
23928 support to use, unless this is overridden by an explicit
23929 @option{-mmul=none} option on the command line. Thus specifying
23930 @option{-mcpu=g13} enables the use of the G13 hardware multiply
23931 peripheral and specifying @option{-mcpu=g10} disables the use of
23932 hardware multiplications altogether.
23934 Note, although the RL78/G14 core is the default target, specifying
23935 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
23936 change the behavior of the toolchain since it also enables G14
23937 hardware multiply support. If these options are not specified on the
23938 command line then software multiplication routines will be used even
23939 though the code targets the RL78 core. This is for backwards
23940 compatibility with older toolchains which did not have hardware
23941 multiply and divide support.
23943 In addition a C preprocessor macro is defined, based upon the setting
23944 of this option. Possible values are: @code{__RL78_G10__},
23945 @code{__RL78_G13__} or @code{__RL78_G14__}.
23955 These are aliases for the corresponding @option{-mcpu=} option. They
23956 are provided for backwards compatibility.
23960 Allow the compiler to use all of the available registers. By default
23961 registers @code{r24..r31} are reserved for use in interrupt handlers.
23962 With this option enabled these registers can be used in ordinary
23965 @item -m64bit-doubles
23966 @itemx -m32bit-doubles
23967 @opindex m64bit-doubles
23968 @opindex m32bit-doubles
23969 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
23970 or 32 bits (@option{-m32bit-doubles}) in size. The default is
23971 @option{-m32bit-doubles}.
23973 @item -msave-mduc-in-interrupts
23974 @itemx -mno-save-mduc-in-interrupts
23975 @opindex msave-mduc-in-interrupts
23976 @opindex mno-save-mduc-in-interrupts
23977 Specifies that interrupt handler functions should preserve the
23978 MDUC registers. This is only necessary if normal code might use
23979 the MDUC registers, for example because it performs multiplication
23980 and division operations. The default is to ignore the MDUC registers
23981 as this makes the interrupt handlers faster. The target option -mg13
23982 needs to be passed for this to work as this feature is only available
23983 on the G13 target (S2 core). The MDUC registers will only be saved
23984 if the interrupt handler performs a multiplication or division
23985 operation or it calls another function.
23989 @node RS/6000 and PowerPC Options
23990 @subsection IBM RS/6000 and PowerPC Options
23991 @cindex RS/6000 and PowerPC Options
23992 @cindex IBM RS/6000 and PowerPC Options
23994 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
23996 @item -mpowerpc-gpopt
23997 @itemx -mno-powerpc-gpopt
23998 @itemx -mpowerpc-gfxopt
23999 @itemx -mno-powerpc-gfxopt
24002 @itemx -mno-powerpc64
24006 @itemx -mno-popcntb
24008 @itemx -mno-popcntd
24015 @itemx -mno-hard-dfp
24016 @opindex mpowerpc-gpopt
24017 @opindex mno-powerpc-gpopt
24018 @opindex mpowerpc-gfxopt
24019 @opindex mno-powerpc-gfxopt
24020 @opindex mpowerpc64
24021 @opindex mno-powerpc64
24025 @opindex mno-popcntb
24027 @opindex mno-popcntd
24033 @opindex mno-hard-dfp
24034 You use these options to specify which instructions are available on the
24035 processor you are using. The default value of these options is
24036 determined when configuring GCC@. Specifying the
24037 @option{-mcpu=@var{cpu_type}} overrides the specification of these
24038 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
24039 rather than the options listed above.
24041 Specifying @option{-mpowerpc-gpopt} allows
24042 GCC to use the optional PowerPC architecture instructions in the
24043 General Purpose group, including floating-point square root. Specifying
24044 @option{-mpowerpc-gfxopt} allows GCC to
24045 use the optional PowerPC architecture instructions in the Graphics
24046 group, including floating-point select.
24048 The @option{-mmfcrf} option allows GCC to generate the move from
24049 condition register field instruction implemented on the POWER4
24050 processor and other processors that support the PowerPC V2.01
24052 The @option{-mpopcntb} option allows GCC to generate the popcount and
24053 double-precision FP reciprocal estimate instruction implemented on the
24054 POWER5 processor and other processors that support the PowerPC V2.02
24056 The @option{-mpopcntd} option allows GCC to generate the popcount
24057 instruction implemented on the POWER7 processor and other processors
24058 that support the PowerPC V2.06 architecture.
24059 The @option{-mfprnd} option allows GCC to generate the FP round to
24060 integer instructions implemented on the POWER5+ processor and other
24061 processors that support the PowerPC V2.03 architecture.
24062 The @option{-mcmpb} option allows GCC to generate the compare bytes
24063 instruction implemented on the POWER6 processor and other processors
24064 that support the PowerPC V2.05 architecture.
24065 The @option{-mhard-dfp} option allows GCC to generate the decimal
24066 floating-point instructions implemented on some POWER processors.
24068 The @option{-mpowerpc64} option allows GCC to generate the additional
24069 64-bit instructions that are found in the full PowerPC64 architecture
24070 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
24071 @option{-mno-powerpc64}.
24073 @item -mcpu=@var{cpu_type}
24075 Set architecture type, register usage, and
24076 instruction scheduling parameters for machine type @var{cpu_type}.
24077 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
24078 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
24079 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
24080 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
24081 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
24082 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
24083 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
24084 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
24085 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
24086 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
24087 @samp{power9}, @samp{future}, @samp{powerpc}, @samp{powerpc64},
24088 @samp{powerpc64le}, @samp{rs64}, and @samp{native}.
24090 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
24091 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
24092 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
24093 architecture machine types, with an appropriate, generic processor
24094 model assumed for scheduling purposes.
24096 Specifying @samp{native} as cpu type detects and selects the
24097 architecture option that corresponds to the host processor of the
24098 system performing the compilation.
24099 @option{-mcpu=native} has no effect if GCC does not recognize the
24102 The other options specify a specific processor. Code generated under
24103 those options runs best on that processor, and may not run at all on
24106 The @option{-mcpu} options automatically enable or disable the
24109 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
24110 -mpopcntb -mpopcntd -mpowerpc64 @gol
24111 -mpowerpc-gpopt -mpowerpc-gfxopt @gol
24112 -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
24113 -mcrypto -mhtm -mpower8-fusion -mpower8-vector @gol
24114 -mquad-memory -mquad-memory-atomic -mfloat128 -mfloat128-hardware}
24116 The particular options set for any particular CPU varies between
24117 compiler versions, depending on what setting seems to produce optimal
24118 code for that CPU; it doesn't necessarily reflect the actual hardware's
24119 capabilities. If you wish to set an individual option to a particular
24120 value, you may specify it after the @option{-mcpu} option, like
24121 @option{-mcpu=970 -mno-altivec}.
24123 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
24124 not enabled or disabled by the @option{-mcpu} option at present because
24125 AIX does not have full support for these options. You may still
24126 enable or disable them individually if you're sure it'll work in your
24129 @item -mtune=@var{cpu_type}
24131 Set the instruction scheduling parameters for machine type
24132 @var{cpu_type}, but do not set the architecture type or register usage,
24133 as @option{-mcpu=@var{cpu_type}} does. The same
24134 values for @var{cpu_type} are used for @option{-mtune} as for
24135 @option{-mcpu}. If both are specified, the code generated uses the
24136 architecture and registers set by @option{-mcpu}, but the
24137 scheduling parameters set by @option{-mtune}.
24139 @item -mcmodel=small
24140 @opindex mcmodel=small
24141 Generate PowerPC64 code for the small model: The TOC is limited to
24144 @item -mcmodel=medium
24145 @opindex mcmodel=medium
24146 Generate PowerPC64 code for the medium model: The TOC and other static
24147 data may be up to a total of 4G in size. This is the default for 64-bit
24150 @item -mcmodel=large
24151 @opindex mcmodel=large
24152 Generate PowerPC64 code for the large model: The TOC may be up to 4G
24153 in size. Other data and code is only limited by the 64-bit address
24157 @itemx -mno-altivec
24159 @opindex mno-altivec
24160 Generate code that uses (does not use) AltiVec instructions, and also
24161 enable the use of built-in functions that allow more direct access to
24162 the AltiVec instruction set. You may also need to set
24163 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
24166 When @option{-maltivec} is used, the element order for AltiVec intrinsics
24167 such as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
24168 match array element order corresponding to the endianness of the
24169 target. That is, element zero identifies the leftmost element in a
24170 vector register when targeting a big-endian platform, and identifies
24171 the rightmost element in a vector register when targeting a
24172 little-endian platform.
24177 @opindex mno-vrsave
24178 Generate VRSAVE instructions when generating AltiVec code.
24181 @opindex msecure-plt
24182 Generate code that allows @command{ld} and @command{ld.so}
24183 to build executables and shared
24184 libraries with non-executable @code{.plt} and @code{.got} sections.
24186 32-bit SYSV ABI option.
24190 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
24192 requires @code{.plt} and @code{.got}
24193 sections that are both writable and executable.
24194 This is a PowerPC 32-bit SYSV ABI option.
24200 This switch enables or disables the generation of ISEL instructions.
24206 Generate code that uses (does not use) vector/scalar (VSX)
24207 instructions, and also enable the use of built-in functions that allow
24208 more direct access to the VSX instruction set.
24213 @opindex mno-crypto
24214 Enable the use (disable) of the built-in functions that allow direct
24215 access to the cryptographic instructions that were added in version
24216 2.07 of the PowerPC ISA.
24222 Enable (disable) the use of the built-in functions that allow direct
24223 access to the Hardware Transactional Memory (HTM) instructions that
24224 were added in version 2.07 of the PowerPC ISA.
24226 @item -mpower8-fusion
24227 @itemx -mno-power8-fusion
24228 @opindex mpower8-fusion
24229 @opindex mno-power8-fusion
24230 Generate code that keeps (does not keeps) some integer operations
24231 adjacent so that the instructions can be fused together on power8 and
24234 @item -mpower8-vector
24235 @itemx -mno-power8-vector
24236 @opindex mpower8-vector
24237 @opindex mno-power8-vector
24238 Generate code that uses (does not use) the vector and scalar
24239 instructions that were added in version 2.07 of the PowerPC ISA. Also
24240 enable the use of built-in functions that allow more direct access to
24241 the vector instructions.
24243 @item -mquad-memory
24244 @itemx -mno-quad-memory
24245 @opindex mquad-memory
24246 @opindex mno-quad-memory
24247 Generate code that uses (does not use) the non-atomic quad word memory
24248 instructions. The @option{-mquad-memory} option requires use of
24251 @item -mquad-memory-atomic
24252 @itemx -mno-quad-memory-atomic
24253 @opindex mquad-memory-atomic
24254 @opindex mno-quad-memory-atomic
24255 Generate code that uses (does not use) the atomic quad word memory
24256 instructions. The @option{-mquad-memory-atomic} option requires use of
24260 @itemx -mno-float128
24262 @opindex mno-float128
24263 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
24264 and use either software emulation for IEEE 128-bit floating point or
24265 hardware instructions.
24267 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7},
24268 @option{-mcpu=power8}), or @option{-mcpu=power9} must be enabled to
24269 use the IEEE 128-bit floating point support. The IEEE 128-bit
24270 floating point support only works on PowerPC Linux systems.
24272 The default for @option{-mfloat128} is enabled on PowerPC Linux
24273 systems using the VSX instruction set, and disabled on other systems.
24275 If you use the ISA 3.0 instruction set (@option{-mpower9-vector} or
24276 @option{-mcpu=power9}) on a 64-bit system, the IEEE 128-bit floating
24277 point support will also enable the generation of ISA 3.0 IEEE 128-bit
24278 floating point instructions. Otherwise, if you do not specify to
24279 generate ISA 3.0 instructions or you are targeting a 32-bit big endian
24280 system, IEEE 128-bit floating point will be done with software
24283 @item -mfloat128-hardware
24284 @itemx -mno-float128-hardware
24285 @opindex mfloat128-hardware
24286 @opindex mno-float128-hardware
24287 Enable/disable using ISA 3.0 hardware instructions to support the
24288 @var{__float128} data type.
24290 The default for @option{-mfloat128-hardware} is enabled on PowerPC
24291 Linux systems using the ISA 3.0 instruction set, and disabled on other
24298 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
24299 targets (including GNU/Linux). The 32-bit environment sets int, long
24300 and pointer to 32 bits and generates code that runs on any PowerPC
24301 variant. The 64-bit environment sets int to 32 bits and long and
24302 pointer to 64 bits, and generates code for PowerPC64, as for
24303 @option{-mpowerpc64}.
24306 @itemx -mno-fp-in-toc
24307 @itemx -mno-sum-in-toc
24308 @itemx -mminimal-toc
24310 @opindex mno-fp-in-toc
24311 @opindex mno-sum-in-toc
24312 @opindex mminimal-toc
24313 Modify generation of the TOC (Table Of Contents), which is created for
24314 every executable file. The @option{-mfull-toc} option is selected by
24315 default. In that case, GCC allocates at least one TOC entry for
24316 each unique non-automatic variable reference in your program. GCC
24317 also places floating-point constants in the TOC@. However, only
24318 16,384 entries are available in the TOC@.
24320 If you receive a linker error message that saying you have overflowed
24321 the available TOC space, you can reduce the amount of TOC space used
24322 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
24323 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
24324 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
24325 generate code to calculate the sum of an address and a constant at
24326 run time instead of putting that sum into the TOC@. You may specify one
24327 or both of these options. Each causes GCC to produce very slightly
24328 slower and larger code at the expense of conserving TOC space.
24330 If you still run out of space in the TOC even when you specify both of
24331 these options, specify @option{-mminimal-toc} instead. This option causes
24332 GCC to make only one TOC entry for every file. When you specify this
24333 option, GCC produces code that is slower and larger but which
24334 uses extremely little TOC space. You may wish to use this option
24335 only on files that contain less frequently-executed code.
24341 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
24342 @code{long} type, and the infrastructure needed to support them.
24343 Specifying @option{-maix64} implies @option{-mpowerpc64},
24344 while @option{-maix32} disables the 64-bit ABI and
24345 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
24348 @itemx -mno-xl-compat
24349 @opindex mxl-compat
24350 @opindex mno-xl-compat
24351 Produce code that conforms more closely to IBM XL compiler semantics
24352 when using AIX-compatible ABI@. Pass floating-point arguments to
24353 prototyped functions beyond the register save area (RSA) on the stack
24354 in addition to argument FPRs. Do not assume that most significant
24355 double in 128-bit long double value is properly rounded when comparing
24356 values and converting to double. Use XL symbol names for long double
24359 The AIX calling convention was extended but not initially documented to
24360 handle an obscure K&R C case of calling a function that takes the
24361 address of its arguments with fewer arguments than declared. IBM XL
24362 compilers access floating-point arguments that do not fit in the
24363 RSA from the stack when a subroutine is compiled without
24364 optimization. Because always storing floating-point arguments on the
24365 stack is inefficient and rarely needed, this option is not enabled by
24366 default and only is necessary when calling subroutines compiled by IBM
24367 XL compilers without optimization.
24371 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
24372 application written to use message passing with special startup code to
24373 enable the application to run. The system must have PE installed in the
24374 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
24375 must be overridden with the @option{-specs=} option to specify the
24376 appropriate directory location. The Parallel Environment does not
24377 support threads, so the @option{-mpe} option and the @option{-pthread}
24378 option are incompatible.
24380 @item -malign-natural
24381 @itemx -malign-power
24382 @opindex malign-natural
24383 @opindex malign-power
24384 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
24385 @option{-malign-natural} overrides the ABI-defined alignment of larger
24386 types, such as floating-point doubles, on their natural size-based boundary.
24387 The option @option{-malign-power} instructs GCC to follow the ABI-specified
24388 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
24390 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
24394 @itemx -mhard-float
24395 @opindex msoft-float
24396 @opindex mhard-float
24397 Generate code that does not use (uses) the floating-point register set.
24398 Software floating-point emulation is provided if you use the
24399 @option{-msoft-float} option, and pass the option to GCC when linking.
24402 @itemx -mno-multiple
24404 @opindex mno-multiple
24405 Generate code that uses (does not use) the load multiple word
24406 instructions and the store multiple word instructions. These
24407 instructions are generated by default on POWER systems, and not
24408 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
24409 PowerPC systems, since those instructions do not work when the
24410 processor is in little-endian mode. The exceptions are PPC740 and
24411 PPC750 which permit these instructions in little-endian mode.
24416 @opindex mno-update
24417 Generate code that uses (does not use) the load or store instructions
24418 that update the base register to the address of the calculated memory
24419 location. These instructions are generated by default. If you use
24420 @option{-mno-update}, there is a small window between the time that the
24421 stack pointer is updated and the address of the previous frame is
24422 stored, which means code that walks the stack frame across interrupts or
24423 signals may get corrupted data.
24425 @item -mavoid-indexed-addresses
24426 @itemx -mno-avoid-indexed-addresses
24427 @opindex mavoid-indexed-addresses
24428 @opindex mno-avoid-indexed-addresses
24429 Generate code that tries to avoid (not avoid) the use of indexed load
24430 or store instructions. These instructions can incur a performance
24431 penalty on Power6 processors in certain situations, such as when
24432 stepping through large arrays that cross a 16M boundary. This option
24433 is enabled by default when targeting Power6 and disabled otherwise.
24436 @itemx -mno-fused-madd
24437 @opindex mfused-madd
24438 @opindex mno-fused-madd
24439 Generate code that uses (does not use) the floating-point multiply and
24440 accumulate instructions. These instructions are generated by default
24441 if hardware floating point is used. The machine-dependent
24442 @option{-mfused-madd} option is now mapped to the machine-independent
24443 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
24444 mapped to @option{-ffp-contract=off}.
24450 Generate code that uses (does not use) the half-word multiply and
24451 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
24452 These instructions are generated by default when targeting those
24459 Generate code that uses (does not use) the string-search @samp{dlmzb}
24460 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
24461 generated by default when targeting those processors.
24463 @item -mno-bit-align
24465 @opindex mno-bit-align
24466 @opindex mbit-align
24467 On System V.4 and embedded PowerPC systems do not (do) force structures
24468 and unions that contain bit-fields to be aligned to the base type of the
24471 For example, by default a structure containing nothing but 8
24472 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
24473 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
24474 the structure is aligned to a 1-byte boundary and is 1 byte in
24477 @item -mno-strict-align
24478 @itemx -mstrict-align
24479 @opindex mno-strict-align
24480 @opindex mstrict-align
24481 On System V.4 and embedded PowerPC systems do not (do) assume that
24482 unaligned memory references are handled by the system.
24484 @item -mrelocatable
24485 @itemx -mno-relocatable
24486 @opindex mrelocatable
24487 @opindex mno-relocatable
24488 Generate code that allows (does not allow) a static executable to be
24489 relocated to a different address at run time. A simple embedded
24490 PowerPC system loader should relocate the entire contents of
24491 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
24492 a table of 32-bit addresses generated by this option. For this to
24493 work, all objects linked together must be compiled with
24494 @option{-mrelocatable} or @option{-mrelocatable-lib}.
24495 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
24497 @item -mrelocatable-lib
24498 @itemx -mno-relocatable-lib
24499 @opindex mrelocatable-lib
24500 @opindex mno-relocatable-lib
24501 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
24502 @code{.fixup} section to allow static executables to be relocated at
24503 run time, but @option{-mrelocatable-lib} does not use the smaller stack
24504 alignment of @option{-mrelocatable}. Objects compiled with
24505 @option{-mrelocatable-lib} may be linked with objects compiled with
24506 any combination of the @option{-mrelocatable} options.
24512 On System V.4 and embedded PowerPC systems do not (do) assume that
24513 register 2 contains a pointer to a global area pointing to the addresses
24514 used in the program.
24517 @itemx -mlittle-endian
24519 @opindex mlittle-endian
24520 On System V.4 and embedded PowerPC systems compile code for the
24521 processor in little-endian mode. The @option{-mlittle-endian} option is
24522 the same as @option{-mlittle}.
24525 @itemx -mbig-endian
24527 @opindex mbig-endian
24528 On System V.4 and embedded PowerPC systems compile code for the
24529 processor in big-endian mode. The @option{-mbig-endian} option is
24530 the same as @option{-mbig}.
24532 @item -mdynamic-no-pic
24533 @opindex mdynamic-no-pic
24534 On Darwin and Mac OS X systems, compile code so that it is not
24535 relocatable, but that its external references are relocatable. The
24536 resulting code is suitable for applications, but not shared
24539 @item -msingle-pic-base
24540 @opindex msingle-pic-base
24541 Treat the register used for PIC addressing as read-only, rather than
24542 loading it in the prologue for each function. The runtime system is
24543 responsible for initializing this register with an appropriate value
24544 before execution begins.
24546 @item -mprioritize-restricted-insns=@var{priority}
24547 @opindex mprioritize-restricted-insns
24548 This option controls the priority that is assigned to
24549 dispatch-slot restricted instructions during the second scheduling
24550 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
24551 or @samp{2} to assign no, highest, or second-highest (respectively)
24552 priority to dispatch-slot restricted
24555 @item -msched-costly-dep=@var{dependence_type}
24556 @opindex msched-costly-dep
24557 This option controls which dependences are considered costly
24558 by the target during instruction scheduling. The argument
24559 @var{dependence_type} takes one of the following values:
24563 No dependence is costly.
24566 All dependences are costly.
24568 @item @samp{true_store_to_load}
24569 A true dependence from store to load is costly.
24571 @item @samp{store_to_load}
24572 Any dependence from store to load is costly.
24575 Any dependence for which the latency is greater than or equal to
24576 @var{number} is costly.
24579 @item -minsert-sched-nops=@var{scheme}
24580 @opindex minsert-sched-nops
24581 This option controls which NOP insertion scheme is used during
24582 the second scheduling pass. The argument @var{scheme} takes one of the
24590 Pad with NOPs any dispatch group that has vacant issue slots,
24591 according to the scheduler's grouping.
24593 @item @samp{regroup_exact}
24594 Insert NOPs to force costly dependent insns into
24595 separate groups. Insert exactly as many NOPs as needed to force an insn
24596 to a new group, according to the estimated processor grouping.
24599 Insert NOPs to force costly dependent insns into
24600 separate groups. Insert @var{number} NOPs to force an insn to a new group.
