1 @c Copyright (C) 1988-2019 Free Software Foundation, Inc.
2 @c This is part of the GCC manual.
3 @c For copying conditions, see the file gcc.texi.
10 @c man begin COPYRIGHT
11 Copyright @copyright{} 1988-2019 Free Software Foundation, Inc.
13 Permission is granted to copy, distribute and/or modify this document
14 under the terms of the GNU Free Documentation License, Version 1.3 or
15 any later version published by the Free Software Foundation; with the
16 Invariant Sections being ``GNU General Public License'' and ``Funding
17 Free Software'', the Front-Cover texts being (a) (see below), and with
18 the Back-Cover Texts being (b) (see below). A copy of the license is
19 included in the gfdl(7) man page.
21 (a) The FSF's Front-Cover Text is:
25 (b) The FSF's Back-Cover Text is:
27 You have freedom to copy and modify this GNU Manual, like GNU
28 software. Copies published by the Free Software Foundation raise
29 funds for GNU development.
31 @c Set file name and title for the man page.
33 @settitle GNU project C and C++ compiler
35 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
36 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
37 [@option{-W}@var{warn}@dots{}] [@option{-Wpedantic}]
38 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
39 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
40 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
41 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
43 Only the most useful options are listed here; see below for the
44 remainder. @command{g++} accepts mostly the same options as @command{gcc}.
47 gpl(7), gfdl(7), fsf-funding(7),
48 cpp(1), gcov(1), as(1), ld(1), gdb(1), dbx(1)
49 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
50 @file{ld}, @file{binutils} and @file{gdb}.
53 For instructions on reporting bugs, see
57 See the Info entry for @command{gcc}, or
58 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
59 for contributors to GCC@.
64 @chapter GCC Command Options
65 @cindex GCC command options
66 @cindex command options
67 @cindex options, GCC command
69 @c man begin DESCRIPTION
70 When you invoke GCC, it normally does preprocessing, compilation,
71 assembly and linking. The ``overall options'' allow you to stop this
72 process at an intermediate stage. For example, the @option{-c} option
73 says not to run the linker. Then the output consists of object files
74 output by the assembler.
75 @xref{Overall Options,,Options Controlling the Kind of Output}.
77 Other options are passed on to one or more stages of processing. Some options
78 control the preprocessor and others the compiler itself. Yet other
79 options control the assembler and linker; most of these are not
80 documented here, since you rarely need to use any of them.
82 @cindex C compilation options
83 Most of the command-line options that you can use with GCC are useful
84 for C programs; when an option is only useful with another language
85 (usually C++), the explanation says so explicitly. If the description
86 for a particular option does not mention a source language, you can use
87 that option with all supported languages.
89 @cindex cross compiling
90 @cindex specifying machine version
91 @cindex specifying compiler version and target machine
92 @cindex compiler version, specifying
93 @cindex target machine, specifying
94 The usual way to run GCC is to run the executable called @command{gcc}, or
95 @command{@var{machine}-gcc} when cross-compiling, or
96 @command{@var{machine}-gcc-@var{version}} to run a specific version of GCC.
97 When you compile C++ programs, you should invoke GCC as @command{g++}
98 instead. @xref{Invoking G++,,Compiling C++ Programs},
99 for information about the differences in behavior between @command{gcc}
100 and @code{g++} when compiling C++ programs.
102 @cindex grouping options
103 @cindex options, grouping
104 The @command{gcc} program accepts options and file names as operands. Many
105 options have multi-letter names; therefore multiple single-letter options
106 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
109 @cindex order of options
110 @cindex options, order
111 You can mix options and other arguments. For the most part, the order
112 you use doesn't matter. Order does matter when you use several
113 options of the same kind; for example, if you specify @option{-L} more
114 than once, the directories are searched in the order specified. Also,
115 the placement of the @option{-l} option is significant.
117 Many options have long names starting with @samp{-f} or with
118 @samp{-W}---for example,
119 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
120 these have both positive and negative forms; the negative form of
121 @option{-ffoo} is @option{-fno-foo}. This manual documents
122 only one of these two forms, whichever one is not the default.
124 Some options take one or more arguments typically separated either
125 by a space or by the equals sign (@samp{=}) from the option name.
126 Unless documented otherwise, an argument can be either numeric or
127 a string. Numeric arguments must typically be small unsigned decimal
128 or hexadecimal integers. Hexadecimal arguments must begin with
129 the @samp{0x} prefix. Arguments to options that specify a size
130 threshold of some sort may be arbitrarily large decimal or hexadecimal
131 integers followed by a byte size suffix designating a multiple of bytes
132 such as @code{kB} and @code{KiB} for kilobyte and kibibyte, respectively,
133 @code{MB} and @code{MiB} for megabyte and mebibyte, @code{GB} and
134 @code{GiB} for gigabyte and gigibyte, and so on. Such arguments are
135 designated by @var{byte-size} in the following text. Refer to the NIST,
136 IEC, and other relevant national and international standards for the full
137 listing and explanation of the binary and decimal byte size prefixes.
141 @xref{Option Index}, for an index to GCC's options.
144 * Option Summary:: Brief list of all options, without explanations.
145 * Overall Options:: Controlling the kind of output:
146 an executable, object files, assembler files,
147 or preprocessed source.
148 * Invoking G++:: Compiling C++ programs.
149 * C Dialect Options:: Controlling the variant of C language compiled.
150 * C++ Dialect Options:: Variations on C++.
151 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
153 * Diagnostic Message Formatting Options:: Controlling how diagnostics should
155 * Warning Options:: How picky should the compiler be?
156 * Debugging Options:: Producing debuggable code.
157 * Optimize Options:: How much optimization?
158 * Instrumentation Options:: Enabling profiling and extra run-time error checking.
159 * Preprocessor Options:: Controlling header files and macro definitions.
160 Also, getting dependency information for Make.
161 * Assembler Options:: Passing options to the assembler.
162 * Link Options:: Specifying libraries and so on.
163 * Directory Options:: Where to find header files and libraries.
164 Where to find the compiler executable files.
165 * Code Gen Options:: Specifying conventions for function calls, data layout
167 * Developer Options:: Printing GCC configuration info, statistics, and
169 * Submodel Options:: Target-specific options, such as compiling for a
170 specific processor variant.
171 * Spec Files:: How to pass switches to sub-processes.
172 * Environment Variables:: Env vars that affect GCC.
173 * Precompiled Headers:: Compiling a header once, and using it many times.
179 @section Option Summary
181 Here is a summary of all the options, grouped by type. Explanations are
182 in the following sections.
185 @item Overall Options
186 @xref{Overall Options,,Options Controlling the Kind of Output}.
187 @gccoptlist{-c -S -E -o @var{file} -x @var{language} @gol
188 -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help --version @gol
189 -pass-exit-codes -pipe -specs=@var{file} -wrapper @gol
190 @@@var{file} -ffile-prefix-map=@var{old}=@var{new} @gol
191 -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
192 -fdump-ada-spec@r{[}-slim@r{]} -fada-spec-parent=@var{unit} -fdump-go-spec=@var{file}}
194 @item C Language Options
195 @xref{C Dialect Options,,Options Controlling C Dialect}.
196 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
197 -fpermitted-flt-eval-methods=@var{standard} @gol
198 -aux-info @var{filename} -fallow-parameterless-variadic-functions @gol
199 -fno-asm -fno-builtin -fno-builtin-@var{function} -fgimple@gol
200 -fhosted -ffreestanding @gol
201 -fopenacc -fopenacc-dim=@var{geom} @gol
202 -fopenmp -fopenmp-simd @gol
203 -fms-extensions -fplan9-extensions -fsso-struct=@var{endianness} @gol
204 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
205 -fsigned-bitfields -fsigned-char @gol
206 -funsigned-bitfields -funsigned-char}
208 @item C++ Language Options
209 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
210 @gccoptlist{-fabi-version=@var{n} -fno-access-control @gol
211 -faligned-new=@var{n} -fargs-in-order=@var{n} -fchar8_t -fcheck-new @gol
212 -fconstexpr-depth=@var{n} -fconstexpr-cache-depth=@var{n} @gol
213 -fconstexpr-loop-limit=@var{n} -fconstexpr-ops-limit=@var{n} @gol
214 -fno-elide-constructors @gol
215 -fno-enforce-eh-specs @gol
216 -fno-gnu-keywords @gol
217 -fno-implicit-templates @gol
218 -fno-implicit-inline-templates @gol
219 -fno-implement-inlines -fms-extensions @gol
220 -fnew-inheriting-ctors @gol
221 -fnew-ttp-matching @gol
222 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
223 -fno-optional-diags -fpermissive @gol
224 -fno-pretty-templates @gol
225 -frepo -fno-rtti -fsized-deallocation @gol
226 -ftemplate-backtrace-limit=@var{n} @gol
227 -ftemplate-depth=@var{n} @gol
228 -fno-threadsafe-statics -fuse-cxa-atexit @gol
229 -fno-weak -nostdinc++ @gol
230 -fvisibility-inlines-hidden @gol
231 -fvisibility-ms-compat @gol
232 -fext-numeric-literals @gol
233 -Wabi=@var{n} -Wabi-tag -Wcomma-subscript -Wconversion-null @gol
234 -Wctor-dtor-privacy @gol
235 -Wdelete-non-virtual-dtor -Wdeprecated-copy -Wdeprecated-copy-dtor @gol
236 -Wliteral-suffix @gol
237 -Wmultiple-inheritance -Wno-init-list-lifetime @gol
238 -Wnamespaces -Wnarrowing @gol
239 -Wpessimizing-move -Wredundant-move @gol
240 -Wnoexcept -Wnoexcept-type -Wclass-memaccess @gol
241 -Wnon-virtual-dtor -Wreorder -Wregister @gol
242 -Weffc++ -Wstrict-null-sentinel -Wtemplates @gol
243 -Wno-non-template-friend -Wold-style-cast @gol
244 -Woverloaded-virtual -Wno-pmf-conversions @gol
245 -Wno-class-conversion -Wno-terminate @gol
246 -Wsign-promo -Wvirtual-inheritance -Wvolatile}
248 @item Objective-C and Objective-C++ Language Options
249 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
250 Objective-C and Objective-C++ Dialects}.
251 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
252 -fgnu-runtime -fnext-runtime @gol
253 -fno-nil-receivers @gol
254 -fobjc-abi-version=@var{n} @gol
255 -fobjc-call-cxx-cdtors @gol
256 -fobjc-direct-dispatch @gol
257 -fobjc-exceptions @gol
260 -fobjc-std=objc1 @gol
261 -fno-local-ivars @gol
262 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
263 -freplace-objc-classes @gol
266 -Wassign-intercept @gol
267 -Wno-protocol -Wselector @gol
268 -Wstrict-selector-match @gol
269 -Wundeclared-selector}
271 @item Diagnostic Message Formatting Options
272 @xref{Diagnostic Message Formatting Options,,Options to Control Diagnostic Messages Formatting}.
273 @gccoptlist{-fmessage-length=@var{n} @gol
274 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
275 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
276 -fdiagnostics-format=@r{[}text@r{|}json@r{]} @gol
277 -fno-diagnostics-show-option -fno-diagnostics-show-caret @gol
278 -fno-diagnostics-show-labels -fno-diagnostics-show-line-numbers @gol
279 -fdiagnostics-minimum-margin-width=@var{width} @gol
280 -fdiagnostics-parseable-fixits -fdiagnostics-generate-patch @gol
281 -fdiagnostics-show-template-tree -fno-elide-type @gol
284 @item Warning Options
285 @xref{Warning Options,,Options to Request or Suppress Warnings}.
286 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
287 -pedantic-errors @gol
288 -w -Wextra -Wall -Waddress -Waddress-of-packed-member @gol
289 -Waggregate-return -Waligned-new @gol
290 -Walloc-zero -Walloc-size-larger-than=@var{byte-size} @gol
291 -Walloca -Walloca-larger-than=@var{byte-size} @gol
292 -Wno-aggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
293 -Wno-attributes -Wattribute-alias=@var{n} @gol
294 -Wbool-compare -Wbool-operation @gol
295 -Wno-builtin-declaration-mismatch @gol
296 -Wno-builtin-macro-redefined -Wc90-c99-compat -Wc99-c11-compat @gol
297 -Wc++-compat -Wc++11-compat -Wc++14-compat -Wc++17-compat @gol
299 -Wcast-align -Wcast-align=strict -Wcast-function-type -Wcast-qual @gol
300 -Wchar-subscripts -Wcatch-value -Wcatch-value=@var{n} @gol
301 -Wclobbered -Wcomment -Wconditionally-supported @gol
302 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wdangling-else -Wdate-time @gol
303 -Wdelete-incomplete @gol
304 -Wno-attribute-warning @gol
305 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
306 -Wdisabled-optimization @gol
307 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
308 -Wno-div-by-zero -Wdouble-promotion @gol
309 -Wduplicated-branches -Wduplicated-cond @gol
310 -Wempty-body -Wenum-compare -Wenum-conversion @gol
311 -Wno-endif-labels -Wexpansion-to-defined @gol
312 -Werror -Werror=* -Wextra-semi -Wfatal-errors @gol
313 -Wfloat-equal -Wformat -Wformat=2 @gol
314 -Wno-format-contains-nul -Wno-format-extra-args @gol
315 -Wformat-nonliteral -Wformat-overflow=@var{n} @gol
316 -Wformat-security -Wformat-signedness -Wformat-truncation=@var{n} @gol
317 -Wformat-y2k -Wframe-address @gol
318 -Wframe-larger-than=@var{byte-size} -Wno-free-nonheap-object @gol
319 -Wjump-misses-init @gol
320 -Whsa -Wif-not-aligned @gol
321 -Wignored-qualifiers -Wignored-attributes -Wincompatible-pointer-types @gol
322 -Wimplicit -Wimplicit-fallthrough -Wimplicit-fallthrough=@var{n} @gol
323 -Wimplicit-function-declaration -Wimplicit-int @gol
324 -Winaccessible-base @gol
325 -Winit-self -Winline -Wno-int-conversion -Wint-in-bool-context @gol
326 -Wno-int-to-pointer-cast -Winvalid-memory-model -Wno-invalid-offsetof @gol
327 -Winvalid-pch -Wlarger-than=@var{byte-size} @gol
328 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
329 -Wmain -Wmaybe-uninitialized -Wmemset-elt-size -Wmemset-transposed-args @gol
330 -Wmisleading-indentation -Wmissing-attributes -Wmissing-braces @gol
331 -Wmissing-field-initializers -Wmissing-format-attribute @gol
332 -Wmissing-include-dirs -Wmissing-noreturn -Wmissing-profile @gol
333 -Wno-multichar -Wmultistatement-macros -Wnonnull -Wnonnull-compare @gol
334 -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
335 -Wnull-dereference -Wodr -Wno-overflow -Wopenmp-simd @gol
336 -Woverride-init-side-effects -Woverlength-strings @gol
337 -Wpacked -Wpacked-bitfield-compat -Wpacked-not-aligned -Wpadded @gol
338 -Wparentheses -Wno-pedantic-ms-format @gol
339 -Wplacement-new -Wplacement-new=@var{n} @gol
340 -Wpointer-arith -Wpointer-compare -Wno-pointer-to-int-cast @gol
341 -Wno-pragmas -Wno-prio-ctor-dtor -Wredundant-decls @gol
342 -Wrestrict -Wno-return-local-addr @gol
343 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
344 -Wshadow=global, -Wshadow=local, -Wshadow=compatible-local @gol
345 -Wshift-overflow -Wshift-overflow=@var{n} @gol
346 -Wshift-count-negative -Wshift-count-overflow -Wshift-negative-value @gol
347 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
348 -Wno-scalar-storage-order -Wsizeof-pointer-div @gol
349 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
350 -Wstack-protector -Wstack-usage=@var{byte-size} -Wstrict-aliasing @gol
351 -Wstrict-aliasing=n -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
352 -Wstringop-overflow=@var{n} -Wstringop-truncation -Wsubobject-linkage @gol
353 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{|}malloc@r{]} @gol
354 -Wsuggest-final-types @gol -Wsuggest-final-methods -Wsuggest-override @gol
355 -Wswitch -Wswitch-bool -Wswitch-default -Wswitch-enum @gol
356 -Wswitch-unreachable -Wsync-nand @gol
357 -Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs @gol
358 -Wtype-limits -Wundef @gol
359 -Wuninitialized -Wunknown-pragmas @gol
360 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
361 -Wunused-label -Wunused-local-typedefs -Wunused-macros @gol
362 -Wunused-parameter -Wno-unused-result @gol
363 -Wunused-value -Wunused-variable @gol
364 -Wunused-const-variable -Wunused-const-variable=@var{n} @gol
365 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
366 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
367 -Wvla -Wvla-larger-than=@var{byte-size} -Wvolatile-register-var @gol
369 -Wzero-as-null-pointer-constant}
371 @item C and Objective-C-only Warning Options
372 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
373 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
374 -Wold-style-declaration -Wold-style-definition @gol
375 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
376 -Wdeclaration-after-statement -Wpointer-sign}
378 @item Debugging Options
379 @xref{Debugging Options,,Options for Debugging Your Program}.
380 @gccoptlist{-g -g@var{level} -gdwarf -gdwarf-@var{version} @gol
381 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
382 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
383 -gas-loc-support -gno-as-loc-support @gol
384 -gas-locview-support -gno-as-locview-support @gol
385 -gcolumn-info -gno-column-info @gol
386 -gstatement-frontiers -gno-statement-frontiers @gol
387 -gvariable-location-views -gno-variable-location-views @gol
388 -ginternal-reset-location-views -gno-internal-reset-location-views @gol
389 -ginline-points -gno-inline-points @gol
390 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
391 -gsplit-dwarf -gdescribe-dies -gno-describe-dies @gol
392 -fdebug-prefix-map=@var{old}=@var{new} -fdebug-types-section @gol
393 -fno-eliminate-unused-debug-types @gol
394 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
395 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
396 -fno-eliminate-unused-debug-symbols -femit-class-debug-always @gol
397 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
398 -fvar-tracking -fvar-tracking-assignments}
400 @item Optimization Options
401 @xref{Optimize Options,,Options that Control Optimization}.
402 @gccoptlist{-faggressive-loop-optimizations @gol
403 -falign-functions[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
404 -falign-jumps[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
405 -falign-labels[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
406 -falign-loops[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
407 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
408 -fauto-inc-dec -fbranch-probabilities @gol
409 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
410 -fbtr-bb-exclusive -fcaller-saves @gol
411 -fcombine-stack-adjustments -fconserve-stack @gol
412 -fcompare-elim -fcprop-registers -fcrossjumping @gol
413 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
414 -fcx-limited-range @gol
415 -fdata-sections -fdce -fdelayed-branch @gol
416 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
417 -fdevirtualize-at-ltrans -fdse @gol
418 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
419 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
421 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
422 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
423 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
424 -fif-conversion2 -findirect-inlining @gol
425 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
426 -finline-small-functions -fipa-cp -fipa-cp-clone @gol
427 -fipa-bit-cp -fipa-vrp -fipa-pta -fipa-profile -fipa-pure-const @gol
428 -fipa-reference -fipa-reference-addressable @gol
429 -fipa-stack-alignment -fipa-icf -fira-algorithm=@var{algorithm} @gol
430 -flive-patching=@var{level} @gol
431 -fira-region=@var{region} -fira-hoist-pressure @gol
432 -fira-loop-pressure -fno-ira-share-save-slots @gol
433 -fno-ira-share-spill-slots @gol
434 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
435 -fivopts -fkeep-inline-functions -fkeep-static-functions @gol
436 -fkeep-static-consts -flimit-function-alignment -flive-range-shrinkage @gol
437 -floop-block -floop-interchange -floop-strip-mine @gol
438 -floop-unroll-and-jam -floop-nest-optimize @gol
439 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
440 -flto-partition=@var{alg} -fmerge-all-constants @gol
441 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
442 -fmove-loop-invariants -fno-branch-count-reg @gol
443 -fno-defer-pop -fno-fp-int-builtin-inexact -fno-function-cse @gol
444 -fno-guess-branch-probability -fno-inline -fno-math-errno -fno-peephole @gol
445 -fno-peephole2 -fno-printf-return-value -fno-sched-interblock @gol
446 -fno-sched-spec -fno-signed-zeros @gol
447 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
448 -fomit-frame-pointer -foptimize-sibling-calls @gol
449 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
450 -fprefetch-loop-arrays @gol
451 -fprofile-correction @gol
452 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
453 -fprofile-reorder-functions @gol
454 -freciprocal-math -free -frename-registers -freorder-blocks @gol
455 -freorder-blocks-algorithm=@var{algorithm} @gol
456 -freorder-blocks-and-partition -freorder-functions @gol
457 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
458 -frounding-math -fsave-optimization-record @gol
459 -fsched2-use-superblocks -fsched-pressure @gol
460 -fsched-spec-load -fsched-spec-load-dangerous @gol
461 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
462 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
463 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
464 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
465 -fschedule-fusion @gol
466 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
467 -fselective-scheduling -fselective-scheduling2 @gol
468 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
469 -fsemantic-interposition -fshrink-wrap -fshrink-wrap-separate @gol
470 -fsignaling-nans @gol
471 -fsingle-precision-constant -fsplit-ivs-in-unroller -fsplit-loops@gol
473 -fsplit-wide-types -fsplit-wide-types-early -fssa-backprop -fssa-phiopt @gol
474 -fstdarg-opt -fstore-merging -fstrict-aliasing @gol
475 -fthread-jumps -ftracer -ftree-bit-ccp @gol
476 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
477 -ftree-coalesce-vars -ftree-copy-prop -ftree-dce -ftree-dominator-opts @gol
478 -ftree-dse -ftree-forwprop -ftree-fre -fcode-hoisting @gol
479 -ftree-loop-if-convert -ftree-loop-im @gol
480 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
481 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
482 -ftree-loop-vectorize @gol
483 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
484 -ftree-reassoc -ftree-scev-cprop -ftree-sink -ftree-slsr -ftree-sra @gol
485 -ftree-switch-conversion -ftree-tail-merge @gol
486 -ftree-ter -ftree-vectorize -ftree-vrp -funconstrained-commons @gol
487 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
488 -funsafe-math-optimizations -funswitch-loops @gol
489 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
490 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
491 --param @var{name}=@var{value}
492 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
494 @item Program Instrumentation Options
495 @xref{Instrumentation Options,,Program Instrumentation Options}.
496 @gccoptlist{-p -pg -fprofile-arcs --coverage -ftest-coverage @gol
497 -fprofile-abs-path @gol
498 -fprofile-dir=@var{path} -fprofile-generate -fprofile-generate=@var{path} @gol
499 -fprofile-note=@var{path} -fprofile-update=@var{method} @gol
500 -fprofile-filter-files=@var{regex} -fprofile-exclude-files=@var{regex} @gol
501 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
502 -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
503 -fsanitize-undefined-trap-on-error -fbounds-check @gol
504 -fcf-protection=@r{[}full@r{|}branch@r{|}return@r{|}none@r{]} @gol
505 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
506 -fstack-protector-explicit -fstack-check @gol
507 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
508 -fno-stack-limit -fsplit-stack @gol
509 -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} @gol
510 -fvtv-counts -fvtv-debug @gol
511 -finstrument-functions @gol
512 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
513 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}}
515 @item Preprocessor Options
516 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
517 @gccoptlist{-A@var{question}=@var{answer} @gol
518 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
519 -C -CC -D@var{macro}@r{[}=@var{defn}@r{]} @gol
520 -dD -dI -dM -dN -dU @gol
521 -fdebug-cpp -fdirectives-only -fdollars-in-identifiers @gol
522 -fexec-charset=@var{charset} -fextended-identifiers @gol
523 -finput-charset=@var{charset} -fmacro-prefix-map=@var{old}=@var{new} @gol
524 -fmax-include-depth=@var{depth} @gol
525 -fno-canonical-system-headers -fpch-deps -fpch-preprocess @gol
526 -fpreprocessed -ftabstop=@var{width} -ftrack-macro-expansion @gol
527 -fwide-exec-charset=@var{charset} -fworking-directory @gol
528 -H -imacros @var{file} -include @var{file} @gol
529 -M -MD -MF -MG -MM -MMD -MP -MQ -MT @gol
530 -no-integrated-cpp -P -pthread -remap @gol
531 -traditional -traditional-cpp -trigraphs @gol
532 -U@var{macro} -undef @gol
533 -Wp,@var{option} -Xpreprocessor @var{option}}
535 @item Assembler Options
536 @xref{Assembler Options,,Passing Options to the Assembler}.
537 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
540 @xref{Link Options,,Options for Linking}.
541 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
542 -nostartfiles -nodefaultlibs -nolibc -nostdlib @gol
543 -e @var{entry} --entry=@var{entry} @gol
544 -pie -pthread -r -rdynamic @gol
545 -s -static -static-pie -static-libgcc -static-libstdc++ @gol
546 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
547 -shared -shared-libgcc -symbolic @gol
548 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
549 -u @var{symbol} -z @var{keyword}}
551 @item Directory Options
552 @xref{Directory Options,,Options for Directory Search}.
553 @gccoptlist{-B@var{prefix} -I@var{dir} -I- @gol
554 -idirafter @var{dir} @gol
555 -imacros @var{file} -imultilib @var{dir} @gol
556 -iplugindir=@var{dir} -iprefix @var{file} @gol
557 -iquote @var{dir} -isysroot @var{dir} -isystem @var{dir} @gol
558 -iwithprefix @var{dir} -iwithprefixbefore @var{dir} @gol
559 -L@var{dir} -no-canonical-prefixes --no-sysroot-suffix @gol
560 -nostdinc -nostdinc++ --sysroot=@var{dir}}
562 @item Code Generation Options
563 @xref{Code Gen Options,,Options for Code Generation Conventions}.
564 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
565 -ffixed-@var{reg} -fexceptions @gol
566 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
567 -fasynchronous-unwind-tables @gol
569 -finhibit-size-directive -fno-common -fno-ident @gol
570 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol
571 -fno-jump-tables @gol
572 -frecord-gcc-switches @gol
573 -freg-struct-return -fshort-enums -fshort-wchar @gol
574 -fverbose-asm -fpack-struct[=@var{n}] @gol
575 -fleading-underscore -ftls-model=@var{model} @gol
576 -fstack-reuse=@var{reuse_level} @gol
577 -ftrampolines -ftrapv -fwrapv @gol
578 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
579 -fstrict-volatile-bitfields -fsync-libcalls}
581 @item Developer Options
582 @xref{Developer Options,,GCC Developer Options}.
583 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
584 -dumpfullversion -fchecking -fchecking=@var{n} -fdbg-cnt-list @gol
585 -fdbg-cnt=@var{counter-value-list} @gol
586 -fdisable-ipa-@var{pass_name} @gol
587 -fdisable-rtl-@var{pass_name} @gol
588 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
589 -fdisable-tree-@var{pass_name} @gol
590 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
591 -fdump-debug -fdump-earlydebug @gol
592 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
593 -fdump-final-insns@r{[}=@var{file}@r{]} @gol
594 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
596 -fdump-lang-@var{switch} @gol
597 -fdump-lang-@var{switch}-@var{options} @gol
598 -fdump-lang-@var{switch}-@var{options}=@var{filename} @gol
600 -fdump-rtl-@var{pass} -fdump-rtl-@var{pass}=@var{filename} @gol
601 -fdump-statistics @gol
603 -fdump-tree-@var{switch} @gol
604 -fdump-tree-@var{switch}-@var{options} @gol
605 -fdump-tree-@var{switch}-@var{options}=@var{filename} @gol
606 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
607 -fenable-@var{kind}-@var{pass} @gol
608 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
609 -fira-verbose=@var{n} @gol
610 -flto-report -flto-report-wpa -fmem-report-wpa @gol
611 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report @gol
612 -fopt-info -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
613 -fprofile-report @gol
614 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
615 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
616 -fstats -fstack-usage -ftime-report -ftime-report-details @gol
617 -fvar-tracking-assignments-toggle -gtoggle @gol
618 -print-file-name=@var{library} -print-libgcc-file-name @gol
619 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
620 -print-prog-name=@var{program} -print-search-dirs -Q @gol
621 -print-sysroot -print-sysroot-headers-suffix @gol
622 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
624 @item Machine-Dependent Options
625 @xref{Submodel Options,,Machine-Dependent Options}.
626 @c This list is ordered alphanumerically by subsection name.
627 @c Try and put the significant identifier (CPU or system) first,
628 @c so users have a clue at guessing where the ones they want will be.
630 @emph{AArch64 Options}
631 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
632 -mgeneral-regs-only @gol
633 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
634 -mstrict-align -mno-strict-align @gol
635 -momit-leaf-frame-pointer @gol
636 -mtls-dialect=desc -mtls-dialect=traditional @gol
637 -mtls-size=@var{size} @gol
638 -mfix-cortex-a53-835769 -mfix-cortex-a53-843419 @gol
639 -mlow-precision-recip-sqrt -mlow-precision-sqrt -mlow-precision-div @gol
640 -mpc-relative-literal-loads @gol
641 -msign-return-address=@var{scope} @gol
642 -mbranch-protection=@var{none}|@var{standard}|@var{pac-ret}[+@var{leaf}
643 +@var{b-key}]|@var{bti} @gol
644 -march=@var{name} -mcpu=@var{name} -mtune=@var{name} @gol
645 -moverride=@var{string} -mverbose-cost-dump @gol
646 -mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{sysreg} @gol
647 -mstack-protector-guard-offset=@var{offset} -mtrack-speculation }
649 @emph{Adapteva Epiphany Options}
650 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
651 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
652 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
653 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
654 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
655 -msplit-vecmove-early -m1reg-@var{reg}}
657 @emph{AMD GCN Options}
658 @gccoptlist{-march=@var{gpu} -mtune=@var{gpu} -mstack-size=@var{bytes}}
661 @gccoptlist{-mbarrel-shifter -mjli-always @gol
662 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
663 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
664 -mea -mno-mpy -mmul32x16 -mmul64 -matomic @gol
665 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
666 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
667 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
668 -mlong-calls -mmedium-calls -msdata -mirq-ctrl-saved @gol
669 -mrgf-banked-regs -mlpc-width=@var{width} -G @var{num} @gol
670 -mvolatile-cache -mtp-regno=@var{regno} @gol
671 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
672 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
673 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
674 -mlra-priority-compact mlra-priority-noncompact -mmillicode @gol
675 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
676 -mtune=@var{cpu} -mmultcost=@var{num} -mcode-density-frame @gol
677 -munalign-prob-threshold=@var{probability} -mmpy-option=@var{multo} @gol
678 -mdiv-rem -mcode-density -mll64 -mfpu=@var{fpu} -mrf16 -mbranch-index}
681 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
682 -mabi=@var{name} @gol
683 -mapcs-stack-check -mno-apcs-stack-check @gol
684 -mapcs-reentrant -mno-apcs-reentrant @gol
685 -mgeneral-regs-only @gol
686 -msched-prolog -mno-sched-prolog @gol
687 -mlittle-endian -mbig-endian @gol
689 -mfloat-abi=@var{name} @gol
690 -mfp16-format=@var{name}
691 -mthumb-interwork -mno-thumb-interwork @gol
692 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
693 -mtune=@var{name} -mprint-tune-info @gol
694 -mstructure-size-boundary=@var{n} @gol
695 -mabort-on-noreturn @gol
696 -mlong-calls -mno-long-calls @gol
697 -msingle-pic-base -mno-single-pic-base @gol
698 -mpic-register=@var{reg} @gol
699 -mnop-fun-dllimport @gol
700 -mpoke-function-name @gol
701 -mthumb -marm -mflip-thumb @gol
702 -mtpcs-frame -mtpcs-leaf-frame @gol
703 -mcaller-super-interworking -mcallee-super-interworking @gol
704 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
705 -mword-relocations @gol
706 -mfix-cortex-m3-ldrd @gol
707 -munaligned-access @gol
708 -mneon-for-64bits @gol
709 -mslow-flash-data @gol
710 -masm-syntax-unified @gol
712 -mverbose-cost-dump @gol
717 @gccoptlist{-mmcu=@var{mcu} -mabsdata -maccumulate-args @gol
718 -mbranch-cost=@var{cost} @gol
719 -mcall-prologues -mgas-isr-prologues -mint8 @gol
720 -mn_flash=@var{size} -mno-interrupts @gol
721 -mmain-is-OS_task -mrelax -mrmw -mstrict-X -mtiny-stack @gol
722 -mfract-convert-truncate @gol
723 -mshort-calls -nodevicelib @gol
724 -Waddr-space-convert -Wmisspelled-isr}
726 @emph{Blackfin Options}
727 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
728 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
729 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
730 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
731 -mno-id-shared-library -mshared-library-id=@var{n} @gol
732 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
733 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
734 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
738 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
739 -msim -msdata=@var{sdata-type}}
742 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
743 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
744 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
745 -mstack-align -mdata-align -mconst-align @gol
746 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
747 -melf -maout -melinux -mlinux -sim -sim2 @gol
748 -mmul-bug-workaround -mno-mul-bug-workaround}
751 @gccoptlist{-mmac @gol
752 -mcr16cplus -mcr16c @gol
753 -msim -mint32 -mbit-ops
754 -mdata-model=@var{model}}
757 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} @gol
758 -mbig-endian -EB -mlittle-endian -EL @gol
759 -mhard-float -msoft-float -mfpu=@var{fpu} -mdouble-float -mfdivdu @gol
760 -melrw -mistack -mmp -mcp -mcache -msecurity -mtrust @gol
761 -mdsp -medsp -mvdsp @gol
762 -mdiv -msmart -mhigh-registers -manchor @gol
763 -mpushpop -mmultiple-stld -mconstpool -mstack-size -mccrt @gol
764 -mbranch-cost=@var{n} -mcse-cc -msched-prolog}
766 @emph{Darwin Options}
767 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
768 -arch_only -bind_at_load -bundle -bundle_loader @gol
769 -client_name -compatibility_version -current_version @gol
771 -dependency-file -dylib_file -dylinker_install_name @gol
772 -dynamic -dynamiclib -exported_symbols_list @gol
773 -filelist -flat_namespace -force_cpusubtype_ALL @gol
774 -force_flat_namespace -headerpad_max_install_names @gol
776 -image_base -init -install_name -keep_private_externs @gol
777 -multi_module -multiply_defined -multiply_defined_unused @gol
778 -noall_load -no_dead_strip_inits_and_terms @gol
779 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
780 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
781 -private_bundle -read_only_relocs -sectalign @gol
782 -sectobjectsymbols -whyload -seg1addr @gol
783 -sectcreate -sectobjectsymbols -sectorder @gol
784 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
785 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
786 -segprot -segs_read_only_addr -segs_read_write_addr @gol
787 -single_module -static -sub_library -sub_umbrella @gol
788 -twolevel_namespace -umbrella -undefined @gol
789 -unexported_symbols_list -weak_reference_mismatches @gol
790 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
791 -mkernel -mone-byte-bool}
793 @emph{DEC Alpha Options}
794 @gccoptlist{-mno-fp-regs -msoft-float @gol
795 -mieee -mieee-with-inexact -mieee-conformant @gol
796 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
797 -mtrap-precision=@var{mode} -mbuild-constants @gol
798 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
799 -mbwx -mmax -mfix -mcix @gol
800 -mfloat-vax -mfloat-ieee @gol
801 -mexplicit-relocs -msmall-data -mlarge-data @gol
802 -msmall-text -mlarge-text @gol
803 -mmemory-latency=@var{time}}
806 @gccoptlist{-msmall-model -mno-lsim}
809 @gccoptlist{-msim -mlra -mnodiv -mft32b -mcompress -mnopm}
812 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
813 -mhard-float -msoft-float @gol
814 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
815 -mdouble -mno-double @gol
816 -mmedia -mno-media -mmuladd -mno-muladd @gol
817 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
818 -mlinked-fp -mlong-calls -malign-labels @gol
819 -mlibrary-pic -macc-4 -macc-8 @gol
820 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
821 -moptimize-membar -mno-optimize-membar @gol
822 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
823 -mvliw-branch -mno-vliw-branch @gol
824 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
825 -mno-nested-cond-exec -mtomcat-stats @gol
829 @emph{GNU/Linux Options}
830 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
831 -tno-android-cc -tno-android-ld}
833 @emph{H8/300 Options}
834 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
837 @gccoptlist{-march=@var{architecture-type} @gol
838 -mcaller-copies -mdisable-fpregs -mdisable-indexing @gol
839 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
840 -mfixed-range=@var{register-range} @gol
841 -mjump-in-delay -mlinker-opt -mlong-calls @gol
842 -mlong-load-store -mno-disable-fpregs @gol
843 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
844 -mno-jump-in-delay -mno-long-load-store @gol
845 -mno-portable-runtime -mno-soft-float @gol
846 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
847 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
848 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
849 -munix=@var{unix-std} -nolibdld -static -threads}
852 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
853 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
854 -mconstant-gp -mauto-pic -mfused-madd @gol
855 -minline-float-divide-min-latency @gol
856 -minline-float-divide-max-throughput @gol
857 -mno-inline-float-divide @gol
858 -minline-int-divide-min-latency @gol
859 -minline-int-divide-max-throughput @gol
860 -mno-inline-int-divide @gol
861 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
862 -mno-inline-sqrt @gol
863 -mdwarf2-asm -mearly-stop-bits @gol
864 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
865 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
866 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
867 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
868 -msched-spec-ldc -msched-spec-control-ldc @gol
869 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
870 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
871 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
872 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
875 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
876 -msign-extend-enabled -muser-enabled}
878 @emph{M32R/D Options}
879 @gccoptlist{-m32r2 -m32rx -m32r @gol
881 -malign-loops -mno-align-loops @gol
882 -missue-rate=@var{number} @gol
883 -mbranch-cost=@var{number} @gol
884 -mmodel=@var{code-size-model-type} @gol
885 -msdata=@var{sdata-type} @gol
886 -mno-flush-func -mflush-func=@var{name} @gol
887 -mno-flush-trap -mflush-trap=@var{number} @gol
891 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
893 @emph{M680x0 Options}
894 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
895 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
896 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
897 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
898 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
899 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
900 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
901 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
902 -mxgot -mno-xgot -mlong-jump-table-offsets}
905 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
906 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
907 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
908 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
909 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
912 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
913 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
914 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
915 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
918 @emph{MicroBlaze Options}
919 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
920 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
921 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
922 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
923 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model} @gol
924 -mpic-data-is-text-relative}
927 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
928 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
929 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
930 -mips16 -mno-mips16 -mflip-mips16 @gol
931 -minterlink-compressed -mno-interlink-compressed @gol
932 -minterlink-mips16 -mno-interlink-mips16 @gol
933 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
934 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
935 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
936 -mno-float -msingle-float -mdouble-float @gol
937 -modd-spreg -mno-odd-spreg @gol
938 -mabs=@var{mode} -mnan=@var{encoding} @gol
939 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
942 -mvirt -mno-virt @gol
945 -mginv -mno-ginv @gol
946 -mmicromips -mno-micromips @gol
948 -mloongson-mmi -mno-loongson-mmi @gol
949 -mloongson-ext -mno-loongson-ext @gol
950 -mloongson-ext2 -mno-loongson-ext2 @gol
951 -mfpu=@var{fpu-type} @gol
952 -msmartmips -mno-smartmips @gol
953 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
954 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
955 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
956 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
957 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
958 -membedded-data -mno-embedded-data @gol
959 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
960 -mcode-readable=@var{setting} @gol
961 -msplit-addresses -mno-split-addresses @gol
962 -mexplicit-relocs -mno-explicit-relocs @gol
963 -mcheck-zero-division -mno-check-zero-division @gol
964 -mdivide-traps -mdivide-breaks @gol
965 -mload-store-pairs -mno-load-store-pairs @gol
966 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
967 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
968 -mfix-24k -mno-fix-24k @gol
969 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
970 -mfix-r5900 -mno-fix-r5900 @gol
971 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
972 -mfix-vr4120 -mno-fix-vr4120 @gol
973 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
974 -mflush-func=@var{func} -mno-flush-func @gol
975 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
976 -mcompact-branches=@var{policy} @gol
977 -mfp-exceptions -mno-fp-exceptions @gol
978 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
979 -mlxc1-sxc1 -mno-lxc1-sxc1 -mmadd4 -mno-madd4 @gol
980 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address @gol
981 -mframe-header-opt -mno-frame-header-opt}
984 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
985 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
986 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
987 -mno-base-addresses -msingle-exit -mno-single-exit}
989 @emph{MN10300 Options}
990 @gccoptlist{-mmult-bug -mno-mult-bug @gol
991 -mno-am33 -mam33 -mam33-2 -mam34 @gol
992 -mtune=@var{cpu-type} @gol
993 -mreturn-pointer-on-d0 @gol
994 -mno-crt0 -mrelax -mliw -msetlb}
997 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
999 @emph{MSP430 Options}
1000 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
1002 -mcode-region= -mdata-region= @gol
1003 -msilicon-errata= -msilicon-errata-warn= @gol
1006 @emph{NDS32 Options}
1007 @gccoptlist{-mbig-endian -mlittle-endian @gol
1008 -mreduced-regs -mfull-regs @gol
1009 -mcmov -mno-cmov @gol
1010 -mext-perf -mno-ext-perf @gol
1011 -mext-perf2 -mno-ext-perf2 @gol
1012 -mext-string -mno-ext-string @gol
1013 -mv3push -mno-v3push @gol
1014 -m16bit -mno-16bit @gol
1015 -misr-vector-size=@var{num} @gol
1016 -mcache-block-size=@var{num} @gol
1017 -march=@var{arch} @gol
1018 -mcmodel=@var{code-model} @gol
1019 -mctor-dtor -mrelax}
1021 @emph{Nios II Options}
1022 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
1023 -mgprel-sec=@var{regexp} -mr0rel-sec=@var{regexp} @gol
1025 -mno-bypass-cache -mbypass-cache @gol
1026 -mno-cache-volatile -mcache-volatile @gol
1027 -mno-fast-sw-div -mfast-sw-div @gol
1028 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
1029 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
1030 -mcustom-fpu-cfg=@var{name} @gol
1031 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
1032 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
1034 @emph{Nvidia PTX Options}
1035 @gccoptlist{-m32 -m64 -mmainkernel -moptimize}
1037 @emph{OpenRISC Options}
1038 @gccoptlist{-mboard=@var{name} -mnewlib -mhard-mul -mhard-div @gol
1039 -msoft-mul -msoft-div @gol
1040 -msoft-float -mhard-float -mdouble-float -munordered-float @gol
1041 -mcmov -mror -mrori -msext -msfimm -mshftimm}
1043 @emph{PDP-11 Options}
1044 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
1045 -mint32 -mno-int16 -mint16 -mno-int32 @gol
1046 -msplit -munix-asm -mdec-asm -mgnu-asm -mlra}
1048 @emph{picoChip Options}
1049 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
1050 -msymbol-as-address -mno-inefficient-warnings}
1052 @emph{PowerPC Options}
1053 See RS/6000 and PowerPC Options.
1056 @gccoptlist{-mmcu=@var{mcu} -minrt -mno-relax -mloop @gol
1057 -mabi=@var{variant} @gol}
1059 @emph{RISC-V Options}
1060 @gccoptlist{-mbranch-cost=@var{N-instruction} @gol
1062 -mabi=@var{ABI-string} @gol
1063 -mfdiv -mno-fdiv @gol
1065 -march=@var{ISA-string} @gol
1066 -mtune=@var{processor-string} @gol
1067 -mpreferred-stack-boundary=@var{num} @gol
1068 -msmall-data-limit=@var{N-bytes} @gol
1069 -msave-restore -mno-save-restore @gol
1070 -mstrict-align -mno-strict-align @gol
1071 -mcmodel=medlow -mcmodel=medany @gol
1072 -mexplicit-relocs -mno-explicit-relocs @gol
1073 -mrelax -mno-relax @gol
1074 -mriscv-attribute -mmo-riscv-attribute @gol
1075 -malign-data=@var{type}}
1078 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
1079 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
1080 -m64bit-doubles -m32bit-doubles -msave-mduc-in-interrupts}
1082 @emph{RS/6000 and PowerPC Options}
1083 @gccoptlist{-mcpu=@var{cpu-type} @gol
1084 -mtune=@var{cpu-type} @gol
1085 -mcmodel=@var{code-model} @gol
1087 -maltivec -mno-altivec @gol
1088 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
1089 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
1090 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
1091 -mfprnd -mno-fprnd @gol
1092 -mcmpb -mno-cmpb -mhard-dfp -mno-hard-dfp @gol
1093 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
1094 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
1095 -malign-power -malign-natural @gol
1096 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
1097 -mupdate -mno-update @gol
1098 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
1099 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
1100 -mstrict-align -mno-strict-align -mrelocatable @gol
1101 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
1102 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
1103 -mdynamic-no-pic -mswdiv -msingle-pic-base @gol
1104 -mprioritize-restricted-insns=@var{priority} @gol
1105 -msched-costly-dep=@var{dependence_type} @gol
1106 -minsert-sched-nops=@var{scheme} @gol
1107 -mcall-aixdesc -mcall-eabi -mcall-freebsd @gol
1108 -mcall-linux -mcall-netbsd -mcall-openbsd @gol
1109 -mcall-sysv -mcall-sysv-eabi -mcall-sysv-noeabi @gol
1110 -mtraceback=@var{traceback_type} @gol
1111 -maix-struct-return -msvr4-struct-return @gol
1112 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
1113 -mlongcall -mno-longcall -mpltseq -mno-pltseq @gol
1114 -mblock-move-inline-limit=@var{num} @gol
1115 -mblock-compare-inline-limit=@var{num} @gol
1116 -mblock-compare-inline-loop-limit=@var{num} @gol
1117 -mstring-compare-inline-limit=@var{num} @gol
1118 -misel -mno-isel @gol
1119 -mvrsave -mno-vrsave @gol
1120 -mmulhw -mno-mulhw @gol
1121 -mdlmzb -mno-dlmzb @gol
1122 -mprototype -mno-prototype @gol
1123 -msim -mmvme -mads -myellowknife -memb -msdata @gol
1124 -msdata=@var{opt} -mreadonly-in-sdata -mvxworks -G @var{num} @gol
1125 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
1126 -mno-recip-precision @gol
1127 -mveclibabi=@var{type} -mfriz -mno-friz @gol
1128 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
1129 -msave-toc-indirect -mno-save-toc-indirect @gol
1130 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
1131 -mcrypto -mno-crypto -mhtm -mno-htm @gol
1132 -mquad-memory -mno-quad-memory @gol
1133 -mquad-memory-atomic -mno-quad-memory-atomic @gol
1134 -mcompat-align-parm -mno-compat-align-parm @gol
1135 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol
1136 -mgnu-attribute -mno-gnu-attribute @gol
1137 -mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{reg} @gol
1138 -mstack-protector-guard-offset=@var{offset} -mpcrel -mno-pcrel}
1141 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
1143 -mbig-endian-data -mlittle-endian-data @gol
1146 -mas100-syntax -mno-as100-syntax@gol
1148 -mmax-constant-size=@gol
1151 -mallow-string-insns -mno-allow-string-insns@gol
1153 -mno-warn-multiple-fast-interrupts@gol
1154 -msave-acc-in-interrupts}
1156 @emph{S/390 and zSeries Options}
1157 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1158 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
1159 -mlong-double-64 -mlong-double-128 @gol
1160 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
1161 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
1162 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
1163 -mhtm -mvx -mzvector @gol
1164 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
1165 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
1166 -mhotpatch=@var{halfwords},@var{halfwords}}
1168 @emph{Score Options}
1169 @gccoptlist{-meb -mel @gol
1173 -mscore5 -mscore5u -mscore7 -mscore7d}
1176 @gccoptlist{-m1 -m2 -m2e @gol
1177 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
1179 -m4-nofpu -m4-single-only -m4-single -m4 @gol
1180 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
1181 -mb -ml -mdalign -mrelax @gol
1182 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
1183 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
1184 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1185 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1186 -maccumulate-outgoing-args @gol
1187 -matomic-model=@var{atomic-model} @gol
1188 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1189 -mcbranch-force-delay-slot @gol
1190 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1191 -mpretend-cmove -mtas}
1193 @emph{Solaris 2 Options}
1194 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1197 @emph{SPARC Options}
1198 @gccoptlist{-mcpu=@var{cpu-type} @gol
1199 -mtune=@var{cpu-type} @gol
1200 -mcmodel=@var{code-model} @gol
1201 -mmemory-model=@var{mem-model} @gol
1202 -m32 -m64 -mapp-regs -mno-app-regs @gol
1203 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1204 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1205 -mhard-quad-float -msoft-quad-float @gol
1206 -mstack-bias -mno-stack-bias @gol
1207 -mstd-struct-return -mno-std-struct-return @gol
1208 -munaligned-doubles -mno-unaligned-doubles @gol
1209 -muser-mode -mno-user-mode @gol
1210 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1211 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1212 -mvis4 -mno-vis4 -mvis4b -mno-vis4b @gol
1213 -mcbcond -mno-cbcond -mfmaf -mno-fmaf -mfsmuld -mno-fsmuld @gol
1214 -mpopc -mno-popc -msubxc -mno-subxc @gol
1215 -mfix-at697f -mfix-ut699 -mfix-ut700 -mfix-gr712rc @gol
1218 @emph{System V Options}
1219 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1221 @emph{TILE-Gx Options}
1222 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1223 -mcmodel=@var{code-model}}
1225 @emph{TILEPro Options}
1226 @gccoptlist{-mcpu=@var{cpu} -m32}
1229 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1230 -mprolog-function -mno-prolog-function -mspace @gol
1231 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1232 -mapp-regs -mno-app-regs @gol
1233 -mdisable-callt -mno-disable-callt @gol
1234 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1235 -mv850e -mv850 -mv850e3v5 @gol
1246 @gccoptlist{-mg -mgnu -munix}
1248 @emph{Visium Options}
1249 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1250 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1253 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1254 -mpointer-size=@var{size}}
1256 @emph{VxWorks Options}
1257 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1258 -Xbind-lazy -Xbind-now}
1261 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1262 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1263 -mfpmath=@var{unit} @gol
1264 -masm=@var{dialect} -mno-fancy-math-387 @gol
1265 -mno-fp-ret-in-387 -m80387 -mhard-float -msoft-float @gol
1266 -mno-wide-multiply -mrtd -malign-double @gol
1267 -mpreferred-stack-boundary=@var{num} @gol
1268 -mincoming-stack-boundary=@var{num} @gol
1269 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1270 -mrecip -mrecip=@var{opt} @gol
1271 -mvzeroupper -mprefer-avx128 -mprefer-vector-width=@var{opt} @gol
1272 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1273 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1274 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1275 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma -mpconfig -mwbnoinvd @gol
1276 -mptwrite -mprefetchwt1 -mclflushopt -mclwb -mxsavec -mxsaves @gol
1277 -msse4a -m3dnow -m3dnowa -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop @gol
1278 -madx -mlzcnt -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mhle -mlwp @gol
1279 -mmwaitx -mclzero -mpku -mthreads -mgfni -mvaes -mwaitpkg @gol
1280 -mshstk -mmanual-endbr -mforce-indirect-call -mavx512vbmi2 -mavx512bf16 -menqcmd @gol
1281 -mvpclmulqdq -mavx512bitalg -mmovdiri -mmovdir64b -mavx512vpopcntdq @gol
1282 -mavx5124fmaps -mavx512vnni -mavx5124vnniw -mprfchw -mrdpid @gol
1283 -mrdseed -msgx -mavx512vp2intersect@gol
1284 -mcldemote -mms-bitfields -mno-align-stringops -minline-all-stringops @gol
1285 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1286 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1287 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1288 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1289 -mregparm=@var{num} -msseregparm @gol
1290 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1291 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1292 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1293 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1294 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1295 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1296 -minstrument-return=@var{type} -mfentry-name=@var{name} -mfentry-section=@var{name} @gol
1297 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1298 -malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol
1299 -mstack-protector-guard-reg=@var{reg} @gol
1300 -mstack-protector-guard-offset=@var{offset} @gol
1301 -mstack-protector-guard-symbol=@var{symbol} @gol
1302 -mgeneral-regs-only -mcall-ms2sysv-xlogues @gol
1303 -mindirect-branch=@var{choice} -mfunction-return=@var{choice} @gol
1304 -mindirect-branch-register}
1306 @emph{x86 Windows Options}
1307 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1308 -mnop-fun-dllimport -mthread @gol
1309 -municode -mwin32 -mwindows -fno-set-stack-executable}
1311 @emph{Xstormy16 Options}
1314 @emph{Xtensa Options}
1315 @gccoptlist{-mconst16 -mno-const16 @gol
1316 -mfused-madd -mno-fused-madd @gol
1318 -mserialize-volatile -mno-serialize-volatile @gol
1319 -mtext-section-literals -mno-text-section-literals @gol
1320 -mauto-litpools -mno-auto-litpools @gol
1321 -mtarget-align -mno-target-align @gol
1322 -mlongcalls -mno-longcalls}
1324 @emph{zSeries Options}
1325 See S/390 and zSeries Options.
1329 @node Overall Options
1330 @section Options Controlling the Kind of Output
1332 Compilation can involve up to four stages: preprocessing, compilation
1333 proper, assembly and linking, always in that order. GCC is capable of
1334 preprocessing and compiling several files either into several
1335 assembler input files, or into one assembler input file; then each
1336 assembler input file produces an object file, and linking combines all
1337 the object files (those newly compiled, and those specified as input)
1338 into an executable file.
1340 @cindex file name suffix
1341 For any given input file, the file name suffix determines what kind of
1342 compilation is done:
1346 C source code that must be preprocessed.
1349 C source code that should not be preprocessed.
1352 C++ source code that should not be preprocessed.
1355 Objective-C source code. Note that you must link with the @file{libobjc}
1356 library to make an Objective-C program work.
1359 Objective-C source code that should not be preprocessed.
1363 Objective-C++ source code. Note that you must link with the @file{libobjc}
1364 library to make an Objective-C++ program work. Note that @samp{.M} refers
1365 to a literal capital M@.
1367 @item @var{file}.mii
1368 Objective-C++ source code that should not be preprocessed.
1371 C, C++, Objective-C or Objective-C++ header file to be turned into a
1372 precompiled header (default), or C, C++ header file to be turned into an
1373 Ada spec (via the @option{-fdump-ada-spec} switch).
1376 @itemx @var{file}.cp
1377 @itemx @var{file}.cxx
1378 @itemx @var{file}.cpp
1379 @itemx @var{file}.CPP
1380 @itemx @var{file}.c++
1382 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1383 the last two letters must both be literally @samp{x}. Likewise,
1384 @samp{.C} refers to a literal capital C@.
1388 Objective-C++ source code that must be preprocessed.
1390 @item @var{file}.mii
1391 Objective-C++ source code that should not be preprocessed.
1395 @itemx @var{file}.hp
1396 @itemx @var{file}.hxx
1397 @itemx @var{file}.hpp
1398 @itemx @var{file}.HPP
1399 @itemx @var{file}.h++
1400 @itemx @var{file}.tcc
1401 C++ header file to be turned into a precompiled header or Ada spec.
1404 @itemx @var{file}.for
1405 @itemx @var{file}.ftn
1406 Fixed form Fortran source code that should not be preprocessed.
1409 @itemx @var{file}.FOR
1410 @itemx @var{file}.fpp
1411 @itemx @var{file}.FPP
1412 @itemx @var{file}.FTN
1413 Fixed form Fortran source code that must be preprocessed (with the traditional
1416 @item @var{file}.f90
1417 @itemx @var{file}.f95
1418 @itemx @var{file}.f03
1419 @itemx @var{file}.f08
1420 Free form Fortran source code that should not be preprocessed.
1422 @item @var{file}.F90
1423 @itemx @var{file}.F95
1424 @itemx @var{file}.F03
1425 @itemx @var{file}.F08
1426 Free form Fortran source code that must be preprocessed (with the
1427 traditional preprocessor).
1432 @item @var{file}.brig
1433 BRIG files (binary representation of HSAIL).
1442 D documentation code (Ddoc).
1444 @item @var{file}.ads
1445 Ada source code file that contains a library unit declaration (a
1446 declaration of a package, subprogram, or generic, or a generic
1447 instantiation), or a library unit renaming declaration (a package,
1448 generic, or subprogram renaming declaration). Such files are also
1451 @item @var{file}.adb
1452 Ada source code file containing a library unit body (a subprogram or
1453 package body). Such files are also called @dfn{bodies}.
1455 @c GCC also knows about some suffixes for languages not yet included:
1463 @itemx @var{file}.sx
1464 Assembler code that must be preprocessed.
1467 An object file to be fed straight into linking.
1468 Any file name with no recognized suffix is treated this way.
1472 You can specify the input language explicitly with the @option{-x} option:
1475 @item -x @var{language}
1476 Specify explicitly the @var{language} for the following input files
1477 (rather than letting the compiler choose a default based on the file
1478 name suffix). This option applies to all following input files until
1479 the next @option{-x} option. Possible values for @var{language} are:
1481 c c-header cpp-output
1482 c++ c++-header c++-cpp-output
1483 objective-c objective-c-header objective-c-cpp-output
1484 objective-c++ objective-c++-header objective-c++-cpp-output
1485 assembler assembler-with-cpp
1488 f77 f77-cpp-input f95 f95-cpp-input
1494 Turn off any specification of a language, so that subsequent files are
1495 handled according to their file name suffixes (as they are if @option{-x}
1496 has not been used at all).
1499 If you only want some of the stages of compilation, you can use
1500 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1501 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1502 @command{gcc} is to stop. Note that some combinations (for example,
1503 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1508 Compile or assemble the source files, but do not link. The linking
1509 stage simply is not done. The ultimate output is in the form of an
1510 object file for each source file.
1512 By default, the object file name for a source file is made by replacing
1513 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1515 Unrecognized input files, not requiring compilation or assembly, are
1520 Stop after the stage of compilation proper; do not assemble. The output
1521 is in the form of an assembler code file for each non-assembler input
1524 By default, the assembler file name for a source file is made by
1525 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1527 Input files that don't require compilation are ignored.
1531 Stop after the preprocessing stage; do not run the compiler proper. The
1532 output is in the form of preprocessed source code, which is sent to the
1535 Input files that don't require preprocessing are ignored.
1537 @cindex output file option
1540 Place output in file @var{file}. This applies to whatever
1541 sort of output is being produced, whether it be an executable file,
1542 an object file, an assembler file or preprocessed C code.
1544 If @option{-o} is not specified, the default is to put an executable
1545 file in @file{a.out}, the object file for
1546 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1547 assembler file in @file{@var{source}.s}, a precompiled header file in
1548 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1553 Print (on standard error output) the commands executed to run the stages
1554 of compilation. Also print the version number of the compiler driver
1555 program and of the preprocessor and the compiler proper.
1559 Like @option{-v} except the commands are not executed and arguments
1560 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1561 This is useful for shell scripts to capture the driver-generated command lines.
1565 Print (on the standard output) a description of the command-line options
1566 understood by @command{gcc}. If the @option{-v} option is also specified
1567 then @option{--help} is also passed on to the various processes
1568 invoked by @command{gcc}, so that they can display the command-line options
1569 they accept. If the @option{-Wextra} option has also been specified
1570 (prior to the @option{--help} option), then command-line options that
1571 have no documentation associated with them are also displayed.
1574 @opindex target-help
1575 Print (on the standard output) a description of target-specific command-line
1576 options for each tool. For some targets extra target-specific
1577 information may also be printed.
1579 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1580 Print (on the standard output) a description of the command-line
1581 options understood by the compiler that fit into all specified classes
1582 and qualifiers. These are the supported classes:
1585 @item @samp{optimizers}
1586 Display all of the optimization options supported by the
1589 @item @samp{warnings}
1590 Display all of the options controlling warning messages
1591 produced by the compiler.
1594 Display target-specific options. Unlike the
1595 @option{--target-help} option however, target-specific options of the
1596 linker and assembler are not displayed. This is because those
1597 tools do not currently support the extended @option{--help=} syntax.
1600 Display the values recognized by the @option{--param}
1603 @item @var{language}
1604 Display the options supported for @var{language}, where
1605 @var{language} is the name of one of the languages supported in this
1609 Display the options that are common to all languages.
1612 These are the supported qualifiers:
1615 @item @samp{undocumented}
1616 Display only those options that are undocumented.
1619 Display options taking an argument that appears after an equal
1620 sign in the same continuous piece of text, such as:
1621 @samp{--help=target}.
1623 @item @samp{separate}
1624 Display options taking an argument that appears as a separate word
1625 following the original option, such as: @samp{-o output-file}.
1628 Thus for example to display all the undocumented target-specific
1629 switches supported by the compiler, use:
1632 --help=target,undocumented
1635 The sense of a qualifier can be inverted by prefixing it with the
1636 @samp{^} character, so for example to display all binary warning
1637 options (i.e., ones that are either on or off and that do not take an
1638 argument) that have a description, use:
1641 --help=warnings,^joined,^undocumented
1644 The argument to @option{--help=} should not consist solely of inverted
1647 Combining several classes is possible, although this usually
1648 restricts the output so much that there is nothing to display. One
1649 case where it does work, however, is when one of the classes is
1650 @var{target}. For example, to display all the target-specific
1651 optimization options, use:
1654 --help=target,optimizers
1657 The @option{--help=} option can be repeated on the command line. Each
1658 successive use displays its requested class of options, skipping
1659 those that have already been displayed. If @option{--help} is also
1660 specified anywhere on the command line then this takes precedence
1661 over any @option{--help=} option.
1663 If the @option{-Q} option appears on the command line before the
1664 @option{--help=} option, then the descriptive text displayed by
1665 @option{--help=} is changed. Instead of describing the displayed
1666 options, an indication is given as to whether the option is enabled,
1667 disabled or set to a specific value (assuming that the compiler
1668 knows this at the point where the @option{--help=} option is used).
1670 Here is a truncated example from the ARM port of @command{gcc}:
1673 % gcc -Q -mabi=2 --help=target -c
1674 The following options are target specific:
1676 -mabort-on-noreturn [disabled]
1680 The output is sensitive to the effects of previous command-line
1681 options, so for example it is possible to find out which optimizations
1682 are enabled at @option{-O2} by using:
1685 -Q -O2 --help=optimizers
1688 Alternatively you can discover which binary optimizations are enabled
1689 by @option{-O3} by using:
1692 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1693 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1694 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1699 Display the version number and copyrights of the invoked GCC@.
1701 @item -pass-exit-codes
1702 @opindex pass-exit-codes
1703 Normally the @command{gcc} program exits with the code of 1 if any
1704 phase of the compiler returns a non-success return code. If you specify
1705 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1706 the numerically highest error produced by any phase returning an error
1707 indication. The C, C++, and Fortran front ends return 4 if an internal
1708 compiler error is encountered.
1712 Use pipes rather than temporary files for communication between the
1713 various stages of compilation. This fails to work on some systems where
1714 the assembler is unable to read from a pipe; but the GNU assembler has
1717 @item -specs=@var{file}
1719 Process @var{file} after the compiler reads in the standard @file{specs}
1720 file, in order to override the defaults which the @command{gcc} driver
1721 program uses when determining what switches to pass to @command{cc1},
1722 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
1723 @option{-specs=@var{file}} can be specified on the command line, and they
1724 are processed in order, from left to right. @xref{Spec Files}, for
1725 information about the format of the @var{file}.
1729 Invoke all subcommands under a wrapper program. The name of the
1730 wrapper program and its parameters are passed as a comma separated
1734 gcc -c t.c -wrapper gdb,--args
1738 This invokes all subprograms of @command{gcc} under
1739 @samp{gdb --args}, thus the invocation of @command{cc1} is
1740 @samp{gdb --args cc1 @dots{}}.
1742 @item -ffile-prefix-map=@var{old}=@var{new}
1743 @opindex ffile-prefix-map
1744 When compiling files residing in directory @file{@var{old}}, record
1745 any references to them in the result of the compilation as if the
1746 files resided in directory @file{@var{new}} instead. Specifying this
1747 option is equivalent to specifying all the individual
1748 @option{-f*-prefix-map} options. This can be used to make reproducible
1749 builds that are location independent. See also
1750 @option{-fmacro-prefix-map} and @option{-fdebug-prefix-map}.
1752 @item -fplugin=@var{name}.so
1754 Load the plugin code in file @var{name}.so, assumed to be a
1755 shared object to be dlopen'd by the compiler. The base name of
1756 the shared object file is used to identify the plugin for the
1757 purposes of argument parsing (See
1758 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1759 Each plugin should define the callback functions specified in the
1762 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1763 @opindex fplugin-arg
1764 Define an argument called @var{key} with a value of @var{value}
1765 for the plugin called @var{name}.
1767 @item -fdump-ada-spec@r{[}-slim@r{]}
1768 @opindex fdump-ada-spec
1769 For C and C++ source and include files, generate corresponding Ada specs.
1770 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1771 GNAT User's Guide}, which provides detailed documentation on this feature.
1773 @item -fada-spec-parent=@var{unit}
1774 @opindex fada-spec-parent
1775 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1776 Ada specs as child units of parent @var{unit}.
1778 @item -fdump-go-spec=@var{file}
1779 @opindex fdump-go-spec
1780 For input files in any language, generate corresponding Go
1781 declarations in @var{file}. This generates Go @code{const},
1782 @code{type}, @code{var}, and @code{func} declarations which may be a
1783 useful way to start writing a Go interface to code written in some
1786 @include @value{srcdir}/../libiberty/at-file.texi
1790 @section Compiling C++ Programs
1792 @cindex suffixes for C++ source
1793 @cindex C++ source file suffixes
1794 C++ source files conventionally use one of the suffixes @samp{.C},
1795 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1796 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1797 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1798 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1799 files with these names and compiles them as C++ programs even if you
1800 call the compiler the same way as for compiling C programs (usually
1801 with the name @command{gcc}).
1805 However, the use of @command{gcc} does not add the C++ library.
1806 @command{g++} is a program that calls GCC and automatically specifies linking
1807 against the C++ library. It treats @samp{.c},
1808 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1809 files unless @option{-x} is used. This program is also useful when
1810 precompiling a C header file with a @samp{.h} extension for use in C++
1811 compilations. On many systems, @command{g++} is also installed with
1812 the name @command{c++}.
1814 @cindex invoking @command{g++}
1815 When you compile C++ programs, you may specify many of the same
1816 command-line options that you use for compiling programs in any
1817 language; or command-line options meaningful for C and related
1818 languages; or options that are meaningful only for C++ programs.
1819 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1820 explanations of options for languages related to C@.
1821 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1822 explanations of options that are meaningful only for C++ programs.
1824 @node C Dialect Options
1825 @section Options Controlling C Dialect
1826 @cindex dialect options
1827 @cindex language dialect options
1828 @cindex options, dialect
1830 The following options control the dialect of C (or languages derived
1831 from C, such as C++, Objective-C and Objective-C++) that the compiler
1835 @cindex ANSI support
1839 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1840 equivalent to @option{-std=c++98}.
1842 This turns off certain features of GCC that are incompatible with ISO
1843 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1844 such as the @code{asm} and @code{typeof} keywords, and
1845 predefined macros such as @code{unix} and @code{vax} that identify the
1846 type of system you are using. It also enables the undesirable and
1847 rarely used ISO trigraph feature. For the C compiler,
1848 it disables recognition of C++ style @samp{//} comments as well as
1849 the @code{inline} keyword.
1851 The alternate keywords @code{__asm__}, @code{__extension__},
1852 @code{__inline__} and @code{__typeof__} continue to work despite
1853 @option{-ansi}. You would not want to use them in an ISO C program, of
1854 course, but it is useful to put them in header files that might be included
1855 in compilations done with @option{-ansi}. Alternate predefined macros
1856 such as @code{__unix__} and @code{__vax__} are also available, with or
1857 without @option{-ansi}.
1859 The @option{-ansi} option does not cause non-ISO programs to be
1860 rejected gratuitously. For that, @option{-Wpedantic} is required in
1861 addition to @option{-ansi}. @xref{Warning Options}.
1863 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1864 option is used. Some header files may notice this macro and refrain
1865 from declaring certain functions or defining certain macros that the
1866 ISO standard doesn't call for; this is to avoid interfering with any
1867 programs that might use these names for other things.
1869 Functions that are normally built in but do not have semantics
1870 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1871 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1872 built-in functions provided by GCC}, for details of the functions
1877 Determine the language standard. @xref{Standards,,Language Standards
1878 Supported by GCC}, for details of these standard versions. This option
1879 is currently only supported when compiling C or C++.
1881 The compiler can accept several base standards, such as @samp{c90} or
1882 @samp{c++98}, and GNU dialects of those standards, such as
1883 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1884 compiler accepts all programs following that standard plus those
1885 using GNU extensions that do not contradict it. For example,
1886 @option{-std=c90} turns off certain features of GCC that are
1887 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1888 keywords, but not other GNU extensions that do not have a meaning in
1889 ISO C90, such as omitting the middle term of a @code{?:}
1890 expression. On the other hand, when a GNU dialect of a standard is
1891 specified, all features supported by the compiler are enabled, even when
1892 those features change the meaning of the base standard. As a result, some
1893 strict-conforming programs may be rejected. The particular standard
1894 is used by @option{-Wpedantic} to identify which features are GNU
1895 extensions given that version of the standard. For example
1896 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1897 comments, while @option{-std=gnu99 -Wpedantic} does not.
1899 A value for this option must be provided; possible values are
1905 Support all ISO C90 programs (certain GNU extensions that conflict
1906 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1908 @item iso9899:199409
1909 ISO C90 as modified in amendment 1.
1915 ISO C99. This standard is substantially completely supported, modulo
1916 bugs and floating-point issues
1917 (mainly but not entirely relating to optional C99 features from
1918 Annexes F and G). See
1919 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1920 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1925 ISO C11, the 2011 revision of the ISO C standard. This standard is
1926 substantially completely supported, modulo bugs, floating-point issues
1927 (mainly but not entirely relating to optional C11 features from
1928 Annexes F and G) and the optional Annexes K (Bounds-checking
1929 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1935 ISO C17, the 2017 revision of the ISO C standard
1936 (published in 2018). This standard is
1937 same as C11 except for corrections of defects (all of which are also
1938 applied with @option{-std=c11}) and a new value of
1939 @code{__STDC_VERSION__}, and so is supported to the same extent as C11.
1942 The next version of the ISO C standard, still under development. The
1943 support for this version is experimental and incomplete.
1947 GNU dialect of ISO C90 (including some C99 features).
1951 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1955 GNU dialect of ISO C11.