24604 @opindex mcall-sysv
24605 On System V.4 and embedded PowerPC systems compile code using calling
24606 conventions that adhere to the March 1995 draft of the System V
24607 Application Binary Interface, PowerPC processor supplement. This is the
24608 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
24610 @item -mcall-sysv-eabi
24612 @opindex mcall-sysv-eabi
24613 @opindex mcall-eabi
24614 Specify both @option{-mcall-sysv} and @option{-meabi} options.
24616 @item -mcall-sysv-noeabi
24617 @opindex mcall-sysv-noeabi
24618 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
24620 @item -mcall-aixdesc
24622 On System V.4 and embedded PowerPC systems compile code for the AIX
24626 @opindex mcall-linux
24627 On System V.4 and embedded PowerPC systems compile code for the
24628 Linux-based GNU system.
24630 @item -mcall-freebsd
24631 @opindex mcall-freebsd
24632 On System V.4 and embedded PowerPC systems compile code for the
24633 FreeBSD operating system.
24635 @item -mcall-netbsd
24636 @opindex mcall-netbsd
24637 On System V.4 and embedded PowerPC systems compile code for the
24638 NetBSD operating system.
24640 @item -mcall-openbsd
24641 @opindex mcall-netbsd
24642 On System V.4 and embedded PowerPC systems compile code for the
24643 OpenBSD operating system.
24645 @item -mtraceback=@var{traceback_type}
24646 @opindex mtraceback
24647 Select the type of traceback table. Valid values for @var{traceback_type}
24648 are @samp{full}, @samp{part}, and @samp{no}.
24650 @item -maix-struct-return
24651 @opindex maix-struct-return
24652 Return all structures in memory (as specified by the AIX ABI)@.
24654 @item -msvr4-struct-return
24655 @opindex msvr4-struct-return
24656 Return structures smaller than 8 bytes in registers (as specified by the
24659 @item -mabi=@var{abi-type}
24661 Extend the current ABI with a particular extension, or remove such extension.
24662 Valid values are @samp{altivec}, @samp{no-altivec},
24663 @samp{ibmlongdouble}, @samp{ieeelongdouble},
24664 @samp{elfv1}, @samp{elfv2}@.
24666 @item -mabi=ibmlongdouble
24667 @opindex mabi=ibmlongdouble
24668 Change the current ABI to use IBM extended-precision long double.
24669 This is not likely to work if your system defaults to using IEEE
24670 extended-precision long double. If you change the long double type
24671 from IEEE extended-precision, the compiler will issue a warning unless
24672 you use the @option{-Wno-psabi} option. Requires @option{-mlong-double-128}
24675 @item -mabi=ieeelongdouble
24676 @opindex mabi=ieeelongdouble
24677 Change the current ABI to use IEEE extended-precision long double.
24678 This is not likely to work if your system defaults to using IBM
24679 extended-precision long double. If you change the long double type
24680 from IBM extended-precision, the compiler will issue a warning unless
24681 you use the @option{-Wno-psabi} option. Requires @option{-mlong-double-128}
24685 @opindex mabi=elfv1
24686 Change the current ABI to use the ELFv1 ABI.
24687 This is the default ABI for big-endian PowerPC 64-bit Linux.
24688 Overriding the default ABI requires special system support and is
24689 likely to fail in spectacular ways.
24692 @opindex mabi=elfv2
24693 Change the current ABI to use the ELFv2 ABI.
24694 This is the default ABI for little-endian PowerPC 64-bit Linux.
24695 Overriding the default ABI requires special system support and is
24696 likely to fail in spectacular ways.
24698 @item -mgnu-attribute
24699 @itemx -mno-gnu-attribute
24700 @opindex mgnu-attribute
24701 @opindex mno-gnu-attribute
24702 Emit .gnu_attribute assembly directives to set tag/value pairs in a
24703 .gnu.attributes section that specify ABI variations in function
24704 parameters or return values.
24707 @itemx -mno-prototype
24708 @opindex mprototype
24709 @opindex mno-prototype
24710 On System V.4 and embedded PowerPC systems assume that all calls to
24711 variable argument functions are properly prototyped. Otherwise, the
24712 compiler must insert an instruction before every non-prototyped call to
24713 set or clear bit 6 of the condition code register (@code{CR}) to
24714 indicate whether floating-point values are passed in the floating-point
24715 registers in case the function takes variable arguments. With
24716 @option{-mprototype}, only calls to prototyped variable argument functions
24717 set or clear the bit.
24721 On embedded PowerPC systems, assume that the startup module is called
24722 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
24723 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
24728 On embedded PowerPC systems, assume that the startup module is called
24729 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
24734 On embedded PowerPC systems, assume that the startup module is called
24735 @file{crt0.o} and the standard C libraries are @file{libads.a} and
24738 @item -myellowknife
24739 @opindex myellowknife
24740 On embedded PowerPC systems, assume that the startup module is called
24741 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
24746 On System V.4 and embedded PowerPC systems, specify that you are
24747 compiling for a VxWorks system.
24751 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
24752 header to indicate that @samp{eabi} extended relocations are used.
24758 On System V.4 and embedded PowerPC systems do (do not) adhere to the
24759 Embedded Applications Binary Interface (EABI), which is a set of
24760 modifications to the System V.4 specifications. Selecting @option{-meabi}
24761 means that the stack is aligned to an 8-byte boundary, a function
24762 @code{__eabi} is called from @code{main} to set up the EABI
24763 environment, and the @option{-msdata} option can use both @code{r2} and
24764 @code{r13} to point to two separate small data areas. Selecting
24765 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
24766 no EABI initialization function is called from @code{main}, and the
24767 @option{-msdata} option only uses @code{r13} to point to a single
24768 small data area. The @option{-meabi} option is on by default if you
24769 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
24772 @opindex msdata=eabi
24773 On System V.4 and embedded PowerPC systems, put small initialized
24774 @code{const} global and static data in the @code{.sdata2} section, which
24775 is pointed to by register @code{r2}. Put small initialized
24776 non-@code{const} global and static data in the @code{.sdata} section,
24777 which is pointed to by register @code{r13}. Put small uninitialized
24778 global and static data in the @code{.sbss} section, which is adjacent to
24779 the @code{.sdata} section. The @option{-msdata=eabi} option is
24780 incompatible with the @option{-mrelocatable} option. The
24781 @option{-msdata=eabi} option also sets the @option{-memb} option.
24784 @opindex msdata=sysv
24785 On System V.4 and embedded PowerPC systems, put small global and static
24786 data in the @code{.sdata} section, which is pointed to by register
24787 @code{r13}. Put small uninitialized global and static data in the
24788 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
24789 The @option{-msdata=sysv} option is incompatible with the
24790 @option{-mrelocatable} option.
24792 @item -msdata=default
24794 @opindex msdata=default
24796 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
24797 compile code the same as @option{-msdata=eabi}, otherwise compile code the
24798 same as @option{-msdata=sysv}.
24801 @opindex msdata=data
24802 On System V.4 and embedded PowerPC systems, put small global
24803 data in the @code{.sdata} section. Put small uninitialized global
24804 data in the @code{.sbss} section. Do not use register @code{r13}
24805 to address small data however. This is the default behavior unless
24806 other @option{-msdata} options are used.
24810 @opindex msdata=none
24812 On embedded PowerPC systems, put all initialized global and static data
24813 in the @code{.data} section, and all uninitialized data in the
24814 @code{.bss} section.
24816 @item -mreadonly-in-sdata
24817 @opindex mreadonly-in-sdata
24818 @opindex mno-readonly-in-sdata
24819 Put read-only objects in the @code{.sdata} section as well. This is the
24822 @item -mblock-move-inline-limit=@var{num}
24823 @opindex mblock-move-inline-limit
24824 Inline all block moves (such as calls to @code{memcpy} or structure
24825 copies) less than or equal to @var{num} bytes. The minimum value for
24826 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
24827 targets. The default value is target-specific.
24829 @item -mblock-compare-inline-limit=@var{num}
24830 @opindex mblock-compare-inline-limit
24831 Generate non-looping inline code for all block compares (such as calls
24832 to @code{memcmp} or structure compares) less than or equal to @var{num}
24833 bytes. If @var{num} is 0, all inline expansion (non-loop and loop) of
24834 block compare is disabled. The default value is target-specific.
24836 @item -mblock-compare-inline-loop-limit=@var{num}
24837 @opindex mblock-compare-inline-loop-limit
24838 Generate an inline expansion using loop code for all block compares that
24839 are less than or equal to @var{num} bytes, but greater than the limit
24840 for non-loop inline block compare expansion. If the block length is not
24841 constant, at most @var{num} bytes will be compared before @code{memcmp}
24842 is called to compare the remainder of the block. The default value is
24845 @item -mstring-compare-inline-limit=@var{num}
24846 @opindex mstring-compare-inline-limit
24847 Compare at most @var{num} string bytes with inline code.
24848 If the difference or end of string is not found at the
24849 end of the inline compare a call to @code{strcmp} or @code{strncmp} will
24850 take care of the rest of the comparison. The default is 64 bytes.
24854 @cindex smaller data references (PowerPC)
24855 @cindex .sdata/.sdata2 references (PowerPC)
24856 On embedded PowerPC systems, put global and static items less than or
24857 equal to @var{num} bytes into the small data or BSS sections instead of
24858 the normal data or BSS section. By default, @var{num} is 8. The
24859 @option{-G @var{num}} switch is also passed to the linker.
24860 All modules should be compiled with the same @option{-G @var{num}} value.
24863 @itemx -mno-regnames
24865 @opindex mno-regnames
24866 On System V.4 and embedded PowerPC systems do (do not) emit register
24867 names in the assembly language output using symbolic forms.
24870 @itemx -mno-longcall
24872 @opindex mno-longcall
24873 By default assume that all calls are far away so that a longer and more
24874 expensive calling sequence is required. This is required for calls
24875 farther than 32 megabytes (33,554,432 bytes) from the current location.
24876 A short call is generated if the compiler knows
24877 the call cannot be that far away. This setting can be overridden by
24878 the @code{shortcall} function attribute, or by @code{#pragma
24881 Some linkers are capable of detecting out-of-range calls and generating
24882 glue code on the fly. On these systems, long calls are unnecessary and
24883 generate slower code. As of this writing, the AIX linker can do this,
24884 as can the GNU linker for PowerPC/64. It is planned to add this feature
24885 to the GNU linker for 32-bit PowerPC systems as well.
24887 On PowerPC64 ELFv2 and 32-bit PowerPC systems with newer GNU linkers,
24888 GCC can generate long calls using an inline PLT call sequence (see
24889 @option{-mpltseq}). PowerPC with @option{-mbss-plt} and PowerPC64
24890 ELFv1 (big-endian) do not support inline PLT calls.
24892 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
24893 callee, L42}, plus a @dfn{branch island} (glue code). The two target
24894 addresses represent the callee and the branch island. The
24895 Darwin/PPC linker prefers the first address and generates a @code{bl
24896 callee} if the PPC @code{bl} instruction reaches the callee directly;
24897 otherwise, the linker generates @code{bl L42} to call the branch
24898 island. The branch island is appended to the body of the
24899 calling function; it computes the full 32-bit address of the callee
24902 On Mach-O (Darwin) systems, this option directs the compiler emit to
24903 the glue for every direct call, and the Darwin linker decides whether
24904 to use or discard it.
24906 In the future, GCC may ignore all longcall specifications
24907 when the linker is known to generate glue.
24912 @opindex mno-pltseq
24913 Implement (do not implement) -fno-plt and long calls using an inline
24914 PLT call sequence that supports lazy linking and long calls to
24915 functions in dlopen'd shared libraries. Inline PLT calls are only
24916 supported on PowerPC64 ELFv2 and 32-bit PowerPC systems with newer GNU
24917 linkers, and are enabled by default if the support is detected when
24918 configuring GCC, and, in the case of 32-bit PowerPC, if GCC is
24919 configured with @option{--enable-secureplt}. @option{-mpltseq} code
24920 and @option{-mbss-plt} 32-bit PowerPC relocatable objects may not be
24923 @item -mtls-markers
24924 @itemx -mno-tls-markers
24925 @opindex mtls-markers
24926 @opindex mno-tls-markers
24927 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
24928 specifying the function argument. The relocation allows the linker to
24929 reliably associate function call with argument setup instructions for
24930 TLS optimization, which in turn allows GCC to better schedule the
24936 This option enables use of the reciprocal estimate and
24937 reciprocal square root estimate instructions with additional
24938 Newton-Raphson steps to increase precision instead of doing a divide or
24939 square root and divide for floating-point arguments. You should use
24940 the @option{-ffast-math} option when using @option{-mrecip} (or at
24941 least @option{-funsafe-math-optimizations},
24942 @option{-ffinite-math-only}, @option{-freciprocal-math} and
24943 @option{-fno-trapping-math}). Note that while the throughput of the
24944 sequence is generally higher than the throughput of the non-reciprocal
24945 instruction, the precision of the sequence can be decreased by up to 2
24946 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
24949 @item -mrecip=@var{opt}
24950 @opindex mrecip=opt
24951 This option controls which reciprocal estimate instructions
24952 may be used. @var{opt} is a comma-separated list of options, which may
24953 be preceded by a @code{!} to invert the option:
24958 Enable all estimate instructions.
24961 Enable the default instructions, equivalent to @option{-mrecip}.
24964 Disable all estimate instructions, equivalent to @option{-mno-recip}.
24967 Enable the reciprocal approximation instructions for both
24968 single and double precision.
24971 Enable the single-precision reciprocal approximation instructions.
24974 Enable the double-precision reciprocal approximation instructions.
24977 Enable the reciprocal square root approximation instructions for both
24978 single and double precision.
24981 Enable the single-precision reciprocal square root approximation instructions.
24984 Enable the double-precision reciprocal square root approximation instructions.
24988 So, for example, @option{-mrecip=all,!rsqrtd} enables
24989 all of the reciprocal estimate instructions, except for the
24990 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
24991 which handle the double-precision reciprocal square root calculations.
24993 @item -mrecip-precision
24994 @itemx -mno-recip-precision
24995 @opindex mrecip-precision
24996 Assume (do not assume) that the reciprocal estimate instructions
24997 provide higher-precision estimates than is mandated by the PowerPC
24998 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
24999 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
25000 The double-precision square root estimate instructions are not generated by
25001 default on low-precision machines, since they do not provide an
25002 estimate that converges after three steps.
25004 @item -mveclibabi=@var{type}
25005 @opindex mveclibabi
25006 Specifies the ABI type to use for vectorizing intrinsics using an
25007 external library. The only type supported at present is @samp{mass},
25008 which specifies to use IBM's Mathematical Acceleration Subsystem
25009 (MASS) libraries for vectorizing intrinsics using external libraries.
25010 GCC currently emits calls to @code{acosd2}, @code{acosf4},
25011 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
25012 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
25013 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
25014 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
25015 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
25016 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
25017 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
25018 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
25019 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
25020 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
25021 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
25022 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
25023 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
25024 for power7. Both @option{-ftree-vectorize} and
25025 @option{-funsafe-math-optimizations} must also be enabled. The MASS
25026 libraries must be specified at link time.
25031 Generate (do not generate) the @code{friz} instruction when the
25032 @option{-funsafe-math-optimizations} option is used to optimize
25033 rounding of floating-point values to 64-bit integer and back to floating
25034 point. The @code{friz} instruction does not return the same value if
25035 the floating-point number is too large to fit in an integer.
25037 @item -mpointers-to-nested-functions
25038 @itemx -mno-pointers-to-nested-functions
25039 @opindex mpointers-to-nested-functions
25040 Generate (do not generate) code to load up the static chain register
25041 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
25042 systems where a function pointer points to a 3-word descriptor giving
25043 the function address, TOC value to be loaded in register @code{r2}, and
25044 static chain value to be loaded in register @code{r11}. The
25045 @option{-mpointers-to-nested-functions} is on by default. You cannot
25046 call through pointers to nested functions or pointers
25047 to functions compiled in other languages that use the static chain if
25048 you use @option{-mno-pointers-to-nested-functions}.
25050 @item -msave-toc-indirect
25051 @itemx -mno-save-toc-indirect
25052 @opindex msave-toc-indirect
25053 Generate (do not generate) code to save the TOC value in the reserved
25054 stack location in the function prologue if the function calls through
25055 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
25056 saved in the prologue, it is saved just before the call through the
25057 pointer. The @option{-mno-save-toc-indirect} option is the default.
25059 @item -mcompat-align-parm
25060 @itemx -mno-compat-align-parm
25061 @opindex mcompat-align-parm
25062 Generate (do not generate) code to pass structure parameters with a
25063 maximum alignment of 64 bits, for compatibility with older versions
25066 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
25067 structure parameter on a 128-bit boundary when that structure contained
25068 a member requiring 128-bit alignment. This is corrected in more
25069 recent versions of GCC. This option may be used to generate code
25070 that is compatible with functions compiled with older versions of
25073 The @option{-mno-compat-align-parm} option is the default.
25075 @item -mstack-protector-guard=@var{guard}
25076 @itemx -mstack-protector-guard-reg=@var{reg}
25077 @itemx -mstack-protector-guard-offset=@var{offset}
25078 @itemx -mstack-protector-guard-symbol=@var{symbol}
25079 @opindex mstack-protector-guard
25080 @opindex mstack-protector-guard-reg
25081 @opindex mstack-protector-guard-offset
25082 @opindex mstack-protector-guard-symbol
25083 Generate stack protection code using canary at @var{guard}. Supported
25084 locations are @samp{global} for global canary or @samp{tls} for per-thread
25085 canary in the TLS block (the default with GNU libc version 2.4 or later).
25087 With the latter choice the options
25088 @option{-mstack-protector-guard-reg=@var{reg}} and
25089 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
25090 which register to use as base register for reading the canary, and from what
25091 offset from that base register. The default for those is as specified in the
25092 relevant ABI. @option{-mstack-protector-guard-symbol=@var{symbol}} overrides
25093 the offset with a symbol reference to a canary in the TLS block.
25099 Generate (do not generate) pc-relative addressing when the option
25100 @option{-mcpu=future} is used.
25104 @subsection RX Options
25107 These command-line options are defined for RX targets:
25110 @item -m64bit-doubles
25111 @itemx -m32bit-doubles
25112 @opindex m64bit-doubles
25113 @opindex m32bit-doubles
25114 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
25115 or 32 bits (@option{-m32bit-doubles}) in size. The default is
25116 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
25117 works on 32-bit values, which is why the default is
25118 @option{-m32bit-doubles}.
25124 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
25125 floating-point hardware. The default is enabled for the RX600
25126 series and disabled for the RX200 series.
25128 Floating-point instructions are only generated for 32-bit floating-point
25129 values, however, so the FPU hardware is not used for doubles if the
25130 @option{-m64bit-doubles} option is used.
25132 @emph{Note} If the @option{-fpu} option is enabled then
25133 @option{-funsafe-math-optimizations} is also enabled automatically.
25134 This is because the RX FPU instructions are themselves unsafe.
25136 @item -mcpu=@var{name}
25138 Selects the type of RX CPU to be targeted. Currently three types are
25139 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
25140 the specific @samp{RX610} CPU. The default is @samp{RX600}.
25142 The only difference between @samp{RX600} and @samp{RX610} is that the
25143 @samp{RX610} does not support the @code{MVTIPL} instruction.
25145 The @samp{RX200} series does not have a hardware floating-point unit
25146 and so @option{-nofpu} is enabled by default when this type is
25149 @item -mbig-endian-data
25150 @itemx -mlittle-endian-data
25151 @opindex mbig-endian-data
25152 @opindex mlittle-endian-data
25153 Store data (but not code) in the big-endian format. The default is
25154 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
25157 @item -msmall-data-limit=@var{N}
25158 @opindex msmall-data-limit
25159 Specifies the maximum size in bytes of global and static variables
25160 which can be placed into the small data area. Using the small data
25161 area can lead to smaller and faster code, but the size of area is
25162 limited and it is up to the programmer to ensure that the area does
25163 not overflow. Also when the small data area is used one of the RX's
25164 registers (usually @code{r13}) is reserved for use pointing to this
25165 area, so it is no longer available for use by the compiler. This
25166 could result in slower and/or larger code if variables are pushed onto
25167 the stack instead of being held in this register.
25169 Note, common variables (variables that have not been initialized) and
25170 constants are not placed into the small data area as they are assigned
25171 to other sections in the output executable.
25173 The default value is zero, which disables this feature. Note, this
25174 feature is not enabled by default with higher optimization levels
25175 (@option{-O2} etc) because of the potentially detrimental effects of
25176 reserving a register. It is up to the programmer to experiment and
25177 discover whether this feature is of benefit to their program. See the
25178 description of the @option{-mpid} option for a description of how the
25179 actual register to hold the small data area pointer is chosen.
25185 Use the simulator runtime. The default is to use the libgloss
25186 board-specific runtime.
25188 @item -mas100-syntax
25189 @itemx -mno-as100-syntax
25190 @opindex mas100-syntax
25191 @opindex mno-as100-syntax
25192 When generating assembler output use a syntax that is compatible with
25193 Renesas's AS100 assembler. This syntax can also be handled by the GAS
25194 assembler, but it has some restrictions so it is not generated by default.
25196 @item -mmax-constant-size=@var{N}
25197 @opindex mmax-constant-size
25198 Specifies the maximum size, in bytes, of a constant that can be used as
25199 an operand in a RX instruction. Although the RX instruction set does
25200 allow constants of up to 4 bytes in length to be used in instructions,
25201 a longer value equates to a longer instruction. Thus in some
25202 circumstances it can be beneficial to restrict the size of constants
25203 that are used in instructions. Constants that are too big are instead
25204 placed into a constant pool and referenced via register indirection.
25206 The value @var{N} can be between 0 and 4. A value of 0 (the default)
25207 or 4 means that constants of any size are allowed.
25211 Enable linker relaxation. Linker relaxation is a process whereby the
25212 linker attempts to reduce the size of a program by finding shorter
25213 versions of various instructions. Disabled by default.
25215 @item -mint-register=@var{N}
25216 @opindex mint-register
25217 Specify the number of registers to reserve for fast interrupt handler
25218 functions. The value @var{N} can be between 0 and 4. A value of 1
25219 means that register @code{r13} is reserved for the exclusive use
25220 of fast interrupt handlers. A value of 2 reserves @code{r13} and
25221 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
25222 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
25223 A value of 0, the default, does not reserve any registers.
25225 @item -msave-acc-in-interrupts
25226 @opindex msave-acc-in-interrupts
25227 Specifies that interrupt handler functions should preserve the
25228 accumulator register. This is only necessary if normal code might use
25229 the accumulator register, for example because it performs 64-bit
25230 multiplications. The default is to ignore the accumulator as this
25231 makes the interrupt handlers faster.
25237 Enables the generation of position independent data. When enabled any
25238 access to constant data is done via an offset from a base address
25239 held in a register. This allows the location of constant data to be
25240 determined at run time without requiring the executable to be
25241 relocated, which is a benefit to embedded applications with tight
25242 memory constraints. Data that can be modified is not affected by this
25245 Note, using this feature reserves a register, usually @code{r13}, for
25246 the constant data base address. This can result in slower and/or
25247 larger code, especially in complicated functions.
25249 The actual register chosen to hold the constant data base address
25250 depends upon whether the @option{-msmall-data-limit} and/or the
25251 @option{-mint-register} command-line options are enabled. Starting
25252 with register @code{r13} and proceeding downwards, registers are
25253 allocated first to satisfy the requirements of @option{-mint-register},
25254 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
25255 is possible for the small data area register to be @code{r8} if both
25256 @option{-mint-register=4} and @option{-mpid} are specified on the
25259 By default this feature is not enabled. The default can be restored
25260 via the @option{-mno-pid} command-line option.