1956 The name @samp{gnu1x} is deprecated.
1960 GNU dialect of ISO C17. This is the default for C code.
1963 The next version of the ISO C standard, still under development, plus
1964 GNU extensions. The support for this version is experimental and
1969 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1970 additional defect reports. Same as @option{-ansi} for C++ code.
1974 GNU dialect of @option{-std=c++98}.
1978 The 2011 ISO C++ standard plus amendments.
1979 The name @samp{c++0x} is deprecated.
1983 GNU dialect of @option{-std=c++11}.
1984 The name @samp{gnu++0x} is deprecated.
1988 The 2014 ISO C++ standard plus amendments.
1989 The name @samp{c++1y} is deprecated.
1993 GNU dialect of @option{-std=c++14}.
1994 This is the default for C++ code.
1995 The name @samp{gnu++1y} is deprecated.
1999 The 2017 ISO C++ standard plus amendments.
2000 The name @samp{c++1z} is deprecated.
2004 GNU dialect of @option{-std=c++17}.
2005 The name @samp{gnu++1z} is deprecated.
2008 The next revision of the ISO C++ standard, tentatively planned for
2009 2020. Support is highly experimental, and will almost certainly
2010 change in incompatible ways in future releases.
2013 GNU dialect of @option{-std=c++2a}. Support is highly experimental,
2014 and will almost certainly change in incompatible ways in future
2018 @item -fgnu89-inline
2019 @opindex fgnu89-inline
2020 The option @option{-fgnu89-inline} tells GCC to use the traditional
2021 GNU semantics for @code{inline} functions when in C99 mode.
2022 @xref{Inline,,An Inline Function is As Fast As a Macro}.
2023 Using this option is roughly equivalent to adding the
2024 @code{gnu_inline} function attribute to all inline functions
2025 (@pxref{Function Attributes}).
2027 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
2028 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
2029 specifies the default behavior).
2030 This option is not supported in @option{-std=c90} or
2031 @option{-std=gnu90} mode.
2033 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
2034 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
2035 in effect for @code{inline} functions. @xref{Common Predefined
2036 Macros,,,cpp,The C Preprocessor}.
2038 @item -fpermitted-flt-eval-methods=@var{style}
2039 @opindex fpermitted-flt-eval-methods
2040 @opindex fpermitted-flt-eval-methods=c11
2041 @opindex fpermitted-flt-eval-methods=ts-18661-3
2042 ISO/IEC TS 18661-3 defines new permissible values for
2043 @code{FLT_EVAL_METHOD} that indicate that operations and constants with
2044 a semantic type that is an interchange or extended format should be
2045 evaluated to the precision and range of that type. These new values are
2046 a superset of those permitted under C99/C11, which does not specify the
2047 meaning of other positive values of @code{FLT_EVAL_METHOD}. As such, code
2048 conforming to C11 may not have been written expecting the possibility of
2051 @option{-fpermitted-flt-eval-methods} specifies whether the compiler
2052 should allow only the values of @code{FLT_EVAL_METHOD} specified in C99/C11,
2053 or the extended set of values specified in ISO/IEC TS 18661-3.
2055 @var{style} is either @code{c11} or @code{ts-18661-3} as appropriate.
2057 The default when in a standards compliant mode (@option{-std=c11} or similar)
2058 is @option{-fpermitted-flt-eval-methods=c11}. The default when in a GNU
2059 dialect (@option{-std=gnu11} or similar) is
2060 @option{-fpermitted-flt-eval-methods=ts-18661-3}.
2062 @item -aux-info @var{filename}
2064 Output to the given filename prototyped declarations for all functions
2065 declared and/or defined in a translation unit, including those in header
2066 files. This option is silently ignored in any language other than C@.
2068 Besides declarations, the file indicates, in comments, the origin of
2069 each declaration (source file and line), whether the declaration was
2070 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
2071 @samp{O} for old, respectively, in the first character after the line
2072 number and the colon), and whether it came from a declaration or a
2073 definition (@samp{C} or @samp{F}, respectively, in the following
2074 character). In the case of function definitions, a K&R-style list of
2075 arguments followed by their declarations is also provided, inside
2076 comments, after the declaration.
2078 @item -fallow-parameterless-variadic-functions
2079 @opindex fallow-parameterless-variadic-functions
2080 Accept variadic functions without named parameters.
2082 Although it is possible to define such a function, this is not very
2083 useful as it is not possible to read the arguments. This is only
2084 supported for C as this construct is allowed by C++.
2089 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
2090 keyword, so that code can use these words as identifiers. You can use
2091 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
2092 instead. @option{-ansi} implies @option{-fno-asm}.
2094 In C++, this switch only affects the @code{typeof} keyword, since
2095 @code{asm} and @code{inline} are standard keywords. You may want to
2096 use the @option{-fno-gnu-keywords} flag instead, which has the same
2097 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
2098 switch only affects the @code{asm} and @code{typeof} keywords, since
2099 @code{inline} is a standard keyword in ISO C99.
2102 @itemx -fno-builtin-@var{function}
2103 @opindex fno-builtin
2105 @cindex built-in functions
2106 Don't recognize built-in functions that do not begin with
2107 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
2108 functions provided by GCC}, for details of the functions affected,
2109 including those which are not built-in functions when @option{-ansi} or
2110 @option{-std} options for strict ISO C conformance are used because they
2111 do not have an ISO standard meaning.
2113 GCC normally generates special code to handle certain built-in functions
2114 more efficiently; for instance, calls to @code{alloca} may become single
2115 instructions which adjust the stack directly, and calls to @code{memcpy}
2116 may become inline copy loops. The resulting code is often both smaller
2117 and faster, but since the function calls no longer appear as such, you
2118 cannot set a breakpoint on those calls, nor can you change the behavior
2119 of the functions by linking with a different library. In addition,
2120 when a function is recognized as a built-in function, GCC may use
2121 information about that function to warn about problems with calls to
2122 that function, or to generate more efficient code, even if the
2123 resulting code still contains calls to that function. For example,
2124 warnings are given with @option{-Wformat} for bad calls to
2125 @code{printf} when @code{printf} is built in and @code{strlen} is
2126 known not to modify global memory.
2128 With the @option{-fno-builtin-@var{function}} option
2129 only the built-in function @var{function} is
2130 disabled. @var{function} must not begin with @samp{__builtin_}. If a
2131 function is named that is not built-in in this version of GCC, this
2132 option is ignored. There is no corresponding
2133 @option{-fbuiltin-@var{function}} option; if you wish to enable
2134 built-in functions selectively when using @option{-fno-builtin} or
2135 @option{-ffreestanding}, you may define macros such as:
2138 #define abs(n) __builtin_abs ((n))
2139 #define strcpy(d, s) __builtin_strcpy ((d), (s))
2145 Enable parsing of function definitions marked with @code{__GIMPLE}.
2146 This is an experimental feature that allows unit testing of GIMPLE
2151 @cindex hosted environment
2153 Assert that compilation targets a hosted environment. This implies
2154 @option{-fbuiltin}. A hosted environment is one in which the
2155 entire standard library is available, and in which @code{main} has a return
2156 type of @code{int}. Examples are nearly everything except a kernel.
2157 This is equivalent to @option{-fno-freestanding}.
2159 @item -ffreestanding
2160 @opindex ffreestanding
2161 @cindex hosted environment
2163 Assert that compilation targets a freestanding environment. This
2164 implies @option{-fno-builtin}. A freestanding environment
2165 is one in which the standard library may not exist, and program startup may
2166 not necessarily be at @code{main}. The most obvious example is an OS kernel.
2167 This is equivalent to @option{-fno-hosted}.
2169 @xref{Standards,,Language Standards Supported by GCC}, for details of
2170 freestanding and hosted environments.
2174 @cindex OpenACC accelerator programming
2175 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
2176 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
2177 compiler generates accelerated code according to the OpenACC Application
2178 Programming Interface v2.0 @w{@uref{https://www.openacc.org}}. This option
2179 implies @option{-pthread}, and thus is only supported on targets that
2180 have support for @option{-pthread}.
2182 @item -fopenacc-dim=@var{geom}
2183 @opindex fopenacc-dim
2184 @cindex OpenACC accelerator programming
2185 Specify default compute dimensions for parallel offload regions that do
2186 not explicitly specify. The @var{geom} value is a triple of
2187 ':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A size
2188 can be omitted, to use a target-specific default value.
2192 @cindex OpenMP parallel
2193 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
2194 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
2195 compiler generates parallel code according to the OpenMP Application
2196 Program Interface v4.5 @w{@uref{https://www.openmp.org}}. This option
2197 implies @option{-pthread}, and thus is only supported on targets that
2198 have support for @option{-pthread}. @option{-fopenmp} implies
2199 @option{-fopenmp-simd}.
2202 @opindex fopenmp-simd
2205 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
2206 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
2211 When the option @option{-fgnu-tm} is specified, the compiler
2212 generates code for the Linux variant of Intel's current Transactional
2213 Memory ABI specification document (Revision 1.1, May 6 2009). This is
2214 an experimental feature whose interface may change in future versions
2215 of GCC, as the official specification changes. Please note that not
2216 all architectures are supported for this feature.
2218 For more information on GCC's support for transactional memory,
2219 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
2220 Transactional Memory Library}.
2222 Note that the transactional memory feature is not supported with
2223 non-call exceptions (@option{-fnon-call-exceptions}).
2225 @item -fms-extensions
2226 @opindex fms-extensions
2227 Accept some non-standard constructs used in Microsoft header files.
2229 In C++ code, this allows member names in structures to be similar
2230 to previous types declarations.
2239 Some cases of unnamed fields in structures and unions are only
2240 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2241 fields within structs/unions}, for details.
2243 Note that this option is off for all targets except for x86
2244 targets using ms-abi.
2246 @item -fplan9-extensions
2247 @opindex fplan9-extensions
2248 Accept some non-standard constructs used in Plan 9 code.
2250 This enables @option{-fms-extensions}, permits passing pointers to
2251 structures with anonymous fields to functions that expect pointers to
2252 elements of the type of the field, and permits referring to anonymous
2253 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2254 struct/union fields within structs/unions}, for details. This is only
2255 supported for C, not C++.
2257 @item -fcond-mismatch
2258 @opindex fcond-mismatch
2259 Allow conditional expressions with mismatched types in the second and
2260 third arguments. The value of such an expression is void. This option
2261 is not supported for C++.
2263 @item -flax-vector-conversions
2264 @opindex flax-vector-conversions
2265 Allow implicit conversions between vectors with differing numbers of
2266 elements and/or incompatible element types. This option should not be
2269 @item -funsigned-char
2270 @opindex funsigned-char
2271 Let the type @code{char} be unsigned, like @code{unsigned char}.
2273 Each kind of machine has a default for what @code{char} should
2274 be. It is either like @code{unsigned char} by default or like
2275 @code{signed char} by default.
2277 Ideally, a portable program should always use @code{signed char} or
2278 @code{unsigned char} when it depends on the signedness of an object.
2279 But many programs have been written to use plain @code{char} and
2280 expect it to be signed, or expect it to be unsigned, depending on the
2281 machines they were written for. This option, and its inverse, let you
2282 make such a program work with the opposite default.
2284 The type @code{char} is always a distinct type from each of
2285 @code{signed char} or @code{unsigned char}, even though its behavior
2286 is always just like one of those two.
2289 @opindex fsigned-char
2290 Let the type @code{char} be signed, like @code{signed char}.
2292 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2293 the negative form of @option{-funsigned-char}. Likewise, the option
2294 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2296 @item -fsigned-bitfields
2297 @itemx -funsigned-bitfields
2298 @itemx -fno-signed-bitfields
2299 @itemx -fno-unsigned-bitfields
2300 @opindex fsigned-bitfields
2301 @opindex funsigned-bitfields
2302 @opindex fno-signed-bitfields
2303 @opindex fno-unsigned-bitfields
2304 These options control whether a bit-field is signed or unsigned, when the
2305 declaration does not use either @code{signed} or @code{unsigned}. By
2306 default, such a bit-field is signed, because this is consistent: the
2307 basic integer types such as @code{int} are signed types.
2309 @item -fsso-struct=@var{endianness}
2310 @opindex fsso-struct
2311 Set the default scalar storage order of structures and unions to the
2312 specified endianness. The accepted values are @samp{big-endian},
2313 @samp{little-endian} and @samp{native} for the native endianness of
2314 the target (the default). This option is not supported for C++.
2316 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2317 code that is not binary compatible with code generated without it if the
2318 specified endianness is not the native endianness of the target.
2321 @node C++ Dialect Options
2322 @section Options Controlling C++ Dialect
2324 @cindex compiler options, C++
2325 @cindex C++ options, command-line
2326 @cindex options, C++
2327 This section describes the command-line options that are only meaningful
2328 for C++ programs. You can also use most of the GNU compiler options
2329 regardless of what language your program is in. For example, you
2330 might compile a file @file{firstClass.C} like this:
2333 g++ -g -fstrict-enums -O -c firstClass.C
2337 In this example, only @option{-fstrict-enums} is an option meant
2338 only for C++ programs; you can use the other options with any
2339 language supported by GCC@.
2341 Some options for compiling C programs, such as @option{-std}, are also
2342 relevant for C++ programs.
2343 @xref{C Dialect Options,,Options Controlling C Dialect}.
2345 Here is a list of options that are @emph{only} for compiling C++ programs:
2349 @item -fabi-version=@var{n}
2350 @opindex fabi-version
2351 Use version @var{n} of the C++ ABI@. The default is version 0.
2353 Version 0 refers to the version conforming most closely to
2354 the C++ ABI specification. Therefore, the ABI obtained using version 0
2355 will change in different versions of G++ as ABI bugs are fixed.
2357 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2359 Version 2 is the version of the C++ ABI that first appeared in G++
2360 3.4, and was the default through G++ 4.9.
2362 Version 3 corrects an error in mangling a constant address as a
2365 Version 4, which first appeared in G++ 4.5, implements a standard
2366 mangling for vector types.
2368 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2369 attribute const/volatile on function pointer types, decltype of a
2370 plain decl, and use of a function parameter in the declaration of
2373 Version 6, which first appeared in G++ 4.7, corrects the promotion
2374 behavior of C++11 scoped enums and the mangling of template argument
2375 packs, const/static_cast, prefix ++ and --, and a class scope function
2376 used as a template argument.
2378 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2379 builtin type and corrects the mangling of lambdas in default argument
2382 Version 8, which first appeared in G++ 4.9, corrects the substitution
2383 behavior of function types with function-cv-qualifiers.
2385 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2388 Version 10, which first appeared in G++ 6.1, adds mangling of
2389 attributes that affect type identity, such as ia32 calling convention
2390 attributes (e.g.@: @samp{stdcall}).
2392 Version 11, which first appeared in G++ 7, corrects the mangling of
2393 sizeof... expressions and operator names. For multiple entities with
2394 the same name within a function, that are declared in different scopes,
2395 the mangling now changes starting with the twelfth occurrence. It also
2396 implies @option{-fnew-inheriting-ctors}.
2398 Version 12, which first appeared in G++ 8, corrects the calling
2399 conventions for empty classes on the x86_64 target and for classes
2400 with only deleted copy/move constructors. It accidentally changes the
2401 calling convention for classes with a deleted copy constructor and a
2402 trivial move constructor.
2404 Version 13, which first appeared in G++ 8.2, fixes the accidental
2405 change in version 12.
2407 See also @option{-Wabi}.
2409 @item -fabi-compat-version=@var{n}
2410 @opindex fabi-compat-version
2411 On targets that support strong aliases, G++
2412 works around mangling changes by creating an alias with the correct
2413 mangled name when defining a symbol with an incorrect mangled name.
2414 This switch specifies which ABI version to use for the alias.
2416 With @option{-fabi-version=0} (the default), this defaults to 11 (GCC 7
2417 compatibility). If another ABI version is explicitly selected, this
2418 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2419 use @option{-fabi-compat-version=2}.
2421 If this option is not provided but @option{-Wabi=@var{n}} is, that
2422 version is used for compatibility aliases. If this option is provided
2423 along with @option{-Wabi} (without the version), the version from this
2424 option is used for the warning.
2426 @item -fno-access-control
2427 @opindex fno-access-control
2428 @opindex faccess-control
2429 Turn off all access checking. This switch is mainly useful for working
2430 around bugs in the access control code.
2433 @opindex faligned-new
2434 Enable support for C++17 @code{new} of types that require more
2435 alignment than @code{void* ::operator new(std::size_t)} provides. A
2436 numeric argument such as @code{-faligned-new=32} can be used to
2437 specify how much alignment (in bytes) is provided by that function,
2438 but few users will need to override the default of
2439 @code{alignof(std::max_align_t)}.
2441 This flag is enabled by default for @option{-std=c++17}.
2446 @opindex fno-char8_t
2447 Enable support for @code{char8_t} as adopted for C++2a. This includes
2448 the addition of a new @code{char8_t} fundamental type, changes to the
2449 types of UTF-8 string and character literals, new signatures for
2450 user-defined literals, associated standard library updates, and new
2451 @code{__cpp_char8_t} and @code{__cpp_lib_char8_t} feature test macros.
2453 This option enables functions to be overloaded for ordinary and UTF-8
2457 int f(const char *); // #1
2458 int f(const char8_t *); // #2
2459 int v1 = f("text"); // Calls #1
2460 int v2 = f(u8"text"); // Calls #2
2464 and introduces new signatures for user-defined literals:
2467 int operator""_udl1(char8_t);
2468 int v3 = u8'x'_udl1;
2469 int operator""_udl2(const char8_t*, std::size_t);
2470 int v4 = u8"text"_udl2;
2471 template<typename T, T...> int operator""_udl3();
2472 int v5 = u8"text"_udl3;
2476 The change to the types of UTF-8 string and character literals introduces
2477 incompatibilities with ISO C++11 and later standards. For example, the
2478 following code is well-formed under ISO C++11, but is ill-formed when
2479 @option{-fchar8_t} is specified.
2482 char ca[] = u8"xx"; // error: char-array initialized from wide
2484 const char *cp = u8"xx";// error: invalid conversion from
2485 // `const char8_t*' to `const char*'
2487 auto v = f(u8"xx"); // error: invalid conversion from
2488 // `const char8_t*' to `const char*'
2489 std::string s@{u8"xx"@}; // error: no matching function for call to
2490 // `std::basic_string<char>::basic_string()'
2491 using namespace std::literals;
2492 s = u8"xx"s; // error: conversion from
2493 // `basic_string<char8_t>' to non-scalar
2494 // type `basic_string<char>' requested
2499 Check that the pointer returned by @code{operator new} is non-null
2500 before attempting to modify the storage allocated. This check is
2501 normally unnecessary because the C++ standard specifies that
2502 @code{operator new} only returns @code{0} if it is declared
2503 @code{throw()}, in which case the compiler always checks the
2504 return value even without this option. In all other cases, when
2505 @code{operator new} has a non-empty exception specification, memory
2506 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2507 @samp{new (nothrow)}.
2511 Enable support for the C++ Extensions for Concepts Technical
2512 Specification, ISO 19217 (2015), which allows code like
2515 template <class T> concept bool Addable = requires (T t) @{ t + t; @};
2516 template <Addable T> T add (T a, T b) @{ return a + b; @}
2519 @item -fconstexpr-depth=@var{n}
2520 @opindex fconstexpr-depth
2521 Set the maximum nested evaluation depth for C++11 constexpr functions
2522 to @var{n}. A limit is needed to detect endless recursion during
2523 constant expression evaluation. The minimum specified by the standard
2526 @item -fconstexpr-cache-depth=@var{n}
2527 @opindex fconstexpr-cache-depth
2528 Set the maximum level of nested evaluation depth for C++11 constexpr
2529 functions that will be cached to @var{n}. This is a heuristic that
2530 trades off compilation speed (when the cache avoids repeated
2531 calculations) against memory consumption (when the cache grows very
2532 large from highly recursive evaluations). The default is 8. Very few
2533 users are likely to want to adjust it, but if your code does heavy
2534 constexpr calculations you might want to experiment to find which
2535 value works best for you.
2537 @item -fconstexpr-loop-limit=@var{n}
2538 @opindex fconstexpr-loop-limit
2539 Set the maximum number of iterations for a loop in C++14 constexpr functions
2540 to @var{n}. A limit is needed to detect infinite loops during
2541 constant expression evaluation. The default is 262144 (1<<18).
2543 @item -fconstexpr-ops-limit=@var{n}
2544 @opindex fconstexpr-ops-limit
2545 Set the maximum number of operations during a single constexpr evaluation.
2546 Even when number of iterations of a single loop is limited with the above limit,
2547 if there are several nested loops and each of them has many iterations but still
2548 smaller than the above limit, or if in a body of some loop or even outside
2549 of a loop too many expressions need to be evaluated, the resulting constexpr
2550 evaluation might take too long.
2551 The default is 33554432 (1<<25).
2553 @item -fno-elide-constructors
2554 @opindex fno-elide-constructors
2555 @opindex felide-constructors
2556 The C++ standard allows an implementation to omit creating a temporary
2557 that is only used to initialize another object of the same type.
2558 Specifying this option disables that optimization, and forces G++ to
2559 call the copy constructor in all cases. This option also causes G++
2560 to call trivial member functions which otherwise would be expanded inline.
2562 In C++17, the compiler is required to omit these temporaries, but this
2563 option still affects trivial member functions.
2565 @item -fno-enforce-eh-specs
2566 @opindex fno-enforce-eh-specs
2567 @opindex fenforce-eh-specs
2568 Don't generate code to check for violation of exception specifications
2569 at run time. This option violates the C++ standard, but may be useful
2570 for reducing code size in production builds, much like defining
2571 @code{NDEBUG}. This does not give user code permission to throw
2572 exceptions in violation of the exception specifications; the compiler
2573 still optimizes based on the specifications, so throwing an
2574 unexpected exception results in undefined behavior at run time.
2576 @item -fextern-tls-init
2577 @itemx -fno-extern-tls-init
2578 @opindex fextern-tls-init
2579 @opindex fno-extern-tls-init
2580 The C++11 and OpenMP standards allow @code{thread_local} and
2581 @code{threadprivate} variables to have dynamic (runtime)
2582 initialization. To support this, any use of such a variable goes
2583 through a wrapper function that performs any necessary initialization.
2584 When the use and definition of the variable are in the same
2585 translation unit, this overhead can be optimized away, but when the
2586 use is in a different translation unit there is significant overhead
2587 even if the variable doesn't actually need dynamic initialization. If
2588 the programmer can be sure that no use of the variable in a
2589 non-defining TU needs to trigger dynamic initialization (either
2590 because the variable is statically initialized, or a use of the
2591 variable in the defining TU will be executed before any uses in
2592 another TU), they can avoid this overhead with the
2593 @option{-fno-extern-tls-init} option.
2595 On targets that support symbol aliases, the default is
2596 @option{-fextern-tls-init}. On targets that do not support symbol
2597 aliases, the default is @option{-fno-extern-tls-init}.
2599 @item -fno-gnu-keywords
2600 @opindex fno-gnu-keywords
2601 @opindex fgnu-keywords
2602 Do not recognize @code{typeof} as a keyword, so that code can use this
2603 word as an identifier. You can use the keyword @code{__typeof__} instead.
2604 This option is implied by the strict ISO C++ dialects: @option{-ansi},
2605 @option{-std=c++98}, @option{-std=c++11}, etc.
2607 @item -fno-implicit-templates
2608 @opindex fno-implicit-templates
2609 @opindex fimplicit-templates
2610 Never emit code for non-inline templates that are instantiated
2611 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2612 If you use this option, you must take care to structure your code to
2613 include all the necessary explicit instantiations to avoid getting
2614 undefined symbols at link time.
2615 @xref{Template Instantiation}, for more information.
2617 @item -fno-implicit-inline-templates
2618 @opindex fno-implicit-inline-templates
2619 @opindex fimplicit-inline-templates
2620 Don't emit code for implicit instantiations of inline templates, either.
2621 The default is to handle inlines differently so that compiles with and
2622 without optimization need the same set of explicit instantiations.
2624 @item -fno-implement-inlines
2625 @opindex fno-implement-inlines
2626 @opindex fimplement-inlines
2627 To save space, do not emit out-of-line copies of inline functions
2628 controlled by @code{#pragma implementation}. This causes linker
2629 errors if these functions are not inlined everywhere they are called.
2631 @item -fms-extensions
2632 @opindex fms-extensions
2633 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2634 int and getting a pointer to member function via non-standard syntax.
2636 @item -fnew-inheriting-ctors
2637 @opindex fnew-inheriting-ctors
2638 Enable the P0136 adjustment to the semantics of C++11 constructor
2639 inheritance. This is part of C++17 but also considered to be a Defect
2640 Report against C++11 and C++14. This flag is enabled by default
2641 unless @option{-fabi-version=10} or lower is specified.
2643 @item -fnew-ttp-matching
2644 @opindex fnew-ttp-matching
2645 Enable the P0522 resolution to Core issue 150, template template
2646 parameters and default arguments: this allows a template with default
2647 template arguments as an argument for a template template parameter
2648 with fewer template parameters. This flag is enabled by default for
2649 @option{-std=c++17}.
2651 @item -fno-nonansi-builtins
2652 @opindex fno-nonansi-builtins
2653 @opindex fnonansi-builtins
2654 Disable built-in declarations of functions that are not mandated by
2655 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2656 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2659 @opindex fnothrow-opt
2660 Treat a @code{throw()} exception specification as if it were a
2661 @code{noexcept} specification to reduce or eliminate the text size
2662 overhead relative to a function with no exception specification. If
2663 the function has local variables of types with non-trivial
2664 destructors, the exception specification actually makes the
2665 function smaller because the EH cleanups for those variables can be
2666 optimized away. The semantic effect is that an exception thrown out of
2667 a function with such an exception specification results in a call
2668 to @code{terminate} rather than @code{unexpected}.
2670 @item -fno-operator-names
2671 @opindex fno-operator-names
2672 @opindex foperator-names
2673 Do not treat the operator name keywords @code{and}, @code{bitand},
2674 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2675 synonyms as keywords.
2677 @item -fno-optional-diags
2678 @opindex fno-optional-diags
2679 @opindex foptional-diags
2680 Disable diagnostics that the standard says a compiler does not need to
2681 issue. Currently, the only such diagnostic issued by G++ is the one for
2682 a name having multiple meanings within a class.
2685 @opindex fpermissive
2686 Downgrade some diagnostics about nonconformant code from errors to
2687 warnings. Thus, using @option{-fpermissive} allows some
2688 nonconforming code to compile.
2690 @item -fno-pretty-templates
2691 @opindex fno-pretty-templates
2692 @opindex fpretty-templates
2693 When an error message refers to a specialization of a function
2694 template, the compiler normally prints the signature of the
2695 template followed by the template arguments and any typedefs or
2696 typenames in the signature (e.g.@: @code{void f(T) [with T = int]}
2697 rather than @code{void f(int)}) so that it's clear which template is
2698 involved. When an error message refers to a specialization of a class
2699 template, the compiler omits any template arguments that match
2700 the default template arguments for that template. If either of these
2701 behaviors make it harder to understand the error message rather than
2702 easier, you can use @option{-fno-pretty-templates} to disable them.
2706 Enable automatic template instantiation at link time. This option also
2707 implies @option{-fno-implicit-templates}. @xref{Template
2708 Instantiation}, for more information.
2713 Disable generation of information about every class with virtual
2714 functions for use by the C++ run-time type identification features
2715 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2716 of the language, you can save some space by using this flag. Note that
2717 exception handling uses the same information, but G++ generates it as
2718 needed. The @code{dynamic_cast} operator can still be used for casts that
2719 do not require run-time type information, i.e.@: casts to @code{void *} or to
2720 unambiguous base classes.
2722 Mixing code compiled with @option{-frtti} with that compiled with
2723 @option{-fno-rtti} may not work. For example, programs may
2724 fail to link if a class compiled with @option{-fno-rtti} is used as a base
2725 for a class compiled with @option{-frtti}.
2727 @item -fsized-deallocation
2728 @opindex fsized-deallocation
2729 Enable the built-in global declarations
2731 void operator delete (void *, std::size_t) noexcept;
2732 void operator delete[] (void *, std::size_t) noexcept;
2734 as introduced in C++14. This is useful for user-defined replacement
2735 deallocation functions that, for example, use the size of the object
2736 to make deallocation faster. Enabled by default under
2737 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2738 warns about places that might want to add a definition.
2740 @item -fstrict-enums
2741 @opindex fstrict-enums
2742 Allow the compiler to optimize using the assumption that a value of
2743 enumerated type can only be one of the values of the enumeration (as
2744 defined in the C++ standard; basically, a value that can be
2745 represented in the minimum number of bits needed to represent all the
2746 enumerators). This assumption may not be valid if the program uses a
2747 cast to convert an arbitrary integer value to the enumerated type.
2749 @item -fstrong-eval-order
2750 @opindex fstrong-eval-order
2751 Evaluate member access, array subscripting, and shift expressions in
2752 left-to-right order, and evaluate assignment in right-to-left order,
2753 as adopted for C++17. Enabled by default with @option{-std=c++17}.
2754 @option{-fstrong-eval-order=some} enables just the ordering of member
2755 access and shift expressions, and is the default without
2756 @option{-std=c++17}.
2758 @item -ftemplate-backtrace-limit=@var{n}
2759 @opindex ftemplate-backtrace-limit
2760 Set the maximum number of template instantiation notes for a single
2761 warning or error to @var{n}. The default value is 10.
2763 @item -ftemplate-depth=@var{n}
2764 @opindex ftemplate-depth
2765 Set the maximum instantiation depth for template classes to @var{n}.
2766 A limit on the template instantiation depth is needed to detect
2767 endless recursions during template class instantiation. ANSI/ISO C++
2768 conforming programs must not rely on a maximum depth greater than 17
2769 (changed to 1024 in C++11). The default value is 900, as the compiler
2770 can run out of stack space before hitting 1024 in some situations.
2772 @item -fno-threadsafe-statics
2773 @opindex fno-threadsafe-statics
2774 @opindex fthreadsafe-statics
2775 Do not emit the extra code to use the routines specified in the C++
2776 ABI for thread-safe initialization of local statics. You can use this
2777 option to reduce code size slightly in code that doesn't need to be
2780 @item -fuse-cxa-atexit
2781 @opindex fuse-cxa-atexit
2782 Register destructors for objects with static storage duration with the
2783 @code{__cxa_atexit} function rather than the @code{atexit} function.
2784 This option is required for fully standards-compliant handling of static
2785 destructors, but only works if your C library supports
2786 @code{__cxa_atexit}.
2788 @item -fno-use-cxa-get-exception-ptr
2789 @opindex fno-use-cxa-get-exception-ptr
2790 @opindex fuse-cxa-get-exception-ptr
2791 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2792 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2793 if the runtime routine is not available.
2795 @item -fvisibility-inlines-hidden
2796 @opindex fvisibility-inlines-hidden
2797 This switch declares that the user does not attempt to compare
2798 pointers to inline functions or methods where the addresses of the two functions
2799 are taken in different shared objects.
2801 The effect of this is that GCC may, effectively, mark inline methods with
2802 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2803 appear in the export table of a DSO and do not require a PLT indirection
2804 when used within the DSO@. Enabling this option can have a dramatic effect
2805 on load and link times of a DSO as it massively reduces the size of the
2806 dynamic export table when the library makes heavy use of templates.
2808 The behavior of this switch is not quite the same as marking the
2809 methods as hidden directly, because it does not affect static variables
2810 local to the function or cause the compiler to deduce that
2811 the function is defined in only one shared object.
2813 You may mark a method as having a visibility explicitly to negate the
2814 effect of the switch for that method. For example, if you do want to
2815 compare pointers to a particular inline method, you might mark it as
2816 having default visibility. Marking the enclosing class with explicit
2817 visibility has no effect.
2819 Explicitly instantiated inline methods are unaffected by this option
2820 as their linkage might otherwise cross a shared library boundary.
2821 @xref{Template Instantiation}.
2823 @item -fvisibility-ms-compat
2824 @opindex fvisibility-ms-compat
2825 This flag attempts to use visibility settings to make GCC's C++
2826 linkage model compatible with that of Microsoft Visual Studio.
2828 The flag makes these changes to GCC's linkage model:
2832 It sets the default visibility to @code{hidden}, like
2833 @option{-fvisibility=hidden}.
2836 Types, but not their members, are not hidden by default.
2839 The One Definition Rule is relaxed for types without explicit
2840 visibility specifications that are defined in more than one
2841 shared object: those declarations are permitted if they are
2842 permitted when this option is not used.
2845 In new code it is better to use @option{-fvisibility=hidden} and
2846 export those classes that are intended to be externally visible.
2847 Unfortunately it is possible for code to rely, perhaps accidentally,
2848 on the Visual Studio behavior.
2850 Among the consequences of these changes are that static data members
2851 of the same type with the same name but defined in different shared
2852 objects are different, so changing one does not change the other;
2853 and that pointers to function members defined in different shared
2854 objects may not compare equal. When this flag is given, it is a
2855 violation of the ODR to define types with the same name differently.
2860 Do not use weak symbol support, even if it is provided by the linker.
2861 By default, G++ uses weak symbols if they are available. This
2862 option exists only for testing, and should not be used by end-users;
2863 it results in inferior code and has no benefits. This option may
2864 be removed in a future release of G++.
2868 Do not search for header files in the standard directories specific to
2869 C++, but do still search the other standard directories. (This option
2870 is used when building the C++ library.)
2873 In addition, these optimization, warning, and code generation options
2874 have meanings only for C++ programs:
2877 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2880 Warn when G++ it generates code that is probably not compatible with
2881 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2882 ABI with each major release, normally @option{-Wabi} will warn only if
2883 there is a check added later in a release series for an ABI issue
2884 discovered since the initial release. @option{-Wabi} will warn about
2885 more things if an older ABI version is selected (with
2886 @option{-fabi-version=@var{n}}).
2888 @option{-Wabi} can also be used with an explicit version number to
2889 warn about compatibility with a particular @option{-fabi-version}
2890 level, e.g.@: @option{-Wabi=2} to warn about changes relative to
2891 @option{-fabi-version=2}.
2893 If an explicit version number is provided and
2894 @option{-fabi-compat-version} is not specified, the version number
2895 from this option is used for compatibility aliases. If no explicit
2896 version number is provided with this option, but
2897 @option{-fabi-compat-version} is specified, that version number is
2898 used for ABI warnings.
2900 Although an effort has been made to warn about
2901 all such cases, there are probably some cases that are not warned about,
2902 even though G++ is generating incompatible code. There may also be
2903 cases where warnings are emitted even though the code that is generated
2906 You should rewrite your code to avoid these warnings if you are
2907 concerned about the fact that code generated by G++ may not be binary
2908 compatible with code generated by other compilers.
2910 Known incompatibilities in @option{-fabi-version=2} (which was the
2911 default from GCC 3.4 to 4.9) include:
2916 A template with a non-type template parameter of reference type was
2917 mangled incorrectly:
2920 template <int &> struct S @{@};
2924 This was fixed in @option{-fabi-version=3}.
2927 SIMD vector types declared using @code{__attribute ((vector_size))} were
2928 mangled in a non-standard way that does not allow for overloading of
2929 functions taking vectors of different sizes.
2931 The mangling was changed in @option{-fabi-version=4}.
2934 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2935 qualifiers, and @code{decltype} of a plain declaration was folded away.
2937 These mangling issues were fixed in @option{-fabi-version=5}.
2940 Scoped enumerators passed as arguments to a variadic function are
2941 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2942 On most targets this does not actually affect the parameter passing
2943 ABI, as there is no way to pass an argument smaller than @code{int}.
2945 Also, the ABI changed the mangling of template argument packs,
2946 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2947 a class scope function used as a template argument.
2949 These issues were corrected in @option{-fabi-version=6}.
2952 Lambdas in default argument scope were mangled incorrectly, and the
2953 ABI changed the mangling of @code{nullptr_t}.
2955 These issues were corrected in @option{-fabi-version=7}.
2958 When mangling a function type with function-cv-qualifiers, the
2959 un-qualified function type was incorrectly treated as a substitution
2962 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2965 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2966 unaligned accesses. Note that this did not affect the ABI of a
2967 function with a @code{nullptr_t} parameter, as parameters have a
2970 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2973 Target-specific attributes that affect the identity of a type, such as
2974 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2975 did not affect the mangled name, leading to name collisions when
2976 function pointers were used as template arguments.
2978 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2982 It also warns about psABI-related changes. The known psABI changes at this
2988 For SysV/x86-64, unions with @code{long double} members are
2989 passed in memory as specified in psABI. For example:
2999 @code{union U} is always passed in memory.
3003 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
3006 Warn when a type with an ABI tag is used in a context that does not
3007 have that ABI tag. See @ref{C++ Attributes} for more information
3010 @item -Wcomma-subscript @r{(C++ and Objective-C++ only)}
3011 @opindex Wcomma-subscript
3012 @opindex Wno-comma-subscript
3013 Warn about uses of a comma expression within a subscripting expression.
3014 This usage was deprecated in C++2a. However, a comma expression wrapped
3015 in @code{( )} is not deprecated. Example:
3019 void f(int *a, int b, int c) @{
3020 a[b,c]; // deprecated
3026 Enabled by default with @option{-std=c++2a}.
3028 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
3029 @opindex Wctor-dtor-privacy
3030 @opindex Wno-ctor-dtor-privacy
3031 Warn when a class seems unusable because all the constructors or
3032 destructors in that class are private, and it has neither friends nor
3033 public static member functions. Also warn if there are no non-private
3034 methods, and there's at least one private member function that isn't
3035 a constructor or destructor.
3037 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
3038 @opindex Wdelete-non-virtual-dtor
3039 @opindex Wno-delete-non-virtual-dtor
3040 Warn when @code{delete} is used to destroy an instance of a class that
3041 has virtual functions and non-virtual destructor. It is unsafe to delete
3042 an instance of a derived class through a pointer to a base class if the
3043 base class does not have a virtual destructor. This warning is enabled
3046 @item -Wdeprecated-copy @r{(C++ and Objective-C++ only)}
3047 @opindex Wdeprecated-copy
3048 @opindex Wno-deprecated-copy
3049 Warn that the implicit declaration of a copy constructor or copy
3050 assignment operator is deprecated if the class has a user-provided
3051 copy constructor or copy assignment operator, in C++11 and up. This
3052 warning is enabled by @option{-Wextra}. With
3053 @option{-Wdeprecated-copy-dtor}, also deprecate if the class has a
3054 user-provided destructor.
3056 @item -Wno-init-list-lifetime @r{(C++ and Objective-C++ only)}
3057 @opindex Winit-list-lifetime
3058 @opindex Wno-init-list-lifetime
3059 Do not warn about uses of @code{std::initializer_list} that are likely
3060 to result in dangling pointers. Since the underlying array for an
3061 @code{initializer_list} is handled like a normal C++ temporary object,
3062 it is easy to inadvertently keep a pointer to the array past the end
3063 of the array's lifetime. For example:
3067 If a function returns a temporary @code{initializer_list}, or a local
3068 @code{initializer_list} variable, the array's lifetime ends at the end
3069 of the return statement, so the value returned has a dangling pointer.
3072 If a new-expression creates an @code{initializer_list}, the array only
3073 lives until the end of the enclosing full-expression, so the
3074 @code{initializer_list} in the heap has a dangling pointer.
3077 When an @code{initializer_list} variable is assigned from a
3078 brace-enclosed initializer list, the temporary array created for the
3079 right side of the assignment only lives until the end of the
3080 full-expression, so at the next statement the @code{initializer_list}
3081 variable has a dangling pointer.
3084 // li's initial underlying array lives as long as li
3085 std::initializer_list<int> li = @{ 1,2,3 @};
3086 // assignment changes li to point to a temporary array
3088 // now the temporary is gone and li has a dangling pointer
3089 int i = li.begin()[0] // undefined behavior
3093 When a list constructor stores the @code{begin} pointer from the
3094 @code{initializer_list} argument, this doesn't extend the lifetime of
3095 the array, so if a class variable is constructed from a temporary
3096 @code{initializer_list}, the pointer is left dangling by the end of
3097 the variable declaration statement.
3101 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
3102 @opindex Wliteral-suffix
3103 @opindex Wno-literal-suffix
3104 Warn when a string or character literal is followed by a ud-suffix which does
3105 not begin with an underscore. As a conforming extension, GCC treats such
3106 suffixes as separate preprocessing tokens in order to maintain backwards
3107 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
3111 #define __STDC_FORMAT_MACROS
3112 #include <inttypes.h>
3117 printf("My int64: %" PRId64"\n", i64);
3121 In this case, @code{PRId64} is treated as a separate preprocessing token.
3123 Additionally, warn when a user-defined literal operator is declared with
3124 a literal suffix identifier that doesn't begin with an underscore. Literal
3125 suffix identifiers that don't begin with an underscore are reserved for
3126 future standardization.
3128 This warning is enabled by default.
3130 @item -Wlto-type-mismatch
3131 @opindex Wlto-type-mismatch
3132 @opindex Wno-lto-type-mismatch
3134 During the link-time optimization warn about type mismatches in
3135 global declarations from different compilation units.
3136 Requires @option{-flto} to be enabled. Enabled by default.
3138 @item -Wno-narrowing @r{(C++ and Objective-C++ only)}
3140 @opindex Wno-narrowing
3141 For C++11 and later standards, narrowing conversions are diagnosed by default,
3142 as required by the standard. A narrowing conversion from a constant produces
3143 an error, and a narrowing conversion from a non-constant produces a warning,
3144 but @option{-Wno-narrowing} suppresses the diagnostic.
3145 Note that this does not affect the meaning of well-formed code;
3146 narrowing conversions are still considered ill-formed in SFINAE contexts.
3148 With @option{-Wnarrowing} in C++98, warn when a narrowing
3149 conversion prohibited by C++11 occurs within
3153 int i = @{ 2.2 @}; // error: narrowing from double to int
3156 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
3158 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
3160 @opindex Wno-noexcept
3161 Warn when a noexcept-expression evaluates to false because of a call
3162 to a function that does not have a non-throwing exception
3163 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
3164 the compiler to never throw an exception.
3166 @item -Wnoexcept-type @r{(C++ and Objective-C++ only)}
3167 @opindex Wnoexcept-type
3168 @opindex Wno-noexcept-type
3169 Warn if the C++17 feature making @code{noexcept} part of a function
3170 type changes the mangled name of a symbol relative to C++14. Enabled
3171 by @option{-Wabi} and @option{-Wc++17-compat}.
3176 template <class T> void f(T t) @{ t(); @};
3178 void h() @{ f(g); @}
3182 In C++14, @code{f} calls @code{f<void(*)()>}, but in
3183 C++17 it calls @code{f<void(*)()noexcept>}.
3185 @item -Wclass-memaccess @r{(C++ and Objective-C++ only)}
3186 @opindex Wclass-memaccess
3187 @opindex Wno-class-memaccess
3188 Warn when the destination of a call to a raw memory function such as
3189 @code{memset} or @code{memcpy} is an object of class type, and when writing
3190 into such an object might bypass the class non-trivial or deleted constructor
3191 or copy assignment, violate const-correctness or encapsulation, or corrupt
3192 virtual table pointers. Modifying the representation of such objects may
3193 violate invariants maintained by member functions of the class. For example,
3194 the call to @code{memset} below is undefined because it modifies a non-trivial
3195 class object and is, therefore, diagnosed. The safe way to either initialize
3196 or clear the storage of objects of such types is by using the appropriate
3197 constructor or assignment operator, if one is available.
3199 std::string str = "abc";
3200 memset (&str, 0, sizeof str);
3202 The @option{-Wclass-memaccess} option is enabled by @option{-Wall}.
3203 Explicitly casting the pointer to the class object to @code{void *} or
3204 to a type that can be safely accessed by the raw memory function suppresses
3207 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
3208 @opindex Wnon-virtual-dtor
3209 @opindex Wno-non-virtual-dtor
3210 Warn when a class has virtual functions and an accessible non-virtual
3211 destructor itself or in an accessible polymorphic base class, in which
3212 case it is possible but unsafe to delete an instance of a derived
3213 class through a pointer to the class itself or base class. This
3214 warning is automatically enabled if @option{-Weffc++} is specified.
3216 @item -Wregister @r{(C++ and Objective-C++ only)}
3218 @opindex Wno-register
3219 Warn on uses of the @code{register} storage class specifier, except
3220 when it is part of the GNU @ref{Explicit Register Variables} extension.
3221 The use of the @code{register} keyword as storage class specifier has
3222 been deprecated in C++11 and removed in C++17.
3223 Enabled by default with @option{-std=c++17}.
3225 @item -Wreorder @r{(C++ and Objective-C++ only)}
3227 @opindex Wno-reorder
3228 @cindex reordering, warning
3229 @cindex warning for reordering of member initializers
3230 Warn when the order of member initializers given in the code does not
3231 match the order in which they must be executed. For instance:
3237 A(): j (0), i (1) @{ @}
3242 The compiler rearranges the member initializers for @code{i}
3243 and @code{j} to match the declaration order of the members, emitting
3244 a warning to that effect. This warning is enabled by @option{-Wall}.
3246 @item -Wno-pessimizing-move @r{(C++ and Objective-C++ only)}
3247 @opindex Wpessimizing-move
3248 @opindex Wno-pessimizing-move
3249 This warning warns when a call to @code{std::move} prevents copy
3250 elision. A typical scenario when copy elision can occur is when returning in
3251 a function with a class return type, when the expression being returned is the
3252 name of a non-volatile automatic object, and is not a function parameter, and
3253 has the same type as the function return type.
3263 return std::move (t);
3267 But in this example, the @code{std::move} call prevents copy elision.
3269 This warning is enabled by @option{-Wall}.
3271 @item -Wno-redundant-move @r{(C++ and Objective-C++ only)}
3272 @opindex Wredundant-move
3273 @opindex Wno-redundant-move
3274 This warning warns about redundant calls to @code{std::move}; that is, when
3275 a move operation would have been performed even without the @code{std::move}
3276 call. This happens because the compiler is forced to treat the object as if
3277 it were an rvalue in certain situations such as returning a local variable,
3278 where copy elision isn't applicable. Consider:
3287 return std::move (t);
3291 Here, the @code{std::move} call is redundant. Because G++ implements Core
3292 Issue 1579, another example is:
3295 struct T @{ // convertible to U
3305 return std::move (t);
3308 In this example, copy elision isn't applicable because the type of the
3309 expression being returned and the function return type differ, yet G++
3310 treats the return value as if it were designated by an rvalue.
3312 This warning is enabled by @option{-Wextra}.
3314 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
3315 @opindex fext-numeric-literals
3316 @opindex fno-ext-numeric-literals
3317 Accept imaginary, fixed-point, or machine-defined
3318 literal number suffixes as GNU extensions.
3319 When this option is turned off these suffixes are treated
3320 as C++11 user-defined literal numeric suffixes.
3321 This is on by default for all pre-C++11 dialects and all GNU dialects:
3322 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
3323 @option{-std=gnu++14}.
3324 This option is off by default
3325 for ISO C++11 onwards (@option{-std=c++11}, ...).
3328 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
3331 @item -Weffc++ @r{(C++ and Objective-C++ only)}
3334 Warn about violations of the following style guidelines from Scott Meyers'
3335 @cite{Effective C++} series of books:
3339 Define a copy constructor and an assignment operator for classes
3340 with dynamically-allocated memory.
3343 Prefer initialization to assignment in constructors.
3346 Have @code{operator=} return a reference to @code{*this}.
3349 Don't try to return a reference when you must return an object.
3352 Distinguish between prefix and postfix forms of increment and
3353 decrement operators.
3356 Never overload @code{&&}, @code{||}, or @code{,}.
3360 This option also enables @option{-Wnon-virtual-dtor}, which is also
3361 one of the effective C++ recommendations. However, the check is
3362 extended to warn about the lack of virtual destructor in accessible
3363 non-polymorphic bases classes too.
3365 When selecting this option, be aware that the standard library
3366 headers do not obey all of these guidelines; use @samp{grep -v}
3367 to filter out those warnings.
3369 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
3370 @opindex Wstrict-null-sentinel
3371 @opindex Wno-strict-null-sentinel
3372 Warn about the use of an uncasted @code{NULL} as sentinel. When
3373 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
3374 to @code{__null}. Although it is a null pointer constant rather than a
3375 null pointer, it is guaranteed to be of the same size as a pointer.
3376 But this use is not portable across different compilers.
3378 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
3379 @opindex Wno-non-template-friend
3380 @opindex Wnon-template-friend
3381 Disable warnings when non-template friend functions are declared
3382 within a template. In very old versions of GCC that predate implementation
3383 of the ISO standard, declarations such as
3384 @samp{friend int foo(int)}, where the name of the friend is an unqualified-id,
3385 could be interpreted as a particular specialization of a template
3386 function; the warning exists to diagnose compatibility problems,
3387 and is enabled by default.
3389 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
3390 @opindex Wold-style-cast
3391 @opindex Wno-old-style-cast
3392 Warn if an old-style (C-style) cast to a non-void type is used within
3393 a C++ program. The new-style casts (@code{dynamic_cast},
3394 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
3395 less vulnerable to unintended effects and much easier to search for.
3397 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
3398 @opindex Woverloaded-virtual
3399 @opindex Wno-overloaded-virtual
3400 @cindex overloaded virtual function, warning
3401 @cindex warning for overloaded virtual function
3402 Warn when a function declaration hides virtual functions from a
3403 base class. For example, in:
3410 struct B: public A @{
3415 the @code{A} class version of @code{f} is hidden in @code{B}, and code
3426 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
3427 @opindex Wno-pmf-conversions
3428 @opindex Wpmf-conversions
3429 Disable the diagnostic for converting a bound pointer to member function
3432 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
3433 @opindex Wsign-promo
3434 @opindex Wno-sign-promo
3435 Warn when overload resolution chooses a promotion from unsigned or
3436 enumerated type to a signed type, over a conversion to an unsigned type of
3437 the same size. Previous versions of G++ tried to preserve
3438 unsignedness, but the standard mandates the current behavior.
3440 @item -Wtemplates @r{(C++ and Objective-C++ only)}
3442 @opindex Wno-templates
3443 Warn when a primary template declaration is encountered. Some coding
3444 rules disallow templates, and this may be used to enforce that rule.
3445 The warning is inactive inside a system header file, such as the STL, so
3446 one can still use the STL. One may also instantiate or specialize
3449 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
3450 @opindex Wmultiple-inheritance
3451 @opindex Wno-multiple-inheritance
3452 Warn when a class is defined with multiple direct base classes. 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 multiple inheritance.
3458 @item -Wvirtual-inheritance
3459 @opindex Wvirtual-inheritance
3460 @opindex Wno-virtual-inheritance
3461 Warn when a class is defined with a virtual direct base class. Some
3462 coding rules disallow multiple inheritance, and this may be used to
3463 enforce that rule. The warning is inactive inside a system header file,
3464 such as the STL, so one can still use the STL. One may also define
3465 classes that indirectly use virtual inheritance.
3468 @opindex Wnamespaces
3469 @opindex Wno-namespaces
3470 Warn when a namespace definition is opened. Some coding rules disallow
3471 namespaces, and this may be used to enforce that rule. The warning is
3472 inactive inside a system header file, such as the STL, so one can still
3473 use the STL. One may also use using directives and qualified names.
3475 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
3477 @opindex Wno-terminate
3478 Disable the warning about a throw-expression that will immediately
3479 result in a call to @code{terminate}.
3481 @item -Wno-class-conversion @r{(C++ and Objective-C++ only)}
3482 @opindex Wno-class-conversion
3483 @opindex Wclass-conversion
3484 Disable the warning about the case when a conversion function converts an
3485 object to the same type, to a base class of that type, or to void; such
3486 a conversion function will never be called.
3488 @item -Wvolatile @r{(C++ and Objective-C++ only)}
3490 @opindex Wno-volatile
3491 Warn about deprecated uses of the @code{volatile} qualifier. This includes
3492 postfix and prefix @code{++} and @code{--} expressions of
3493 @code{volatile}-qualified types, using simple assignments where the left
3494 operand is a @code{volatile}-qualified non-class type for their value,
3495 compound assignments where the left operand is a @code{volatile}-qualified
3496 non-class type, @code{volatile}-qualified function return type,
3497 @code{volatile}-qualified parameter type, and structured bindings of a
3498 @code{volatile}-qualified type. This usage was deprecated in C++20.
3500 Enabled by default with @option{-std=c++2a}.
3503 @node Objective-C and Objective-C++ Dialect Options
3504 @section Options Controlling Objective-C and Objective-C++ Dialects
3506 @cindex compiler options, Objective-C and Objective-C++
3507 @cindex Objective-C and Objective-C++ options, command-line
3508 @cindex options, Objective-C and Objective-C++
3509 (NOTE: This manual does not describe the Objective-C and Objective-C++
3510 languages themselves. @xref{Standards,,Language Standards
3511 Supported by GCC}, for references.)
3513 This section describes the command-line options that are only meaningful
3514 for Objective-C and Objective-C++ programs. You can also use most of
3515 the language-independent GNU compiler options.
3516 For example, you might compile a file @file{some_class.m} like this:
3519 gcc -g -fgnu-runtime -O -c some_class.m
3523 In this example, @option{-fgnu-runtime} is an option meant only for
3524 Objective-C and Objective-C++ programs; you can use the other options with
3525 any language supported by GCC@.
3527 Note that since Objective-C is an extension of the C language, Objective-C
3528 compilations may also use options specific to the C front-end (e.g.,
3529 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3530 C++-specific options (e.g., @option{-Wabi}).
3532 Here is a list of options that are @emph{only} for compiling Objective-C
3533 and Objective-C++ programs:
3536 @item -fconstant-string-class=@var{class-name}
3537 @opindex fconstant-string-class
3538 Use @var{class-name} as the name of the class to instantiate for each
3539 literal string specified with the syntax @code{@@"@dots{}"}. The default
3540 class name is @code{NXConstantString} if the GNU runtime is being used, and
3541 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3542 @option{-fconstant-cfstrings} option, if also present, overrides the
3543 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3544 to be laid out as constant CoreFoundation strings.
3547 @opindex fgnu-runtime
3548 Generate object code compatible with the standard GNU Objective-C
3549 runtime. This is the default for most types of systems.
3551 @item -fnext-runtime
3552 @opindex fnext-runtime
3553 Generate output compatible with the NeXT runtime. This is the default
3554 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3555 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3558 @item -fno-nil-receivers
3559 @opindex fno-nil-receivers
3560 @opindex fnil-receivers
3561 Assume that all Objective-C message dispatches (@code{[receiver
3562 message:arg]}) in this translation unit ensure that the receiver is
3563 not @code{nil}. This allows for more efficient entry points in the
3564 runtime to be used. This option is only available in conjunction with
3565 the NeXT runtime and ABI version 0 or 1.
3567 @item -fobjc-abi-version=@var{n}
3568 @opindex fobjc-abi-version
3569 Use version @var{n} of the Objective-C ABI for the selected runtime.
3570 This option is currently supported only for the NeXT runtime. In that
3571 case, Version 0 is the traditional (32-bit) ABI without support for
3572 properties and other Objective-C 2.0 additions. Version 1 is the
3573 traditional (32-bit) ABI with support for properties and other
3574 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3575 nothing is specified, the default is Version 0 on 32-bit target
3576 machines, and Version 2 on 64-bit target machines.
3578 @item -fobjc-call-cxx-cdtors
3579 @opindex fobjc-call-cxx-cdtors
3580 For each Objective-C class, check if any of its instance variables is a
3581 C++ object with a non-trivial default constructor. If so, synthesize a
3582 special @code{- (id) .cxx_construct} instance method which runs
3583 non-trivial default constructors on any such instance variables, in order,
3584 and then return @code{self}. Similarly, check if any instance variable
3585 is a C++ object with a non-trivial destructor, and if so, synthesize a
3586 special @code{- (void) .cxx_destruct} method which runs
3587 all such default destructors, in reverse order.
3589 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3590 methods thusly generated only operate on instance variables
3591 declared in the current Objective-C class, and not those inherited
3592 from superclasses. It is the responsibility of the Objective-C
3593 runtime to invoke all such methods in an object's inheritance
3594 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3595 by the runtime immediately after a new object instance is allocated;
3596 the @code{- (void) .cxx_destruct} methods are invoked immediately
3597 before the runtime deallocates an object instance.
3599 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3600 support for invoking the @code{- (id) .cxx_construct} and
3601 @code{- (void) .cxx_destruct} methods.
3603 @item -fobjc-direct-dispatch
3604 @opindex fobjc-direct-dispatch
3605 Allow fast jumps to the message dispatcher. On Darwin this is
3606 accomplished via the comm page.
3608 @item -fobjc-exceptions
3609 @opindex fobjc-exceptions
3610 Enable syntactic support for structured exception handling in
3611 Objective-C, similar to what is offered by C++. This option
3612 is required to use the Objective-C keywords @code{@@try},
3613 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3614 @code{@@synchronized}. This option is available with both the GNU
3615 runtime and the NeXT runtime (but not available in conjunction with
3616 the NeXT runtime on Mac OS X 10.2 and earlier).
3620 Enable garbage collection (GC) in Objective-C and Objective-C++
3621 programs. This option is only available with the NeXT runtime; the
3622 GNU runtime has a different garbage collection implementation that
3623 does not require special compiler flags.
3625 @item -fobjc-nilcheck
3626 @opindex fobjc-nilcheck
3627 For the NeXT runtime with version 2 of the ABI, check for a nil
3628 receiver in method invocations before doing the actual method call.
3629 This is the default and can be disabled using
3630 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3631 checked for nil in this way no matter what this flag is set to.
3632 Currently this flag does nothing when the GNU runtime, or an older
3633 version of the NeXT runtime ABI, is used.
3635 @item -fobjc-std=objc1
3637 Conform to the language syntax of Objective-C 1.0, the language
3638 recognized by GCC 4.0. This only affects the Objective-C additions to
3639 the C/C++ language; it does not affect conformance to C/C++ standards,
3640 which is controlled by the separate C/C++ dialect option flags. When
3641 this option is used with the Objective-C or Objective-C++ compiler,
3642 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3643 This is useful if you need to make sure that your Objective-C code can
3644 be compiled with older versions of GCC@.
3646 @item -freplace-objc-classes
3647 @opindex freplace-objc-classes
3648 Emit a special marker instructing @command{ld(1)} not to statically link in
3649 the resulting object file, and allow @command{dyld(1)} to load it in at
3650 run time instead. This is used in conjunction with the Fix-and-Continue
3651 debugging mode, where the object file in question may be recompiled and
3652 dynamically reloaded in the course of program execution, without the need
3653 to restart the program itself. Currently, Fix-and-Continue functionality
3654 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3659 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3660 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3661 compile time) with static class references that get initialized at load time,
3662 which improves run-time performance. Specifying the @option{-fzero-link} flag
3663 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3664 to be retained. This is useful in Zero-Link debugging mode, since it allows
3665 for individual class implementations to be modified during program execution.
3666 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3667 regardless of command-line options.
3669 @item -fno-local-ivars
3670 @opindex fno-local-ivars
3671 @opindex flocal-ivars
3672 By default instance variables in Objective-C can be accessed as if
3673 they were local variables from within the methods of the class they're
3674 declared in. This can lead to shadowing between instance variables
3675 and other variables declared either locally inside a class method or
3676 globally with the same name. Specifying the @option{-fno-local-ivars}
3677 flag disables this behavior thus avoiding variable shadowing issues.
3679 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3680 @opindex fivar-visibility
3681 Set the default instance variable visibility to the specified option
3682 so that instance variables declared outside the scope of any access
3683 modifier directives default to the specified visibility.
3687 Dump interface declarations for all classes seen in the source file to a
3688 file named @file{@var{sourcename}.decl}.
3690 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3691 @opindex Wassign-intercept
3692 @opindex Wno-assign-intercept
3693 Warn whenever an Objective-C assignment is being intercepted by the
3696 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3697 @opindex Wno-protocol
3699 If a class is declared to implement a protocol, a warning is issued for
3700 every method in the protocol that is not implemented by the class. The
3701 default behavior is to issue a warning for every method not explicitly
3702 implemented in the class, even if a method implementation is inherited
3703 from the superclass. If you use the @option{-Wno-protocol} option, then
3704 methods inherited from the superclass are considered to be implemented,
3705 and no warning is issued for them.
3707 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3709 @opindex Wno-selector
3710 Warn if multiple methods of different types for the same selector are
3711 found during compilation. The check is performed on the list of methods
3712 in the final stage of compilation. Additionally, a check is performed
3713 for each selector appearing in a @code{@@selector(@dots{})}
3714 expression, and a corresponding method for that selector has been found
3715 during compilation. Because these checks scan the method table only at
3716 the end of compilation, these warnings are not produced if the final
3717 stage of compilation is not reached, for example because an error is
3718 found during compilation, or because the @option{-fsyntax-only} option is
3721 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3722 @opindex Wstrict-selector-match
3723 @opindex Wno-strict-selector-match
3724 Warn if multiple methods with differing argument and/or return types are
3725 found for a given selector when attempting to send a message using this
3726 selector to a receiver of type @code{id} or @code{Class}. When this flag
3727 is off (which is the default behavior), the compiler omits such warnings
3728 if any differences found are confined to types that share the same size
3731 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3732 @opindex Wundeclared-selector
3733 @opindex Wno-undeclared-selector
3734 Warn if a @code{@@selector(@dots{})} expression referring to an
3735 undeclared selector is found. A selector is considered undeclared if no
3736 method with that name has been declared before the
3737 @code{@@selector(@dots{})} expression, either explicitly in an
3738 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3739 an @code{@@implementation} section. This option always performs its
3740 checks as soon as a @code{@@selector(@dots{})} expression is found,
3741 while @option{-Wselector} only performs its checks in the final stage of
3742 compilation. This also enforces the coding style convention
3743 that methods and selectors must be declared before being used.
3745 @item -print-objc-runtime-info
3746 @opindex print-objc-runtime-info
3747 Generate C header describing the largest structure that is passed by
3752 @node Diagnostic Message Formatting Options
3753 @section Options to Control Diagnostic Messages Formatting
3754 @cindex options to control diagnostics formatting
3755 @cindex diagnostic messages
3756 @cindex message formatting
3758 Traditionally, diagnostic messages have been formatted irrespective of
3759 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3760 options described below
3761 to control the formatting algorithm for diagnostic messages,
3762 e.g.@: how many characters per line, how often source location
3763 information should be reported. Note that some language front ends may not
3764 honor these options.
3767 @item -fmessage-length=@var{n}
3768 @opindex fmessage-length
3769 Try to format error messages so that they fit on lines of about
3770 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3771 done; each error message appears on a single line. This is the
3772 default for all front ends.
3774 Note - this option also affects the display of the @samp{#error} and
3775 @samp{#warning} pre-processor directives, and the @samp{deprecated}
3776 function/type/variable attribute. It does not however affect the
3777 @samp{pragma GCC warning} and @samp{pragma GCC error} pragmas.
3779 @item -fdiagnostics-show-location=once
3780 @opindex fdiagnostics-show-location
3781 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3782 reporter to emit source location information @emph{once}; that is, in
3783 case the message is too long to fit on a single physical line and has to
3784 be wrapped, the source location won't be emitted (as prefix) again,
3785 over and over, in subsequent continuation lines. This is the default
3788 @item -fdiagnostics-show-location=every-line
3789 Only meaningful in line-wrapping mode. Instructs the diagnostic
3790 messages reporter to emit the same source location information (as
3791 prefix) for physical lines that result from the process of breaking
3792 a message which is too long to fit on a single line.
3794 @item -fdiagnostics-color[=@var{WHEN}]
3795 @itemx -fno-diagnostics-color
3796 @opindex fdiagnostics-color
3797 @cindex highlight, color
3798 @vindex GCC_COLORS @r{environment variable}
3799 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3800 or @samp{auto}. The default depends on how the compiler has been configured,
3801 it can be any of the above @var{WHEN} options or also @samp{never}
3802 if @env{GCC_COLORS} environment variable isn't present in the environment,
3803 and @samp{auto} otherwise.
3804 @samp{auto} means to use color only when the standard error is a terminal.
3805 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3806 aliases for @option{-fdiagnostics-color=always} and
3807 @option{-fdiagnostics-color=never}, respectively.
3809 The colors are defined by the environment variable @env{GCC_COLORS}.
3810 Its value is a colon-separated list of capabilities and Select Graphic
3811 Rendition (SGR) substrings. SGR commands are interpreted by the
3812 terminal or terminal emulator. (See the section in the documentation
3813 of your text terminal for permitted values and their meanings as
3814 character attributes.) These substring values are integers in decimal
3815 representation and can be concatenated with semicolons.
3816 Common values to concatenate include
3818 @samp{4} for underline,
3820 @samp{7} for inverse,
3821 @samp{39} for default foreground color,
3822 @samp{30} to @samp{37} for foreground colors,
3823 @samp{90} to @samp{97} for 16-color mode foreground colors,
3824 @samp{38;5;0} to @samp{38;5;255}
3825 for 88-color and 256-color modes foreground colors,
3826 @samp{49} for default background color,
3827 @samp{40} to @samp{47} for background colors,
3828 @samp{100} to @samp{107} for 16-color mode background colors,
3829 and @samp{48;5;0} to @samp{48;5;255}
3830 for 88-color and 256-color modes background colors.
3832 The default @env{GCC_COLORS} is
3834 error=01;31:warning=01;35:note=01;36:range1=32:range2=34:locus=01:\
3835 quote=01:fixit-insert=32:fixit-delete=31:\
3836 diff-filename=01:diff-hunk=32:diff-delete=31:diff-insert=32:\
3840 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3841 @samp{01;36} is bold cyan, @samp{32} is green, @samp{34} is blue,
3842 @samp{01} is bold, and @samp{31} is red.
3843 Setting @env{GCC_COLORS} to the empty string disables colors.
3844 Supported capabilities are as follows.
3848 @vindex error GCC_COLORS @r{capability}
3849 SGR substring for error: markers.
3852 @vindex warning GCC_COLORS @r{capability}
3853 SGR substring for warning: markers.
3856 @vindex note GCC_COLORS @r{capability}
3857 SGR substring for note: markers.
3860 @vindex range1 GCC_COLORS @r{capability}
3861 SGR substring for first additional range.
3864 @vindex range2 GCC_COLORS @r{capability}
3865 SGR substring for second additional range.
3868 @vindex locus GCC_COLORS @r{capability}
3869 SGR substring for location information, @samp{file:line} or
3870 @samp{file:line:column} etc.
3873 @vindex quote GCC_COLORS @r{capability}
3874 SGR substring for information printed within quotes.
3877 @vindex fixit-insert GCC_COLORS @r{capability}
3878 SGR substring for fix-it hints suggesting text to
3879 be inserted or replaced.
3882 @vindex fixit-delete GCC_COLORS @r{capability}
3883 SGR substring for fix-it hints suggesting text to
3886 @item diff-filename=
3887 @vindex diff-filename GCC_COLORS @r{capability}
3888 SGR substring for filename headers within generated patches.
3891 @vindex diff-hunk GCC_COLORS @r{capability}
3892 SGR substring for the starts of hunks within generated patches.
3895 @vindex diff-delete GCC_COLORS @r{capability}
3896 SGR substring for deleted lines within generated patches.
3899 @vindex diff-insert GCC_COLORS @r{capability}
3900 SGR substring for inserted lines within generated patches.
3903 @vindex type-diff GCC_COLORS @r{capability}
3904 SGR substring for highlighting mismatching types within template
3905 arguments in the C++ frontend.
3908 @item -fno-diagnostics-show-option
3909 @opindex fno-diagnostics-show-option
3910 @opindex fdiagnostics-show-option
3911 By default, each diagnostic emitted includes text indicating the
3912 command-line option that directly controls the diagnostic (if such an
3913 option is known to the diagnostic machinery). Specifying the
3914 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3916 @item -fno-diagnostics-show-caret
3917 @opindex fno-diagnostics-show-caret
3918 @opindex fdiagnostics-show-caret
3919 By default, each diagnostic emitted includes the original source line
3920 and a caret @samp{^} indicating the column. This option suppresses this
3921 information. The source line is truncated to @var{n} characters, if
3922 the @option{-fmessage-length=n} option is given. When the output is done
3923 to the terminal, the width is limited to the width given by the
3924 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3926 @item -fno-diagnostics-show-labels
3927 @opindex fno-diagnostics-show-labels
3928 @opindex fdiagnostics-show-labels
3929 By default, when printing source code (via @option{-fdiagnostics-show-caret}),
3930 diagnostics can label ranges of source code with pertinent information, such
3931 as the types of expressions:
3934 printf ("foo %s bar", long_i + long_j);
3940 This option suppresses the printing of these labels (in the example above,
3941 the vertical bars and the ``char *'' and ``long int'' text).
3943 @item -fno-diagnostics-show-line-numbers
3944 @opindex fno-diagnostics-show-line-numbers
3945 @opindex fdiagnostics-show-line-numbers
3946 By default, when printing source code (via @option{-fdiagnostics-show-caret}),
3947 a left margin is printed, showing line numbers. This option suppresses this
3950 @item -fdiagnostics-minimum-margin-width=@var{width}
3951 @opindex fdiagnostics-minimum-margin-width
3952 This option controls the minimum width of the left margin printed by
3953 @option{-fdiagnostics-show-line-numbers}. It defaults to 6.
3955 @item -fdiagnostics-parseable-fixits
3956 @opindex fdiagnostics-parseable-fixits
3957 Emit fix-it hints in a machine-parseable format, suitable for consumption
3958 by IDEs. For each fix-it, a line will be printed after the relevant
3959 diagnostic, starting with the string ``fix-it:''. For example:
3962 fix-it:"test.c":@{45:3-45:21@}:"gtk_widget_show_all"
3965 The location is expressed as a half-open range, expressed as a count of
3966 bytes, starting at byte 1 for the initial column. In the above example,
3967 bytes 3 through 20 of line 45 of ``test.c'' are to be replaced with the
3971 00000000011111111112222222222
3972 12345678901234567890123456789
3973 gtk_widget_showall (dlg);
3978 The filename and replacement string escape backslash as ``\\", tab as ``\t'',
3979 newline as ``\n'', double quotes as ``\"'', non-printable characters as octal
3980 (e.g. vertical tab as ``\013'').
3982 An empty replacement string indicates that the given range is to be removed.
3983 An empty range (e.g. ``45:3-45:3'') indicates that the string is to
3984 be inserted at the given position.