25262 @item -mno-warn-multiple-fast-interrupts
25263 @itemx -mwarn-multiple-fast-interrupts
25264 @opindex mno-warn-multiple-fast-interrupts
25265 @opindex mwarn-multiple-fast-interrupts
25266 Prevents GCC from issuing a warning message if it finds more than one
25267 fast interrupt handler when it is compiling a file. The default is to
25268 issue a warning for each extra fast interrupt handler found, as the RX
25269 only supports one such interrupt.
25271 @item -mallow-string-insns
25272 @itemx -mno-allow-string-insns
25273 @opindex mallow-string-insns
25274 @opindex mno-allow-string-insns
25275 Enables or disables the use of the string manipulation instructions
25276 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
25277 @code{SWHILE} and also the @code{RMPA} instruction. These
25278 instructions may prefetch data, which is not safe to do if accessing
25279 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
25280 for more information).
25282 The default is to allow these instructions, but it is not possible for
25283 GCC to reliably detect all circumstances where a string instruction
25284 might be used to access an I/O register, so their use cannot be
25285 disabled automatically. Instead it is reliant upon the programmer to
25286 use the @option{-mno-allow-string-insns} option if their program
25287 accesses I/O space.
25289 When the instructions are enabled GCC defines the C preprocessor
25290 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
25291 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
25297 Use only (or not only) @code{JSR} instructions to access functions.
25298 This option can be used when code size exceeds the range of @code{BSR}
25299 instructions. Note that @option{-mno-jsr} does not mean to not use
25300 @code{JSR} but instead means that any type of branch may be used.
25303 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
25304 has special significance to the RX port when used with the
25305 @code{interrupt} function attribute. This attribute indicates a
25306 function intended to process fast interrupts. GCC ensures
25307 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
25308 and/or @code{r13} and only provided that the normal use of the
25309 corresponding registers have been restricted via the
25310 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
25313 @node S/390 and zSeries Options
25314 @subsection S/390 and zSeries Options
25315 @cindex S/390 and zSeries Options
25317 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
25321 @itemx -msoft-float
25322 @opindex mhard-float
25323 @opindex msoft-float
25324 Use (do not use) the hardware floating-point instructions and registers
25325 for floating-point operations. When @option{-msoft-float} is specified,
25326 functions in @file{libgcc.a} are used to perform floating-point
25327 operations. When @option{-mhard-float} is specified, the compiler
25328 generates IEEE floating-point instructions. This is the default.
25331 @itemx -mno-hard-dfp
25333 @opindex mno-hard-dfp
25334 Use (do not use) the hardware decimal-floating-point instructions for
25335 decimal-floating-point operations. When @option{-mno-hard-dfp} is
25336 specified, functions in @file{libgcc.a} are used to perform
25337 decimal-floating-point operations. When @option{-mhard-dfp} is
25338 specified, the compiler generates decimal-floating-point hardware
25339 instructions. This is the default for @option{-march=z9-ec} or higher.
25341 @item -mlong-double-64
25342 @itemx -mlong-double-128
25343 @opindex mlong-double-64
25344 @opindex mlong-double-128
25345 These switches control the size of @code{long double} type. A size
25346 of 64 bits makes the @code{long double} type equivalent to the @code{double}
25347 type. This is the default.
25350 @itemx -mno-backchain
25351 @opindex mbackchain
25352 @opindex mno-backchain
25353 Store (do not store) the address of the caller's frame as backchain pointer
25354 into the callee's stack frame.
25355 A backchain may be needed to allow debugging using tools that do not understand
25356 DWARF call frame information.
25357 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
25358 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
25359 the backchain is placed into the topmost word of the 96/160 byte register
25362 In general, code compiled with @option{-mbackchain} is call-compatible with
25363 code compiled with @option{-mmo-backchain}; however, use of the backchain
25364 for debugging purposes usually requires that the whole binary is built with
25365 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
25366 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
25367 to build a linux kernel use @option{-msoft-float}.
25369 The default is to not maintain the backchain.
25371 @item -mpacked-stack
25372 @itemx -mno-packed-stack
25373 @opindex mpacked-stack
25374 @opindex mno-packed-stack
25375 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
25376 specified, the compiler uses the all fields of the 96/160 byte register save
25377 area only for their default purpose; unused fields still take up stack space.
25378 When @option{-mpacked-stack} is specified, register save slots are densely
25379 packed at the top of the register save area; unused space is reused for other
25380 purposes, allowing for more efficient use of the available stack space.
25381 However, when @option{-mbackchain} is also in effect, the topmost word of
25382 the save area is always used to store the backchain, and the return address
25383 register is always saved two words below the backchain.
25385 As long as the stack frame backchain is not used, code generated with
25386 @option{-mpacked-stack} is call-compatible with code generated with
25387 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
25388 S/390 or zSeries generated code that uses the stack frame backchain at run
25389 time, not just for debugging purposes. Such code is not call-compatible
25390 with code compiled with @option{-mpacked-stack}. Also, note that the
25391 combination of @option{-mbackchain},
25392 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
25393 to build a linux kernel use @option{-msoft-float}.
25395 The default is to not use the packed stack layout.
25398 @itemx -mno-small-exec
25399 @opindex msmall-exec
25400 @opindex mno-small-exec
25401 Generate (or do not generate) code using the @code{bras} instruction
25402 to do subroutine calls.
25403 This only works reliably if the total executable size does not
25404 exceed 64k. The default is to use the @code{basr} instruction instead,
25405 which does not have this limitation.
25411 When @option{-m31} is specified, generate code compliant to the
25412 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
25413 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
25414 particular to generate 64-bit instructions. For the @samp{s390}
25415 targets, the default is @option{-m31}, while the @samp{s390x}
25416 targets default to @option{-m64}.
25422 When @option{-mzarch} is specified, generate code using the
25423 instructions available on z/Architecture.
25424 When @option{-mesa} is specified, generate code using the
25425 instructions available on ESA/390. Note that @option{-mesa} is
25426 not possible with @option{-m64}.
25427 When generating code compliant to the GNU/Linux for S/390 ABI,
25428 the default is @option{-mesa}. When generating code compliant
25429 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
25435 The @option{-mhtm} option enables a set of builtins making use of
25436 instructions available with the transactional execution facility
25437 introduced with the IBM zEnterprise EC12 machine generation
25438 @ref{S/390 System z Built-in Functions}.
25439 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
25445 When @option{-mvx} is specified, generate code using the instructions
25446 available with the vector extension facility introduced with the IBM
25447 z13 machine generation.
25448 This option changes the ABI for some vector type values with regard to
25449 alignment and calling conventions. In case vector type values are
25450 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
25451 command will be added to mark the resulting binary with the ABI used.
25452 @option{-mvx} is enabled by default when using @option{-march=z13}.
25455 @itemx -mno-zvector
25457 @opindex mno-zvector
25458 The @option{-mzvector} option enables vector language extensions and
25459 builtins using instructions available with the vector extension
25460 facility introduced with the IBM z13 machine generation.
25461 This option adds support for @samp{vector} to be used as a keyword to
25462 define vector type variables and arguments. @samp{vector} is only
25463 available when GNU extensions are enabled. It will not be expanded
25464 when requesting strict standard compliance e.g.@: with @option{-std=c99}.
25465 In addition to the GCC low-level builtins @option{-mzvector} enables
25466 a set of builtins added for compatibility with AltiVec-style
25467 implementations like Power and Cell. In order to make use of these
25468 builtins the header file @file{vecintrin.h} needs to be included.
25469 @option{-mzvector} is disabled by default.
25475 Generate (or do not generate) code using the @code{mvcle} instruction
25476 to perform block moves. When @option{-mno-mvcle} is specified,
25477 use a @code{mvc} loop instead. This is the default unless optimizing for
25484 Print (or do not print) additional debug information when compiling.
25485 The default is to not print debug information.
25487 @item -march=@var{cpu-type}
25489 Generate code that runs on @var{cpu-type}, which is the name of a
25490 system representing a certain processor type. Possible values for
25491 @var{cpu-type} are @samp{z900}/@samp{arch5}, @samp{z990}/@samp{arch6},
25492 @samp{z9-109}, @samp{z9-ec}/@samp{arch7}, @samp{z10}/@samp{arch8},
25493 @samp{z196}/@samp{arch9}, @samp{zEC12}, @samp{z13}/@samp{arch11},
25494 @samp{z14}/@samp{arch12}, and @samp{native}.
25496 The default is @option{-march=z900}.
25498 Specifying @samp{native} as cpu type can be used to select the best
25499 architecture option for the host processor.
25500 @option{-march=native} has no effect if GCC does not recognize the
25503 @item -mtune=@var{cpu-type}
25505 Tune to @var{cpu-type} everything applicable about the generated code,
25506 except for the ABI and the set of available instructions.
25507 The list of @var{cpu-type} values is the same as for @option{-march}.
25508 The default is the value used for @option{-march}.
25511 @itemx -mno-tpf-trace
25512 @opindex mtpf-trace
25513 @opindex mno-tpf-trace
25514 Generate code that adds (does not add) in TPF OS specific branches to trace
25515 routines in the operating system. This option is off by default, even
25516 when compiling for the TPF OS@.
25519 @itemx -mno-fused-madd
25520 @opindex mfused-madd
25521 @opindex mno-fused-madd
25522 Generate code that uses (does not use) the floating-point multiply and
25523 accumulate instructions. These instructions are generated by default if
25524 hardware floating point is used.
25526 @item -mwarn-framesize=@var{framesize}
25527 @opindex mwarn-framesize
25528 Emit a warning if the current function exceeds the given frame size. Because
25529 this is a compile-time check it doesn't need to be a real problem when the program
25530 runs. It is intended to identify functions that most probably cause
25531 a stack overflow. It is useful to be used in an environment with limited stack
25532 size e.g.@: the linux kernel.
25534 @item -mwarn-dynamicstack
25535 @opindex mwarn-dynamicstack
25536 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
25537 arrays. This is generally a bad idea with a limited stack size.
25539 @item -mstack-guard=@var{stack-guard}
25540 @itemx -mstack-size=@var{stack-size}
25541 @opindex mstack-guard
25542 @opindex mstack-size
25543 If these options are provided the S/390 back end emits additional instructions in
25544 the function prologue that trigger a trap if the stack size is @var{stack-guard}
25545 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
25546 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
25547 the frame size of the compiled function is chosen.
25548 These options are intended to be used to help debugging stack overflow problems.
25549 The additionally emitted code causes only little overhead and hence can also be
25550 used in production-like systems without greater performance degradation. The given
25551 values have to be exact powers of 2 and @var{stack-size} has to be greater than
25552 @var{stack-guard} without exceeding 64k.
25553 In order to be efficient the extra code makes the assumption that the stack starts
25554 at an address aligned to the value given by @var{stack-size}.
25555 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
25557 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
25559 If the hotpatch option is enabled, a ``hot-patching'' function
25560 prologue is generated for all functions in the compilation unit.
25561 The funtion label is prepended with the given number of two-byte
25562 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
25563 the label, 2 * @var{post-halfwords} bytes are appended, using the
25564 largest NOP like instructions the architecture allows (maximum
25567 If both arguments are zero, hotpatching is disabled.
25569 This option can be overridden for individual functions with the
25570 @code{hotpatch} attribute.
25573 @node Score Options
25574 @subsection Score Options
25575 @cindex Score Options
25577 These options are defined for Score implementations:
25582 Compile code for big-endian mode. This is the default.
25586 Compile code for little-endian mode.
25590 Disable generation of @code{bcnz} instructions.
25594 Enable generation of unaligned load and store instructions.
25598 Enable the use of multiply-accumulate instructions. Disabled by default.
25602 Specify the SCORE5 as the target architecture.
25606 Specify the SCORE5U of the target architecture.
25610 Specify the SCORE7 as the target architecture. This is the default.
25614 Specify the SCORE7D as the target architecture.
25618 @subsection SH Options
25620 These @samp{-m} options are defined for the SH implementations:
25625 Generate code for the SH1.
25629 Generate code for the SH2.
25632 Generate code for the SH2e.
25636 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
25637 that the floating-point unit is not used.
25639 @item -m2a-single-only
25640 @opindex m2a-single-only
25641 Generate code for the SH2a-FPU, in such a way that no double-precision
25642 floating-point operations are used.
25645 @opindex m2a-single
25646 Generate code for the SH2a-FPU assuming the floating-point unit is in
25647 single-precision mode by default.
25651 Generate code for the SH2a-FPU assuming the floating-point unit is in
25652 double-precision mode by default.
25656 Generate code for the SH3.
25660 Generate code for the SH3e.
25664 Generate code for the SH4 without a floating-point unit.
25666 @item -m4-single-only
25667 @opindex m4-single-only
25668 Generate code for the SH4 with a floating-point unit that only
25669 supports single-precision arithmetic.
25673 Generate code for the SH4 assuming the floating-point unit is in
25674 single-precision mode by default.
25678 Generate code for the SH4.
25682 Generate code for SH4-100.
25684 @item -m4-100-nofpu
25685 @opindex m4-100-nofpu
25686 Generate code for SH4-100 in such a way that the
25687 floating-point unit is not used.
25689 @item -m4-100-single
25690 @opindex m4-100-single
25691 Generate code for SH4-100 assuming the floating-point unit is in
25692 single-precision mode by default.
25694 @item -m4-100-single-only
25695 @opindex m4-100-single-only
25696 Generate code for SH4-100 in such a way that no double-precision
25697 floating-point operations are used.
25701 Generate code for SH4-200.
25703 @item -m4-200-nofpu
25704 @opindex m4-200-nofpu
25705 Generate code for SH4-200 without in such a way that the
25706 floating-point unit is not used.
25708 @item -m4-200-single
25709 @opindex m4-200-single
25710 Generate code for SH4-200 assuming the floating-point unit is in
25711 single-precision mode by default.
25713 @item -m4-200-single-only
25714 @opindex m4-200-single-only
25715 Generate code for SH4-200 in such a way that no double-precision
25716 floating-point operations are used.
25720 Generate code for SH4-300.
25722 @item -m4-300-nofpu
25723 @opindex m4-300-nofpu
25724 Generate code for SH4-300 without in such a way that the
25725 floating-point unit is not used.
25727 @item -m4-300-single
25728 @opindex m4-300-single
25729 Generate code for SH4-300 in such a way that no double-precision
25730 floating-point operations are used.
25732 @item -m4-300-single-only
25733 @opindex m4-300-single-only
25734 Generate code for SH4-300 in such a way that no double-precision
25735 floating-point operations are used.
25739 Generate code for SH4-340 (no MMU, no FPU).
25743 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
25748 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
25749 floating-point unit is not used.
25751 @item -m4a-single-only
25752 @opindex m4a-single-only
25753 Generate code for the SH4a, in such a way that no double-precision
25754 floating-point operations are used.
25757 @opindex m4a-single
25758 Generate code for the SH4a assuming the floating-point unit is in
25759 single-precision mode by default.
25763 Generate code for the SH4a.
25767 Same as @option{-m4a-nofpu}, except that it implicitly passes
25768 @option{-dsp} to the assembler. GCC doesn't generate any DSP
25769 instructions at the moment.
25773 Compile code for the processor in big-endian mode.
25777 Compile code for the processor in little-endian mode.
25781 Align doubles at 64-bit boundaries. Note that this changes the calling
25782 conventions, and thus some functions from the standard C library do
25783 not work unless you recompile it first with @option{-mdalign}.
25787 Shorten some address references at link time, when possible; uses the
25788 linker option @option{-relax}.
25792 Use 32-bit offsets in @code{switch} tables. The default is to use
25797 Enable the use of bit manipulation instructions on SH2A.
25801 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
25802 alignment constraints.
25806 Comply with the calling conventions defined by Renesas.
25809 @opindex mno-renesas
25810 Comply with the calling conventions defined for GCC before the Renesas
25811 conventions were available. This option is the default for all
25812 targets of the SH toolchain.
25815 @opindex mnomacsave
25816 Mark the @code{MAC} register as call-clobbered, even if
25817 @option{-mrenesas} is given.
25823 Control the IEEE compliance of floating-point comparisons, which affects the
25824 handling of cases where the result of a comparison is unordered. By default
25825 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
25826 enabled @option{-mno-ieee} is implicitly set, which results in faster
25827 floating-point greater-equal and less-equal comparisons. The implicit settings
25828 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
25830 @item -minline-ic_invalidate
25831 @opindex minline-ic_invalidate
25832 Inline code to invalidate instruction cache entries after setting up
25833 nested function trampolines.
25834 This option has no effect if @option{-musermode} is in effect and the selected
25835 code generation option (e.g.@: @option{-m4}) does not allow the use of the @code{icbi}
25837 If the selected code generation option does not allow the use of the @code{icbi}
25838 instruction, and @option{-musermode} is not in effect, the inlined code
25839 manipulates the instruction cache address array directly with an associative
25840 write. This not only requires privileged mode at run time, but it also
25841 fails if the cache line had been mapped via the TLB and has become unmapped.
25845 Dump instruction size and location in the assembly code.
25848 @opindex mpadstruct
25849 This option is deprecated. It pads structures to multiple of 4 bytes,
25850 which is incompatible with the SH ABI@.
25852 @item -matomic-model=@var{model}
25853 @opindex matomic-model=@var{model}
25854 Sets the model of atomic operations and additional parameters as a comma
25855 separated list. For details on the atomic built-in functions see
25856 @ref{__atomic Builtins}. The following models and parameters are supported:
25861 Disable compiler generated atomic sequences and emit library calls for atomic
25862 operations. This is the default if the target is not @code{sh*-*-linux*}.
25865 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
25866 built-in functions. The generated atomic sequences require additional support
25867 from the interrupt/exception handling code of the system and are only suitable
25868 for SH3* and SH4* single-core systems. This option is enabled by default when
25869 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
25870 this option also partially utilizes the hardware atomic instructions
25871 @code{movli.l} and @code{movco.l} to create more efficient code, unless
25872 @samp{strict} is specified.
25875 Generate software atomic sequences that use a variable in the thread control
25876 block. This is a variation of the gUSA sequences which can also be used on
25877 SH1* and SH2* targets. The generated atomic sequences require additional
25878 support from the interrupt/exception handling code of the system and are only
25879 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
25880 parameter has to be specified as well.
25883 Generate software atomic sequences that temporarily disable interrupts by
25884 setting @code{SR.IMASK = 1111}. This model works only when the program runs
25885 in privileged mode and is only suitable for single-core systems. Additional
25886 support from the interrupt/exception handling code of the system is not
25887 required. This model is enabled by default when the target is
25888 @code{sh*-*-linux*} and SH1* or SH2*.
25891 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
25892 instructions only. This is only available on SH4A and is suitable for
25893 multi-core systems. Since the hardware instructions support only 32 bit atomic
25894 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
25895 Code compiled with this option is also compatible with other software
25896 atomic model interrupt/exception handling systems if executed on an SH4A
25897 system. Additional support from the interrupt/exception handling code of the
25898 system is not required for this model.
25901 This parameter specifies the offset in bytes of the variable in the thread
25902 control block structure that should be used by the generated atomic sequences
25903 when the @samp{soft-tcb} model has been selected. For other models this
25904 parameter is ignored. The specified value must be an integer multiple of four
25905 and in the range 0-1020.
25908 This parameter prevents mixed usage of multiple atomic models, even if they
25909 are compatible, and makes the compiler generate atomic sequences of the
25910 specified model only.
25916 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
25917 Notice that depending on the particular hardware and software configuration
25918 this can degrade overall performance due to the operand cache line flushes
25919 that are implied by the @code{tas.b} instruction. On multi-core SH4A
25920 processors the @code{tas.b} instruction must be used with caution since it
25921 can result in data corruption for certain cache configurations.
25924 @opindex mprefergot
25925 When generating position-independent code, emit function calls using
25926 the Global Offset Table instead of the Procedure Linkage Table.
25929 @itemx -mno-usermode
25931 @opindex mno-usermode
25932 Don't allow (allow) the compiler generating privileged mode code. Specifying
25933 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
25934 inlined code would not work in user mode. @option{-musermode} is the default
25935 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
25936 @option{-musermode} has no effect, since there is no user mode.
25938 @item -multcost=@var{number}
25939 @opindex multcost=@var{number}
25940 Set the cost to assume for a multiply insn.
25942 @item -mdiv=@var{strategy}
25943 @opindex mdiv=@var{strategy}
25944 Set the division strategy to be used for integer division operations.
25945 @var{strategy} can be one of:
25950 Calls a library function that uses the single-step division instruction
25951 @code{div1} to perform the operation. Division by zero calculates an
25952 unspecified result and does not trap. This is the default except for SH4,
25953 SH2A and SHcompact.
25956 Calls a library function that performs the operation in double precision
25957 floating point. Division by zero causes a floating-point exception. This is
25958 the default for SHcompact with FPU. Specifying this for targets that do not
25959 have a double precision FPU defaults to @code{call-div1}.
25962 Calls a library function that uses a lookup table for small divisors and
25963 the @code{div1} instruction with case distinction for larger divisors. Division
25964 by zero calculates an unspecified result and does not trap. This is the default
25965 for SH4. Specifying this for targets that do not have dynamic shift
25966 instructions defaults to @code{call-div1}.
25970 When a division strategy has not been specified the default strategy is
25971 selected based on the current target. For SH2A the default strategy is to
25972 use the @code{divs} and @code{divu} instructions instead of library function
25975 @item -maccumulate-outgoing-args
25976 @opindex maccumulate-outgoing-args
25977 Reserve space once for outgoing arguments in the function prologue rather
25978 than around each call. Generally beneficial for performance and size. Also
25979 needed for unwinding to avoid changing the stack frame around conditional code.
25981 @item -mdivsi3_libfunc=@var{name}
25982 @opindex mdivsi3_libfunc=@var{name}
25983 Set the name of the library function used for 32-bit signed division to
25985 This only affects the name used in the @samp{call} division strategies, and
25986 the compiler still expects the same sets of input/output/clobbered registers as
25987 if this option were not present.
25989 @item -mfixed-range=@var{register-range}
25990 @opindex mfixed-range
25991 Generate code treating the given register range as fixed registers.
25992 A fixed register is one that the register allocator cannot use. This is
25993 useful when compiling kernel code. A register range is specified as
25994 two registers separated by a dash. Multiple register ranges can be
25995 specified separated by a comma.
25997 @item -mbranch-cost=@var{num}
25998 @opindex mbranch-cost=@var{num}
25999 Assume @var{num} to be the cost for a branch instruction. Higher numbers
26000 make the compiler try to generate more branch-free code if possible.
26001 If not specified the value is selected depending on the processor type that
26002 is being compiled for.