3986 @item -fdiagnostics-generate-patch
3987 @opindex fdiagnostics-generate-patch
3988 Print fix-it hints to stderr in unified diff format, after any diagnostics
3989 are printed. For example:
3996 void show_cb(GtkDialog *dlg)
3998 - gtk_widget_showall(dlg);
3999 + gtk_widget_show_all(dlg);
4004 The diff may or may not be colorized, following the same rules
4005 as for diagnostics (see @option{-fdiagnostics-color}).
4007 @item -fdiagnostics-show-template-tree
4008 @opindex fdiagnostics-show-template-tree
4010 In the C++ frontend, when printing diagnostics showing mismatching
4011 template types, such as:
4014 could not convert 'std::map<int, std::vector<double> >()'
4015 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
4018 the @option{-fdiagnostics-show-template-tree} flag enables printing a
4019 tree-like structure showing the common and differing parts of the types,
4029 The parts that differ are highlighted with color (``double'' and
4030 ``float'' in this case).
4032 @item -fno-elide-type
4033 @opindex fno-elide-type
4034 @opindex felide-type
4035 By default when the C++ frontend prints diagnostics showing mismatching
4036 template types, common parts of the types are printed as ``[...]'' to
4037 simplify the error message. For example:
4040 could not convert 'std::map<int, std::vector<double> >()'
4041 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
4044 Specifying the @option{-fno-elide-type} flag suppresses that behavior.
4045 This flag also affects the output of the
4046 @option{-fdiagnostics-show-template-tree} flag.
4048 @item -fno-show-column
4049 @opindex fno-show-column
4050 @opindex fshow-column
4051 Do not print column numbers in diagnostics. This may be necessary if
4052 diagnostics are being scanned by a program that does not understand the
4053 column numbers, such as @command{dejagnu}.
4055 @item -fdiagnostics-format=@var{FORMAT}
4056 @opindex fdiagnostics-format
4057 Select a different format for printing diagnostics.
4058 @var{FORMAT} is @samp{text} or @samp{json}.
4059 The default is @samp{text}.
4061 The @samp{json} format consists of a top-level JSON array containing JSON
4062 objects representing the diagnostics.
4064 The JSON is emitted as one line, without formatting; the examples below
4065 have been formatted for clarity.
4067 Diagnostics can have child diagnostics. For example, this error and note:
4070 misleading-indentation.c:15:3: warning: this 'if' clause does not
4071 guard... [-Wmisleading-indentation]
4074 misleading-indentation.c:17:5: note: ...this statement, but the latter
4075 is misleadingly indented as if it were guarded by the 'if'
4081 might be printed in JSON form (after formatting) like this:
4091 "file": "misleading-indentation.c",
4096 "file": "misleading-indentation.c",
4101 "message": "this \u2018if\u2019 clause does not guard...",
4102 "option": "-Wmisleading-indentation",
4110 "file": "misleading-indentation.c",
4115 "message": "...this statement, but the latter is @dots{}"
4124 where the @code{note} is a child of the @code{warning}.
4126 A diagnostic has a @code{kind}. If this is @code{warning}, then there is
4127 an @code{option} key describing the command-line option controlling the
4130 A diagnostic can contain zero or more locations. Each location has up
4131 to three positions within it: a @code{caret} position and optional
4132 @code{start} and @code{finish} positions. A location can also have
4133 an optional @code{label} string. For example, this error:
4136 bad-binary-ops.c:64:23: error: invalid operands to binary + (have 'S' @{aka
4137 'struct s'@} and 'T' @{aka 'struct t'@})
4138 64 | return callee_4a () + callee_4b ();
4139 | ~~~~~~~~~~~~ ^ ~~~~~~~~~~~~
4141 | | T @{aka struct t@}
4142 | S @{aka struct s@}
4146 has three locations. Its primary location is at the ``+'' token at column
4147 23. It has two secondary locations, describing the left and right-hand sides
4148 of the expression, which have labels. It might be printed in JSON form as:
4157 "column": 23, "file": "bad-binary-ops.c", "line": 64
4162 "column": 10, "file": "bad-binary-ops.c", "line": 64
4165 "column": 21, "file": "bad-binary-ops.c", "line": 64
4167 "label": "S @{aka struct s@}"
4171 "column": 25, "file": "bad-binary-ops.c", "line": 64
4174 "column": 36, "file": "bad-binary-ops.c", "line": 64
4176 "label": "T @{aka struct t@}"
4179 "message": "invalid operands to binary + @dots{}"
4183 If a diagnostic contains fix-it hints, it has a @code{fixits} array,
4184 consisting of half-open intervals, similar to the output of
4185 @option{-fdiagnostics-parseable-fixits}. For example, this diagnostic
4186 with a replacement fix-it hint:
4189 demo.c:8:15: error: 'struct s' has no member named 'colour'; did you
4191 8 | return ptr->colour;
4197 might be printed in JSON form as:
4232 "message": "\u2018struct s\u2019 has no member named @dots{}"
4237 where the fix-it hint suggests replacing the text from @code{start} up
4238 to but not including @code{next} with @code{string}'s value. Deletions
4239 are expressed via an empty value for @code{string}, insertions by
4240 having @code{start} equal @code{next}.
4244 @node Warning Options
4245 @section Options to Request or Suppress Warnings
4246 @cindex options to control warnings
4247 @cindex warning messages
4248 @cindex messages, warning
4249 @cindex suppressing warnings
4251 Warnings are diagnostic messages that report constructions that
4252 are not inherently erroneous but that are risky or suggest there
4253 may have been an error.
4255 The following language-independent options do not enable specific
4256 warnings but control the kinds of diagnostics produced by GCC@.
4259 @cindex syntax checking
4261 @opindex fsyntax-only
4262 Check the code for syntax errors, but don't do anything beyond that.
4264 @item -fmax-errors=@var{n}
4265 @opindex fmax-errors
4266 Limits the maximum number of error messages to @var{n}, at which point
4267 GCC bails out rather than attempting to continue processing the source
4268 code. If @var{n} is 0 (the default), there is no limit on the number
4269 of error messages produced. If @option{-Wfatal-errors} is also
4270 specified, then @option{-Wfatal-errors} takes precedence over this
4275 Inhibit all warning messages.
4280 Make all warnings into errors.
4285 Make the specified warning into an error. The specifier for a warning
4286 is appended; for example @option{-Werror=switch} turns the warnings
4287 controlled by @option{-Wswitch} into errors. This switch takes a
4288 negative form, to be used to negate @option{-Werror} for specific
4289 warnings; for example @option{-Wno-error=switch} makes
4290 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
4293 The warning message for each controllable warning includes the
4294 option that controls the warning. That option can then be used with
4295 @option{-Werror=} and @option{-Wno-error=} as described above.
4296 (Printing of the option in the warning message can be disabled using the
4297 @option{-fno-diagnostics-show-option} flag.)
4299 Note that specifying @option{-Werror=}@var{foo} automatically implies
4300 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
4303 @item -Wfatal-errors
4304 @opindex Wfatal-errors
4305 @opindex Wno-fatal-errors
4306 This option causes the compiler to abort compilation on the first error
4307 occurred rather than trying to keep going and printing further error
4312 You can request many specific warnings with options beginning with
4313 @samp{-W}, for example @option{-Wimplicit} to request warnings on
4314 implicit declarations. Each of these specific warning options also
4315 has a negative form beginning @samp{-Wno-} to turn off warnings; for
4316 example, @option{-Wno-implicit}. This manual lists only one of the
4317 two forms, whichever is not the default. For further
4318 language-specific options also refer to @ref{C++ Dialect Options} and
4319 @ref{Objective-C and Objective-C++ Dialect Options}.
4321 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
4322 options, such as @option{-Wunused}, which may turn on further options,
4323 such as @option{-Wunused-value}. The combined effect of positive and
4324 negative forms is that more specific options have priority over less
4325 specific ones, independently of their position in the command-line. For
4326 options of the same specificity, the last one takes effect. Options
4327 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
4328 as if they appeared at the end of the command-line.
4330 When an unrecognized warning option is requested (e.g.,
4331 @option{-Wunknown-warning}), GCC emits a diagnostic stating
4332 that the option is not recognized. However, if the @option{-Wno-} form
4333 is used, the behavior is slightly different: no diagnostic is
4334 produced for @option{-Wno-unknown-warning} unless other diagnostics
4335 are being produced. This allows the use of new @option{-Wno-} options
4336 with old compilers, but if something goes wrong, the compiler
4337 warns that an unrecognized option is present.
4344 @opindex Wno-pedantic
4345 Issue all the warnings demanded by strict ISO C and ISO C++;
4346 reject all programs that use forbidden extensions, and some other
4347 programs that do not follow ISO C and ISO C++. For ISO C, follows the
4348 version of the ISO C standard specified by any @option{-std} option used.
4350 Valid ISO C and ISO C++ programs should compile properly with or without
4351 this option (though a rare few require @option{-ansi} or a
4352 @option{-std} option specifying the required version of ISO C)@. However,
4353 without this option, certain GNU extensions and traditional C and C++
4354 features are supported as well. With this option, they are rejected.
4356 @option{-Wpedantic} does not cause warning messages for use of the
4357 alternate keywords whose names begin and end with @samp{__}. This alternate
4358 format can also be used to disable warnings for non-ISO @samp{__intN} types,
4359 i.e. @samp{__intN__}.
4360 Pedantic warnings are also disabled in the expression that follows
4361 @code{__extension__}. However, only system header files should use
4362 these escape routes; application programs should avoid them.
4363 @xref{Alternate Keywords}.
4365 Some users try to use @option{-Wpedantic} to check programs for strict ISO
4366 C conformance. They soon find that it does not do quite what they want:
4367 it finds some non-ISO practices, but not all---only those for which
4368 ISO C @emph{requires} a diagnostic, and some others for which
4369 diagnostics have been added.
4371 A feature to report any failure to conform to ISO C might be useful in
4372 some instances, but would require considerable additional work and would
4373 be quite different from @option{-Wpedantic}. We don't have plans to
4374 support such a feature in the near future.
4376 Where the standard specified with @option{-std} represents a GNU
4377 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
4378 corresponding @dfn{base standard}, the version of ISO C on which the GNU
4379 extended dialect is based. Warnings from @option{-Wpedantic} are given
4380 where they are required by the base standard. (It does not make sense
4381 for such warnings to be given only for features not in the specified GNU
4382 C dialect, since by definition the GNU dialects of C include all
4383 features the compiler supports with the given option, and there would be
4384 nothing to warn about.)
4386 @item -pedantic-errors
4387 @opindex pedantic-errors
4388 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
4389 requires a diagnostic, in some cases where there is undefined behavior
4390 at compile-time and in some other cases that do not prevent compilation
4391 of programs that are valid according to the standard. This is not
4392 equivalent to @option{-Werror=pedantic}, since there are errors enabled
4393 by this option and not enabled by the latter and vice versa.
4398 This enables all the warnings about constructions that some users
4399 consider questionable, and that are easy to avoid (or modify to
4400 prevent the warning), even in conjunction with macros. This also
4401 enables some language-specific warnings described in @ref{C++ Dialect
4402 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
4404 @option{-Wall} turns on the following warning flags:
4406 @gccoptlist{-Waddress @gol
4407 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
4409 -Wbool-operation @gol
4410 -Wc++11-compat -Wc++14-compat @gol
4411 -Wcatch-value @r{(C++ and Objective-C++ only)} @gol
4412 -Wchar-subscripts @gol
4414 -Wduplicate-decl-specifier @r{(C and Objective-C only)} @gol
4415 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
4416 -Wenum-conversion @r{in C/ObjC;} @gol
4418 -Wint-in-bool-context @gol
4419 -Wimplicit @r{(C and Objective-C only)} @gol
4420 -Wimplicit-int @r{(C and Objective-C only)} @gol
4421 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
4422 -Winit-self @r{(only for C++)} @gol
4423 -Wlogical-not-parentheses @gol
4424 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
4425 -Wmaybe-uninitialized @gol
4426 -Wmemset-elt-size @gol
4427 -Wmemset-transposed-args @gol
4428 -Wmisleading-indentation @r{(only for C/C++)} @gol
4429 -Wmissing-attributes @gol
4430 -Wmissing-braces @r{(only for C/ObjC)} @gol
4431 -Wmultistatement-macros @gol
4432 -Wnarrowing @r{(only for C++)} @gol
4434 -Wnonnull-compare @gol
4437 -Wpessimizing-move @r{(only for C++)} @gol
4442 -Wsequence-point @gol
4443 -Wsign-compare @r{(only in C++)} @gol
4444 -Wsizeof-pointer-div @gol
4445 -Wsizeof-pointer-memaccess @gol
4446 -Wstrict-aliasing @gol
4447 -Wstrict-overflow=1 @gol
4449 -Wtautological-compare @gol
4451 -Wuninitialized @gol
4452 -Wunknown-pragmas @gol
4453 -Wunused-function @gol
4456 -Wunused-variable @gol
4457 -Wvolatile-register-var}
4459 Note that some warning flags are not implied by @option{-Wall}. Some of
4460 them warn about constructions that users generally do not consider
4461 questionable, but which occasionally you might wish to check for;
4462 others warn about constructions that are necessary or hard to avoid in
4463 some cases, and there is no simple way to modify the code to suppress
4464 the warning. Some of them are enabled by @option{-Wextra} but many of
4465 them must be enabled individually.
4471 This enables some extra warning flags that are not enabled by
4472 @option{-Wall}. (This option used to be called @option{-W}. The older
4473 name is still supported, but the newer name is more descriptive.)
4475 @gccoptlist{-Wclobbered @gol
4476 -Wcast-function-type @gol
4477 -Wdeprecated-copy @r{(C++ only)} @gol
4479 -Wignored-qualifiers @gol
4480 -Wimplicit-fallthrough=3 @gol
4481 -Wmissing-field-initializers @gol
4482 -Wmissing-parameter-type @r{(C only)} @gol
4483 -Wold-style-declaration @r{(C only)} @gol
4484 -Woverride-init @gol
4485 -Wsign-compare @r{(C only)} @gol
4486 -Wredundant-move @r{(only for C++)} @gol
4488 -Wuninitialized @gol
4489 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
4490 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
4491 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)}}
4494 The option @option{-Wextra} also prints warning messages for the
4500 A pointer is compared against integer zero with @code{<}, @code{<=},
4501 @code{>}, or @code{>=}.
4504 (C++ only) An enumerator and a non-enumerator both appear in a
4505 conditional expression.
4508 (C++ only) Ambiguous virtual bases.
4511 (C++ only) Subscripting an array that has been declared @code{register}.
4514 (C++ only) Taking the address of a variable that has been declared
4518 (C++ only) A base class is not initialized in the copy constructor
4523 @item -Wchar-subscripts
4524 @opindex Wchar-subscripts
4525 @opindex Wno-char-subscripts
4526 Warn if an array subscript has type @code{char}. This is a common cause
4527 of error, as programmers often forget that this type is signed on some
4529 This warning is enabled by @option{-Wall}.
4531 @item -Wno-coverage-mismatch
4532 @opindex Wno-coverage-mismatch
4533 @opindex Wcoverage-mismatch
4534 Warn if feedback profiles do not match when using the
4535 @option{-fprofile-use} option.
4536 If a source file is changed between compiling with @option{-fprofile-generate}
4537 and with @option{-fprofile-use}, the files with the profile feedback can fail
4538 to match the source file and GCC cannot use the profile feedback
4539 information. By default, this warning is enabled and is treated as an
4540 error. @option{-Wno-coverage-mismatch} can be used to disable the
4541 warning or @option{-Wno-error=coverage-mismatch} can be used to
4542 disable the error. Disabling the error for this warning can result in
4543 poorly optimized code and is useful only in the
4544 case of very minor changes such as bug fixes to an existing code-base.
4545 Completely disabling the warning is not recommended.
4548 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
4550 Suppress warning messages emitted by @code{#warning} directives.
4552 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
4553 @opindex Wdouble-promotion
4554 @opindex Wno-double-promotion
4555 Give a warning when a value of type @code{float} is implicitly
4556 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
4557 floating-point unit implement @code{float} in hardware, but emulate
4558 @code{double} in software. On such a machine, doing computations
4559 using @code{double} values is much more expensive because of the
4560 overhead required for software emulation.
4562 It is easy to accidentally do computations with @code{double} because
4563 floating-point literals are implicitly of type @code{double}. For
4567 float area(float radius)
4569 return 3.14159 * radius * radius;
4573 the compiler performs the entire computation with @code{double}
4574 because the floating-point literal is a @code{double}.
4576 @item -Wduplicate-decl-specifier @r{(C and Objective-C only)}
4577 @opindex Wduplicate-decl-specifier
4578 @opindex Wno-duplicate-decl-specifier
4579 Warn if a declaration has duplicate @code{const}, @code{volatile},
4580 @code{restrict} or @code{_Atomic} specifier. This warning is enabled by
4584 @itemx -Wformat=@var{n}
4587 @opindex ffreestanding
4588 @opindex fno-builtin
4590 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
4591 the arguments supplied have types appropriate to the format string
4592 specified, and that the conversions specified in the format string make
4593 sense. This includes standard functions, and others specified by format
4594 attributes (@pxref{Function Attributes}), in the @code{printf},
4595 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
4596 not in the C standard) families (or other target-specific families).
4597 Which functions are checked without format attributes having been
4598 specified depends on the standard version selected, and such checks of
4599 functions without the attribute specified are disabled by
4600 @option{-ffreestanding} or @option{-fno-builtin}.
4602 The formats are checked against the format features supported by GNU
4603 libc version 2.2. These include all ISO C90 and C99 features, as well
4604 as features from the Single Unix Specification and some BSD and GNU
4605 extensions. Other library implementations may not support all these
4606 features; GCC does not support warning about features that go beyond a
4607 particular library's limitations. However, if @option{-Wpedantic} is used
4608 with @option{-Wformat}, warnings are given about format features not
4609 in the selected standard version (but not for @code{strfmon} formats,
4610 since those are not in any version of the C standard). @xref{C Dialect
4611 Options,,Options Controlling C Dialect}.
4618 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
4619 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
4620 @option{-Wformat} also checks for null format arguments for several
4621 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
4622 aspects of this level of format checking can be disabled by the
4623 options: @option{-Wno-format-contains-nul},
4624 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
4625 @option{-Wformat} is enabled by @option{-Wall}.
4627 @item -Wno-format-contains-nul
4628 @opindex Wno-format-contains-nul
4629 @opindex Wformat-contains-nul
4630 If @option{-Wformat} is specified, do not warn about format strings that
4633 @item -Wno-format-extra-args
4634 @opindex Wno-format-extra-args
4635 @opindex Wformat-extra-args
4636 If @option{-Wformat} is specified, do not warn about excess arguments to a
4637 @code{printf} or @code{scanf} format function. The C standard specifies
4638 that such arguments are ignored.
4640 Where the unused arguments lie between used arguments that are
4641 specified with @samp{$} operand number specifications, normally
4642 warnings are still given, since the implementation could not know what
4643 type to pass to @code{va_arg} to skip the unused arguments. However,
4644 in the case of @code{scanf} formats, this option suppresses the
4645 warning if the unused arguments are all pointers, since the Single
4646 Unix Specification says that such unused arguments are allowed.
4648 @item -Wformat-overflow
4649 @itemx -Wformat-overflow=@var{level}
4650 @opindex Wformat-overflow
4651 @opindex Wno-format-overflow
4652 Warn about calls to formatted input/output functions such as @code{sprintf}
4653 and @code{vsprintf} that might overflow the destination buffer. When the
4654 exact number of bytes written by a format directive cannot be determined
4655 at compile-time it is estimated based on heuristics that depend on the
4656 @var{level} argument and on optimization. While enabling optimization
4657 will in most cases improve the accuracy of the warning, it may also
4658 result in false positives.
4661 @item -Wformat-overflow
4662 @itemx -Wformat-overflow=1
4663 @opindex Wformat-overflow
4664 @opindex Wno-format-overflow
4665 Level @var{1} of @option{-Wformat-overflow} enabled by @option{-Wformat}
4666 employs a conservative approach that warns only about calls that most
4667 likely overflow the buffer. At this level, numeric arguments to format
4668 directives with unknown values are assumed to have the value of one, and
4669 strings of unknown length to be empty. Numeric arguments that are known
4670 to be bounded to a subrange of their type, or string arguments whose output
4671 is bounded either by their directive's precision or by a finite set of
4672 string literals, are assumed to take on the value within the range that
4673 results in the most bytes on output. For example, the call to @code{sprintf}
4674 below is diagnosed because even with both @var{a} and @var{b} equal to zero,
4675 the terminating NUL character (@code{'\0'}) appended by the function
4676 to the destination buffer will be written past its end. Increasing
4677 the size of the buffer by a single byte is sufficient to avoid the
4678 warning, though it may not be sufficient to avoid the overflow.
4681 void f (int a, int b)
4684 sprintf (buf, "a = %i, b = %i\n", a, b);
4688 @item -Wformat-overflow=2
4689 Level @var{2} warns also about calls that might overflow the destination
4690 buffer given an argument of sufficient length or magnitude. At level
4691 @var{2}, unknown numeric arguments are assumed to have the minimum
4692 representable value for signed types with a precision greater than 1, and
4693 the maximum representable value otherwise. Unknown string arguments whose
4694 length cannot be assumed to be bounded either by the directive's precision,
4695 or by a finite set of string literals they may evaluate to, or the character
4696 array they may point to, are assumed to be 1 character long.
4698 At level @var{2}, the call in the example above is again diagnosed, but
4699 this time because with @var{a} equal to a 32-bit @code{INT_MIN} the first
4700 @code{%i} directive will write some of its digits beyond the end of
4701 the destination buffer. To make the call safe regardless of the values
4702 of the two variables, the size of the destination buffer must be increased
4703 to at least 34 bytes. GCC includes the minimum size of the buffer in
4704 an informational note following the warning.
4706 An alternative to increasing the size of the destination buffer is to
4707 constrain the range of formatted values. The maximum length of string
4708 arguments can be bounded by specifying the precision in the format
4709 directive. When numeric arguments of format directives can be assumed
4710 to be bounded by less than the precision of their type, choosing
4711 an appropriate length modifier to the format specifier will reduce
4712 the required buffer size. For example, if @var{a} and @var{b} in the
4713 example above can be assumed to be within the precision of
4714 the @code{short int} type then using either the @code{%hi} format
4715 directive or casting the argument to @code{short} reduces the maximum
4716 required size of the buffer to 24 bytes.
4719 void f (int a, int b)
4722 sprintf (buf, "a = %hi, b = %i\n", a, (short)b);
4727 @item -Wno-format-zero-length
4728 @opindex Wno-format-zero-length
4729 @opindex Wformat-zero-length
4730 If @option{-Wformat} is specified, do not warn about zero-length formats.
4731 The C standard specifies that zero-length formats are allowed.
4736 Enable @option{-Wformat} plus additional format checks. Currently
4737 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
4740 @item -Wformat-nonliteral
4741 @opindex Wformat-nonliteral
4742 @opindex Wno-format-nonliteral
4743 If @option{-Wformat} is specified, also warn if the format string is not a
4744 string literal and so cannot be checked, unless the format function
4745 takes its format arguments as a @code{va_list}.
4747 @item -Wformat-security
4748 @opindex Wformat-security
4749 @opindex Wno-format-security
4750 If @option{-Wformat} is specified, also warn about uses of format
4751 functions that represent possible security problems. At present, this
4752 warns about calls to @code{printf} and @code{scanf} functions where the
4753 format string is not a string literal and there are no format arguments,
4754 as in @code{printf (foo);}. This may be a security hole if the format
4755 string came from untrusted input and contains @samp{%n}. (This is
4756 currently a subset of what @option{-Wformat-nonliteral} warns about, but
4757 in future warnings may be added to @option{-Wformat-security} that are not
4758 included in @option{-Wformat-nonliteral}.)
4760 @item -Wformat-signedness
4761 @opindex Wformat-signedness
4762 @opindex Wno-format-signedness
4763 If @option{-Wformat} is specified, also warn if the format string
4764 requires an unsigned argument and the argument is signed and vice versa.
4766 @item -Wformat-truncation
4767 @itemx -Wformat-truncation=@var{level}
4768 @opindex Wformat-truncation
4769 @opindex Wno-format-truncation
4770 Warn about calls to formatted input/output functions such as @code{snprintf}
4771 and @code{vsnprintf} that might result in output truncation. When the exact
4772 number of bytes written by a format directive cannot be determined at
4773 compile-time it is estimated based on heuristics that depend on
4774 the @var{level} argument and on optimization. While enabling optimization
4775 will in most cases improve the accuracy of the warning, it may also result
4776 in false positives. Except as noted otherwise, the option uses the same
4777 logic @option{-Wformat-overflow}.
4780 @item -Wformat-truncation
4781 @itemx -Wformat-truncation=1
4782 @opindex Wformat-truncation
4783 @opindex Wno-format-truncation
4784 Level @var{1} of @option{-Wformat-truncation} enabled by @option{-Wformat}
4785 employs a conservative approach that warns only about calls to bounded
4786 functions whose return value is unused and that will most likely result
4787 in output truncation.
4789 @item -Wformat-truncation=2
4790 Level @var{2} warns also about calls to bounded functions whose return
4791 value is used and that might result in truncation given an argument of
4792 sufficient length or magnitude.
4796 @opindex Wformat-y2k
4797 @opindex Wno-format-y2k
4798 If @option{-Wformat} is specified, also warn about @code{strftime}
4799 formats that may yield only a two-digit year.
4804 @opindex Wno-nonnull
4805 Warn about passing a null pointer for arguments marked as
4806 requiring a non-null value by the @code{nonnull} function attribute.
4808 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
4809 can be disabled with the @option{-Wno-nonnull} option.
4811 @item -Wnonnull-compare
4812 @opindex Wnonnull-compare
4813 @opindex Wno-nonnull-compare
4814 Warn when comparing an argument marked with the @code{nonnull}
4815 function attribute against null inside the function.
4817 @option{-Wnonnull-compare} is included in @option{-Wall}. It
4818 can be disabled with the @option{-Wno-nonnull-compare} option.
4820 @item -Wnull-dereference
4821 @opindex Wnull-dereference
4822 @opindex Wno-null-dereference
4823 Warn if the compiler detects paths that trigger erroneous or
4824 undefined behavior due to dereferencing a null pointer. This option
4825 is only active when @option{-fdelete-null-pointer-checks} is active,
4826 which is enabled by optimizations in most targets. The precision of
4827 the warnings depends on the optimization options used.
4829 @item -Winaccessible-base @r{(C++, Objective-C++ only)}
4830 @opindex Winaccessible-base
4831 @opindex Wno-inaccessible-base
4832 Warn when a base class is inaccessible in a class derived from it due to
4833 ambiguity. The warning is enabled by default. Note the warning for virtual
4834 bases is enabled by the @option{-Wextra} option.
4837 struct A @{ int a; @};
4841 struct C : B, A @{ @};
4845 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
4847 @opindex Wno-init-self
4848 Warn about uninitialized variables that are initialized with themselves.
4849 Note this option can only be used with the @option{-Wuninitialized} option.
4851 For example, GCC warns about @code{i} being uninitialized in the
4852 following snippet only when @option{-Winit-self} has been specified:
4863 This warning is enabled by @option{-Wall} in C++.
4865 @item -Wimplicit-int @r{(C and Objective-C only)}
4866 @opindex Wimplicit-int
4867 @opindex Wno-implicit-int
4868 Warn when a declaration does not specify a type.
4869 This warning is enabled by @option{-Wall}.
4871 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
4872 @opindex Wimplicit-function-declaration
4873 @opindex Wno-implicit-function-declaration
4874 Give a warning whenever a function is used before being declared. In
4875 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
4876 enabled by default and it is made into an error by
4877 @option{-pedantic-errors}. This warning is also enabled by
4880 @item -Wimplicit @r{(C and Objective-C only)}
4882 @opindex Wno-implicit
4883 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
4884 This warning is enabled by @option{-Wall}.
4886 @item -Wimplicit-fallthrough
4887 @opindex Wimplicit-fallthrough
4888 @opindex Wno-implicit-fallthrough
4889 @option{-Wimplicit-fallthrough} is the same as @option{-Wimplicit-fallthrough=3}
4890 and @option{-Wno-implicit-fallthrough} is the same as
4891 @option{-Wimplicit-fallthrough=0}.
4893 @item -Wimplicit-fallthrough=@var{n}
4894 @opindex Wimplicit-fallthrough=
4895 Warn when a switch case falls through. For example:
4913 This warning does not warn when the last statement of a case cannot
4914 fall through, e.g. when there is a return statement or a call to function
4915 declared with the noreturn attribute. @option{-Wimplicit-fallthrough=}
4916 also takes into account control flow statements, such as ifs, and only
4917 warns when appropriate. E.g.@:
4927 @} else if (i < 1) @{
4937 Since there are occasions where a switch case fall through is desirable,
4938 GCC provides an attribute, @code{__attribute__ ((fallthrough))}, that is
4939 to be used along with a null statement to suppress this warning that
4940 would normally occur:
4948 __attribute__ ((fallthrough));
4955 C++17 provides a standard way to suppress the @option{-Wimplicit-fallthrough}
4956 warning using @code{[[fallthrough]];} instead of the GNU attribute. In C++11
4957 or C++14 users can use @code{[[gnu::fallthrough]];}, which is a GNU extension.
4958 Instead of these attributes, it is also possible to add a fallthrough comment
4959 to silence the warning. The whole body of the C or C++ style comment should
4960 match the given regular expressions listed below. The option argument @var{n}
4961 specifies what kind of comments are accepted:
4965 @item @option{-Wimplicit-fallthrough=0} disables the warning altogether.
4967 @item @option{-Wimplicit-fallthrough=1} matches @code{.*} regular
4968 expression, any comment is used as fallthrough comment.
4970 @item @option{-Wimplicit-fallthrough=2} case insensitively matches
4971 @code{.*falls?[ \t-]*thr(ough|u).*} regular expression.
4973 @item @option{-Wimplicit-fallthrough=3} case sensitively matches one of the
4974 following regular expressions:
4978 @item @code{-fallthrough}
4980 @item @code{@@fallthrough@@}
4982 @item @code{lint -fallthrough[ \t]*}
4984 @item @code{[ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?@*FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)?}
4986 @item @code{[ \t.!]*(Else,? |Intentional(ly)? )?@*Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4988 @item @code{[ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?@*fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4992 @item @option{-Wimplicit-fallthrough=4} case sensitively matches one of the
4993 following regular expressions:
4997 @item @code{-fallthrough}
4999 @item @code{@@fallthrough@@}
5001 @item @code{lint -fallthrough[ \t]*}
5003 @item @code{[ \t]*FALLTHR(OUGH|U)[ \t]*}
5007 @item @option{-Wimplicit-fallthrough=5} doesn't recognize any comments as
5008 fallthrough comments, only attributes disable the warning.
5012 The comment needs to be followed after optional whitespace and other comments
5013 by @code{case} or @code{default} keywords or by a user label that precedes some
5014 @code{case} or @code{default} label.
5029 The @option{-Wimplicit-fallthrough=3} warning is enabled by @option{-Wextra}.
5031 @item -Wif-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
5032 @opindex Wif-not-aligned
5033 @opindex Wno-if-not-aligned
5034 Control if warning triggered by the @code{warn_if_not_aligned} attribute
5035 should be issued. This is enabled by default.
5036 Use @option{-Wno-if-not-aligned} to disable it.
5038 @item -Wignored-qualifiers @r{(C and C++ only)}
5039 @opindex Wignored-qualifiers
5040 @opindex Wno-ignored-qualifiers
5041 Warn if the return type of a function has a type qualifier
5042 such as @code{const}. For ISO C such a type qualifier has no effect,
5043 since the value returned by a function is not an lvalue.
5044 For C++, the warning is only emitted for scalar types or @code{void}.
5045 ISO C prohibits qualified @code{void} return types on function
5046 definitions, so such return types always receive a warning
5047 even without this option.
5049 This warning is also enabled by @option{-Wextra}.
5051 @item -Wignored-attributes @r{(C and C++ only)}
5052 @opindex Wignored-attributes
5053 @opindex Wno-ignored-attributes
5054 Warn when an attribute is ignored. This is different from the
5055 @option{-Wattributes} option in that it warns whenever the compiler decides
5056 to drop an attribute, not that the attribute is either unknown, used in a
5057 wrong place, etc. This warning is enabled by default.
5062 Warn if the type of @code{main} is suspicious. @code{main} should be
5063 a function with external linkage, returning int, taking either zero
5064 arguments, two, or three arguments of appropriate types. This warning
5065 is enabled by default in C++ and is enabled by either @option{-Wall}
5066 or @option{-Wpedantic}.
5068 @item -Wmisleading-indentation @r{(C and C++ only)}
5069 @opindex Wmisleading-indentation
5070 @opindex Wno-misleading-indentation
5071 Warn when the indentation of the code does not reflect the block structure.
5072 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
5073 @code{for} clauses with a guarded statement that does not use braces,
5074 followed by an unguarded statement with the same indentation.
5076 In the following example, the call to ``bar'' is misleadingly indented as
5077 if it were guarded by the ``if'' conditional.
5080 if (some_condition ())
5082 bar (); /* Gotcha: this is not guarded by the "if". */
5085 In the case of mixed tabs and spaces, the warning uses the
5086 @option{-ftabstop=} option to determine if the statements line up
5089 The warning is not issued for code involving multiline preprocessor logic
5090 such as the following example.
5095 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
5101 The warning is not issued after a @code{#line} directive, since this
5102 typically indicates autogenerated code, and no assumptions can be made
5103 about the layout of the file that the directive references.
5105 This warning is enabled by @option{-Wall} in C and C++.
5107 @item -Wmissing-attributes
5108 @opindex Wmissing-attributes
5109 @opindex Wno-missing-attributes
5110 Warn when a declaration of a function is missing one or more attributes
5111 that a related function is declared with and whose absence may adversely
5112 affect the correctness or efficiency of generated code. For example,
5113 the warning is issued for declarations of aliases that use attributes
5114 to specify less restrictive requirements than those of their targets.
5115 This typically represents a potential optimization opportunity.
5116 By contrast, the @option{-Wattribute-alias=2} option controls warnings
5117 issued when the alias is more restrictive than the target, which could
5118 lead to incorrect code generation.
5119 Attributes considered include @code{alloc_align}, @code{alloc_size},
5120 @code{cold}, @code{const}, @code{hot}, @code{leaf}, @code{malloc},
5121 @code{nonnull}, @code{noreturn}, @code{nothrow}, @code{pure},
5122 @code{returns_nonnull}, and @code{returns_twice}.
5124 In C++, the warning is issued when an explicit specialization of a primary
5125 template declared with attribute @code{alloc_align}, @code{alloc_size},
5126 @code{assume_aligned}, @code{format}, @code{format_arg}, @code{malloc},
5127 or @code{nonnull} is declared without it. Attributes @code{deprecated},
5128 @code{error}, and @code{warning} suppress the warning.
5129 (@pxref{Function Attributes}).
5131 You can use the @code{copy} attribute to apply the same
5132 set of attributes to a declaration as that on another declaration without
5133 explicitly enumerating the attributes. This attribute can be applied
5134 to declarations of functions (@pxref{Common Function Attributes}),
5135 variables (@pxref{Common Variable Attributes}), or types
5136 (@pxref{Common Type Attributes}).
5138 @option{-Wmissing-attributes} is enabled by @option{-Wall}.
5140 For example, since the declaration of the primary function template
5141 below makes use of both attribute @code{malloc} and @code{alloc_size}
5142 the declaration of the explicit specialization of the template is
5143 diagnosed because it is missing one of the attributes.
5147 T* __attribute__ ((malloc, alloc_size (1)))
5151 void* __attribute__ ((malloc)) // missing alloc_size
5152 allocate<void> (size_t);
5155 @item -Wmissing-braces
5156 @opindex Wmissing-braces
5157 @opindex Wno-missing-braces
5158 Warn if an aggregate or union initializer is not fully bracketed. In
5159 the following example, the initializer for @code{a} is not fully
5160 bracketed, but that for @code{b} is fully bracketed. This warning is
5161 enabled by @option{-Wall} in C.
5164 int a[2][2] = @{ 0, 1, 2, 3 @};
5165 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
5168 This warning is enabled by @option{-Wall}.
5170 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
5171 @opindex Wmissing-include-dirs
5172 @opindex Wno-missing-include-dirs
5173 Warn if a user-supplied include directory does not exist.
5175 @item -Wmissing-profile
5176 @opindex Wmissing-profile
5177 @opindex Wno-missing-profile
5178 Warn if feedback profiles are missing when using the
5179 @option{-fprofile-use} option.
5180 This option diagnoses those cases where a new function or a new file is added
5181 to the user code between compiling with @option{-fprofile-generate} and with
5182 @option{-fprofile-use}, without regenerating the profiles. In these cases, the
5183 profile feedback data files do not contain any profile feedback information for
5184 the newly added function or file respectively. Also, in the case when profile
5185 count data (.gcda) files are removed, GCC cannot use any profile feedback
5186 information. In all these cases, warnings are issued to inform the user that a
5187 profile generation step is due. @option{-Wno-missing-profile} can be used to
5188 disable the warning. Ignoring the warning can result in poorly optimized code.
5189 Completely disabling the warning is not recommended and should be done only
5190 when non-existent profile data is justified.
5192 @item -Wmultistatement-macros
5193 @opindex Wmultistatement-macros
5194 @opindex Wno-multistatement-macros
5195 Warn about unsafe multiple statement macros that appear to be guarded
5196 by a clause such as @code{if}, @code{else}, @code{for}, @code{switch}, or
5197 @code{while}, in which only the first statement is actually guarded after
5198 the macro is expanded.
5203 #define DOIT x++; y++
5208 will increment @code{y} unconditionally, not just when @code{c} holds.
5209 The can usually be fixed by wrapping the macro in a do-while loop:
5211 #define DOIT do @{ x++; y++; @} while (0)
5216 This warning is enabled by @option{-Wall} in C and C++.
5219 @opindex Wparentheses
5220 @opindex Wno-parentheses
5221 Warn if parentheses are omitted in certain contexts, such
5222 as when there is an assignment in a context where a truth value
5223 is expected, or when operators are nested whose precedence people
5224 often get confused about.
5226 Also warn if a comparison like @code{x<=y<=z} appears; this is
5227 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
5228 interpretation from that of ordinary mathematical notation.
5230 Also warn for dangerous uses of the GNU extension to
5231 @code{?:} with omitted middle operand. When the condition
5232 in the @code{?}: operator is a boolean expression, the omitted value is
5233 always 1. Often programmers expect it to be a value computed
5234 inside the conditional expression instead.
5236 For C++ this also warns for some cases of unnecessary parentheses in
5237 declarations, which can indicate an attempt at a function call instead
5241 // Declares a local variable called mymutex.
5242 std::unique_lock<std::mutex> (mymutex);
5243 // User meant std::unique_lock<std::mutex> lock (mymutex);
5247 This warning is enabled by @option{-Wall}.
5249 @item -Wsequence-point
5250 @opindex Wsequence-point
5251 @opindex Wno-sequence-point
5252 Warn about code that may have undefined semantics because of violations
5253 of sequence point rules in the C and C++ standards.
5255 The C and C++ standards define the order in which expressions in a C/C++
5256 program are evaluated in terms of @dfn{sequence points}, which represent
5257 a partial ordering between the execution of parts of the program: those
5258 executed before the sequence point, and those executed after it. These
5259 occur after the evaluation of a full expression (one which is not part
5260 of a larger expression), after the evaluation of the first operand of a
5261 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
5262 function is called (but after the evaluation of its arguments and the
5263 expression denoting the called function), and in certain other places.
5264 Other than as expressed by the sequence point rules, the order of
5265 evaluation of subexpressions of an expression is not specified. All
5266 these rules describe only a partial order rather than a total order,
5267 since, for example, if two functions are called within one expression
5268 with no sequence point between them, the order in which the functions
5269 are called is not specified. However, the standards committee have
5270 ruled that function calls do not overlap.
5272 It is not specified when between sequence points modifications to the
5273 values of objects take effect. Programs whose behavior depends on this
5274 have undefined behavior; the C and C++ standards specify that ``Between
5275 the previous and next sequence point an object shall have its stored
5276 value modified at most once by the evaluation of an expression.
5277 Furthermore, the prior value shall be read only to determine the value
5278 to be stored.''. If a program breaks these rules, the results on any
5279 particular implementation are entirely unpredictable.
5281 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
5282 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
5283 diagnosed by this option, and it may give an occasional false positive
5284 result, but in general it has been found fairly effective at detecting
5285 this sort of problem in programs.
5287 The C++17 standard will define the order of evaluation of operands in
5288 more cases: in particular it requires that the right-hand side of an
5289 assignment be evaluated before the left-hand side, so the above
5290 examples are no longer undefined. But this warning will still warn
5291 about them, to help people avoid writing code that is undefined in C
5292 and earlier revisions of C++.
5294 The standard is worded confusingly, therefore there is some debate
5295 over the precise meaning of the sequence point rules in subtle cases.
5296 Links to discussions of the problem, including proposed formal
5297 definitions, may be found on the GCC readings page, at
5298 @uref{http://gcc.gnu.org/@/readings.html}.
5300 This warning is enabled by @option{-Wall} for C and C++.
5302 @item -Wno-return-local-addr
5303 @opindex Wno-return-local-addr
5304 @opindex Wreturn-local-addr
5305 Do not warn about returning a pointer (or in C++, a reference) to a
5306 variable that goes out of scope after the function returns.
5309 @opindex Wreturn-type
5310 @opindex Wno-return-type
5311 Warn whenever a function is defined with a return type that defaults
5312 to @code{int}. Also warn about any @code{return} statement with no
5313 return value in a function whose return type is not @code{void}
5314 (falling off the end of the function body is considered returning
5317 For C only, warn about a @code{return} statement with an expression in a
5318 function whose return type is @code{void}, unless the expression type is
5319 also @code{void}. As a GNU extension, the latter case is accepted
5320 without a warning unless @option{-Wpedantic} is used. Attempting
5321 to use the return value of a non-@code{void} function other than @code{main}
5322 that flows off the end by reaching the closing curly brace that terminates
5323 the function is undefined.
5325 Unlike in C, in C++, flowing off the end of a non-@code{void} function other
5326 than @code{main} results in undefined behavior even when the value of
5327 the function is not used.
5329 This warning is enabled by default in C++ and by @option{-Wall} otherwise.
5331 @item -Wshift-count-negative
5332 @opindex Wshift-count-negative
5333 @opindex Wno-shift-count-negative
5334 Warn if shift count is negative. This warning is enabled by default.
5336 @item -Wshift-count-overflow
5337 @opindex Wshift-count-overflow
5338 @opindex Wno-shift-count-overflow
5339 Warn if shift count >= width of type. This warning is enabled by default.
5341 @item -Wshift-negative-value
5342 @opindex Wshift-negative-value
5343 @opindex Wno-shift-negative-value
5344 Warn if left shifting a negative value. This warning is enabled by
5345 @option{-Wextra} in C99 and C++11 modes (and newer).
5347 @item -Wshift-overflow
5348 @itemx -Wshift-overflow=@var{n}
5349 @opindex Wshift-overflow
5350 @opindex Wno-shift-overflow
5351 Warn about left shift overflows. This warning is enabled by
5352 default in C99 and C++11 modes (and newer).
5355 @item -Wshift-overflow=1
5356 This is the warning level of @option{-Wshift-overflow} and is enabled
5357 by default in C99 and C++11 modes (and newer). This warning level does
5358 not warn about left-shifting 1 into the sign bit. (However, in C, such
5359 an overflow is still rejected in contexts where an integer constant expression
5360 is required.) No warning is emitted in C++2A mode (and newer), as signed left
5363 @item -Wshift-overflow=2
5364 This warning level also warns about left-shifting 1 into the sign bit,
5365 unless C++14 mode (or newer) is active.
5371 Warn whenever a @code{switch} statement has an index of enumerated type
5372 and lacks a @code{case} for one or more of the named codes of that
5373 enumeration. (The presence of a @code{default} label prevents this
5374 warning.) @code{case} labels outside the enumeration range also
5375 provoke warnings when this option is used (even if there is a
5376 @code{default} label).
5377 This warning is enabled by @option{-Wall}.
5379 @item -Wswitch-default
5380 @opindex Wswitch-default
5381 @opindex Wno-switch-default
5382 Warn whenever a @code{switch} statement does not have a @code{default}
5386 @opindex Wswitch-enum
5387 @opindex Wno-switch-enum
5388 Warn whenever a @code{switch} statement has an index of enumerated type
5389 and lacks a @code{case} for one or more of the named codes of that
5390 enumeration. @code{case} labels outside the enumeration range also
5391 provoke warnings when this option is used. The only difference
5392 between @option{-Wswitch} and this option is that this option gives a
5393 warning about an omitted enumeration code even if there is a
5394 @code{default} label.
5397 @opindex Wswitch-bool
5398 @opindex Wno-switch-bool
5399 Warn whenever a @code{switch} statement has an index of boolean type
5400 and the case values are outside the range of a boolean type.
5401 It is possible to suppress this warning by casting the controlling
5402 expression to a type other than @code{bool}. For example:
5405 switch ((int) (a == 4))
5411 This warning is enabled by default for C and C++ programs.
5413 @item -Wswitch-outside-range
5414 @opindex Wswitch-outside-range
5415 @opindex Wno-switch-outside-range
5416 Warn whenever a @code{switch} case has a value that is outside of its
5417 respective type range. This warning is enabled by default for
5420 @item -Wswitch-unreachable
5421 @opindex Wswitch-unreachable
5422 @opindex Wno-switch-unreachable
5423 Warn whenever a @code{switch} statement contains statements between the
5424 controlling expression and the first case label, which will never be
5425 executed. For example:
5437 @option{-Wswitch-unreachable} does not warn if the statement between the
5438 controlling expression and the first case label is just a declaration:
5451 This warning is enabled by default for C and C++ programs.
5453 @item -Wsync-nand @r{(C and C++ only)}
5455 @opindex Wno-sync-nand
5456 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
5457 built-in functions are used. These functions changed semantics in GCC 4.4.
5459 @item -Wunused-but-set-parameter
5460 @opindex Wunused-but-set-parameter
5461 @opindex Wno-unused-but-set-parameter
5462 Warn whenever a function parameter is assigned to, but otherwise unused
5463 (aside from its declaration).
5465 To suppress this warning use the @code{unused} attribute
5466 (@pxref{Variable Attributes}).
5468 This warning is also enabled by @option{-Wunused} together with
5471 @item -Wunused-but-set-variable
5472 @opindex Wunused-but-set-variable
5473 @opindex Wno-unused-but-set-variable
5474 Warn whenever a local variable is assigned to, but otherwise unused
5475 (aside from its declaration).
5476 This warning is enabled by @option{-Wall}.
5478 To suppress this warning use the @code{unused} attribute
5479 (@pxref{Variable Attributes}).
5481 This warning is also enabled by @option{-Wunused}, which is enabled
5484 @item -Wunused-function
5485 @opindex Wunused-function
5486 @opindex Wno-unused-function
5487 Warn whenever a static function is declared but not defined or a
5488 non-inline static function is unused.
5489 This warning is enabled by @option{-Wall}.
5491 @item -Wunused-label
5492 @opindex Wunused-label
5493 @opindex Wno-unused-label
5494 Warn whenever a label is declared but not used.
5495 This warning is enabled by @option{-Wall}.
5497 To suppress this warning use the @code{unused} attribute
5498 (@pxref{Variable Attributes}).
5500 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
5501 @opindex Wunused-local-typedefs
5502 @opindex Wno-unused-local-typedefs
5503 Warn when a typedef locally defined in a function is not used.
5504 This warning is enabled by @option{-Wall}.
5506 @item -Wunused-parameter
5507 @opindex Wunused-parameter
5508 @opindex Wno-unused-parameter
5509 Warn whenever a function parameter is unused aside from its declaration.
5511 To suppress this warning use the @code{unused} attribute
5512 (@pxref{Variable Attributes}).
5514 @item -Wno-unused-result
5515 @opindex Wunused-result
5516 @opindex Wno-unused-result
5517 Do not warn if a caller of a function marked with attribute
5518 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
5519 its return value. The default is @option{-Wunused-result}.
5521 @item -Wunused-variable
5522 @opindex Wunused-variable
5523 @opindex Wno-unused-variable
5524 Warn whenever a local or static variable is unused aside from its
5525 declaration. This option implies @option{-Wunused-const-variable=1} for C,
5526 but not for C++. This warning is enabled by @option{-Wall}.
5528 To suppress this warning use the @code{unused} attribute
5529 (@pxref{Variable Attributes}).
5531 @item -Wunused-const-variable
5532 @itemx -Wunused-const-variable=@var{n}
5533 @opindex Wunused-const-variable
5534 @opindex Wno-unused-const-variable
5535 Warn whenever a constant static variable is unused aside from its declaration.
5536 @option{-Wunused-const-variable=1} is enabled by @option{-Wunused-variable}
5537 for C, but not for C++. In C this declares variable storage, but in C++ this
5538 is not an error since const variables take the place of @code{#define}s.
5540 To suppress this warning use the @code{unused} attribute
5541 (@pxref{Variable Attributes}).
5544 @item -Wunused-const-variable=1
5545 This is the warning level that is enabled by @option{-Wunused-variable} for
5546 C. It warns only about unused static const variables defined in the main
5547 compilation unit, but not about static const variables declared in any
5550 @item -Wunused-const-variable=2
5551 This warning level also warns for unused constant static variables in
5552 headers (excluding system headers). This is the warning level of
5553 @option{-Wunused-const-variable} and must be explicitly requested since
5554 in C++ this isn't an error and in C it might be harder to clean up all
5558 @item -Wunused-value
5559 @opindex Wunused-value
5560 @opindex Wno-unused-value
5561 Warn whenever a statement computes a result that is explicitly not
5562 used. To suppress this warning cast the unused expression to
5563 @code{void}. This includes an expression-statement or the left-hand
5564 side of a comma expression that contains no side effects. For example,
5565 an expression such as @code{x[i,j]} causes a warning, while
5566 @code{x[(void)i,j]} does not.
5568 This warning is enabled by @option{-Wall}.
5573 All the above @option{-Wunused} options combined.
5575 In order to get a warning about an unused function parameter, you must
5576 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
5577 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
5579 @item -Wuninitialized
5580 @opindex Wuninitialized
5581 @opindex Wno-uninitialized
5582 Warn if an automatic variable is used without first being initialized
5583 or if a variable may be clobbered by a @code{setjmp} call. In C++,
5584 warn if a non-static reference or non-static @code{const} member
5585 appears in a class without constructors.
5587 If you want to warn about code that uses the uninitialized value of the
5588 variable in its own initializer, use the @option{-Winit-self} option.
5590 These warnings occur for individual uninitialized or clobbered
5591 elements of structure, union or array variables as well as for
5592 variables that are uninitialized or clobbered as a whole. They do
5593 not occur for variables or elements declared @code{volatile}. Because
5594 these warnings depend on optimization, the exact variables or elements
5595 for which there are warnings depends on the precise optimization
5596 options and version of GCC used.
5598 Note that there may be no warning about a variable that is used only
5599 to compute a value that itself is never used, because such
5600 computations may be deleted by data flow analysis before the warnings
5603 @item -Winvalid-memory-model
5604 @opindex Winvalid-memory-model
5605 @opindex Wno-invalid-memory-model
5606 Warn for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins},
5607 and the C11 atomic generic functions with a memory consistency argument
5608 that is either invalid for the operation or outside the range of values
5609 of the @code{memory_order} enumeration. For example, since the
5610 @code{__atomic_store} and @code{__atomic_store_n} built-ins are only
5611 defined for the relaxed, release, and sequentially consistent memory
5612 orders the following code is diagnosed:
5617 __atomic_store_n (i, 0, memory_order_consume);
5621 @option{-Winvalid-memory-model} is enabled by default.
5623 @item -Wmaybe-uninitialized
5624 @opindex Wmaybe-uninitialized
5625 @opindex Wno-maybe-uninitialized
5626 For an automatic (i.e.@: local) variable, if there exists a path from the
5627 function entry to a use of the variable that is initialized, but there exist
5628 some other paths for which the variable is not initialized, the compiler
5629 emits a warning if it cannot prove the uninitialized paths are not
5630 executed at run time.
5632 These warnings are only possible in optimizing compilation, because otherwise
5633 GCC does not keep track of the state of variables.
5635 These warnings are made optional because GCC may not be able to determine when
5636 the code is correct in spite of appearing to have an error. Here is one
5637 example of how this can happen:
5657 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
5658 always initialized, but GCC doesn't know this. To suppress the
5659 warning, you need to provide a default case with assert(0) or
5662 @cindex @code{longjmp} warnings
5663 This option also warns when a non-volatile automatic variable might be
5664 changed by a call to @code{longjmp}.
5665 The compiler sees only the calls to @code{setjmp}. It cannot know
5666 where @code{longjmp} will be called; in fact, a signal handler could
5667 call it at any point in the code. As a result, you may get a warning
5668 even when there is in fact no problem because @code{longjmp} cannot
5669 in fact be called at the place that would cause a problem.
5671 Some spurious warnings can be avoided if you declare all the functions
5672 you use that never return as @code{noreturn}. @xref{Function
5675 This warning is enabled by @option{-Wall} or @option{-Wextra}.
5677 @item -Wunknown-pragmas
5678 @opindex Wunknown-pragmas
5679 @opindex Wno-unknown-pragmas
5680 @cindex warning for unknown pragmas
5681 @cindex unknown pragmas, warning
5682 @cindex pragmas, warning of unknown
5683 Warn when a @code{#pragma} directive is encountered that is not understood by
5684 GCC@. If this command-line option is used, warnings are even issued
5685 for unknown pragmas in system header files. This is not the case if
5686 the warnings are only enabled by the @option{-Wall} command-line option.
5689 @opindex Wno-pragmas
5691 Do not warn about misuses of pragmas, such as incorrect parameters,
5692 invalid syntax, or conflicts between pragmas. See also
5693 @option{-Wunknown-pragmas}.
5695 @item -Wno-prio-ctor-dtor
5696 @opindex Wno-prio-ctor-dtor
5697 @opindex Wprio-ctor-dtor
5698 Do not warn if a priority from 0 to 100 is used for constructor or destructor.
5699 The use of constructor and destructor attributes allow you to assign a
5700 priority to the constructor/destructor to control its order of execution
5701 before @code{main} is called or after it returns. The priority values must be
5702 greater than 100 as the compiler reserves priority values between 0--100 for
5705 @item -Wstrict-aliasing
5706 @opindex Wstrict-aliasing
5707 @opindex Wno-strict-aliasing
5708 This option is only active when @option{-fstrict-aliasing} is active.
5709 It warns about code that might break the strict aliasing rules that the
5710 compiler is using for optimization. The warning does not catch all
5711 cases, but does attempt to catch the more common pitfalls. It is
5712 included in @option{-Wall}.
5713 It is equivalent to @option{-Wstrict-aliasing=3}
5715 @item -Wstrict-aliasing=n
5716 @opindex Wstrict-aliasing=n
5717 This option is only active when @option{-fstrict-aliasing} is active.
5718 It warns about code that might break the strict aliasing rules that the
5719 compiler is using for optimization.
5720 Higher levels correspond to higher accuracy (fewer false positives).
5721 Higher levels also correspond to more effort, similar to the way @option{-O}
5723 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
5725 Level 1: Most aggressive, quick, least accurate.
5726 Possibly useful when higher levels
5727 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
5728 false negatives. However, it has many false positives.
5729 Warns for all pointer conversions between possibly incompatible types,
5730 even if never dereferenced. Runs in the front end only.
5732 Level 2: Aggressive, quick, not too precise.
5733 May still have many false positives (not as many as level 1 though),
5734 and few false negatives (but possibly more than level 1).
5735 Unlike level 1, it only warns when an address is taken. Warns about
5736 incomplete types. Runs in the front end only.
5738 Level 3 (default for @option{-Wstrict-aliasing}):
5739 Should have very few false positives and few false
5740 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
5741 Takes care of the common pun+dereference pattern in the front end:
5742 @code{*(int*)&some_float}.
5743 If optimization is enabled, it also runs in the back end, where it deals
5744 with multiple statement cases using flow-sensitive points-to information.
5745 Only warns when the converted pointer is dereferenced.
5746 Does not warn about incomplete types.
5748 @item -Wstrict-overflow
5749 @itemx -Wstrict-overflow=@var{n}
5750 @opindex Wstrict-overflow
5751 @opindex Wno-strict-overflow
5752 This option is only active when signed overflow is undefined.
5753 It warns about cases where the compiler optimizes based on the
5754 assumption that signed overflow does not occur. Note that it does not
5755 warn about all cases where the code might overflow: it only warns
5756 about cases where the compiler implements some optimization. Thus
5757 this warning depends on the optimization level.
5759 An optimization that assumes that signed overflow does not occur is
5760 perfectly safe if the values of the variables involved are such that
5761 overflow never does, in fact, occur. Therefore this warning can
5762 easily give a false positive: a warning about code that is not
5763 actually a problem. To help focus on important issues, several
5764 warning levels are defined. No warnings are issued for the use of
5765 undefined signed overflow when estimating how many iterations a loop
5766 requires, in particular when determining whether a loop will be
5770 @item -Wstrict-overflow=1
5771 Warn about cases that are both questionable and easy to avoid. For
5772 example the compiler simplifies
5773 @code{x + 1 > x} to @code{1}. This level of
5774 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
5775 are not, and must be explicitly requested.
5777 @item -Wstrict-overflow=2
5778 Also warn about other cases where a comparison is simplified to a
5779 constant. For example: @code{abs (x) >= 0}. This can only be
5780 simplified when signed integer overflow is undefined, because
5781 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
5782 zero. @option{-Wstrict-overflow} (with no level) is the same as
5783 @option{-Wstrict-overflow=2}.
5785 @item -Wstrict-overflow=3
5786 Also warn about other cases where a comparison is simplified. For
5787 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
5789 @item -Wstrict-overflow=4
5790 Also warn about other simplifications not covered by the above cases.
5791 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
5793 @item -Wstrict-overflow=5
5794 Also warn about cases where the compiler reduces the magnitude of a
5795 constant involved in a comparison. For example: @code{x + 2 > y} is
5796 simplified to @code{x + 1 >= y}. This is reported only at the
5797 highest warning level because this simplification applies to many
5798 comparisons, so this warning level gives a very large number of
5802 @item -Wstringop-overflow
5803 @itemx -Wstringop-overflow=@var{type}
5804 @opindex Wstringop-overflow
5805 @opindex Wno-stringop-overflow
5806 Warn for calls to string manipulation functions such as @code{memcpy} and
5807 @code{strcpy} that are determined to overflow the destination buffer. The
5808 optional argument is one greater than the type of Object Size Checking to
5809 perform to determine the size of the destination. @xref{Object Size Checking}.
5810 The argument is meaningful only for functions that operate on character arrays
5811 but not for raw memory functions like @code{memcpy} which always make use
5812 of Object Size type-0. The option also warns for calls that specify a size
5813 in excess of the largest possible object or at most @code{SIZE_MAX / 2} bytes.
5814 The option produces the best results with optimization enabled but can detect
5815 a small subset of simple buffer overflows even without optimization in
5816 calls to the GCC built-in functions like @code{__builtin_memcpy} that
5817 correspond to the standard functions. In any case, the option warns about
5818 just a subset of buffer overflows detected by the corresponding overflow
5819 checking built-ins. For example, the option will issue a warning for
5820 the @code{strcpy} call below because it copies at least 5 characters
5821 (the string @code{"blue"} including the terminating NUL) into the buffer
5825 enum Color @{ blue, purple, yellow @};
5826 const char* f (enum Color clr)
5828 static char buf [4];
5832 case blue: str = "blue"; break;
5833 case purple: str = "purple"; break;
5834 case yellow: str = "yellow"; break;
5837 return strcpy (buf, str); // warning here
5841 Option @option{-Wstringop-overflow=2} is enabled by default.
5844 @item -Wstringop-overflow
5845 @itemx -Wstringop-overflow=1
5846 @opindex Wstringop-overflow
5847 @opindex Wno-stringop-overflow
5848 The @option{-Wstringop-overflow=1} option uses type-zero Object Size Checking
5849 to determine the sizes of destination objects. This is the default setting
5850 of the option. At this setting the option will not warn for writes past
5851 the end of subobjects of larger objects accessed by pointers unless the
5852 size of the largest surrounding object is known. When the destination may
5853 be one of several objects it is assumed to be the largest one of them. On
5854 Linux systems, when optimization is enabled at this setting the option warns
5855 for the same code as when the @code{_FORTIFY_SOURCE} macro is defined to
5858 @item -Wstringop-overflow=2
5859 The @option{-Wstringop-overflow=2} option uses type-one Object Size Checking
5860 to determine the sizes of destination objects. At this setting the option
5861 will warn about overflows when writing to members of the largest complete
5862 objects whose exact size is known. It will, however, not warn for excessive
5863 writes to the same members of unknown objects referenced by pointers since
5864 they may point to arrays containing unknown numbers of elements.
5866 @item -Wstringop-overflow=3
5867 The @option{-Wstringop-overflow=3} option uses type-two Object Size Checking
5868 to determine the sizes of destination objects. At this setting the option
5869 warns about overflowing the smallest object or data member. This is the
5870 most restrictive setting of the option that may result in warnings for safe
5873 @item -Wstringop-overflow=4
5874 The @option{-Wstringop-overflow=4} option uses type-three Object Size Checking
5875 to determine the sizes of destination objects. At this setting the option
5876 will warn about overflowing any data members, and when the destination is
5877 one of several objects it uses the size of the largest of them to decide
5878 whether to issue a warning. Similarly to @option{-Wstringop-overflow=3} this
5879 setting of the option may result in warnings for benign code.
5882 @item -Wstringop-truncation
5883 @opindex Wstringop-truncation
5884 @opindex Wno-stringop-truncation
5885 Warn for calls to bounded string manipulation functions such as @code{strncat},
5886 @code{strncpy}, and @code{stpncpy} that may either truncate the copied string
5887 or leave the destination unchanged.
5889 In the following example, the call to @code{strncat} specifies a bound that
5890 is less than the length of the source string. As a result, the copy of
5891 the source will be truncated and so the call is diagnosed. To avoid the
5892 warning use @code{bufsize - strlen (buf) - 1)} as the bound.
5895 void append (char *buf, size_t bufsize)
5897 strncat (buf, ".txt", 3);
5901 As another example, the following call to @code{strncpy} results in copying
5902 to @code{d} just the characters preceding the terminating NUL, without
5903 appending the NUL to the end. Assuming the result of @code{strncpy} is
5904 necessarily a NUL-terminated string is a common mistake, and so the call
5905 is diagnosed. To avoid the warning when the result is not expected to be
5906 NUL-terminated, call @code{memcpy} instead.
5909 void copy (char *d, const char *s)
5911 strncpy (d, s, strlen (s));
5915 In the following example, the call to @code{strncpy} specifies the size
5916 of the destination buffer as the bound. If the length of the source
5917 string is equal to or greater than this size the result of the copy will
5918 not be NUL-terminated. Therefore, the call is also diagnosed. To avoid
5919 the warning, specify @code{sizeof buf - 1} as the bound and set the last
5920 element of the buffer to @code{NUL}.
5923 void copy (const char *s)
5926 strncpy (buf, s, sizeof buf);
5931 In situations where a character array is intended to store a sequence
5932 of bytes with no terminating @code{NUL} such an array may be annotated
5933 with attribute @code{nonstring} to avoid this warning. Such arrays,
5934 however, are not suitable arguments to functions that expect
5935 @code{NUL}-terminated strings. To help detect accidental misuses of
5936 such arrays GCC issues warnings unless it can prove that the use is
5937 safe. @xref{Common Variable Attributes}.
5939 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{|}cold@r{|}malloc@r{]}
5940 @opindex Wsuggest-attribute=
5941 @opindex Wno-suggest-attribute=
5942 Warn for cases where adding an attribute may be beneficial. The
5943 attributes currently supported are listed below.
5946 @item -Wsuggest-attribute=pure
5947 @itemx -Wsuggest-attribute=const
5948 @itemx -Wsuggest-attribute=noreturn
5949 @itemx -Wmissing-noreturn
5950 @itemx -Wsuggest-attribute=malloc
5951 @opindex Wsuggest-attribute=pure
5952 @opindex Wno-suggest-attribute=pure
5953 @opindex Wsuggest-attribute=const
5954 @opindex Wno-suggest-attribute=const
5955 @opindex Wsuggest-attribute=noreturn
5956 @opindex Wno-suggest-attribute=noreturn
5957 @opindex Wmissing-noreturn
5958 @opindex Wno-missing-noreturn
5959 @opindex Wsuggest-attribute=malloc
5960 @opindex Wno-suggest-attribute=malloc
5962 Warn about functions that might be candidates for attributes
5963 @code{pure}, @code{const} or @code{noreturn} or @code{malloc}. The compiler
5964 only warns for functions visible in other compilation units or (in the case of
5965 @code{pure} and @code{const}) if it cannot prove that the function returns
5966 normally. A function returns normally if it doesn't contain an infinite loop or
5967 return abnormally by throwing, calling @code{abort} or trapping. This analysis
5968 requires option @option{-fipa-pure-const}, which is enabled by default at
5969 @option{-O} and higher. Higher optimization levels improve the accuracy
5972 @item -Wsuggest-attribute=format
5973 @itemx -Wmissing-format-attribute
5974 @opindex Wsuggest-attribute=format
5975 @opindex Wmissing-format-attribute
5976 @opindex Wno-suggest-attribute=format
5977 @opindex Wno-missing-format-attribute
5981 Warn about function pointers that might be candidates for @code{format}
5982 attributes. Note these are only possible candidates, not absolute ones.
5983 GCC guesses that function pointers with @code{format} attributes that
5984 are used in assignment, initialization, parameter passing or return
5985 statements should have a corresponding @code{format} attribute in the
5986 resulting type. I.e.@: the left-hand side of the assignment or
5987 initialization, the type of the parameter variable, or the return type
5988 of the containing function respectively should also have a @code{format}
5989 attribute to avoid the warning.
5991 GCC also warns about function definitions that might be
5992 candidates for @code{format} attributes. Again, these are only
5993 possible candidates. GCC guesses that @code{format} attributes
5994 might be appropriate for any function that calls a function like
5995 @code{vprintf} or @code{vscanf}, but this might not always be the
5996 case, and some functions for which @code{format} attributes are
5997 appropriate may not be detected.
5999 @item -Wsuggest-attribute=cold
6000 @opindex Wsuggest-attribute=cold
6001 @opindex Wno-suggest-attribute=cold
6003 Warn about functions that might be candidates for @code{cold} attribute. This
6004 is based on static detection and generally will only warn about functions which
6005 always leads to a call to another @code{cold} function such as wrappers of
6006 C++ @code{throw} or fatal error reporting functions leading to @code{abort}.
6009 @item -Wsuggest-final-types
6010 @opindex Wno-suggest-final-types
6011 @opindex Wsuggest-final-types
6012 Warn about types with virtual methods where code quality would be improved
6013 if the type were declared with the C++11 @code{final} specifier,
6015 declared in an anonymous namespace. This allows GCC to more aggressively
6016 devirtualize the polymorphic calls. This warning is more effective with
6017 link-time optimization,
6018 where the information about the class hierarchy graph is
6021 @item -Wsuggest-final-methods
6022 @opindex Wno-suggest-final-methods
6023 @opindex Wsuggest-final-methods
6024 Warn about virtual methods where code quality would be improved if the method
6025 were declared with the C++11 @code{final} specifier,
6026 or, if possible, its type were
6027 declared in an anonymous namespace or with the @code{final} specifier.
6029 more effective with link-time optimization, where the information about the
6030 class hierarchy graph is more complete. It is recommended to first consider
6031 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
6034 @item -Wsuggest-override
6035 Warn about overriding virtual functions that are not marked with the override
6039 @opindex Wno-alloc-zero
6040 @opindex Walloc-zero
6041 Warn about calls to allocation functions decorated with attribute
6042 @code{alloc_size} that specify zero bytes, including those to the built-in
6043 forms of the functions @code{aligned_alloc}, @code{alloca}, @code{calloc},
6044 @code{malloc}, and @code{realloc}. Because the behavior of these functions
6045 when called with a zero size differs among implementations (and in the case
6046 of @code{realloc} has been deprecated) relying on it may result in subtle
6047 portability bugs and should be avoided.
6049 @item -Walloc-size-larger-than=@var{byte-size}
6050 @opindex Walloc-size-larger-than=
6051 @opindex Wno-alloc-size-larger-than
6052 Warn about calls to functions decorated with attribute @code{alloc_size}
6053 that attempt to allocate objects larger than the specified number of bytes,
6054 or where the result of the size computation in an integer type with infinite
6055 precision would exceed the value of @samp{PTRDIFF_MAX} on the target.
6056 @option{-Walloc-size-larger-than=}@samp{PTRDIFF_MAX} is enabled by default.
6057 Warnings controlled by the option can be disabled either by specifying
6058 @var{byte-size} of @samp{SIZE_MAX} or more or by
6059 @option{-Wno-alloc-size-larger-than}.
6060 @xref{Function Attributes}.
6062 @item -Wno-alloc-size-larger-than
6063 @opindex Wno-alloc-size-larger-than
6064 Disable @option{-Walloc-size-larger-than=} warnings. The option is
6065 equivalent to @option{-Walloc-size-larger-than=}@samp{SIZE_MAX} or
6071 This option warns on all uses of @code{alloca} in the source.
6073 @item -Walloca-larger-than=@var{byte-size}
6074 @opindex Walloca-larger-than=
6075 @opindex Wno-alloca-larger-than
6076 This option warns on calls to @code{alloca} with an integer argument whose
6077 value is either zero, or that is not bounded by a controlling predicate
6078 that limits its value to at most @var{byte-size}. It also warns for calls
6079 to @code{alloca} where the bound value is unknown. Arguments of non-integer
6080 types are considered unbounded even if they appear to be constrained to
6083 For example, a bounded case of @code{alloca} could be:
6086 void func (size_t n)
6097 In the above example, passing @code{-Walloca-larger-than=1000} would not
6098 issue a warning because the call to @code{alloca} is known to be at most
6099 1000 bytes. However, if @code{-Walloca-larger-than=500} were passed,
6100 the compiler would emit a warning.
6102 Unbounded uses, on the other hand, are uses of @code{alloca} with no
6103 controlling predicate constraining its integer argument. For example:
6108 void *p = alloca (n);
6113 If @code{-Walloca-larger-than=500} were passed, the above would trigger
6114 a warning, but this time because of the lack of bounds checking.
6116 Note, that even seemingly correct code involving signed integers could
6120 void func (signed int n)
6130 In the above example, @var{n} could be negative, causing a larger than
6131 expected argument to be implicitly cast into the @code{alloca} call.
6133 This option also warns when @code{alloca} is used in a loop.
6135 @option{-Walloca-larger-than=}@samp{PTRDIFF_MAX} is enabled by default
6136 but is usually only effective when @option{-ftree-vrp} is active (default
6137 for @option{-O2} and above).
6139 See also @option{-Wvla-larger-than=}@samp{byte-size}.
6141 @item -Wno-alloca-larger-than
6142 @opindex Wno-alloca-larger-than
6143 Disable @option{-Walloca-larger-than=} warnings. The option is
6144 equivalent to @option{-Walloca-larger-than=}@samp{SIZE_MAX} or larger.
6146 @item -Warray-bounds
6147 @itemx -Warray-bounds=@var{n}
6148 @opindex Wno-array-bounds
6149 @opindex Warray-bounds
6150 This option is only active when @option{-ftree-vrp} is active
6151 (default for @option{-O2} and above). It warns about subscripts to arrays
6152 that are always out of bounds. This warning is enabled by @option{-Wall}.
6155 @item -Warray-bounds=1
6156 This is the warning level of @option{-Warray-bounds} and is enabled
6157 by @option{-Wall}; higher levels are not, and must be explicitly requested.
6159 @item -Warray-bounds=2
6160 This warning level also warns about out of bounds access for
6161 arrays at the end of a struct and for arrays accessed through
6162 pointers. This warning level may give a larger number of
6163 false positives and is deactivated by default.
6166 @item -Wattribute-alias=@var{n}
6167 @itemx -Wno-attribute-alias
6168 @opindex Wattribute-alias
6169 @opindex Wno-attribute-alias
6170 Warn about declarations using the @code{alias} and similar attributes whose
6171 target is incompatible with the type of the alias.
6172 @xref{Function Attributes,,Declaring Attributes of Functions}.
6175 @item -Wattribute-alias=1
6176 The default warning level of the @option{-Wattribute-alias} option diagnoses
6177 incompatibilities between the type of the alias declaration and that of its
6178 target. Such incompatibilities are typically indicative of bugs.
6180 @item -Wattribute-alias=2
6182 At this level @option{-Wattribute-alias} also diagnoses cases where
6183 the attributes of the alias declaration are more restrictive than the
6184 attributes applied to its target. These mismatches can potentially
6185 result in incorrect code generation. In other cases they may be
6186 benign and could be resolved simply by adding the missing attribute to
6187 the target. For comparison, see the @option{-Wmissing-attributes}
6188 option, which controls diagnostics when the alias declaration is less
6189 restrictive than the target, rather than more restrictive.
6191 Attributes considered include @code{alloc_align}, @code{alloc_size},
6192 @code{cold}, @code{const}, @code{hot}, @code{leaf}, @code{malloc},
6193 @code{nonnull}, @code{noreturn}, @code{nothrow}, @code{pure},
6194 @code{returns_nonnull}, and @code{returns_twice}.
6197 @option{-Wattribute-alias} is equivalent to @option{-Wattribute-alias=1}.
6198 This is the default. You can disable these warnings with either
6199 @option{-Wno-attribute-alias} or @option{-Wattribute-alias=0}.
6201 @item -Wbool-compare
6202 @opindex Wno-bool-compare
6203 @opindex Wbool-compare
6204 Warn about boolean expression compared with an integer value different from
6205 @code{true}/@code{false}. For instance, the following comparison is
6210 if ((n > 1) == 2) @{ @dots{} @}
6212 This warning is enabled by @option{-Wall}.
6214 @item -Wbool-operation
6215 @opindex Wno-bool-operation
6216 @opindex Wbool-operation
6217 Warn about suspicious operations on expressions of a boolean type. For
6218 instance, bitwise negation of a boolean is very likely a bug in the program.
6219 For C, this warning also warns about incrementing or decrementing a boolean,
6220 which rarely makes sense. (In C++, decrementing a boolean is always invalid.
6221 Incrementing a boolean is invalid in C++17, and deprecated otherwise.)
6223 This warning is enabled by @option{-Wall}.
6225 @item -Wduplicated-branches
6226 @opindex Wno-duplicated-branches
6227 @opindex Wduplicated-branches
6228 Warn when an if-else has identical branches. This warning detects cases like
6235 It doesn't warn when both branches contain just a null statement. This warning
6236 also warn for conditional operators:
6238 int i = x ? *p : *p;
6241 @item -Wduplicated-cond
6242 @opindex Wno-duplicated-cond
6243 @opindex Wduplicated-cond
6244 Warn about duplicated conditions in an if-else-if chain. For instance,
6245 warn for the following code:
6247 if (p->q != NULL) @{ @dots{} @}
6248 else if (p->q != NULL) @{ @dots{} @}
6251 @item -Wframe-address
6252 @opindex Wno-frame-address
6253 @opindex Wframe-address
6254 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
6255 is called with an argument greater than 0. Such calls may return indeterminate
6256 values or crash the program. The warning is included in @option{-Wall}.
6258 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
6259 @opindex Wno-discarded-qualifiers
6260 @opindex Wdiscarded-qualifiers
6261 Do not warn if type qualifiers on pointers are being discarded.
6262 Typically, the compiler warns if a @code{const char *} variable is
6263 passed to a function that takes a @code{char *} parameter. This option
6264 can be used to suppress such a warning.
6266 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
6267 @opindex Wno-discarded-array-qualifiers
6268 @opindex Wdiscarded-array-qualifiers
6269 Do not warn if type qualifiers on arrays which are pointer targets
6270 are being discarded. Typically, the compiler warns if a
6271 @code{const int (*)[]} variable is passed to a function that
6272 takes a @code{int (*)[]} parameter. This option can be used to
6273 suppress such a warning.
6275 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
6276 @opindex Wno-incompatible-pointer-types
6277 @opindex Wincompatible-pointer-types
6278 Do not warn when there is a conversion between pointers that have incompatible
6279 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
6280 which warns for pointer argument passing or assignment with different
6283 @item -Wno-int-conversion @r{(C and Objective-C only)}
6284 @opindex Wno-int-conversion
6285 @opindex Wint-conversion
6286 Do not warn about incompatible integer to pointer and pointer to integer
6287 conversions. This warning is about implicit conversions; for explicit
6288 conversions the warnings @option{-Wno-int-to-pointer-cast} and
6289 @option{-Wno-pointer-to-int-cast} may be used.
6291 @item -Wno-div-by-zero
6292 @opindex Wno-div-by-zero
6293 @opindex Wdiv-by-zero
6294 Do not warn about compile-time integer division by zero. Floating-point
6295 division by zero is not warned about, as it can be a legitimate way of
6296 obtaining infinities and NaNs.
6298 @item -Wsystem-headers
6299 @opindex Wsystem-headers
6300 @opindex Wno-system-headers
6301 @cindex warnings from system headers
6302 @cindex system headers, warnings from
6303 Print warning messages for constructs found in system header files.
6304 Warnings from system headers are normally suppressed, on the assumption
6305 that they usually do not indicate real problems and would only make the
6306 compiler output harder to read. Using this command-line option tells
6307 GCC to emit warnings from system headers as if they occurred in user
6308 code. However, note that using @option{-Wall} in conjunction with this
6309 option does @emph{not} warn about unknown pragmas in system
6310 headers---for that, @option{-Wunknown-pragmas} must also be used.
6312 @item -Wtautological-compare
6313 @opindex Wtautological-compare
6314 @opindex Wno-tautological-compare
6315 Warn if a self-comparison always evaluates to true or false. This
6316 warning detects various mistakes such as:
6320 if (i > i) @{ @dots{} @}
6323 This warning also warns about bitwise comparisons that always evaluate
6324 to true or false, for instance:
6326 if ((a & 16) == 10) @{ @dots{} @}
6328 will always be false.
6330 This warning is enabled by @option{-Wall}.
6333 @opindex Wtrampolines
6334 @opindex Wno-trampolines
6335 Warn about trampolines generated for pointers to nested functions.
6336 A trampoline is a small piece of data or code that is created at run
6337 time on the stack when the address of a nested function is taken, and is
6338 used to call the nested function indirectly. For some targets, it is
6339 made up of data only and thus requires no special treatment. But, for
6340 most targets, it is made up of code and thus requires the stack to be
6341 made executable in order for the program to work properly.
6344 @opindex Wfloat-equal
6345 @opindex Wno-float-equal
6346 Warn if floating-point values are used in equality comparisons.
6348 The idea behind this is that sometimes it is convenient (for the
6349 programmer) to consider floating-point values as approximations to
6350 infinitely precise real numbers. If you are doing this, then you need
6351 to compute (by analyzing the code, or in some other way) the maximum or
6352 likely maximum error that the computation introduces, and allow for it
6353 when performing comparisons (and when producing output, but that's a
6354 different problem). In particular, instead of testing for equality, you
6355 should check to see whether the two values have ranges that overlap; and
6356 this is done with the relational operators, so equality comparisons are
6359 @item -Wtraditional @r{(C and Objective-C only)}
6360 @opindex Wtraditional
6361 @opindex Wno-traditional
6362 Warn about certain constructs that behave differently in traditional and
6363 ISO C@. Also warn about ISO C constructs that have no traditional C
6364 equivalent, and/or problematic constructs that should be avoided.
6368 Macro parameters that appear within string literals in the macro body.
6369 In traditional C macro replacement takes place within string literals,
6370 but in ISO C it does not.
6373 In traditional C, some preprocessor directives did not exist.
6374 Traditional preprocessors only considered a line to be a directive
6375 if the @samp{#} appeared in column 1 on the line. Therefore
6376 @option{-Wtraditional} warns about directives that traditional C
6377 understands but ignores because the @samp{#} does not appear as the
6378 first character on the line. It also suggests you hide directives like
6379 @code{#pragma} not understood by traditional C by indenting them. Some
6380 traditional implementations do not recognize @code{#elif}, so this option
6381 suggests avoiding it altogether.
6384 A function-like macro that appears without arguments.
6387 The unary plus operator.
6390 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
6391 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
6392 constants.) Note, these suffixes appear in macros defined in the system
6393 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
6394 Use of these macros in user code might normally lead to spurious
6395 warnings, however GCC's integrated preprocessor has enough context to
6396 avoid warning in these cases.
6399 A function declared external in one block and then used after the end of
6403 A @code{switch} statement has an operand of type @code{long}.
6406 A non-@code{static} function declaration follows a @code{static} one.
6407 This construct is not accepted by some traditional C compilers.
6410 The ISO type of an integer constant has a different width or
6411 signedness from its traditional type. This warning is only issued if
6412 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
6413 typically represent bit patterns, are not warned about.
6416 Usage of ISO string concatenation is detected.
6419 Initialization of automatic aggregates.
6422 Identifier conflicts with labels. Traditional C lacks a separate
6423 namespace for labels.
6426 Initialization of unions. If the initializer is zero, the warning is
6427 omitted. This is done under the assumption that the zero initializer in
6428 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
6429 initializer warnings and relies on default initialization to zero in the
6433 Conversions by prototypes between fixed/floating-point values and vice
6434 versa. The absence of these prototypes when compiling with traditional
6435 C causes serious problems. This is a subset of the possible
6436 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
6439 Use of ISO C style function definitions. This warning intentionally is
6440 @emph{not} issued for prototype declarations or variadic functions
6441 because these ISO C features appear in your code when using
6442 libiberty's traditional C compatibility macros, @code{PARAMS} and
6443 @code{VPARAMS}. This warning is also bypassed for nested functions
6444 because that feature is already a GCC extension and thus not relevant to
6445 traditional C compatibility.
6448 @item -Wtraditional-conversion @r{(C and Objective-C only)}
6449 @opindex Wtraditional-conversion
6450 @opindex Wno-traditional-conversion
6451 Warn if a prototype causes a type conversion that is different from what
6452 would happen to the same argument in the absence of a prototype. This
6453 includes conversions of fixed point to floating and vice versa, and
6454 conversions changing the width or signedness of a fixed-point argument
6455 except when the same as the default promotion.
6457 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
6458 @opindex Wdeclaration-after-statement
6459 @opindex Wno-declaration-after-statement
6460 Warn when a declaration is found after a statement in a block. This
6461 construct, known from C++, was introduced with ISO C99 and is by default
6462 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
6467 Warn whenever a local variable or type declaration shadows another
6468 variable, parameter, type, class member (in C++), or instance variable
6469 (in Objective-C) or whenever a built-in function is shadowed. Note
6470 that in C++, the compiler warns if a local variable shadows an
6471 explicit typedef, but not if it shadows a struct/class/enum.
6472 Same as @option{-Wshadow=global}.
6474 @item -Wno-shadow-ivar @r{(Objective-C only)}
6475 @opindex Wno-shadow-ivar
6476 @opindex Wshadow-ivar
6477 Do not warn whenever a local variable shadows an instance variable in an
6480 @item -Wshadow=global
6481 @opindex Wshadow=local
6482 The default for @option{-Wshadow}. Warns for any (global) shadowing.
6484 @item -Wshadow=local
6485 @opindex Wshadow=local
6486 Warn when a local variable shadows another local variable or parameter.
6487 This warning is enabled by @option{-Wshadow=global}.
6489 @item -Wshadow=compatible-local
6490 @opindex Wshadow=compatible-local
6491 Warn when a local variable shadows another local variable or parameter
6492 whose type is compatible with that of the shadowing variable. In C++,
6493 type compatibility here means the type of the shadowing variable can be
6494 converted to that of the shadowed variable. The creation of this flag
6495 (in addition to @option{-Wshadow=local}) is based on the idea that when
6496 a local variable shadows another one of incompatible type, it is most
6497 likely intentional, not a bug or typo, as shown in the following example:
6501 for (SomeIterator i = SomeObj.begin(); i != SomeObj.end(); ++i)
6503 for (int i = 0; i < N; ++i)
6512 Since the two variable @code{i} in the example above have incompatible types,
6513 enabling only @option{-Wshadow=compatible-local} will not emit a warning.
6514 Because their types are incompatible, if a programmer accidentally uses one
6515 in place of the other, type checking will catch that and emit an error or
6516 warning. So not warning (about shadowing) in this case will not lead to
6517 undetected bugs. Use of this flag instead of @option{-Wshadow=local} can
6518 possibly reduce the number of warnings triggered by intentional shadowing.
6520 This warning is enabled by @option{-Wshadow=local}.
6522 @item -Wlarger-than=@var{byte-size}
6523 @opindex Wlarger-than=
6524 @opindex Wlarger-than-@var{byte-size}
6525 Warn whenever an object is defined whose size exceeds @var{byte-size}.
6526 @option{-Wlarger-than=}@samp{PTRDIFF_MAX} is enabled by default.
6527 Warnings controlled by the option can be disabled either by specifying
6528 @var{byte-size} of @samp{SIZE_MAX} or more or by
6529 @option{-Wno-larger-than}.
6531 @item -Wno-larger-than
6532 @opindex Wno-larger-than
6533 Disable @option{-Wlarger-than=} warnings. The option is equivalent
6534 to @option{-Wlarger-than=}@samp{SIZE_MAX} or larger.
6536 @item -Wframe-larger-than=@var{byte-size}
6537 @opindex Wframe-larger-than=
6538 @opindex Wno-frame-larger-than
6539 Warn if the size of a function frame exceeds @var{byte-size}.
6540 The computation done to determine the stack frame size is approximate
6541 and not conservative.
6542 The actual requirements may be somewhat greater than @var{byte-size}
6543 even if you do not get a warning. In addition, any space allocated
6544 via @code{alloca}, variable-length arrays, or related constructs
6545 is not included by the compiler when determining
6546 whether or not to issue a warning.
6547 @option{-Wframe-larger-than=}@samp{PTRDIFF_MAX} is enabled by default.
6548 Warnings controlled by the option can be disabled either by specifying
6549 @var{byte-size} of @samp{SIZE_MAX} or more or by
6550 @option{-Wno-frame-larger-than}.
6552 @item -Wno-frame-larger-than
6553 @opindex Wno-frame-larger-than
6554 Disable @option{-Wframe-larger-than=} warnings. The option is equivalent
6555 to @option{-Wframe-larger-than=}@samp{SIZE_MAX} or larger.
6557 @item -Wno-free-nonheap-object
6558 @opindex Wno-free-nonheap-object
6559 @opindex Wfree-nonheap-object
6560 Do not warn when attempting to free an object that was not allocated
6563 @item -Wstack-usage=@var{byte-size}
6564 @opindex Wstack-usage
6565 @opindex Wno-stack-usage
6566 Warn if the stack usage of a function might exceed @var{byte-size}.
6567 The computation done to determine the stack usage is conservative.
6568 Any space allocated via @code{alloca}, variable-length arrays, or related
6569 constructs is included by the compiler when determining whether or not to
6572 The message is in keeping with the output of @option{-fstack-usage}.
6576 If the stack usage is fully static but exceeds the specified amount, it's:
6579 warning: stack usage is 1120 bytes
6582 If the stack usage is (partly) dynamic but bounded, it's:
6585 warning: stack usage might be 1648 bytes
6588 If the stack usage is (partly) dynamic and not bounded, it's:
6591 warning: stack usage might be unbounded
6595 @option{-Wstack-usage=}@samp{PTRDIFF_MAX} is enabled by default.
6596 Warnings controlled by the option can be disabled either by specifying
6597 @var{byte-size} of @samp{SIZE_MAX} or more or by
6598 @option{-Wno-stack-usage}.
6600 @item -Wno-stack-usage
6601 @opindex Wno-stack-usage
6602 Disable @option{-Wstack-usage=} warnings. The option is equivalent
6603 to @option{-Wstack-usage=}@samp{SIZE_MAX} or larger.
6605 @item -Wunsafe-loop-optimizations
6606 @opindex Wunsafe-loop-optimizations
6607 @opindex Wno-unsafe-loop-optimizations
6608 Warn if the loop cannot be optimized because the compiler cannot
6609 assume anything on the bounds of the loop indices. With
6610 @option{-funsafe-loop-optimizations} warn if the compiler makes
6613 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
6614 @opindex Wno-pedantic-ms-format
6615 @opindex Wpedantic-ms-format
6616 When used in combination with @option{-Wformat}
6617 and @option{-pedantic} without GNU extensions, this option
6618 disables the warnings about non-ISO @code{printf} / @code{scanf} format
6619 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
6620 which depend on the MS runtime.
6623 @opindex Waligned-new
6624 @opindex Wno-aligned-new
6625 Warn about a new-expression of a type that requires greater alignment
6626 than the @code{alignof(std::max_align_t)} but uses an allocation
6627 function without an explicit alignment parameter. This option is
6628 enabled by @option{-Wall}.
6630 Normally this only warns about global allocation functions, but
6631 @option{-Waligned-new=all} also warns about class member allocation
6634 @item -Wplacement-new
6635 @itemx -Wplacement-new=@var{n}
6636 @opindex Wplacement-new
6637 @opindex Wno-placement-new
6638 Warn about placement new expressions with undefined behavior, such as
6639 constructing an object in a buffer that is smaller than the type of
6640 the object. For example, the placement new expression below is diagnosed
6641 because it attempts to construct an array of 64 integers in a buffer only
6647 This warning is enabled by default.
6650 @item -Wplacement-new=1
6651 This is the default warning level of @option{-Wplacement-new}. At this
6652 level the warning is not issued for some strictly undefined constructs that
6653 GCC allows as extensions for compatibility with legacy code. For example,
6654 the following @code{new} expression is not diagnosed at this level even
6655 though it has undefined behavior according to the C++ standard because
6656 it writes past the end of the one-element array.
6658 struct S @{ int n, a[1]; @};
6659 S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]);
6660 new (s->a)int [32]();
6663 @item -Wplacement-new=2
6664 At this level, in addition to diagnosing all the same constructs as at level
6665 1, a diagnostic is also issued for placement new expressions that construct
6666 an object in the last member of structure whose type is an array of a single
6667 element and whose size is less than the size of the object being constructed.
6668 While the previous example would be diagnosed, the following construct makes
6669 use of the flexible member array extension to avoid the warning at level 2.
6671 struct S @{ int n, a[]; @};
6672 S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]);
6673 new (s->a)int [32]();
6678 @item -Wpointer-arith
6679 @opindex Wpointer-arith
6680 @opindex Wno-pointer-arith
6681 Warn about anything that depends on the ``size of'' a function type or
6682 of @code{void}. GNU C assigns these types a size of 1, for
6683 convenience in calculations with @code{void *} pointers and pointers
6684 to functions. In C++, warn also when an arithmetic operation involves
6685 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
6687 @item -Wpointer-compare
6688 @opindex Wpointer-compare
6689 @opindex Wno-pointer-compare
6690 Warn if a pointer is compared with a zero character constant. This usually
6691 means that the pointer was meant to be dereferenced. For example:
6694 const char *p = foo ();
6699 Note that the code above is invalid in C++11.
6701 This warning is enabled by default.
6704 @opindex Wtype-limits
6705 @opindex Wno-type-limits
6706 Warn if a comparison is always true or always false due to the limited
6707 range of the data type, but do not warn for constant expressions. For
6708 example, warn if an unsigned variable is compared against zero with
6709 @code{<} or @code{>=}. This warning is also enabled by
6712 @item -Wabsolute-value @r{(C and Objective-C only)}
6713 @opindex Wabsolute-value
6714 @opindex Wno-absolute-value
6715 Warn for calls to standard functions that compute the absolute value
6716 of an argument when a more appropriate standard function is available.
6717 For example, calling @code{abs(3.14)} triggers the warning because the
6718 appropriate function to call to compute the absolute value of a double
6719 argument is @code{fabs}. The option also triggers warnings when the
6720 argument in a call to such a function has an unsigned type. This
6721 warning can be suppressed with an explicit type cast and it is also
6722 enabled by @option{-Wextra}.
6724 @include cppwarnopts.texi
6726 @item -Wbad-function-cast @r{(C and Objective-C only)}
6727 @opindex Wbad-function-cast
6728 @opindex Wno-bad-function-cast
6729 Warn when a function call is cast to a non-matching type.
6730 For example, warn if a call to a function returning an integer type
6731 is cast to a pointer type.
6733 @item -Wc90-c99-compat @r{(C and Objective-C only)}
6734 @opindex Wc90-c99-compat
6735 @opindex Wno-c90-c99-compat
6736 Warn about features not present in ISO C90, but present in ISO C99.
6737 For instance, warn about use of variable length arrays, @code{long long}
6738 type, @code{bool} type, compound literals, designated initializers, and so
6739 on. This option is independent of the standards mode. Warnings are disabled
6740 in the expression that follows @code{__extension__}.
6742 @item -Wc99-c11-compat @r{(C and Objective-C only)}
6743 @opindex Wc99-c11-compat
6744 @opindex Wno-c99-c11-compat
6745 Warn about features not present in ISO C99, but present in ISO C11.
6746 For instance, warn about use of anonymous structures and unions,
6747 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
6748 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
6749 and so on. This option is independent of the standards mode. Warnings are
6750 disabled in the expression that follows @code{__extension__}.
6752 @item -Wc++-compat @r{(C and Objective-C only)}
6753 @opindex Wc++-compat
6754 @opindex Wno-c++-compat
6755 Warn about ISO C constructs that are outside of the common subset of
6756 ISO C and ISO C++, e.g.@: request for implicit conversion from
6757 @code{void *} to a pointer to non-@code{void} type.
6759 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
6760 @opindex Wc++11-compat
6761 @opindex Wno-c++11-compat
6762 Warn about C++ constructs whose meaning differs between ISO C++ 1998
6763 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
6764 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
6765 enabled by @option{-Wall}.
6767 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
6768 @opindex Wc++14-compat
6769 @opindex Wno-c++14-compat
6770 Warn about C++ constructs whose meaning differs between ISO C++ 2011
6771 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
6773 @item -Wc++17-compat @r{(C++ and Objective-C++ only)}
6774 @opindex Wc++17-compat
6775 @opindex Wno-c++17-compat
6776 Warn about C++ constructs whose meaning differs between ISO C++ 2014
6777 and ISO C++ 2017. This warning is enabled by @option{-Wall}.
6779 @item -Wc++20-compat @r{(C++ and Objective-C++ only)}
6780 @opindex Wc++20-compat
6781 @opindex Wno-c++20-compat
6782 Warn about C++ constructs whose meaning differs between ISO C++ 2017
6783 and ISO C++ 2020. This warning is enabled by @option{-Wall}.
6787 @opindex Wno-cast-qual
6788 Warn whenever a pointer is cast so as to remove a type qualifier from
6789 the target type. For example, warn if a @code{const char *} is cast
6790 to an ordinary @code{char *}.
6792 Also warn when making a cast that introduces a type qualifier in an
6793 unsafe way. For example, casting @code{char **} to @code{const char **}
6794 is unsafe, as in this example:
6797 /* p is char ** value. */
6798 const char **q = (const char **) p;
6799 /* Assignment of readonly string to const char * is OK. */
6801 /* Now char** pointer points to read-only memory. */
6806 @opindex Wcast-align
6807 @opindex Wno-cast-align
6808 Warn whenever a pointer is cast such that the required alignment of the
6809 target is increased. For example, warn if a @code{char *} is cast to
6810 an @code{int *} on machines where integers can only be accessed at
6811 two- or four-byte boundaries.
6813 @item -Wcast-align=strict
6814 @opindex Wcast-align=strict
6815 Warn whenever a pointer is cast such that the required alignment of the
6816 target is increased. For example, warn if a @code{char *} is cast to
6817 an @code{int *} regardless of the target machine.
6819 @item -Wcast-function-type
6820 @opindex Wcast-function-type
6821 @opindex Wno-cast-function-type
6822 Warn when a function pointer is cast to an incompatible function pointer.
6823 In a cast involving function types with a variable argument list only
6824 the types of initial arguments that are provided are considered.
6825 Any parameter of pointer-type matches any other pointer-type. Any benign
6826 differences in integral types are ignored, like @code{int} vs.@: @code{long}
6827 on ILP32 targets. Likewise type qualifiers are ignored. The function
6828 type @code{void (*) (void)} is special and matches everything, which can
6829 be used to suppress this warning.
6830 In a cast involving pointer to member types this warning warns whenever
6831 the type cast is changing the pointer to member type.
6832 This warning is enabled by @option{-Wextra}.
6834 @item -Wwrite-strings
6835 @opindex Wwrite-strings
6836 @opindex Wno-write-strings
6837 When compiling C, give string constants the type @code{const
6838 char[@var{length}]} so that copying the address of one into a
6839 non-@code{const} @code{char *} pointer produces a warning. These
6840 warnings help you find at compile time code that can try to write
6841 into a string constant, but only if you have been very careful about
6842 using @code{const} in declarations and prototypes. Otherwise, it is
6843 just a nuisance. This is why we did not make @option{-Wall} request
6846 When compiling C++, warn about the deprecated conversion from string
6847 literals to @code{char *}. This warning is enabled by default for C++
6851 @itemx -Wcatch-value=@var{n} @r{(C++ and Objective-C++ only)}
6852 @opindex Wcatch-value
6853 @opindex Wno-catch-value
6854 Warn about catch handlers that do not catch via reference.
6855 With @option{-Wcatch-value=1} (or @option{-Wcatch-value} for short)
6856 warn about polymorphic class types that are caught by value.
6857 With @option{-Wcatch-value=2} warn about all class types that are caught
6858 by value. With @option{-Wcatch-value=3} warn about all types that are
6859 not caught by reference. @option{-Wcatch-value} is enabled by @option{-Wall}.
6863 @opindex Wno-clobbered
6864 Warn for variables that might be changed by @code{longjmp} or
6865 @code{vfork}. This warning is also enabled by @option{-Wextra}.
6867 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
6868 @opindex Wconditionally-supported
6869 @opindex Wno-conditionally-supported
6870 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
6873 @opindex Wconversion
6874 @opindex Wno-conversion
6875 Warn for implicit conversions that may alter a value. This includes
6876 conversions between real and integer, like @code{abs (x)} when
6877 @code{x} is @code{double}; conversions between signed and unsigned,
6878 like @code{unsigned ui = -1}; and conversions to smaller types, like
6879 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
6880 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
6881 changed by the conversion like in @code{abs (2.0)}. Warnings about
6882 conversions between signed and unsigned integers can be disabled by
6883 using @option{-Wno-sign-conversion}.
6885 For C++, also warn for confusing overload resolution for user-defined
6886 conversions; and conversions that never use a type conversion
6887 operator: conversions to @code{void}, the same type, a base class or a
6888 reference to them. Warnings about conversions between signed and
6889 unsigned integers are disabled by default in C++ unless
6890 @option{-Wsign-conversion} is explicitly enabled.
6892 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
6893 @opindex Wconversion-null
6894 @opindex Wno-conversion-null
6895 Do not warn for conversions between @code{NULL} and non-pointer
6896 types. @option{-Wconversion-null} is enabled by default.
6898 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
6899 @opindex Wzero-as-null-pointer-constant
6900 @opindex Wno-zero-as-null-pointer-constant
6901 Warn when a literal @samp{0} is used as null pointer constant. This can
6902 be useful to facilitate the conversion to @code{nullptr} in C++11.
6904 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
6905 @opindex Wsubobject-linkage
6906 @opindex Wno-subobject-linkage
6907 Warn if a class type has a base or a field whose type uses the anonymous
6908 namespace or depends on a type with no linkage. If a type A depends on
6909 a type B with no or internal linkage, defining it in multiple
6910 translation units would be an ODR violation because the meaning of B
6911 is different in each translation unit. If A only appears in a single
6912 translation unit, the best way to silence the warning is to give it
6913 internal linkage by putting it in an anonymous namespace as well. The
6914 compiler doesn't give this warning for types defined in the main .C
6915 file, as those are unlikely to have multiple definitions.
6916 @option{-Wsubobject-linkage} is enabled by default.
6918 @item -Wdangling-else
6919 @opindex Wdangling-else
6920 @opindex Wno-dangling-else
6921 Warn about constructions where there may be confusion to which
6922 @code{if} statement an @code{else} branch belongs. Here is an example of
6937 In C/C++, every @code{else} branch belongs to the innermost possible
6938 @code{if} statement, which in this example is @code{if (b)}. This is
6939 often not what the programmer expected, as illustrated in the above
6940 example by indentation the programmer chose. When there is the
6941 potential for this confusion, GCC issues a warning when this flag
6942 is specified. To eliminate the warning, add explicit braces around
6943 the innermost @code{if} statement so there is no way the @code{else}
6944 can belong to the enclosing @code{if}. The resulting code
6961 This warning is enabled by @option{-Wparentheses}.
6965 @opindex Wno-date-time
6966 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
6967 are encountered as they might prevent bit-wise-identical reproducible
6970 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
6971 @opindex Wdelete-incomplete
6972 @opindex Wno-delete-incomplete
6973 Warn when deleting a pointer to incomplete type, which may cause
6974 undefined behavior at runtime. This warning is enabled by default.
6976 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
6977 @opindex Wuseless-cast
6978 @opindex Wno-useless-cast
6979 Warn when an expression is casted to its own type.
6982 @opindex Wempty-body
6983 @opindex Wno-empty-body
6984 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
6985 while} statement. This warning is also enabled by @option{-Wextra}.
6987 @item -Wenum-compare
6988 @opindex Wenum-compare
6989 @opindex Wno-enum-compare
6990 Warn about a comparison between values of different enumerated types.
6991 In C++ enumerated type mismatches in conditional expressions are also
6992 diagnosed and the warning is enabled by default. In C this warning is
6993 enabled by @option{-Wall}.
6995 @item -Wenum-conversion @r{(C, Objective-C only)}
6996 @opindex Wenum-conversion
6997 @opindex Wno-enum-conversion
6998 Warn when a value of enumerated type is implicitly converted to a
6999 different enumerated type. This warning is enabled by @option{-Wextra}.
7001 @item -Wextra-semi @r{(C++, Objective-C++ only)}
7002 @opindex Wextra-semi
7003 @opindex Wno-extra-semi
7004 Warn about redundant semicolon after in-class function definition.
7006 @item -Wjump-misses-init @r{(C, Objective-C only)}
7007 @opindex Wjump-misses-init
7008 @opindex Wno-jump-misses-init
7009 Warn if a @code{goto} statement or a @code{switch} statement jumps
7010 forward across the initialization of a variable, or jumps backward to a
7011 label after the variable has been initialized. This only warns about
7012 variables that are initialized when they are declared. This warning is
7013 only supported for C and Objective-C; in C++ this sort of branch is an
7016 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
7017 can be disabled with the @option{-Wno-jump-misses-init} option.
7019 @item -Wsign-compare
7020 @opindex Wsign-compare
7021 @opindex Wno-sign-compare
7022 @cindex warning for comparison of signed and unsigned values
7023 @cindex comparison of signed and unsigned values, warning
7024 @cindex signed and unsigned values, comparison warning
7025 Warn when a comparison between signed and unsigned values could produce
7026 an incorrect result when the signed value is converted to unsigned.
7027 In C++, this warning is also enabled by @option{-Wall}. In C, it is
7028 also enabled by @option{-Wextra}.
7030 @item -Wsign-conversion
7031 @opindex Wsign-conversion
7032 @opindex Wno-sign-conversion
7033 Warn for implicit conversions that may change the sign of an integer
7034 value, like assigning a signed integer expression to an unsigned
7035 integer variable. An explicit cast silences the warning. In C, this
7036 option is enabled also by @option{-Wconversion}.
7038 @item -Wfloat-conversion
7039 @opindex Wfloat-conversion
7040 @opindex Wno-float-conversion
7041 Warn for implicit conversions that reduce the precision of a real value.
7042 This includes conversions from real to integer, and from higher precision
7043 real to lower precision real values. This option is also enabled by
7044 @option{-Wconversion}.
7046 @item -Wno-scalar-storage-order
7047 @opindex Wno-scalar-storage-order
7048 @opindex Wscalar-storage-order
7049 Do not warn on suspicious constructs involving reverse scalar storage order.
7051 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
7052 @opindex Wsized-deallocation
7053 @opindex Wno-sized-deallocation
7054 Warn about a definition of an unsized deallocation function
7056 void operator delete (void *) noexcept;
7057 void operator delete[] (void *) noexcept;
7059 without a definition of the corresponding sized deallocation function
7061 void operator delete (void *, std::size_t) noexcept;
7062 void operator delete[] (void *, std::size_t) noexcept;
7064 or vice versa. Enabled by @option{-Wextra} along with
7065 @option{-fsized-deallocation}.
7067 @item -Wsizeof-pointer-div
7068 @opindex Wsizeof-pointer-div
7069 @opindex Wno-sizeof-pointer-div
7070 Warn for suspicious divisions of two sizeof expressions that divide
7071 the pointer size by the element size, which is the usual way to compute
7072 the array size but won't work out correctly with pointers. This warning
7073 warns e.g.@: about @code{sizeof (ptr) / sizeof (ptr[0])} if @code{ptr} is
7074 not an array, but a pointer. This warning is enabled by @option{-Wall}.
7076 @item -Wsizeof-pointer-memaccess
7077 @opindex Wsizeof-pointer-memaccess
7078 @opindex Wno-sizeof-pointer-memaccess
7079 Warn for suspicious length parameters to certain string and memory built-in
7080 functions if the argument uses @code{sizeof}. This warning triggers for
7081 example for @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not
7082 an array, but a pointer, and suggests a possible fix, or about
7083 @code{memcpy (&foo, ptr, sizeof (&foo));}. @option{-Wsizeof-pointer-memaccess}
7084 also warns about calls to bounded string copy functions like @code{strncat}
7085 or @code{strncpy} that specify as the bound a @code{sizeof} expression of
7086 the source array. For example, in the following function the call to
7087 @code{strncat} specifies the size of the source string as the bound. That
7088 is almost certainly a mistake and so the call is diagnosed.
7090 void make_file (const char *name)
7092 char path[PATH_MAX];
7093 strncpy (path, name, sizeof path - 1);
7094 strncat (path, ".text", sizeof ".text");
7099 The @option{-Wsizeof-pointer-memaccess} option is enabled by @option{-Wall}.
7101 @item -Wsizeof-array-argument
7102 @opindex Wsizeof-array-argument
7103 @opindex Wno-sizeof-array-argument
7104 Warn when the @code{sizeof} operator is applied to a parameter that is
7105 declared as an array in a function definition. This warning is enabled by
7106 default for C and C++ programs.
7108 @item -Wmemset-elt-size
7109 @opindex Wmemset-elt-size
7110 @opindex Wno-memset-elt-size
7111 Warn for suspicious calls to the @code{memset} built-in function, if the
7112 first argument references an array, and the third argument is a number
7113 equal to the number of elements, but not equal to the size of the array
7114 in memory. This indicates that the user has omitted a multiplication by
7115 the element size. This warning is enabled by @option{-Wall}.
7117 @item -Wmemset-transposed-args
7118 @opindex Wmemset-transposed-args
7119 @opindex Wno-memset-transposed-args
7120 Warn for suspicious calls to the @code{memset} built-in function where
7121 the second argument is not zero and the third argument is zero. For
7122 example, the call @code{memset (buf, sizeof buf, 0)} is diagnosed because
7123 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostic
7124 is only emitted if the third argument is a literal zero. Otherwise, if
7125 it is an expression that is folded to zero, or a cast of zero to some
7126 type, it is far less likely that the arguments have been mistakenly
7127 transposed and no warning is emitted. This warning is enabled
7132 @opindex Wno-address
7133 Warn about suspicious uses of memory addresses. These include using
7134 the address of a function in a conditional expression, such as
7135 @code{void func(void); if (func)}, and comparisons against the memory
7136 address of a string literal, such as @code{if (x == "abc")}. Such
7137 uses typically indicate a programmer error: the address of a function
7138 always evaluates to true, so their use in a conditional usually
7139 indicate that the programmer forgot the parentheses in a function
7140 call; and comparisons against string literals result in unspecified
7141 behavior and are not portable in C, so they usually indicate that the
7142 programmer intended to use @code{strcmp}. This warning is enabled by
7145 @item -Waddress-of-packed-member
7146 @opindex Waddress-of-packed-member
7147 @opindex Wno-address-of-packed-member
7148 Warn when the address of packed member of struct or union is taken,
7149 which usually results in an unaligned pointer value. This is
7153 @opindex Wlogical-op
7154 @opindex Wno-logical-op
7155 Warn about suspicious uses of logical operators in expressions.
7156 This includes using logical operators in contexts where a
7157 bit-wise operator is likely to be expected. Also warns when
7158 the operands of a logical operator are the same:
7161 if (a < 0 && a < 0) @{ @dots{} @}
7164 @item -Wlogical-not-parentheses
7165 @opindex Wlogical-not-parentheses
7166 @opindex Wno-logical-not-parentheses
7167 Warn about logical not used on the left hand side operand of a comparison.
7168 This option does not warn if the right operand is considered to be a boolean
7169 expression. Its purpose is to detect suspicious code like the following:
7173 if (!a > 1) @{ @dots{} @}
7176 It is possible to suppress the warning by wrapping the LHS into
7179 if ((!a) > 1) @{ @dots{} @}
7182 This warning is enabled by @option{-Wall}.
7184 @item -Waggregate-return
7185 @opindex Waggregate-return
7186 @opindex Wno-aggregate-return
7187 Warn if any functions that return structures or unions are defined or
7188 called. (In languages where you can return an array, this also elicits
7191 @item -Wno-aggressive-loop-optimizations
7192 @opindex Wno-aggressive-loop-optimizations
7193 @opindex Waggressive-loop-optimizations
7194 Warn if in a loop with constant number of iterations the compiler detects
7195 undefined behavior in some statement during one or more of the iterations.
7197 @item -Wno-attributes
7198 @opindex Wno-attributes
7199 @opindex Wattributes
7200 Do not warn if an unexpected @code{__attribute__} is used, such as
7201 unrecognized attributes, function attributes applied to variables,
7202 etc. This does not stop errors for incorrect use of supported
7205 @item -Wno-builtin-declaration-mismatch
7206 @opindex Wno-builtin-declaration-mismatch
7207 @opindex Wbuiltin-declaration-mismatch
7208 Warn if a built-in function is declared with an incompatible signature
7209 or as a non-function, or when a built-in function declared with a type
7210 that does not include a prototype is called with arguments whose promoted
7211 types do not match those expected by the function. When @option{-Wextra}
7212 is specified, also warn when a built-in function that takes arguments is
7213 declared without a prototype. The @option{-Wno-builtin-declaration-mismatch}
7214 warning is enabled by default. To avoid the warning include the appropriate
7215 header to bring the prototypes of built-in functions into scope.
7217 For example, the call to @code{memset} below is diagnosed by the warning
7218 because the function expects a value of type @code{size_t} as its argument
7219 but the type of @code{32} is @code{int}. With @option{-Wextra},
7220 the declaration of the function is diagnosed as well.
7222 extern void* memset ();
7225 memset (d, '\0', 32);
7229 @item -Wno-builtin-macro-redefined
7230 @opindex Wno-builtin-macro-redefined
7231 @opindex Wbuiltin-macro-redefined
7232 Do not warn if certain built-in macros are redefined. This suppresses
7233 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
7234 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
7236 @item -Wstrict-prototypes @r{(C and Objective-C only)}
7237 @opindex Wstrict-prototypes
7238 @opindex Wno-strict-prototypes
7239 Warn if a function is declared or defined without specifying the
7240 argument types. (An old-style function definition is permitted without
7241 a warning if preceded by a declaration that specifies the argument
7244 @item -Wold-style-declaration @r{(C and Objective-C only)}
7245 @opindex Wold-style-declaration
7246 @opindex Wno-old-style-declaration
7247 Warn for obsolescent usages, according to the C Standard, in a
7248 declaration. For example, warn if storage-class specifiers like
7249 @code{static} are not the first things in a declaration. This warning
7250 is also enabled by @option{-Wextra}.
7252 @item -Wold-style-definition @r{(C and Objective-C only)}
7253 @opindex Wold-style-definition
7254 @opindex Wno-old-style-definition
7255 Warn if an old-style function definition is used. A warning is given
7256 even if there is a previous prototype.
7258 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
7259 @opindex Wmissing-parameter-type
7260 @opindex Wno-missing-parameter-type
7261 A function parameter is declared without a type specifier in K&R-style
7268 This warning is also enabled by @option{-Wextra}.
7270 @item -Wmissing-prototypes @r{(C and Objective-C only)}
7271 @opindex Wmissing-prototypes
7272 @opindex Wno-missing-prototypes
7273 Warn if a global function is defined without a previous prototype
7274 declaration. This warning is issued even if the definition itself
7275 provides a prototype. Use this option to detect global functions
7276 that do not have a matching prototype declaration in a header file.
7277 This option is not valid for C++ because all function declarations
7278 provide prototypes and a non-matching declaration declares an
7279 overload rather than conflict with an earlier declaration.
7280 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
7282 @item -Wmissing-declarations
7283 @opindex Wmissing-declarations
7284 @opindex Wno-missing-declarations
7285 Warn if a global function is defined without a previous declaration.
7286 Do so even if the definition itself provides a prototype.
7287 Use this option to detect global functions that are not declared in
7288 header files. In C, no warnings are issued for functions with previous
7289 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
7290 missing prototypes. In C++, no warnings are issued for function templates,
7291 or for inline functions, or for functions in anonymous namespaces.
7293 @item -Wmissing-field-initializers
7294 @opindex Wmissing-field-initializers
7295 @opindex Wno-missing-field-initializers
7299 Warn if a structure's initializer has some fields missing. For
7300 example, the following code causes such a warning, because
7301 @code{x.h} is implicitly zero:
7304 struct s @{ int f, g, h; @};
7305 struct s x = @{ 3, 4 @};
7308 This option does not warn about designated initializers, so the following
7309 modification does not trigger a warning:
7312 struct s @{ int f, g, h; @};
7313 struct s x = @{ .f = 3, .g = 4 @};
7316 In C this option does not warn about the universal zero initializer
7320 struct s @{ int f, g, h; @};
7321 struct s x = @{ 0 @};
7324 Likewise, in C++ this option does not warn about the empty @{ @}
7325 initializer, for example:
7328 struct s @{ int f, g, h; @};
7332 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
7333 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
7335 @item -Wno-multichar
7336 @opindex Wno-multichar
7338 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
7339 Usually they indicate a typo in the user's code, as they have
7340 implementation-defined values, and should not be used in portable code.
7342 @item -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]}
7343 @opindex Wnormalized=
7344 @opindex Wnormalized
7345 @opindex Wno-normalized
7348 @cindex character set, input normalization
7349 In ISO C and ISO C++, two identifiers are different if they are
7350 different sequences of characters. However, sometimes when characters
7351 outside the basic ASCII character set are used, you can have two
7352 different character sequences that look the same. To avoid confusion,
7353 the ISO 10646 standard sets out some @dfn{normalization rules} which
7354 when applied ensure that two sequences that look the same are turned into
7355 the same sequence. GCC can warn you if you are using identifiers that
7356 have not been normalized; this option controls that warning.
7358 There are four levels of warning supported by GCC@. The default is
7359 @option{-Wnormalized=nfc}, which warns about any identifier that is
7360 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
7361 recommended form for most uses. It is equivalent to
7362 @option{-Wnormalized}.
7364 Unfortunately, there are some characters allowed in identifiers by
7365 ISO C and ISO C++ that, when turned into NFC, are not allowed in
7366 identifiers. That is, there's no way to use these symbols in portable
7367 ISO C or C++ and have all your identifiers in NFC@.
7368 @option{-Wnormalized=id} suppresses the warning for these characters.
7369 It is hoped that future versions of the standards involved will correct
7370 this, which is why this option is not the default.
7372 You can switch the warning off for all characters by writing
7373 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
7374 only do this if you are using some other normalization scheme (like
7375 ``D''), because otherwise you can easily create bugs that are
7376 literally impossible to see.
7378 Some characters in ISO 10646 have distinct meanings but look identical
7379 in some fonts or display methodologies, especially once formatting has
7380 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
7381 LETTER N'', displays just like a regular @code{n} that has been
7382 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
7383 normalization scheme to convert all these into a standard form as
7384 well, and GCC warns if your code is not in NFKC if you use
7385 @option{-Wnormalized=nfkc}. This warning is comparable to warning
7386 about every identifier that contains the letter O because it might be
7387 confused with the digit 0, and so is not the default, but may be
7388 useful as a local coding convention if the programming environment
7389 cannot be fixed to display these characters distinctly.
7391 @item -Wno-attribute-warning
7392 @opindex Wno-attribute-warning
7393 @opindex Wattribute-warning
7394 Do not warn about usage of functions (@pxref{Function Attributes})
7395 declared with @code{warning} attribute. By default, this warning is
7396 enabled. @option{-Wno-attribute-warning} can be used to disable the
7397 warning or @option{-Wno-error=attribute-warning} can be used to
7398 disable the error when compiled with @option{-Werror} flag.
7400 @item -Wno-deprecated
7401 @opindex Wno-deprecated
7402 @opindex Wdeprecated
7403 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
7405 @item -Wno-deprecated-declarations
7406 @opindex Wno-deprecated-declarations
7407 @opindex Wdeprecated-declarations
7408 Do not warn about uses of functions (@pxref{Function Attributes}),
7409 variables (@pxref{Variable Attributes}), and types (@pxref{Type
7410 Attributes}) marked as deprecated by using the @code{deprecated}
7414 @opindex Wno-overflow
7416 Do not warn about compile-time overflow in constant expressions.
7421 Warn about One Definition Rule violations during link-time optimization.
7425 @opindex Wopenmp-simd
7426 @opindex Wno-openmp-simd
7427 Warn if the vectorizer cost model overrides the OpenMP
7428 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
7429 option can be used to relax the cost model.
7431 @item -Woverride-init @r{(C and Objective-C only)}
7432 @opindex Woverride-init
7433 @opindex Wno-override-init
7437 Warn if an initialized field without side effects is overridden when
7438 using designated initializers (@pxref{Designated Inits, , Designated
7441 This warning is included in @option{-Wextra}. To get other
7442 @option{-Wextra} warnings without this one, use @option{-Wextra
7443 -Wno-override-init}.
7445 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
7446 @opindex Woverride-init-side-effects
7447 @opindex Wno-override-init-side-effects
7448 Warn if an initialized field with side effects is overridden when
7449 using designated initializers (@pxref{Designated Inits, , Designated
7450 Initializers}). This warning is enabled by default.
7455 Warn if a structure is given the packed attribute, but the packed
7456 attribute has no effect on the layout or size of the structure.
7457 Such structures may be mis-aligned for little benefit. For
7458 instance, in this code, the variable @code{f.x} in @code{struct bar}
7459 is misaligned even though @code{struct bar} does not itself
7460 have the packed attribute:
7467 @} __attribute__((packed));
7475 @item -Wpacked-bitfield-compat
7476 @opindex Wpacked-bitfield-compat
7477 @opindex Wno-packed-bitfield-compat
7478 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
7479 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
7480 the change can lead to differences in the structure layout. GCC
7481 informs you when the offset of such a field has changed in GCC 4.4.
7482 For example there is no longer a 4-bit padding between field @code{a}
7483 and @code{b} in this structure:
7490 @} __attribute__ ((packed));
7493 This warning is enabled by default. Use
7494 @option{-Wno-packed-bitfield-compat} to disable this warning.
7496 @item -Wpacked-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
7497 @opindex Wpacked-not-aligned
7498 @opindex Wno-packed-not-aligned
7499 Warn if a structure field with explicitly specified alignment in a
7500 packed struct or union is misaligned. For example, a warning will
7501 be issued on @code{struct S}, like, @code{warning: alignment 1 of
7502 'struct S' is less than 8}, in this code:
7506 struct __attribute__ ((aligned (8))) S8 @{ char a[8]; @};
7507 struct __attribute__ ((packed)) S @{
7513 This warning is enabled by @option{-Wall}.
7518 Warn if padding is included in a structure, either to align an element
7519 of the structure or to align the whole structure. Sometimes when this
7520 happens it is possible to rearrange the fields of the structure to
7521 reduce the padding and so make the structure smaller.
7523 @item -Wredundant-decls
7524 @opindex Wredundant-decls
7525 @opindex Wno-redundant-decls
7526 Warn if anything is declared more than once in the same scope, even in
7527 cases where multiple declaration is valid and changes nothing.
7531 @opindex Wno-restrict
7532 Warn when an object referenced by a @code{restrict}-qualified parameter
7533 (or, in C++, a @code{__restrict}-qualified parameter) is aliased by another
7534 argument, or when copies between such objects overlap. For example,
7535 the call to the @code{strcpy} function below attempts to truncate the string
7536 by replacing its initial characters with the last four. However, because
7537 the call writes the terminating NUL into @code{a[4]}, the copies overlap and
7538 the call is diagnosed.
7543 char a[] = "abcd1234";
7548 The @option{-Wrestrict} option detects some instances of simple overlap
7549 even without optimization but works best at @option{-O2} and above. It
7550 is included in @option{-Wall}.
7552 @item -Wnested-externs @r{(C and Objective-C only)}
7553 @opindex Wnested-externs
7554 @opindex Wno-nested-externs
7555 Warn if an @code{extern} declaration is encountered within a function.
7557 @item -Wno-inherited-variadic-ctor
7558 @opindex Winherited-variadic-ctor
7559 @opindex Wno-inherited-variadic-ctor
7560 Suppress warnings about use of C++11 inheriting constructors when the
7561 base class inherited from has a C variadic constructor; the warning is
7562 on by default because the ellipsis is not inherited.
7567 Warn if a function that is declared as inline cannot be inlined.
7568 Even with this option, the compiler does not warn about failures to
7569 inline functions declared in system headers.
7571 The compiler uses a variety of heuristics to determine whether or not
7572 to inline a function. For example, the compiler takes into account
7573 the size of the function being inlined and the amount of inlining
7574 that has already been done in the current function. Therefore,
7575 seemingly insignificant changes in the source program can cause the
7576 warnings produced by @option{-Winline} to appear or disappear.
7578 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
7579 @opindex Wno-invalid-offsetof
7580 @opindex Winvalid-offsetof
7581 Suppress warnings from applying the @code{offsetof} macro to a non-POD
7582 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
7583 to a non-standard-layout type is undefined. In existing C++ implementations,
7584 however, @code{offsetof} typically gives meaningful results.
7585 This flag is for users who are aware that they are
7586 writing nonportable code and who have deliberately chosen to ignore the
7589 The restrictions on @code{offsetof} may be relaxed in a future version
7590 of the C++ standard.
7592 @item -Wint-in-bool-context
7593 @opindex Wint-in-bool-context
7594 @opindex Wno-int-in-bool-context
7595 Warn for suspicious use of integer values where boolean values are expected,
7596 such as conditional expressions (?:) using non-boolean integer constants in
7597 boolean context, like @code{if (a <= b ? 2 : 3)}. Or left shifting of signed
7598 integers in boolean context, like @code{for (a = 0; 1 << a; a++);}. Likewise
7599 for all kinds of multiplications regardless of the data type.
7600 This warning is enabled by @option{-Wall}.
7602 @item -Wno-int-to-pointer-cast
7603 @opindex Wno-int-to-pointer-cast
7604 @opindex Wint-to-pointer-cast
7605 Suppress warnings from casts to pointer type of an integer of a
7606 different size. In C++, casting to a pointer type of smaller size is
7607 an error. @option{Wint-to-pointer-cast} is enabled by default.
7610 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
7611 @opindex Wno-pointer-to-int-cast
7612 @opindex Wpointer-to-int-cast
7613 Suppress warnings from casts from a pointer to an integer type of a
7617 @opindex Winvalid-pch
7618 @opindex Wno-invalid-pch
7619 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
7620 the search path but cannot be used.
7624 @opindex Wno-long-long
7625 Warn if @code{long long} type is used. This is enabled by either
7626 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
7627 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
7629 @item -Wvariadic-macros
7630 @opindex Wvariadic-macros
7631 @opindex Wno-variadic-macros
7632 Warn if variadic macros are used in ISO C90 mode, or if the GNU
7633 alternate syntax is used in ISO C99 mode. This is enabled by either
7634 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
7635 messages, use @option{-Wno-variadic-macros}.
7639 @opindex Wno-varargs
7640 Warn upon questionable usage of the macros used to handle variable
7641 arguments like @code{va_start}. This is default. To inhibit the
7642 warning messages, use @option{-Wno-varargs}.
7644 @item -Wvector-operation-performance
7645 @opindex Wvector-operation-performance
7646 @opindex Wno-vector-operation-performance
7647 Warn if vector operation is not implemented via SIMD capabilities of the
7648 architecture. Mainly useful for the performance tuning.
7649 Vector operation can be implemented @code{piecewise}, which means that the
7650 scalar operation is performed on every vector element;
7651 @code{in parallel}, which means that the vector operation is implemented
7652 using scalars of wider type, which normally is more performance efficient;
7653 and @code{as a single scalar}, which means that vector fits into a
7656 @item -Wno-virtual-move-assign
7657 @opindex Wvirtual-move-assign
7658 @opindex Wno-virtual-move-assign
7659 Suppress warnings about inheriting from a virtual base with a
7660 non-trivial C++11 move assignment operator. This is dangerous because
7661 if the virtual base is reachable along more than one path, it is
7662 moved multiple times, which can mean both objects end up in the
7663 moved-from state. If the move assignment operator is written to avoid
7664 moving from a moved-from object, this warning can be disabled.
7669 Warn if a variable-length array is used in the code.
7670 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
7671 the variable-length array.
7673 @item -Wvla-larger-than=@var{byte-size}
7674 @opindex Wvla-larger-than=
7675 @opindex Wno-vla-larger-than
7676 If this option is used, the compiler will warn for declarations of
7677 variable-length arrays whose size is either unbounded, or bounded
7678 by an argument that allows the array size to exceed @var{byte-size}
7679 bytes. This is similar to how @option{-Walloca-larger-than=}@var{byte-size}
7680 works, but with variable-length arrays.
7682 Note that GCC may optimize small variable-length arrays of a known
7683 value into plain arrays, so this warning may not get triggered for
7686 @option{-Wvla-larger-than=}@samp{PTRDIFF_MAX} is enabled by default but
7687 is typically only effective when @option{-ftree-vrp} is active (default
7688 for @option{-O2} and above).
7690 See also @option{-Walloca-larger-than=@var{byte-size}}.
7692 @item -Wno-vla-larger-than
7693 @opindex Wno-vla-larger-than
7694 Disable @option{-Wvla-larger-than=} warnings. The option is equivalent
7695 to @option{-Wvla-larger-than=}@samp{SIZE_MAX} or larger.
7697 @item -Wvolatile-register-var
7698 @opindex Wvolatile-register-var
7699 @opindex Wno-volatile-register-var
7700 Warn if a register variable is declared volatile. The volatile
7701 modifier does not inhibit all optimizations that may eliminate reads
7702 and/or writes to register variables. This warning is enabled by
7705 @item -Wdisabled-optimization
7706 @opindex Wdisabled-optimization
7707 @opindex Wno-disabled-optimization
7708 Warn if a requested optimization pass is disabled. This warning does
7709 not generally indicate that there is anything wrong with your code; it
7710 merely indicates that GCC's optimizers are unable to handle the code
7711 effectively. Often, the problem is that your code is too big or too
7712 complex; GCC refuses to optimize programs when the optimization
7713 itself is likely to take inordinate amounts of time.
7715 @item -Wpointer-sign @r{(C and Objective-C only)}
7716 @opindex Wpointer-sign
7717 @opindex Wno-pointer-sign
7718 Warn for pointer argument passing or assignment with different signedness.
7719 This option is only supported for C and Objective-C@. It is implied by
7720 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
7721 @option{-Wno-pointer-sign}.
7723 @item -Wstack-protector
7724 @opindex Wstack-protector
7725 @opindex Wno-stack-protector
7726 This option is only active when @option{-fstack-protector} is active. It
7727 warns about functions that are not protected against stack smashing.
7729 @item -Woverlength-strings
7730 @opindex Woverlength-strings
7731 @opindex Wno-overlength-strings
7732 Warn about string constants that are longer than the ``minimum
7733 maximum'' length specified in the C standard. Modern compilers
7734 generally allow string constants that are much longer than the
7735 standard's minimum limit, but very portable programs should avoid
7736 using longer strings.
7738 The limit applies @emph{after} string constant concatenation, and does
7739 not count the trailing NUL@. In C90, the limit was 509 characters; in
7740 C99, it was raised to 4095. C++98 does not specify a normative
7741 minimum maximum, so we do not diagnose overlength strings in C++@.
7743 This option is implied by @option{-Wpedantic}, and can be disabled with
7744 @option{-Wno-overlength-strings}.
7746 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
7747 @opindex Wunsuffixed-float-constants
7748 @opindex Wno-unsuffixed-float-constants
7750 Issue a warning for any floating constant that does not have
7751 a suffix. When used together with @option{-Wsystem-headers} it
7752 warns about such constants in system header files. This can be useful
7753 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
7754 from the decimal floating-point extension to C99.
7756 @item -Wno-designated-init @r{(C and Objective-C only)}
7757 Suppress warnings when a positional initializer is used to initialize
7758 a structure that has been marked with the @code{designated_init}
7762 Issue a warning when HSAIL cannot be emitted for the compiled function or
7767 @node Debugging Options
7768 @section Options for Debugging Your Program
7769 @cindex options, debugging
7770 @cindex debugging information options
7772 To tell GCC to emit extra information for use by a debugger, in almost
7773 all cases you need only to add @option{-g} to your other options.
7775 GCC allows you to use @option{-g} with
7776 @option{-O}. The shortcuts taken by optimized code may occasionally
7777 be surprising: some variables you declared may not exist
7778 at all; flow of control may briefly move where you did not expect it;
7779 some statements may not be executed because they compute constant
7780 results or their values are already at hand; some statements may
7781 execute in different places because they have been moved out of loops.
7782 Nevertheless it is possible to debug optimized output. This makes
7783 it reasonable to use the optimizer for programs that might have bugs.
7785 If you are not using some other optimization option, consider
7786 using @option{-Og} (@pxref{Optimize Options}) with @option{-g}.
7787 With no @option{-O} option at all, some compiler passes that collect
7788 information useful for debugging do not run at all, so that
7789 @option{-Og} may result in a better debugging experience.
7794 Produce debugging information in the operating system's native format
7795 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
7798 On most systems that use stabs format, @option{-g} enables use of extra
7799 debugging information that only GDB can use; this extra information
7800 makes debugging work better in GDB but probably makes other debuggers
7802 refuse to read the program. If you want to control for certain whether
7803 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
7804 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
7808 Produce debugging information for use by GDB@. This means to use the
7809 most expressive format available (DWARF, stabs, or the native format
7810 if neither of those are supported), including GDB extensions if at all
7814 @itemx -gdwarf-@var{version}
7816 Produce debugging information in DWARF format (if that is supported).
7817 The value of @var{version} may be either 2, 3, 4 or 5; the default version
7818 for most targets is 4. DWARF Version 5 is only experimental.
7820 Note that with DWARF Version 2, some ports require and always
7821 use some non-conflicting DWARF 3 extensions in the unwind tables.
7823 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
7824 for maximum benefit.
7826 GCC no longer supports DWARF Version 1, which is substantially
7827 different than Version 2 and later. For historical reasons, some
7828 other DWARF-related options such as
7829 @option{-fno-dwarf2-cfi-asm}) retain a reference to DWARF Version 2
7830 in their names, but apply to all currently-supported versions of DWARF.
7834 Produce debugging information in stabs format (if that is supported),
7835 without GDB extensions. This is the format used by DBX on most BSD
7836 systems. On MIPS, Alpha and System V Release 4 systems this option
7837 produces stabs debugging output that is not understood by DBX@.
7838 On System V Release 4 systems this option requires the GNU assembler.
7842 Produce debugging information in stabs format (if that is supported),
7843 using GNU extensions understood only by the GNU debugger (GDB)@. The
7844 use of these extensions is likely to make other debuggers crash or
7845 refuse to read the program.
7849 Produce debugging information in XCOFF format (if that is supported).
7850 This is the format used by the DBX debugger on IBM RS/6000 systems.
7854 Produce debugging information in XCOFF format (if that is supported),
7855 using GNU extensions understood only by the GNU debugger (GDB)@. The
7856 use of these extensions is likely to make other debuggers crash or
7857 refuse to read the program, and may cause assemblers other than the GNU
7858 assembler (GAS) to fail with an error.
7862 Produce debugging information in Alpha/VMS debug format (if that is
7863 supported). This is the format used by DEBUG on Alpha/VMS systems.
7866 @itemx -ggdb@var{level}
7867 @itemx -gstabs@var{level}
7868 @itemx -gxcoff@var{level}
7869 @itemx -gvms@var{level}
7870 Request debugging information and also use @var{level} to specify how
7871 much information. The default level is 2.
7873 Level 0 produces no debug information at all. Thus, @option{-g0} negates
7876 Level 1 produces minimal information, enough for making backtraces in
7877 parts of the program that you don't plan to debug. This includes
7878 descriptions of functions and external variables, and line number
7879 tables, but no information about local variables.
7881 Level 3 includes extra information, such as all the macro definitions
7882 present in the program. Some debuggers support macro expansion when
7883 you use @option{-g3}.
7885 If you use multiple @option{-g} options, with or without level numbers,
7886 the last such option is the one that is effective.
7888 @option{-gdwarf} does not accept a concatenated debug level, to avoid
7889 confusion with @option{-gdwarf-@var{level}}.
7890 Instead use an additional @option{-g@var{level}} option to change the
7891 debug level for DWARF.
7893 @item -fno-eliminate-unused-debug-symbols
7894 @opindex feliminate-unused-debug-symbols
7895 @opindex fno-eliminate-unused-debug-symbols
7896 By default, no debug information is produced for symbols that are not actually
7897 used. Use this option if you want debug information for all symbols.
7899 @item -femit-class-debug-always
7900 @opindex femit-class-debug-always
7901 Instead of emitting debugging information for a C++ class in only one
7902 object file, emit it in all object files using the class. This option
7903 should be used only with debuggers that are unable to handle the way GCC
7904 normally emits debugging information for classes because using this
7905 option increases the size of debugging information by as much as a
7908 @item -fno-merge-debug-strings
7909 @opindex fmerge-debug-strings
7910 @opindex fno-merge-debug-strings
7911 Direct the linker to not merge together strings in the debugging
7912 information that are identical in different object files. Merging is
7913 not supported by all assemblers or linkers. Merging decreases the size
7914 of the debug information in the output file at the cost of increasing
7915 link processing time. Merging is enabled by default.
7917 @item -fdebug-prefix-map=@var{old}=@var{new}
7918 @opindex fdebug-prefix-map
7919 When compiling files residing in directory @file{@var{old}}, record
7920 debugging information describing them as if the files resided in
7921 directory @file{@var{new}} instead. This can be used to replace a
7922 build-time path with an install-time path in the debug info. It can
7923 also be used to change an absolute path to a relative path by using
7924 @file{.} for @var{new}. This can give more reproducible builds, which
7925 are location independent, but may require an extra command to tell GDB
7926 where to find the source files. See also @option{-ffile-prefix-map}.
7928 @item -fvar-tracking
7929 @opindex fvar-tracking
7930 Run variable tracking pass. It computes where variables are stored at each
7931 position in code. Better debugging information is then generated
7932 (if the debugging information format supports this information).
7934 It is enabled by default when compiling with optimization (@option{-Os},
7935 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7936 the debug info format supports it.
7938 @item -fvar-tracking-assignments
7939 @opindex fvar-tracking-assignments
7940 @opindex fno-var-tracking-assignments
7941 Annotate assignments to user variables early in the compilation and
7942 attempt to carry the annotations over throughout the compilation all the
7943 way to the end, in an attempt to improve debug information while
7944 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7946 It can be enabled even if var-tracking is disabled, in which case
7947 annotations are created and maintained, but discarded at the end.
7948 By default, this flag is enabled together with @option{-fvar-tracking},
7949 except when selective scheduling is enabled.
7952 @opindex gsplit-dwarf
7953 Separate as much DWARF debugging information as possible into a
7954 separate output file with the extension @file{.dwo}. This option allows
7955 the build system to avoid linking files with debug information. To
7956 be useful, this option requires a debugger capable of reading @file{.dwo}
7959 @item -gdescribe-dies
7960 @opindex gdescribe-dies
7961 Add description attributes to some DWARF DIEs that have no name attribute,
7962 such as artificial variables, external references and call site
7967 Generate DWARF @code{.debug_pubnames} and @code{.debug_pubtypes} sections.
7969 @item -ggnu-pubnames
7970 @opindex ggnu-pubnames
7971 Generate @code{.debug_pubnames} and @code{.debug_pubtypes} sections in a format
7972 suitable for conversion into a GDB@ index. This option is only useful
7973 with a linker that can produce GDB@ index version 7.
7975 @item -fdebug-types-section
7976 @opindex fdebug-types-section
7977 @opindex fno-debug-types-section
7978 When using DWARF Version 4 or higher, type DIEs can be put into
7979 their own @code{.debug_types} section instead of making them part of the
7980 @code{.debug_info} section. It is more efficient to put them in a separate
7981 comdat section since the linker can then remove duplicates.
7982 But not all DWARF consumers support @code{.debug_types} sections yet
7983 and on some objects @code{.debug_types} produces larger instead of smaller
7984 debugging information.
7986 @item -grecord-gcc-switches
7987 @itemx -gno-record-gcc-switches
7988 @opindex grecord-gcc-switches
7989 @opindex gno-record-gcc-switches
7990 This switch causes the command-line options used to invoke the
7991 compiler that may affect code generation to be appended to the
7992 DW_AT_producer attribute in DWARF debugging information. The options
7993 are concatenated with spaces separating them from each other and from
7994 the compiler version.
7995 It is enabled by default.
7996 See also @option{-frecord-gcc-switches} for another
7997 way of storing compiler options into the object file.
7999 @item -gstrict-dwarf
8000 @opindex gstrict-dwarf
8001 Disallow using extensions of later DWARF standard version than selected
8002 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
8003 DWARF extensions from later standard versions is allowed.
8005 @item -gno-strict-dwarf
8006 @opindex gno-strict-dwarf
8007 Allow using extensions of later DWARF standard version than selected with
8008 @option{-gdwarf-@var{version}}.
8010 @item -gas-loc-support
8011 @opindex gas-loc-support
8012 Inform the compiler that the assembler supports @code{.loc} directives.
8013 It may then use them for the assembler to generate DWARF2+ line number
8016 This is generally desirable, because assembler-generated line-number
8017 tables are a lot more compact than those the compiler can generate
8020 This option will be enabled by default if, at GCC configure time, the
8021 assembler was found to support such directives.
8023 @item -gno-as-loc-support
8024 @opindex gno-as-loc-support
8025 Force GCC to generate DWARF2+ line number tables internally, if DWARF2+
8026 line number tables are to be generated.
8028 @item -gas-locview-support
8029 @opindex gas-locview-support
8030 Inform the compiler that the assembler supports @code{view} assignment
8031 and reset assertion checking in @code{.loc} directives.
8033 This option will be enabled by default if, at GCC configure time, the
8034 assembler was found to support them.
8036 @item -gno-as-locview-support
8037 Force GCC to assign view numbers internally, if
8038 @option{-gvariable-location-views} are explicitly requested.
8041 @itemx -gno-column-info
8042 @opindex gcolumn-info
8043 @opindex gno-column-info
8044 Emit location column information into DWARF debugging information, rather
8045 than just file and line.
8046 This option is enabled by default.
8048 @item -gstatement-frontiers
8049 @itemx -gno-statement-frontiers
8050 @opindex gstatement-frontiers
8051 @opindex gno-statement-frontiers
8052 This option causes GCC to create markers in the internal representation
8053 at the beginning of statements, and to keep them roughly in place
8054 throughout compilation, using them to guide the output of @code{is_stmt}
8055 markers in the line number table. This is enabled by default when
8056 compiling with optimization (@option{-Os}, @option{-O}, @option{-O2},
8057 @dots{}), and outputting DWARF 2 debug information at the normal level.