26005 @itemx -mno-zdcbranch
26006 @opindex mzdcbranch
26007 @opindex mno-zdcbranch
26008 Assume (do not assume) that zero displacement conditional branch instructions
26009 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
26010 compiler prefers zero displacement branch code sequences. This is
26011 enabled by default when generating code for SH4 and SH4A. It can be explicitly
26012 disabled by specifying @option{-mno-zdcbranch}.
26014 @item -mcbranch-force-delay-slot
26015 @opindex mcbranch-force-delay-slot
26016 Force the usage of delay slots for conditional branches, which stuffs the delay
26017 slot with a @code{nop} if a suitable instruction cannot be found. By default
26018 this option is disabled. It can be enabled to work around hardware bugs as
26019 found in the original SH7055.
26022 @itemx -mno-fused-madd
26023 @opindex mfused-madd
26024 @opindex mno-fused-madd
26025 Generate code that uses (does not use) the floating-point multiply and
26026 accumulate instructions. These instructions are generated by default
26027 if hardware floating point is used. The machine-dependent
26028 @option{-mfused-madd} option is now mapped to the machine-independent
26029 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
26030 mapped to @option{-ffp-contract=off}.
26036 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
26037 and cosine approximations. The option @option{-mfsca} must be used in
26038 combination with @option{-funsafe-math-optimizations}. It is enabled by default
26039 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
26040 approximations even if @option{-funsafe-math-optimizations} is in effect.
26046 Allow or disallow the compiler to emit the @code{fsrra} instruction for
26047 reciprocal square root approximations. The option @option{-mfsrra} must be used
26048 in combination with @option{-funsafe-math-optimizations} and
26049 @option{-ffinite-math-only}. It is enabled by default when generating code for
26050 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
26051 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
26054 @item -mpretend-cmove
26055 @opindex mpretend-cmove
26056 Prefer zero-displacement conditional branches for conditional move instruction
26057 patterns. This can result in faster code on the SH4 processor.
26061 Generate code using the FDPIC ABI.
26065 @node Solaris 2 Options
26066 @subsection Solaris 2 Options
26067 @cindex Solaris 2 options
26069 These @samp{-m} options are supported on Solaris 2:
26072 @item -mclear-hwcap
26073 @opindex mclear-hwcap
26074 @option{-mclear-hwcap} tells the compiler to remove the hardware
26075 capabilities generated by the Solaris assembler. This is only necessary
26076 when object files use ISA extensions not supported by the current
26077 machine, but check at runtime whether or not to use them.
26079 @item -mimpure-text
26080 @opindex mimpure-text
26081 @option{-mimpure-text}, used in addition to @option{-shared}, tells
26082 the compiler to not pass @option{-z text} to the linker when linking a
26083 shared object. Using this option, you can link position-dependent
26084 code into a shared object.
26086 @option{-mimpure-text} suppresses the ``relocations remain against
26087 allocatable but non-writable sections'' linker error message.
26088 However, the necessary relocations trigger copy-on-write, and the
26089 shared object is not actually shared across processes. Instead of
26090 using @option{-mimpure-text}, you should compile all source code with
26091 @option{-fpic} or @option{-fPIC}.
26095 These switches are supported in addition to the above on Solaris 2:
26100 This is a synonym for @option{-pthread}.
26103 @node SPARC Options
26104 @subsection SPARC Options
26105 @cindex SPARC options
26107 These @samp{-m} options are supported on the SPARC:
26110 @item -mno-app-regs
26112 @opindex mno-app-regs
26114 Specify @option{-mapp-regs} to generate output using the global registers
26115 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
26116 global register 1, each global register 2 through 4 is then treated as an
26117 allocable register that is clobbered by function calls. This is the default.
26119 To be fully SVR4 ABI-compliant at the cost of some performance loss,
26120 specify @option{-mno-app-regs}. You should compile libraries and system
26121 software with this option.
26127 With @option{-mflat}, the compiler does not generate save/restore instructions
26128 and uses a ``flat'' or single register window model. This model is compatible
26129 with the regular register window model. The local registers and the input
26130 registers (0--5) are still treated as ``call-saved'' registers and are
26131 saved on the stack as needed.
26133 With @option{-mno-flat} (the default), the compiler generates save/restore
26134 instructions (except for leaf functions). This is the normal operating mode.
26137 @itemx -mhard-float
26139 @opindex mhard-float
26140 Generate output containing floating-point instructions. This is the
26144 @itemx -msoft-float
26146 @opindex msoft-float
26147 Generate output containing library calls for floating point.
26148 @strong{Warning:} the requisite libraries are not available for all SPARC
26149 targets. Normally the facilities of the machine's usual C compiler are
26150 used, but this cannot be done directly in cross-compilation. You must make
26151 your own arrangements to provide suitable library functions for
26152 cross-compilation. The embedded targets @samp{sparc-*-aout} and
26153 @samp{sparclite-*-*} do provide software floating-point support.
26155 @option{-msoft-float} changes the calling convention in the output file;
26156 therefore, it is only useful if you compile @emph{all} of a program with
26157 this option. In particular, you need to compile @file{libgcc.a}, the
26158 library that comes with GCC, with @option{-msoft-float} in order for
26161 @item -mhard-quad-float
26162 @opindex mhard-quad-float
26163 Generate output containing quad-word (long double) floating-point
26166 @item -msoft-quad-float
26167 @opindex msoft-quad-float
26168 Generate output containing library calls for quad-word (long double)
26169 floating-point instructions. The functions called are those specified
26170 in the SPARC ABI@. This is the default.
26172 As of this writing, there are no SPARC implementations that have hardware
26173 support for the quad-word floating-point instructions. They all invoke
26174 a trap handler for one of these instructions, and then the trap handler
26175 emulates the effect of the instruction. Because of the trap handler overhead,
26176 this is much slower than calling the ABI library routines. Thus the
26177 @option{-msoft-quad-float} option is the default.
26179 @item -mno-unaligned-doubles
26180 @itemx -munaligned-doubles
26181 @opindex mno-unaligned-doubles
26182 @opindex munaligned-doubles
26183 Assume that doubles have 8-byte alignment. This is the default.
26185 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
26186 alignment only if they are contained in another type, or if they have an
26187 absolute address. Otherwise, it assumes they have 4-byte alignment.
26188 Specifying this option avoids some rare compatibility problems with code
26189 generated by other compilers. It is not the default because it results
26190 in a performance loss, especially for floating-point code.
26193 @itemx -mno-user-mode
26194 @opindex muser-mode
26195 @opindex mno-user-mode
26196 Do not generate code that can only run in supervisor mode. This is relevant
26197 only for the @code{casa} instruction emitted for the LEON3 processor. This
26200 @item -mfaster-structs
26201 @itemx -mno-faster-structs
26202 @opindex mfaster-structs
26203 @opindex mno-faster-structs
26204 With @option{-mfaster-structs}, the compiler assumes that structures
26205 should have 8-byte alignment. This enables the use of pairs of
26206 @code{ldd} and @code{std} instructions for copies in structure
26207 assignment, in place of twice as many @code{ld} and @code{st} pairs.
26208 However, the use of this changed alignment directly violates the SPARC
26209 ABI@. Thus, it's intended only for use on targets where the developer
26210 acknowledges that their resulting code is not directly in line with
26211 the rules of the ABI@.
26213 @item -mstd-struct-return
26214 @itemx -mno-std-struct-return
26215 @opindex mstd-struct-return
26216 @opindex mno-std-struct-return
26217 With @option{-mstd-struct-return}, the compiler generates checking code
26218 in functions returning structures or unions to detect size mismatches
26219 between the two sides of function calls, as per the 32-bit ABI@.
26221 The default is @option{-mno-std-struct-return}. This option has no effect
26228 Enable Local Register Allocation. This is the default for SPARC since GCC 7
26229 so @option{-mno-lra} needs to be passed to get old Reload.
26231 @item -mcpu=@var{cpu_type}
26233 Set the instruction set, register set, and instruction scheduling parameters
26234 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
26235 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
26236 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
26237 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
26238 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
26239 @samp{niagara3}, @samp{niagara4}, @samp{niagara7} and @samp{m8}.
26241 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
26242 which selects the best architecture option for the host processor.
26243 @option{-mcpu=native} has no effect if GCC does not recognize
26246 Default instruction scheduling parameters are used for values that select
26247 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
26248 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
26250 Here is a list of each supported architecture and their supported
26258 supersparc, hypersparc, leon, leon3
26261 f930, f934, sparclite86x
26267 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4,
26271 By default (unless configured otherwise), GCC generates code for the V7
26272 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
26273 additionally optimizes it for the Cypress CY7C602 chip, as used in the
26274 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
26275 SPARCStation 1, 2, IPX etc.
26277 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
26278 architecture. The only difference from V7 code is that the compiler emits
26279 the integer multiply and integer divide instructions which exist in SPARC-V8
26280 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
26281 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
26284 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
26285 the SPARC architecture. This adds the integer multiply, integer divide step
26286 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
26287 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
26288 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
26289 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
26290 MB86934 chip, which is the more recent SPARClite with FPU@.
26292 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
26293 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
26294 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
26295 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
26296 optimizes it for the TEMIC SPARClet chip.
26298 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
26299 architecture. This adds 64-bit integer and floating-point move instructions,
26300 3 additional floating-point condition code registers and conditional move
26301 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
26302 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
26303 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
26304 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
26305 @option{-mcpu=niagara}, the compiler additionally optimizes it for
26306 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
26307 additionally optimizes it for Sun UltraSPARC T2 chips. With
26308 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
26309 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
26310 additionally optimizes it for Sun UltraSPARC T4 chips. With
26311 @option{-mcpu=niagara7}, the compiler additionally optimizes it for
26312 Oracle SPARC M7 chips. With @option{-mcpu=m8}, the compiler
26313 additionally optimizes it for Oracle M8 chips.
26315 @item -mtune=@var{cpu_type}
26317 Set the instruction scheduling parameters for machine type
26318 @var{cpu_type}, but do not set the instruction set or register set that the
26319 option @option{-mcpu=@var{cpu_type}} does.
26321 The same values for @option{-mcpu=@var{cpu_type}} can be used for
26322 @option{-mtune=@var{cpu_type}}, but the only useful values are those
26323 that select a particular CPU implementation. Those are
26324 @samp{cypress}, @samp{supersparc}, @samp{hypersparc}, @samp{leon},
26325 @samp{leon3}, @samp{leon3v7}, @samp{f930}, @samp{f934},
26326 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
26327 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
26328 @samp{niagara4}, @samp{niagara7} and @samp{m8}. With native Solaris
26329 and GNU/Linux toolchains, @samp{native} can also be used.
26334 @opindex mno-v8plus
26335 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
26336 difference from the V8 ABI is that the global and out registers are
26337 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
26338 mode for all SPARC-V9 processors.
26344 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
26345 Visual Instruction Set extensions. The default is @option{-mno-vis}.
26351 With @option{-mvis2}, GCC generates code that takes advantage of
26352 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
26353 default is @option{-mvis2} when targeting a cpu that supports such
26354 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
26355 also sets @option{-mvis}.
26361 With @option{-mvis3}, GCC generates code that takes advantage of
26362 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
26363 default is @option{-mvis3} when targeting a cpu that supports such
26364 instructions, such as niagara-3 and later. Setting @option{-mvis3}
26365 also sets @option{-mvis2} and @option{-mvis}.
26371 With @option{-mvis4}, GCC generates code that takes advantage of
26372 version 4.0 of the UltraSPARC Visual Instruction Set extensions. The
26373 default is @option{-mvis4} when targeting a cpu that supports such
26374 instructions, such as niagara-7 and later. Setting @option{-mvis4}
26375 also sets @option{-mvis3}, @option{-mvis2} and @option{-mvis}.
26381 With @option{-mvis4b}, GCC generates code that takes advantage of
26382 version 4.0 of the UltraSPARC Visual Instruction Set extensions, plus
26383 the additional VIS instructions introduced in the Oracle SPARC
26384 Architecture 2017. The default is @option{-mvis4b} when targeting a
26385 cpu that supports such instructions, such as m8 and later. Setting
26386 @option{-mvis4b} also sets @option{-mvis4}, @option{-mvis3},
26387 @option{-mvis2} and @option{-mvis}.
26392 @opindex mno-cbcond
26393 With @option{-mcbcond}, GCC generates code that takes advantage of the UltraSPARC
26394 Compare-and-Branch-on-Condition instructions. The default is @option{-mcbcond}
26395 when targeting a CPU that supports such instructions, such as Niagara-4 and
26402 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
26403 Fused Multiply-Add Floating-point instructions. The default is @option{-mfmaf}
26404 when targeting a CPU that supports such instructions, such as Niagara-3 and
26410 @opindex mno-fsmuld
26411 With @option{-mfsmuld}, GCC generates code that takes advantage of the
26412 Floating-point Multiply Single to Double (FsMULd) instruction. The default is
26413 @option{-mfsmuld} when targeting a CPU supporting the architecture versions V8
26414 or V9 with FPU except @option{-mcpu=leon}.
26420 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
26421 Population Count instruction. The default is @option{-mpopc}
26422 when targeting a CPU that supports such an instruction, such as Niagara-2 and
26429 With @option{-msubxc}, GCC generates code that takes advantage of the UltraSPARC
26430 Subtract-Extended-with-Carry instruction. The default is @option{-msubxc}
26431 when targeting a CPU that supports such an instruction, such as Niagara-7 and
26435 @opindex mfix-at697f
26436 Enable the documented workaround for the single erratum of the Atmel AT697F
26437 processor (which corresponds to erratum #13 of the AT697E processor).
26440 @opindex mfix-ut699
26441 Enable the documented workarounds for the floating-point errata and the data
26442 cache nullify errata of the UT699 processor.
26445 @opindex mfix-ut700
26446 Enable the documented workaround for the back-to-back store errata of
26447 the UT699E/UT700 processor.
26449 @item -mfix-gr712rc
26450 @opindex mfix-gr712rc
26451 Enable the documented workaround for the back-to-back store errata of
26452 the GR712RC processor.
26455 These @samp{-m} options are supported in addition to the above
26456 on SPARC-V9 processors in 64-bit environments:
26463 Generate code for a 32-bit or 64-bit environment.
26464 The 32-bit environment sets int, long and pointer to 32 bits.
26465 The 64-bit environment sets int to 32 bits and long and pointer
26468 @item -mcmodel=@var{which}
26470 Set the code model to one of
26474 The Medium/Low code model: 64-bit addresses, programs
26475 must be linked in the low 32 bits of memory. Programs can be statically
26476 or dynamically linked.
26479 The Medium/Middle code model: 64-bit addresses, programs
26480 must be linked in the low 44 bits of memory, the text and data segments must
26481 be less than 2GB in size and the data segment must be located within 2GB of
26485 The Medium/Anywhere code model: 64-bit addresses, programs
26486 may be linked anywhere in memory, the text and data segments must be less
26487 than 2GB in size and the data segment must be located within 2GB of the
26491 The Medium/Anywhere code model for embedded systems:
26492 64-bit addresses, the text and data segments must be less than 2GB in
26493 size, both starting anywhere in memory (determined at link time). The
26494 global register %g4 points to the base of the data segment. Programs
26495 are statically linked and PIC is not supported.
26498 @item -mmemory-model=@var{mem-model}
26499 @opindex mmemory-model
26500 Set the memory model in force on the processor to one of
26504 The default memory model for the processor and operating system.
26507 Relaxed Memory Order
26510 Partial Store Order
26516 Sequential Consistency
26519 These memory models are formally defined in Appendix D of the SPARC-V9
26520 architecture manual, as set in the processor's @code{PSTATE.MM} field.
26523 @itemx -mno-stack-bias
26524 @opindex mstack-bias
26525 @opindex mno-stack-bias
26526 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
26527 frame pointer if present, are offset by @minus{}2047 which must be added back
26528 when making stack frame references. This is the default in 64-bit mode.
26529 Otherwise, assume no such offset is present.
26533 @subsection SPU Options
26534 @cindex SPU options
26536 These @samp{-m} options are supported on the SPU:
26540 @itemx -merror-reloc
26541 @opindex mwarn-reloc
26542 @opindex merror-reloc
26544 The loader for SPU does not handle dynamic relocations. By default, GCC
26545 gives an error when it generates code that requires a dynamic
26546 relocation. @option{-mno-error-reloc} disables the error,
26547 @option{-mwarn-reloc} generates a warning instead.
26550 @itemx -munsafe-dma
26552 @opindex munsafe-dma
26554 Instructions that initiate or test completion of DMA must not be
26555 reordered with respect to loads and stores of the memory that is being
26557 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
26558 memory accesses, but that can lead to inefficient code in places where the
26559 memory is known to not change. Rather than mark the memory as volatile,
26560 you can use @option{-msafe-dma} to tell the compiler to treat
26561 the DMA instructions as potentially affecting all memory.
26563 @item -mbranch-hints
26564 @opindex mbranch-hints
26566 By default, GCC generates a branch hint instruction to avoid
26567 pipeline stalls for always-taken or probably-taken branches. A hint
26568 is not generated closer than 8 instructions away from its branch.
26569 There is little reason to disable them, except for debugging purposes,
26570 or to make an object a little bit smaller.
26574 @opindex msmall-mem
26575 @opindex mlarge-mem
26577 By default, GCC generates code assuming that addresses are never larger
26578 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
26579 a full 32-bit address.
26584 By default, GCC links against startup code that assumes the SPU-style
26585 main function interface (which has an unconventional parameter list).
26586 With @option{-mstdmain}, GCC links your program against startup
26587 code that assumes a C99-style interface to @code{main}, including a
26588 local copy of @code{argv} strings.
26590 @item -mfixed-range=@var{register-range}
26591 @opindex mfixed-range
26592 Generate code treating the given register range as fixed registers.
26593 A fixed register is one that the register allocator cannot use. This is
26594 useful when compiling kernel code. A register range is specified as
26595 two registers separated by a dash. Multiple register ranges can be
26596 specified separated by a comma.
26602 Compile code assuming that pointers to the PPU address space accessed
26603 via the @code{__ea} named address space qualifier are either 32 or 64
26604 bits wide. The default is 32 bits. As this is an ABI-changing option,
26605 all object code in an executable must be compiled with the same setting.
26607 @item -maddress-space-conversion
26608 @itemx -mno-address-space-conversion
26609 @opindex maddress-space-conversion
26610 @opindex mno-address-space-conversion
26611 Allow/disallow treating the @code{__ea} address space as superset
26612 of the generic address space. This enables explicit type casts
26613 between @code{__ea} and generic pointer as well as implicit
26614 conversions of generic pointers to @code{__ea} pointers. The
26615 default is to allow address space pointer conversions.
26617 @item -mcache-size=@var{cache-size}
26618 @opindex mcache-size
26619 This option controls the version of libgcc that the compiler links to an
26620 executable and selects a software-managed cache for accessing variables
26621 in the @code{__ea} address space with a particular cache size. Possible
26622 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
26623 and @samp{128}. The default cache size is 64KB.
26625 @item -matomic-updates
26626 @itemx -mno-atomic-updates
26627 @opindex matomic-updates
26628 @opindex mno-atomic-updates
26629 This option controls the version of libgcc that the compiler links to an
26630 executable and selects whether atomic updates to the software-managed
26631 cache of PPU-side variables are used. If you use atomic updates, changes
26632 to a PPU variable from SPU code using the @code{__ea} named address space
26633 qualifier do not interfere with changes to other PPU variables residing
26634 in the same cache line from PPU code. If you do not use atomic updates,
26635 such interference may occur; however, writing back cache lines is
26636 more efficient. The default behavior is to use atomic updates.
26639 @itemx -mdual-nops=@var{n}
26640 @opindex mdual-nops
26641 By default, GCC inserts NOPs to increase dual issue when it expects
26642 it to increase performance. @var{n} can be a value from 0 to 10. A
26643 smaller @var{n} inserts fewer NOPs. 10 is the default, 0 is the
26644 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
26646 @item -mhint-max-nops=@var{n}
26647 @opindex mhint-max-nops
26648 Maximum number of NOPs to insert for a branch hint. A branch hint must
26649 be at least 8 instructions away from the branch it is affecting. GCC
26650 inserts up to @var{n} NOPs to enforce this, otherwise it does not
26651 generate the branch hint.
26653 @item -mhint-max-distance=@var{n}
26654 @opindex mhint-max-distance
26655 The encoding of the branch hint instruction limits the hint to be within
26656 256 instructions of the branch it is affecting. By default, GCC makes
26657 sure it is within 125.
26660 @opindex msafe-hints
26661 Work around a hardware bug that causes the SPU to stall indefinitely.
26662 By default, GCC inserts the @code{hbrp} instruction to make sure
26663 this stall won't happen.
26667 @node System V Options
26668 @subsection Options for System V
26670 These additional options are available on System V Release 4 for
26671 compatibility with other compilers on those systems:
26676 Create a shared object.
26677 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
26681 Identify the versions of each tool used by the compiler, in a
26682 @code{.ident} assembler directive in the output.
26686 Refrain from adding @code{.ident} directives to the output file (this is
26689 @item -YP,@var{dirs}
26691 Search the directories @var{dirs}, and no others, for libraries
26692 specified with @option{-l}.
26694 @item -Ym,@var{dir}
26696 Look in the directory @var{dir} to find the M4 preprocessor.
26697 The assembler uses this option.
26698 @c This is supposed to go with a -Yd for predefined M4 macro files, but
26699 @c the generic assembler that comes with Solaris takes just -Ym.
26702 @node TILE-Gx Options
26703 @subsection TILE-Gx Options
26704 @cindex TILE-Gx options
26706 These @samp{-m} options are supported on the TILE-Gx:
26709 @item -mcmodel=small
26710 @opindex mcmodel=small
26711 Generate code for the small model. The distance for direct calls is
26712 limited to 500M in either direction. PC-relative addresses are 32
26713 bits. Absolute addresses support the full address range.
26715 @item -mcmodel=large
26716 @opindex mcmodel=large
26717 Generate code for the large model. There is no limitation on call
26718 distance, pc-relative addresses, or absolute addresses.
26720 @item -mcpu=@var{name}
26722 Selects the type of CPU to be targeted. Currently the only supported
26723 type is @samp{tilegx}.
26729 Generate code for a 32-bit or 64-bit environment. The 32-bit
26730 environment sets int, long, and pointer to 32 bits. The 64-bit
26731 environment sets int to 32 bits and long and pointer to 64 bits.
26734 @itemx -mlittle-endian
26735 @opindex mbig-endian
26736 @opindex mlittle-endian
26737 Generate code in big/little endian mode, respectively.
26740 @node TILEPro Options
26741 @subsection TILEPro Options
26742 @cindex TILEPro options
26744 These @samp{-m} options are supported on the TILEPro:
26747 @item -mcpu=@var{name}
26749 Selects the type of CPU to be targeted. Currently the only supported
26750 type is @samp{tilepro}.