8059 @item -gvariable-location-views
8060 @itemx -gvariable-location-views=incompat5
8061 @itemx -gno-variable-location-views
8062 @opindex gvariable-location-views
8063 @opindex gvariable-location-views=incompat5
8064 @opindex gno-variable-location-views
8065 Augment variable location lists with progressive view numbers implied
8066 from the line number table. This enables debug information consumers to
8067 inspect state at certain points of the program, even if no instructions
8068 associated with the corresponding source locations are present at that
8069 point. If the assembler lacks support for view numbers in line number
8070 tables, this will cause the compiler to emit the line number table,
8071 which generally makes them somewhat less compact. The augmented line
8072 number tables and location lists are fully backward-compatible, so they
8073 can be consumed by debug information consumers that are not aware of
8074 these augmentations, but they won't derive any benefit from them either.
8076 This is enabled by default when outputting DWARF 2 debug information at
8077 the normal level, as long as there is assembler support,
8078 @option{-fvar-tracking-assignments} is enabled and
8079 @option{-gstrict-dwarf} is not. When assembler support is not
8080 available, this may still be enabled, but it will force GCC to output
8081 internal line number tables, and if
8082 @option{-ginternal-reset-location-views} is not enabled, that will most
8083 certainly lead to silently mismatching location views.
8085 There is a proposed representation for view numbers that is not backward
8086 compatible with the location list format introduced in DWARF 5, that can
8087 be enabled with @option{-gvariable-location-views=incompat5}. This
8088 option may be removed in the future, is only provided as a reference
8089 implementation of the proposed representation. Debug information
8090 consumers are not expected to support this extended format, and they
8091 would be rendered unable to decode location lists using it.
8093 @item -ginternal-reset-location-views
8094 @itemx -gnointernal-reset-location-views
8095 @opindex ginternal-reset-location-views
8096 @opindex gno-internal-reset-location-views
8097 Attempt to determine location views that can be omitted from location
8098 view lists. This requires the compiler to have very accurate insn
8099 length estimates, which isn't always the case, and it may cause
8100 incorrect view lists to be generated silently when using an assembler
8101 that does not support location view lists. The GNU assembler will flag
8102 any such error as a @code{view number mismatch}. This is only enabled
8103 on ports that define a reliable estimation function.
8105 @item -ginline-points
8106 @itemx -gno-inline-points
8107 @opindex ginline-points
8108 @opindex gno-inline-points
8109 Generate extended debug information for inlined functions. Location
8110 view tracking markers are inserted at inlined entry points, so that
8111 address and view numbers can be computed and output in debug
8112 information. This can be enabled independently of location views, in
8113 which case the view numbers won't be output, but it can only be enabled
8114 along with statement frontiers, and it is only enabled by default if
8115 location views are enabled.
8117 @item -gz@r{[}=@var{type}@r{]}
8119 Produce compressed debug sections in DWARF format, if that is supported.
8120 If @var{type} is not given, the default type depends on the capabilities
8121 of the assembler and linker used. @var{type} may be one of
8122 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
8123 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
8124 compression in traditional GNU format). If the linker doesn't support
8125 writing compressed debug sections, the option is rejected. Otherwise,
8126 if the assembler does not support them, @option{-gz} is silently ignored
8127 when producing object files.
8129 @item -femit-struct-debug-baseonly
8130 @opindex femit-struct-debug-baseonly
8131 Emit debug information for struct-like types
8132 only when the base name of the compilation source file
8133 matches the base name of file in which the struct is defined.
8135 This option substantially reduces the size of debugging information,
8136 but at significant potential loss in type information to the debugger.
8137 See @option{-femit-struct-debug-reduced} for a less aggressive option.
8138 See @option{-femit-struct-debug-detailed} for more detailed control.
8140 This option works only with DWARF debug output.
8142 @item -femit-struct-debug-reduced
8143 @opindex femit-struct-debug-reduced
8144 Emit debug information for struct-like types
8145 only when the base name of the compilation source file
8146 matches the base name of file in which the type is defined,
8147 unless the struct is a template or defined in a system header.
8149 This option significantly reduces the size of debugging information,
8150 with some potential loss in type information to the debugger.
8151 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
8152 See @option{-femit-struct-debug-detailed} for more detailed control.
8154 This option works only with DWARF debug output.
8156 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
8157 @opindex femit-struct-debug-detailed
8158 Specify the struct-like types
8159 for which the compiler generates debug information.
8160 The intent is to reduce duplicate struct debug information
8161 between different object files within the same program.
8163 This option is a detailed version of
8164 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
8165 which serves for most needs.
8167 A specification has the syntax@*
8168 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
8170 The optional first word limits the specification to
8171 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
8172 A struct type is used directly when it is the type of a variable, member.
8173 Indirect uses arise through pointers to structs.
8174 That is, when use of an incomplete struct is valid, the use is indirect.
8176 @samp{struct one direct; struct two * indirect;}.
8178 The optional second word limits the specification to
8179 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
8180 Generic structs are a bit complicated to explain.
8181 For C++, these are non-explicit specializations of template classes,
8182 or non-template classes within the above.
8183 Other programming languages have generics,
8184 but @option{-femit-struct-debug-detailed} does not yet implement them.
8186 The third word specifies the source files for those
8187 structs for which the compiler should emit debug information.
8188 The values @samp{none} and @samp{any} have the normal meaning.
8189 The value @samp{base} means that
8190 the base of name of the file in which the type declaration appears
8191 must match the base of the name of the main compilation file.
8192 In practice, this means that when compiling @file{foo.c}, debug information
8193 is generated for types declared in that file and @file{foo.h},
8194 but not other header files.
8195 The value @samp{sys} means those types satisfying @samp{base}
8196 or declared in system or compiler headers.
8198 You may need to experiment to determine the best settings for your application.
8200 The default is @option{-femit-struct-debug-detailed=all}.
8202 This option works only with DWARF debug output.
8204 @item -fno-dwarf2-cfi-asm
8205 @opindex fdwarf2-cfi-asm
8206 @opindex fno-dwarf2-cfi-asm
8207 Emit DWARF unwind info as compiler generated @code{.eh_frame} section
8208 instead of using GAS @code{.cfi_*} directives.
8210 @item -fno-eliminate-unused-debug-types
8211 @opindex feliminate-unused-debug-types
8212 @opindex fno-eliminate-unused-debug-types
8213 Normally, when producing DWARF output, GCC avoids producing debug symbol
8214 output for types that are nowhere used in the source file being compiled.
8215 Sometimes it is useful to have GCC emit debugging
8216 information for all types declared in a compilation
8217 unit, regardless of whether or not they are actually used
8218 in that compilation unit, for example
8219 if, in the debugger, you want to cast a value to a type that is
8220 not actually used in your program (but is declared). More often,
8221 however, this results in a significant amount of wasted space.
8224 @node Optimize Options
8225 @section Options That Control Optimization
8226 @cindex optimize options
8227 @cindex options, optimization
8229 These options control various sorts of optimizations.
8231 Without any optimization option, the compiler's goal is to reduce the
8232 cost of compilation and to make debugging produce the expected
8233 results. Statements are independent: if you stop the program with a
8234 breakpoint between statements, you can then assign a new value to any
8235 variable or change the program counter to any other statement in the
8236 function and get exactly the results you expect from the source
8239 Turning on optimization flags makes the compiler attempt to improve
8240 the performance and/or code size at the expense of compilation time
8241 and possibly the ability to debug the program.
8243 The compiler performs optimization based on the knowledge it has of the
8244 program. Compiling multiple files at once to a single output file mode allows
8245 the compiler to use information gained from all of the files when compiling
8248 Not all optimizations are controlled directly by a flag. Only
8249 optimizations that have a flag are listed in this section.
8251 Most optimizations are completely disabled at @option{-O0} or if an
8252 @option{-O} level is not set on the command line, even if individual
8253 optimization flags are specified. Similarly, @option{-Og} suppresses
8254 many optimization passes.
8256 Depending on the target and how GCC was configured, a slightly different
8257 set of optimizations may be enabled at each @option{-O} level than
8258 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
8259 to find out the exact set of optimizations that are enabled at each level.
8260 @xref{Overall Options}, for examples.
8267 Optimize. Optimizing compilation takes somewhat more time, and a lot
8268 more memory for a large function.
8270 With @option{-O}, the compiler tries to reduce code size and execution
8271 time, without performing any optimizations that take a great deal of
8274 @c Note that in addition to the default_options_table list in opts.c,
8275 @c several optimization flags default to true but control optimization
8276 @c passes that are explicitly disabled at -O0.
8278 @option{-O} turns on the following optimization flags:
8280 @c Please keep the following list alphabetized.
8281 @gccoptlist{-fauto-inc-dec @gol
8282 -fbranch-count-reg @gol
8283 -fcombine-stack-adjustments @gol
8285 -fcprop-registers @gol
8288 -fdelayed-branch @gol
8290 -fforward-propagate @gol
8291 -fguess-branch-probability @gol
8292 -fif-conversion @gol
8293 -fif-conversion2 @gol
8294 -finline-functions-called-once @gol
8296 -fipa-pure-const @gol
8297 -fipa-reference @gol
8298 -fipa-reference-addressable @gol
8299 -fmerge-constants @gol
8300 -fmove-loop-invariants @gol
8301 -fomit-frame-pointer @gol
8302 -freorder-blocks @gol
8304 -fshrink-wrap-separate @gol
8305 -fsplit-wide-types @gol
8311 -ftree-coalesce-vars @gol
8312 -ftree-copy-prop @gol
8314 -ftree-dominator-opts @gol
8316 -ftree-forwprop @gol
8320 -ftree-scev-cprop @gol
8329 Optimize even more. GCC performs nearly all supported optimizations
8330 that do not involve a space-speed tradeoff.
8331 As compared to @option{-O}, this option increases both compilation time
8332 and the performance of the generated code.
8334 @option{-O2} turns on all optimization flags specified by @option{-O}. It
8335 also turns on the following optimization flags:
8337 @c Please keep the following list alphabetized!
8338 @gccoptlist{-falign-functions -falign-jumps @gol
8339 -falign-labels -falign-loops @gol
8341 -fcode-hoisting @gol
8343 -fcse-follow-jumps -fcse-skip-blocks @gol
8344 -fdelete-null-pointer-checks @gol
8345 -fdevirtualize -fdevirtualize-speculatively @gol
8346 -fexpensive-optimizations @gol
8348 -fgcse -fgcse-lm @gol
8349 -fhoist-adjacent-loads @gol
8350 -finline-small-functions @gol
8351 -findirect-inlining @gol
8352 -fipa-bit-cp -fipa-cp -fipa-icf @gol
8353 -fipa-ra -fipa-sra -fipa-vrp @gol
8354 -fisolate-erroneous-paths-dereference @gol
8356 -foptimize-sibling-calls @gol
8357 -foptimize-strlen @gol
8358 -fpartial-inlining @gol
8360 -freorder-blocks-algorithm=stc @gol
8361 -freorder-blocks-and-partition -freorder-functions @gol
8362 -frerun-cse-after-loop @gol
8363 -fschedule-insns -fschedule-insns2 @gol
8364 -fsched-interblock -fsched-spec @gol
8365 -fstore-merging @gol
8366 -fstrict-aliasing @gol
8368 -ftree-builtin-call-dce @gol
8370 -ftree-switch-conversion -ftree-tail-merge @gol
8373 Please note the warning under @option{-fgcse} about
8374 invoking @option{-O2} on programs that use computed gotos.
8378 Optimize yet more. @option{-O3} turns on all optimizations specified
8379 by @option{-O2} and also turns on the following optimization flags:
8381 @c Please keep the following list alphabetized!
8382 @gccoptlist{-fgcse-after-reload @gol
8383 -finline-functions @gol
8385 -floop-interchange @gol
8386 -floop-unroll-and-jam @gol
8388 -fpredictive-commoning @gol
8390 -ftree-loop-distribute-patterns @gol
8391 -ftree-loop-distribution @gol
8392 -ftree-loop-vectorize @gol
8393 -ftree-partial-pre @gol
8394 -ftree-slp-vectorize @gol
8395 -funswitch-loops @gol
8396 -fvect-cost-model @gol
8397 -fversion-loops-for-strides}
8401 Reduce compilation time and make debugging produce the expected
8402 results. This is the default.
8406 Optimize for size. @option{-Os} enables all @option{-O2} optimizations
8407 except those that often increase code size:
8409 @gccoptlist{-falign-functions -falign-jumps @gol
8410 -falign-labels -falign-loops @gol
8411 -fprefetch-loop-arrays -freorder-blocks-algorithm=stc}
8413 It also enables @option{-finline-functions}, causes the compiler to tune for
8414 code size rather than execution speed, and performs further optimizations
8415 designed to reduce code size.
8419 Disregard strict standards compliance. @option{-Ofast} enables all
8420 @option{-O3} optimizations. It also enables optimizations that are not
8421 valid for all standard-compliant programs.
8422 It turns on @option{-ffast-math} and the Fortran-specific
8423 @option{-fstack-arrays}, unless @option{-fmax-stack-var-size} is
8424 specified, and @option{-fno-protect-parens}.
8428 Optimize debugging experience. @option{-Og} should be the optimization
8429 level of choice for the standard edit-compile-debug cycle, offering
8430 a reasonable level of optimization while maintaining fast compilation
8431 and a good debugging experience. It is a better choice than @option{-O0}
8432 for producing debuggable code because some compiler passes
8433 that collect debug information are disabled at @option{-O0}.
8435 Like @option{-O0}, @option{-Og} completely disables a number of
8436 optimization passes so that individual options controlling them have
8437 no effect. Otherwise @option{-Og} enables all @option{-O1}
8438 optimization flags except for those that may interfere with debugging:
8440 @gccoptlist{-fbranch-count-reg -fdelayed-branch @gol
8441 -fdse -fif-conversion -fif-conversion2 @gol
8442 -finline-functions-called-once @gol
8443 -fmove-loop-invariants -fssa-phiopt @gol
8444 -ftree-bit-ccp -ftree-dse -ftree-pta -ftree-sra}
8448 If you use multiple @option{-O} options, with or without level numbers,
8449 the last such option is the one that is effective.
8451 Options of the form @option{-f@var{flag}} specify machine-independent
8452 flags. Most flags have both positive and negative forms; the negative
8453 form of @option{-ffoo} is @option{-fno-foo}. In the table
8454 below, only one of the forms is listed---the one you typically
8455 use. You can figure out the other form by either removing @samp{no-}
8458 The following options control specific optimizations. They are either
8459 activated by @option{-O} options or are related to ones that are. You
8460 can use the following flags in the rare cases when ``fine-tuning'' of
8461 optimizations to be performed is desired.
8464 @item -fno-defer-pop
8465 @opindex fno-defer-pop
8467 For machines that must pop arguments after a function call, always pop
8468 the arguments as soon as each function returns.
8469 At levels @option{-O1} and higher, @option{-fdefer-pop} is the default;
8470 this allows the compiler to let arguments accumulate on the stack for several
8471 function calls and pop them all at once.
8473 @item -fforward-propagate
8474 @opindex fforward-propagate
8475 Perform a forward propagation pass on RTL@. The pass tries to combine two
8476 instructions and checks if the result can be simplified. If loop unrolling
8477 is active, two passes are performed and the second is scheduled after
8480 This option is enabled by default at optimization levels @option{-O},
8481 @option{-O2}, @option{-O3}, @option{-Os}.
8483 @item -ffp-contract=@var{style}
8484 @opindex ffp-contract
8485 @option{-ffp-contract=off} disables floating-point expression contraction.
8486 @option{-ffp-contract=fast} enables floating-point expression contraction
8487 such as forming of fused multiply-add operations if the target has
8488 native support for them.
8489 @option{-ffp-contract=on} enables floating-point expression contraction
8490 if allowed by the language standard. This is currently not implemented
8491 and treated equal to @option{-ffp-contract=off}.
8493 The default is @option{-ffp-contract=fast}.
8495 @item -fomit-frame-pointer
8496 @opindex fomit-frame-pointer
8497 Omit the frame pointer in functions that don't need one. This avoids the
8498 instructions to save, set up and restore the frame pointer; on many targets
8499 it also makes an extra register available.
8501 On some targets this flag has no effect because the standard calling sequence
8502 always uses a frame pointer, so it cannot be omitted.
8504 Note that @option{-fno-omit-frame-pointer} doesn't guarantee the frame pointer
8505 is used in all functions. Several targets always omit the frame pointer in
8508 Enabled by default at @option{-O} and higher.
8510 @item -foptimize-sibling-calls
8511 @opindex foptimize-sibling-calls
8512 Optimize sibling and tail recursive calls.
8514 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8516 @item -foptimize-strlen
8517 @opindex foptimize-strlen
8518 Optimize various standard C string functions (e.g.@: @code{strlen},
8519 @code{strchr} or @code{strcpy}) and
8520 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
8522 Enabled at levels @option{-O2}, @option{-O3}.
8527 Do not expand any functions inline apart from those marked with
8528 the @code{always_inline} attribute. This is the default when not
8531 Single functions can be exempted from inlining by marking them
8532 with the @code{noinline} attribute.
8534 @item -finline-small-functions
8535 @opindex finline-small-functions
8536 Integrate functions into their callers when their body is smaller than expected
8537 function call code (so overall size of program gets smaller). The compiler
8538 heuristically decides which functions are simple enough to be worth integrating
8539 in this way. This inlining applies to all functions, even those not declared
8542 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8544 @item -findirect-inlining
8545 @opindex findirect-inlining
8546 Inline also indirect calls that are discovered to be known at compile
8547 time thanks to previous inlining. This option has any effect only
8548 when inlining itself is turned on by the @option{-finline-functions}
8549 or @option{-finline-small-functions} options.
8551 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8553 @item -finline-functions
8554 @opindex finline-functions
8555 Consider all functions for inlining, even if they are not declared inline.
8556 The compiler heuristically decides which functions are worth integrating
8559 If all calls to a given function are integrated, and the function is
8560 declared @code{static}, then the function is normally not output as
8561 assembler code in its own right.
8563 Enabled at levels @option{-O3}, @option{-Os}. Also enabled
8564 by @option{-fprofile-use} and @option{-fauto-profile}.
8566 @item -finline-functions-called-once
8567 @opindex finline-functions-called-once
8568 Consider all @code{static} functions called once for inlining into their
8569 caller even if they are not marked @code{inline}. If a call to a given
8570 function is integrated, then the function is not output as assembler code
8573 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os},
8574 but not @option{-Og}.
8576 @item -fearly-inlining
8577 @opindex fearly-inlining
8578 Inline functions marked by @code{always_inline} and functions whose body seems
8579 smaller than the function call overhead early before doing
8580 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
8581 makes profiling significantly cheaper and usually inlining faster on programs
8582 having large chains of nested wrapper functions.
8588 Perform interprocedural scalar replacement of aggregates, removal of
8589 unused parameters and replacement of parameters passed by reference
8590 by parameters passed by value.
8592 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
8594 @item -finline-limit=@var{n}
8595 @opindex finline-limit
8596 By default, GCC limits the size of functions that can be inlined. This flag
8597 allows coarse control of this limit. @var{n} is the size of functions that
8598 can be inlined in number of pseudo instructions.
8600 Inlining is actually controlled by a number of parameters, which may be
8601 specified individually by using @option{--param @var{name}=@var{value}}.
8602 The @option{-finline-limit=@var{n}} option sets some of these parameters
8606 @item max-inline-insns-single
8607 is set to @var{n}/2.
8608 @item max-inline-insns-auto
8609 is set to @var{n}/2.
8612 See below for a documentation of the individual
8613 parameters controlling inlining and for the defaults of these parameters.
8615 @emph{Note:} there may be no value to @option{-finline-limit} that results
8616 in default behavior.
8618 @emph{Note:} pseudo instruction represents, in this particular context, an
8619 abstract measurement of function's size. In no way does it represent a count
8620 of assembly instructions and as such its exact meaning might change from one
8621 release to an another.
8623 @item -fno-keep-inline-dllexport
8624 @opindex fno-keep-inline-dllexport
8625 @opindex fkeep-inline-dllexport
8626 This is a more fine-grained version of @option{-fkeep-inline-functions},
8627 which applies only to functions that are declared using the @code{dllexport}
8628 attribute or declspec. @xref{Function Attributes,,Declaring Attributes of
8631 @item -fkeep-inline-functions
8632 @opindex fkeep-inline-functions
8633 In C, emit @code{static} functions that are declared @code{inline}
8634 into the object file, even if the function has been inlined into all
8635 of its callers. This switch does not affect functions using the
8636 @code{extern inline} extension in GNU C90@. In C++, emit any and all
8637 inline functions into the object file.
8639 @item -fkeep-static-functions
8640 @opindex fkeep-static-functions
8641 Emit @code{static} functions into the object file, even if the function
8644 @item -fkeep-static-consts
8645 @opindex fkeep-static-consts
8646 Emit variables declared @code{static const} when optimization isn't turned
8647 on, even if the variables aren't referenced.
8649 GCC enables this option by default. If you want to force the compiler to
8650 check if a variable is referenced, regardless of whether or not
8651 optimization is turned on, use the @option{-fno-keep-static-consts} option.
8653 @item -fmerge-constants
8654 @opindex fmerge-constants
8655 Attempt to merge identical constants (string constants and floating-point
8656 constants) across compilation units.
8658 This option is the default for optimized compilation if the assembler and
8659 linker support it. Use @option{-fno-merge-constants} to inhibit this
8662 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8664 @item -fmerge-all-constants
8665 @opindex fmerge-all-constants
8666 Attempt to merge identical constants and identical variables.
8668 This option implies @option{-fmerge-constants}. In addition to
8669 @option{-fmerge-constants} this considers e.g.@: even constant initialized
8670 arrays or initialized constant variables with integral or floating-point
8671 types. Languages like C or C++ require each variable, including multiple
8672 instances of the same variable in recursive calls, to have distinct locations,
8673 so using this option results in non-conforming
8676 @item -fmodulo-sched
8677 @opindex fmodulo-sched
8678 Perform swing modulo scheduling immediately before the first scheduling
8679 pass. This pass looks at innermost loops and reorders their
8680 instructions by overlapping different iterations.
8682 @item -fmodulo-sched-allow-regmoves
8683 @opindex fmodulo-sched-allow-regmoves
8684 Perform more aggressive SMS-based modulo scheduling with register moves
8685 allowed. By setting this flag certain anti-dependences edges are
8686 deleted, which triggers the generation of reg-moves based on the
8687 life-range analysis. This option is effective only with
8688 @option{-fmodulo-sched} enabled.
8690 @item -fno-branch-count-reg
8691 @opindex fno-branch-count-reg
8692 @opindex fbranch-count-reg
8693 Disable the optimization pass that scans for opportunities to use
8694 ``decrement and branch'' instructions on a count register instead of
8695 instruction sequences that decrement a register, compare it against zero, and
8696 then branch based upon the result. This option is only meaningful on
8697 architectures that support such instructions, which include x86, PowerPC,
8698 IA-64 and S/390. Note that the @option{-fno-branch-count-reg} option
8699 doesn't remove the decrement and branch instructions from the generated
8700 instruction stream introduced by other optimization passes.
8702 The default is @option{-fbranch-count-reg} at @option{-O1} and higher,
8703 except for @option{-Og}.
8705 @item -fno-function-cse
8706 @opindex fno-function-cse
8707 @opindex ffunction-cse
8708 Do not put function addresses in registers; make each instruction that
8709 calls a constant function contain the function's address explicitly.
8711 This option results in less efficient code, but some strange hacks
8712 that alter the assembler output may be confused by the optimizations
8713 performed when this option is not used.
8715 The default is @option{-ffunction-cse}
8717 @item -fno-zero-initialized-in-bss
8718 @opindex fno-zero-initialized-in-bss
8719 @opindex fzero-initialized-in-bss
8720 If the target supports a BSS section, GCC by default puts variables that
8721 are initialized to zero into BSS@. This can save space in the resulting
8724 This option turns off this behavior because some programs explicitly
8725 rely on variables going to the data section---e.g., so that the
8726 resulting executable can find the beginning of that section and/or make
8727 assumptions based on that.
8729 The default is @option{-fzero-initialized-in-bss}.
8731 @item -fthread-jumps
8732 @opindex fthread-jumps
8733 Perform optimizations that check to see if a jump branches to a
8734 location where another comparison subsumed by the first is found. If
8735 so, the first branch is redirected to either the destination of the
8736 second branch or a point immediately following it, depending on whether
8737 the condition is known to be true or false.
8739 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8741 @item -fsplit-wide-types
8742 @opindex fsplit-wide-types
8743 When using a type that occupies multiple registers, such as @code{long
8744 long} on a 32-bit system, split the registers apart and allocate them
8745 independently. This normally generates better code for those types,
8746 but may make debugging more difficult.
8748 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
8751 @item -fsplit-wide-types-early
8752 @opindex fsplit-wide-types-early
8753 Fully split wide types early, instead of very late.
8754 This option has no effect unless @option{-fsplit-wide-types} is turned on.
8756 This is the default on some targets.
8758 @item -fcse-follow-jumps
8759 @opindex fcse-follow-jumps
8760 In common subexpression elimination (CSE), scan through jump instructions
8761 when the target of the jump is not reached by any other path. For
8762 example, when CSE encounters an @code{if} statement with an
8763 @code{else} clause, CSE follows the jump when the condition
8766 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8768 @item -fcse-skip-blocks
8769 @opindex fcse-skip-blocks
8770 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
8771 follow jumps that conditionally skip over blocks. When CSE
8772 encounters a simple @code{if} statement with no else clause,
8773 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
8774 body of the @code{if}.
8776 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8778 @item -frerun-cse-after-loop
8779 @opindex frerun-cse-after-loop
8780 Re-run common subexpression elimination after loop optimizations are
8783 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8787 Perform a global common subexpression elimination pass.
8788 This pass also performs global constant and copy propagation.
8790 @emph{Note:} When compiling a program using computed gotos, a GCC
8791 extension, you may get better run-time performance if you disable
8792 the global common subexpression elimination pass by adding
8793 @option{-fno-gcse} to the command line.
8795 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8799 When @option{-fgcse-lm} is enabled, global common subexpression elimination
8800 attempts to move loads that are only killed by stores into themselves. This
8801 allows a loop containing a load/store sequence to be changed to a load outside
8802 the loop, and a copy/store within the loop.
8804 Enabled by default when @option{-fgcse} is enabled.
8808 When @option{-fgcse-sm} is enabled, a store motion pass is run after
8809 global common subexpression elimination. This pass attempts to move
8810 stores out of loops. When used in conjunction with @option{-fgcse-lm},
8811 loops containing a load/store sequence can be changed to a load before
8812 the loop and a store after the loop.
8814 Not enabled at any optimization level.
8818 When @option{-fgcse-las} is enabled, the global common subexpression
8819 elimination pass eliminates redundant loads that come after stores to the
8820 same memory location (both partial and full redundancies).
8822 Not enabled at any optimization level.
8824 @item -fgcse-after-reload
8825 @opindex fgcse-after-reload
8826 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
8827 pass is performed after reload. The purpose of this pass is to clean up
8830 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
8832 @item -faggressive-loop-optimizations
8833 @opindex faggressive-loop-optimizations
8834 This option tells the loop optimizer to use language constraints to
8835 derive bounds for the number of iterations of a loop. This assumes that
8836 loop code does not invoke undefined behavior by for example causing signed
8837 integer overflows or out-of-bound array accesses. The bounds for the
8838 number of iterations of a loop are used to guide loop unrolling and peeling
8839 and loop exit test optimizations.
8840 This option is enabled by default.
8842 @item -funconstrained-commons
8843 @opindex funconstrained-commons
8844 This option tells the compiler that variables declared in common blocks
8845 (e.g.@: Fortran) may later be overridden with longer trailing arrays. This
8846 prevents certain optimizations that depend on knowing the array bounds.
8848 @item -fcrossjumping
8849 @opindex fcrossjumping
8850 Perform cross-jumping transformation.
8851 This transformation unifies equivalent code and saves code size. The
8852 resulting code may or may not perform better than without cross-jumping.
8854 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8856 @item -fauto-inc-dec
8857 @opindex fauto-inc-dec
8858 Combine increments or decrements of addresses with memory accesses.
8859 This pass is always skipped on architectures that do not have
8860 instructions to support this. Enabled by default at @option{-O} and
8861 higher on architectures that support this.
8865 Perform dead code elimination (DCE) on RTL@.
8866 Enabled by default at @option{-O} and higher.
8870 Perform dead store elimination (DSE) on RTL@.
8871 Enabled by default at @option{-O} and higher.
8873 @item -fif-conversion
8874 @opindex fif-conversion
8875 Attempt to transform conditional jumps into branch-less equivalents. This
8876 includes use of conditional moves, min, max, set flags and abs instructions, and
8877 some tricks doable by standard arithmetics. The use of conditional execution
8878 on chips where it is available is controlled by @option{-fif-conversion2}.
8880 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}, but
8881 not with @option{-Og}.
8883 @item -fif-conversion2
8884 @opindex fif-conversion2
8885 Use conditional execution (where available) to transform conditional jumps into
8886 branch-less equivalents.
8888 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}, but
8889 not with @option{-Og}.
8891 @item -fdeclone-ctor-dtor
8892 @opindex fdeclone-ctor-dtor
8893 The C++ ABI requires multiple entry points for constructors and
8894 destructors: one for a base subobject, one for a complete object, and
8895 one for a virtual destructor that calls operator delete afterwards.
8896 For a hierarchy with virtual bases, the base and complete variants are
8897 clones, which means two copies of the function. With this option, the
8898 base and complete variants are changed to be thunks that call a common
8901 Enabled by @option{-Os}.
8903 @item -fdelete-null-pointer-checks
8904 @opindex fdelete-null-pointer-checks
8905 Assume that programs cannot safely dereference null pointers, and that
8906 no code or data element resides at address zero.
8907 This option enables simple constant
8908 folding optimizations at all optimization levels. In addition, other
8909 optimization passes in GCC use this flag to control global dataflow
8910 analyses that eliminate useless checks for null pointers; these assume
8911 that a memory access to address zero always results in a trap, so
8912 that if a pointer is checked after it has already been dereferenced,
8915 Note however that in some environments this assumption is not true.
8916 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
8917 for programs that depend on that behavior.
8919 This option is enabled by default on most targets. On Nios II ELF, it
8920 defaults to off. On AVR, CR16, and MSP430, this option is completely disabled.
8922 Passes that use the dataflow information
8923 are enabled independently at different optimization levels.
8925 @item -fdevirtualize
8926 @opindex fdevirtualize
8927 Attempt to convert calls to virtual functions to direct calls. This
8928 is done both within a procedure and interprocedurally as part of
8929 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
8930 propagation (@option{-fipa-cp}).
8931 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8933 @item -fdevirtualize-speculatively
8934 @opindex fdevirtualize-speculatively
8935 Attempt to convert calls to virtual functions to speculative direct calls.
8936 Based on the analysis of the type inheritance graph, determine for a given call
8937 the set of likely targets. If the set is small, preferably of size 1, change
8938 the call into a conditional deciding between direct and indirect calls. The
8939 speculative calls enable more optimizations, such as inlining. When they seem
8940 useless after further optimization, they are converted back into original form.
8942 @item -fdevirtualize-at-ltrans
8943 @opindex fdevirtualize-at-ltrans
8944 Stream extra information needed for aggressive devirtualization when running
8945 the link-time optimizer in local transformation mode.
8946 This option enables more devirtualization but
8947 significantly increases the size of streamed data. For this reason it is
8948 disabled by default.
8950 @item -fexpensive-optimizations
8951 @opindex fexpensive-optimizations
8952 Perform a number of minor optimizations that are relatively expensive.
8954 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8958 Attempt to remove redundant extension instructions. This is especially
8959 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
8960 registers after writing to their lower 32-bit half.
8962 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
8963 @option{-O3}, @option{-Os}.
8965 @item -fno-lifetime-dse
8966 @opindex fno-lifetime-dse
8967 @opindex flifetime-dse
8968 In C++ the value of an object is only affected by changes within its
8969 lifetime: when the constructor begins, the object has an indeterminate
8970 value, and any changes during the lifetime of the object are dead when
8971 the object is destroyed. Normally dead store elimination will take
8972 advantage of this; if your code relies on the value of the object
8973 storage persisting beyond the lifetime of the object, you can use this
8974 flag to disable this optimization. To preserve stores before the
8975 constructor starts (e.g.@: because your operator new clears the object
8976 storage) but still treat the object as dead after the destructor you,
8977 can use @option{-flifetime-dse=1}. The default behavior can be
8978 explicitly selected with @option{-flifetime-dse=2}.
8979 @option{-flifetime-dse=0} is equivalent to @option{-fno-lifetime-dse}.
8981 @item -flive-range-shrinkage
8982 @opindex flive-range-shrinkage
8983 Attempt to decrease register pressure through register live range
8984 shrinkage. This is helpful for fast processors with small or moderate
8987 @item -fira-algorithm=@var{algorithm}
8988 @opindex fira-algorithm
8989 Use the specified coloring algorithm for the integrated register
8990 allocator. The @var{algorithm} argument can be @samp{priority}, which
8991 specifies Chow's priority coloring, or @samp{CB}, which specifies
8992 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
8993 for all architectures, but for those targets that do support it, it is
8994 the default because it generates better code.
8996 @item -fira-region=@var{region}
8997 @opindex fira-region
8998 Use specified regions for the integrated register allocator. The
8999 @var{region} argument should be one of the following:
9004 Use all loops as register allocation regions.
9005 This can give the best results for machines with a small and/or
9006 irregular register set.
9009 Use all loops except for loops with small register pressure
9010 as the regions. This value usually gives
9011 the best results in most cases and for most architectures,
9012 and is enabled by default when compiling with optimization for speed
9013 (@option{-O}, @option{-O2}, @dots{}).
9016 Use all functions as a single region.
9017 This typically results in the smallest code size, and is enabled by default for
9018 @option{-Os} or @option{-O0}.
9022 @item -fira-hoist-pressure
9023 @opindex fira-hoist-pressure
9024 Use IRA to evaluate register pressure in the code hoisting pass for
9025 decisions to hoist expressions. This option usually results in smaller
9026 code, but it can slow the compiler down.
9028 This option is enabled at level @option{-Os} for all targets.
9030 @item -fira-loop-pressure
9031 @opindex fira-loop-pressure
9032 Use IRA to evaluate register pressure in loops for decisions to move
9033 loop invariants. This option usually results in generation
9034 of faster and smaller code on machines with large register files (>= 32
9035 registers), but it can slow the compiler down.
9037 This option is enabled at level @option{-O3} for some targets.
9039 @item -fno-ira-share-save-slots
9040 @opindex fno-ira-share-save-slots
9041 @opindex fira-share-save-slots
9042 Disable sharing of stack slots used for saving call-used hard
9043 registers living through a call. Each hard register gets a
9044 separate stack slot, and as a result function stack frames are
9047 @item -fno-ira-share-spill-slots
9048 @opindex fno-ira-share-spill-slots
9049 @opindex fira-share-spill-slots
9050 Disable sharing of stack slots allocated for pseudo-registers. Each
9051 pseudo-register that does not get a hard register gets a separate
9052 stack slot, and as a result function stack frames are larger.
9056 Enable CFG-sensitive rematerialization in LRA. Instead of loading
9057 values of spilled pseudos, LRA tries to rematerialize (recalculate)
9058 values if it is profitable.
9060 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9062 @item -fdelayed-branch
9063 @opindex fdelayed-branch
9064 If supported for the target machine, attempt to reorder instructions
9065 to exploit instruction slots available after delayed branch
9068 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os},
9069 but not at @option{-Og}.
9071 @item -fschedule-insns
9072 @opindex fschedule-insns
9073 If supported for the target machine, attempt to reorder instructions to
9074 eliminate execution stalls due to required data being unavailable. This
9075 helps machines that have slow floating point or memory load instructions
9076 by allowing other instructions to be issued until the result of the load
9077 or floating-point instruction is required.
9079 Enabled at levels @option{-O2}, @option{-O3}.
9081 @item -fschedule-insns2
9082 @opindex fschedule-insns2
9083 Similar to @option{-fschedule-insns}, but requests an additional pass of
9084 instruction scheduling after register allocation has been done. This is
9085 especially useful on machines with a relatively small number of
9086 registers and where memory load instructions take more than one cycle.
9088 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9090 @item -fno-sched-interblock
9091 @opindex fno-sched-interblock
9092 @opindex fsched-interblock
9093 Disable instruction scheduling across basic blocks, which
9094 is normally enabled when scheduling before register allocation, i.e.@:
9095 with @option{-fschedule-insns} or at @option{-O2} or higher.
9097 @item -fno-sched-spec
9098 @opindex fno-sched-spec
9099 @opindex fsched-spec
9100 Disable speculative motion of non-load instructions, which
9101 is normally enabled when scheduling before register allocation, i.e.@:
9102 with @option{-fschedule-insns} or at @option{-O2} or higher.
9104 @item -fsched-pressure
9105 @opindex fsched-pressure
9106 Enable register pressure sensitive insn scheduling before register
9107 allocation. This only makes sense when scheduling before register
9108 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
9109 @option{-O2} or higher. Usage of this option can improve the
9110 generated code and decrease its size by preventing register pressure
9111 increase above the number of available hard registers and subsequent
9112 spills in register allocation.
9114 @item -fsched-spec-load
9115 @opindex fsched-spec-load
9116 Allow speculative motion of some load instructions. This only makes
9117 sense when scheduling before register allocation, i.e.@: with
9118 @option{-fschedule-insns} or at @option{-O2} or higher.
9120 @item -fsched-spec-load-dangerous
9121 @opindex fsched-spec-load-dangerous
9122 Allow speculative motion of more load instructions. This only makes
9123 sense when scheduling before register allocation, i.e.@: with
9124 @option{-fschedule-insns} or at @option{-O2} or higher.
9126 @item -fsched-stalled-insns
9127 @itemx -fsched-stalled-insns=@var{n}
9128 @opindex fsched-stalled-insns
9129 Define how many insns (if any) can be moved prematurely from the queue
9130 of stalled insns into the ready list during the second scheduling pass.
9131 @option{-fno-sched-stalled-insns} means that no insns are moved
9132 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
9133 on how many queued insns can be moved prematurely.
9134 @option{-fsched-stalled-insns} without a value is equivalent to
9135 @option{-fsched-stalled-insns=1}.
9137 @item -fsched-stalled-insns-dep
9138 @itemx -fsched-stalled-insns-dep=@var{n}
9139 @opindex fsched-stalled-insns-dep
9140 Define how many insn groups (cycles) are examined for a dependency
9141 on a stalled insn that is a candidate for premature removal from the queue
9142 of stalled insns. This has an effect only during the second scheduling pass,
9143 and only if @option{-fsched-stalled-insns} is used.
9144 @option{-fno-sched-stalled-insns-dep} is equivalent to
9145 @option{-fsched-stalled-insns-dep=0}.
9146 @option{-fsched-stalled-insns-dep} without a value is equivalent to
9147 @option{-fsched-stalled-insns-dep=1}.
9149 @item -fsched2-use-superblocks
9150 @opindex fsched2-use-superblocks
9151 When scheduling after register allocation, use superblock scheduling.
9152 This allows motion across basic block boundaries,
9153 resulting in faster schedules. This option is experimental, as not all machine
9154 descriptions used by GCC model the CPU closely enough to avoid unreliable
9155 results from the algorithm.
9157 This only makes sense when scheduling after register allocation, i.e.@: with
9158 @option{-fschedule-insns2} or at @option{-O2} or higher.
9160 @item -fsched-group-heuristic
9161 @opindex fsched-group-heuristic
9162 Enable the group heuristic in the scheduler. This heuristic favors
9163 the instruction that belongs to a schedule group. This is enabled
9164 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
9165 or @option{-fschedule-insns2} or at @option{-O2} or higher.
9167 @item -fsched-critical-path-heuristic
9168 @opindex fsched-critical-path-heuristic
9169 Enable the critical-path heuristic in the scheduler. This heuristic favors
9170 instructions on the critical path. This is enabled by default when
9171 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
9172 or @option{-fschedule-insns2} or at @option{-O2} or higher.
9174 @item -fsched-spec-insn-heuristic
9175 @opindex fsched-spec-insn-heuristic
9176 Enable the speculative instruction heuristic in the scheduler. This
9177 heuristic favors speculative instructions with greater dependency weakness.
9178 This is enabled by default when scheduling is enabled, i.e.@:
9179 with @option{-fschedule-insns} or @option{-fschedule-insns2}
9180 or at @option{-O2} or higher.
9182 @item -fsched-rank-heuristic
9183 @opindex fsched-rank-heuristic
9184 Enable the rank heuristic in the scheduler. This heuristic favors
9185 the instruction belonging to a basic block with greater size or frequency.
9186 This is enabled by default when scheduling is enabled, i.e.@:
9187 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
9188 at @option{-O2} or higher.
9190 @item -fsched-last-insn-heuristic
9191 @opindex fsched-last-insn-heuristic
9192 Enable the last-instruction heuristic in the scheduler. This heuristic
9193 favors the instruction that is less dependent on the last instruction
9194 scheduled. This is enabled by default when scheduling is enabled,
9195 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
9196 at @option{-O2} or higher.
9198 @item -fsched-dep-count-heuristic
9199 @opindex fsched-dep-count-heuristic
9200 Enable the dependent-count heuristic in the scheduler. This heuristic
9201 favors the instruction that has more instructions depending on it.
9202 This is enabled by default when scheduling is enabled, i.e.@:
9203 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
9204 at @option{-O2} or higher.
9206 @item -freschedule-modulo-scheduled-loops
9207 @opindex freschedule-modulo-scheduled-loops
9208 Modulo scheduling is performed before traditional scheduling. If a loop
9209 is modulo scheduled, later scheduling passes may change its schedule.
9210 Use this option to control that behavior.
9212 @item -fselective-scheduling
9213 @opindex fselective-scheduling
9214 Schedule instructions using selective scheduling algorithm. Selective
9215 scheduling runs instead of the first scheduler pass.
9217 @item -fselective-scheduling2
9218 @opindex fselective-scheduling2
9219 Schedule instructions using selective scheduling algorithm. Selective
9220 scheduling runs instead of the second scheduler pass.
9222 @item -fsel-sched-pipelining
9223 @opindex fsel-sched-pipelining
9224 Enable software pipelining of innermost loops during selective scheduling.
9225 This option has no effect unless one of @option{-fselective-scheduling} or
9226 @option{-fselective-scheduling2} is turned on.
9228 @item -fsel-sched-pipelining-outer-loops
9229 @opindex fsel-sched-pipelining-outer-loops
9230 When pipelining loops during selective scheduling, also pipeline outer loops.
9231 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
9233 @item -fsemantic-interposition
9234 @opindex fsemantic-interposition
9235 Some object formats, like ELF, allow interposing of symbols by the
9237 This means that for symbols exported from the DSO, the compiler cannot perform
9238 interprocedural propagation, inlining and other optimizations in anticipation
9239 that the function or variable in question may change. While this feature is
9240 useful, for example, to rewrite memory allocation functions by a debugging
9241 implementation, it is expensive in the terms of code quality.
9242 With @option{-fno-semantic-interposition} the compiler assumes that
9243 if interposition happens for functions the overwriting function will have
9244 precisely the same semantics (and side effects).
9245 Similarly if interposition happens
9246 for variables, the constructor of the variable will be the same. The flag
9247 has no effect for functions explicitly declared inline
9248 (where it is never allowed for interposition to change semantics)
9249 and for symbols explicitly declared weak.
9252 @opindex fshrink-wrap
9253 Emit function prologues only before parts of the function that need it,
9254 rather than at the top of the function. This flag is enabled by default at
9255 @option{-O} and higher.
9257 @item -fshrink-wrap-separate
9258 @opindex fshrink-wrap-separate
9259 Shrink-wrap separate parts of the prologue and epilogue separately, so that
9260 those parts are only executed when needed.
9261 This option is on by default, but has no effect unless @option{-fshrink-wrap}
9262 is also turned on and the target supports this.
9264 @item -fcaller-saves
9265 @opindex fcaller-saves
9266 Enable allocation of values to registers that are clobbered by
9267 function calls, by emitting extra instructions to save and restore the
9268 registers around such calls. Such allocation is done only when it
9269 seems to result in better code.
9271 This option is always enabled by default on certain machines, usually
9272 those which have no call-preserved registers to use instead.
9274 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9276 @item -fcombine-stack-adjustments
9277 @opindex fcombine-stack-adjustments
9278 Tracks stack adjustments (pushes and pops) and stack memory references
9279 and then tries to find ways to combine them.
9281 Enabled by default at @option{-O1} and higher.
9285 Use caller save registers for allocation if those registers are not used by
9286 any called function. In that case it is not necessary to save and restore
9287 them around calls. This is only possible if called functions are part of
9288 same compilation unit as current function and they are compiled before it.
9290 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}, however the option
9291 is disabled if generated code will be instrumented for profiling
9292 (@option{-p}, or @option{-pg}) or if callee's register usage cannot be known
9293 exactly (this happens on targets that do not expose prologues
9294 and epilogues in RTL).
9296 @item -fconserve-stack
9297 @opindex fconserve-stack
9298 Attempt to minimize stack usage. The compiler attempts to use less
9299 stack space, even if that makes the program slower. This option
9300 implies setting the @option{large-stack-frame} parameter to 100
9301 and the @option{large-stack-frame-growth} parameter to 400.
9303 @item -ftree-reassoc
9304 @opindex ftree-reassoc
9305 Perform reassociation on trees. This flag is enabled by default
9306 at @option{-O} and higher.
9308 @item -fcode-hoisting
9309 @opindex fcode-hoisting
9310 Perform code hoisting. Code hoisting tries to move the
9311 evaluation of expressions executed on all paths to the function exit
9312 as early as possible. This is especially useful as a code size
9313 optimization, but it often helps for code speed as well.
9314 This flag is enabled by default at @option{-O2} and higher.
9318 Perform partial redundancy elimination (PRE) on trees. This flag is
9319 enabled by default at @option{-O2} and @option{-O3}.
9321 @item -ftree-partial-pre
9322 @opindex ftree-partial-pre
9323 Make partial redundancy elimination (PRE) more aggressive. This flag is
9324 enabled by default at @option{-O3}.
9326 @item -ftree-forwprop
9327 @opindex ftree-forwprop
9328 Perform forward propagation on trees. This flag is enabled by default
9329 at @option{-O} and higher.
9333 Perform full redundancy elimination (FRE) on trees. The difference
9334 between FRE and PRE is that FRE only considers expressions
9335 that are computed on all paths leading to the redundant computation.
9336 This analysis is faster than PRE, though it exposes fewer redundancies.
9337 This flag is enabled by default at @option{-O} and higher.
9339 @item -ftree-phiprop
9340 @opindex ftree-phiprop
9341 Perform hoisting of loads from conditional pointers on trees. This
9342 pass is enabled by default at @option{-O} and higher.
9344 @item -fhoist-adjacent-loads
9345 @opindex fhoist-adjacent-loads
9346 Speculatively hoist loads from both branches of an if-then-else if the
9347 loads are from adjacent locations in the same structure and the target
9348 architecture has a conditional move instruction. This flag is enabled
9349 by default at @option{-O2} and higher.
9351 @item -ftree-copy-prop
9352 @opindex ftree-copy-prop
9353 Perform copy propagation on trees. This pass eliminates unnecessary
9354 copy operations. This flag is enabled by default at @option{-O} and
9357 @item -fipa-pure-const
9358 @opindex fipa-pure-const
9359 Discover which functions are pure or constant.
9360 Enabled by default at @option{-O} and higher.
9362 @item -fipa-reference
9363 @opindex fipa-reference
9364 Discover which static variables do not escape the
9366 Enabled by default at @option{-O} and higher.
9368 @item -fipa-reference-addressable
9369 @opindex fipa-reference-addressable
9370 Discover read-only, write-only and non-addressable static variables.
9371 Enabled by default at @option{-O} and higher.
9373 @item -fipa-stack-alignment
9374 @opindex fipa-stack-alignment
9375 Reduce stack alignment on call sites if possible.
9380 Perform interprocedural pointer analysis and interprocedural modification
9381 and reference analysis. This option can cause excessive memory and
9382 compile-time usage on large compilation units. It is not enabled by
9383 default at any optimization level.
9386 @opindex fipa-profile
9387 Perform interprocedural profile propagation. The functions called only from
9388 cold functions are marked as cold. Also functions executed once (such as
9389 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
9390 functions and loop less parts of functions executed once are then optimized for
9392 Enabled by default at @option{-O} and higher.
9396 Perform interprocedural constant propagation.
9397 This optimization analyzes the program to determine when values passed
9398 to functions are constants and then optimizes accordingly.
9399 This optimization can substantially increase performance
9400 if the application has constants passed to functions.
9401 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
9402 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9404 @item -fipa-cp-clone
9405 @opindex fipa-cp-clone
9406 Perform function cloning to make interprocedural constant propagation stronger.
9407 When enabled, interprocedural constant propagation performs function cloning
9408 when externally visible function can be called with constant arguments.
9409 Because this optimization can create multiple copies of functions,
9410 it may significantly increase code size
9411 (see @option{--param ipcp-unit-growth=@var{value}}).
9412 This flag is enabled by default at @option{-O3}.
9413 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9416 @opindex fipa-bit-cp
9417 When enabled, perform interprocedural bitwise constant
9418 propagation. This flag is enabled by default at @option{-O2} and
9419 by @option{-fprofile-use} and @option{-fauto-profile}.
9420 It requires that @option{-fipa-cp} is enabled.
9424 When enabled, perform interprocedural propagation of value
9425 ranges. This flag is enabled by default at @option{-O2}. It requires
9426 that @option{-fipa-cp} is enabled.
9430 Perform Identical Code Folding for functions and read-only variables.
9431 The optimization reduces code size and may disturb unwind stacks by replacing
9432 a function by equivalent one with a different name. The optimization works
9433 more effectively with link-time optimization enabled.
9435 Although the behavior is similar to the Gold Linker's ICF optimization, GCC ICF
9436 works on different levels and thus the optimizations are not same - there are
9437 equivalences that are found only by GCC and equivalences found only by Gold.
9439 This flag is enabled by default at @option{-O2} and @option{-Os}.
9441 @item -flive-patching=@var{level}
9442 @opindex flive-patching
9443 Control GCC's optimizations to produce output suitable for live-patching.
9445 If the compiler's optimization uses a function's body or information extracted
9446 from its body to optimize/change another function, the latter is called an
9447 impacted function of the former. If a function is patched, its impacted
9448 functions should be patched too.
9450 The impacted functions are determined by the compiler's interprocedural
9451 optimizations. For example, a caller is impacted when inlining a function
9453 cloning a function and changing its caller to call this new clone,
9454 or extracting a function's pureness/constness information to optimize
9455 its direct or indirect callers, etc.
9457 Usually, the more IPA optimizations enabled, the larger the number of
9458 impacted functions for each function. In order to control the number of
9459 impacted functions and more easily compute the list of impacted function,
9460 IPA optimizations can be partially enabled at two different levels.
9462 The @var{level} argument should be one of the following:
9468 Only enable inlining and cloning optimizations, which includes inlining,
9469 cloning, interprocedural scalar replacement of aggregates and partial inlining.
9470 As a result, when patching a function, all its callers and its clones'
9471 callers are impacted, therefore need to be patched as well.
9473 @option{-flive-patching=inline-clone} disables the following optimization flags:
9474 @gccoptlist{-fwhole-program -fipa-pta -fipa-reference -fipa-ra @gol
9475 -fipa-icf -fipa-icf-functions -fipa-icf-variables @gol
9476 -fipa-bit-cp -fipa-vrp -fipa-pure-const -fipa-reference-addressable @gol
9477 -fipa-stack-alignment}
9479 @item inline-only-static
9481 Only enable inlining of static functions.
9482 As a result, when patching a static function, all its callers are impacted
9483 and so need to be patched as well.
9485 In addition to all the flags that @option{-flive-patching=inline-clone}
9487 @option{-flive-patching=inline-only-static} disables the following additional
9489 @gccoptlist{-fipa-cp-clone -fipa-sra -fpartial-inlining -fipa-cp}
9493 When @option{-flive-patching} is specified without any value, the default value
9494 is @var{inline-clone}.
9496 This flag is disabled by default.
9498 Note that @option{-flive-patching} is not supported with link-time optimization
9501 @item -fisolate-erroneous-paths-dereference
9502 @opindex fisolate-erroneous-paths-dereference
9503 Detect paths that trigger erroneous or undefined behavior due to
9504 dereferencing a null pointer. Isolate those paths from the main control
9505 flow and turn the statement with erroneous or undefined behavior into a trap.
9506 This flag is enabled by default at @option{-O2} and higher and depends on
9507 @option{-fdelete-null-pointer-checks} also being enabled.
9509 @item -fisolate-erroneous-paths-attribute
9510 @opindex fisolate-erroneous-paths-attribute
9511 Detect paths that trigger erroneous or undefined behavior due to a null value
9512 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
9513 attribute. Isolate those paths from the main control flow and turn the
9514 statement with erroneous or undefined behavior into a trap. This is not
9515 currently enabled, but may be enabled by @option{-O2} in the future.
9519 Perform forward store motion on trees. This flag is
9520 enabled by default at @option{-O} and higher.
9522 @item -ftree-bit-ccp
9523 @opindex ftree-bit-ccp
9524 Perform sparse conditional bit constant propagation on trees and propagate
9525 pointer alignment information.
9526 This pass only operates on local scalar variables and is enabled by default
9527 at @option{-O1} and higher, except for @option{-Og}.
9528 It requires that @option{-ftree-ccp} is enabled.
9532 Perform sparse conditional constant propagation (CCP) on trees. This
9533 pass only operates on local scalar variables and is enabled by default
9534 at @option{-O} and higher.
9536 @item -fssa-backprop
9537 @opindex fssa-backprop
9538 Propagate information about uses of a value up the definition chain
9539 in order to simplify the definitions. For example, this pass strips
9540 sign operations if the sign of a value never matters. The flag is
9541 enabled by default at @option{-O} and higher.
9544 @opindex fssa-phiopt
9545 Perform pattern matching on SSA PHI nodes to optimize conditional
9546 code. This pass is enabled by default at @option{-O1} and higher,
9547 except for @option{-Og}.
9549 @item -ftree-switch-conversion
9550 @opindex ftree-switch-conversion
9551 Perform conversion of simple initializations in a switch to
9552 initializations from a scalar array. This flag is enabled by default
9553 at @option{-O2} and higher.
9555 @item -ftree-tail-merge
9556 @opindex ftree-tail-merge
9557 Look for identical code sequences. When found, replace one with a jump to the
9558 other. This optimization is known as tail merging or cross jumping. This flag
9559 is enabled by default at @option{-O2} and higher. The compilation time
9561 be limited using @option{max-tail-merge-comparisons} parameter and
9562 @option{max-tail-merge-iterations} parameter.
9566 Perform dead code elimination (DCE) on trees. This flag is enabled by
9567 default at @option{-O} and higher.
9569 @item -ftree-builtin-call-dce
9570 @opindex ftree-builtin-call-dce
9571 Perform conditional dead code elimination (DCE) for calls to built-in functions
9572 that may set @code{errno} but are otherwise free of side effects. This flag is
9573 enabled by default at @option{-O2} and higher if @option{-Os} is not also
9576 @item -ffinite-loops
9577 @opindex ffinite-loops
9578 @opindex fno-finite-loops
9579 Assume that a loop with an exit will eventually take the exit and not loop
9580 indefinitely. This allows the compiler to remove loops that otherwise have
9581 no side-effects, not considering eventual endless looping as such.
9583 This option is enabled by default at @option{-O2}.
9585 @item -ftree-dominator-opts
9586 @opindex ftree-dominator-opts
9587 Perform a variety of simple scalar cleanups (constant/copy
9588 propagation, redundancy elimination, range propagation and expression
9589 simplification) based on a dominator tree traversal. This also
9590 performs jump threading (to reduce jumps to jumps). This flag is
9591 enabled by default at @option{-O} and higher.
9595 Perform dead store elimination (DSE) on trees. A dead store is a store into
9596 a memory location that is later overwritten by another store without
9597 any intervening loads. In this case the earlier store can be deleted. This
9598 flag is enabled by default at @option{-O} and higher.
9602 Perform loop header copying on trees. This is beneficial since it increases
9603 effectiveness of code motion optimizations. It also saves one jump. This flag
9604 is enabled by default at @option{-O} and higher. It is not enabled
9605 for @option{-Os}, since it usually increases code size.
9607 @item -ftree-loop-optimize
9608 @opindex ftree-loop-optimize
9609 Perform loop optimizations on trees. This flag is enabled by default
9610 at @option{-O} and higher.
9612 @item -ftree-loop-linear
9613 @itemx -floop-strip-mine
9615 @opindex ftree-loop-linear
9616 @opindex floop-strip-mine
9617 @opindex floop-block
9618 Perform loop nest optimizations. Same as
9619 @option{-floop-nest-optimize}. To use this code transformation, GCC has
9620 to be configured with @option{--with-isl} to enable the Graphite loop
9621 transformation infrastructure.
9623 @item -fgraphite-identity
9624 @opindex fgraphite-identity
9625 Enable the identity transformation for graphite. For every SCoP we generate
9626 the polyhedral representation and transform it back to gimple. Using
9627 @option{-fgraphite-identity} we can check the costs or benefits of the
9628 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
9629 are also performed by the code generator isl, like index splitting and
9630 dead code elimination in loops.
9632 @item -floop-nest-optimize
9633 @opindex floop-nest-optimize
9634 Enable the isl based loop nest optimizer. This is a generic loop nest
9635 optimizer based on the Pluto optimization algorithms. It calculates a loop
9636 structure optimized for data-locality and parallelism. This option
9639 @item -floop-parallelize-all
9640 @opindex floop-parallelize-all
9641 Use the Graphite data dependence analysis to identify loops that can
9642 be parallelized. Parallelize all the loops that can be analyzed to
9643 not contain loop carried dependences without checking that it is
9644 profitable to parallelize the loops.
9646 @item -ftree-coalesce-vars
9647 @opindex ftree-coalesce-vars
9648 While transforming the program out of the SSA representation, attempt to
9649 reduce copying by coalescing versions of different user-defined
9650 variables, instead of just compiler temporaries. This may severely
9651 limit the ability to debug an optimized program compiled with
9652 @option{-fno-var-tracking-assignments}. In the negated form, this flag
9653 prevents SSA coalescing of user variables. This option is enabled by
9654 default if optimization is enabled, and it does very little otherwise.
9656 @item -ftree-loop-if-convert
9657 @opindex ftree-loop-if-convert
9658 Attempt to transform conditional jumps in the innermost loops to
9659 branch-less equivalents. The intent is to remove control-flow from
9660 the innermost loops in order to improve the ability of the
9661 vectorization pass to handle these loops. This is enabled by default
9662 if vectorization is enabled.
9664 @item -ftree-loop-distribution
9665 @opindex ftree-loop-distribution
9666 Perform loop distribution. This flag can improve cache performance on
9667 big loop bodies and allow further loop optimizations, like
9668 parallelization or vectorization, to take place. For example, the loop
9684 This flag is enabled by default at @option{-O3}.
9685 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9687 @item -ftree-loop-distribute-patterns
9688 @opindex ftree-loop-distribute-patterns
9689 Perform loop distribution of patterns that can be code generated with
9690 calls to a library. This flag is enabled by default at @option{-O3}, and
9691 by @option{-fprofile-use} and @option{-fauto-profile}.
9693 This pass distributes the initialization loops and generates a call to
9694 memset zero. For example, the loop
9710 and the initialization loop is transformed into a call to memset zero.
9711 This flag is enabled by default at @option{-O3}.
9712 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9714 @item -floop-interchange
9715 @opindex floop-interchange
9716 Perform loop interchange outside of graphite. This flag can improve cache
9717 performance on loop nest and allow further loop optimizations, like
9718 vectorization, to take place. For example, the loop
9720 for (int i = 0; i < N; i++)
9721 for (int j = 0; j < N; j++)
9722 for (int k = 0; k < N; k++)
9723 c[i][j] = c[i][j] + a[i][k]*b[k][j];
9727 for (int i = 0; i < N; i++)
9728 for (int k = 0; k < N; k++)
9729 for (int j = 0; j < N; j++)
9730 c[i][j] = c[i][j] + a[i][k]*b[k][j];
9732 This flag is enabled by default at @option{-O3}.
9733 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9735 @item -floop-unroll-and-jam
9736 @opindex floop-unroll-and-jam
9737 Apply unroll and jam transformations on feasible loops. In a loop
9738 nest this unrolls the outer loop by some factor and fuses the resulting
9739 multiple inner loops. This flag is enabled by default at @option{-O3}.
9740 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9742 @item -ftree-loop-im
9743 @opindex ftree-loop-im
9744 Perform loop invariant motion on trees. This pass moves only invariants that
9745 are hard to handle at RTL level (function calls, operations that expand to
9746 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
9747 operands of conditions that are invariant out of the loop, so that we can use
9748 just trivial invariantness analysis in loop unswitching. The pass also includes
9751 @item -ftree-loop-ivcanon
9752 @opindex ftree-loop-ivcanon
9753 Create a canonical counter for number of iterations in loops for which
9754 determining number of iterations requires complicated analysis. Later
9755 optimizations then may determine the number easily. Useful especially
9756 in connection with unrolling.
9758 @item -ftree-scev-cprop
9759 @opindex ftree-scev-cprop
9760 Perform final value replacement. If a variable is modified in a loop
9761 in such a way that its value when exiting the loop can be determined using
9762 only its initial value and the number of loop iterations, replace uses of
9763 the final value by such a computation, provided it is sufficiently cheap.
9764 This reduces data dependencies and may allow further simplifications.
9765 Enabled by default at @option{-O} and higher.
9769 Perform induction variable optimizations (strength reduction, induction
9770 variable merging and induction variable elimination) on trees.
9772 @item -ftree-parallelize-loops=n
9773 @opindex ftree-parallelize-loops
9774 Parallelize loops, i.e., split their iteration space to run in n threads.
9775 This is only possible for loops whose iterations are independent
9776 and can be arbitrarily reordered. The optimization is only
9777 profitable on multiprocessor machines, for loops that are CPU-intensive,
9778 rather than constrained e.g.@: by memory bandwidth. This option
9779 implies @option{-pthread}, and thus is only supported on targets
9780 that have support for @option{-pthread}.
9784 Perform function-local points-to analysis on trees. This flag is
9785 enabled by default at @option{-O1} and higher, except for @option{-Og}.
9789 Perform scalar replacement of aggregates. This pass replaces structure
9790 references with scalars to prevent committing structures to memory too
9791 early. This flag is enabled by default at @option{-O1} and higher,
9792 except for @option{-Og}.
9794 @item -fstore-merging
9795 @opindex fstore-merging
9796 Perform merging of narrow stores to consecutive memory addresses. This pass
9797 merges contiguous stores of immediate values narrower than a word into fewer
9798 wider stores to reduce the number of instructions. This is enabled by default
9799 at @option{-O2} and higher as well as @option{-Os}.
9803 Perform temporary expression replacement during the SSA->normal phase. Single
9804 use/single def temporaries are replaced at their use location with their
9805 defining expression. This results in non-GIMPLE code, but gives the expanders
9806 much more complex trees to work on resulting in better RTL generation. This is
9807 enabled by default at @option{-O} and higher.
9811 Perform straight-line strength reduction on trees. This recognizes related
9812 expressions involving multiplications and replaces them by less expensive
9813 calculations when possible. This is enabled by default at @option{-O} and
9816 @item -ftree-vectorize
9817 @opindex ftree-vectorize
9818 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
9819 and @option{-ftree-slp-vectorize} if not explicitly specified.
9821 @item -ftree-loop-vectorize
9822 @opindex ftree-loop-vectorize
9823 Perform loop vectorization on trees. This flag is enabled by default at
9824 @option{-O3} and by @option{-ftree-vectorize}, @option{-fprofile-use},
9825 and @option{-fauto-profile}.
9827 @item -ftree-slp-vectorize
9828 @opindex ftree-slp-vectorize
9829 Perform basic block vectorization on trees. This flag is enabled by default at
9830 @option{-O3} and by @option{-ftree-vectorize}, @option{-fprofile-use},
9831 and @option{-fauto-profile}.
9833 @item -fvect-cost-model=@var{model}
9834 @opindex fvect-cost-model
9835 Alter the cost model used for vectorization. The @var{model} argument
9836 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
9837 With the @samp{unlimited} model the vectorized code-path is assumed
9838 to be profitable while with the @samp{dynamic} model a runtime check
9839 guards the vectorized code-path to enable it only for iteration
9840 counts that will likely execute faster than when executing the original
9841 scalar loop. The @samp{cheap} model disables vectorization of
9842 loops where doing so would be cost prohibitive for example due to
9843 required runtime checks for data dependence or alignment but otherwise
9844 is equal to the @samp{dynamic} model.
9845 The default cost model depends on other optimization flags and is
9846 either @samp{dynamic} or @samp{cheap}.
9848 @item -fsimd-cost-model=@var{model}
9849 @opindex fsimd-cost-model
9850 Alter the cost model used for vectorization of loops marked with the OpenMP
9851 simd directive. The @var{model} argument should be one of
9852 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
9853 have the same meaning as described in @option{-fvect-cost-model} and by
9854 default a cost model defined with @option{-fvect-cost-model} is used.
9858 Perform Value Range Propagation on trees. This is similar to the
9859 constant propagation pass, but instead of values, ranges of values are
9860 propagated. This allows the optimizers to remove unnecessary range
9861 checks like array bound checks and null pointer checks. This is
9862 enabled by default at @option{-O2} and higher. Null pointer check
9863 elimination is only done if @option{-fdelete-null-pointer-checks} is
9867 @opindex fsplit-paths
9868 Split paths leading to loop backedges. This can improve dead code
9869 elimination and common subexpression elimination. This is enabled by
9870 default at @option{-O3} and above.
9872 @item -fsplit-ivs-in-unroller
9873 @opindex fsplit-ivs-in-unroller
9874 Enables expression of values of induction variables in later iterations
9875 of the unrolled loop using the value in the first iteration. This breaks
9876 long dependency chains, thus improving efficiency of the scheduling passes.
9878 A combination of @option{-fweb} and CSE is often sufficient to obtain the
9879 same effect. However, that is not reliable in cases where the loop body
9880 is more complicated than a single basic block. It also does not work at all
9881 on some architectures due to restrictions in the CSE pass.
9883 This optimization is enabled by default.
9885 @item -fvariable-expansion-in-unroller
9886 @opindex fvariable-expansion-in-unroller
9887 With this option, the compiler creates multiple copies of some
9888 local variables when unrolling a loop, which can result in superior code.
9890 This optimization is enabled by default for PowerPC targets, but disabled
9891 by default otherwise.
9893 @item -fpartial-inlining
9894 @opindex fpartial-inlining
9895 Inline parts of functions. This option has any effect only
9896 when inlining itself is turned on by the @option{-finline-functions}
9897 or @option{-finline-small-functions} options.
9899 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9901 @item -fpredictive-commoning
9902 @opindex fpredictive-commoning
9903 Perform predictive commoning optimization, i.e., reusing computations
9904 (especially memory loads and stores) performed in previous
9905 iterations of loops.
9907 This option is enabled at level @option{-O3}.
9908 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9910 @item -fprefetch-loop-arrays
9911 @opindex fprefetch-loop-arrays
9912 If supported by the target machine, generate instructions to prefetch
9913 memory to improve the performance of loops that access large arrays.
9915 This option may generate better or worse code; results are highly
9916 dependent on the structure of loops within the source code.
9918 Disabled at level @option{-Os}.
9920 @item -fno-printf-return-value
9921 @opindex fno-printf-return-value
9922 @opindex fprintf-return-value
9923 Do not substitute constants for known return value of formatted output
9924 functions such as @code{sprintf}, @code{snprintf}, @code{vsprintf}, and
9925 @code{vsnprintf} (but not @code{printf} of @code{fprintf}). This
9926 transformation allows GCC to optimize or even eliminate branches based
9927 on the known return value of these functions called with arguments that
9928 are either constant, or whose values are known to be in a range that
9929 makes determining the exact return value possible. For example, when
9930 @option{-fprintf-return-value} is in effect, both the branch and the
9931 body of the @code{if} statement (but not the call to @code{snprint})
9932 can be optimized away when @code{i} is a 32-bit or smaller integer
9933 because the return value is guaranteed to be at most 8.
9937 if (snprintf (buf, "%08x", i) >= sizeof buf)
9941 The @option{-fprintf-return-value} option relies on other optimizations
9942 and yields best results with @option{-O2} and above. It works in tandem
9943 with the @option{-Wformat-overflow} and @option{-Wformat-truncation}
9944 options. The @option{-fprintf-return-value} option is enabled by default.
9947 @itemx -fno-peephole2
9948 @opindex fno-peephole
9950 @opindex fno-peephole2
9952 Disable any machine-specific peephole optimizations. The difference
9953 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
9954 are implemented in the compiler; some targets use one, some use the
9955 other, a few use both.
9957 @option{-fpeephole} is enabled by default.
9958 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9960 @item -fno-guess-branch-probability
9961 @opindex fno-guess-branch-probability
9962 @opindex fguess-branch-probability
9963 Do not guess branch probabilities using heuristics.
9965 GCC uses heuristics to guess branch probabilities if they are
9966 not provided by profiling feedback (@option{-fprofile-arcs}). These
9967 heuristics are based on the control flow graph. If some branch probabilities
9968 are specified by @code{__builtin_expect}, then the heuristics are
9969 used to guess branch probabilities for the rest of the control flow graph,
9970 taking the @code{__builtin_expect} info into account. The interactions
9971 between the heuristics and @code{__builtin_expect} can be complex, and in
9972 some cases, it may be useful to disable the heuristics so that the effects
9973 of @code{__builtin_expect} are easier to understand.
9975 It is also possible to specify expected probability of the expression
9976 with @code{__builtin_expect_with_probability} built-in function.
9978 The default is @option{-fguess-branch-probability} at levels
9979 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9981 @item -freorder-blocks
9982 @opindex freorder-blocks
9983 Reorder basic blocks in the compiled function in order to reduce number of
9984 taken branches and improve code locality.
9986 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9988 @item -freorder-blocks-algorithm=@var{algorithm}
9989 @opindex freorder-blocks-algorithm
9990 Use the specified algorithm for basic block reordering. The
9991 @var{algorithm} argument can be @samp{simple}, which does not increase
9992 code size (except sometimes due to secondary effects like alignment),
9993 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
9994 put all often executed code together, minimizing the number of branches
9995 executed by making extra copies of code.
9997 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
9998 @samp{stc} at levels @option{-O2}, @option{-O3}.
10000 @item -freorder-blocks-and-partition
10001 @opindex freorder-blocks-and-partition
10002 In addition to reordering basic blocks in the compiled function, in order
10003 to reduce number of taken branches, partitions hot and cold basic blocks
10004 into separate sections of the assembly and @file{.o} files, to improve
10005 paging and cache locality performance.
10007 This optimization is automatically turned off in the presence of
10008 exception handling or unwind tables (on targets using setjump/longjump or target specific scheme), for linkonce sections, for functions with a user-defined
10009 section attribute and on any architecture that does not support named
10010 sections. When @option{-fsplit-stack} is used this option is not
10011 enabled by default (to avoid linker errors), but may be enabled
10012 explicitly (if using a working linker).
10014 Enabled for x86 at levels @option{-O2}, @option{-O3}, @option{-Os}.
10016 @item -freorder-functions
10017 @opindex freorder-functions
10018 Reorder functions in the object file in order to
10019 improve code locality. This is implemented by using special
10020 subsections @code{.text.hot} for most frequently executed functions and
10021 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
10022 the linker so object file format must support named sections and linker must
10023 place them in a reasonable way.
10025 This option isn't effective unless you either provide profile feedback
10026 (see @option{-fprofile-arcs} for details) or manually annotate functions with
10027 @code{hot} or @code{cold} attributes (@pxref{Common Function Attributes}).
10029 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
10031 @item -fstrict-aliasing
10032 @opindex fstrict-aliasing
10033 Allow the compiler to assume the strictest aliasing rules applicable to
10034 the language being compiled. For C (and C++), this activates
10035 optimizations based on the type of expressions. In particular, an
10036 object of one type is assumed never to reside at the same address as an
10037 object of a different type, unless the types are almost the same. For
10038 example, an @code{unsigned int} can alias an @code{int}, but not a
10039 @code{void*} or a @code{double}. A character type may alias any other
10042 @anchor{Type-punning}Pay special attention to code like this:
10055 The practice of reading from a different union member than the one most
10056 recently written to (called ``type-punning'') is common. Even with
10057 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
10058 is accessed through the union type. So, the code above works as
10059 expected. @xref{Structures unions enumerations and bit-fields
10060 implementation}. However, this code might not:
10071 Similarly, access by taking the address, casting the resulting pointer
10072 and dereferencing the result has undefined behavior, even if the cast
10073 uses a union type, e.g.:
10077 return ((union a_union *) &d)->i;
10081 The @option{-fstrict-aliasing} option is enabled at levels
10082 @option{-O2}, @option{-O3}, @option{-Os}.
10084 @item -falign-functions
10085 @itemx -falign-functions=@var{n}
10086 @itemx -falign-functions=@var{n}:@var{m}
10087 @itemx -falign-functions=@var{n}:@var{m}:@var{n2}
10088 @itemx -falign-functions=@var{n}:@var{m}:@var{n2}:@var{m2}
10089 @opindex falign-functions
10090 Align the start of functions to the next power-of-two greater than
10091 @var{n}, skipping up to @var{m}-1 bytes. This ensures that at least
10092 the first @var{m} bytes of the function can be fetched by the CPU
10093 without crossing an @var{n}-byte alignment boundary.
10095 If @var{m} is not specified, it defaults to @var{n}.
10097 Examples: @option{-falign-functions=32} aligns functions to the next
10098 32-byte boundary, @option{-falign-functions=24} aligns to the next
10099 32-byte boundary only if this can be done by skipping 23 bytes or less,
10100 @option{-falign-functions=32:7} aligns to the next
10101 32-byte boundary only if this can be done by skipping 6 bytes or less.
10103 The second pair of @var{n2}:@var{m2} values allows you to specify
10104 a secondary alignment: @option{-falign-functions=64:7:32:3} aligns to
10105 the next 64-byte boundary if this can be done by skipping 6 bytes or less,
10106 otherwise aligns to the next 32-byte boundary if this can be done
10107 by skipping 2 bytes or less.
10108 If @var{m2} is not specified, it defaults to @var{n2}.
10110 Some assemblers only support this flag when @var{n} is a power of two;
10111 in that case, it is rounded up.
10113 @option{-fno-align-functions} and @option{-falign-functions=1} are
10114 equivalent and mean that functions are not aligned.
10116 If @var{n} is not specified or is zero, use a machine-dependent default.
10117 The maximum allowed @var{n} option value is 65536.
10119 Enabled at levels @option{-O2}, @option{-O3}.
10121 @item -flimit-function-alignment
10122 If this option is enabled, the compiler tries to avoid unnecessarily
10123 overaligning functions. It attempts to instruct the assembler to align
10124 by the amount specified by @option{-falign-functions}, but not to
10125 skip more bytes than the size of the function.
10127 @item -falign-labels
10128 @itemx -falign-labels=@var{n}
10129 @itemx -falign-labels=@var{n}:@var{m}
10130 @itemx -falign-labels=@var{n}:@var{m}:@var{n2}
10131 @itemx -falign-labels=@var{n}:@var{m}:@var{n2}:@var{m2}
10132 @opindex falign-labels
10133 Align all branch targets to a power-of-two boundary.
10135 Parameters of this option are analogous to the @option{-falign-functions} option.
10136 @option{-fno-align-labels} and @option{-falign-labels=1} are
10137 equivalent and mean that labels are not aligned.
10139 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
10140 are greater than this value, then their values are used instead.
10142 If @var{n} is not specified or is zero, use a machine-dependent default
10143 which is very likely to be @samp{1}, meaning no alignment.
10144 The maximum allowed @var{n} option value is 65536.
10146 Enabled at levels @option{-O2}, @option{-O3}.
10148 @item -falign-loops
10149 @itemx -falign-loops=@var{n}
10150 @itemx -falign-loops=@var{n}:@var{m}
10151 @itemx -falign-loops=@var{n}:@var{m}:@var{n2}
10152 @itemx -falign-loops=@var{n}:@var{m}:@var{n2}:@var{m2}
10153 @opindex falign-loops
10154 Align loops to a power-of-two boundary. If the loops are executed
10155 many times, this makes up for any execution of the dummy padding
10158 Parameters of this option are analogous to the @option{-falign-functions} option.
10159 @option{-fno-align-loops} and @option{-falign-loops=1} are
10160 equivalent and mean that loops are not aligned.
10161 The maximum allowed @var{n} option value is 65536.
10163 If @var{n} is not specified or is zero, use a machine-dependent default.
10165 Enabled at levels @option{-O2}, @option{-O3}.
10167 @item -falign-jumps
10168 @itemx -falign-jumps=@var{n}
10169 @itemx -falign-jumps=@var{n}:@var{m}
10170 @itemx -falign-jumps=@var{n}:@var{m}:@var{n2}
10171 @itemx -falign-jumps=@var{n}:@var{m}:@var{n2}:@var{m2}
10172 @opindex falign-jumps
10173 Align branch targets to a power-of-two boundary, for branch targets
10174 where the targets can only be reached by jumping. In this case,
10175 no dummy operations need be executed.
10177 Parameters of this option are analogous to the @option{-falign-functions} option.
10178 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
10179 equivalent and mean that loops are not aligned.
10181 If @var{n} is not specified or is zero, use a machine-dependent default.
10182 The maximum allowed @var{n} option value is 65536.
10184 Enabled at levels @option{-O2}, @option{-O3}.
10186 @item -funit-at-a-time
10187 @opindex funit-at-a-time
10188 This option is left for compatibility reasons. @option{-funit-at-a-time}
10189 has no effect, while @option{-fno-unit-at-a-time} implies
10190 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
10192 Enabled by default.
10194 @item -fno-toplevel-reorder
10195 @opindex fno-toplevel-reorder
10196 @opindex ftoplevel-reorder
10197 Do not reorder top-level functions, variables, and @code{asm}
10198 statements. Output them in the same order that they appear in the
10199 input file. When this option is used, unreferenced static variables
10200 are not removed. This option is intended to support existing code
10201 that relies on a particular ordering. For new code, it is better to
10202 use attributes when possible.
10204 @option{-ftoplevel-reorder} is the default at @option{-O1} and higher, and
10205 also at @option{-O0} if @option{-fsection-anchors} is explicitly requested.
10206 Additionally @option{-fno-toplevel-reorder} implies
10207 @option{-fno-section-anchors}.
10211 Constructs webs as commonly used for register allocation purposes and assign
10212 each web individual pseudo register. This allows the register allocation pass
10213 to operate on pseudos directly, but also strengthens several other optimization
10214 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
10215 however, make debugging impossible, since variables no longer stay in a
10218 Enabled by default with @option{-funroll-loops}.
10220 @item -fwhole-program
10221 @opindex fwhole-program
10222 Assume that the current compilation unit represents the whole program being
10223 compiled. All public functions and variables with the exception of @code{main}
10224 and those merged by attribute @code{externally_visible} become static functions
10225 and in effect are optimized more aggressively by interprocedural optimizers.
10227 This option should not be used in combination with @option{-flto}.
10228 Instead relying on a linker plugin should provide safer and more precise
10231 @item -flto[=@var{n}]
10233 This option runs the standard link-time optimizer. When invoked
10234 with source code, it generates GIMPLE (one of GCC's internal
10235 representations) and writes it to special ELF sections in the object
10236 file. When the object files are linked together, all the function
10237 bodies are read from these ELF sections and instantiated as if they
10238 had been part of the same translation unit.
10240 To use the link-time optimizer, @option{-flto} and optimization
10241 options should be specified at compile time and during the final link.
10242 It is recommended that you compile all the files participating in the
10243 same link with the same options and also specify those options at
10248 gcc -c -O2 -flto foo.c
10249 gcc -c -O2 -flto bar.c
10250 gcc -o myprog -flto -O2 foo.o bar.o
10253 The first two invocations to GCC save a bytecode representation
10254 of GIMPLE into special ELF sections inside @file{foo.o} and
10255 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
10256 @file{foo.o} and @file{bar.o}, merges the two files into a single
10257 internal image, and compiles the result as usual. Since both
10258 @file{foo.o} and @file{bar.o} are merged into a single image, this
10259 causes all the interprocedural analyses and optimizations in GCC to
10260 work across the two files as if they were a single one. This means,
10261 for example, that the inliner is able to inline functions in
10262 @file{bar.o} into functions in @file{foo.o} and vice-versa.
10264 Another (simpler) way to enable link-time optimization is:
10267 gcc -o myprog -flto -O2 foo.c bar.c
10270 The above generates bytecode for @file{foo.c} and @file{bar.c},
10271 merges them together into a single GIMPLE representation and optimizes
10272 them as usual to produce @file{myprog}.
10274 The important thing to keep in mind is that to enable link-time
10275 optimizations you need to use the GCC driver to perform the link step.
10276 GCC automatically performs link-time optimization if any of the
10277 objects involved were compiled with the @option{-flto} command-line option.
10278 You can always override
10279 the automatic decision to do link-time optimization
10280 by passing @option{-fno-lto} to the link command.
10282 To make whole program optimization effective, it is necessary to make
10283 certain whole program assumptions. The compiler needs to know
10284 what functions and variables can be accessed by libraries and runtime
10285 outside of the link-time optimized unit. When supported by the linker,
10286 the linker plugin (see @option{-fuse-linker-plugin}) passes information
10287 to the compiler about used and externally visible symbols. When
10288 the linker plugin is not available, @option{-fwhole-program} should be
10289 used to allow the compiler to make these assumptions, which leads
10290 to more aggressive optimization decisions.
10292 When a file is compiled with @option{-flto} without
10293 @option{-fuse-linker-plugin}, the generated object file is larger than
10294 a regular object file because it contains GIMPLE bytecodes and the usual
10295 final code (see @option{-ffat-lto-objects}. This means that
10296 object files with LTO information can be linked as normal object
10297 files; if @option{-fno-lto} is passed to the linker, no
10298 interprocedural optimizations are applied. Note that when
10299 @option{-fno-fat-lto-objects} is enabled the compile stage is faster
10300 but you cannot perform a regular, non-LTO link on them.
10302 When producing the final binary, GCC only
10303 applies link-time optimizations to those files that contain bytecode.
10304 Therefore, you can mix and match object files and libraries with
10305 GIMPLE bytecodes and final object code. GCC automatically selects
10306 which files to optimize in LTO mode and which files to link without
10307 further processing.
10309 Generally, options specified at link time override those
10310 specified at compile time, although in some cases GCC attempts to infer
10311 link-time options from the settings used to compile the input files.
10313 If you do not specify an optimization level option @option{-O} at
10314 link time, then GCC uses the highest optimization level
10315 used when compiling the object files. Note that it is generally
10316 ineffective to specify an optimization level option only at link time and
10317 not at compile time, for two reasons. First, compiling without
10318 optimization suppresses compiler passes that gather information
10319 needed for effective optimization at link time. Second, some early
10320 optimization passes can be performed only at compile time and
10323 There are some code generation flags preserved by GCC when
10324 generating bytecodes, as they need to be used during the final link.
10325 Currently, the following options and their settings are taken from
10326 the first object file that explicitly specifies them:
10327 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
10328 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
10329 and all the @option{-m} target flags.
10331 Certain ABI-changing flags are required to match in all compilation units,
10332 and trying to override this at link time with a conflicting value
10333 is ignored. This includes options such as @option{-freg-struct-return}
10334 and @option{-fpcc-struct-return}.
10336 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
10337 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
10338 are passed through to the link stage and merged conservatively for
10339 conflicting translation units. Specifically
10340 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
10341 precedence; and for example @option{-ffp-contract=off} takes precedence
10342 over @option{-ffp-contract=fast}. You can override them at link time.
10344 If LTO encounters objects with C linkage declared with incompatible
10345 types in separate translation units to be linked together (undefined
10346 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
10347 issued. The behavior is still undefined at run time. Similar
10348 diagnostics may be raised for other languages.
10350 Another feature of LTO is that it is possible to apply interprocedural
10351 optimizations on files written in different languages:
10355 g++ -c -flto bar.cc
10356 gfortran -c -flto baz.f90
10357 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
10360 Notice that the final link is done with @command{g++} to get the C++
10361 runtime libraries and @option{-lgfortran} is added to get the Fortran
10362 runtime libraries. In general, when mixing languages in LTO mode, you
10363 should use the same link command options as when mixing languages in a
10364 regular (non-LTO) compilation.
10366 If object files containing GIMPLE bytecode are stored in a library archive, say
10367 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
10368 are using a linker with plugin support. To create static libraries suitable
10369 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
10370 and @command{ranlib};
10371 to show the symbols of object files with GIMPLE bytecode, use
10372 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
10373 and @command{nm} have been compiled with plugin support. At link time, use the
10374 flag @option{-fuse-linker-plugin} to ensure that the library participates in
10375 the LTO optimization process:
10378 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
10381 With the linker plugin enabled, the linker extracts the needed
10382 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
10383 to make them part of the aggregated GIMPLE image to be optimized.
10385 If you are not using a linker with plugin support and/or do not
10386 enable the linker plugin, then the objects inside @file{libfoo.a}
10387 are extracted and linked as usual, but they do not participate
10388 in the LTO optimization process. In order to make a static library suitable
10389 for both LTO optimization and usual linkage, compile its object files with
10390 @option{-flto} @option{-ffat-lto-objects}.
10392 Link-time optimizations do not require the presence of the whole program to
10393 operate. If the program does not require any symbols to be exported, it is
10394 possible to combine @option{-flto} and @option{-fwhole-program} to allow
10395 the interprocedural optimizers to use more aggressive assumptions which may
10396 lead to improved optimization opportunities.
10397 Use of @option{-fwhole-program} is not needed when linker plugin is
10398 active (see @option{-fuse-linker-plugin}).
10400 The current implementation of LTO makes no
10401 attempt to generate bytecode that is portable between different
10402 types of hosts. The bytecode files are versioned and there is a
10403 strict version check, so bytecode files generated in one version of
10404 GCC do not work with an older or newer version of GCC.
10406 Link-time optimization does not work well with generation of debugging
10407 information on systems other than those using a combination of ELF and
10410 If you specify the optional @var{n}, the optimization and code
10411 generation done at link time is executed in parallel using @var{n}
10412 parallel jobs by utilizing an installed @command{make} program. The
10413 environment variable @env{MAKE} may be used to override the program
10416 You can also specify @option{-flto=jobserver} to use GNU make's
10417 job server mode to determine the number of parallel jobs. This
10418 is useful when the Makefile calling GCC is already executing in parallel.
10419 You must prepend a @samp{+} to the command recipe in the parent Makefile
10420 for this to work. This option likely only works if @env{MAKE} is
10421 GNU make. Even without the option value, GCC tries to automatically
10422 detect a running GNU make's job server.
10424 Use @option{-flto=auto} to use GNU make's job server, if available,
10425 or otherwise fall back to autodetection of the number of CPU threads
10426 present in your system.
10428 @item -flto-partition=@var{alg}
10429 @opindex flto-partition
10430 Specify the partitioning algorithm used by the link-time optimizer.
10431 The value is either @samp{1to1} to specify a partitioning mirroring
10432 the original source files or @samp{balanced} to specify partitioning
10433 into equally sized chunks (whenever possible) or @samp{max} to create
10434 new partition for every symbol where possible. Specifying @samp{none}
10435 as an algorithm disables partitioning and streaming completely.
10436 The default value is @samp{balanced}. While @samp{1to1} can be used
10437 as an workaround for various code ordering issues, the @samp{max}
10438 partitioning is intended for internal testing only.
10439 The value @samp{one} specifies that exactly one partition should be
10440 used while the value @samp{none} bypasses partitioning and executes
10441 the link-time optimization step directly from the WPA phase.
10443 @item -flto-compression-level=@var{n}
10444 @opindex flto-compression-level
10445 This option specifies the level of compression used for intermediate
10446 language written to LTO object files, and is only meaningful in
10447 conjunction with LTO mode (@option{-flto}). Valid
10448 values are 0 (no compression) to 9 (maximum compression). Values
10449 outside this range are clamped to either 0 or 9. If the option is not
10450 given, a default balanced compression setting is used.
10452 @item -fuse-linker-plugin
10453 @opindex fuse-linker-plugin
10454 Enables the use of a linker plugin during link-time optimization. This
10455 option relies on plugin support in the linker, which is available in gold
10456 or in GNU ld 2.21 or newer.
10458 This option enables the extraction of object files with GIMPLE bytecode out
10459 of library archives. This improves the quality of optimization by exposing
10460 more code to the link-time optimizer. This information specifies what
10461 symbols can be accessed externally (by non-LTO object or during dynamic
10462 linking). Resulting code quality improvements on binaries (and shared
10463 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
10464 See @option{-flto} for a description of the effect of this flag and how to
10467 This option is enabled by default when LTO support in GCC is enabled
10468 and GCC was configured for use with
10469 a linker supporting plugins (GNU ld 2.21 or newer or gold).
10471 @item -ffat-lto-objects
10472 @opindex ffat-lto-objects
10473 Fat LTO objects are object files that contain both the intermediate language
10474 and the object code. This makes them usable for both LTO linking and normal
10475 linking. This option is effective only when compiling with @option{-flto}
10476 and is ignored at link time.
10478 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
10479 requires the complete toolchain to be aware of LTO. It requires a linker with
10480 linker plugin support for basic functionality. Additionally,
10481 @command{nm}, @command{ar} and @command{ranlib}
10482 need to support linker plugins to allow a full-featured build environment
10483 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
10484 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
10485 to these tools. With non fat LTO makefiles need to be modified to use them.
10487 Note that modern binutils provide plugin auto-load mechanism.
10488 Installing the linker plugin into @file{$libdir/bfd-plugins} has the same
10489 effect as usage of the command wrappers (@command{gcc-ar}, @command{gcc-nm} and
10490 @command{gcc-ranlib}).
10492 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
10495 @item -fcompare-elim
10496 @opindex fcompare-elim
10497 After register allocation and post-register allocation instruction splitting,
10498 identify arithmetic instructions that compute processor flags similar to a
10499 comparison operation based on that arithmetic. If possible, eliminate the
10500 explicit comparison operation.
10502 This pass only applies to certain targets that cannot explicitly represent
10503 the comparison operation before register allocation is complete.
10505 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
10507 @item -fcprop-registers
10508 @opindex fcprop-registers
10509 After register allocation and post-register allocation instruction splitting,
10510 perform a copy-propagation pass to try to reduce scheduling dependencies
10511 and occasionally eliminate the copy.
10513 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
10515 @item -fprofile-correction
10516 @opindex fprofile-correction
10517 Profiles collected using an instrumented binary for multi-threaded programs may
10518 be inconsistent due to missed counter updates. When this option is specified,
10519 GCC uses heuristics to correct or smooth out such inconsistencies. By
10520 default, GCC emits an error message when an inconsistent profile is detected.
10522 This option is enabled by @option{-fauto-profile}.
10524 @item -fprofile-use
10525 @itemx -fprofile-use=@var{path}
10526 @opindex fprofile-use
10527 Enable profile feedback-directed optimizations,
10528 and the following optimizations, many of which
10529 are generally profitable only with profile feedback available:
10531 @gccoptlist{-fbranch-probabilities -fprofile-values @gol
10532 -funroll-loops -fpeel-loops -ftracer -fvpt @gol
10533 -finline-functions -fipa-cp -fipa-cp-clone -fipa-bit-cp @gol
10534 -fpredictive-commoning -fsplit-loops -funswitch-loops @gol
10535 -fgcse-after-reload -ftree-loop-vectorize -ftree-slp-vectorize @gol
10536 -fvect-cost-model=dynamic -ftree-loop-distribute-patterns @gol
10537 -fprofile-reorder-functions}
10539 Before you can use this option, you must first generate profiling information.
10540 @xref{Instrumentation Options}, for information about the
10541 @option{-fprofile-generate} option.
10543 By default, GCC emits an error message if the feedback profiles do not
10544 match the source code. This error can be turned into a warning by using
10545 @option{-Wno-error=coverage-mismatch}. Note this may result in poorly
10546 optimized code. Additionally, by default, GCC also emits a warning message if
10547 the feedback profiles do not exist (see @option{-Wmissing-profile}).
10549 If @var{path} is specified, GCC looks at the @var{path} to find
10550 the profile feedback data files. See @option{-fprofile-dir}.
10552 @item -fauto-profile
10553 @itemx -fauto-profile=@var{path}
10554 @opindex fauto-profile
10555 Enable sampling-based feedback-directed optimizations,
10556 and the following optimizations,
10557 many of which are generally profitable only with profile feedback available:
10559 @gccoptlist{-fbranch-probabilities -fprofile-values @gol
10560 -funroll-loops -fpeel-loops -ftracer -fvpt @gol
10561 -finline-functions -fipa-cp -fipa-cp-clone -fipa-bit-cp @gol
10562 -fpredictive-commoning -fsplit-loops -funswitch-loops @gol
10563 -fgcse-after-reload -ftree-loop-vectorize -ftree-slp-vectorize @gol
10564 -fvect-cost-model=dynamic -ftree-loop-distribute-patterns @gol
10565 -fprofile-correction}
10567 @var{path} is the name of a file containing AutoFDO profile information.
10568 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
10570 Producing an AutoFDO profile data file requires running your program
10571 with the @command{perf} utility on a supported GNU/Linux target system.
10572 For more information, see @uref{https://perf.wiki.kernel.org/}.
10576 perf record -e br_inst_retired:near_taken -b -o perf.data \
10580 Then use the @command{create_gcov} tool to convert the raw profile data
10581 to a format that can be used by GCC.@ You must also supply the
10582 unstripped binary for your program to this tool.
10583 See @uref{https://github.com/google/autofdo}.
10587 create_gcov --binary=your_program.unstripped --profile=perf.data \
10588 --gcov=profile.afdo
10592 The following options control compiler behavior regarding floating-point
10593 arithmetic. These options trade off between speed and
10594 correctness. All must be specifically enabled.
10597 @item -ffloat-store
10598 @opindex ffloat-store
10599 Do not store floating-point variables in registers, and inhibit other
10600 options that might change whether a floating-point value is taken from a
10601 register or memory.
10603 @cindex floating-point precision
10604 This option prevents undesirable excess precision on machines such as
10605 the 68000 where the floating registers (of the 68881) keep more
10606 precision than a @code{double} is supposed to have. Similarly for the
10607 x86 architecture. For most programs, the excess precision does only
10608 good, but a few programs rely on the precise definition of IEEE floating
10609 point. Use @option{-ffloat-store} for such programs, after modifying
10610 them to store all pertinent intermediate computations into variables.
10612 @item -fexcess-precision=@var{style}
10613 @opindex fexcess-precision
10614 This option allows further control over excess precision on machines
10615 where floating-point operations occur in a format with more precision or
10616 range than the IEEE standard and interchange floating-point types. By
10617 default, @option{-fexcess-precision=fast} is in effect; this means that
10618 operations may be carried out in a wider precision than the types specified
10619 in the source if that would result in faster code, and it is unpredictable
10620 when rounding to the types specified in the source code takes place.
10621 When compiling C, if @option{-fexcess-precision=standard} is specified then
10622 excess precision follows the rules specified in ISO C99; in particular,
10623 both casts and assignments cause values to be rounded to their
10624 semantic types (whereas @option{-ffloat-store} only affects
10625 assignments). This option is enabled by default for C if a strict
10626 conformance option such as @option{-std=c99} is used.
10627 @option{-ffast-math} enables @option{-fexcess-precision=fast} by default
10628 regardless of whether a strict conformance option is used.
10631 @option{-fexcess-precision=standard} is not implemented for languages
10632 other than C. On the x86, it has no effect if @option{-mfpmath=sse}
10633 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
10634 semantics apply without excess precision, and in the latter, rounding
10638 @opindex ffast-math
10639 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
10640 @option{-ffinite-math-only}, @option{-fno-rounding-math},
10641 @option{-fno-signaling-nans}, @option{-fcx-limited-range} and
10642 @option{-fexcess-precision=fast}.
10644 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
10646 This option is not turned on by any @option{-O} option besides
10647 @option{-Ofast} since it can result in incorrect output for programs
10648 that depend on an exact implementation of IEEE or ISO rules/specifications
10649 for math functions. It may, however, yield faster code for programs
10650 that do not require the guarantees of these specifications.
10652 @item -fno-math-errno
10653 @opindex fno-math-errno
10654 @opindex fmath-errno
10655 Do not set @code{errno} after calling math functions that are executed
10656 with a single instruction, e.g., @code{sqrt}. A program that relies on
10657 IEEE exceptions for math error handling may want to use this flag
10658 for speed while maintaining IEEE arithmetic compatibility.
10660 This option is not turned on by any @option{-O} option since
10661 it can result in incorrect output for programs that depend on
10662 an exact implementation of IEEE or ISO rules/specifications for
10663 math functions. It may, however, yield faster code for programs
10664 that do not require the guarantees of these specifications.
10666 The default is @option{-fmath-errno}.
10668 On Darwin systems, the math library never sets @code{errno}. There is
10669 therefore no reason for the compiler to consider the possibility that
10670 it might, and @option{-fno-math-errno} is the default.
10672 @item -funsafe-math-optimizations
10673 @opindex funsafe-math-optimizations
10675 Allow optimizations for floating-point arithmetic that (a) assume
10676 that arguments and results are valid and (b) may violate IEEE or
10677 ANSI standards. When used at link time, it may include libraries
10678 or startup files that change the default FPU control word or other
10679 similar optimizations.
10681 This option is not turned on by any @option{-O} option since
10682 it can result in incorrect output for programs that depend on
10683 an exact implementation of IEEE or ISO rules/specifications for
10684 math functions. It may, however, yield faster code for programs
10685 that do not require the guarantees of these specifications.
10686 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
10687 @option{-fassociative-math} and @option{-freciprocal-math}.
10689 The default is @option{-fno-unsafe-math-optimizations}.
10691 @item -fassociative-math
10692 @opindex fassociative-math
10694 Allow re-association of operands in series of floating-point operations.
10695 This violates the ISO C and C++ language standard by possibly changing
10696 computation result. NOTE: re-ordering may change the sign of zero as
10697 well as ignore NaNs and inhibit or create underflow or overflow (and
10698 thus cannot be used on code that relies on rounding behavior like
10699 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
10700 and thus may not be used when ordered comparisons are required.
10701 This option requires that both @option{-fno-signed-zeros} and
10702 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
10703 much sense with @option{-frounding-math}. For Fortran the option
10704 is automatically enabled when both @option{-fno-signed-zeros} and
10705 @option{-fno-trapping-math} are in effect.
10707 The default is @option{-fno-associative-math}.
10709 @item -freciprocal-math
10710 @opindex freciprocal-math
10712 Allow the reciprocal of a value to be used instead of dividing by
10713 the value if this enables optimizations. For example @code{x / y}
10714 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
10715 is subject to common subexpression elimination. Note that this loses
10716 precision and increases the number of flops operating on the value.
10718 The default is @option{-fno-reciprocal-math}.
10720 @item -ffinite-math-only
10721 @opindex ffinite-math-only
10722 Allow optimizations for floating-point arithmetic that assume
10723 that arguments and results are not NaNs or +-Infs.
10725 This option is not turned on by any @option{-O} option since
10726 it can result in incorrect output for programs that depend on
10727 an exact implementation of IEEE or ISO rules/specifications for
10728 math functions. It may, however, yield faster code for programs
10729 that do not require the guarantees of these specifications.
10731 The default is @option{-fno-finite-math-only}.
10733 @item -fno-signed-zeros
10734 @opindex fno-signed-zeros
10735 @opindex fsigned-zeros
10736 Allow optimizations for floating-point arithmetic that ignore the
10737 signedness of zero. IEEE arithmetic specifies the behavior of
10738 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
10739 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
10740 This option implies that the sign of a zero result isn't significant.
10742 The default is @option{-fsigned-zeros}.
10744 @item -fno-trapping-math
10745 @opindex fno-trapping-math
10746 @opindex ftrapping-math
10747 Compile code assuming that floating-point operations cannot generate
10748 user-visible traps. These traps include division by zero, overflow,
10749 underflow, inexact result and invalid operation. This option requires
10750 that @option{-fno-signaling-nans} be in effect. Setting this option may
10751 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
10753 This option should never be turned on by any @option{-O} option since
10754 it can result in incorrect output for programs that depend on
10755 an exact implementation of IEEE or ISO rules/specifications for
10758 The default is @option{-ftrapping-math}.
10760 @item -frounding-math
10761 @opindex frounding-math
10762 Disable transformations and optimizations that assume default floating-point
10763 rounding behavior. This is round-to-zero for all floating point
10764 to integer conversions, and round-to-nearest for all other arithmetic
10765 truncations. This option should be specified for programs that change
10766 the FP rounding mode dynamically, or that may be executed with a
10767 non-default rounding mode. This option disables constant folding of
10768 floating-point expressions at compile time (which may be affected by
10769 rounding mode) and arithmetic transformations that are unsafe in the
10770 presence of sign-dependent rounding modes.
10772 The default is @option{-fno-rounding-math}.
10774 This option is experimental and does not currently guarantee to
10775 disable all GCC optimizations that are affected by rounding mode.
10776 Future versions of GCC may provide finer control of this setting
10777 using C99's @code{FENV_ACCESS} pragma. This command-line option
10778 will be used to specify the default state for @code{FENV_ACCESS}.
10780 @item -fsignaling-nans
10781 @opindex fsignaling-nans
10782 Compile code assuming that IEEE signaling NaNs may generate user-visible
10783 traps during floating-point operations. Setting this option disables
10784 optimizations that may change the number of exceptions visible with
10785 signaling NaNs. This option implies @option{-ftrapping-math}.
10787 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
10790 The default is @option{-fno-signaling-nans}.
10792 This option is experimental and does not currently guarantee to
10793 disable all GCC optimizations that affect signaling NaN behavior.
10795 @item -fno-fp-int-builtin-inexact
10796 @opindex fno-fp-int-builtin-inexact
10797 @opindex ffp-int-builtin-inexact
10798 Do not allow the built-in functions @code{ceil}, @code{floor},
10799 @code{round} and @code{trunc}, and their @code{float} and @code{long
10800 double} variants, to generate code that raises the ``inexact''
10801 floating-point exception for noninteger arguments. ISO C99 and C11
10802 allow these functions to raise the ``inexact'' exception, but ISO/IEC
10803 TS 18661-1:2014, the C bindings to IEEE 754-2008, does not allow these
10804 functions to do so.
10806 The default is @option{-ffp-int-builtin-inexact}, allowing the
10807 exception to be raised. This option does nothing unless
10808 @option{-ftrapping-math} is in effect.
10810 Even if @option{-fno-fp-int-builtin-inexact} is used, if the functions
10811 generate a call to a library function then the ``inexact'' exception
10812 may be raised if the library implementation does not follow TS 18661.
10814 @item -fsingle-precision-constant
10815 @opindex fsingle-precision-constant
10816 Treat floating-point constants as single precision instead of
10817 implicitly converting them to double-precision constants.
10819 @item -fcx-limited-range
10820 @opindex fcx-limited-range
10821 When enabled, this option states that a range reduction step is not
10822 needed when performing complex division. Also, there is no checking
10823 whether the result of a complex multiplication or division is @code{NaN
10824 + I*NaN}, with an attempt to rescue the situation in that case. The
10825 default is @option{-fno-cx-limited-range}, but is enabled by
10826 @option{-ffast-math}.
10828 This option controls the default setting of the ISO C99
10829 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
10832 @item -fcx-fortran-rules
10833 @opindex fcx-fortran-rules
10834 Complex multiplication and division follow Fortran rules. Range
10835 reduction is done as part of complex division, but there is no checking
10836 whether the result of a complex multiplication or division is @code{NaN
10837 + I*NaN}, with an attempt to rescue the situation in that case.
10839 The default is @option{-fno-cx-fortran-rules}.
10843 The following options control optimizations that may improve
10844 performance, but are not enabled by any @option{-O} options. This
10845 section includes experimental options that may produce broken code.
10848 @item -fbranch-probabilities
10849 @opindex fbranch-probabilities
10850 After running a program compiled with @option{-fprofile-arcs}
10851 (@pxref{Instrumentation Options}),
10852 you can compile it a second time using
10853 @option{-fbranch-probabilities}, to improve optimizations based on
10854 the number of times each branch was taken. When a program
10855 compiled with @option{-fprofile-arcs} exits, it saves arc execution
10856 counts to a file called @file{@var{sourcename}.gcda} for each source
10857 file. The information in this data file is very dependent on the
10858 structure of the generated code, so you must use the same source code
10859 and the same optimization options for both compilations.
10861 With @option{-fbranch-probabilities}, GCC puts a
10862 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
10863 These can be used to improve optimization. Currently, they are only
10864 used in one place: in @file{reorg.c}, instead of guessing which path a
10865 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
10866 exactly determine which path is taken more often.
10868 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10870 @item -fprofile-values
10871 @opindex fprofile-values
10872 If combined with @option{-fprofile-arcs}, it adds code so that some
10873 data about values of expressions in the program is gathered.
10875 With @option{-fbranch-probabilities}, it reads back the data gathered
10876 from profiling values of expressions for usage in optimizations.
10878 Enabled by @option{-fprofile-generate}, @option{-fprofile-use}, and
10879 @option{-fauto-profile}.
10881 @item -fprofile-reorder-functions
10882 @opindex fprofile-reorder-functions
10883 Function reordering based on profile instrumentation collects
10884 first time of execution of a function and orders these functions
10885 in ascending order.
10887 Enabled with @option{-fprofile-use}.
10891 If combined with @option{-fprofile-arcs}, this option instructs the compiler
10892 to add code to gather information about values of expressions.
10894 With @option{-fbranch-probabilities}, it reads back the data gathered
10895 and actually performs the optimizations based on them.
10896 Currently the optimizations include specialization of division operations
10897 using the knowledge about the value of the denominator.
10899 Enabled with @option{-fprofile-use} and @option{-fauto-profile}.
10901 @item -frename-registers
10902 @opindex frename-registers
10903 Attempt to avoid false dependencies in scheduled code by making use
10904 of registers left over after register allocation. This optimization
10905 most benefits processors with lots of registers. Depending on the
10906 debug information format adopted by the target, however, it can
10907 make debugging impossible, since variables no longer stay in
10908 a ``home register''.
10910 Enabled by default with @option{-funroll-loops}.
10912 @item -fschedule-fusion
10913 @opindex fschedule-fusion
10914 Performs a target dependent pass over the instruction stream to schedule
10915 instructions of same type together because target machine can execute them
10916 more efficiently if they are adjacent to each other in the instruction flow.
10918 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
10922 Perform tail duplication to enlarge superblock size. This transformation
10923 simplifies the control flow of the function allowing other optimizations to do
10926 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10928 @item -funroll-loops
10929 @opindex funroll-loops
10930 Unroll loops whose number of iterations can be determined at compile time or
10931 upon entry to the loop. @option{-funroll-loops} implies
10932 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
10933 It also turns on complete loop peeling (i.e.@: complete removal of loops with
10934 a small constant number of iterations). This option makes code larger, and may
10935 or may not make it run faster.
10937 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10939 @item -funroll-all-loops
10940 @opindex funroll-all-loops
10941 Unroll all loops, even if their number of iterations is uncertain when
10942 the loop is entered. This usually makes programs run more slowly.
10943 @option{-funroll-all-loops} implies the same options as
10944 @option{-funroll-loops}.
10947 @opindex fpeel-loops
10948 Peels loops for which there is enough information that they do not
10949 roll much (from profile feedback or static analysis). It also turns on
10950 complete loop peeling (i.e.@: complete removal of loops with small constant
10951 number of iterations).
10953 Enabled by @option{-O3}, @option{-fprofile-use}, and @option{-fauto-profile}.
10955 @item -fmove-loop-invariants
10956 @opindex fmove-loop-invariants
10957 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
10958 at level @option{-O1} and higher, except for @option{-Og}.
10960 @item -fsplit-loops
10961 @opindex fsplit-loops
10962 Split a loop into two if it contains a condition that's always true
10963 for one side of the iteration space and false for the other.
10965 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10967 @item -funswitch-loops
10968 @opindex funswitch-loops
10969 Move branches with loop invariant conditions out of the loop, with duplicates
10970 of the loop on both branches (modified according to result of the condition).
10972 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10974 @item -fversion-loops-for-strides
10975 @opindex fversion-loops-for-strides
10976 If a loop iterates over an array with a variable stride, create another
10977 version of the loop that assumes the stride is always one. For example:
10980 for (int i = 0; i < n; ++i)
10981 x[i * stride] = @dots{};
10988 for (int i = 0; i < n; ++i)
10991 for (int i = 0; i < n; ++i)
10992 x[i * stride] = @dots{};
10995 This is particularly useful for assumed-shape arrays in Fortran where
10996 (for example) it allows better vectorization assuming contiguous accesses.
10997 This flag is enabled by default at @option{-O3}.
10998 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
11000 @item -ffunction-sections
11001 @itemx -fdata-sections
11002 @opindex ffunction-sections
11003 @opindex fdata-sections
11004 Place each function or data item into its own section in the output
11005 file if the target supports arbitrary sections. The name of the
11006 function or the name of the data item determines the section's name
11007 in the output file.
11009 Use these options on systems where the linker can perform optimizations to
11010 improve locality of reference in the instruction space. Most systems using the
11011 ELF object format have linkers with such optimizations. On AIX, the linker
11012 rearranges sections (CSECTs) based on the call graph. The performance impact
11015 Together with a linker garbage collection (linker @option{--gc-sections}
11016 option) these options may lead to smaller statically-linked executables (after
11019 On ELF/DWARF systems these options do not degenerate the quality of the debug
11020 information. There could be issues with other object files/debug info formats.
11022 Only use these options when there are significant benefits from doing so. When
11023 you specify these options, the assembler and linker create larger object and
11024 executable files and are also slower. These options affect code generation.
11025 They prevent optimizations by the compiler and assembler using relative
11026 locations inside a translation unit since the locations are unknown until
11027 link time. An example of such an optimization is relaxing calls to short call
11030 @item -fbranch-target-load-optimize
11031 @opindex fbranch-target-load-optimize
11032 Perform branch target register load optimization before prologue / epilogue
11034 The use of target registers can typically be exposed only during reload,
11035 thus hoisting loads out of loops and doing inter-block scheduling needs
11036 a separate optimization pass.
11038 @item -fbranch-target-load-optimize2
11039 @opindex fbranch-target-load-optimize2
11040 Perform branch target register load optimization after prologue / epilogue
11043 @item -fbtr-bb-exclusive
11044 @opindex fbtr-bb-exclusive
11045 When performing branch target register load optimization, don't reuse
11046 branch target registers within any basic block.
11049 @opindex fstdarg-opt
11050 Optimize the prologue of variadic argument functions with respect to usage of
11053 @item -fsection-anchors
11054 @opindex fsection-anchors
11055 Try to reduce the number of symbolic address calculations by using
11056 shared ``anchor'' symbols to address nearby objects. This transformation
11057 can help to reduce the number of GOT entries and GOT accesses on some
11060 For example, the implementation of the following function @code{foo}:
11063 static int a, b, c;
11064 int foo (void) @{ return a + b + c; @}
11068 usually calculates the addresses of all three variables, but if you
11069 compile it with @option{-fsection-anchors}, it accesses the variables
11070 from a common anchor point instead. The effect is similar to the
11071 following pseudocode (which isn't valid C):
11076 register int *xr = &x;
11077 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
11081 Not all targets support this option.
11083 @item --param @var{name}=@var{value}
11085 In some places, GCC uses various constants to control the amount of
11086 optimization that is done. For example, GCC does not inline functions
11087 that contain more than a certain number of instructions. You can
11088 control some of these constants on the command line using the
11089 @option{--param} option.
11091 The names of specific parameters, and the meaning of the values, are
11092 tied to the internals of the compiler, and are subject to change
11093 without notice in future releases.
11095 In order to get minimal, maximal and default value of a parameter,
11096 one can use @option{--help=param -Q} options.
11098 In each case, the @var{value} is an integer. The allowable choices for
11102 @item predictable-branch-outcome
11103 When branch is predicted to be taken with probability lower than this threshold
11104 (in percent), then it is considered well predictable.
11106 @item max-rtl-if-conversion-insns
11107 RTL if-conversion tries to remove conditional branches around a block and
11108 replace them with conditionally executed instructions. This parameter
11109 gives the maximum number of instructions in a block which should be
11110 considered for if-conversion. The compiler will
11111 also use other heuristics to decide whether if-conversion is likely to be
11114 @item max-rtl-if-conversion-predictable-cost
11115 @itemx max-rtl-if-conversion-unpredictable-cost
11116 RTL if-conversion will try to remove conditional branches around a block
11117 and replace them with conditionally executed instructions. These parameters
11118 give the maximum permissible cost for the sequence that would be generated
11119 by if-conversion depending on whether the branch is statically determined
11120 to be predictable or not. The units for this parameter are the same as
11121 those for the GCC internal seq_cost metric. The compiler will try to
11122 provide a reasonable default for this parameter using the BRANCH_COST
11125 @item max-crossjump-edges
11126 The maximum number of incoming edges to consider for cross-jumping.
11127 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
11128 the number of edges incoming to each block. Increasing values mean
11129 more aggressive optimization, making the compilation time increase with
11130 probably small improvement in executable size.
11132 @item min-crossjump-insns
11133 The minimum number of instructions that must be matched at the end
11134 of two blocks before cross-jumping is performed on them. This
11135 value is ignored in the case where all instructions in the block being
11136 cross-jumped from are matched.
11138 @item max-grow-copy-bb-insns
11139 The maximum code size expansion factor when copying basic blocks
11140 instead of jumping. The expansion is relative to a jump instruction.
11142 @item max-goto-duplication-insns
11143 The maximum number of instructions to duplicate to a block that jumps
11144 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
11145 passes, GCC factors computed gotos early in the compilation process,
11146 and unfactors them as late as possible. Only computed jumps at the
11147 end of a basic blocks with no more than max-goto-duplication-insns are
11150 @item max-delay-slot-insn-search
11151 The maximum number of instructions to consider when looking for an
11152 instruction to fill a delay slot. If more than this arbitrary number of
11153 instructions are searched, the time savings from filling the delay slot
11154 are minimal, so stop searching. Increasing values mean more
11155 aggressive optimization, making the compilation time increase with probably
11156 small improvement in execution time.
11158 @item max-delay-slot-live-search
11159 When trying to fill delay slots, the maximum number of instructions to
11160 consider when searching for a block with valid live register
11161 information. Increasing this arbitrarily chosen value means more
11162 aggressive optimization, increasing the compilation time. This parameter
11163 should be removed when the delay slot code is rewritten to maintain the
11164 control-flow graph.
11166 @item max-gcse-memory
11167 The approximate maximum amount of memory that can be allocated in
11168 order to perform the global common subexpression elimination
11169 optimization. If more memory than specified is required, the
11170 optimization is not done.
11172 @item max-gcse-insertion-ratio
11173 If the ratio of expression insertions to deletions is larger than this value
11174 for any expression, then RTL PRE inserts or removes the expression and thus
11175 leaves partially redundant computations in the instruction stream.
11177 @item max-pending-list-length
11178 The maximum number of pending dependencies scheduling allows
11179 before flushing the current state and starting over. Large functions
11180 with few branches or calls can create excessively large lists which
11181 needlessly consume memory and resources.
11183 @item max-modulo-backtrack-attempts
11184 The maximum number of backtrack attempts the scheduler should make
11185 when modulo scheduling a loop. Larger values can exponentially increase
11188 @item max-inline-insns-single
11189 Several parameters control the tree inliner used in GCC@.
11190 This number sets the maximum number of instructions (counted in GCC's
11191 internal representation) in a single function that the tree inliner
11192 considers for inlining. This only affects functions declared
11193 inline and methods implemented in a class declaration (C++).
11195 @item max-inline-insns-auto
11196 When you use @option{-finline-functions} (included in @option{-O3}),
11197 a lot of functions that would otherwise not be considered for inlining
11198 by the compiler are investigated. To those functions, a different
11199 (more restrictive) limit compared to functions declared inline can
11202 @item max-inline-insns-small
11203 This is bound applied to calls which are considered relevant with
11204 @option{-finline-small-functions}.
11206 @item max-inline-insns-size
11207 This is bound applied to calls which are optimized for size. Small growth
11208 may be desirable to anticipate optimization oppurtunities exposed by inlining.
11210 @item uninlined-function-insns
11211 Number of instructions accounted by inliner for function overhead such as
11212 function prologue and epilogue.
11214 @item uninlined-function-time
11215 Extra time accounted by inliner for function overhead such as time needed to
11216 execute function prologue and epilogue
11218 @item uninlined-thunk-insns
11219 @item uninlined-thunk-time
11220 Same as @option{--param uninlined-function-insns} and
11221 @option{--param uninlined-function-time} but applied to function thunks
11223 @item inline-min-speedup
11224 When estimated performance improvement of caller + callee runtime exceeds this
11225 threshold (in percent), the function can be inlined regardless of the limit on
11226 @option{--param max-inline-insns-single} and @option{--param
11227 max-inline-insns-auto}.
11229 @item large-function-insns
11230 The limit specifying really large functions. For functions larger than this
11231 limit after inlining, inlining is constrained by
11232 @option{--param large-function-growth}. This parameter is useful primarily
11233 to avoid extreme compilation time caused by non-linear algorithms used by the
11236 @item large-function-growth
11237 Specifies maximal growth of large function caused by inlining in percents.
11238 For example, parameter value 100 limits large function growth to 2.0 times
11241 @item large-unit-insns
11242 The limit specifying large translation unit. Growth caused by inlining of
11243 units larger than this limit is limited by @option{--param inline-unit-growth}.
11244 For small units this might be too tight.
11245 For example, consider a unit consisting of function A
11246 that is inline and B that just calls A three times. If B is small relative to
11247 A, the growth of unit is 300\% and yet such inlining is very sane. For very
11248 large units consisting of small inlineable functions, however, the overall unit
11249 growth limit is needed to avoid exponential explosion of code size. Thus for
11250 smaller units, the size is increased to @option{--param large-unit-insns}
11251 before applying @option{--param inline-unit-growth}.
11253 @item inline-unit-growth
11254 Specifies maximal overall growth of the compilation unit caused by inlining.
11255 For example, parameter value 20 limits unit growth to 1.2 times the original
11256 size. Cold functions (either marked cold via an attribute or by profile
11257 feedback) are not accounted into the unit size.
11259 @item ipcp-unit-growth
11260 Specifies maximal overall growth of the compilation unit caused by
11261 interprocedural constant propagation. For example, parameter value 10 limits
11262 unit growth to 1.1 times the original size.
11264 @item large-stack-frame
11265 The limit specifying large stack frames. While inlining the algorithm is trying
11266 to not grow past this limit too much.
11268 @item large-stack-frame-growth
11269 Specifies maximal growth of large stack frames caused by inlining in percents.
11270 For example, parameter value 1000 limits large stack frame growth to 11 times
11273 @item max-inline-insns-recursive
11274 @itemx max-inline-insns-recursive-auto
11275 Specifies the maximum number of instructions an out-of-line copy of a
11276 self-recursive inline
11277 function can grow into by performing recursive inlining.
11279 @option{--param max-inline-insns-recursive} applies to functions
11281 For functions not declared inline, recursive inlining
11282 happens only when @option{-finline-functions} (included in @option{-O3}) is
11283 enabled; @option{--param max-inline-insns-recursive-auto} applies instead.
11285 @item max-inline-recursive-depth
11286 @itemx max-inline-recursive-depth-auto
11287 Specifies the maximum recursion depth used for recursive inlining.
11289 @option{--param max-inline-recursive-depth} applies to functions
11290 declared inline. For functions not declared inline, recursive inlining
11291 happens only when @option{-finline-functions} (included in @option{-O3}) is
11292 enabled; @option{--param max-inline-recursive-depth-auto} applies instead.
11294 @item min-inline-recursive-probability
11295 Recursive inlining is profitable only for function having deep recursion
11296 in average and can hurt for function having little recursion depth by
11297 increasing the prologue size or complexity of function body to other
11300 When profile feedback is available (see @option{-fprofile-generate}) the actual
11301 recursion depth can be guessed from the probability that function recurses
11302 via a given call expression. This parameter limits inlining only to call
11303 expressions whose probability exceeds the given threshold (in percents).
11305 @item early-inlining-insns
11306 Specify growth that the early inliner can make. In effect it increases
11307 the amount of inlining for code having a large abstraction penalty.
11309 @item max-early-inliner-iterations
11310 Limit of iterations of the early inliner. This basically bounds
11311 the number of nested indirect calls the early inliner can resolve.
11312 Deeper chains are still handled by late inlining.
11314 @item comdat-sharing-probability
11315 Probability (in percent) that C++ inline function with comdat visibility
11316 are shared across multiple compilation units.
11318 @item profile-func-internal-id
11319 A parameter to control whether to use function internal id in profile
11320 database lookup. If the value is 0, the compiler uses an id that
11321 is based on function assembler name and filename, which makes old profile
11322 data more tolerant to source changes such as function reordering etc.
11324 @item min-vect-loop-bound
11325 The minimum number of iterations under which loops are not vectorized
11326 when @option{-ftree-vectorize} is used. The number of iterations after
11327 vectorization needs to be greater than the value specified by this option
11328 to allow vectorization.
11330 @item gcse-cost-distance-ratio
11331 Scaling factor in calculation of maximum distance an expression
11332 can be moved by GCSE optimizations. This is currently supported only in the
11333 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
11334 is with simple expressions, i.e., the expressions that have cost
11335 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
11336 hoisting of simple expressions.
11338 @item gcse-unrestricted-cost
11339 Cost, roughly measured as the cost of a single typical machine
11340 instruction, at which GCSE optimizations do not constrain
11341 the distance an expression can travel. This is currently
11342 supported only in the code hoisting pass. The lesser the cost,
11343 the more aggressive code hoisting is. Specifying 0
11344 allows all expressions to travel unrestricted distances.
11346 @item max-hoist-depth
11347 The depth of search in the dominator tree for expressions to hoist.
11348 This is used to avoid quadratic behavior in hoisting algorithm.
11349 The value of 0 does not limit on the search, but may slow down compilation
11352 @item max-tail-merge-comparisons
11353 The maximum amount of similar bbs to compare a bb with. This is used to
11354 avoid quadratic behavior in tree tail merging.
11356 @item max-tail-merge-iterations
11357 The maximum amount of iterations of the pass over the function. This is used to
11358 limit compilation time in tree tail merging.
11360 @item store-merging-allow-unaligned
11361 Allow the store merging pass to introduce unaligned stores if it is legal to
11364 @item max-stores-to-merge
11365 The maximum number of stores to attempt to merge into wider stores in the store
11368 @item max-unrolled-insns
11369 The maximum number of instructions that a loop may have to be unrolled.
11370 If a loop is unrolled, this parameter also determines how many times
11371 the loop code is unrolled.
11373 @item max-average-unrolled-insns
11374 The maximum number of instructions biased by probabilities of their execution
11375 that a loop may have to be unrolled. If a loop is unrolled,
11376 this parameter also determines how many times the loop code is unrolled.
11378 @item max-unroll-times
11379 The maximum number of unrollings of a single loop.
11381 @item max-peeled-insns
11382 The maximum number of instructions that a loop may have to be peeled.
11383 If a loop is peeled, this parameter also determines how many times
11384 the loop code is peeled.
11386 @item max-peel-times
11387 The maximum number of peelings of a single loop.
11389 @item max-peel-branches
11390 The maximum number of branches on the hot path through the peeled sequence.
11392 @item max-completely-peeled-insns
11393 The maximum number of insns of a completely peeled loop.
11395 @item max-completely-peel-times
11396 The maximum number of iterations of a loop to be suitable for complete peeling.
11398 @item max-completely-peel-loop-nest-depth
11399 The maximum depth of a loop nest suitable for complete peeling.
11401 @item max-unswitch-insns
11402 The maximum number of insns of an unswitched loop.
11404 @item max-unswitch-level
11405 The maximum number of branches unswitched in a single loop.
11407 @item lim-expensive
11408 The minimum cost of an expensive expression in the loop invariant motion.
11410 @item iv-consider-all-candidates-bound
11411 Bound on number of candidates for induction variables, below which
11412 all candidates are considered for each use in induction variable
11413 optimizations. If there are more candidates than this,
11414 only the most relevant ones are considered to avoid quadratic time complexity.
11416 @item iv-max-considered-uses
11417 The induction variable optimizations give up on loops that contain more
11418 induction variable uses.
11420 @item iv-always-prune-cand-set-bound
11421 If the number of candidates in the set is smaller than this value,
11422 always try to remove unnecessary ivs from the set
11423 when adding a new one.
11425 @item avg-loop-niter
11426 Average number of iterations of a loop.
11428 @item dse-max-object-size
11429 Maximum size (in bytes) of objects tracked bytewise by dead store elimination.
11430 Larger values may result in larger compilation times.
11432 @item dse-max-alias-queries-per-store
11433 Maximum number of queries into the alias oracle per store.
11434 Larger values result in larger compilation times and may result in more
11435 removed dead stores.
11437 @item scev-max-expr-size
11438 Bound on size of expressions used in the scalar evolutions analyzer.
11439 Large expressions slow the analyzer.
11441 @item scev-max-expr-complexity
11442 Bound on the complexity of the expressions in the scalar evolutions analyzer.
11443 Complex expressions slow the analyzer.
11445 @item max-tree-if-conversion-phi-args
11446 Maximum number of arguments in a PHI supported by TREE if conversion
11447 unless the loop is marked with simd pragma.
11449 @item vect-max-version-for-alignment-checks
11450 The maximum number of run-time checks that can be performed when
11451 doing loop versioning for alignment in the vectorizer.
11453 @item vect-max-version-for-alias-checks
11454 The maximum number of run-time checks that can be performed when
11455 doing loop versioning for alias in the vectorizer.
11457 @item vect-max-peeling-for-alignment
11458 The maximum number of loop peels to enhance access alignment
11459 for vectorizer. Value -1 means no limit.
11461 @item max-iterations-to-track
11462 The maximum number of iterations of a loop the brute-force algorithm
11463 for analysis of the number of iterations of the loop tries to evaluate.
11465 @item hot-bb-count-fraction
11466 The denominator n of fraction 1/n of the maximal execution count of a
11467 basic block in the entire program that a basic block needs to at least
11468 have in order to be considered hot. The default is 10000, which means
11469 that a basic block is considered hot if its execution count is greater
11470 than 1/10000 of the maximal execution count. 0 means that it is never
11471 considered hot. Used in non-LTO mode.
11473 @item hot-bb-count-ws-permille
11474 The number of most executed permilles, ranging from 0 to 1000, of the
11475 profiled execution of the entire program to which the execution count
11476 of a basic block must be part of in order to be considered hot. The
11477 default is 990, which means that a basic block is considered hot if
11478 its execution count contributes to the upper 990 permilles, or 99.0%,
11479 of the profiled execution of the entire program. 0 means that it is
11480 never considered hot. Used in LTO mode.
11482 @item hot-bb-frequency-fraction
11483 The denominator n of fraction 1/n of the execution frequency of the
11484 entry block of a function that a basic block of this function needs
11485 to at least have in order to be considered hot. The default is 1000,
11486 which means that a basic block is considered hot in a function if it
11487 is executed more frequently than 1/1000 of the frequency of the entry
11488 block of the function. 0 means that it is never considered hot.
11490 @item unlikely-bb-count-fraction
11491 The denominator n of fraction 1/n of the number of profiled runs of
11492 the entire program below which the execution count of a basic block
11493 must be in order for the basic block to be considered unlikely executed.
11494 The default is 20, which means that a basic block is considered unlikely
11495 executed if it is executed in fewer than 1/20, or 5%, of the runs of
11496 the program. 0 means that it is always considered unlikely executed.
11498 @item max-predicted-iterations
11499 The maximum number of loop iterations we predict statically. This is useful
11500 in cases where a function contains a single loop with known bound and
11501 another loop with unknown bound.
11502 The known number of iterations is predicted correctly, while
11503 the unknown number of iterations average to roughly 10. This means that the
11504 loop without bounds appears artificially cold relative to the other one.
11506 @item builtin-expect-probability
11507 Control the probability of the expression having the specified value. This
11508 parameter takes a percentage (i.e.@: 0 ... 100) as input.
11510 @item builtin-string-cmp-inline-length
11511 The maximum length of a constant string for a builtin string cmp call
11512 eligible for inlining.
11514 @item align-threshold
11516 Select fraction of the maximal frequency of executions of a basic block in
11517 a function to align the basic block.
11519 @item align-loop-iterations
11521 A loop expected to iterate at least the selected number of iterations is
11524 @item tracer-dynamic-coverage
11525 @itemx tracer-dynamic-coverage-feedback
11527 This value is used to limit superblock formation once the given percentage of
11528 executed instructions is covered. This limits unnecessary code size
11531 The @option{tracer-dynamic-coverage-feedback} parameter
11532 is used only when profile
11533 feedback is available. The real profiles (as opposed to statically estimated
11534 ones) are much less balanced allowing the threshold to be larger value.
11536 @item tracer-max-code-growth
11537 Stop tail duplication once code growth has reached given percentage. This is
11538 a rather artificial limit, as most of the duplicates are eliminated later in
11539 cross jumping, so it may be set to much higher values than is the desired code
11542 @item tracer-min-branch-ratio
11544 Stop reverse growth when the reverse probability of best edge is less than this
11545 threshold (in percent).
11547 @item tracer-min-branch-probability
11548 @itemx tracer-min-branch-probability-feedback
11550 Stop forward growth if the best edge has probability lower than this
11553 Similarly to @option{tracer-dynamic-coverage} two parameters are
11554 provided. @option{tracer-min-branch-probability-feedback} is used for
11555 compilation with profile feedback and @option{tracer-min-branch-probability}
11556 compilation without. The value for compilation with profile feedback
11557 needs to be more conservative (higher) in order to make tracer
11560 @item stack-clash-protection-guard-size
11561 Specify the size of the operating system provided stack guard as
11562 2 raised to @var{num} bytes. Higher values may reduce the
11563 number of explicit probes, but a value larger than the operating system
11564 provided guard will leave code vulnerable to stack clash style attacks.
11566 @item stack-clash-protection-probe-interval
11567 Stack clash protection involves probing stack space as it is allocated. This
11568 param controls the maximum distance between probes into the stack as 2 raised
11569 to @var{num} bytes. Higher values may reduce the number of explicit probes, but a value
11570 larger than the operating system provided guard will leave code vulnerable to
11571 stack clash style attacks.
11573 @item max-cse-path-length
11575 The maximum number of basic blocks on path that CSE considers.
11577 @item max-cse-insns
11578 The maximum number of instructions CSE processes before flushing.
11580 @item ggc-min-expand
11582 GCC uses a garbage collector to manage its own memory allocation. This
11583 parameter specifies the minimum percentage by which the garbage
11584 collector's heap should be allowed to expand between collections.
11585 Tuning this may improve compilation speed; it has no effect on code
11588 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
11589 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
11590 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
11591 GCC is not able to calculate RAM on a particular platform, the lower
11592 bound of 30% is used. Setting this parameter and
11593 @option{ggc-min-heapsize} to zero causes a full collection to occur at
11594 every opportunity. This is extremely slow, but can be useful for
11597 @item ggc-min-heapsize
11599 Minimum size of the garbage collector's heap before it begins bothering
11600 to collect garbage. The first collection occurs after the heap expands
11601 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
11602 tuning this may improve compilation speed, and has no effect on code
11605 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
11606 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
11607 with a lower bound of 4096 (four megabytes) and an upper bound of
11608 131072 (128 megabytes). If GCC is not able to calculate RAM on a
11609 particular platform, the lower bound is used. Setting this parameter
11610 very large effectively disables garbage collection. Setting this
11611 parameter and @option{ggc-min-expand} to zero causes a full collection
11612 to occur at every opportunity.
11614 @item max-reload-search-insns
11615 The maximum number of instruction reload should look backward for equivalent
11616 register. Increasing values mean more aggressive optimization, making the
11617 compilation time increase with probably slightly better performance.
11619 @item max-cselib-memory-locations
11620 The maximum number of memory locations cselib should take into account.
11621 Increasing values mean more aggressive optimization, making the compilation time
11622 increase with probably slightly better performance.
11624 @item max-sched-ready-insns
11625 The maximum number of instructions ready to be issued the scheduler should
11626 consider at any given time during the first scheduling pass. Increasing
11627 values mean more thorough searches, making the compilation time increase
11628 with probably little benefit.
11630 @item max-sched-region-blocks
11631 The maximum number of blocks in a region to be considered for
11632 interblock scheduling.
11634 @item max-pipeline-region-blocks
11635 The maximum number of blocks in a region to be considered for
11636 pipelining in the selective scheduler.
11638 @item max-sched-region-insns
11639 The maximum number of insns in a region to be considered for
11640 interblock scheduling.
11642 @item max-pipeline-region-insns
11643 The maximum number of insns in a region to be considered for
11644 pipelining in the selective scheduler.
11646 @item min-spec-prob
11647 The minimum probability (in percents) of reaching a source block
11648 for interblock speculative scheduling.
11650 @item max-sched-extend-regions-iters
11651 The maximum number of iterations through CFG to extend regions.
11652 A value of 0 disables region extensions.
11654 @item max-sched-insn-conflict-delay
11655 The maximum conflict delay for an insn to be considered for speculative motion.
11657 @item sched-spec-prob-cutoff
11658 The minimal probability of speculation success (in percents), so that
11659 speculative insns are scheduled.
11661 @item sched-state-edge-prob-cutoff
11662 The minimum probability an edge must have for the scheduler to save its
11665 @item sched-mem-true-dep-cost
11666 Minimal distance (in CPU cycles) between store and load targeting same
11669 @item selsched-max-lookahead
11670 The maximum size of the lookahead window of selective scheduling. It is a
11671 depth of search for available instructions.
11673 @item selsched-max-sched-times
11674 The maximum number of times that an instruction is scheduled during
11675 selective scheduling. This is the limit on the number of iterations
11676 through which the instruction may be pipelined.
11678 @item selsched-insns-to-rename
11679 The maximum number of best instructions in the ready list that are considered
11680 for renaming in the selective scheduler.
11683 The minimum value of stage count that swing modulo scheduler
11686 @item max-last-value-rtl
11687 The maximum size measured as number of RTLs that can be recorded in an expression
11688 in combiner for a pseudo register as last known value of that register.
11690 @item max-combine-insns
11691 The maximum number of instructions the RTL combiner tries to combine.
11693 @item integer-share-limit
11694 Small integer constants can use a shared data structure, reducing the
11695 compiler's memory usage and increasing its speed. This sets the maximum
11696 value of a shared integer constant.
11698 @item ssp-buffer-size
11699 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
11700 protection when @option{-fstack-protection} is used.
11702 @item min-size-for-stack-sharing
11703 The minimum size of variables taking part in stack slot sharing when not
11706 @item max-jump-thread-duplication-stmts
11707 Maximum number of statements allowed in a block that needs to be
11708 duplicated when threading jumps.
11710 @item max-fields-for-field-sensitive
11711 Maximum number of fields in a structure treated in
11712 a field sensitive manner during pointer analysis.
11714 @item prefetch-latency
11715 Estimate on average number of instructions that are executed before
11716 prefetch finishes. The distance prefetched ahead is proportional
11717 to this constant. Increasing this number may also lead to less
11718 streams being prefetched (see @option{simultaneous-prefetches}).
11720 @item simultaneous-prefetches
11721 Maximum number of prefetches that can run at the same time.
11723 @item l1-cache-line-size
11724 The size of cache line in L1 data cache, in bytes.
11726 @item l1-cache-size
11727 The size of L1 data cache, in kilobytes.
11729 @item l2-cache-size
11730 The size of L2 data cache, in kilobytes.
11732 @item prefetch-dynamic-strides
11733 Whether the loop array prefetch pass should issue software prefetch hints
11734 for strides that are non-constant. In some cases this may be
11735 beneficial, though the fact the stride is non-constant may make it
11736 hard to predict when there is clear benefit to issuing these hints.
11738 Set to 1 if the prefetch hints should be issued for non-constant
11739 strides. Set to 0 if prefetch hints should be issued only for strides that
11740 are known to be constant and below @option{prefetch-minimum-stride}.
11742 @item prefetch-minimum-stride
11743 Minimum constant stride, in bytes, to start using prefetch hints for. If
11744 the stride is less than this threshold, prefetch hints will not be issued.
11746 This setting is useful for processors that have hardware prefetchers, in
11747 which case there may be conflicts between the hardware prefetchers and
11748 the software prefetchers. If the hardware prefetchers have a maximum
11749 stride they can handle, it should be used here to improve the use of
11750 software prefetchers.
11752 A value of -1 means we don't have a threshold and therefore
11753 prefetch hints can be issued for any constant stride.
11755 This setting is only useful for strides that are known and constant.
11757 @item loop-interchange-max-num-stmts
11758 The maximum number of stmts in a loop to be interchanged.
11760 @item loop-interchange-stride-ratio
11761 The minimum ratio between stride of two loops for interchange to be profitable.
11763 @item min-insn-to-prefetch-ratio
11764 The minimum ratio between the number of instructions and the
11765 number of prefetches to enable prefetching in a loop.
11767 @item prefetch-min-insn-to-mem-ratio
11768 The minimum ratio between the number of instructions and the
11769 number of memory references to enable prefetching in a loop.
11771 @item use-canonical-types
11772 Whether the compiler should use the ``canonical'' type system.
11773 Should always be 1, which uses a more efficient internal
11774 mechanism for comparing types in C++ and Objective-C++. However, if
11775 bugs in the canonical type system are causing compilation failures,
11776 set this value to 0 to disable canonical types.
11778 @item switch-conversion-max-branch-ratio
11779 Switch initialization conversion refuses to create arrays that are
11780 bigger than @option{switch-conversion-max-branch-ratio} times the number of
11781 branches in the switch.
11783 @item max-partial-antic-length
11784 Maximum length of the partial antic set computed during the tree
11785 partial redundancy elimination optimization (@option{-ftree-pre}) when
11786 optimizing at @option{-O3} and above. For some sorts of source code
11787 the enhanced partial redundancy elimination optimization can run away,
11788 consuming all of the memory available on the host machine. This
11789 parameter sets a limit on the length of the sets that are computed,
11790 which prevents the runaway behavior. Setting a value of 0 for
11791 this parameter allows an unlimited set length.
11793 @item rpo-vn-max-loop-depth
11794 Maximum loop depth that is value-numbered optimistically.
11795 When the limit hits the innermost
11796 @var{rpo-vn-max-loop-depth} loops and the outermost loop in the
11797 loop nest are value-numbered optimistically and the remaining ones not.
11799 @item sccvn-max-alias-queries-per-access
11800 Maximum number of alias-oracle queries we perform when looking for
11801 redundancies for loads and stores. If this limit is hit the search
11802 is aborted and the load or store is not considered redundant. The
11803 number of queries is algorithmically limited to the number of
11804 stores on all paths from the load to the function entry.
11806 @item ira-max-loops-num
11807 IRA uses regional register allocation by default. If a function
11808 contains more loops than the number given by this parameter, only at most
11809 the given number of the most frequently-executed loops form regions
11810 for regional register allocation.
11812 @item ira-max-conflict-table-size
11813 Although IRA uses a sophisticated algorithm to compress the conflict
11814 table, the table can still require excessive amounts of memory for
11815 huge functions. If the conflict table for a function could be more
11816 than the size in MB given by this parameter, the register allocator
11817 instead uses a faster, simpler, and lower-quality
11818 algorithm that does not require building a pseudo-register conflict table.
11820 @item ira-loop-reserved-regs
11821 IRA can be used to evaluate more accurate register pressure in loops
11822 for decisions to move loop invariants (see @option{-O3}). The number
11823 of available registers reserved for some other purposes is given
11824 by this parameter. Default of the parameter
11825 is the best found from numerous experiments.
11827 @item lra-inheritance-ebb-probability-cutoff
11828 LRA tries to reuse values reloaded in registers in subsequent insns.
11829 This optimization is called inheritance. EBB is used as a region to
11830 do this optimization. The parameter defines a minimal fall-through
11831 edge probability in percentage used to add BB to inheritance EBB in
11832 LRA. The default value was chosen
11833 from numerous runs of SPEC2000 on x86-64.
11835 @item loop-invariant-max-bbs-in-loop
11836 Loop invariant motion can be very expensive, both in compilation time and
11837 in amount of needed compile-time memory, with very large loops. Loops
11838 with more basic blocks than this parameter won't have loop invariant
11839 motion optimization performed on them.