26754 Generate code for a 32-bit environment, which sets int, long, and
26755 pointer to 32 bits. This is the only supported behavior so the flag
26756 is essentially ignored.
26760 @subsection V850 Options
26761 @cindex V850 Options
26763 These @samp{-m} options are defined for V850 implementations:
26767 @itemx -mno-long-calls
26768 @opindex mlong-calls
26769 @opindex mno-long-calls
26770 Treat all calls as being far away (near). If calls are assumed to be
26771 far away, the compiler always loads the function's address into a
26772 register, and calls indirect through the pointer.
26778 Do not optimize (do optimize) basic blocks that use the same index
26779 pointer 4 or more times to copy pointer into the @code{ep} register, and
26780 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
26781 option is on by default if you optimize.
26783 @item -mno-prolog-function
26784 @itemx -mprolog-function
26785 @opindex mno-prolog-function
26786 @opindex mprolog-function
26787 Do not use (do use) external functions to save and restore registers
26788 at the prologue and epilogue of a function. The external functions
26789 are slower, but use less code space if more than one function saves
26790 the same number of registers. The @option{-mprolog-function} option
26791 is on by default if you optimize.
26795 Try to make the code as small as possible. At present, this just turns
26796 on the @option{-mep} and @option{-mprolog-function} options.
26798 @item -mtda=@var{n}
26800 Put static or global variables whose size is @var{n} bytes or less into
26801 the tiny data area that register @code{ep} points to. The tiny data
26802 area can hold up to 256 bytes in total (128 bytes for byte references).
26804 @item -msda=@var{n}
26806 Put static or global variables whose size is @var{n} bytes or less into
26807 the small data area that register @code{gp} points to. The small data
26808 area can hold up to 64 kilobytes.
26810 @item -mzda=@var{n}
26812 Put static or global variables whose size is @var{n} bytes or less into
26813 the first 32 kilobytes of memory.
26817 Specify that the target processor is the V850.
26821 Specify that the target processor is the V850E3V5. The preprocessor
26822 constant @code{__v850e3v5__} is defined if this option is used.
26826 Specify that the target processor is the V850E3V5. This is an alias for
26827 the @option{-mv850e3v5} option.
26831 Specify that the target processor is the V850E2V3. The preprocessor
26832 constant @code{__v850e2v3__} is defined if this option is used.
26836 Specify that the target processor is the V850E2. The preprocessor
26837 constant @code{__v850e2__} is defined if this option is used.
26841 Specify that the target processor is the V850E1. The preprocessor
26842 constants @code{__v850e1__} and @code{__v850e__} are defined if
26843 this option is used.
26847 Specify that the target processor is the V850ES. This is an alias for
26848 the @option{-mv850e1} option.
26852 Specify that the target processor is the V850E@. The preprocessor
26853 constant @code{__v850e__} is defined if this option is used.
26855 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
26856 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
26857 are defined then a default target processor is chosen and the
26858 relevant @samp{__v850*__} preprocessor constant is defined.
26860 The preprocessor constants @code{__v850} and @code{__v851__} are always
26861 defined, regardless of which processor variant is the target.
26863 @item -mdisable-callt
26864 @itemx -mno-disable-callt
26865 @opindex mdisable-callt
26866 @opindex mno-disable-callt
26867 This option suppresses generation of the @code{CALLT} instruction for the
26868 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
26871 This option is enabled by default when the RH850 ABI is
26872 in use (see @option{-mrh850-abi}), and disabled by default when the
26873 GCC ABI is in use. If @code{CALLT} instructions are being generated
26874 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
26880 Pass on (or do not pass on) the @option{-mrelax} command-line option
26884 @itemx -mno-long-jumps
26885 @opindex mlong-jumps
26886 @opindex mno-long-jumps
26887 Disable (or re-enable) the generation of PC-relative jump instructions.
26890 @itemx -mhard-float
26891 @opindex msoft-float
26892 @opindex mhard-float
26893 Disable (or re-enable) the generation of hardware floating point
26894 instructions. This option is only significant when the target
26895 architecture is @samp{V850E2V3} or higher. If hardware floating point
26896 instructions are being generated then the C preprocessor symbol
26897 @code{__FPU_OK__} is defined, otherwise the symbol
26898 @code{__NO_FPU__} is defined.
26902 Enables the use of the e3v5 LOOP instruction. The use of this
26903 instruction is not enabled by default when the e3v5 architecture is
26904 selected because its use is still experimental.
26908 @opindex mrh850-abi
26910 Enables support for the RH850 version of the V850 ABI. This is the
26911 default. With this version of the ABI the following rules apply:
26915 Integer sized structures and unions are returned via a memory pointer
26916 rather than a register.
26919 Large structures and unions (more than 8 bytes in size) are passed by
26923 Functions are aligned to 16-bit boundaries.
26926 The @option{-m8byte-align} command-line option is supported.
26929 The @option{-mdisable-callt} command-line option is enabled by
26930 default. The @option{-mno-disable-callt} command-line option is not
26934 When this version of the ABI is enabled the C preprocessor symbol
26935 @code{__V850_RH850_ABI__} is defined.
26939 Enables support for the old GCC version of the V850 ABI. With this
26940 version of the ABI the following rules apply:
26944 Integer sized structures and unions are returned in register @code{r10}.
26947 Large structures and unions (more than 8 bytes in size) are passed by
26951 Functions are aligned to 32-bit boundaries, unless optimizing for
26955 The @option{-m8byte-align} command-line option is not supported.
26958 The @option{-mdisable-callt} command-line option is supported but not
26959 enabled by default.
26962 When this version of the ABI is enabled the C preprocessor symbol
26963 @code{__V850_GCC_ABI__} is defined.
26965 @item -m8byte-align
26966 @itemx -mno-8byte-align
26967 @opindex m8byte-align
26968 @opindex mno-8byte-align
26969 Enables support for @code{double} and @code{long long} types to be
26970 aligned on 8-byte boundaries. The default is to restrict the
26971 alignment of all objects to at most 4-bytes. When
26972 @option{-m8byte-align} is in effect the C preprocessor symbol
26973 @code{__V850_8BYTE_ALIGN__} is defined.
26976 @opindex mbig-switch
26977 Generate code suitable for big switch tables. Use this option only if
26978 the assembler/linker complain about out of range branches within a switch
26983 This option causes r2 and r5 to be used in the code generated by
26984 the compiler. This setting is the default.
26986 @item -mno-app-regs
26987 @opindex mno-app-regs
26988 This option causes r2 and r5 to be treated as fixed registers.
26993 @subsection VAX Options
26994 @cindex VAX options
26996 These @samp{-m} options are defined for the VAX:
27001 Do not output certain jump instructions (@code{aobleq} and so on)
27002 that the Unix assembler for the VAX cannot handle across long
27007 Do output those jump instructions, on the assumption that the
27008 GNU assembler is being used.
27012 Output code for G-format floating-point numbers instead of D-format.
27015 @node Visium Options
27016 @subsection Visium Options
27017 @cindex Visium options
27023 A program which performs file I/O and is destined to run on an MCM target
27024 should be linked with this option. It causes the libraries libc.a and
27025 libdebug.a to be linked. The program should be run on the target under
27026 the control of the GDB remote debugging stub.
27030 A program which performs file I/O and is destined to run on the simulator
27031 should be linked with option. This causes libraries libc.a and libsim.a to
27035 @itemx -mhard-float
27037 @opindex mhard-float
27038 Generate code containing floating-point instructions. This is the
27042 @itemx -msoft-float
27044 @opindex msoft-float
27045 Generate code containing library calls for floating-point.
27047 @option{-msoft-float} changes the calling convention in the output file;
27048 therefore, it is only useful if you compile @emph{all} of a program with
27049 this option. In particular, you need to compile @file{libgcc.a}, the
27050 library that comes with GCC, with @option{-msoft-float} in order for
27053 @item -mcpu=@var{cpu_type}
27055 Set the instruction set, register set, and instruction scheduling parameters
27056 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
27057 @samp{mcm}, @samp{gr5} and @samp{gr6}.
27059 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
27061 By default (unless configured otherwise), GCC generates code for the GR5
27062 variant of the Visium architecture.
27064 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
27065 architecture. The only difference from GR5 code is that the compiler will
27066 generate block move instructions.
27068 @item -mtune=@var{cpu_type}
27070 Set the instruction scheduling parameters for machine type @var{cpu_type},
27071 but do not set the instruction set or register set that the option
27072 @option{-mcpu=@var{cpu_type}} would.
27076 Generate code for the supervisor mode, where there are no restrictions on
27077 the access to general registers. This is the default.
27080 @opindex muser-mode
27081 Generate code for the user mode, where the access to some general registers
27082 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
27083 mode; on the GR6, only registers r29 to r31 are affected.
27087 @subsection VMS Options
27089 These @samp{-m} options are defined for the VMS implementations:
27092 @item -mvms-return-codes
27093 @opindex mvms-return-codes
27094 Return VMS condition codes from @code{main}. The default is to return POSIX-style
27095 condition (e.g.@: error) codes.
27097 @item -mdebug-main=@var{prefix}
27098 @opindex mdebug-main=@var{prefix}
27099 Flag the first routine whose name starts with @var{prefix} as the main
27100 routine for the debugger.
27104 Default to 64-bit memory allocation routines.
27106 @item -mpointer-size=@var{size}
27107 @opindex mpointer-size=@var{size}
27108 Set the default size of pointers. Possible options for @var{size} are
27109 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
27110 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
27111 The later option disables @code{pragma pointer_size}.
27114 @node VxWorks Options
27115 @subsection VxWorks Options
27116 @cindex VxWorks Options
27118 The options in this section are defined for all VxWorks targets.
27119 Options specific to the target hardware are listed with the other
27120 options for that target.
27125 GCC can generate code for both VxWorks kernels and real time processes
27126 (RTPs). This option switches from the former to the latter. It also
27127 defines the preprocessor macro @code{__RTP__}.
27130 @opindex non-static
27131 Link an RTP executable against shared libraries rather than static
27132 libraries. The options @option{-static} and @option{-shared} can
27133 also be used for RTPs (@pxref{Link Options}); @option{-static}
27140 These options are passed down to the linker. They are defined for
27141 compatibility with Diab.
27144 @opindex Xbind-lazy
27145 Enable lazy binding of function calls. This option is equivalent to
27146 @option{-Wl,-z,now} and is defined for compatibility with Diab.
27150 Disable lazy binding of function calls. This option is the default and
27151 is defined for compatibility with Diab.
27155 @subsection x86 Options
27156 @cindex x86 Options
27158 These @samp{-m} options are defined for the x86 family of computers.
27162 @item -march=@var{cpu-type}
27164 Generate instructions for the machine type @var{cpu-type}. In contrast to
27165 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
27166 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
27167 to generate code that may not run at all on processors other than the one
27168 indicated. Specifying @option{-march=@var{cpu-type}} implies
27169 @option{-mtune=@var{cpu-type}}.
27171 The choices for @var{cpu-type} are:
27175 This selects the CPU to generate code for at compilation time by determining
27176 the processor type of the compiling machine. Using @option{-march=native}
27177 enables all instruction subsets supported by the local machine (hence
27178 the result might not run on different machines). Using @option{-mtune=native}
27179 produces code optimized for the local machine under the constraints
27180 of the selected instruction set.
27183 A generic CPU with 64-bit extensions.
27186 Original Intel i386 CPU@.
27189 Intel i486 CPU@. (No scheduling is implemented for this chip.)
27193 Intel Pentium CPU with no MMX support.
27196 Intel Lakemont MCU, based on Intel Pentium CPU.
27199 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
27202 Intel Pentium Pro CPU@.
27205 When used with @option{-march}, the Pentium Pro
27206 instruction set is used, so the code runs on all i686 family chips.
27207 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
27210 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
27215 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
27219 Intel Pentium M; low-power version of Intel Pentium III CPU
27220 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
27224 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
27227 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
27231 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
27232 SSE2 and SSE3 instruction set support.
27235 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
27236 instruction set support.
27239 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27240 SSE4.1, SSE4.2 and POPCNT instruction set support.
27243 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27244 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
27247 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27248 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
27251 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27252 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
27253 instruction set support.
27256 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27257 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27258 BMI, BMI2 and F16C instruction set support.
27261 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27262 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27263 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
27266 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27267 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27268 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
27269 XSAVES instruction set support.
27272 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
27273 instruction set support.
27276 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27277 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
27280 Intel Goldmont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27281 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL, RDRND, XSAVE, XSAVEOPT and FSGSBASE
27282 instruction set support.
27284 @item goldmont-plus
27285 Intel Goldmont Plus CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27286 SSSE3, SSE4.1, SSE4.2, POPCNT, AES, PCLMUL, RDRND, XSAVE, XSAVEOPT, FSGSBASE,
27287 PTWRITE, RDPID, SGX and UMIP instruction set support.
27290 Intel Tremont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27291 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL, RDRND, XSAVE, XSAVEOPT, FSGSBASE, PTWRITE,
27292 RDPID, SGX, UMIP, GFNI-SSE, CLWB and ENCLV instruction set support.
27295 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27296 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27297 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
27298 AVX512CD instruction set support.
27301 Intel Knights Mill CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27302 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27303 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER, AVX512CD,
27304 AVX5124VNNIW, AVX5124FMAPS and AVX512VPOPCNTDQ instruction set support.
27306 @item skylake-avx512
27307 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27308 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27309 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
27310 CLWB, AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
27313 Intel Cannonlake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
27314 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
27315 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
27316 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
27317 AVX512IFMA, SHA and UMIP instruction set support.
27319 @item icelake-client
27320 Intel Icelake Client CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
27321 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
27322 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
27323 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
27324 AVX512IFMA, SHA, CLWB, UMIP, RDPID, GFNI, AVX512VBMI2, AVX512VPOPCNTDQ,
27325 AVX512BITALG, AVX512VNNI, VPCLMULQDQ, VAES instruction set support.
27327 @item icelake-server
27328 Intel Icelake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
27329 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
27330 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
27331 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
27332 AVX512IFMA, SHA, CLWB, UMIP, RDPID, GFNI, AVX512VBMI2, AVX512VPOPCNTDQ,
27333 AVX512BITALG, AVX512VNNI, VPCLMULQDQ, VAES, PCONFIG and WBNOINVD instruction
27337 Intel Cascadelake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27338 SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI,
27339 BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F, CLWB,
27340 AVX512VL, AVX512BW, AVX512DQ, AVX512CD and AVX512VNNI instruction set support.
27343 AMD K6 CPU with MMX instruction set support.
27347 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
27350 @itemx athlon-tbird
27351 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
27357 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
27358 instruction set support.
27364 Processors based on the AMD K8 core with x86-64 instruction set support,
27365 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
27366 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
27367 instruction set extensions.)
27370 @itemx opteron-sse3
27371 @itemx athlon64-sse3
27372 Improved versions of AMD K8 cores with SSE3 instruction set support.
27376 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
27377 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
27378 instruction set extensions.)
27381 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
27382 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
27383 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
27385 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
27386 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
27387 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
27390 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
27391 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
27392 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
27393 64-bit instruction set extensions.
27395 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
27396 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
27397 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
27398 SSE4.2, ABM and 64-bit instruction set extensions.
27401 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
27402 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
27403 SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
27404 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
27405 instruction set extensions.
27407 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
27408 supersets BMI, BMI2, ,CLWB, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED,
27409 MWAITX, SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A,
27410 SSSE3, SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
27411 instruction set extensions.)
27415 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
27416 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
27417 instruction set extensions.)
27420 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
27421 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
27422 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
27425 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
27429 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
27430 instruction set support.
27433 VIA C3 CPU with MMX and 3DNow!@: instruction set support.
27434 (No scheduling is implemented for this chip.)
27437 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
27438 (No scheduling is implemented for this chip.)
27441 VIA C7 (Esther) CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
27442 (No scheduling is implemented for this chip.)
27445 VIA Eden Samuel 2 CPU with MMX and 3DNow!@: instruction set support.
27446 (No scheduling is implemented for this chip.)
27449 VIA Eden Nehemiah CPU with MMX and SSE instruction set support.
27450 (No scheduling is implemented for this chip.)
27453 VIA Eden Esther CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
27454 (No scheduling is implemented for this chip.)
27457 VIA Eden X2 CPU with x86-64, MMX, SSE, SSE2 and SSE3 instruction set support.
27458 (No scheduling is implemented for this chip.)
27461 VIA Eden X4 CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2,
27462 AVX and AVX2 instruction set support.
27463 (No scheduling is implemented for this chip.)
27466 Generic VIA Nano CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
27467 instruction set support.
27468 (No scheduling is implemented for this chip.)
27471 VIA Nano 1xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
27472 instruction set support.
27473 (No scheduling is implemented for this chip.)
27476 VIA Nano 2xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
27477 instruction set support.
27478 (No scheduling is implemented for this chip.)
27481 VIA Nano 3xxx CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
27482 instruction set support.
27483 (No scheduling is implemented for this chip.)
27486 VIA Nano Dual Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
27487 instruction set support.
27488 (No scheduling is implemented for this chip.)
27491 VIA Nano Quad Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
27492 instruction set support.
27493 (No scheduling is implemented for this chip.)
27496 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
27499 @item -mtune=@var{cpu-type}
27501 Tune to @var{cpu-type} everything applicable about the generated code, except
27502 for the ABI and the set of available instructions.
27503 While picking a specific @var{cpu-type} schedules things appropriately
27504 for that particular chip, the compiler does not generate any code that
27505 cannot run on the default machine type unless you use a
27506 @option{-march=@var{cpu-type}} option.
27507 For example, if GCC is configured for i686-pc-linux-gnu
27508 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
27509 but still runs on i686 machines.
27511 The choices for @var{cpu-type} are the same as for @option{-march}.
27512 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
27516 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
27517 If you know the CPU on which your code will run, then you should use
27518 the corresponding @option{-mtune} or @option{-march} option instead of
27519 @option{-mtune=generic}. But, if you do not know exactly what CPU users
27520 of your application will have, then you should use this option.
27522 As new processors are deployed in the marketplace, the behavior of this
27523 option will change. Therefore, if you upgrade to a newer version of
27524 GCC, code generation controlled by this option will change to reflect
27526 that are most common at the time that version of GCC is released.
27528 There is no @option{-march=generic} option because @option{-march}
27529 indicates the instruction set the compiler can use, and there is no
27530 generic instruction set applicable to all processors. In contrast,
27531 @option{-mtune} indicates the processor (or, in this case, collection of
27532 processors) for which the code is optimized.
27535 Produce code optimized for the most current Intel processors, which are
27536 Haswell and Silvermont for this version of GCC. If you know the CPU
27537 on which your code will run, then you should use the corresponding
27538 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
27539 But, if you want your application performs better on both Haswell and
27540 Silvermont, then you should use this option.
27542 As new Intel processors are deployed in the marketplace, the behavior of
27543 this option will change. Therefore, if you upgrade to a newer version of
27544 GCC, code generation controlled by this option will change to reflect
27545 the most current Intel processors at the time that version of GCC is
27548 There is no @option{-march=intel} option because @option{-march} indicates
27549 the instruction set the compiler can use, and there is no common
27550 instruction set applicable to all processors. In contrast,
27551 @option{-mtune} indicates the processor (or, in this case, collection of
27552 processors) for which the code is optimized.
27555 @item -mcpu=@var{cpu-type}
27557 A deprecated synonym for @option{-mtune}.
27559 @item -mfpmath=@var{unit}
27561 Generate floating-point arithmetic for selected unit @var{unit}. The choices
27562 for @var{unit} are:
27566 Use the standard 387 floating-point coprocessor present on the majority of chips and
27567 emulated otherwise. Code compiled with this option runs almost everywhere.
27568 The temporary results are computed in 80-bit precision instead of the precision
27569 specified by the type, resulting in slightly different results compared to most
27570 of other chips. See @option{-ffloat-store} for more detailed description.
27572 This is the default choice for non-Darwin x86-32 targets.
27575 Use scalar floating-point instructions present in the SSE instruction set.
27576 This instruction set is supported by Pentium III and newer chips,
27577 and in the AMD line
27578 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
27579 instruction set supports only single-precision arithmetic, thus the double and
27580 extended-precision arithmetic are still done using 387. A later version, present
27581 only in Pentium 4 and AMD x86-64 chips, supports double-precision
27584 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
27585 or @option{-msse2} switches to enable SSE extensions and make this option
27586 effective. For the x86-64 compiler, these extensions are enabled by default.
27588 The resulting code should be considerably faster in the majority of cases and avoid
27589 the numerical instability problems of 387 code, but may break some existing
27590 code that expects temporaries to be 80 bits.
27592 This is the default choice for the x86-64 compiler, Darwin x86-32 targets,
27593 and the default choice for x86-32 targets with the SSE2 instruction set
27594 when @option{-ffast-math} is enabled.
27599 Attempt to utilize both instruction sets at once. This effectively doubles the
27600 amount of available registers, and on chips with separate execution units for
27601 387 and SSE the execution resources too. Use this option with care, as it is
27602 still experimental, because the GCC register allocator does not model separate
27603 functional units well, resulting in unstable performance.
27606 @item -masm=@var{dialect}
27607 @opindex masm=@var{dialect}
27608 Output assembly instructions using selected @var{dialect}. Also affects
27609 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
27610 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
27611 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
27612 not support @samp{intel}.
27615 @itemx -mno-ieee-fp
27617 @opindex mno-ieee-fp
27618 Control whether or not the compiler uses IEEE floating-point
27619 comparisons. These correctly handle the case where the result of a
27620 comparison is unordered.
27623 @itemx -mhard-float
27625 @opindex mhard-float
27626 Generate output containing 80387 instructions for floating point.
27629 @itemx -msoft-float
27631 @opindex msoft-float
27632 Generate output containing library calls for floating point.
27634 @strong{Warning:} the requisite libraries are not part of GCC@.
27635 Normally the facilities of the machine's usual C compiler are used, but
27636 this cannot be done directly in cross-compilation. You must make your
27637 own arrangements to provide suitable library functions for
27640 On machines where a function returns floating-point results in the 80387
27641 register stack, some floating-point opcodes may be emitted even if
27642 @option{-msoft-float} is used.
27644 @item -mno-fp-ret-in-387
27645 @opindex mno-fp-ret-in-387
27646 @opindex mfp-ret-in-387
27647 Do not use the FPU registers for return values of functions.
27649 The usual calling convention has functions return values of types
27650 @code{float} and @code{double} in an FPU register, even if there
27651 is no FPU@. The idea is that the operating system should emulate
27654 The option @option{-mno-fp-ret-in-387} causes such values to be returned
27655 in ordinary CPU registers instead.
27657 @item -mno-fancy-math-387
27658 @opindex mno-fancy-math-387
27659 @opindex mfancy-math-387
27660 Some 387 emulators do not support the @code{sin}, @code{cos} and
27661 @code{sqrt} instructions for the 387. Specify this option to avoid
27662 generating those instructions.
27663 This option is overridden when @option{-march}
27664 indicates that the target CPU always has an FPU and so the
27665 instruction does not need emulation. These
27666 instructions are not generated unless you also use the
27667 @option{-funsafe-math-optimizations} switch.
27669 @item -malign-double
27670 @itemx -mno-align-double
27671 @opindex malign-double
27672 @opindex mno-align-double
27673 Control whether GCC aligns @code{double}, @code{long double}, and
27674 @code{long long} variables on a two-word boundary or a one-word
27675 boundary. Aligning @code{double} variables on a two-word boundary
27676 produces code that runs somewhat faster on a Pentium at the
27677 expense of more memory.
27679 On x86-64, @option{-malign-double} is enabled by default.
27681 @strong{Warning:} if you use the @option{-malign-double} switch,
27682 structures containing the above types are aligned differently than
27683 the published application binary interface specifications for the x86-32
27684 and are not binary compatible with structures in code compiled
27685 without that switch.
27687 @item -m96bit-long-double
27688 @itemx -m128bit-long-double
27689 @opindex m96bit-long-double
27690 @opindex m128bit-long-double
27691 These switches control the size of @code{long double} type. The x86-32
27692 application binary interface specifies the size to be 96 bits,
27693 so @option{-m96bit-long-double} is the default in 32-bit mode.
27695 Modern architectures (Pentium and newer) prefer @code{long double}
27696 to be aligned to an 8- or 16-byte boundary. In arrays or structures
27697 conforming to the ABI, this is not possible. So specifying
27698 @option{-m128bit-long-double} aligns @code{long double}
27699 to a 16-byte boundary by padding the @code{long double} with an additional
27702 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
27703 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
27705 Notice that neither of these options enable any extra precision over the x87
27706 standard of 80 bits for a @code{long double}.
27708 @strong{Warning:} if you override the default value for your target ABI, this
27709 changes the size of
27710 structures and arrays containing @code{long double} variables,
27711 as well as modifying the function calling convention for functions taking
27712 @code{long double}. Hence they are not binary-compatible
27713 with code compiled without that switch.
27715 @item -mlong-double-64
27716 @itemx -mlong-double-80
27717 @itemx -mlong-double-128
27718 @opindex mlong-double-64
27719 @opindex mlong-double-80
27720 @opindex mlong-double-128
27721 These switches control the size of @code{long double} type. A size
27722 of 64 bits makes the @code{long double} type equivalent to the @code{double}
27723 type. This is the default for 32-bit Bionic C library. A size
27724 of 128 bits makes the @code{long double} type equivalent to the
27725 @code{__float128} type. This is the default for 64-bit Bionic C library.
27727 @strong{Warning:} if you override the default value for your target ABI, this
27728 changes the size of
27729 structures and arrays containing @code{long double} variables,
27730 as well as modifying the function calling convention for functions taking
27731 @code{long double}. Hence they are not binary-compatible
27732 with code compiled without that switch.
27734 @item -malign-data=@var{type}
27735 @opindex malign-data
27736 Control how GCC aligns variables. Supported values for @var{type} are
27737 @samp{compat} uses increased alignment value compatible uses GCC 4.8
27738 and earlier, @samp{abi} uses alignment value as specified by the
27739 psABI, and @samp{cacheline} uses increased alignment value to match
27740 the cache line size. @samp{compat} is the default.
27742 @item -mlarge-data-threshold=@var{threshold}
27743 @opindex mlarge-data-threshold
27744 When @option{-mcmodel=medium} is specified, data objects larger than
27745 @var{threshold} are placed in the large data section. This value must be the
27746 same across all objects linked into the binary, and defaults to 65535.
27750 Use a different function-calling convention, in which functions that
27751 take a fixed number of arguments return with the @code{ret @var{num}}
27752 instruction, which pops their arguments while returning. This saves one
27753 instruction in the caller since there is no need to pop the arguments
27756 You can specify that an individual function is called with this calling
27757 sequence with the function attribute @code{stdcall}. You can also
27758 override the @option{-mrtd} option by using the function attribute
27759 @code{cdecl}. @xref{Function Attributes}.
27761 @strong{Warning:} this calling convention is incompatible with the one
27762 normally used on Unix, so you cannot use it if you need to call
27763 libraries compiled with the Unix compiler.
27765 Also, you must provide function prototypes for all functions that
27766 take variable numbers of arguments (including @code{printf});
27767 otherwise incorrect code is generated for calls to those
27770 In addition, seriously incorrect code results if you call a
27771 function with too many arguments. (Normally, extra arguments are
27772 harmlessly ignored.)
27774 @item -mregparm=@var{num}
27776 Control how many registers are used to pass integer arguments. By
27777 default, no registers are used to pass arguments, and at most 3
27778 registers can be used. You can control this behavior for a specific
27779 function by using the function attribute @code{regparm}.
27780 @xref{Function Attributes}.
27782 @strong{Warning:} if you use this switch, and
27783 @var{num} is nonzero, then you must build all modules with the same
27784 value, including any libraries. This includes the system libraries and
27788 @opindex msseregparm
27789 Use SSE register passing conventions for float and double arguments
27790 and return values. You can control this behavior for a specific
27791 function by using the function attribute @code{sseregparm}.
27792 @xref{Function Attributes}.
27794 @strong{Warning:} if you use this switch then you must build all
27795 modules with the same value, including any libraries. This includes
27796 the system libraries and startup modules.
27798 @item -mvect8-ret-in-mem
27799 @opindex mvect8-ret-in-mem
27800 Return 8-byte vectors in memory instead of MMX registers. This is the
27801 default on VxWorks to match the ABI of the Sun Studio compilers until
27802 version 12. @emph{Only} use this option if you need to remain
27803 compatible with existing code produced by those previous compiler
27804 versions or older versions of GCC@.
27813 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
27814 is specified, the significands of results of floating-point operations are
27815 rounded to 24 bits (single precision); @option{-mpc64} rounds the
27816 significands of results of floating-point operations to 53 bits (double
27817 precision) and @option{-mpc80} rounds the significands of results of
27818 floating-point operations to 64 bits (extended double precision), which is
27819 the default. When this option is used, floating-point operations in higher
27820 precisions are not available to the programmer without setting the FPU
27821 control word explicitly.
27823 Setting the rounding of floating-point operations to less than the default
27824 80 bits can speed some programs by 2% or more. Note that some mathematical
27825 libraries assume that extended-precision (80-bit) floating-point operations
27826 are enabled by default; routines in such libraries could suffer significant
27827 loss of accuracy, typically through so-called ``catastrophic cancellation'',
27828 when this option is used to set the precision to less than extended precision.
27830 @item -mstackrealign
27831 @opindex mstackrealign
27832 Realign the stack at entry. On the x86, the @option{-mstackrealign}
27833 option generates an alternate prologue and epilogue that realigns the
27834 run-time stack if necessary. This supports mixing legacy codes that keep
27835 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
27836 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
27837 applicable to individual functions.
27839 @item -mpreferred-stack-boundary=@var{num}
27840 @opindex mpreferred-stack-boundary
27841 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
27842 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
27843 the default is 4 (16 bytes or 128 bits).
27845 @strong{Warning:} When generating code for the x86-64 architecture with
27846 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
27847 used to keep the stack boundary aligned to 8 byte boundary. Since
27848 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
27849 intended to be used in controlled environment where stack space is
27850 important limitation. This option leads to wrong code when functions
27851 compiled with 16 byte stack alignment (such as functions from a standard
27852 library) are called with misaligned stack. In this case, SSE
27853 instructions may lead to misaligned memory access traps. In addition,
27854 variable arguments are handled incorrectly for 16 byte aligned
27855 objects (including x87 long double and __int128), leading to wrong
27856 results. You must build all modules with
27857 @option{-mpreferred-stack-boundary=3}, including any libraries. This
27858 includes the system libraries and startup modules.
27860 @item -mincoming-stack-boundary=@var{num}
27861 @opindex mincoming-stack-boundary
27862 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
27863 boundary. If @option{-mincoming-stack-boundary} is not specified,
27864 the one specified by @option{-mpreferred-stack-boundary} is used.
27866 On Pentium and Pentium Pro, @code{double} and @code{long double} values
27867 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
27868 suffer significant run time performance penalties. On Pentium III, the
27869 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
27870 properly if it is not 16-byte aligned.
27872 To ensure proper alignment of this values on the stack, the stack boundary
27873 must be as aligned as that required by any value stored on the stack.
27874 Further, every function must be generated such that it keeps the stack
27875 aligned. Thus calling a function compiled with a higher preferred
27876 stack boundary from a function compiled with a lower preferred stack
27877 boundary most likely misaligns the stack. It is recommended that
27878 libraries that use callbacks always use the default setting.
27880 This extra alignment does consume extra stack space, and generally
27881 increases code size. Code that is sensitive to stack space usage, such
27882 as embedded systems and operating system kernels, may want to reduce the
27883 preferred alignment to @option{-mpreferred-stack-boundary=2}.
27940 @itemx -mavx512ifma
27941 @opindex mavx512ifma
27943 @itemx -mavx512vbmi
27944 @opindex mavx512vbmi
27955 @itemx -mclflushopt
27956 @opindex mclflushopt
27991 @itemx -mprefetchwt1
27992 @opindex mprefetchwt1
28063 @itemx -mavx512vbmi2
28064 @opindex mavx512vbmi2
28066 @itemx -mavx512bf16
28067 @opindex mavx512bf16
28078 @itemx -mvpclmulqdq
28079 @opindex mvpclmulqdq
28081 @itemx -mavx512bitalg
28082 @opindex mavx512bitalg
28088 @opindex mmovdir64b
28093 @itemx -mavx512vpopcntdq
28094 @opindex mavx512vpopcntdq
28096 @itemx -mavx5124fmaps
28097 @opindex mavx5124fmaps
28099 @itemx -mavx512vnni
28100 @opindex mavx512vnni
28102 @itemx -mavx5124vnniw
28103 @opindex mavx5124vnniw
28107 These switches enable the use of instructions in the MMX, SSE,
28108 SSE2, SSE3, SSSE3, SSE4, SSE4A, SSE4.1, SSE4.2, AVX, AVX2, AVX512F, AVX512PF,
28109 AVX512ER, AVX512CD, AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA, AVX512VBMI, SHA,
28110 AES, PCLMUL, CLFLUSHOPT, CLWB, FSGSBASE, PTWRITE, RDRND, F16C, FMA, PCONFIG,
28111 WBNOINVD, FMA4, PREFETCHW, RDPID, PREFETCHWT1, RDSEED, SGX, XOP, LWP,
28112 3DNow!@:, enhanced 3DNow!@:, POPCNT, ABM, ADX, BMI, BMI2, LZCNT, FXSR, XSAVE,
28113 XSAVEOPT, XSAVEC, XSAVES, RTM, HLE, TBM, MWAITX, CLZERO, PKU, AVX512VBMI2,
28114 GFNI, VAES, WAITPKG, VPCLMULQDQ, AVX512BITALG, MOVDIRI, MOVDIR64B, AVX512BF16,
28115 ENQCMD, AVX512VPOPCNTDQ, AVX5124FMAPS, AVX512VNNI, AVX5124VNNIW, or CLDEMOTE
28116 extended instruction sets. Each has a corresponding @option{-mno-} option to
28117 disable use of these instructions.
28119 These extensions are also available as built-in functions: see
28120 @ref{x86 Built-in Functions}, for details of the functions enabled and
28121 disabled by these switches.
28123 To generate SSE/SSE2 instructions automatically from floating-point
28124 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
28126 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
28127 generates new AVX instructions or AVX equivalence for all SSEx instructions
28130 These options enable GCC to use these extended instructions in
28131 generated code, even without @option{-mfpmath=sse}. Applications that
28132 perform run-time CPU detection must compile separate files for each
28133 supported architecture, using the appropriate flags. In particular,
28134 the file containing the CPU detection code should be compiled without
28137 @item -mdump-tune-features
28138 @opindex mdump-tune-features
28139 This option instructs GCC to dump the names of the x86 performance
28140 tuning features and default settings. The names can be used in
28141 @option{-mtune-ctrl=@var{feature-list}}.
28143 @item -mtune-ctrl=@var{feature-list}
28144 @opindex mtune-ctrl=@var{feature-list}
28145 This option is used to do fine grain control of x86 code generation features.
28146 @var{feature-list} is a comma separated list of @var{feature} names. See also
28147 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
28148 on if it is not preceded with @samp{^}, otherwise, it is turned off.
28149 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
28150 developers. Using it may lead to code paths not covered by testing and can
28151 potentially result in compiler ICEs or runtime errors.
28154 @opindex mno-default
28155 This option instructs GCC to turn off all tunable features. See also
28156 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
28160 This option instructs GCC to emit a @code{cld} instruction in the prologue
28161 of functions that use string instructions. String instructions depend on
28162 the DF flag to select between autoincrement or autodecrement mode. While the
28163 ABI specifies the DF flag to be cleared on function entry, some operating
28164 systems violate this specification by not clearing the DF flag in their
28165 exception dispatchers. The exception handler can be invoked with the DF flag
28166 set, which leads to wrong direction mode when string instructions are used.
28167 This option can be enabled by default on 32-bit x86 targets by configuring
28168 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
28169 instructions can be suppressed with the @option{-mno-cld} compiler option
28173 @opindex mvzeroupper
28174 This option instructs GCC to emit a @code{vzeroupper} instruction
28175 before a transfer of control flow out of the function to minimize
28176 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
28179 @item -mprefer-avx128
28180 @opindex mprefer-avx128
28181 This option instructs GCC to use 128-bit AVX instructions instead of
28182 256-bit AVX instructions in the auto-vectorizer.
28184 @item -mprefer-vector-width=@var{opt}
28185 @opindex mprefer-vector-width
28186 This option instructs GCC to use @var{opt}-bit vector width in instructions
28187 instead of default on the selected platform.
28191 No extra limitations applied to GCC other than defined by the selected platform.
28194 Prefer 128-bit vector width for instructions.
28197 Prefer 256-bit vector width for instructions.
28200 Prefer 512-bit vector width for instructions.
28205 This option enables GCC to generate @code{CMPXCHG16B} instructions in 64-bit
28206 code to implement compare-and-exchange operations on 16-byte aligned 128-bit
28207 objects. This is useful for atomic updates of data structures exceeding one
28208 machine word in size. The compiler uses this instruction to implement
28209 @ref{__sync Builtins}. However, for @ref{__atomic Builtins} operating on
28210 128-bit integers, a library call is always used.
28214 This option enables generation of @code{SAHF} instructions in 64-bit code.
28215 Early Intel Pentium 4 CPUs with Intel 64 support,
28216 prior to the introduction of Pentium 4 G1 step in December 2005,
28217 lacked the @code{LAHF} and @code{SAHF} instructions
28218 which are supported by AMD64.
28219 These are load and store instructions, respectively, for certain status flags.
28220 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
28221 @code{drem}, and @code{remainder} built-in functions;
28222 see @ref{Other Builtins} for details.
28226 This option enables use of the @code{movbe} instruction to implement
28227 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
28231 The @option{-mshstk} option enables shadow stack built-in functions
28232 from x86 Control-flow Enforcement Technology (CET).
28236 This option enables built-in functions @code{__builtin_ia32_crc32qi},
28237 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
28238 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
28242 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
28243 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
28244 with an additional Newton-Raphson step
28245 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
28246 (and their vectorized
28247 variants) for single-precision floating-point arguments. These instructions
28248 are generated only when @option{-funsafe-math-optimizations} is enabled
28249 together with @option{-ffinite-math-only} and @option{-fno-trapping-math}.
28250 Note that while the throughput of the sequence is higher than the throughput
28251 of the non-reciprocal instruction, the precision of the sequence can be
28252 decreased by up to 2 ulp (i.e.@: the inverse of 1.0 equals 0.99999994).
28254 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
28255 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
28256 combination), and doesn't need @option{-mrecip}.
28258 Also note that GCC emits the above sequence with additional Newton-Raphson step
28259 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
28260 already with @option{-ffast-math} (or the above option combination), and
28261 doesn't need @option{-mrecip}.
28263 @item -mrecip=@var{opt}
28264 @opindex mrecip=opt
28265 This option controls which reciprocal estimate instructions
28266 may be used. @var{opt} is a comma-separated list of options, which may
28267 be preceded by a @samp{!} to invert the option:
28271 Enable all estimate instructions.
28274 Enable the default instructions, equivalent to @option{-mrecip}.
28277 Disable all estimate instructions, equivalent to @option{-mno-recip}.
28280 Enable the approximation for scalar division.
28283 Enable the approximation for vectorized division.
28286 Enable the approximation for scalar square root.
28289 Enable the approximation for vectorized square root.
28292 So, for example, @option{-mrecip=all,!sqrt} enables
28293 all of the reciprocal approximations, except for square root.
28295 @item -mveclibabi=@var{type}
28296 @opindex mveclibabi
28297 Specifies the ABI type to use for vectorizing intrinsics using an
28298 external library. Supported values for @var{type} are @samp{svml}
28299 for the Intel short
28300 vector math library and @samp{acml} for the AMD math core library.
28301 To use this option, both @option{-ftree-vectorize} and
28302 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
28303 ABI-compatible library must be specified at link time.
28305 GCC currently emits calls to @code{vmldExp2},
28306 @code{vmldLn2}, @code{vmldLog102}, @code{vmldPow2},
28307 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
28308 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
28309 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
28310 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4},
28311 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
28312 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
28313 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
28314 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
28315 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
28316 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
28317 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
28318 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
28319 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
28320 when @option{-mveclibabi=acml} is used.
28322 @item -mabi=@var{name}
28324 Generate code for the specified calling convention. Permissible values
28325 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
28326 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
28327 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
28328 You can control this behavior for specific functions by
28329 using the function attributes @code{ms_abi} and @code{sysv_abi}.
28330 @xref{Function Attributes}.
28332 @item -mforce-indirect-call
28333 @opindex mforce-indirect-call
28334 Force all calls to functions to be indirect. This is useful
28335 when using Intel Processor Trace where it generates more precise timing
28336 information for function calls.
28338 @item -mmanual-endbr
28339 @opindex mmanual-endbr
28340 Insert ENDBR instruction at function entry only via the @code{cf_check}
28341 function attribute. This is useful when used with the option
28342 @option{-fcf-protection=branch} to control ENDBR insertion at the
28345 @item -mcall-ms2sysv-xlogues
28346 @opindex mcall-ms2sysv-xlogues
28347 @opindex mno-call-ms2sysv-xlogues
28348 Due to differences in 64-bit ABIs, any Microsoft ABI function that calls a
28349 System V ABI function must consider RSI, RDI and XMM6-15 as clobbered. By
28350 default, the code for saving and restoring these registers is emitted inline,
28351 resulting in fairly lengthy prologues and epilogues. Using
28352 @option{-mcall-ms2sysv-xlogues} emits prologues and epilogues that
28353 use stubs in the static portion of libgcc to perform these saves and restores,
28354 thus reducing function size at the cost of a few extra instructions.
28356 @item -mtls-dialect=@var{type}
28357 @opindex mtls-dialect
28358 Generate code to access thread-local storage using the @samp{gnu} or
28359 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
28360 @samp{gnu2} is more efficient, but it may add compile- and run-time
28361 requirements that cannot be satisfied on all systems.
28364 @itemx -mno-push-args
28365 @opindex mpush-args
28366 @opindex mno-push-args
28367 Use PUSH operations to store outgoing parameters. This method is shorter
28368 and usually equally fast as method using SUB/MOV operations and is enabled
28369 by default. In some cases disabling it may improve performance because of
28370 improved scheduling and reduced dependencies.
28372 @item -maccumulate-outgoing-args
28373 @opindex maccumulate-outgoing-args
28374 If enabled, the maximum amount of space required for outgoing arguments is
28375 computed in the function prologue. This is faster on most modern CPUs
28376 because of reduced dependencies, improved scheduling and reduced stack usage
28377 when the preferred stack boundary is not equal to 2. The drawback is a notable
28378 increase in code size. This switch implies @option{-mno-push-args}.
28382 Support thread-safe exception handling on MinGW. Programs that rely
28383 on thread-safe exception handling must compile and link all code with the
28384 @option{-mthreads} option. When compiling, @option{-mthreads} defines
28385 @option{-D_MT}; when linking, it links in a special thread helper library
28386 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
28388 @item -mms-bitfields
28389 @itemx -mno-ms-bitfields
28390 @opindex mms-bitfields
28391 @opindex mno-ms-bitfields
28393 Enable/disable bit-field layout compatible with the native Microsoft
28396 If @code{packed} is used on a structure, or if bit-fields are used,
28397 it may be that the Microsoft ABI lays out the structure differently
28398 than the way GCC normally does. Particularly when moving packed
28399 data between functions compiled with GCC and the native Microsoft compiler
28400 (either via function call or as data in a file), it may be necessary to access
28403 This option is enabled by default for Microsoft Windows
28404 targets. This behavior can also be controlled locally by use of variable
28405 or type attributes. For more information, see @ref{x86 Variable Attributes}
28406 and @ref{x86 Type Attributes}.
28408 The Microsoft structure layout algorithm is fairly simple with the exception
28409 of the bit-field packing.
28410 The padding and alignment of members of structures and whether a bit-field
28411 can straddle a storage-unit boundary are determine by these rules:
28414 @item Structure members are stored sequentially in the order in which they are
28415 declared: the first member has the lowest memory address and the last member
28418 @item Every data object has an alignment requirement. The alignment requirement
28419 for all data except structures, unions, and arrays is either the size of the
28420 object or the current packing size (specified with either the
28421 @code{aligned} attribute or the @code{pack} pragma),
28422 whichever is less. For structures, unions, and arrays,
28423 the alignment requirement is the largest alignment requirement of its members.
28424 Every object is allocated an offset so that:
28427 offset % alignment_requirement == 0
28430 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
28431 unit if the integral types are the same size and if the next bit-field fits
28432 into the current allocation unit without crossing the boundary imposed by the
28433 common alignment requirements of the bit-fields.
28436 MSVC interprets zero-length bit-fields in the following ways:
28439 @item If a zero-length bit-field is inserted between two bit-fields that
28440 are normally coalesced, the bit-fields are not coalesced.
28447 unsigned long bf_1 : 12;
28449 unsigned long bf_2 : 12;
28454 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
28455 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
28457 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
28458 alignment of the zero-length bit-field is greater than the member that follows it,
28459 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
28480 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
28481 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
28482 bit-field does not affect the alignment of @code{bar} or, as a result, the size
28485 Taking this into account, it is important to note the following:
28488 @item If a zero-length bit-field follows a normal bit-field, the type of the
28489 zero-length bit-field may affect the alignment of the structure as whole. For
28490 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
28491 normal bit-field, and is of type short.