11841 @item loop-max-datarefs-for-datadeps
11842 Building data dependencies is expensive for very large loops. This
11843 parameter limits the number of data references in loops that are
11844 considered for data dependence analysis. These large loops are no
11845 handled by the optimizations using loop data dependencies.
11847 @item max-vartrack-size
11848 Sets a maximum number of hash table slots to use during variable
11849 tracking dataflow analysis of any function. If this limit is exceeded
11850 with variable tracking at assignments enabled, analysis for that
11851 function is retried without it, after removing all debug insns from
11852 the function. If the limit is exceeded even without debug insns, var
11853 tracking analysis is completely disabled for the function. Setting
11854 the parameter to zero makes it unlimited.
11856 @item max-vartrack-expr-depth
11857 Sets a maximum number of recursion levels when attempting to map
11858 variable names or debug temporaries to value expressions. This trades
11859 compilation time for more complete debug information. If this is set too
11860 low, value expressions that are available and could be represented in
11861 debug information may end up not being used; setting this higher may
11862 enable the compiler to find more complex debug expressions, but compile
11863 time and memory use may grow.
11865 @item max-debug-marker-count
11866 Sets a threshold on the number of debug markers (e.g.@: begin stmt
11867 markers) to avoid complexity explosion at inlining or expanding to RTL.
11868 If a function has more such gimple stmts than the set limit, such stmts
11869 will be dropped from the inlined copy of a function, and from its RTL
11872 @item min-nondebug-insn-uid
11873 Use uids starting at this parameter for nondebug insns. The range below
11874 the parameter is reserved exclusively for debug insns created by
11875 @option{-fvar-tracking-assignments}, but debug insns may get
11876 (non-overlapping) uids above it if the reserved range is exhausted.
11878 @item ipa-sra-ptr-growth-factor
11879 IPA-SRA replaces a pointer to an aggregate with one or more new
11880 parameters only when their cumulative size is less or equal to
11881 @option{ipa-sra-ptr-growth-factor} times the size of the original
11884 @item sra-max-scalarization-size-Ospeed
11885 @itemx sra-max-scalarization-size-Osize
11886 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
11887 replace scalar parts of aggregates with uses of independent scalar
11888 variables. These parameters control the maximum size, in storage units,
11889 of aggregate which is considered for replacement when compiling for
11891 (@option{sra-max-scalarization-size-Ospeed}) or size
11892 (@option{sra-max-scalarization-size-Osize}) respectively.
11894 @item tm-max-aggregate-size
11895 When making copies of thread-local variables in a transaction, this
11896 parameter specifies the size in bytes after which variables are
11897 saved with the logging functions as opposed to save/restore code
11898 sequence pairs. This option only applies when using
11901 @item graphite-max-nb-scop-params
11902 To avoid exponential effects in the Graphite loop transforms, the
11903 number of parameters in a Static Control Part (SCoP) is bounded.
11904 A value of zero can be used to lift
11905 the bound. A variable whose value is unknown at compilation time and
11906 defined outside a SCoP is a parameter of the SCoP.
11908 @item loop-block-tile-size
11909 Loop blocking or strip mining transforms, enabled with
11910 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
11911 loop in the loop nest by a given number of iterations. The strip
11912 length can be changed using the @option{loop-block-tile-size}
11915 @item ipa-cp-value-list-size
11916 IPA-CP attempts to track all possible values and types passed to a function's
11917 parameter in order to propagate them and perform devirtualization.
11918 @option{ipa-cp-value-list-size} is the maximum number of values and types it
11919 stores per one formal parameter of a function.
11921 @item ipa-cp-eval-threshold
11922 IPA-CP calculates its own score of cloning profitability heuristics
11923 and performs those cloning opportunities with scores that exceed
11924 @option{ipa-cp-eval-threshold}.
11926 @item ipa-cp-recursion-penalty
11927 Percentage penalty the recursive functions will receive when they
11928 are evaluated for cloning.
11930 @item ipa-cp-single-call-penalty
11931 Percentage penalty functions containing a single call to another
11932 function will receive when they are evaluated for cloning.
11934 @item ipa-max-agg-items
11935 IPA-CP is also capable to propagate a number of scalar values passed
11936 in an aggregate. @option{ipa-max-agg-items} controls the maximum
11937 number of such values per one parameter.
11939 @item ipa-cp-loop-hint-bonus
11940 When IPA-CP determines that a cloning candidate would make the number
11941 of iterations of a loop known, it adds a bonus of
11942 @option{ipa-cp-loop-hint-bonus} to the profitability score of
11945 @item ipa-max-aa-steps
11946 During its analysis of function bodies, IPA-CP employs alias analysis
11947 in order to track values pointed to by function parameters. In order
11948 not spend too much time analyzing huge functions, it gives up and
11949 consider all memory clobbered after examining
11950 @option{ipa-max-aa-steps} statements modifying memory.
11952 @item lto-partitions
11953 Specify desired number of partitions produced during WHOPR compilation.
11954 The number of partitions should exceed the number of CPUs used for compilation.
11956 @item lto-min-partition
11957 Size of minimal partition for WHOPR (in estimated instructions).
11958 This prevents expenses of splitting very small programs into too many
11961 @item lto-max-partition
11962 Size of max partition for WHOPR (in estimated instructions).
11963 to provide an upper bound for individual size of partition.
11964 Meant to be used only with balanced partitioning.
11966 @item lto-max-streaming-parallelism
11967 Maximal number of parallel processes used for LTO streaming.
11969 @item cxx-max-namespaces-for-diagnostic-help
11970 The maximum number of namespaces to consult for suggestions when C++
11971 name lookup fails for an identifier.
11973 @item sink-frequency-threshold
11974 The maximum relative execution frequency (in percents) of the target block
11975 relative to a statement's original block to allow statement sinking of a
11976 statement. Larger numbers result in more aggressive statement sinking.
11977 A small positive adjustment is applied for
11978 statements with memory operands as those are even more profitable so sink.
11980 @item max-stores-to-sink
11981 The maximum number of conditional store pairs that can be sunk. Set to 0
11982 if either vectorization (@option{-ftree-vectorize}) or if-conversion
11983 (@option{-ftree-loop-if-convert}) is disabled.
11985 @item allow-store-data-races
11986 Allow optimizers to introduce new data races on stores.
11987 Set to 1 to allow, otherwise to 0.
11989 @item case-values-threshold
11990 The smallest number of different values for which it is best to use a
11991 jump-table instead of a tree of conditional branches. If the value is
11992 0, use the default for the machine.
11994 @item jump-table-max-growth-ratio-for-size
11995 The maximum code size growth ratio when expanding
11996 into a jump table (in percent). The parameter is used when
11997 optimizing for size.
11999 @item jump-table-max-growth-ratio-for-speed
12000 The maximum code size growth ratio when expanding
12001 into a jump table (in percent). The parameter is used when
12002 optimizing for speed.
12004 @item tree-reassoc-width
12005 Set the maximum number of instructions executed in parallel in
12006 reassociated tree. This parameter overrides target dependent
12007 heuristics used by default if has non zero value.
12009 @item sched-pressure-algorithm
12010 Choose between the two available implementations of
12011 @option{-fsched-pressure}. Algorithm 1 is the original implementation
12012 and is the more likely to prevent instructions from being reordered.
12013 Algorithm 2 was designed to be a compromise between the relatively
12014 conservative approach taken by algorithm 1 and the rather aggressive
12015 approach taken by the default scheduler. It relies more heavily on
12016 having a regular register file and accurate register pressure classes.
12017 See @file{haifa-sched.c} in the GCC sources for more details.
12019 The default choice depends on the target.
12021 @item max-slsr-cand-scan
12022 Set the maximum number of existing candidates that are considered when
12023 seeking a basis for a new straight-line strength reduction candidate.
12026 Enable buffer overflow detection for global objects. This kind
12027 of protection is enabled by default if you are using
12028 @option{-fsanitize=address} option.
12029 To disable global objects protection use @option{--param asan-globals=0}.
12032 Enable buffer overflow detection for stack objects. This kind of
12033 protection is enabled by default when using @option{-fsanitize=address}.
12034 To disable stack protection use @option{--param asan-stack=0} option.
12036 @item asan-instrument-reads
12037 Enable buffer overflow detection for memory reads. This kind of
12038 protection is enabled by default when using @option{-fsanitize=address}.
12039 To disable memory reads protection use
12040 @option{--param asan-instrument-reads=0}.
12042 @item asan-instrument-writes
12043 Enable buffer overflow detection for memory writes. This kind of
12044 protection is enabled by default when using @option{-fsanitize=address}.
12045 To disable memory writes protection use
12046 @option{--param asan-instrument-writes=0} option.
12048 @item asan-memintrin
12049 Enable detection for built-in functions. This kind of protection
12050 is enabled by default when using @option{-fsanitize=address}.
12051 To disable built-in functions protection use
12052 @option{--param asan-memintrin=0}.
12054 @item asan-use-after-return
12055 Enable detection of use-after-return. This kind of protection
12056 is enabled by default when using the @option{-fsanitize=address} option.
12057 To disable it use @option{--param asan-use-after-return=0}.
12059 Note: By default the check is disabled at run time. To enable it,
12060 add @code{detect_stack_use_after_return=1} to the environment variable
12061 @env{ASAN_OPTIONS}.
12063 @item asan-instrumentation-with-call-threshold
12064 If number of memory accesses in function being instrumented
12065 is greater or equal to this number, use callbacks instead of inline checks.
12066 E.g. to disable inline code use
12067 @option{--param asan-instrumentation-with-call-threshold=0}.
12069 @item use-after-scope-direct-emission-threshold
12070 If the size of a local variable in bytes is smaller or equal to this
12071 number, directly poison (or unpoison) shadow memory instead of using
12072 run-time callbacks.
12074 @item max-fsm-thread-path-insns
12075 Maximum number of instructions to copy when duplicating blocks on a
12076 finite state automaton jump thread path.
12078 @item max-fsm-thread-length
12079 Maximum number of basic blocks on a finite state automaton jump thread
12082 @item max-fsm-thread-paths
12083 Maximum number of new jump thread paths to create for a finite state
12086 @item parloops-chunk-size
12087 Chunk size of omp schedule for loops parallelized by parloops.
12089 @item parloops-schedule
12090 Schedule type of omp schedule for loops parallelized by parloops (static,
12091 dynamic, guided, auto, runtime).
12093 @item parloops-min-per-thread
12094 The minimum number of iterations per thread of an innermost parallelized
12095 loop for which the parallelized variant is preferred over the single threaded
12096 one. Note that for a parallelized loop nest the
12097 minimum number of iterations of the outermost loop per thread is two.
12099 @item max-ssa-name-query-depth
12100 Maximum depth of recursion when querying properties of SSA names in things
12101 like fold routines. One level of recursion corresponds to following a
12104 @item hsa-gen-debug-stores
12105 Enable emission of special debug stores within HSA kernels which are
12106 then read and reported by libgomp plugin. Generation of these stores
12107 is disabled by default, use @option{--param hsa-gen-debug-stores=1} to
12110 @item max-speculative-devirt-maydefs
12111 The maximum number of may-defs we analyze when looking for a must-def
12112 specifying the dynamic type of an object that invokes a virtual call
12113 we may be able to devirtualize speculatively.
12115 @item max-vrp-switch-assertions
12116 The maximum number of assertions to add along the default edge of a switch
12117 statement during VRP.
12119 @item unroll-jam-min-percent
12120 The minimum percentage of memory references that must be optimized
12121 away for the unroll-and-jam transformation to be considered profitable.
12123 @item unroll-jam-max-unroll
12124 The maximum number of times the outer loop should be unrolled by
12125 the unroll-and-jam transformation.
12127 @item max-rtl-if-conversion-unpredictable-cost
12128 Maximum permissible cost for the sequence that would be generated
12129 by the RTL if-conversion pass for a branch that is considered unpredictable.
12131 @item max-variable-expansions-in-unroller
12132 If @option{-fvariable-expansion-in-unroller} is used, the maximum number
12133 of times that an individual variable will be expanded during loop unrolling.
12135 @item tracer-min-branch-probability-feedback
12136 Stop forward growth if the probability of best edge is less than
12137 this threshold (in percent). Used when profile feedback is available.
12139 @item partial-inlining-entry-probability
12140 Maximum probability of the entry BB of split region
12141 (in percent relative to entry BB of the function)
12142 to make partial inlining happen.
12144 @item max-tracked-strlens
12145 Maximum number of strings for which strlen optimization pass will
12146 track string lengths.
12148 @item gcse-after-reload-partial-fraction
12149 The threshold ratio for performing partial redundancy
12150 elimination after reload.
12152 @item gcse-after-reload-critical-fraction
12153 The threshold ratio of critical edges execution count that
12154 permit performing redundancy elimination after reload.
12156 @item max-loop-header-insns
12157 The maximum number of insns in loop header duplicated
12158 by the copy loop headers pass.
12160 @item vect-epilogues-nomask
12161 Enable loop epilogue vectorization using smaller vector size.
12163 @item slp-max-insns-in-bb
12164 Maximum number of instructions in basic block to be
12165 considered for SLP vectorization.
12167 @item avoid-fma-max-bits
12168 Maximum number of bits for which we avoid creating FMAs.
12170 @item sms-loop-average-count-threshold
12171 A threshold on the average loop count considered by the swing modulo scheduler.
12173 @item sms-dfa-history
12174 The number of cycles the swing modulo scheduler considers when checking
12175 conflicts using DFA.
12177 @item max-inline-insns-recursive-auto
12178 The maximum number of instructions non-inline function
12179 can grow to via recursive inlining.
12181 @item graphite-allow-codegen-errors
12182 Whether codegen errors should be ICEs when @option{-fchecking}.
12184 @item sms-max-ii-factor
12185 A factor for tuning the upper bound that swing modulo scheduler
12186 uses for scheduling a loop.
12188 @item lra-max-considered-reload-pseudos
12189 The max number of reload pseudos which are considered during
12190 spilling a non-reload pseudo.
12192 @item max-pow-sqrt-depth
12193 Maximum depth of sqrt chains to use when synthesizing exponentiation
12194 by a real constant.
12196 @item max-dse-active-local-stores
12197 Maximum number of active local stores in RTL dead store elimination.
12199 @item asan-instrument-allocas
12200 Enable asan allocas/VLAs protection.
12202 @item max-iterations-computation-cost
12203 Bound on the cost of an expression to compute the number of iterations.
12205 @item max-isl-operations
12206 Maximum number of isl operations, 0 means unlimited.
12208 @item graphite-max-arrays-per-scop
12209 Maximum number of arrays per scop.
12211 @item max-vartrack-reverse-op-size
12212 Max. size of loc list for which reverse ops should be added.
12214 @item tracer-dynamic-coverage-feedback
12215 The percentage of function, weighted by execution frequency,
12216 that must be covered by trace formation.
12217 Used when profile feedback is available.
12219 @item max-inline-recursive-depth-auto
12220 The maximum depth of recursive inlining for non-inline functions.
12222 @item fsm-scale-path-stmts
12223 Scale factor to apply to the number of statements in a threading path
12224 when comparing to the number of (scaled) blocks.
12226 @item fsm-maximum-phi-arguments
12227 Maximum number of arguments a PHI may have before the FSM threader
12228 will not try to thread through its block.
12230 @item uninit-control-dep-attempts
12231 Maximum number of nested calls to search for control dependencies
12232 during uninitialized variable analysis.
12234 @item max-once-peeled-insns
12235 The maximum number of insns of a peeled loop that rolls only once.
12237 @item sra-max-scalarization-size-Osize
12238 Maximum size, in storage units, of an aggregate
12239 which should be considered for scalarization when compiling for size.
12241 @item fsm-scale-path-blocks
12242 Scale factor to apply to the number of blocks in a threading path
12243 when comparing to the number of (scaled) statements.
12245 @item sched-autopref-queue-depth
12246 Hardware autoprefetcher scheduler model control flag.
12247 Number of lookahead cycles the model looks into; at '
12248 ' only enable instruction sorting heuristic.
12250 @item loop-versioning-max-inner-insns
12251 The maximum number of instructions that an inner loop can have
12252 before the loop versioning pass considers it too big to copy.
12254 @item loop-versioning-max-outer-insns
12255 The maximum number of instructions that an outer loop can have
12256 before the loop versioning pass considers it too big to copy,
12257 discounting any instructions in inner loops that directly benefit
12260 @item ssa-name-def-chain-limit
12261 The maximum number of SSA_NAME assignments to follow in determining
12262 a property of a variable such as its value. This limits the number
12263 of iterations or recursive calls GCC performs when optimizing certain
12264 statements or when determining their validity prior to issuing
12270 @node Instrumentation Options
12271 @section Program Instrumentation Options
12272 @cindex instrumentation options
12273 @cindex program instrumentation options
12274 @cindex run-time error checking options
12275 @cindex profiling options
12276 @cindex options, program instrumentation
12277 @cindex options, run-time error checking
12278 @cindex options, profiling
12280 GCC supports a number of command-line options that control adding
12281 run-time instrumentation to the code it normally generates.
12282 For example, one purpose of instrumentation is collect profiling
12283 statistics for use in finding program hot spots, code coverage
12284 analysis, or profile-guided optimizations.
12285 Another class of program instrumentation is adding run-time checking
12286 to detect programming errors like invalid pointer
12287 dereferences or out-of-bounds array accesses, as well as deliberately
12288 hostile attacks such as stack smashing or C++ vtable hijacking.
12289 There is also a general hook which can be used to implement other
12290 forms of tracing or function-level instrumentation for debug or
12291 program analysis purposes.
12294 @cindex @command{prof}
12295 @cindex @command{gprof}
12300 Generate extra code to write profile information suitable for the
12301 analysis program @command{prof} (for @option{-p}) or @command{gprof}
12302 (for @option{-pg}). You must use this option when compiling
12303 the source files you want data about, and you must also use it when
12306 You can use the function attribute @code{no_instrument_function} to
12307 suppress profiling of individual functions when compiling with these options.
12308 @xref{Common Function Attributes}.
12310 @item -fprofile-arcs
12311 @opindex fprofile-arcs
12312 Add code so that program flow @dfn{arcs} are instrumented. During
12313 execution the program records how many times each branch and call is
12314 executed and how many times it is taken or returns. On targets that support
12315 constructors with priority support, profiling properly handles constructors,
12316 destructors and C++ constructors (and destructors) of classes which are used
12317 as a type of a global variable.
12320 program exits it saves this data to a file called
12321 @file{@var{auxname}.gcda} for each source file. The data may be used for
12322 profile-directed optimizations (@option{-fbranch-probabilities}), or for
12323 test coverage analysis (@option{-ftest-coverage}). Each object file's
12324 @var{auxname} is generated from the name of the output file, if
12325 explicitly specified and it is not the final executable, otherwise it is
12326 the basename of the source file. In both cases any suffix is removed
12327 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
12328 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
12329 @xref{Cross-profiling}.
12331 @cindex @command{gcov}
12335 This option is used to compile and link code instrumented for coverage
12336 analysis. The option is a synonym for @option{-fprofile-arcs}
12337 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
12338 linking). See the documentation for those options for more details.
12343 Compile the source files with @option{-fprofile-arcs} plus optimization
12344 and code generation options. For test coverage analysis, use the
12345 additional @option{-ftest-coverage} option. You do not need to profile
12346 every source file in a program.
12349 Compile the source files additionally with @option{-fprofile-abs-path}
12350 to create absolute path names in the @file{.gcno} files. This allows
12351 @command{gcov} to find the correct sources in projects where compilations
12352 occur with different working directories.
12355 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
12356 (the latter implies the former).
12359 Run the program on a representative workload to generate the arc profile
12360 information. This may be repeated any number of times. You can run
12361 concurrent instances of your program, and provided that the file system
12362 supports locking, the data files will be correctly updated. Unless
12363 a strict ISO C dialect option is in effect, @code{fork} calls are
12364 detected and correctly handled without double counting.
12367 For profile-directed optimizations, compile the source files again with
12368 the same optimization and code generation options plus
12369 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
12370 Control Optimization}).
12373 For test coverage analysis, use @command{gcov} to produce human readable
12374 information from the @file{.gcno} and @file{.gcda} files. Refer to the
12375 @command{gcov} documentation for further information.
12379 With @option{-fprofile-arcs}, for each function of your program GCC
12380 creates a program flow graph, then finds a spanning tree for the graph.
12381 Only arcs that are not on the spanning tree have to be instrumented: the
12382 compiler adds code to count the number of times that these arcs are
12383 executed. When an arc is the only exit or only entrance to a block, the
12384 instrumentation code can be added to the block; otherwise, a new basic
12385 block must be created to hold the instrumentation code.
12388 @item -ftest-coverage
12389 @opindex ftest-coverage
12390 Produce a notes file that the @command{gcov} code-coverage utility
12391 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
12392 show program coverage. Each source file's note file is called
12393 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
12394 above for a description of @var{auxname} and instructions on how to
12395 generate test coverage data. Coverage data matches the source files
12396 more closely if you do not optimize.
12398 @item -fprofile-abs-path
12399 @opindex fprofile-abs-path
12400 Automatically convert relative source file names to absolute path names
12401 in the @file{.gcno} files. This allows @command{gcov} to find the correct
12402 sources in projects where compilations occur with different working
12405 @item -fprofile-dir=@var{path}
12406 @opindex fprofile-dir
12408 Set the directory to search for the profile data files in to @var{path}.
12409 This option affects only the profile data generated by
12410 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
12411 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
12412 and its related options. Both absolute and relative paths can be used.
12413 By default, GCC uses the current directory as @var{path}, thus the
12414 profile data file appears in the same directory as the object file.
12415 In order to prevent the file name clashing, if the object file name is
12416 not an absolute path, we mangle the absolute path of the
12417 @file{@var{sourcename}.gcda} file and use it as the file name of a
12418 @file{.gcda} file. See similar option @option{-fprofile-note}.
12420 When an executable is run in a massive parallel environment, it is recommended
12421 to save profile to different folders. That can be done with variables
12422 in @var{path} that are exported during run-time:
12430 value of environment variable @var{VAR}
12434 @item -fprofile-generate
12435 @itemx -fprofile-generate=@var{path}
12436 @opindex fprofile-generate
12438 Enable options usually used for instrumenting application to produce
12439 profile useful for later recompilation with profile feedback based
12440 optimization. You must use @option{-fprofile-generate} both when
12441 compiling and when linking your program.
12443 The following options are enabled:
12444 @option{-fprofile-arcs}, @option{-fprofile-values},
12445 @option{-finline-functions}, and @option{-fipa-bit-cp}.
12447 If @var{path} is specified, GCC looks at the @var{path} to find
12448 the profile feedback data files. See @option{-fprofile-dir}.
12450 To optimize the program based on the collected profile information, use
12451 @option{-fprofile-use}. @xref{Optimize Options}, for more information.
12453 @item -fprofile-note=@var{path}
12454 @opindex fprofile-note
12456 If @var{path} is specified, GCC saves @file{.gcno} file into @var{path}
12457 location. If you combine the option with multiple source files,
12458 the @file{.gcno} file will be overwritten.
12460 @item -fprofile-update=@var{method}
12461 @opindex fprofile-update
12463 Alter the update method for an application instrumented for profile
12464 feedback based optimization. The @var{method} argument should be one of
12465 @samp{single}, @samp{atomic} or @samp{prefer-atomic}.
12466 The first one is useful for single-threaded applications,
12467 while the second one prevents profile corruption by emitting thread-safe code.
12469 @strong{Warning:} When an application does not properly join all threads
12470 (or creates an detached thread), a profile file can be still corrupted.
12472 Using @samp{prefer-atomic} would be transformed either to @samp{atomic},
12473 when supported by a target, or to @samp{single} otherwise. The GCC driver
12474 automatically selects @samp{prefer-atomic} when @option{-pthread}
12475 is present in the command line.
12477 @item -fprofile-filter-files=@var{regex}
12478 @opindex fprofile-filter-files
12480 Instrument only functions from files where names match
12481 any regular expression (separated by a semi-colon).
12483 For example, @option{-fprofile-filter-files=main.c;module.*.c} will instrument
12484 only @file{main.c} and all C files starting with 'module'.
12486 @item -fprofile-exclude-files=@var{regex}
12487 @opindex fprofile-exclude-files
12489 Instrument only functions from files where names do not match
12490 all the regular expressions (separated by a semi-colon).
12492 For example, @option{-fprofile-exclude-files=/usr/*} will prevent instrumentation
12493 of all files that are located in @file{/usr/} folder.
12495 @item -fsanitize=address
12496 @opindex fsanitize=address
12497 Enable AddressSanitizer, a fast memory error detector.
12498 Memory access instructions are instrumented to detect
12499 out-of-bounds and use-after-free bugs.
12500 The option enables @option{-fsanitize-address-use-after-scope}.
12501 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
12502 more details. The run-time behavior can be influenced using the
12503 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
12504 the available options are shown at startup of the instrumented program. See
12505 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
12506 for a list of supported options.
12507 The option cannot be combined with @option{-fsanitize=thread}.
12509 @item -fsanitize=kernel-address
12510 @opindex fsanitize=kernel-address
12511 Enable AddressSanitizer for Linux kernel.
12512 See @uref{https://github.com/google/kasan/wiki} for more details.
12514 @item -fsanitize=pointer-compare
12515 @opindex fsanitize=pointer-compare
12516 Instrument comparison operation (<, <=, >, >=) with pointer operands.
12517 The option must be combined with either @option{-fsanitize=kernel-address} or
12518 @option{-fsanitize=address}
12519 The option cannot be combined with @option{-fsanitize=thread}.
12520 Note: By default the check is disabled at run time. To enable it,
12521 add @code{detect_invalid_pointer_pairs=2} to the environment variable
12522 @env{ASAN_OPTIONS}. Using @code{detect_invalid_pointer_pairs=1} detects
12523 invalid operation only when both pointers are non-null.
12525 @item -fsanitize=pointer-subtract
12526 @opindex fsanitize=pointer-subtract
12527 Instrument subtraction with pointer operands.
12528 The option must be combined with either @option{-fsanitize=kernel-address} or
12529 @option{-fsanitize=address}
12530 The option cannot be combined with @option{-fsanitize=thread}.
12531 Note: By default the check is disabled at run time. To enable it,
12532 add @code{detect_invalid_pointer_pairs=2} to the environment variable
12533 @env{ASAN_OPTIONS}. Using @code{detect_invalid_pointer_pairs=1} detects
12534 invalid operation only when both pointers are non-null.
12536 @item -fsanitize=thread
12537 @opindex fsanitize=thread
12538 Enable ThreadSanitizer, a fast data race detector.
12539 Memory access instructions are instrumented to detect
12540 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
12541 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
12542 environment variable; see
12543 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
12545 The option cannot be combined with @option{-fsanitize=address},
12546 @option{-fsanitize=leak}.
12548 Note that sanitized atomic builtins cannot throw exceptions when
12549 operating on invalid memory addresses with non-call exceptions
12550 (@option{-fnon-call-exceptions}).
12552 @item -fsanitize=leak
12553 @opindex fsanitize=leak
12554 Enable LeakSanitizer, a memory leak detector.
12555 This option only matters for linking of executables and
12556 the executable is linked against a library that overrides @code{malloc}
12557 and other allocator functions. See
12558 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
12559 details. The run-time behavior can be influenced using the
12560 @env{LSAN_OPTIONS} environment variable.
12561 The option cannot be combined with @option{-fsanitize=thread}.
12563 @item -fsanitize=undefined
12564 @opindex fsanitize=undefined
12565 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
12566 Various computations are instrumented to detect undefined behavior
12567 at runtime. Current suboptions are:
12571 @item -fsanitize=shift
12572 @opindex fsanitize=shift
12573 This option enables checking that the result of a shift operation is
12574 not undefined. Note that what exactly is considered undefined differs
12575 slightly between C and C++, as well as between ISO C90 and C99, etc.
12576 This option has two suboptions, @option{-fsanitize=shift-base} and
12577 @option{-fsanitize=shift-exponent}.
12579 @item -fsanitize=shift-exponent
12580 @opindex fsanitize=shift-exponent
12581 This option enables checking that the second argument of a shift operation
12582 is not negative and is smaller than the precision of the promoted first
12585 @item -fsanitize=shift-base
12586 @opindex fsanitize=shift-base
12587 If the second argument of a shift operation is within range, check that the
12588 result of a shift operation is not undefined. Note that what exactly is
12589 considered undefined differs slightly between C and C++, as well as between
12590 ISO C90 and C99, etc.
12592 @item -fsanitize=integer-divide-by-zero
12593 @opindex fsanitize=integer-divide-by-zero
12594 Detect integer division by zero as well as @code{INT_MIN / -1} division.
12596 @item -fsanitize=unreachable
12597 @opindex fsanitize=unreachable
12598 With this option, the compiler turns the @code{__builtin_unreachable}
12599 call into a diagnostics message call instead. When reaching the
12600 @code{__builtin_unreachable} call, the behavior is undefined.
12602 @item -fsanitize=vla-bound
12603 @opindex fsanitize=vla-bound
12604 This option instructs the compiler to check that the size of a variable
12605 length array is positive.
12607 @item -fsanitize=null
12608 @opindex fsanitize=null
12609 This option enables pointer checking. Particularly, the application
12610 built with this option turned on will issue an error message when it
12611 tries to dereference a NULL pointer, or if a reference (possibly an
12612 rvalue reference) is bound to a NULL pointer, or if a method is invoked
12613 on an object pointed by a NULL pointer.
12615 @item -fsanitize=return
12616 @opindex fsanitize=return
12617 This option enables return statement checking. Programs
12618 built with this option turned on will issue an error message
12619 when the end of a non-void function is reached without actually
12620 returning a value. This option works in C++ only.
12622 @item -fsanitize=signed-integer-overflow
12623 @opindex fsanitize=signed-integer-overflow
12624 This option enables signed integer overflow checking. We check that
12625 the result of @code{+}, @code{*}, and both unary and binary @code{-}
12626 does not overflow in the signed arithmetics. Note, integer promotion
12627 rules must be taken into account. That is, the following is not an
12630 signed char a = SCHAR_MAX;
12634 @item -fsanitize=bounds
12635 @opindex fsanitize=bounds
12636 This option enables instrumentation of array bounds. Various out of bounds
12637 accesses are detected. Flexible array members, flexible array member-like
12638 arrays, and initializers of variables with static storage are not instrumented.
12640 @item -fsanitize=bounds-strict
12641 @opindex fsanitize=bounds-strict
12642 This option enables strict instrumentation of array bounds. Most out of bounds
12643 accesses are detected, including flexible array members and flexible array
12644 member-like arrays. Initializers of variables with static storage are not
12647 @item -fsanitize=alignment
12648 @opindex fsanitize=alignment
12650 This option enables checking of alignment of pointers when they are
12651 dereferenced, or when a reference is bound to insufficiently aligned target,
12652 or when a method or constructor is invoked on insufficiently aligned object.
12654 @item -fsanitize=object-size
12655 @opindex fsanitize=object-size
12656 This option enables instrumentation of memory references using the
12657 @code{__builtin_object_size} function. Various out of bounds pointer
12658 accesses are detected.
12660 @item -fsanitize=float-divide-by-zero
12661 @opindex fsanitize=float-divide-by-zero
12662 Detect floating-point division by zero. Unlike other similar options,
12663 @option{-fsanitize=float-divide-by-zero} is not enabled by
12664 @option{-fsanitize=undefined}, since floating-point division by zero can
12665 be a legitimate way of obtaining infinities and NaNs.
12667 @item -fsanitize=float-cast-overflow
12668 @opindex fsanitize=float-cast-overflow
12669 This option enables floating-point type to integer conversion checking.
12670 We check that the result of the conversion does not overflow.
12671 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
12672 not enabled by @option{-fsanitize=undefined}.
12673 This option does not work well with @code{FE_INVALID} exceptions enabled.
12675 @item -fsanitize=nonnull-attribute
12676 @opindex fsanitize=nonnull-attribute
12678 This option enables instrumentation of calls, checking whether null values
12679 are not passed to arguments marked as requiring a non-null value by the
12680 @code{nonnull} function attribute.
12682 @item -fsanitize=returns-nonnull-attribute
12683 @opindex fsanitize=returns-nonnull-attribute
12685 This option enables instrumentation of return statements in functions
12686 marked with @code{returns_nonnull} function attribute, to detect returning
12687 of null values from such functions.
12689 @item -fsanitize=bool
12690 @opindex fsanitize=bool
12692 This option enables instrumentation of loads from bool. If a value other
12693 than 0/1 is loaded, a run-time error is issued.
12695 @item -fsanitize=enum
12696 @opindex fsanitize=enum
12698 This option enables instrumentation of loads from an enum type. If
12699 a value outside the range of values for the enum type is loaded,
12700 a run-time error is issued.
12702 @item -fsanitize=vptr
12703 @opindex fsanitize=vptr
12705 This option enables instrumentation of C++ member function calls, member
12706 accesses and some conversions between pointers to base and derived classes,
12707 to verify the referenced object has the correct dynamic type.
12709 @item -fsanitize=pointer-overflow
12710 @opindex fsanitize=pointer-overflow
12712 This option enables instrumentation of pointer arithmetics. If the pointer
12713 arithmetics overflows, a run-time error is issued.
12715 @item -fsanitize=builtin
12716 @opindex fsanitize=builtin
12718 This option enables instrumentation of arguments to selected builtin
12719 functions. If an invalid value is passed to such arguments, a run-time
12720 error is issued. E.g.@ passing 0 as the argument to @code{__builtin_ctz}
12721 or @code{__builtin_clz} invokes undefined behavior and is diagnosed
12726 While @option{-ftrapv} causes traps for signed overflows to be emitted,
12727 @option{-fsanitize=undefined} gives a diagnostic message.
12728 This currently works only for the C family of languages.
12730 @item -fno-sanitize=all
12731 @opindex fno-sanitize=all
12733 This option disables all previously enabled sanitizers.
12734 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
12737 @item -fasan-shadow-offset=@var{number}
12738 @opindex fasan-shadow-offset
12739 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
12740 It is useful for experimenting with different shadow memory layouts in
12741 Kernel AddressSanitizer.
12743 @item -fsanitize-sections=@var{s1},@var{s2},...
12744 @opindex fsanitize-sections
12745 Sanitize global variables in selected user-defined sections. @var{si} may
12748 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
12749 @opindex fsanitize-recover
12750 @opindex fno-sanitize-recover
12751 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
12752 mentioned in comma-separated list of @var{opts}. Enabling this option
12753 for a sanitizer component causes it to attempt to continue
12754 running the program as if no error happened. This means multiple
12755 runtime errors can be reported in a single program run, and the exit
12756 code of the program may indicate success even when errors
12757 have been reported. The @option{-fno-sanitize-recover=} option
12758 can be used to alter
12759 this behavior: only the first detected error is reported
12760 and program then exits with a non-zero exit code.
12762 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
12763 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
12764 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
12765 @option{-fsanitize=bounds-strict},
12766 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
12767 For these sanitizers error recovery is turned on by default,
12768 except @option{-fsanitize=address}, for which this feature is experimental.
12769 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
12770 accepted, the former enables recovery for all sanitizers that support it,
12771 the latter disables recovery for all sanitizers that support it.
12773 Even if a recovery mode is turned on the compiler side, it needs to be also
12774 enabled on the runtime library side, otherwise the failures are still fatal.
12775 The runtime library defaults to @code{halt_on_error=0} for
12776 ThreadSanitizer and UndefinedBehaviorSanitizer, while default value for
12777 AddressSanitizer is @code{halt_on_error=1}. This can be overridden through
12778 setting the @code{halt_on_error} flag in the corresponding environment variable.
12780 Syntax without an explicit @var{opts} parameter is deprecated. It is
12781 equivalent to specifying an @var{opts} list of:
12784 undefined,float-cast-overflow,float-divide-by-zero,bounds-strict
12787 @item -fsanitize-address-use-after-scope
12788 @opindex fsanitize-address-use-after-scope
12789 Enable sanitization of local variables to detect use-after-scope bugs.
12790 The option sets @option{-fstack-reuse} to @samp{none}.
12792 @item -fsanitize-undefined-trap-on-error
12793 @opindex fsanitize-undefined-trap-on-error
12794 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
12795 report undefined behavior using @code{__builtin_trap} rather than
12796 a @code{libubsan} library routine. The advantage of this is that the
12797 @code{libubsan} library is not needed and is not linked in, so this
12798 is usable even in freestanding environments.
12800 @item -fsanitize-coverage=trace-pc
12801 @opindex fsanitize-coverage=trace-pc
12802 Enable coverage-guided fuzzing code instrumentation.
12803 Inserts a call to @code{__sanitizer_cov_trace_pc} into every basic block.
12805 @item -fsanitize-coverage=trace-cmp
12806 @opindex fsanitize-coverage=trace-cmp
12807 Enable dataflow guided fuzzing code instrumentation.
12808 Inserts a call to @code{__sanitizer_cov_trace_cmp1},
12809 @code{__sanitizer_cov_trace_cmp2}, @code{__sanitizer_cov_trace_cmp4} or
12810 @code{__sanitizer_cov_trace_cmp8} for integral comparison with both operands
12811 variable or @code{__sanitizer_cov_trace_const_cmp1},
12812 @code{__sanitizer_cov_trace_const_cmp2},
12813 @code{__sanitizer_cov_trace_const_cmp4} or
12814 @code{__sanitizer_cov_trace_const_cmp8} for integral comparison with one
12815 operand constant, @code{__sanitizer_cov_trace_cmpf} or
12816 @code{__sanitizer_cov_trace_cmpd} for float or double comparisons and
12817 @code{__sanitizer_cov_trace_switch} for switch statements.
12819 @item -fcf-protection=@r{[}full@r{|}branch@r{|}return@r{|}none@r{]}
12820 @opindex fcf-protection
12821 Enable code instrumentation of control-flow transfers to increase
12822 program security by checking that target addresses of control-flow
12823 transfer instructions (such as indirect function call, function return,
12824 indirect jump) are valid. This prevents diverting the flow of control
12825 to an unexpected target. This is intended to protect against such
12826 threats as Return-oriented Programming (ROP), and similarly
12827 call/jmp-oriented programming (COP/JOP).
12829 The value @code{branch} tells the compiler to implement checking of
12830 validity of control-flow transfer at the point of indirect branch
12831 instructions, i.e.@: call/jmp instructions. The value @code{return}
12832 implements checking of validity at the point of returning from a
12833 function. The value @code{full} is an alias for specifying both
12834 @code{branch} and @code{return}. The value @code{none} turns off
12837 The macro @code{__CET__} is defined when @option{-fcf-protection} is
12838 used. The first bit of @code{__CET__} is set to 1 for the value
12839 @code{branch} and the second bit of @code{__CET__} is set to 1 for
12842 You can also use the @code{nocf_check} attribute to identify
12843 which functions and calls should be skipped from instrumentation
12844 (@pxref{Function Attributes}).
12846 Currently the x86 GNU/Linux target provides an implementation based
12847 on Intel Control-flow Enforcement Technology (CET).
12849 @item -fstack-protector
12850 @opindex fstack-protector
12851 Emit extra code to check for buffer overflows, such as stack smashing
12852 attacks. This is done by adding a guard variable to functions with
12853 vulnerable objects. This includes functions that call @code{alloca}, and
12854 functions with buffers larger than 8 bytes. The guards are initialized
12855 when a function is entered and then checked when the function exits.
12856 If a guard check fails, an error message is printed and the program exits.
12858 @item -fstack-protector-all
12859 @opindex fstack-protector-all
12860 Like @option{-fstack-protector} except that all functions are protected.
12862 @item -fstack-protector-strong
12863 @opindex fstack-protector-strong
12864 Like @option{-fstack-protector} but includes additional functions to
12865 be protected --- those that have local array definitions, or have
12866 references to local frame addresses.
12868 @item -fstack-protector-explicit
12869 @opindex fstack-protector-explicit
12870 Like @option{-fstack-protector} but only protects those functions which
12871 have the @code{stack_protect} attribute.
12873 @item -fstack-check
12874 @opindex fstack-check
12875 Generate code to verify that you do not go beyond the boundary of the
12876 stack. You should specify this flag if you are running in an
12877 environment with multiple threads, but you only rarely need to specify it in
12878 a single-threaded environment since stack overflow is automatically
12879 detected on nearly all systems if there is only one stack.
12881 Note that this switch does not actually cause checking to be done; the
12882 operating system or the language runtime must do that. The switch causes
12883 generation of code to ensure that they see the stack being extended.
12885 You can additionally specify a string parameter: @samp{no} means no
12886 checking, @samp{generic} means force the use of old-style checking,
12887 @samp{specific} means use the best checking method and is equivalent
12888 to bare @option{-fstack-check}.
12890 Old-style checking is a generic mechanism that requires no specific
12891 target support in the compiler but comes with the following drawbacks:
12895 Modified allocation strategy for large objects: they are always
12896 allocated dynamically if their size exceeds a fixed threshold. Note this
12897 may change the semantics of some code.
12900 Fixed limit on the size of the static frame of functions: when it is
12901 topped by a particular function, stack checking is not reliable and
12902 a warning is issued by the compiler.
12905 Inefficiency: because of both the modified allocation strategy and the
12906 generic implementation, code performance is hampered.
12909 Note that old-style stack checking is also the fallback method for
12910 @samp{specific} if no target support has been added in the compiler.
12912 @samp{-fstack-check=} is designed for Ada's needs to detect infinite recursion
12913 and stack overflows. @samp{specific} is an excellent choice when compiling
12914 Ada code. It is not generally sufficient to protect against stack-clash
12915 attacks. To protect against those you want @samp{-fstack-clash-protection}.
12917 @item -fstack-clash-protection
12918 @opindex fstack-clash-protection
12919 Generate code to prevent stack clash style attacks. When this option is
12920 enabled, the compiler will only allocate one page of stack space at a time
12921 and each page is accessed immediately after allocation. Thus, it prevents
12922 allocations from jumping over any stack guard page provided by the
12925 Most targets do not fully support stack clash protection. However, on
12926 those targets @option{-fstack-clash-protection} will protect dynamic stack
12927 allocations. @option{-fstack-clash-protection} may also provide limited
12928 protection for static stack allocations if the target supports
12929 @option{-fstack-check=specific}.
12931 @item -fstack-limit-register=@var{reg}
12932 @itemx -fstack-limit-symbol=@var{sym}
12933 @itemx -fno-stack-limit
12934 @opindex fstack-limit-register
12935 @opindex fstack-limit-symbol
12936 @opindex fno-stack-limit
12937 Generate code to ensure that the stack does not grow beyond a certain value,
12938 either the value of a register or the address of a symbol. If a larger
12939 stack is required, a signal is raised at run time. For most targets,
12940 the signal is raised before the stack overruns the boundary, so
12941 it is possible to catch the signal without taking special precautions.
12943 For instance, if the stack starts at absolute address @samp{0x80000000}
12944 and grows downwards, you can use the flags
12945 @option{-fstack-limit-symbol=__stack_limit} and
12946 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
12947 of 128KB@. Note that this may only work with the GNU linker.
12949 You can locally override stack limit checking by using the
12950 @code{no_stack_limit} function attribute (@pxref{Function Attributes}).
12952 @item -fsplit-stack
12953 @opindex fsplit-stack
12954 Generate code to automatically split the stack before it overflows.
12955 The resulting program has a discontiguous stack which can only
12956 overflow if the program is unable to allocate any more memory. This
12957 is most useful when running threaded programs, as it is no longer
12958 necessary to calculate a good stack size to use for each thread. This
12959 is currently only implemented for the x86 targets running
12962 When code compiled with @option{-fsplit-stack} calls code compiled
12963 without @option{-fsplit-stack}, there may not be much stack space
12964 available for the latter code to run. If compiling all code,
12965 including library code, with @option{-fsplit-stack} is not an option,
12966 then the linker can fix up these calls so that the code compiled
12967 without @option{-fsplit-stack} always has a large stack. Support for
12968 this is implemented in the gold linker in GNU binutils release 2.21
12971 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
12972 @opindex fvtable-verify
12973 This option is only available when compiling C++ code.
12974 It turns on (or off, if using @option{-fvtable-verify=none}) the security
12975 feature that verifies at run time, for every virtual call, that
12976 the vtable pointer through which the call is made is valid for the type of
12977 the object, and has not been corrupted or overwritten. If an invalid vtable
12978 pointer is detected at run time, an error is reported and execution of the
12979 program is immediately halted.
12981 This option causes run-time data structures to be built at program startup,
12982 which are used for verifying the vtable pointers.
12983 The options @samp{std} and @samp{preinit}
12984 control the timing of when these data structures are built. In both cases the
12985 data structures are built before execution reaches @code{main}. Using
12986 @option{-fvtable-verify=std} causes the data structures to be built after
12987 shared libraries have been loaded and initialized.
12988 @option{-fvtable-verify=preinit} causes them to be built before shared
12989 libraries have been loaded and initialized.
12991 If this option appears multiple times in the command line with different
12992 values specified, @samp{none} takes highest priority over both @samp{std} and
12993 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
12996 @opindex fvtv-debug
12997 When used in conjunction with @option{-fvtable-verify=std} or
12998 @option{-fvtable-verify=preinit}, causes debug versions of the
12999 runtime functions for the vtable verification feature to be called.
13000 This flag also causes the compiler to log information about which
13001 vtable pointers it finds for each class.
13002 This information is written to a file named @file{vtv_set_ptr_data.log}
13003 in the directory named by the environment variable @env{VTV_LOGS_DIR}
13004 if that is defined or the current working directory otherwise.
13006 Note: This feature @emph{appends} data to the log file. If you want a fresh log
13007 file, be sure to delete any existing one.
13010 @opindex fvtv-counts
13011 This is a debugging flag. When used in conjunction with
13012 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
13013 causes the compiler to keep track of the total number of virtual calls
13014 it encounters and the number of verifications it inserts. It also
13015 counts the number of calls to certain run-time library functions
13016 that it inserts and logs this information for each compilation unit.
13017 The compiler writes this information to a file named
13018 @file{vtv_count_data.log} in the directory named by the environment
13019 variable @env{VTV_LOGS_DIR} if that is defined or the current working
13020 directory otherwise. It also counts the size of the vtable pointer sets
13021 for each class, and writes this information to @file{vtv_class_set_sizes.log}
13022 in the same directory.
13024 Note: This feature @emph{appends} data to the log files. To get fresh log
13025 files, be sure to delete any existing ones.
13027 @item -finstrument-functions
13028 @opindex finstrument-functions
13029 Generate instrumentation calls for entry and exit to functions. Just
13030 after function entry and just before function exit, the following
13031 profiling functions are called with the address of the current
13032 function and its call site. (On some platforms,
13033 @code{__builtin_return_address} does not work beyond the current
13034 function, so the call site information may not be available to the
13035 profiling functions otherwise.)
13038 void __cyg_profile_func_enter (void *this_fn,
13040 void __cyg_profile_func_exit (void *this_fn,
13044 The first argument is the address of the start of the current function,
13045 which may be looked up exactly in the symbol table.
13047 This instrumentation is also done for functions expanded inline in other
13048 functions. The profiling calls indicate where, conceptually, the
13049 inline function is entered and exited. This means that addressable
13050 versions of such functions must be available. If all your uses of a
13051 function are expanded inline, this may mean an additional expansion of
13052 code size. If you use @code{extern inline} in your C code, an
13053 addressable version of such functions must be provided. (This is
13054 normally the case anyway, but if you get lucky and the optimizer always
13055 expands the functions inline, you might have gotten away without
13056 providing static copies.)
13058 A function may be given the attribute @code{no_instrument_function}, in
13059 which case this instrumentation is not done. This can be used, for
13060 example, for the profiling functions listed above, high-priority
13061 interrupt routines, and any functions from which the profiling functions
13062 cannot safely be called (perhaps signal handlers, if the profiling
13063 routines generate output or allocate memory).
13064 @xref{Common Function Attributes}.
13066 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
13067 @opindex finstrument-functions-exclude-file-list
13069 Set the list of functions that are excluded from instrumentation (see
13070 the description of @option{-finstrument-functions}). If the file that
13071 contains a function definition matches with one of @var{file}, then
13072 that function is not instrumented. The match is done on substrings:
13073 if the @var{file} parameter is a substring of the file name, it is
13074 considered to be a match.
13079 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
13083 excludes any inline function defined in files whose pathnames
13084 contain @file{/bits/stl} or @file{include/sys}.
13086 If, for some reason, you want to include letter @samp{,} in one of
13087 @var{sym}, write @samp{\,}. For example,
13088 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
13089 (note the single quote surrounding the option).
13091 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
13092 @opindex finstrument-functions-exclude-function-list
13094 This is similar to @option{-finstrument-functions-exclude-file-list},
13095 but this option sets the list of function names to be excluded from
13096 instrumentation. The function name to be matched is its user-visible
13097 name, such as @code{vector<int> blah(const vector<int> &)}, not the
13098 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
13099 match is done on substrings: if the @var{sym} parameter is a substring
13100 of the function name, it is considered to be a match. For C99 and C++
13101 extended identifiers, the function name must be given in UTF-8, not
13102 using universal character names.
13104 @item -fpatchable-function-entry=@var{N}[,@var{M}]
13105 @opindex fpatchable-function-entry
13106 Generate @var{N} NOPs right at the beginning
13107 of each function, with the function entry point before the @var{M}th NOP.
13108 If @var{M} is omitted, it defaults to @code{0} so the
13109 function entry points to the address just at the first NOP.
13110 The NOP instructions reserve extra space which can be used to patch in
13111 any desired instrumentation at run time, provided that the code segment
13112 is writable. The amount of space is controllable indirectly via
13113 the number of NOPs; the NOP instruction used corresponds to the instruction
13114 emitted by the internal GCC back-end interface @code{gen_nop}. This behavior
13115 is target-specific and may also depend on the architecture variant and/or
13116 other compilation options.
13118 For run-time identification, the starting addresses of these areas,
13119 which correspond to their respective function entries minus @var{M},
13120 are additionally collected in the @code{__patchable_function_entries}
13121 section of the resulting binary.
13123 Note that the value of @code{__attribute__ ((patchable_function_entry
13124 (N,M)))} takes precedence over command-line option
13125 @option{-fpatchable-function-entry=N,M}. This can be used to increase
13126 the area size or to remove it completely on a single function.
13127 If @code{N=0}, no pad location is recorded.
13129 The NOP instructions are inserted at---and maybe before, depending on
13130 @var{M}---the function entry address, even before the prologue.
13135 @node Preprocessor Options
13136 @section Options Controlling the Preprocessor
13137 @cindex preprocessor options
13138 @cindex options, preprocessor
13140 These options control the C preprocessor, which is run on each C source
13141 file before actual compilation.
13143 If you use the @option{-E} option, nothing is done except preprocessing.
13144 Some of these options make sense only together with @option{-E} because
13145 they cause the preprocessor output to be unsuitable for actual
13148 In addition to the options listed here, there are a number of options
13149 to control search paths for include files documented in
13150 @ref{Directory Options}.
13151 Options to control preprocessor diagnostics are listed in
13152 @ref{Warning Options}.
13155 @include cppopts.texi
13157 @item -Wp,@var{option}
13159 You can use @option{-Wp,@var{option}} to bypass the compiler driver
13160 and pass @var{option} directly through to the preprocessor. If
13161 @var{option} contains commas, it is split into multiple options at the
13162 commas. However, many options are modified, translated or interpreted
13163 by the compiler driver before being passed to the preprocessor, and
13164 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
13165 interface is undocumented and subject to change, so whenever possible
13166 you should avoid using @option{-Wp} and let the driver handle the
13169 @item -Xpreprocessor @var{option}
13170 @opindex Xpreprocessor
13171 Pass @var{option} as an option to the preprocessor. You can use this to
13172 supply system-specific preprocessor options that GCC does not
13175 If you want to pass an option that takes an argument, you must use
13176 @option{-Xpreprocessor} twice, once for the option and once for the argument.
13178 @item -no-integrated-cpp
13179 @opindex no-integrated-cpp
13180 Perform preprocessing as a separate pass before compilation.
13181 By default, GCC performs preprocessing as an integrated part of
13182 input tokenization and parsing.
13183 If this option is provided, the appropriate language front end
13184 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
13185 and Objective-C, respectively) is instead invoked twice,
13186 once for preprocessing only and once for actual compilation
13187 of the preprocessed input.
13188 This option may be useful in conjunction with the @option{-B} or
13189 @option{-wrapper} options to specify an alternate preprocessor or
13190 perform additional processing of the program source between
13191 normal preprocessing and compilation.
13195 @node Assembler Options
13196 @section Passing Options to the Assembler
13198 @c prevent bad page break with this line
13199 You can pass options to the assembler.
13202 @item -Wa,@var{option}
13204 Pass @var{option} as an option to the assembler. If @var{option}
13205 contains commas, it is split into multiple options at the commas.
13207 @item -Xassembler @var{option}
13208 @opindex Xassembler
13209 Pass @var{option} as an option to the assembler. You can use this to
13210 supply system-specific assembler options that GCC does not
13213 If you want to pass an option that takes an argument, you must use
13214 @option{-Xassembler} twice, once for the option and once for the argument.
13219 @section Options for Linking
13220 @cindex link options
13221 @cindex options, linking
13223 These options come into play when the compiler links object files into
13224 an executable output file. They are meaningless if the compiler is
13225 not doing a link step.
13229 @item @var{object-file-name}
13230 A file name that does not end in a special recognized suffix is
13231 considered to name an object file or library. (Object files are
13232 distinguished from libraries by the linker according to the file
13233 contents.) If linking is done, these object files are used as input
13242 If any of these options is used, then the linker is not run, and
13243 object file names should not be used as arguments. @xref{Overall
13246 @item -flinker-output=@var{type}
13247 @opindex flinker-output
13248 This option controls code generation of the link-time optimizer. By
13249 default the linker output is automatically determined by the linker
13250 plugin. For debugging the compiler and if incremental linking with a
13251 non-LTO object file is desired, it may be useful to control the type
13254 If @var{type} is @samp{exec}, code generation produces a static
13255 binary. In this case @option{-fpic} and @option{-fpie} are both
13258 If @var{type} is @samp{dyn}, code generation produces a shared
13259 library. In this case @option{-fpic} or @option{-fPIC} is preserved,
13260 but not enabled automatically. This allows to build shared libraries
13261 without position-independent code on architectures where this is
13262 possible, i.e.@: on x86.
13264 If @var{type} is @samp{pie}, code generation produces an @option{-fpie}
13265 executable. This results in similar optimizations as @samp{exec}
13266 except that @option{-fpie} is not disabled if specified at compilation
13269 If @var{type} is @samp{rel}, the compiler assumes that incremental linking is
13270 done. The sections containing intermediate code for link-time optimization are
13271 merged, pre-optimized, and output to the resulting object file. In addition, if
13272 @option{-ffat-lto-objects} is specified, binary code is produced for future
13273 non-LTO linking. The object file produced by incremental linking is smaller
13274 than a static library produced from the same object files. At link time the
13275 result of incremental linking also loads faster than a static
13276 library assuming that the majority of objects in the library are used.
13278 Finally @samp{nolto-rel} configures the compiler for incremental linking where
13279 code generation is forced, a final binary is produced, and the intermediate
13280 code for later link-time optimization is stripped. When multiple object files
13281 are linked together the resulting code is better optimized than with
13282 link-time optimizations disabled (for example, cross-module inlining
13283 happens), but most of benefits of whole program optimizations are lost.
13285 During the incremental link (by @option{-r}) the linker plugin defaults to
13286 @option{rel}. With current interfaces to GNU Binutils it is however not
13287 possible to incrementally link LTO objects and non-LTO objects into a single
13288 mixed object file. If any of object files in incremental link cannot
13289 be used for link-time optimization, the linker plugin issues a warning and
13290 uses @samp{nolto-rel}. To maintain whole program optimization, it is
13291 recommended to link such objects into static library instead. Alternatively it
13292 is possible to use H.J. Lu's binutils with support for mixed objects.
13295 @opindex fuse-ld=bfd
13296 Use the @command{bfd} linker instead of the default linker.
13298 @item -fuse-ld=gold
13299 @opindex fuse-ld=gold
13300 Use the @command{gold} linker instead of the default linker.
13303 @opindex fuse-ld=lld
13304 Use the LLVM @command{lld} linker instead of the default linker.
13307 @item -l@var{library}
13308 @itemx -l @var{library}
13310 Search the library named @var{library} when linking. (The second
13311 alternative with the library as a separate argument is only for
13312 POSIX compliance and is not recommended.)
13314 The @option{-l} option is passed directly to the linker by GCC. Refer
13315 to your linker documentation for exact details. The general
13316 description below applies to the GNU linker.
13318 The linker searches a standard list of directories for the library.
13319 The directories searched include several standard system directories
13320 plus any that you specify with @option{-L}.
13322 Static libraries are archives of object files, and have file names
13323 like @file{lib@var{library}.a}. Some targets also support shared
13324 libraries, which typically have names like @file{lib@var{library}.so}.
13325 If both static and shared libraries are found, the linker gives
13326 preference to linking with the shared library unless the
13327 @option{-static} option is used.
13329 It makes a difference where in the command you write this option; the
13330 linker searches and processes libraries and object files in the order they
13331 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
13332 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
13333 to functions in @samp{z}, those functions may not be loaded.
13337 You need this special case of the @option{-l} option in order to
13338 link an Objective-C or Objective-C++ program.
13340 @item -nostartfiles
13341 @opindex nostartfiles
13342 Do not use the standard system startup files when linking.
13343 The standard system libraries are used normally, unless @option{-nostdlib},
13344 @option{-nolibc}, or @option{-nodefaultlibs} is used.
13346 @item -nodefaultlibs
13347 @opindex nodefaultlibs
13348 Do not use the standard system libraries when linking.
13349 Only the libraries you specify are passed to the linker, and options
13350 specifying linkage of the system libraries, such as @option{-static-libgcc}
13351 or @option{-shared-libgcc}, are ignored.
13352 The standard startup files are used normally, unless @option{-nostartfiles}
13355 The compiler may generate calls to @code{memcmp},
13356 @code{memset}, @code{memcpy} and @code{memmove}.
13357 These entries are usually resolved by entries in
13358 libc. These entry points should be supplied through some other
13359 mechanism when this option is specified.
13363 Do not use the C library or system libraries tightly coupled with it when
13364 linking. Still link with the startup files, @file{libgcc} or toolchain
13365 provided language support libraries such as @file{libgnat}, @file{libgfortran}
13366 or @file{libstdc++} unless options preventing their inclusion are used as
13367 well. This typically removes @option{-lc} from the link command line, as well
13368 as system libraries that normally go with it and become meaningless when
13369 absence of a C library is assumed, for example @option{-lpthread} or
13370 @option{-lm} in some configurations. This is intended for bare-board
13371 targets when there is indeed no C library available.
13375 Do not use the standard system startup files or libraries when linking.
13376 No startup files and only the libraries you specify are passed to
13377 the linker, and options specifying linkage of the system libraries, such as
13378 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
13380 The compiler may generate calls to @code{memcmp}, @code{memset},
13381 @code{memcpy} and @code{memmove}.
13382 These entries are usually resolved by entries in
13383 libc. These entry points should be supplied through some other
13384 mechanism when this option is specified.
13386 @cindex @option{-lgcc}, use with @option{-nostdlib}
13387 @cindex @option{-nostdlib} and unresolved references
13388 @cindex unresolved references and @option{-nostdlib}
13389 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
13390 @cindex @option{-nodefaultlibs} and unresolved references
13391 @cindex unresolved references and @option{-nodefaultlibs}
13392 One of the standard libraries bypassed by @option{-nostdlib} and
13393 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
13394 which GCC uses to overcome shortcomings of particular machines, or special
13395 needs for some languages.
13396 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
13397 Collection (GCC) Internals},
13398 for more discussion of @file{libgcc.a}.)
13399 In most cases, you need @file{libgcc.a} even when you want to avoid
13400 other standard libraries. In other words, when you specify @option{-nostdlib}
13401 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
13402 This ensures that you have no unresolved references to internal GCC
13403 library subroutines.
13404 (An example of such an internal subroutine is @code{__main}, used to ensure C++
13405 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
13406 GNU Compiler Collection (GCC) Internals}.)
13408 @item -e @var{entry}
13409 @itemx --entry=@var{entry}
13413 Specify that the program entry point is @var{entry}. The argument is
13414 interpreted by the linker; the GNU linker accepts either a symbol name
13419 Produce a dynamically linked position independent executable on targets
13420 that support it. For predictable results, you must also specify the same
13421 set of options used for compilation (@option{-fpie}, @option{-fPIE},
13422 or model suboptions) when you specify this linker option.
13426 Don't produce a dynamically linked position independent executable.
13429 @opindex static-pie
13430 Produce a static position independent executable on targets that support
13431 it. A static position independent executable is similar to a static
13432 executable, but can be loaded at any address without a dynamic linker.
13433 For predictable results, you must also specify the same set of options
13434 used for compilation (@option{-fpie}, @option{-fPIE}, or model
13435 suboptions) when you specify this linker option.
13439 Link with the POSIX threads library. This option is supported on
13440 GNU/Linux targets, most other Unix derivatives, and also on
13441 x86 Cygwin and MinGW targets. On some targets this option also sets
13442 flags for the preprocessor, so it should be used consistently for both
13443 compilation and linking.
13447 Produce a relocatable object as output. This is also known as partial
13452 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
13453 that support it. This instructs the linker to add all symbols, not
13454 only used ones, to the dynamic symbol table. This option is needed
13455 for some uses of @code{dlopen} or to allow obtaining backtraces
13456 from within a program.
13460 Remove all symbol table and relocation information from the executable.
13464 On systems that support dynamic linking, this overrides @option{-pie}
13465 and prevents linking with the shared libraries. On other systems, this
13466 option has no effect.
13470 Produce a shared object which can then be linked with other objects to
13471 form an executable. Not all systems support this option. For predictable
13472 results, you must also specify the same set of options used for compilation
13473 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
13474 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
13475 needs to build supplementary stub code for constructors to work. On
13476 multi-libbed systems, @samp{gcc -shared} must select the correct support
13477 libraries to link against. Failing to supply the correct flags may lead
13478 to subtle defects. Supplying them in cases where they are not necessary
13481 @item -shared-libgcc
13482 @itemx -static-libgcc
13483 @opindex shared-libgcc
13484 @opindex static-libgcc
13485 On systems that provide @file{libgcc} as a shared library, these options
13486 force the use of either the shared or static version, respectively.
13487 If no shared version of @file{libgcc} was built when the compiler was
13488 configured, these options have no effect.
13490 There are several situations in which an application should use the
13491 shared @file{libgcc} instead of the static version. The most common
13492 of these is when the application wishes to throw and catch exceptions
13493 across different shared libraries. In that case, each of the libraries
13494 as well as the application itself should use the shared @file{libgcc}.
13496 Therefore, the G++ driver automatically adds @option{-shared-libgcc}
13497 whenever you build a shared library or a main executable, because C++
13498 programs typically use exceptions, so this is the right thing to do.
13500 If, instead, you use the GCC driver to create shared libraries, you may
13501 find that they are not always linked with the shared @file{libgcc}.
13502 If GCC finds, at its configuration time, that you have a non-GNU linker
13503 or a GNU linker that does not support option @option{--eh-frame-hdr},
13504 it links the shared version of @file{libgcc} into shared libraries
13505 by default. Otherwise, it takes advantage of the linker and optimizes
13506 away the linking with the shared version of @file{libgcc}, linking with
13507 the static version of libgcc by default. This allows exceptions to
13508 propagate through such shared libraries, without incurring relocation
13509 costs at library load time.
13511 However, if a library or main executable is supposed to throw or catch
13512 exceptions, you must link it using the G++ driver, or using the option
13513 @option{-shared-libgcc}, such that it is linked with the shared
13516 @item -static-libasan
13517 @opindex static-libasan
13518 When the @option{-fsanitize=address} option is used to link a program,
13519 the GCC driver automatically links against @option{libasan}. If
13520 @file{libasan} is available as a shared library, and the @option{-static}
13521 option is not used, then this links against the shared version of
13522 @file{libasan}. The @option{-static-libasan} option directs the GCC
13523 driver to link @file{libasan} statically, without necessarily linking
13524 other libraries statically.
13526 @item -static-libtsan
13527 @opindex static-libtsan
13528 When the @option{-fsanitize=thread} option is used to link a program,
13529 the GCC driver automatically links against @option{libtsan}. If
13530 @file{libtsan} is available as a shared library, and the @option{-static}
13531 option is not used, then this links against the shared version of
13532 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
13533 driver to link @file{libtsan} statically, without necessarily linking
13534 other libraries statically.
13536 @item -static-liblsan
13537 @opindex static-liblsan
13538 When the @option{-fsanitize=leak} option is used to link a program,
13539 the GCC driver automatically links against @option{liblsan}. If
13540 @file{liblsan} is available as a shared library, and the @option{-static}
13541 option is not used, then this links against the shared version of
13542 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
13543 driver to link @file{liblsan} statically, without necessarily linking
13544 other libraries statically.
13546 @item -static-libubsan
13547 @opindex static-libubsan
13548 When the @option{-fsanitize=undefined} option is used to link a program,
13549 the GCC driver automatically links against @option{libubsan}. If
13550 @file{libubsan} is available as a shared library, and the @option{-static}
13551 option is not used, then this links against the shared version of
13552 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
13553 driver to link @file{libubsan} statically, without necessarily linking
13554 other libraries statically.
13556 @item -static-libstdc++
13557 @opindex static-libstdc++
13558 When the @command{g++} program is used to link a C++ program, it
13559 normally automatically links against @option{libstdc++}. If
13560 @file{libstdc++} is available as a shared library, and the
13561 @option{-static} option is not used, then this links against the
13562 shared version of @file{libstdc++}. That is normally fine. However, it
13563 is sometimes useful to freeze the version of @file{libstdc++} used by
13564 the program without going all the way to a fully static link. The
13565 @option{-static-libstdc++} option directs the @command{g++} driver to
13566 link @file{libstdc++} statically, without necessarily linking other
13567 libraries statically.
13571 Bind references to global symbols when building a shared object. Warn
13572 about any unresolved references (unless overridden by the link editor
13573 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
13576 @item -T @var{script}
13578 @cindex linker script
13579 Use @var{script} as the linker script. This option is supported by most
13580 systems using the GNU linker. On some targets, such as bare-board
13581 targets without an operating system, the @option{-T} option may be required
13582 when linking to avoid references to undefined symbols.
13584 @item -Xlinker @var{option}
13586 Pass @var{option} as an option to the linker. You can use this to
13587 supply system-specific linker options that GCC does not recognize.
13589 If you want to pass an option that takes a separate argument, you must use
13590 @option{-Xlinker} twice, once for the option and once for the argument.
13591 For example, to pass @option{-assert definitions}, you must write
13592 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
13593 @option{-Xlinker "-assert definitions"}, because this passes the entire
13594 string as a single argument, which is not what the linker expects.
13596 When using the GNU linker, it is usually more convenient to pass
13597 arguments to linker options using the @option{@var{option}=@var{value}}
13598 syntax than as separate arguments. For example, you can specify
13599 @option{-Xlinker -Map=output.map} rather than
13600 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
13601 this syntax for command-line options.
13603 @item -Wl,@var{option}
13605 Pass @var{option} as an option to the linker. If @var{option} contains
13606 commas, it is split into multiple options at the commas. You can use this
13607 syntax to pass an argument to the option.
13608 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
13609 linker. When using the GNU linker, you can also get the same effect with
13610 @option{-Wl,-Map=output.map}.
13612 @item -u @var{symbol}
13614 Pretend the symbol @var{symbol} is undefined, to force linking of
13615 library modules to define it. You can use @option{-u} multiple times with
13616 different symbols to force loading of additional library modules.
13618 @item -z @var{keyword}
13620 @option{-z} is passed directly on to the linker along with the keyword
13621 @var{keyword}. See the section in the documentation of your linker for
13622 permitted values and their meanings.
13625 @node Directory Options
13626 @section Options for Directory Search
13627 @cindex directory options
13628 @cindex options, directory search
13629 @cindex search path
13631 These options specify directories to search for header files, for
13632 libraries and for parts of the compiler:
13635 @include cppdiropts.texi
13637 @item -iplugindir=@var{dir}
13638 @opindex iplugindir=
13639 Set the directory to search for plugins that are passed
13640 by @option{-fplugin=@var{name}} instead of
13641 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
13642 to be used by the user, but only passed by the driver.
13646 Add directory @var{dir} to the list of directories to be searched
13649 @item -B@var{prefix}
13651 This option specifies where to find the executables, libraries,
13652 include files, and data files of the compiler itself.
13654 The compiler driver program runs one or more of the subprograms
13655 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
13656 @var{prefix} as a prefix for each program it tries to run, both with and
13657 without @samp{@var{machine}/@var{version}/} for the corresponding target
13658 machine and compiler version.
13660 For each subprogram to be run, the compiler driver first tries the
13661 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
13662 is not specified, the driver tries two standard prefixes,
13663 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
13664 those results in a file name that is found, the unmodified program
13665 name is searched for using the directories specified in your
13666 @env{PATH} environment variable.
13668 The compiler checks to see if the path provided by @option{-B}
13669 refers to a directory, and if necessary it adds a directory
13670 separator character at the end of the path.
13672 @option{-B} prefixes that effectively specify directory names also apply
13673 to libraries in the linker, because the compiler translates these
13674 options into @option{-L} options for the linker. They also apply to
13675 include files in the preprocessor, because the compiler translates these
13676 options into @option{-isystem} options for the preprocessor. In this case,
13677 the compiler appends @samp{include} to the prefix.
13679 The runtime support file @file{libgcc.a} can also be searched for using
13680 the @option{-B} prefix, if needed. If it is not found there, the two
13681 standard prefixes above are tried, and that is all. The file is left
13682 out of the link if it is not found by those means.
13684 Another way to specify a prefix much like the @option{-B} prefix is to use
13685 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
13688 As a special kludge, if the path provided by @option{-B} is
13689 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
13690 9, then it is replaced by @file{[dir/]include}. This is to help
13691 with boot-strapping the compiler.
13693 @item -no-canonical-prefixes
13694 @opindex no-canonical-prefixes
13695 Do not expand any symbolic links, resolve references to @samp{/../}
13696 or @samp{/./}, or make the path absolute when generating a relative
13699 @item --sysroot=@var{dir}
13701 Use @var{dir} as the logical root directory for headers and libraries.
13702 For example, if the compiler normally searches for headers in
13703 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
13704 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
13706 If you use both this option and the @option{-isysroot} option, then
13707 the @option{--sysroot} option applies to libraries, but the
13708 @option{-isysroot} option applies to header files.