28493 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
28494 still affect the alignment of the structure:
28505 Here, @code{t4} takes up 4 bytes.
28508 @item Zero-length bit-fields following non-bit-field members are ignored:
28520 Here, @code{t5} takes up 2 bytes.
28524 @item -mno-align-stringops
28525 @opindex mno-align-stringops
28526 @opindex malign-stringops
28527 Do not align the destination of inlined string operations. This switch reduces
28528 code size and improves performance in case the destination is already aligned,
28529 but GCC doesn't know about it.
28531 @item -minline-all-stringops
28532 @opindex minline-all-stringops
28533 By default GCC inlines string operations only when the destination is
28534 known to be aligned to least a 4-byte boundary.
28535 This enables more inlining and increases code
28536 size, but may improve performance of code that depends on fast
28537 @code{memcpy} and @code{memset} for short lengths.
28538 The option enables inline expansion of @code{strlen} for all
28539 pointer alignments.
28541 @item -minline-stringops-dynamically
28542 @opindex minline-stringops-dynamically
28543 For string operations of unknown size, use run-time checks with
28544 inline code for small blocks and a library call for large blocks.
28546 @item -mstringop-strategy=@var{alg}
28547 @opindex mstringop-strategy=@var{alg}
28548 Override the internal decision heuristic for the particular algorithm to use
28549 for inlining string operations. The allowed values for @var{alg} are:
28555 Expand using i386 @code{rep} prefix of the specified size.
28559 @itemx unrolled_loop
28560 Expand into an inline loop.
28563 Always use a library call.
28566 @item -mmemcpy-strategy=@var{strategy}
28567 @opindex mmemcpy-strategy=@var{strategy}
28568 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
28569 should be inlined and what inline algorithm to use when the expected size
28570 of the copy operation is known. @var{strategy}
28571 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
28572 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
28573 the max byte size with which inline algorithm @var{alg} is allowed. For the last
28574 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
28575 in the list must be specified in increasing order. The minimal byte size for
28576 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
28579 @item -mmemset-strategy=@var{strategy}
28580 @opindex mmemset-strategy=@var{strategy}
28581 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
28582 @code{__builtin_memset} expansion.
28584 @item -momit-leaf-frame-pointer
28585 @opindex momit-leaf-frame-pointer
28586 Don't keep the frame pointer in a register for leaf functions. This
28587 avoids the instructions to save, set up, and restore frame pointers and
28588 makes an extra register available in leaf functions. The option
28589 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
28590 which might make debugging harder.
28592 @item -mtls-direct-seg-refs
28593 @itemx -mno-tls-direct-seg-refs
28594 @opindex mtls-direct-seg-refs
28595 Controls whether TLS variables may be accessed with offsets from the
28596 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
28597 or whether the thread base pointer must be added. Whether or not this
28598 is valid depends on the operating system, and whether it maps the
28599 segment to cover the entire TLS area.
28601 For systems that use the GNU C Library, the default is on.
28604 @itemx -mno-sse2avx
28606 Specify that the assembler should encode SSE instructions with VEX
28607 prefix. The option @option{-mavx} turns this on by default.
28612 If profiling is active (@option{-pg}), put the profiling
28613 counter call before the prologue.
28614 Note: On x86 architectures the attribute @code{ms_hook_prologue}
28615 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
28617 @item -mrecord-mcount
28618 @itemx -mno-record-mcount
28619 @opindex mrecord-mcount
28620 If profiling is active (@option{-pg}), generate a __mcount_loc section
28621 that contains pointers to each profiling call. This is useful for
28622 automatically patching and out calls.
28625 @itemx -mno-nop-mcount
28626 @opindex mnop-mcount
28627 If profiling is active (@option{-pg}), generate the calls to
28628 the profiling functions as NOPs. This is useful when they
28629 should be patched in later dynamically. This is likely only
28630 useful together with @option{-mrecord-mcount}.
28632 @item -minstrument-return=@var{type}
28633 @opindex minstrument-return
28634 Instrument function exit in -pg -mfentry instrumented functions with
28635 call to specified function. This only instruments true returns ending
28636 with ret, but not sibling calls ending with jump. Valid types
28637 are @var{none} to not instrument, @var{call} to generate a call to __return__,
28638 or @var{nop5} to generate a 5 byte nop.
28640 @item -mrecord-return
28641 @itemx -mno-record-return
28642 @opindex mrecord-return
28643 Generate a __return_loc section pointing to all return instrumentation code.
28645 @item -mfentry-name=@var{name}
28646 @opindex mfentry-name
28647 Set name of __fentry__ symbol called at function entry for -pg -mfentry functions.
28649 @item -mfentry-section=@var{name}
28650 @opindex mfentry-section
28651 Set name of section to record -mrecord-mcount calls (default __mcount_loc).
28653 @item -mskip-rax-setup
28654 @itemx -mno-skip-rax-setup
28655 @opindex mskip-rax-setup
28656 When generating code for the x86-64 architecture with SSE extensions
28657 disabled, @option{-mskip-rax-setup} can be used to skip setting up RAX
28658 register when there are no variable arguments passed in vector registers.
28660 @strong{Warning:} Since RAX register is used to avoid unnecessarily
28661 saving vector registers on stack when passing variable arguments, the
28662 impacts of this option are callees may waste some stack space,
28663 misbehave or jump to a random location. GCC 4.4 or newer don't have
28664 those issues, regardless the RAX register value.
28667 @itemx -mno-8bit-idiv
28668 @opindex m8bit-idiv
28669 On some processors, like Intel Atom, 8-bit unsigned integer divide is
28670 much faster than 32-bit/64-bit integer divide. This option generates a
28671 run-time check. If both dividend and divisor are within range of 0
28672 to 255, 8-bit unsigned integer divide is used instead of
28673 32-bit/64-bit integer divide.
28675 @item -mavx256-split-unaligned-load
28676 @itemx -mavx256-split-unaligned-store
28677 @opindex mavx256-split-unaligned-load
28678 @opindex mavx256-split-unaligned-store
28679 Split 32-byte AVX unaligned load and store.
28681 @item -mstack-protector-guard=@var{guard}
28682 @itemx -mstack-protector-guard-reg=@var{reg}
28683 @itemx -mstack-protector-guard-offset=@var{offset}
28684 @opindex mstack-protector-guard
28685 @opindex mstack-protector-guard-reg
28686 @opindex mstack-protector-guard-offset
28687 Generate stack protection code using canary at @var{guard}. Supported
28688 locations are @samp{global} for global canary or @samp{tls} for per-thread
28689 canary in the TLS block (the default). This option has effect only when
28690 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
28692 With the latter choice the options
28693 @option{-mstack-protector-guard-reg=@var{reg}} and
28694 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
28695 which segment register (@code{%fs} or @code{%gs}) to use as base register
28696 for reading the canary, and from what offset from that base register.
28697 The default for those is as specified in the relevant ABI.
28699 @item -mgeneral-regs-only
28700 @opindex mgeneral-regs-only
28701 Generate code that uses only the general-purpose registers. This
28702 prevents the compiler from using floating-point, vector, mask and bound
28705 @item -mindirect-branch=@var{choice}
28706 @opindex mindirect-branch
28707 Convert indirect call and jump with @var{choice}. The default is
28708 @samp{keep}, which keeps indirect call and jump unmodified.
28709 @samp{thunk} converts indirect call and jump to call and return thunk.
28710 @samp{thunk-inline} converts indirect call and jump to inlined call
28711 and return thunk. @samp{thunk-extern} converts indirect call and jump
28712 to external call and return thunk provided in a separate object file.
28713 You can control this behavior for a specific function by using the
28714 function attribute @code{indirect_branch}. @xref{Function Attributes}.
28716 Note that @option{-mcmodel=large} is incompatible with
28717 @option{-mindirect-branch=thunk} and
28718 @option{-mindirect-branch=thunk-extern} since the thunk function may
28719 not be reachable in the large code model.
28721 Note that @option{-mindirect-branch=thunk-extern} is incompatible with
28722 @option{-fcf-protection=branch} since the external thunk cannot be modified
28723 to disable control-flow check.
28725 @item -mfunction-return=@var{choice}
28726 @opindex mfunction-return
28727 Convert function return with @var{choice}. The default is @samp{keep},
28728 which keeps function return unmodified. @samp{thunk} converts function
28729 return to call and return thunk. @samp{thunk-inline} converts function
28730 return to inlined call and return thunk. @samp{thunk-extern} converts
28731 function return to external call and return thunk provided in a separate
28732 object file. You can control this behavior for a specific function by
28733 using the function attribute @code{function_return}.
28734 @xref{Function Attributes}.
28736 Note that @option{-mcmodel=large} is incompatible with
28737 @option{-mfunction-return=thunk} and
28738 @option{-mfunction-return=thunk-extern} since the thunk function may
28739 not be reachable in the large code model.
28742 @item -mindirect-branch-register
28743 @opindex mindirect-branch-register
28744 Force indirect call and jump via register.
28748 These @samp{-m} switches are supported in addition to the above
28749 on x86-64 processors in 64-bit environments.
28762 Generate code for a 16-bit, 32-bit or 64-bit environment.
28763 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
28765 generates code that runs on any i386 system.
28767 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
28768 types to 64 bits, and generates code for the x86-64 architecture.
28769 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
28770 and @option{-mdynamic-no-pic} options.
28772 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
28774 generates code for the x86-64 architecture.
28776 The @option{-m16} option is the same as @option{-m32}, except for that
28777 it outputs the @code{.code16gcc} assembly directive at the beginning of
28778 the assembly output so that the binary can run in 16-bit mode.
28780 The @option{-miamcu} option generates code which conforms to Intel MCU
28781 psABI. It requires the @option{-m32} option to be turned on.
28783 @item -mno-red-zone
28784 @opindex mno-red-zone
28786 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
28787 by the x86-64 ABI; it is a 128-byte area beyond the location of the
28788 stack pointer that is not modified by signal or interrupt handlers
28789 and therefore can be used for temporary data without adjusting the stack
28790 pointer. The flag @option{-mno-red-zone} disables this red zone.
28792 @item -mcmodel=small
28793 @opindex mcmodel=small
28794 Generate code for the small code model: the program and its symbols must
28795 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
28796 Programs can be statically or dynamically linked. This is the default
28799 @item -mcmodel=kernel
28800 @opindex mcmodel=kernel
28801 Generate code for the kernel code model. The kernel runs in the
28802 negative 2 GB of the address space.
28803 This model has to be used for Linux kernel code.
28805 @item -mcmodel=medium
28806 @opindex mcmodel=medium
28807 Generate code for the medium model: the program is linked in the lower 2
28808 GB of the address space. Small symbols are also placed there. Symbols
28809 with sizes larger than @option{-mlarge-data-threshold} are put into
28810 large data or BSS sections and can be located above 2GB. Programs can
28811 be statically or dynamically linked.
28813 @item -mcmodel=large
28814 @opindex mcmodel=large
28815 Generate code for the large model. This model makes no assumptions
28816 about addresses and sizes of sections.
28818 @item -maddress-mode=long
28819 @opindex maddress-mode=long
28820 Generate code for long address mode. This is only supported for 64-bit
28821 and x32 environments. It is the default address mode for 64-bit
28824 @item -maddress-mode=short
28825 @opindex maddress-mode=short
28826 Generate code for short address mode. This is only supported for 32-bit
28827 and x32 environments. It is the default address mode for 32-bit and
28831 @node x86 Windows Options
28832 @subsection x86 Windows Options
28833 @cindex x86 Windows Options
28834 @cindex Windows Options for x86
28836 These additional options are available for Microsoft Windows targets:
28842 specifies that a console application is to be generated, by
28843 instructing the linker to set the PE header subsystem type
28844 required for console applications.
28845 This option is available for Cygwin and MinGW targets and is
28846 enabled by default on those targets.
28850 This option is available for Cygwin and MinGW targets. It
28851 specifies that a DLL---a dynamic link library---is to be
28852 generated, enabling the selection of the required runtime
28853 startup object and entry point.
28855 @item -mnop-fun-dllimport
28856 @opindex mnop-fun-dllimport
28857 This option is available for Cygwin and MinGW targets. It
28858 specifies that the @code{dllimport} attribute should be ignored.
28862 This option is available for MinGW targets. It specifies
28863 that MinGW-specific thread support is to be used.
28867 This option is available for MinGW-w64 targets. It causes
28868 the @code{UNICODE} preprocessor macro to be predefined, and
28869 chooses Unicode-capable runtime startup code.
28873 This option is available for Cygwin and MinGW targets. It
28874 specifies that the typical Microsoft Windows predefined macros are to
28875 be set in the pre-processor, but does not influence the choice
28876 of runtime library/startup code.
28880 This option is available for Cygwin and MinGW targets. It
28881 specifies that a GUI application is to be generated by
28882 instructing the linker to set the PE header subsystem type
28885 @item -fno-set-stack-executable
28886 @opindex fno-set-stack-executable
28887 @opindex fset-stack-executable
28888 This option is available for MinGW targets. It specifies that
28889 the executable flag for the stack used by nested functions isn't
28890 set. This is necessary for binaries running in kernel mode of
28891 Microsoft Windows, as there the User32 API, which is used to set executable
28892 privileges, isn't available.
28894 @item -fwritable-relocated-rdata
28895 @opindex fno-writable-relocated-rdata
28896 @opindex fwritable-relocated-rdata
28897 This option is available for MinGW and Cygwin targets. It specifies
28898 that relocated-data in read-only section is put into the @code{.data}
28899 section. This is a necessary for older runtimes not supporting
28900 modification of @code{.rdata} sections for pseudo-relocation.
28902 @item -mpe-aligned-commons
28903 @opindex mpe-aligned-commons
28904 This option is available for Cygwin and MinGW targets. It
28905 specifies that the GNU extension to the PE file format that
28906 permits the correct alignment of COMMON variables should be
28907 used when generating code. It is enabled by default if
28908 GCC detects that the target assembler found during configuration
28909 supports the feature.
28912 See also under @ref{x86 Options} for standard options.
28914 @node Xstormy16 Options
28915 @subsection Xstormy16 Options
28916 @cindex Xstormy16 Options
28918 These options are defined for Xstormy16:
28923 Choose startup files and linker script suitable for the simulator.
28926 @node Xtensa Options
28927 @subsection Xtensa Options
28928 @cindex Xtensa Options
28930 These options are supported for Xtensa targets:
28934 @itemx -mno-const16
28936 @opindex mno-const16
28937 Enable or disable use of @code{CONST16} instructions for loading
28938 constant values. The @code{CONST16} instruction is currently not a
28939 standard option from Tensilica. When enabled, @code{CONST16}
28940 instructions are always used in place of the standard @code{L32R}
28941 instructions. The use of @code{CONST16} is enabled by default only if
28942 the @code{L32R} instruction is not available.
28945 @itemx -mno-fused-madd
28946 @opindex mfused-madd
28947 @opindex mno-fused-madd
28948 Enable or disable use of fused multiply/add and multiply/subtract
28949 instructions in the floating-point option. This has no effect if the
28950 floating-point option is not also enabled. Disabling fused multiply/add
28951 and multiply/subtract instructions forces the compiler to use separate
28952 instructions for the multiply and add/subtract operations. This may be
28953 desirable in some cases where strict IEEE 754-compliant results are
28954 required: the fused multiply add/subtract instructions do not round the
28955 intermediate result, thereby producing results with @emph{more} bits of
28956 precision than specified by the IEEE standard. Disabling fused multiply
28957 add/subtract instructions also ensures that the program output is not
28958 sensitive to the compiler's ability to combine multiply and add/subtract
28961 @item -mserialize-volatile
28962 @itemx -mno-serialize-volatile
28963 @opindex mserialize-volatile
28964 @opindex mno-serialize-volatile
28965 When this option is enabled, GCC inserts @code{MEMW} instructions before
28966 @code{volatile} memory references to guarantee sequential consistency.
28967 The default is @option{-mserialize-volatile}. Use
28968 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
28970 @item -mforce-no-pic
28971 @opindex mforce-no-pic
28972 For targets, like GNU/Linux, where all user-mode Xtensa code must be
28973 position-independent code (PIC), this option disables PIC for compiling
28976 @item -mtext-section-literals
28977 @itemx -mno-text-section-literals
28978 @opindex mtext-section-literals
28979 @opindex mno-text-section-literals
28980 These options control the treatment of literal pools. The default is
28981 @option{-mno-text-section-literals}, which places literals in a separate
28982 section in the output file. This allows the literal pool to be placed
28983 in a data RAM/ROM, and it also allows the linker to combine literal
28984 pools from separate object files to remove redundant literals and
28985 improve code size. With @option{-mtext-section-literals}, the literals
28986 are interspersed in the text section in order to keep them as close as
28987 possible to their references. This may be necessary for large assembly
28988 files. Literals for each function are placed right before that function.
28990 @item -mauto-litpools
28991 @itemx -mno-auto-litpools
28992 @opindex mauto-litpools
28993 @opindex mno-auto-litpools
28994 These options control the treatment of literal pools. The default is
28995 @option{-mno-auto-litpools}, which places literals in a separate
28996 section in the output file unless @option{-mtext-section-literals} is
28997 used. With @option{-mauto-litpools} the literals are interspersed in
28998 the text section by the assembler. Compiler does not produce explicit
28999 @code{.literal} directives and loads literals into registers with
29000 @code{MOVI} instructions instead of @code{L32R} to let the assembler
29001 do relaxation and place literals as necessary. This option allows
29002 assembler to create several literal pools per function and assemble
29003 very big functions, which may not be possible with
29004 @option{-mtext-section-literals}.
29006 @item -mtarget-align
29007 @itemx -mno-target-align
29008 @opindex mtarget-align
29009 @opindex mno-target-align
29010 When this option is enabled, GCC instructs the assembler to
29011 automatically align instructions to reduce branch penalties at the
29012 expense of some code density. The assembler attempts to widen density
29013 instructions to align branch targets and the instructions following call
29014 instructions. If there are not enough preceding safe density
29015 instructions to align a target, no widening is performed. The
29016 default is @option{-mtarget-align}. These options do not affect the
29017 treatment of auto-aligned instructions like @code{LOOP}, which the
29018 assembler always aligns, either by widening density instructions or
29019 by inserting NOP instructions.
29022 @itemx -mno-longcalls
29023 @opindex mlongcalls
29024 @opindex mno-longcalls
29025 When this option is enabled, GCC instructs the assembler to translate
29026 direct calls to indirect calls unless it can determine that the target
29027 of a direct call is in the range allowed by the call instruction. This
29028 translation typically occurs for calls to functions in other source
29029 files. Specifically, the assembler translates a direct @code{CALL}
29030 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
29031 The default is @option{-mno-longcalls}. This option should be used in
29032 programs where the call target can potentially be out of range. This
29033 option is implemented in the assembler, not the compiler, so the
29034 assembly code generated by GCC still shows direct call
29035 instructions---look at the disassembled object code to see the actual
29036 instructions. Note that the assembler uses an indirect call for
29037 every cross-file call, not just those that really are out of range.
29040 @node zSeries Options
29041 @subsection zSeries Options
29042 @cindex zSeries options
29044 These are listed under @xref{S/390 and zSeries Options}.
29050 @section Specifying Subprocesses and the Switches to Pass to Them
29053 @command{gcc} is a driver program. It performs its job by invoking a
29054 sequence of other programs to do the work of compiling, assembling and
29055 linking. GCC interprets its command-line parameters and uses these to
29056 deduce which programs it should invoke, and which command-line options
29057 it ought to place on their command lines. This behavior is controlled
29058 by @dfn{spec strings}. In most cases there is one spec string for each
29059 program that GCC can invoke, but a few programs have multiple spec
29060 strings to control their behavior. The spec strings built into GCC can
29061 be overridden by using the @option{-specs=} command-line switch to specify
29064 @dfn{Spec files} are plain-text files that are used to construct spec
29065 strings. They consist of a sequence of directives separated by blank
29066 lines. The type of directive is determined by the first non-whitespace
29067 character on the line, which can be one of the following:
29070 @item %@var{command}
29071 Issues a @var{command} to the spec file processor. The commands that can
29075 @item %include <@var{file}>
29076 @cindex @code{%include}
29077 Search for @var{file} and insert its text at the current point in the
29080 @item %include_noerr <@var{file}>
29081 @cindex @code{%include_noerr}
29082 Just like @samp{%include}, but do not generate an error message if the include
29083 file cannot be found.
29085 @item %rename @var{old_name} @var{new_name}
29086 @cindex @code{%rename}
29087 Rename the spec string @var{old_name} to @var{new_name}.
29091 @item *[@var{spec_name}]:
29092 This tells the compiler to create, override or delete the named spec
29093 string. All lines after this directive up to the next directive or
29094 blank line are considered to be the text for the spec string. If this
29095 results in an empty string then the spec is deleted. (Or, if the
29096 spec did not exist, then nothing happens.) Otherwise, if the spec
29097 does not currently exist a new spec is created. If the spec does
29098 exist then its contents are overridden by the text of this
29099 directive, unless the first character of that text is the @samp{+}
29100 character, in which case the text is appended to the spec.
29102 @item [@var{suffix}]:
29103 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
29104 and up to the next directive or blank line are considered to make up the
29105 spec string for the indicated suffix. When the compiler encounters an
29106 input file with the named suffix, it processes the spec string in
29107 order to work out how to compile that file. For example:
29111 z-compile -input %i
29114 This says that any input file whose name ends in @samp{.ZZ} should be
29115 passed to the program @samp{z-compile}, which should be invoked with the
29116 command-line switch @option{-input} and with the result of performing the
29117 @samp{%i} substitution. (See below.)
29119 As an alternative to providing a spec string, the text following a
29120 suffix directive can be one of the following:
29123 @item @@@var{language}
29124 This says that the suffix is an alias for a known @var{language}. This is
29125 similar to using the @option{-x} command-line switch to GCC to specify a
29126 language explicitly. For example:
29133 Says that .ZZ files are, in fact, C++ source files.
29136 This causes an error messages saying:
29139 @var{name} compiler not installed on this system.
29143 GCC already has an extensive list of suffixes built into it.
29144 This directive adds an entry to the end of the list of suffixes, but
29145 since the list is searched from the end backwards, it is effectively
29146 possible to override earlier entries using this technique.
29150 GCC has the following spec strings built into it. Spec files can
29151 override these strings or create their own. Note that individual
29152 targets can also add their own spec strings to this list.
29155 asm Options to pass to the assembler
29156 asm_final Options to pass to the assembler post-processor
29157 cpp Options to pass to the C preprocessor
29158 cc1 Options to pass to the C compiler
29159 cc1plus Options to pass to the C++ compiler
29160 endfile Object files to include at the end of the link
29161 link Options to pass to the linker
29162 lib Libraries to include on the command line to the linker
29163 libgcc Decides which GCC support library to pass to the linker
29164 linker Sets the name of the linker
29165 predefines Defines to be passed to the C preprocessor
29166 signed_char Defines to pass to CPP to say whether @code{char} is signed
29168 startfile Object files to include at the start of the link
29171 Here is a small example of a spec file:
29174 %rename lib old_lib
29177 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
29180 This example renames the spec called @samp{lib} to @samp{old_lib} and
29181 then overrides the previous definition of @samp{lib} with a new one.
29182 The new definition adds in some extra command-line options before
29183 including the text of the old definition.
29185 @dfn{Spec strings} are a list of command-line options to be passed to their
29186 corresponding program. In addition, the spec strings can contain
29187 @samp{%}-prefixed sequences to substitute variable text or to
29188 conditionally insert text into the command line. Using these constructs
29189 it is possible to generate quite complex command lines.
29191 Here is a table of all defined @samp{%}-sequences for spec
29192 strings. Note that spaces are not generated automatically around the
29193 results of expanding these sequences. Therefore you can concatenate them
29194 together or combine them with constant text in a single argument.
29198 Substitute one @samp{%} into the program name or argument.
29201 Substitute the name of the input file being processed.
29204 Substitute the basename of the input file being processed.
29205 This is the substring up to (and not including) the last period
29206 and not including the directory.
29209 This is the same as @samp{%b}, but include the file suffix (text after
29213 Marks the argument containing or following the @samp{%d} as a
29214 temporary file name, so that that file is deleted if GCC exits
29215 successfully. Unlike @samp{%g}, this contributes no text to the
29218 @item %g@var{suffix}
29219 Substitute a file name that has suffix @var{suffix} and is chosen
29220 once per compilation, and mark the argument in the same way as
29221 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
29222 name is now chosen in a way that is hard to predict even when previously
29223 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
29224 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
29225 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
29226 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
29227 was simply substituted with a file name chosen once per compilation,
29228 without regard to any appended suffix (which was therefore treated
29229 just like ordinary text), making such attacks more likely to succeed.
29231 @item %u@var{suffix}
29232 Like @samp{%g}, but generates a new temporary file name
29233 each time it appears instead of once per compilation.
29235 @item %U@var{suffix}
29236 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
29237 new one if there is no such last file name. In the absence of any
29238 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
29239 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
29240 involves the generation of two distinct file names, one
29241 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
29242 simply substituted with a file name chosen for the previous @samp{%u},
29243 without regard to any appended suffix.
29245 @item %j@var{suffix}
29246 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
29247 writable, and if @option{-save-temps} is not used;
29248 otherwise, substitute the name
29249 of a temporary file, just like @samp{%u}. This temporary file is not
29250 meant for communication between processes, but rather as a junk
29251 disposal mechanism.
29253 @item %|@var{suffix}
29254 @itemx %m@var{suffix}
29255 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
29256 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
29257 all. These are the two most common ways to instruct a program that it
29258 should read from standard input or write to standard output. If you
29259 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
29260 construct: see for example @file{f/lang-specs.h}.
29262 @item %.@var{SUFFIX}
29263 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
29264 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
29265 terminated by the next space or %.
29268 Marks the argument containing or following the @samp{%w} as the
29269 designated output file of this compilation. This puts the argument
29270 into the sequence of arguments that @samp{%o} substitutes.
29273 Substitutes the names of all the output files, with spaces
29274 automatically placed around them. You should write spaces
29275 around the @samp{%o} as well or the results are undefined.
29276 @samp{%o} is for use in the specs for running the linker.
29277 Input files whose names have no recognized suffix are not compiled
29278 at all, but they are included among the output files, so they are
29282 Substitutes the suffix for object files. Note that this is
29283 handled specially when it immediately follows @samp{%g, %u, or %U},
29284 because of the need for those to form complete file names. The
29285 handling is such that @samp{%O} is treated exactly as if it had already
29286 been substituted, except that @samp{%g, %u, and %U} do not currently
29287 support additional @var{suffix} characters following @samp{%O} as they do
29288 following, for example, @samp{.o}.
29291 Substitutes the standard macro predefinitions for the
29292 current target machine. Use this when running @command{cpp}.
29295 Like @samp{%p}, but puts @samp{__} before and after the name of each
29296 predefined macro, except for macros that start with @samp{__} or with
29297 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
29301 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
29302 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
29303 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
29304 and @option{-imultilib} as necessary.
29307 Current argument is the name of a library or startup file of some sort.
29308 Search for that file in a standard list of directories and substitute
29309 the full name found. The current working directory is included in the
29310 list of directories scanned.
29313 Current argument is the name of a linker script. Search for that file
29314 in the current list of directories to scan for libraries. If the file
29315 is located insert a @option{--script} option into the command line
29316 followed by the full path name found. If the file is not found then
29317 generate an error message. Note: the current working directory is not
29321 Print @var{str} as an error message. @var{str} is terminated by a newline.
29322 Use this when inconsistent options are detected.
29324 @item %(@var{name})
29325 Substitute the contents of spec string @var{name} at this point.
29327 @item %x@{@var{option}@}
29328 Accumulate an option for @samp{%X}.
29331 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
29335 Output the accumulated assembler options specified by @option{-Wa}.
29338 Output the accumulated preprocessor options specified by @option{-Wp}.
29341 Process the @code{asm} spec. This is used to compute the
29342 switches to be passed to the assembler.
29345 Process the @code{asm_final} spec. This is a spec string for
29346 passing switches to an assembler post-processor, if such a program is
29350 Process the @code{link} spec. This is the spec for computing the
29351 command line passed to the linker. Typically it makes use of the
29352 @samp{%L %G %S %D and %E} sequences.
29355 Dump out a @option{-L} option for each directory that GCC believes might
29356 contain startup files. If the target supports multilibs then the
29357 current multilib directory is prepended to each of these paths.
29360 Process the @code{lib} spec. This is a spec string for deciding which
29361 libraries are included on the command line to the linker.
29364 Process the @code{libgcc} spec. This is a spec string for deciding
29365 which GCC support library is included on the command line to the linker.
29368 Process the @code{startfile} spec. This is a spec for deciding which
29369 object files are the first ones passed to the linker. Typically
29370 this might be a file named @file{crt0.o}.
29373 Process the @code{endfile} spec. This is a spec string that specifies
29374 the last object files that are passed to the linker.
29377 Process the @code{cpp} spec. This is used to construct the arguments
29378 to be passed to the C preprocessor.
29381 Process the @code{cc1} spec. This is used to construct the options to be
29382 passed to the actual C compiler (@command{cc1}).
29385 Process the @code{cc1plus} spec. This is used to construct the options to be
29386 passed to the actual C++ compiler (@command{cc1plus}).
29389 Substitute the variable part of a matched option. See below.
29390 Note that each comma in the substituted string is replaced by
29394 Remove all occurrences of @code{-S} from the command line. Note---this
29395 command is position dependent. @samp{%} commands in the spec string
29396 before this one see @code{-S}, @samp{%} commands in the spec string
29397 after this one do not.
29399 @item %:@var{function}(@var{args})
29400 Call the named function @var{function}, passing it @var{args}.
29401 @var{args} is first processed as a nested spec string, then split
29402 into an argument vector in the usual fashion. The function returns
29403 a string which is processed as if it had appeared literally as part
29404 of the current spec.
29406 The following built-in spec functions are provided:
29409 @item @code{getenv}
29410 The @code{getenv} spec function takes two arguments: an environment
29411 variable name and a string. If the environment variable is not
29412 defined, a fatal error is issued. Otherwise, the return value is the
29413 value of the environment variable concatenated with the string. For
29414 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
29417 %:getenv(TOPDIR /include)
29420 expands to @file{/path/to/top/include}.
29422 @item @code{if-exists}
29423 The @code{if-exists} spec function takes one argument, an absolute
29424 pathname to a file. If the file exists, @code{if-exists} returns the
29425 pathname. Here is a small example of its usage:
29429 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
29432 @item @code{if-exists-else}
29433 The @code{if-exists-else} spec function is similar to the @code{if-exists}
29434 spec function, except that it takes two arguments. The first argument is
29435 an absolute pathname to a file. If the file exists, @code{if-exists-else}
29436 returns the pathname. If it does not exist, it returns the second argument.
29437 This way, @code{if-exists-else} can be used to select one file or another,
29438 based on the existence of the first. Here is a small example of its usage:
29442 crt0%O%s %:if-exists(crti%O%s) \
29443 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
29446 @item @code{replace-outfile}
29447 The @code{replace-outfile} spec function takes two arguments. It looks for the
29448 first argument in the outfiles array and replaces it with the second argument. Here
29449 is a small example of its usage:
29452 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
29455 @item @code{remove-outfile}
29456 The @code{remove-outfile} spec function takes one argument. It looks for the
29457 first argument in the outfiles array and removes it. Here is a small example
29461 %:remove-outfile(-lm)
29464 @item @code{pass-through-libs}
29465 The @code{pass-through-libs} spec function takes any number of arguments. It
29466 finds any @option{-l} options and any non-options ending in @file{.a} (which it
29467 assumes are the names of linker input library archive files) and returns a
29468 result containing all the found arguments each prepended by
29469 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
29470 intended to be passed to the LTO linker plugin.
29473 %:pass-through-libs(%G %L %G)
29476 @item @code{print-asm-header}
29477 The @code{print-asm-header} function takes no arguments and simply
29478 prints a banner like:
29484 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
29487 It is used to separate compiler options from assembler options
29488 in the @option{--target-help} output.
29492 Substitutes the @code{-S} switch, if that switch is given to GCC@.
29493 If that switch is not specified, this substitutes nothing. Note that
29494 the leading dash is omitted when specifying this option, and it is
29495 automatically inserted if the substitution is performed. Thus the spec
29496 string @samp{%@{foo@}} matches the command-line option @option{-foo}
29497 and outputs the command-line option @option{-foo}.
29500 Like %@{@code{S}@} but mark last argument supplied within as a file to be
29501 deleted on failure.
29504 Substitutes all the switches specified to GCC whose names start
29505 with @code{-S}, but which also take an argument. This is used for
29506 switches like @option{-o}, @option{-D}, @option{-I}, etc.
29507 GCC considers @option{-o foo} as being
29508 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
29509 text, including the space. Thus two arguments are generated.
29512 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
29513 (the order of @code{S} and @code{T} in the spec is not significant).
29514 There can be any number of ampersand-separated variables; for each the
29515 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
29518 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
29521 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
29524 Substitutes @code{X} if one or more switches whose names start with
29525 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
29526 once, no matter how many such switches appeared. However, if @code{%*}
29527 appears somewhere in @code{X}, then @code{X} is substituted once
29528 for each matching switch, with the @code{%*} replaced by the part of
29529 that switch matching the @code{*}.
29531 If @code{%*} appears as the last part of a spec sequence then a space
29532 is added after the end of the last substitution. If there is more
29533 text in the sequence, however, then a space is not generated. This
29534 allows the @code{%*} substitution to be used as part of a larger
29535 string. For example, a spec string like this:
29538 %@{mcu=*:--script=%*/memory.ld@}
29542 when matching an option like @option{-mcu=newchip} produces:
29545 --script=newchip/memory.ld
29549 Substitutes @code{X}, if processing a file with suffix @code{S}.
29552 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
29555 Substitutes @code{X}, if processing a file for language @code{S}.
29558 Substitutes @code{X}, if not processing a file for language @code{S}.
29561 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
29562 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
29563 @code{*} sequences as well, although they have a stronger binding than
29564 the @samp{|}. If @code{%*} appears in @code{X}, all of the
29565 alternatives must be starred, and only the first matching alternative
29568 For example, a spec string like this:
29571 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
29575 outputs the following command-line options from the following input
29576 command-line options:
29581 -d fred.c -foo -baz -boggle
29582 -d jim.d -bar -baz -boggle
29585 @item %@{S:X; T:Y; :D@}
29587 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
29588 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
29589 be as many clauses as you need. This may be combined with @code{.},
29590 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
29595 The switch matching text @code{S} in a @samp{%@{S@}}, @samp{%@{S:X@}}
29596 or similar construct can use a backslash to ignore the special meaning
29597 of the character following it, thus allowing literal matching of a
29598 character that is otherwise specially treated. For example,
29599 @samp{%@{std=iso9899\:1999:X@}} substitutes @code{X} if the
29600 @option{-std=iso9899:1999} option is given.
29602 The conditional text @code{X} in a @samp{%@{S:X@}} or similar
29603 construct may contain other nested @samp{%} constructs or spaces, or
29604 even newlines. They are processed as usual, as described above.
29605 Trailing white space in @code{X} is ignored. White space may also
29606 appear anywhere on the left side of the colon in these constructs,
29607 except between @code{.} or @code{*} and the corresponding word.
29609 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
29610 handled specifically in these constructs. If another value of
29611 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
29612 @option{-W} switch is found later in the command line, the earlier
29613 switch value is ignored, except with @{@code{S}*@} where @code{S} is
29614 just one letter, which passes all matching options.
29616 The character @samp{|} at the beginning of the predicate text is used to
29617 indicate that a command should be piped to the following command, but
29618 only if @option{-pipe} is specified.
29620 It is built into GCC which switches take arguments and which do not.
29621 (You might think it would be useful to generalize this to allow each
29622 compiler's spec to say which switches take arguments. But this cannot
29623 be done in a consistent fashion. GCC cannot even decide which input
29624 files have been specified without knowing which switches take arguments,
29625 and it must know which input files to compile in order to tell which
29628 GCC also knows implicitly that arguments starting in @option{-l} are to be
29629 treated as compiler output files, and passed to the linker in their
29630 proper position among the other output files.
29632 @node Environment Variables
29633 @section Environment Variables Affecting GCC
29634 @cindex environment variables
29636 @c man begin ENVIRONMENT
29637 This section describes several environment variables that affect how GCC
29638 operates. Some of them work by specifying directories or prefixes to use
29639 when searching for various kinds of files. Some are used to specify other
29640 aspects of the compilation environment.
29642 Note that you can also specify places to search using options such as
29643 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
29644 take precedence over places specified using environment variables, which
29645 in turn take precedence over those specified by the configuration of GCC@.
29646 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
29647 GNU Compiler Collection (GCC) Internals}.
29652 @c @itemx LC_COLLATE
29654 @c @itemx LC_MONETARY
29655 @c @itemx LC_NUMERIC
29660 @c @findex LC_COLLATE
29661 @findex LC_MESSAGES
29662 @c @findex LC_MONETARY
29663 @c @findex LC_NUMERIC
29667 These environment variables control the way that GCC uses
29668 localization information which allows GCC to work with different
29669 national conventions. GCC inspects the locale categories
29670 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
29671 so. These locale categories can be set to any value supported by your
29672 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
29673 Kingdom encoded in UTF-8.
29675 The @env{LC_CTYPE} environment variable specifies character
29676 classification. GCC uses it to determine the character boundaries in
29677 a string; this is needed for some multibyte encodings that contain quote
29678 and escape characters that are otherwise interpreted as a string
29681 The @env{LC_MESSAGES} environment variable specifies the language to
29682 use in diagnostic messages.
29684 If the @env{LC_ALL} environment variable is set, it overrides the value
29685 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
29686 and @env{LC_MESSAGES} default to the value of the @env{LANG}
29687 environment variable. If none of these variables are set, GCC
29688 defaults to traditional C English behavior.
29692 If @env{TMPDIR} is set, it specifies the directory to use for temporary
29693 files. GCC uses temporary files to hold the output of one stage of
29694 compilation which is to be used as input to the next stage: for example,
29695 the output of the preprocessor, which is the input to the compiler
29698 @item GCC_COMPARE_DEBUG
29699 @findex GCC_COMPARE_DEBUG
29700 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
29701 @option{-fcompare-debug} to the compiler driver. See the documentation
29702 of this option for more details.
29704 @item GCC_EXEC_PREFIX
29705 @findex GCC_EXEC_PREFIX
29706 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
29707 names of the subprograms executed by the compiler. No slash is added
29708 when this prefix is combined with the name of a subprogram, but you can
29709 specify a prefix that ends with a slash if you wish.
29711 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
29712 an appropriate prefix to use based on the pathname it is invoked with.
29714 If GCC cannot find the subprogram using the specified prefix, it
29715 tries looking in the usual places for the subprogram.
29717 The default value of @env{GCC_EXEC_PREFIX} is
29718 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
29719 the installed compiler. In many cases @var{prefix} is the value
29720 of @code{prefix} when you ran the @file{configure} script.
29722 Other prefixes specified with @option{-B} take precedence over this prefix.
29724 This prefix is also used for finding files such as @file{crt0.o} that are
29727 In addition, the prefix is used in an unusual way in finding the
29728 directories to search for header files. For each of the standard
29729 directories whose name normally begins with @samp{/usr/local/lib/gcc}
29730 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
29731 replacing that beginning with the specified prefix to produce an
29732 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
29733 @file{foo/bar} just before it searches the standard directory
29734 @file{/usr/local/lib/bar}.
29735 If a standard directory begins with the configured
29736 @var{prefix} then the value of @var{prefix} is replaced by
29737 @env{GCC_EXEC_PREFIX} when looking for header files.
29739 @item COMPILER_PATH
29740 @findex COMPILER_PATH
29741 The value of @env{COMPILER_PATH} is a colon-separated list of
29742 directories, much like @env{PATH}. GCC tries the directories thus
29743 specified when searching for subprograms, if it cannot find the
29744 subprograms using @env{GCC_EXEC_PREFIX}.
29747 @findex LIBRARY_PATH
29748 The value of @env{LIBRARY_PATH} is a colon-separated list of
29749 directories, much like @env{PATH}. When configured as a native compiler,
29750 GCC tries the directories thus specified when searching for special
29751 linker files, if it cannot find them using @env{GCC_EXEC_PREFIX}. Linking
29752 using GCC also uses these directories when searching for ordinary
29753 libraries for the @option{-l} option (but directories specified with
29754 @option{-L} come first).
29758 @cindex locale definition
29759 This variable is used to pass locale information to the compiler. One way in
29760 which this information is used is to determine the character set to be used
29761 when character literals, string literals and comments are parsed in C and C++.
29762 When the compiler is configured to allow multibyte characters,
29763 the following values for @env{LANG} are recognized:
29767 Recognize JIS characters.
29769 Recognize SJIS characters.
29771 Recognize EUCJP characters.
29774 If @env{LANG} is not defined, or if it has some other value, then the
29775 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
29776 recognize and translate multibyte characters.
29780 Some additional environment variables affect the behavior of the
29783 @include cppenv.texi
29787 @node Precompiled Headers
29788 @section Using Precompiled Headers
29789 @cindex precompiled headers
29790 @cindex speed of compilation
29792 Often large projects have many header files that are included in every
29793 source file. The time the compiler takes to process these header files
29794 over and over again can account for nearly all of the time required to
29795 build the project. To make builds faster, GCC allows you to
29796 @dfn{precompile} a header file.
29798 To create a precompiled header file, simply compile it as you would any
29799 other file, if necessary using the @option{-x} option to make the driver
29800 treat it as a C or C++ header file. You may want to use a
29801 tool like @command{make} to keep the precompiled header up-to-date when
29802 the headers it contains change.
29804 A precompiled header file is searched for when @code{#include} is
29805 seen in the compilation. As it searches for the included file
29806 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
29807 compiler looks for a precompiled header in each directory just before it
29808 looks for the include file in that directory. The name searched for is
29809 the name specified in the @code{#include} with @samp{.gch} appended. If
29810 the precompiled header file cannot be used, it is ignored.
29812 For instance, if you have @code{#include "all.h"}, and you have
29813 @file{all.h.gch} in the same directory as @file{all.h}, then the
29814 precompiled header file is used if possible, and the original
29815 header is used otherwise.
29817 Alternatively, you might decide to put the precompiled header file in a
29818 directory and use @option{-I} to ensure that directory is searched
29819 before (or instead of) the directory containing the original header.
29820 Then, if you want to check that the precompiled header file is always
29821 used, you can put a file of the same name as the original header in this
29822 directory containing an @code{#error} command.
29824 This also works with @option{-include}. So yet another way to use
29825 precompiled headers, good for projects not designed with precompiled
29826 header files in mind, is to simply take most of the header files used by
29827 a project, include them from another header file, precompile that header
29828 file, and @option{-include} the precompiled header. If the header files
29829 have guards against multiple inclusion, they are skipped because
29830 they've already been included (in the precompiled header).
29832 If you need to precompile the same header file for different
29833 languages, targets, or compiler options, you can instead make a
29834 @emph{directory} named like @file{all.h.gch}, and put each precompiled
29835 header in the directory, perhaps using @option{-o}. It doesn't matter
29836 what you call the files in the directory; every precompiled header in
29837 the directory is considered. The first precompiled header
29838 encountered in the directory that is valid for this compilation is
29839 used; they're searched in no particular order.
29841 There are many other possibilities, limited only by your imagination,
29842 good sense, and the constraints of your build system.
29844 A precompiled header file can be used only when these conditions apply:
29848 Only one precompiled header can be used in a particular compilation.
29851 A precompiled header cannot be used once the first C token is seen. You
29852 can have preprocessor directives before a precompiled header; you cannot
29853 include a precompiled header from inside another header.
29856 The precompiled header file must be produced for the same language as
29857 the current compilation. You cannot use a C precompiled header for a C++
29861 The precompiled header file must have been produced by the same compiler
29862 binary as the current compilation is using.
29865 Any macros defined before the precompiled header is included must
29866 either be defined in the same way as when the precompiled header was
29867 generated, or must not affect the precompiled header, which usually
29868 means that they don't appear in the precompiled header at all.
29870 The @option{-D} option is one way to define a macro before a
29871 precompiled header is included; using a @code{#define} can also do it.
29872 There are also some options that define macros implicitly, like
29873 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
29876 @item If debugging information is output when using the precompiled
29877 header, using @option{-g} or similar, the same kind of debugging information
29878 must have been output when building the precompiled header. However,
29879 a precompiled header built using @option{-g} can be used in a compilation
29880 when no debugging information is being output.
29882 @item The same @option{-m} options must generally be used when building
29883 and using the precompiled header. @xref{Submodel Options},
29884 for any cases where this rule is relaxed.
29886 @item Each of the following options must be the same when building and using
29887 the precompiled header:
29889 @gccoptlist{-fexceptions}
29892 Some other command-line options starting with @option{-f},
29893 @option{-p}, or @option{-O} must be defined in the same way as when
29894 the precompiled header was generated. At present, it's not clear
29895 which options are safe to change and which are not; the safest choice
29896 is to use exactly the same options when generating and using the
29897 precompiled header. The following are known to be safe:
29899 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
29900 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
29901 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
29906 For all of these except the last, the compiler automatically
29907 ignores the precompiled header if the conditions aren't met. If you
29908 find an option combination that doesn't work and doesn't cause the
29909 precompiled header to be ignored, please consider filing a bug report,
29912 If you do use differing options when generating and using the
29913 precompiled header, the actual behavior is a mixture of the
29914 behavior for the options. For instance, if you use @option{-g} to
29915 generate the precompiled header but not when using it, you may or may
29916 not get debugging information for routines in the precompiled header.