13710 The GNU linker (beginning with version 2.16) has the necessary support
13711 for this option. If your linker does not support this option, the
13712 header file aspect of @option{--sysroot} still works, but the
13713 library aspect does not.
13715 @item --no-sysroot-suffix
13716 @opindex no-sysroot-suffix
13717 For some targets, a suffix is added to the root directory specified
13718 with @option{--sysroot}, depending on the other options used, so that
13719 headers may for example be found in
13720 @file{@var{dir}/@var{suffix}/usr/include} instead of
13721 @file{@var{dir}/usr/include}. This option disables the addition of
13726 @node Code Gen Options
13727 @section Options for Code Generation Conventions
13728 @cindex code generation conventions
13729 @cindex options, code generation
13730 @cindex run-time options
13732 These machine-independent options control the interface conventions
13733 used in code generation.
13735 Most of them have both positive and negative forms; the negative form
13736 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
13737 one of the forms is listed---the one that is not the default. You
13738 can figure out the other form by either removing @samp{no-} or adding
13742 @item -fstack-reuse=@var{reuse-level}
13743 @opindex fstack_reuse
13744 This option controls stack space reuse for user declared local/auto variables
13745 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
13746 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
13747 local variables and temporaries, @samp{named_vars} enables the reuse only for
13748 user defined local variables with names, and @samp{none} disables stack reuse
13749 completely. The default value is @samp{all}. The option is needed when the
13750 program extends the lifetime of a scoped local variable or a compiler generated
13751 temporary beyond the end point defined by the language. When a lifetime of
13752 a variable ends, and if the variable lives in memory, the optimizing compiler
13753 has the freedom to reuse its stack space with other temporaries or scoped
13754 local variables whose live range does not overlap with it. Legacy code extending
13755 local lifetime is likely to break with the stack reuse optimization.
13774 if (*p == 10) // out of scope use of local1
13785 A(int k) : i(k), j(k) @{ @}
13792 void foo(const A& ar)
13799 foo(A(10)); // temp object's lifetime ends when foo returns
13805 ap->i+= 10; // ap references out of scope temp whose space
13806 // is reused with a. What is the value of ap->i?
13811 The lifetime of a compiler generated temporary is well defined by the C++
13812 standard. When a lifetime of a temporary ends, and if the temporary lives
13813 in memory, the optimizing compiler has the freedom to reuse its stack
13814 space with other temporaries or scoped local variables whose live range
13815 does not overlap with it. However some of the legacy code relies on
13816 the behavior of older compilers in which temporaries' stack space is
13817 not reused, the aggressive stack reuse can lead to runtime errors. This
13818 option is used to control the temporary stack reuse optimization.
13822 This option generates traps for signed overflow on addition, subtraction,
13823 multiplication operations.
13824 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
13825 @option{-ftrapv} @option{-fwrapv} on the command-line results in
13826 @option{-fwrapv} being effective. Note that only active options override, so
13827 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
13828 results in @option{-ftrapv} being effective.
13832 This option instructs the compiler to assume that signed arithmetic
13833 overflow of addition, subtraction and multiplication wraps around
13834 using twos-complement representation. This flag enables some optimizations
13835 and disables others.
13836 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
13837 @option{-ftrapv} @option{-fwrapv} on the command-line results in
13838 @option{-fwrapv} being effective. Note that only active options override, so
13839 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
13840 results in @option{-ftrapv} being effective.
13842 @item -fwrapv-pointer
13843 @opindex fwrapv-pointer
13844 This option instructs the compiler to assume that pointer arithmetic
13845 overflow on addition and subtraction wraps around using twos-complement
13846 representation. This flag disables some optimizations which assume
13847 pointer overflow is invalid.
13849 @item -fstrict-overflow
13850 @opindex fstrict-overflow
13851 This option implies @option{-fno-wrapv} @option{-fno-wrapv-pointer} and when
13852 negated implies @option{-fwrapv} @option{-fwrapv-pointer}.
13855 @opindex fexceptions
13856 Enable exception handling. Generates extra code needed to propagate
13857 exceptions. For some targets, this implies GCC generates frame
13858 unwind information for all functions, which can produce significant data
13859 size overhead, although it does not affect execution. If you do not
13860 specify this option, GCC enables it by default for languages like
13861 C++ that normally require exception handling, and disables it for
13862 languages like C that do not normally require it. However, you may need
13863 to enable this option when compiling C code that needs to interoperate
13864 properly with exception handlers written in C++. You may also wish to
13865 disable this option if you are compiling older C++ programs that don't
13866 use exception handling.
13868 @item -fnon-call-exceptions
13869 @opindex fnon-call-exceptions
13870 Generate code that allows trapping instructions to throw exceptions.
13871 Note that this requires platform-specific runtime support that does
13872 not exist everywhere. Moreover, it only allows @emph{trapping}
13873 instructions to throw exceptions, i.e.@: memory references or floating-point
13874 instructions. It does not allow exceptions to be thrown from
13875 arbitrary signal handlers such as @code{SIGALRM}.
13877 @item -fdelete-dead-exceptions
13878 @opindex fdelete-dead-exceptions
13879 Consider that instructions that may throw exceptions but don't otherwise
13880 contribute to the execution of the program can be optimized away.
13881 This option is enabled by default for the Ada front end, as permitted by
13882 the Ada language specification.
13883 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
13885 @item -funwind-tables
13886 @opindex funwind-tables
13887 Similar to @option{-fexceptions}, except that it just generates any needed
13888 static data, but does not affect the generated code in any other way.
13889 You normally do not need to enable this option; instead, a language processor
13890 that needs this handling enables it on your behalf.
13892 @item -fasynchronous-unwind-tables
13893 @opindex fasynchronous-unwind-tables
13894 Generate unwind table in DWARF format, if supported by target machine. The
13895 table is exact at each instruction boundary, so it can be used for stack
13896 unwinding from asynchronous events (such as debugger or garbage collector).
13898 @item -fno-gnu-unique
13899 @opindex fno-gnu-unique
13900 @opindex fgnu-unique
13901 On systems with recent GNU assembler and C library, the C++ compiler
13902 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
13903 of template static data members and static local variables in inline
13904 functions are unique even in the presence of @code{RTLD_LOCAL}; this
13905 is necessary to avoid problems with a library used by two different
13906 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
13907 therefore disagreeing with the other one about the binding of the
13908 symbol. But this causes @code{dlclose} to be ignored for affected
13909 DSOs; if your program relies on reinitialization of a DSO via
13910 @code{dlclose} and @code{dlopen}, you can use
13911 @option{-fno-gnu-unique}.
13913 @item -fpcc-struct-return
13914 @opindex fpcc-struct-return
13915 Return ``short'' @code{struct} and @code{union} values in memory like
13916 longer ones, rather than in registers. This convention is less
13917 efficient, but it has the advantage of allowing intercallability between
13918 GCC-compiled files and files compiled with other compilers, particularly
13919 the Portable C Compiler (pcc).
13921 The precise convention for returning structures in memory depends
13922 on the target configuration macros.
13924 Short structures and unions are those whose size and alignment match
13925 that of some integer type.
13927 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
13928 switch is not binary compatible with code compiled with the
13929 @option{-freg-struct-return} switch.
13930 Use it to conform to a non-default application binary interface.
13932 @item -freg-struct-return
13933 @opindex freg-struct-return
13934 Return @code{struct} and @code{union} values in registers when possible.
13935 This is more efficient for small structures than
13936 @option{-fpcc-struct-return}.
13938 If you specify neither @option{-fpcc-struct-return} nor
13939 @option{-freg-struct-return}, GCC defaults to whichever convention is
13940 standard for the target. If there is no standard convention, GCC
13941 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
13942 the principal compiler. In those cases, we can choose the standard, and
13943 we chose the more efficient register return alternative.
13945 @strong{Warning:} code compiled with the @option{-freg-struct-return}
13946 switch is not binary compatible with code compiled with the
13947 @option{-fpcc-struct-return} switch.
13948 Use it to conform to a non-default application binary interface.
13950 @item -fshort-enums
13951 @opindex fshort-enums
13952 Allocate to an @code{enum} type only as many bytes as it needs for the
13953 declared range of possible values. Specifically, the @code{enum} type
13954 is equivalent to the smallest integer type that has enough room.
13956 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
13957 code that is not binary compatible with code generated without that switch.
13958 Use it to conform to a non-default application binary interface.
13960 @item -fshort-wchar
13961 @opindex fshort-wchar
13962 Override the underlying type for @code{wchar_t} to be @code{short
13963 unsigned int} instead of the default for the target. This option is
13964 useful for building programs to run under WINE@.
13966 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
13967 code that is not binary compatible with code generated without that switch.
13968 Use it to conform to a non-default application binary interface.
13971 @opindex fno-common
13973 @cindex tentative definitions
13974 In C code, this option controls the placement of global variables
13975 defined without an initializer, known as @dfn{tentative definitions}
13976 in the C standard. Tentative definitions are distinct from declarations
13977 of a variable with the @code{extern} keyword, which do not allocate storage.
13979 Unix C compilers have traditionally allocated storage for
13980 uninitialized global variables in a common block. This allows the
13981 linker to resolve all tentative definitions of the same variable
13982 in different compilation units to the same object, or to a non-tentative
13984 This is the behavior specified by @option{-fcommon}, and is the default for
13985 GCC on most targets.
13986 On the other hand, this behavior is not required by ISO
13987 C, and on some targets may carry a speed or code size penalty on
13988 variable references.
13990 The @option{-fno-common} option specifies that the compiler should instead
13991 place uninitialized global variables in the BSS section of the object file.
13992 This inhibits the merging of tentative definitions by the linker so
13993 you get a multiple-definition error if the same
13994 variable is defined in more than one compilation unit.
13995 Compiling with @option{-fno-common} is useful on targets for which
13996 it provides better performance, or if you wish to verify that the
13997 program will work on other systems that always treat uninitialized
13998 variable definitions this way.
14003 Ignore the @code{#ident} directive.
14005 @item -finhibit-size-directive
14006 @opindex finhibit-size-directive
14007 Don't output a @code{.size} assembler directive, or anything else that
14008 would cause trouble if the function is split in the middle, and the
14009 two halves are placed at locations far apart in memory. This option is
14010 used when compiling @file{crtstuff.c}; you should not need to use it
14013 @item -fverbose-asm
14014 @opindex fverbose-asm
14015 Put extra commentary information in the generated assembly code to
14016 make it more readable. This option is generally only of use to those
14017 who actually need to read the generated assembly code (perhaps while
14018 debugging the compiler itself).
14020 @option{-fno-verbose-asm}, the default, causes the
14021 extra information to be omitted and is useful when comparing two assembler
14024 The added comments include:
14029 information on the compiler version and command-line options,
14032 the source code lines associated with the assembly instructions,
14033 in the form FILENAME:LINENUMBER:CONTENT OF LINE,
14036 hints on which high-level expressions correspond to
14037 the various assembly instruction operands.
14041 For example, given this C source file:
14049 for (i = 0; i < n; i++)
14056 compiling to (x86_64) assembly via @option{-S} and emitting the result
14057 direct to stdout via @option{-o} @option{-}
14060 gcc -S test.c -fverbose-asm -Os -o -
14063 gives output similar to this:
14067 # GNU C11 (GCC) version 7.0.0 20160809 (experimental) (x86_64-pc-linux-gnu)
14074 .type test, @@function
14078 # test.c:4: int total = 0;
14079 xorl %eax, %eax # <retval>
14080 # test.c:6: for (i = 0; i < n; i++)
14081 xorl %edx, %edx # i
14083 # test.c:6: for (i = 0; i < n; i++)
14084 cmpl %edi, %edx # n, i
14086 # test.c:7: total += i * i;
14087 movl %edx, %ecx # i, tmp92
14088 imull %edx, %ecx # i, tmp92
14089 # test.c:6: for (i = 0; i < n; i++)
14091 # test.c:7: total += i * i;
14092 addl %ecx, %eax # tmp92, <retval>
14100 .ident "GCC: (GNU) 7.0.0 20160809 (experimental)"
14101 .section .note.GNU-stack,"",@@progbits
14104 The comments are intended for humans rather than machines and hence the
14105 precise format of the comments is subject to change.
14107 @item -frecord-gcc-switches
14108 @opindex frecord-gcc-switches
14109 This switch causes the command line used to invoke the
14110 compiler to be recorded into the object file that is being created.
14111 This switch is only implemented on some targets and the exact format
14112 of the recording is target and binary file format dependent, but it
14113 usually takes the form of a section containing ASCII text. This
14114 switch is related to the @option{-fverbose-asm} switch, but that
14115 switch only records information in the assembler output file as
14116 comments, so it never reaches the object file.
14117 See also @option{-grecord-gcc-switches} for another
14118 way of storing compiler options into the object file.
14122 @cindex global offset table
14124 Generate position-independent code (PIC) suitable for use in a shared
14125 library, if supported for the target machine. Such code accesses all
14126 constant addresses through a global offset table (GOT)@. The dynamic
14127 loader resolves the GOT entries when the program starts (the dynamic
14128 loader is not part of GCC; it is part of the operating system). If
14129 the GOT size for the linked executable exceeds a machine-specific
14130 maximum size, you get an error message from the linker indicating that
14131 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
14132 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
14133 on the m68k and RS/6000. The x86 has no such limit.)
14135 Position-independent code requires special support, and therefore works
14136 only on certain machines. For the x86, GCC supports PIC for System V
14137 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
14138 position-independent.
14140 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
14145 If supported for the target machine, emit position-independent code,
14146 suitable for dynamic linking and avoiding any limit on the size of the
14147 global offset table. This option makes a difference on AArch64, m68k,
14148 PowerPC and SPARC@.
14150 Position-independent code requires special support, and therefore works
14151 only on certain machines.
14153 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
14160 These options are similar to @option{-fpic} and @option{-fPIC}, but the
14161 generated position-independent code can be only linked into executables.
14162 Usually these options are used to compile code that will be linked using
14163 the @option{-pie} GCC option.
14165 @option{-fpie} and @option{-fPIE} both define the macros
14166 @code{__pie__} and @code{__PIE__}. The macros have the value 1
14167 for @option{-fpie} and 2 for @option{-fPIE}.
14172 Do not use the PLT for external function calls in position-independent code.
14173 Instead, load the callee address at call sites from the GOT and branch to it.
14174 This leads to more efficient code by eliminating PLT stubs and exposing
14175 GOT loads to optimizations. On architectures such as 32-bit x86 where
14176 PLT stubs expect the GOT pointer in a specific register, this gives more
14177 register allocation freedom to the compiler.
14178 Lazy binding requires use of the PLT;
14179 with @option{-fno-plt} all external symbols are resolved at load time.
14181 Alternatively, the function attribute @code{noplt} can be used to avoid calls
14182 through the PLT for specific external functions.
14184 In position-dependent code, a few targets also convert calls to
14185 functions that are marked to not use the PLT to use the GOT instead.
14187 @item -fno-jump-tables
14188 @opindex fno-jump-tables
14189 @opindex fjump-tables
14190 Do not use jump tables for switch statements even where it would be
14191 more efficient than other code generation strategies. This option is
14192 of use in conjunction with @option{-fpic} or @option{-fPIC} for
14193 building code that forms part of a dynamic linker and cannot
14194 reference the address of a jump table. On some targets, jump tables
14195 do not require a GOT and this option is not needed.
14197 @item -ffixed-@var{reg}
14199 Treat the register named @var{reg} as a fixed register; generated code
14200 should never refer to it (except perhaps as a stack pointer, frame
14201 pointer or in some other fixed role).
14203 @var{reg} must be the name of a register. The register names accepted
14204 are machine-specific and are defined in the @code{REGISTER_NAMES}
14205 macro in the machine description macro file.
14207 This flag does not have a negative form, because it specifies a
14210 @item -fcall-used-@var{reg}
14211 @opindex fcall-used
14212 Treat the register named @var{reg} as an allocable register that is
14213 clobbered by function calls. It may be allocated for temporaries or
14214 variables that do not live across a call. Functions compiled this way
14215 do not save and restore the register @var{reg}.
14217 It is an error to use this flag with the frame pointer or stack pointer.
14218 Use of this flag for other registers that have fixed pervasive roles in
14219 the machine's execution model produces disastrous results.
14221 This flag does not have a negative form, because it specifies a
14224 @item -fcall-saved-@var{reg}
14225 @opindex fcall-saved
14226 Treat the register named @var{reg} as an allocable register saved by
14227 functions. It may be allocated even for temporaries or variables that
14228 live across a call. Functions compiled this way save and restore
14229 the register @var{reg} if they use it.
14231 It is an error to use this flag with the frame pointer or stack pointer.
14232 Use of this flag for other registers that have fixed pervasive roles in
14233 the machine's execution model produces disastrous results.
14235 A different sort of disaster results from the use of this flag for
14236 a register in which function values may be returned.
14238 This flag does not have a negative form, because it specifies a
14241 @item -fpack-struct[=@var{n}]
14242 @opindex fpack-struct
14243 Without a value specified, pack all structure members together without
14244 holes. When a value is specified (which must be a small power of two), pack
14245 structure members according to this value, representing the maximum
14246 alignment (that is, objects with default alignment requirements larger than
14247 this are output potentially unaligned at the next fitting location.
14249 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
14250 code that is not binary compatible with code generated without that switch.
14251 Additionally, it makes the code suboptimal.
14252 Use it to conform to a non-default application binary interface.
14254 @item -fleading-underscore
14255 @opindex fleading-underscore
14256 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
14257 change the way C symbols are represented in the object file. One use
14258 is to help link with legacy assembly code.
14260 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
14261 generate code that is not binary compatible with code generated without that
14262 switch. Use it to conform to a non-default application binary interface.
14263 Not all targets provide complete support for this switch.
14265 @item -ftls-model=@var{model}
14266 @opindex ftls-model
14267 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
14268 The @var{model} argument should be one of @samp{global-dynamic},
14269 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
14270 Note that the choice is subject to optimization: the compiler may use
14271 a more efficient model for symbols not visible outside of the translation
14272 unit, or if @option{-fpic} is not given on the command line.
14274 The default without @option{-fpic} is @samp{initial-exec}; with
14275 @option{-fpic} the default is @samp{global-dynamic}.
14277 @item -ftrampolines
14278 @opindex ftrampolines
14279 For targets that normally need trampolines for nested functions, always
14280 generate them instead of using descriptors. Otherwise, for targets that
14281 do not need them, like for example HP-PA or IA-64, do nothing.
14283 A trampoline is a small piece of code that is created at run time on the
14284 stack when the address of a nested function is taken, and is used to call
14285 the nested function indirectly. Therefore, it requires the stack to be
14286 made executable in order for the program to work properly.
14288 @option{-fno-trampolines} is enabled by default on a language by language
14289 basis to let the compiler avoid generating them, if it computes that this
14290 is safe, and replace them with descriptors. Descriptors are made up of data
14291 only, but the generated code must be prepared to deal with them. As of this
14292 writing, @option{-fno-trampolines} is enabled by default only for Ada.
14294 Moreover, code compiled with @option{-ftrampolines} and code compiled with
14295 @option{-fno-trampolines} are not binary compatible if nested functions are
14296 present. This option must therefore be used on a program-wide basis and be
14297 manipulated with extreme care.
14299 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
14300 @opindex fvisibility
14301 Set the default ELF image symbol visibility to the specified option---all
14302 symbols are marked with this unless overridden within the code.
14303 Using this feature can very substantially improve linking and
14304 load times of shared object libraries, produce more optimized
14305 code, provide near-perfect API export and prevent symbol clashes.
14306 It is @strong{strongly} recommended that you use this in any shared objects
14309 Despite the nomenclature, @samp{default} always means public; i.e.,
14310 available to be linked against from outside the shared object.
14311 @samp{protected} and @samp{internal} are pretty useless in real-world
14312 usage so the only other commonly used option is @samp{hidden}.
14313 The default if @option{-fvisibility} isn't specified is
14314 @samp{default}, i.e., make every symbol public.
14316 A good explanation of the benefits offered by ensuring ELF
14317 symbols have the correct visibility is given by ``How To Write
14318 Shared Libraries'' by Ulrich Drepper (which can be found at
14319 @w{@uref{https://www.akkadia.org/drepper/}})---however a superior
14320 solution made possible by this option to marking things hidden when
14321 the default is public is to make the default hidden and mark things
14322 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
14323 and @code{__attribute__ ((visibility("default")))} instead of
14324 @code{__declspec(dllexport)} you get almost identical semantics with
14325 identical syntax. This is a great boon to those working with
14326 cross-platform projects.
14328 For those adding visibility support to existing code, you may find
14329 @code{#pragma GCC visibility} of use. This works by you enclosing
14330 the declarations you wish to set visibility for with (for example)
14331 @code{#pragma GCC visibility push(hidden)} and
14332 @code{#pragma GCC visibility pop}.
14333 Bear in mind that symbol visibility should be viewed @strong{as
14334 part of the API interface contract} and thus all new code should
14335 always specify visibility when it is not the default; i.e., declarations
14336 only for use within the local DSO should @strong{always} be marked explicitly
14337 as hidden as so to avoid PLT indirection overheads---making this
14338 abundantly clear also aids readability and self-documentation of the code.
14339 Note that due to ISO C++ specification requirements, @code{operator new} and
14340 @code{operator delete} must always be of default visibility.
14342 Be aware that headers from outside your project, in particular system
14343 headers and headers from any other library you use, may not be
14344 expecting to be compiled with visibility other than the default. You
14345 may need to explicitly say @code{#pragma GCC visibility push(default)}
14346 before including any such headers.
14348 @code{extern} declarations are not affected by @option{-fvisibility}, so
14349 a lot of code can be recompiled with @option{-fvisibility=hidden} with
14350 no modifications. However, this means that calls to @code{extern}
14351 functions with no explicit visibility use the PLT, so it is more
14352 effective to use @code{__attribute ((visibility))} and/or
14353 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
14354 declarations should be treated as hidden.
14356 Note that @option{-fvisibility} does affect C++ vague linkage
14357 entities. This means that, for instance, an exception class that is
14358 be thrown between DSOs must be explicitly marked with default
14359 visibility so that the @samp{type_info} nodes are unified between
14362 An overview of these techniques, their benefits and how to use them
14363 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
14365 @item -fstrict-volatile-bitfields
14366 @opindex fstrict-volatile-bitfields
14367 This option should be used if accesses to volatile bit-fields (or other
14368 structure fields, although the compiler usually honors those types
14369 anyway) should use a single access of the width of the
14370 field's type, aligned to a natural alignment if possible. For
14371 example, targets with memory-mapped peripheral registers might require
14372 all such accesses to be 16 bits wide; with this flag you can
14373 declare all peripheral bit-fields as @code{unsigned short} (assuming short
14374 is 16 bits on these targets) to force GCC to use 16-bit accesses
14375 instead of, perhaps, a more efficient 32-bit access.
14377 If this option is disabled, the compiler uses the most efficient
14378 instruction. In the previous example, that might be a 32-bit load
14379 instruction, even though that accesses bytes that do not contain
14380 any portion of the bit-field, or memory-mapped registers unrelated to
14381 the one being updated.
14383 In some cases, such as when the @code{packed} attribute is applied to a
14384 structure field, it may not be possible to access the field with a single
14385 read or write that is correctly aligned for the target machine. In this
14386 case GCC falls back to generating multiple accesses rather than code that
14387 will fault or truncate the result at run time.
14389 Note: Due to restrictions of the C/C++11 memory model, write accesses are
14390 not allowed to touch non bit-field members. It is therefore recommended
14391 to define all bits of the field's type as bit-field members.
14393 The default value of this option is determined by the application binary
14394 interface for the target processor.
14396 @item -fsync-libcalls
14397 @opindex fsync-libcalls
14398 This option controls whether any out-of-line instance of the @code{__sync}
14399 family of functions may be used to implement the C++11 @code{__atomic}
14400 family of functions.
14402 The default value of this option is enabled, thus the only useful form
14403 of the option is @option{-fno-sync-libcalls}. This option is used in
14404 the implementation of the @file{libatomic} runtime library.
14408 @node Developer Options
14409 @section GCC Developer Options
14410 @cindex developer options
14411 @cindex debugging GCC
14412 @cindex debug dump options
14413 @cindex dump options
14414 @cindex compilation statistics
14416 This section describes command-line options that are primarily of
14417 interest to GCC developers, including options to support compiler
14418 testing and investigation of compiler bugs and compile-time
14419 performance problems. This includes options that produce debug dumps
14420 at various points in the compilation; that print statistics such as
14421 memory use and execution time; and that print information about GCC's
14422 configuration, such as where it searches for libraries. You should
14423 rarely need to use any of these options for ordinary compilation and
14426 Many developer options that cause GCC to dump output to a file take an
14427 optional @samp{=@var{filename}} suffix. You can specify @samp{stdout}
14428 or @samp{-} to dump to standard output, and @samp{stderr} for standard
14431 If @samp{=@var{filename}} is omitted, a default dump file name is
14432 constructed by concatenating the base dump file name, a pass number,
14433 phase letter, and pass name. The base dump file name is the name of
14434 output file produced by the compiler if explicitly specified and not
14435 an executable; otherwise it is the source file name.
14436 The pass number is determined by the order passes are registered with
14437 the compiler's pass manager.
14438 This is generally the same as the order of execution, but passes
14439 registered by plugins, target-specific passes, or passes that are
14440 otherwise registered late are numbered higher than the pass named
14441 @samp{final}, even if they are executed earlier. The phase letter is
14442 one of @samp{i} (inter-procedural analysis), @samp{l}
14443 (language-specific), @samp{r} (RTL), or @samp{t} (tree).
14444 The files are created in the directory of the output file.
14448 @item -d@var{letters}
14449 @itemx -fdump-rtl-@var{pass}
14450 @itemx -fdump-rtl-@var{pass}=@var{filename}
14452 @opindex fdump-rtl-@var{pass}
14453 Says to make debugging dumps during compilation at times specified by
14454 @var{letters}. This is used for debugging the RTL-based passes of the
14457 Some @option{-d@var{letters}} switches have different meaning when
14458 @option{-E} is used for preprocessing. @xref{Preprocessor Options},
14459 for information about preprocessor-specific dump options.
14461 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
14462 @option{-d} option @var{letters}. Here are the possible
14463 letters for use in @var{pass} and @var{letters}, and their meanings:
14467 @item -fdump-rtl-alignments
14468 @opindex fdump-rtl-alignments
14469 Dump after branch alignments have been computed.
14471 @item -fdump-rtl-asmcons
14472 @opindex fdump-rtl-asmcons
14473 Dump after fixing rtl statements that have unsatisfied in/out constraints.
14475 @item -fdump-rtl-auto_inc_dec
14476 @opindex fdump-rtl-auto_inc_dec
14477 Dump after auto-inc-dec discovery. This pass is only run on
14478 architectures that have auto inc or auto dec instructions.
14480 @item -fdump-rtl-barriers
14481 @opindex fdump-rtl-barriers
14482 Dump after cleaning up the barrier instructions.
14484 @item -fdump-rtl-bbpart
14485 @opindex fdump-rtl-bbpart
14486 Dump after partitioning hot and cold basic blocks.
14488 @item -fdump-rtl-bbro
14489 @opindex fdump-rtl-bbro
14490 Dump after block reordering.
14492 @item -fdump-rtl-btl1
14493 @itemx -fdump-rtl-btl2
14494 @opindex fdump-rtl-btl2
14495 @opindex fdump-rtl-btl2
14496 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
14497 after the two branch
14498 target load optimization passes.
14500 @item -fdump-rtl-bypass
14501 @opindex fdump-rtl-bypass
14502 Dump after jump bypassing and control flow optimizations.
14504 @item -fdump-rtl-combine
14505 @opindex fdump-rtl-combine
14506 Dump after the RTL instruction combination pass.
14508 @item -fdump-rtl-compgotos
14509 @opindex fdump-rtl-compgotos
14510 Dump after duplicating the computed gotos.
14512 @item -fdump-rtl-ce1
14513 @itemx -fdump-rtl-ce2
14514 @itemx -fdump-rtl-ce3
14515 @opindex fdump-rtl-ce1
14516 @opindex fdump-rtl-ce2
14517 @opindex fdump-rtl-ce3
14518 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
14519 @option{-fdump-rtl-ce3} enable dumping after the three
14520 if conversion passes.
14522 @item -fdump-rtl-cprop_hardreg
14523 @opindex fdump-rtl-cprop_hardreg
14524 Dump after hard register copy propagation.
14526 @item -fdump-rtl-csa
14527 @opindex fdump-rtl-csa
14528 Dump after combining stack adjustments.
14530 @item -fdump-rtl-cse1
14531 @itemx -fdump-rtl-cse2
14532 @opindex fdump-rtl-cse1
14533 @opindex fdump-rtl-cse2
14534 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
14535 the two common subexpression elimination passes.
14537 @item -fdump-rtl-dce
14538 @opindex fdump-rtl-dce
14539 Dump after the standalone dead code elimination passes.
14541 @item -fdump-rtl-dbr
14542 @opindex fdump-rtl-dbr
14543 Dump after delayed branch scheduling.
14545 @item -fdump-rtl-dce1
14546 @itemx -fdump-rtl-dce2
14547 @opindex fdump-rtl-dce1
14548 @opindex fdump-rtl-dce2
14549 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
14550 the two dead store elimination passes.
14552 @item -fdump-rtl-eh
14553 @opindex fdump-rtl-eh
14554 Dump after finalization of EH handling code.
14556 @item -fdump-rtl-eh_ranges
14557 @opindex fdump-rtl-eh_ranges
14558 Dump after conversion of EH handling range regions.
14560 @item -fdump-rtl-expand
14561 @opindex fdump-rtl-expand
14562 Dump after RTL generation.
14564 @item -fdump-rtl-fwprop1
14565 @itemx -fdump-rtl-fwprop2
14566 @opindex fdump-rtl-fwprop1
14567 @opindex fdump-rtl-fwprop2
14568 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
14569 dumping after the two forward propagation passes.
14571 @item -fdump-rtl-gcse1
14572 @itemx -fdump-rtl-gcse2
14573 @opindex fdump-rtl-gcse1
14574 @opindex fdump-rtl-gcse2
14575 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
14576 after global common subexpression elimination.
14578 @item -fdump-rtl-init-regs
14579 @opindex fdump-rtl-init-regs
14580 Dump after the initialization of the registers.
14582 @item -fdump-rtl-initvals
14583 @opindex fdump-rtl-initvals
14584 Dump after the computation of the initial value sets.
14586 @item -fdump-rtl-into_cfglayout
14587 @opindex fdump-rtl-into_cfglayout
14588 Dump after converting to cfglayout mode.
14590 @item -fdump-rtl-ira
14591 @opindex fdump-rtl-ira
14592 Dump after iterated register allocation.
14594 @item -fdump-rtl-jump
14595 @opindex fdump-rtl-jump
14596 Dump after the second jump optimization.
14598 @item -fdump-rtl-loop2
14599 @opindex fdump-rtl-loop2
14600 @option{-fdump-rtl-loop2} enables dumping after the rtl
14601 loop optimization passes.
14603 @item -fdump-rtl-mach
14604 @opindex fdump-rtl-mach
14605 Dump after performing the machine dependent reorganization pass, if that
14608 @item -fdump-rtl-mode_sw
14609 @opindex fdump-rtl-mode_sw
14610 Dump after removing redundant mode switches.
14612 @item -fdump-rtl-rnreg
14613 @opindex fdump-rtl-rnreg
14614 Dump after register renumbering.
14616 @item -fdump-rtl-outof_cfglayout
14617 @opindex fdump-rtl-outof_cfglayout
14618 Dump after converting from cfglayout mode.
14620 @item -fdump-rtl-peephole2
14621 @opindex fdump-rtl-peephole2
14622 Dump after the peephole pass.
14624 @item -fdump-rtl-postreload
14625 @opindex fdump-rtl-postreload
14626 Dump after post-reload optimizations.
14628 @item -fdump-rtl-pro_and_epilogue
14629 @opindex fdump-rtl-pro_and_epilogue
14630 Dump after generating the function prologues and epilogues.
14632 @item -fdump-rtl-sched1
14633 @itemx -fdump-rtl-sched2
14634 @opindex fdump-rtl-sched1
14635 @opindex fdump-rtl-sched2
14636 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
14637 after the basic block scheduling passes.
14639 @item -fdump-rtl-ree
14640 @opindex fdump-rtl-ree
14641 Dump after sign/zero extension elimination.
14643 @item -fdump-rtl-seqabstr
14644 @opindex fdump-rtl-seqabstr
14645 Dump after common sequence discovery.
14647 @item -fdump-rtl-shorten
14648 @opindex fdump-rtl-shorten
14649 Dump after shortening branches.
14651 @item -fdump-rtl-sibling
14652 @opindex fdump-rtl-sibling
14653 Dump after sibling call optimizations.
14655 @item -fdump-rtl-split1
14656 @itemx -fdump-rtl-split2
14657 @itemx -fdump-rtl-split3
14658 @itemx -fdump-rtl-split4
14659 @itemx -fdump-rtl-split5
14660 @opindex fdump-rtl-split1
14661 @opindex fdump-rtl-split2
14662 @opindex fdump-rtl-split3
14663 @opindex fdump-rtl-split4
14664 @opindex fdump-rtl-split5
14665 These options enable dumping after five rounds of
14666 instruction splitting.
14668 @item -fdump-rtl-sms
14669 @opindex fdump-rtl-sms
14670 Dump after modulo scheduling. This pass is only run on some
14673 @item -fdump-rtl-stack
14674 @opindex fdump-rtl-stack
14675 Dump after conversion from GCC's ``flat register file'' registers to the
14676 x87's stack-like registers. This pass is only run on x86 variants.
14678 @item -fdump-rtl-subreg1
14679 @itemx -fdump-rtl-subreg2
14680 @opindex fdump-rtl-subreg1
14681 @opindex fdump-rtl-subreg2
14682 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
14683 the two subreg expansion passes.
14685 @item -fdump-rtl-unshare
14686 @opindex fdump-rtl-unshare
14687 Dump after all rtl has been unshared.
14689 @item -fdump-rtl-vartrack
14690 @opindex fdump-rtl-vartrack
14691 Dump after variable tracking.
14693 @item -fdump-rtl-vregs
14694 @opindex fdump-rtl-vregs
14695 Dump after converting virtual registers to hard registers.
14697 @item -fdump-rtl-web
14698 @opindex fdump-rtl-web
14699 Dump after live range splitting.
14701 @item -fdump-rtl-regclass
14702 @itemx -fdump-rtl-subregs_of_mode_init
14703 @itemx -fdump-rtl-subregs_of_mode_finish
14704 @itemx -fdump-rtl-dfinit
14705 @itemx -fdump-rtl-dfinish
14706 @opindex fdump-rtl-regclass
14707 @opindex fdump-rtl-subregs_of_mode_init
14708 @opindex fdump-rtl-subregs_of_mode_finish
14709 @opindex fdump-rtl-dfinit
14710 @opindex fdump-rtl-dfinish
14711 These dumps are defined but always produce empty files.
14714 @itemx -fdump-rtl-all
14716 @opindex fdump-rtl-all
14717 Produce all the dumps listed above.
14721 Annotate the assembler output with miscellaneous debugging information.
14725 Dump all macro definitions, at the end of preprocessing, in addition to
14730 Produce a core dump whenever an error occurs.
14734 Annotate the assembler output with a comment indicating which
14735 pattern and alternative is used. The length and cost of each instruction are
14740 Dump the RTL in the assembler output as a comment before each instruction.
14741 Also turns on @option{-dp} annotation.
14745 Just generate RTL for a function instead of compiling it. Usually used
14746 with @option{-fdump-rtl-expand}.
14750 @opindex fdump-debug
14751 Dump debugging information generated during the debug
14754 @item -fdump-earlydebug
14755 @opindex fdump-earlydebug
14756 Dump debugging information generated during the early debug
14759 @item -fdump-noaddr
14760 @opindex fdump-noaddr
14761 When doing debugging dumps, suppress address output. This makes it more
14762 feasible to use diff on debugging dumps for compiler invocations with
14763 different compiler binaries and/or different
14764 text / bss / data / heap / stack / dso start locations.
14767 @opindex freport-bug
14768 Collect and dump debug information into a temporary file if an
14769 internal compiler error (ICE) occurs.
14771 @item -fdump-unnumbered
14772 @opindex fdump-unnumbered
14773 When doing debugging dumps, suppress instruction numbers and address output.
14774 This makes it more feasible to use diff on debugging dumps for compiler
14775 invocations with different options, in particular with and without
14778 @item -fdump-unnumbered-links
14779 @opindex fdump-unnumbered-links
14780 When doing debugging dumps (see @option{-d} option above), suppress
14781 instruction numbers for the links to the previous and next instructions
14784 @item -fdump-ipa-@var{switch}
14785 @itemx -fdump-ipa-@var{switch}-@var{options}
14787 Control the dumping at various stages of inter-procedural analysis
14788 language tree to a file. The file name is generated by appending a
14789 switch specific suffix to the source file name, and the file is created
14790 in the same directory as the output file. The following dumps are
14795 Enables all inter-procedural analysis dumps.
14798 Dumps information about call-graph optimization, unused function removal,
14799 and inlining decisions.
14802 Dump after function inlining.
14806 Additionally, the options @option{-optimized}, @option{-missed},
14807 @option{-note}, and @option{-all} can be provided, with the same meaning
14808 as for @option{-fopt-info}, defaulting to @option{-optimized}.
14810 For example, @option{-fdump-ipa-inline-optimized-missed} will emit
14811 information on callsites that were inlined, along with callsites
14812 that were not inlined.
14814 By default, the dump will contain messages about successful
14815 optimizations (equivalent to @option{-optimized}) together with
14816 low-level details about the analysis.
14818 @item -fdump-lang-all
14819 @itemx -fdump-lang-@var{switch}
14820 @itemx -fdump-lang-@var{switch}-@var{options}
14821 @itemx -fdump-lang-@var{switch}-@var{options}=@var{filename}
14822 @opindex fdump-lang-all
14823 @opindex fdump-lang
14824 Control the dumping of language-specific information. The @var{options}
14825 and @var{filename} portions behave as described in the
14826 @option{-fdump-tree} option. The following @var{switch} values are
14832 Enable all language-specific dumps.
14835 Dump class hierarchy information. Virtual table information is emitted
14836 unless '@option{slim}' is specified. This option is applicable to C++ only.
14839 Dump the raw internal tree data. This option is applicable to C++ only.
14843 @item -fdump-passes
14844 @opindex fdump-passes
14845 Print on @file{stderr} the list of optimization passes that are turned
14846 on and off by the current command-line options.
14848 @item -fdump-statistics-@var{option}
14849 @opindex fdump-statistics
14850 Enable and control dumping of pass statistics in a separate file. The
14851 file name is generated by appending a suffix ending in
14852 @samp{.statistics} to the source file name, and the file is created in
14853 the same directory as the output file. If the @samp{-@var{option}}
14854 form is used, @samp{-stats} causes counters to be summed over the
14855 whole compilation unit while @samp{-details} dumps every event as
14856 the passes generate them. The default with no option is to sum
14857 counters for each function compiled.
14859 @item -fdump-tree-all
14860 @itemx -fdump-tree-@var{switch}
14861 @itemx -fdump-tree-@var{switch}-@var{options}
14862 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
14863 @opindex fdump-tree-all
14864 @opindex fdump-tree
14865 Control the dumping at various stages of processing the intermediate
14866 language tree to a file. If the @samp{-@var{options}}
14867 form is used, @var{options} is a list of @samp{-} separated options
14868 which control the details of the dump. Not all options are applicable
14869 to all dumps; those that are not meaningful are ignored. The
14870 following options are available
14874 Print the address of each node. Usually this is not meaningful as it
14875 changes according to the environment and source file. Its primary use
14876 is for tying up a dump file with a debug environment.
14878 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
14879 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
14880 use working backward from mangled names in the assembly file.
14882 When dumping front-end intermediate representations, inhibit dumping
14883 of members of a scope or body of a function merely because that scope
14884 has been reached. Only dump such items when they are directly reachable
14885 by some other path.
14887 When dumping pretty-printed trees, this option inhibits dumping the
14888 bodies of control structures.
14890 When dumping RTL, print the RTL in slim (condensed) form instead of
14891 the default LISP-like representation.
14893 Print a raw representation of the tree. By default, trees are
14894 pretty-printed into a C-like representation.
14896 Enable more detailed dumps (not honored by every dump option). Also
14897 include information from the optimization passes.
14899 Enable dumping various statistics about the pass (not honored by every dump
14902 Enable showing basic block boundaries (disabled in raw dumps).
14904 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
14905 dump a representation of the control flow graph suitable for viewing with
14906 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
14907 the file is pretty-printed as a subgraph, so that GraphViz can render them
14908 all in a single plot.
14910 This option currently only works for RTL dumps, and the RTL is always
14911 dumped in slim form.
14913 Enable showing virtual operands for every statement.
14915 Enable showing line numbers for statements.
14917 Enable showing the unique ID (@code{DECL_UID}) for each variable.
14919 Enable showing the tree dump for each statement.
14921 Enable showing the EH region number holding each statement.
14923 Enable showing scalar evolution analysis details.
14925 Enable showing optimization information (only available in certain
14928 Enable showing missed optimization information (only available in certain
14931 Enable other detailed optimization information (only available in
14934 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
14935 and @option{lineno}.
14937 Turn on all optimization options, i.e., @option{optimized},
14938 @option{missed}, and @option{note}.
14941 To determine what tree dumps are available or find the dump for a pass
14942 of interest follow the steps below.
14946 Invoke GCC with @option{-fdump-passes} and in the @file{stderr} output
14947 look for a code that corresponds to the pass you are interested in.
14948 For example, the codes @code{tree-evrp}, @code{tree-vrp1}, and
14949 @code{tree-vrp2} correspond to the three Value Range Propagation passes.
14950 The number at the end distinguishes distinct invocations of the same pass.
14952 To enable the creation of the dump file, append the pass code to
14953 the @option{-fdump-} option prefix and invoke GCC with it. For example,
14954 to enable the dump from the Early Value Range Propagation pass, invoke
14955 GCC with the @option{-fdump-tree-evrp} option. Optionally, you may
14956 specify the name of the dump file. If you don't specify one, GCC
14957 creates as described below.
14959 Find the pass dump in a file whose name is composed of three components
14960 separated by a period: the name of the source file GCC was invoked to
14961 compile, a numeric suffix indicating the pass number followed by the
14962 letter @samp{t} for tree passes (and the letter @samp{r} for RTL passes),
14963 and finally the pass code. For example, the Early VRP pass dump might
14964 be in a file named @file{myfile.c.038t.evrp} in the current working
14965 directory. Note that the numeric codes are not stable and may change
14966 from one version of GCC to another.
14970 @itemx -fopt-info-@var{options}
14971 @itemx -fopt-info-@var{options}=@var{filename}
14973 Controls optimization dumps from various optimization passes. If the
14974 @samp{-@var{options}} form is used, @var{options} is a list of
14975 @samp{-} separated option keywords to select the dump details and
14978 The @var{options} can be divided into three groups:
14981 options describing what kinds of messages should be emitted,
14983 options describing the verbosity of the dump, and
14985 options describing which optimizations should be included.
14987 The options from each group can be freely mixed as they are
14988 non-overlapping. However, in case of any conflicts,
14989 the later options override the earlier options on the command
14992 The following options control which kinds of messages should be emitted:
14996 Print information when an optimization is successfully applied. It is
14997 up to a pass to decide which information is relevant. For example, the
14998 vectorizer passes print the source location of loops which are
14999 successfully vectorized.
15001 Print information about missed optimizations. Individual passes
15002 control which information to include in the output.
15004 Print verbose information about optimizations, such as certain
15005 transformations, more detailed messages about decisions etc.
15007 Print detailed optimization information. This includes
15008 @samp{optimized}, @samp{missed}, and @samp{note}.
15011 The following option controls the dump verbosity:
15015 By default, only ``high-level'' messages are emitted. This option enables
15016 additional, more detailed, messages, which are likely to only be of interest
15020 One or more of the following option keywords can be used to describe a
15021 group of optimizations:
15025 Enable dumps from all interprocedural optimizations.
15027 Enable dumps from all loop optimizations.
15029 Enable dumps from all inlining optimizations.
15031 Enable dumps from all OMP (Offloading and Multi Processing) optimizations.
15033 Enable dumps from all vectorization optimizations.
15035 Enable dumps from all optimizations. This is a superset of
15036 the optimization groups listed above.
15039 If @var{options} is
15040 omitted, it defaults to @samp{optimized-optall}, which means to dump messages
15041 about successful optimizations from all the passes, omitting messages
15042 that are treated as ``internals''.
15044 If the @var{filename} is provided, then the dumps from all the
15045 applicable optimizations are concatenated into the @var{filename}.
15046 Otherwise the dump is output onto @file{stderr}. Though multiple
15047 @option{-fopt-info} options are accepted, only one of them can include
15048 a @var{filename}. If other filenames are provided then all but the
15049 first such option are ignored.
15051 Note that the output @var{filename} is overwritten
15052 in case of multiple translation units. If a combined output from
15053 multiple translation units is desired, @file{stderr} should be used
15056 In the following example, the optimization info is output to
15065 gcc -O3 -fopt-info-missed=missed.all
15069 outputs missed optimization report from all the passes into
15070 @file{missed.all}, and this one:
15073 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
15077 prints information about missed optimization opportunities from
15078 vectorization passes on @file{stderr}.
15079 Note that @option{-fopt-info-vec-missed} is equivalent to
15080 @option{-fopt-info-missed-vec}. The order of the optimization group
15081 names and message types listed after @option{-fopt-info} does not matter.
15083 As another example,
15085 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
15089 outputs information about missed optimizations as well as
15090 optimized locations from all the inlining passes into
15096 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
15100 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
15101 in conflict since only one output file is allowed. In this case, only
15102 the first option takes effect and the subsequent options are
15103 ignored. Thus only @file{vec.miss} is produced which contains
15104 dumps from the vectorizer about missed opportunities.
15106 @item -fsave-optimization-record
15107 @opindex fsave-optimization-record
15108 Write a SRCFILE.opt-record.json.gz file detailing what optimizations
15109 were performed, for those optimizations that support @option{-fopt-info}.
15111 This option is experimental and the format of the data within the
15112 compressed JSON file is subject to change.
15114 It is roughly equivalent to a machine-readable version of
15115 @option{-fopt-info-all}, as a collection of messages with source file,
15116 line number and column number, with the following additional data for
15122 the execution count of the code being optimized, along with metadata about
15123 whether this was from actual profile data, or just an estimate, allowing
15124 consumers to prioritize messages by code hotness,
15127 the function name of the code being optimized, where applicable,
15130 the ``inlining chain'' for the code being optimized, so that when
15131 a function is inlined into several different places (which might
15132 themselves be inlined), the reader can distinguish between the copies,
15135 objects identifying those parts of the message that refer to expressions,
15136 statements or symbol-table nodes, which of these categories they are, and,
15137 when available, their source code location,
15140 the GCC pass that emitted the message, and
15143 the location in GCC's own code from which the message was emitted
15147 Additionally, some messages are logically nested within other
15148 messages, reflecting implementation details of the optimization
15151 @item -fsched-verbose=@var{n}
15152 @opindex fsched-verbose
15153 On targets that use instruction scheduling, this option controls the
15154 amount of debugging output the scheduler prints to the dump files.
15156 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
15157 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
15158 For @var{n} greater than one, it also output basic block probabilities,
15159 detailed ready list information and unit/insn info. For @var{n} greater
15160 than two, it includes RTL at abort point, control-flow and regions info.
15161 And for @var{n} over four, @option{-fsched-verbose} also includes
15166 @item -fenable-@var{kind}-@var{pass}
15167 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
15171 This is a set of options that are used to explicitly disable/enable
15172 optimization passes. These options are intended for use for debugging GCC.
15173 Compiler users should use regular options for enabling/disabling
15178 @item -fdisable-ipa-@var{pass}
15179 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
15180 statically invoked in the compiler multiple times, the pass name should be
15181 appended with a sequential number starting from 1.
15183 @item -fdisable-rtl-@var{pass}
15184 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
15185 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
15186 statically invoked in the compiler multiple times, the pass name should be
15187 appended with a sequential number starting from 1. @var{range-list} is a
15188 comma-separated list of function ranges or assembler names. Each range is a number
15189 pair separated by a colon. The range is inclusive in both ends. If the range
15190 is trivial, the number pair can be simplified as a single number. If the
15191 function's call graph node's @var{uid} falls within one of the specified ranges,
15192 the @var{pass} is disabled for that function. The @var{uid} is shown in the
15193 function header of a dump file, and the pass names can be dumped by using
15194 option @option{-fdump-passes}.
15196 @item -fdisable-tree-@var{pass}
15197 @itemx -fdisable-tree-@var{pass}=@var{range-list}
15198 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
15201 @item -fenable-ipa-@var{pass}
15202 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
15203 statically invoked in the compiler multiple times, the pass name should be
15204 appended with a sequential number starting from 1.
15206 @item -fenable-rtl-@var{pass}
15207 @itemx -fenable-rtl-@var{pass}=@var{range-list}
15208 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
15209 description and examples.
15211 @item -fenable-tree-@var{pass}
15212 @itemx -fenable-tree-@var{pass}=@var{range-list}
15213 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
15214 of option arguments.
15218 Here are some examples showing uses of these options.
15222 # disable ccp1 for all functions
15223 -fdisable-tree-ccp1
15224 # disable complete unroll for function whose cgraph node uid is 1
15225 -fenable-tree-cunroll=1
15226 # disable gcse2 for functions at the following ranges [1,1],
15227 # [300,400], and [400,1000]
15228 # disable gcse2 for functions foo and foo2
15229 -fdisable-rtl-gcse2=foo,foo2
15230 # disable early inlining
15231 -fdisable-tree-einline
15232 # disable ipa inlining
15233 -fdisable-ipa-inline
15234 # enable tree full unroll
15235 -fenable-tree-unroll
15240 @itemx -fchecking=@var{n}
15242 @opindex fno-checking
15243 Enable internal consistency checking. The default depends on
15244 the compiler configuration. @option{-fchecking=2} enables further
15245 internal consistency checking that might affect code generation.
15247 @item -frandom-seed=@var{string}
15248 @opindex frandom-seed
15249 This option provides a seed that GCC uses in place of
15250 random numbers in generating certain symbol names
15251 that have to be different in every compiled file. It is also used to
15252 place unique stamps in coverage data files and the object files that
15253 produce them. You can use the @option{-frandom-seed} option to produce
15254 reproducibly identical object files.
15256 The @var{string} can either be a number (decimal, octal or hex) or an
15257 arbitrary string (in which case it's converted to a number by
15260 The @var{string} should be different for every file you compile.
15263 @itemx -save-temps=cwd
15264 @opindex save-temps
15265 Store the usual ``temporary'' intermediate files permanently; place them
15266 in the current directory and name them based on the source file. Thus,
15267 compiling @file{foo.c} with @option{-c -save-temps} produces files
15268 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
15269 preprocessed @file{foo.i} output file even though the compiler now
15270 normally uses an integrated preprocessor.
15272 When used in combination with the @option{-x} command-line option,
15273 @option{-save-temps} is sensible enough to avoid over writing an
15274 input source file with the same extension as an intermediate file.
15275 The corresponding intermediate file may be obtained by renaming the
15276 source file before using @option{-save-temps}.
15278 If you invoke GCC in parallel, compiling several different source
15279 files that share a common base name in different subdirectories or the
15280 same source file compiled for multiple output destinations, it is
15281 likely that the different parallel compilers will interfere with each
15282 other, and overwrite the temporary files. For instance:
15285 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
15286 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
15289 may result in @file{foo.i} and @file{foo.o} being written to
15290 simultaneously by both compilers.
15292 @item -save-temps=obj
15293 @opindex save-temps=obj
15294 Store the usual ``temporary'' intermediate files permanently. If the
15295 @option{-o} option is used, the temporary files are based on the
15296 object file. If the @option{-o} option is not used, the
15297 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
15302 gcc -save-temps=obj -c foo.c
15303 gcc -save-temps=obj -c bar.c -o dir/xbar.o
15304 gcc -save-temps=obj foobar.c -o dir2/yfoobar
15308 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
15309 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
15310 @file{dir2/yfoobar.o}.
15312 @item -time@r{[}=@var{file}@r{]}
15314 Report the CPU time taken by each subprocess in the compilation
15315 sequence. For C source files, this is the compiler proper and assembler
15316 (plus the linker if linking is done).
15318 Without the specification of an output file, the output looks like this:
15325 The first number on each line is the ``user time'', that is time spent
15326 executing the program itself. The second number is ``system time'',
15327 time spent executing operating system routines on behalf of the program.
15328 Both numbers are in seconds.
15330 With the specification of an output file, the output is appended to the
15331 named file, and it looks like this:
15334 0.12 0.01 cc1 @var{options}
15335 0.00 0.01 as @var{options}
15338 The ``user time'' and the ``system time'' are moved before the program
15339 name, and the options passed to the program are displayed, so that one
15340 can later tell what file was being compiled, and with which options.
15342 @item -fdump-final-insns@r{[}=@var{file}@r{]}
15343 @opindex fdump-final-insns
15344 Dump the final internal representation (RTL) to @var{file}. If the
15345 optional argument is omitted (or if @var{file} is @code{.}), the name
15346 of the dump file is determined by appending @code{.gkd} to the
15347 compilation output file name.
15349 @item -fcompare-debug@r{[}=@var{opts}@r{]}
15350 @opindex fcompare-debug
15351 @opindex fno-compare-debug
15352 If no error occurs during compilation, run the compiler a second time,
15353 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
15354 passed to the second compilation. Dump the final internal
15355 representation in both compilations, and print an error if they differ.
15357 If the equal sign is omitted, the default @option{-gtoggle} is used.
15359 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
15360 and nonzero, implicitly enables @option{-fcompare-debug}. If
15361 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
15362 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
15365 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
15366 is equivalent to @option{-fno-compare-debug}, which disables the dumping
15367 of the final representation and the second compilation, preventing even
15368 @env{GCC_COMPARE_DEBUG} from taking effect.
15370 To verify full coverage during @option{-fcompare-debug} testing, set
15371 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
15372 which GCC rejects as an invalid option in any actual compilation
15373 (rather than preprocessing, assembly or linking). To get just a
15374 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
15375 not overridden} will do.
15377 @item -fcompare-debug-second
15378 @opindex fcompare-debug-second
15379 This option is implicitly passed to the compiler for the second
15380 compilation requested by @option{-fcompare-debug}, along with options to
15381 silence warnings, and omitting other options that would cause the compiler
15382 to produce output to files or to standard output as a side effect. Dump
15383 files and preserved temporary files are renamed so as to contain the
15384 @code{.gk} additional extension during the second compilation, to avoid
15385 overwriting those generated by the first.
15387 When this option is passed to the compiler driver, it causes the
15388 @emph{first} compilation to be skipped, which makes it useful for little
15389 other than debugging the compiler proper.
15393 Turn off generation of debug info, if leaving out this option
15394 generates it, or turn it on at level 2 otherwise. The position of this
15395 argument in the command line does not matter; it takes effect after all
15396 other options are processed, and it does so only once, no matter how
15397 many times it is given. This is mainly intended to be used with
15398 @option{-fcompare-debug}.
15400 @item -fvar-tracking-assignments-toggle
15401 @opindex fvar-tracking-assignments-toggle
15402 @opindex fno-var-tracking-assignments-toggle
15403 Toggle @option{-fvar-tracking-assignments}, in the same way that
15404 @option{-gtoggle} toggles @option{-g}.
15408 Makes the compiler print out each function name as it is compiled, and
15409 print some statistics about each pass when it finishes.
15411 @item -ftime-report
15412 @opindex ftime-report
15413 Makes the compiler print some statistics about the time consumed by each
15414 pass when it finishes.
15416 @item -ftime-report-details
15417 @opindex ftime-report-details
15418 Record the time consumed by infrastructure parts separately for each pass.
15420 @item -fira-verbose=@var{n}
15421 @opindex fira-verbose
15422 Control the verbosity of the dump file for the integrated register allocator.
15423 The default value is 5. If the value @var{n} is greater or equal to 10,
15424 the dump output is sent to stderr using the same format as @var{n} minus 10.
15427 @opindex flto-report
15428 Prints a report with internal details on the workings of the link-time
15429 optimizer. The contents of this report vary from version to version.
15430 It is meant to be useful to GCC developers when processing object
15431 files in LTO mode (via @option{-flto}).
15433 Disabled by default.
15435 @item -flto-report-wpa
15436 @opindex flto-report-wpa
15437 Like @option{-flto-report}, but only print for the WPA phase of link-time
15441 @opindex fmem-report
15442 Makes the compiler print some statistics about permanent memory
15443 allocation when it finishes.
15445 @item -fmem-report-wpa
15446 @opindex fmem-report-wpa
15447 Makes the compiler print some statistics about permanent memory
15448 allocation for the WPA phase only.
15450 @item -fpre-ipa-mem-report
15451 @opindex fpre-ipa-mem-report
15452 @item -fpost-ipa-mem-report
15453 @opindex fpost-ipa-mem-report
15454 Makes the compiler print some statistics about permanent memory
15455 allocation before or after interprocedural optimization.
15457 @item -fprofile-report
15458 @opindex fprofile-report
15459 Makes the compiler print some statistics about consistency of the
15460 (estimated) profile and effect of individual passes.
15462 @item -fstack-usage
15463 @opindex fstack-usage
15464 Makes the compiler output stack usage information for the program, on a
15465 per-function basis. The filename for the dump is made by appending
15466 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
15467 the output file, if explicitly specified and it is not an executable,
15468 otherwise it is the basename of the source file. An entry is made up
15473 The name of the function.
15477 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
15480 The qualifier @code{static} means that the function manipulates the stack
15481 statically: a fixed number of bytes are allocated for the frame on function
15482 entry and released on function exit; no stack adjustments are otherwise made
15483 in the function. The second field is this fixed number of bytes.
15485 The qualifier @code{dynamic} means that the function manipulates the stack
15486 dynamically: in addition to the static allocation described above, stack
15487 adjustments are made in the body of the function, for example to push/pop
15488 arguments around function calls. If the qualifier @code{bounded} is also
15489 present, the amount of these adjustments is bounded at compile time and
15490 the second field is an upper bound of the total amount of stack used by
15491 the function. If it is not present, the amount of these adjustments is
15492 not bounded at compile time and the second field only represents the
15497 Emit statistics about front-end processing at the end of the compilation.
15498 This option is supported only by the C++ front end, and
15499 the information is generally only useful to the G++ development team.
15501 @item -fdbg-cnt-list
15502 @opindex fdbg-cnt-list
15503 Print the name and the counter upper bound for all debug counters.
15506 @item -fdbg-cnt=@var{counter-value-list}
15508 Set the internal debug counter lower and upper bound. @var{counter-value-list}
15509 is a comma-separated list of @var{name}:@var{lower_bound}:@var{upper_bound}
15510 tuples which sets the lower and the upper bound of each debug
15511 counter @var{name}. The @var{lower_bound} is optional and is zero
15512 initialized if not set.
15513 All debug counters have the initial upper bound of @code{UINT_MAX};
15514 thus @code{dbg_cnt} returns true always unless the upper bound
15515 is set by this option.
15516 For example, with @option{-fdbg-cnt=dce:2:4,tail_call:10},
15517 @code{dbg_cnt(dce)} returns true only for third and fourth invocation.
15518 For @code{dbg_cnt(tail_call)} true is returned for first 10 invocations.
15520 @item -print-file-name=@var{library}
15521 @opindex print-file-name
15522 Print the full absolute name of the library file @var{library} that
15523 would be used when linking---and don't do anything else. With this
15524 option, GCC does not compile or link anything; it just prints the
15527 @item -print-multi-directory
15528 @opindex print-multi-directory
15529 Print the directory name corresponding to the multilib selected by any
15530 other switches present in the command line. This directory is supposed
15531 to exist in @env{GCC_EXEC_PREFIX}.
15533 @item -print-multi-lib
15534 @opindex print-multi-lib
15535 Print the mapping from multilib directory names to compiler switches
15536 that enable them. The directory name is separated from the switches by
15537 @samp{;}, and each switch starts with an @samp{@@} instead of the
15538 @samp{-}, without spaces between multiple switches. This is supposed to
15539 ease shell processing.
15541 @item -print-multi-os-directory
15542 @opindex print-multi-os-directory
15543 Print the path to OS libraries for the selected
15544 multilib, relative to some @file{lib} subdirectory. If OS libraries are
15545 present in the @file{lib} subdirectory and no multilibs are used, this is
15546 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
15547 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
15548 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
15549 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
15551 @item -print-multiarch
15552 @opindex print-multiarch
15553 Print the path to OS libraries for the selected multiarch,
15554 relative to some @file{lib} subdirectory.
15556 @item -print-prog-name=@var{program}
15557 @opindex print-prog-name
15558 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
15560 @item -print-libgcc-file-name
15561 @opindex print-libgcc-file-name
15562 Same as @option{-print-file-name=libgcc.a}.
15564 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
15565 but you do want to link with @file{libgcc.a}. You can do:
15568 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
15571 @item -print-search-dirs
15572 @opindex print-search-dirs
15573 Print the name of the configured installation directory and a list of
15574 program and library directories @command{gcc} searches---and don't do anything else.
15576 This is useful when @command{gcc} prints the error message
15577 @samp{installation problem, cannot exec cpp0: No such file or directory}.
15578 To resolve this you either need to put @file{cpp0} and the other compiler
15579 components where @command{gcc} expects to find them, or you can set the environment
15580 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
15581 Don't forget the trailing @samp{/}.
15582 @xref{Environment Variables}.
15584 @item -print-sysroot
15585 @opindex print-sysroot
15586 Print the target sysroot directory that is used during
15587 compilation. This is the target sysroot specified either at configure
15588 time or using the @option{--sysroot} option, possibly with an extra
15589 suffix that depends on compilation options. If no target sysroot is
15590 specified, the option prints nothing.
15592 @item -print-sysroot-headers-suffix
15593 @opindex print-sysroot-headers-suffix
15594 Print the suffix added to the target sysroot when searching for
15595 headers, or give an error if the compiler is not configured with such
15596 a suffix---and don't do anything else.
15599 @opindex dumpmachine
15600 Print the compiler's target machine (for example,
15601 @samp{i686-pc-linux-gnu})---and don't do anything else.
15604 @opindex dumpversion
15605 Print the compiler version (for example, @code{3.0}, @code{6.3.0} or @code{7})---and don't do
15606 anything else. This is the compiler version used in filesystem paths and
15607 specs. Depending on how the compiler has been configured it can be just
15608 a single number (major version), two numbers separated by a dot (major and
15609 minor version) or three numbers separated by dots (major, minor and patchlevel
15612 @item -dumpfullversion
15613 @opindex dumpfullversion
15614 Print the full compiler version---and don't do anything else. The output is
15615 always three numbers separated by dots, major, minor and patchlevel version.
15619 Print the compiler's built-in specs---and don't do anything else. (This
15620 is used when GCC itself is being built.) @xref{Spec Files}.
15623 @node Submodel Options
15624 @section Machine-Dependent Options
15625 @cindex submodel options
15626 @cindex specifying hardware config
15627 @cindex hardware models and configurations, specifying
15628 @cindex target-dependent options
15629 @cindex machine-dependent options
15631 Each target machine supported by GCC can have its own options---for
15632 example, to allow you to compile for a particular processor variant or
15633 ABI, or to control optimizations specific to that machine. By
15634 convention, the names of machine-specific options start with
15637 Some configurations of the compiler also support additional target-specific
15638 options, usually for compatibility with other compilers on the same
15641 @c This list is ordered alphanumerically by subsection name.
15642 @c It should be the same order and spelling as these options are listed
15643 @c in Machine Dependent Options
15646 * AArch64 Options::
15647 * Adapteva Epiphany Options::
15648 * AMD GCN Options::
15652 * Blackfin Options::
15658 * DEC Alpha Options::
15662 * GNU/Linux Options::
15672 * MicroBlaze Options::
15675 * MN10300 Options::
15679 * Nios II Options::
15680 * Nvidia PTX Options::
15681 * OpenRISC Options::
15683 * picoChip Options::
15684 * PowerPC Options::
15688 * RS/6000 and PowerPC Options::
15690 * S/390 and zSeries Options::
15693 * Solaris 2 Options::
15695 * System V Options::
15696 * TILE-Gx Options::
15697 * TILEPro Options::
15702 * VxWorks Options::
15704 * x86 Windows Options::
15705 * Xstormy16 Options::
15707 * zSeries Options::
15710 @node AArch64 Options
15711 @subsection AArch64 Options
15712 @cindex AArch64 Options
15714 These options are defined for AArch64 implementations:
15718 @item -mabi=@var{name}
15720 Generate code for the specified data model. Permissible values
15721 are @samp{ilp32} for SysV-like data model where int, long int and pointers
15722 are 32 bits, and @samp{lp64} for SysV-like data model where int is 32 bits,
15723 but long int and pointers are 64 bits.
15725 The default depends on the specific target configuration. Note that
15726 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
15727 entire program with the same ABI, and link with a compatible set of libraries.
15730 @opindex mbig-endian
15731 Generate big-endian code. This is the default when GCC is configured for an
15732 @samp{aarch64_be-*-*} target.
15734 @item -mgeneral-regs-only
15735 @opindex mgeneral-regs-only
15736 Generate code which uses only the general-purpose registers. This will prevent
15737 the compiler from using floating-point and Advanced SIMD registers but will not
15738 impose any restrictions on the assembler.
15740 @item -mlittle-endian
15741 @opindex mlittle-endian
15742 Generate little-endian code. This is the default when GCC is configured for an
15743 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
15745 @item -mcmodel=tiny
15746 @opindex mcmodel=tiny
15747 Generate code for the tiny code model. The program and its statically defined
15748 symbols must be within 1MB of each other. Programs can be statically or
15749 dynamically linked.
15751 @item -mcmodel=small
15752 @opindex mcmodel=small
15753 Generate code for the small code model. The program and its statically defined
15754 symbols must be within 4GB of each other. Programs can be statically or
15755 dynamically linked. This is the default code model.
15757 @item -mcmodel=large
15758 @opindex mcmodel=large
15759 Generate code for the large code model. This makes no assumptions about
15760 addresses and sizes of sections. Programs can be statically linked only.
15762 @item -mstrict-align
15763 @itemx -mno-strict-align
15764 @opindex mstrict-align
15765 @opindex mno-strict-align
15766 Avoid or allow generating memory accesses that may not be aligned on a natural
15767 object boundary as described in the architecture specification.
15769 @item -momit-leaf-frame-pointer
15770 @itemx -mno-omit-leaf-frame-pointer
15771 @opindex momit-leaf-frame-pointer
15772 @opindex mno-omit-leaf-frame-pointer
15773 Omit or keep the frame pointer in leaf functions. The former behavior is the
15776 @item -mstack-protector-guard=@var{guard}
15777 @itemx -mstack-protector-guard-reg=@var{reg}
15778 @itemx -mstack-protector-guard-offset=@var{offset}
15779 @opindex mstack-protector-guard
15780 @opindex mstack-protector-guard-reg
15781 @opindex mstack-protector-guard-offset
15782 Generate stack protection code using canary at @var{guard}. Supported
15783 locations are @samp{global} for a global canary or @samp{sysreg} for a
15784 canary in an appropriate system register.
15786 With the latter choice the options
15787 @option{-mstack-protector-guard-reg=@var{reg}} and
15788 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
15789 which system register to use as base register for reading the canary,
15790 and from what offset from that base register. There is no default
15791 register or offset as this is entirely for use within the Linux
15794 @item -mstack-protector-guard=@var{guard}
15795 @itemx -mstack-protector-guard-reg=@var{reg}
15796 @itemx -mstack-protector-guard-offset=@var{offset}
15797 @opindex mstack-protector-guard
15798 @opindex mstack-protector-guard-reg
15799 @opindex mstack-protector-guard-offset
15800 Generate stack protection code using canary at @var{guard}. Supported
15801 locations are @samp{global} for a global canary or @samp{sysreg} for a
15802 canary in an appropriate system register.
15804 With the latter choice the options
15805 @option{-mstack-protector-guard-reg=@var{reg}} and
15806 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
15807 which system register to use as base register for reading the canary,
15808 and from what offset from that base register. There is no default
15809 register or offset as this is entirely for use within the Linux
15812 @item -mtls-dialect=desc
15813 @opindex mtls-dialect=desc
15814 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
15815 of TLS variables. This is the default.
15817 @item -mtls-dialect=traditional
15818 @opindex mtls-dialect=traditional
15819 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
15822 @item -mtls-size=@var{size}
15824 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
15825 This option requires binutils 2.26 or newer.
15827 @item -mfix-cortex-a53-835769
15828 @itemx -mno-fix-cortex-a53-835769
15829 @opindex mfix-cortex-a53-835769
15830 @opindex mno-fix-cortex-a53-835769
15831 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
15832 This involves inserting a NOP instruction between memory instructions and
15833 64-bit integer multiply-accumulate instructions.
15835 @item -mfix-cortex-a53-843419
15836 @itemx -mno-fix-cortex-a53-843419
15837 @opindex mfix-cortex-a53-843419
15838 @opindex mno-fix-cortex-a53-843419
15839 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
15840 This erratum workaround is made at link time and this will only pass the
15841 corresponding flag to the linker.
15843 @item -mlow-precision-recip-sqrt
15844 @itemx -mno-low-precision-recip-sqrt
15845 @opindex mlow-precision-recip-sqrt
15846 @opindex mno-low-precision-recip-sqrt
15847 Enable or disable the reciprocal square root approximation.
15848 This option only has an effect if @option{-ffast-math} or
15849 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
15850 precision of reciprocal square root results to about 16 bits for
15851 single precision and to 32 bits for double precision.
15853 @item -mlow-precision-sqrt
15854 @itemx -mno-low-precision-sqrt
15855 @opindex mlow-precision-sqrt
15856 @opindex mno-low-precision-sqrt
15857 Enable or disable the square root approximation.
15858 This option only has an effect if @option{-ffast-math} or
15859 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
15860 precision of square root results to about 16 bits for
15861 single precision and to 32 bits for double precision.
15862 If enabled, it implies @option{-mlow-precision-recip-sqrt}.
15864 @item -mlow-precision-div
15865 @itemx -mno-low-precision-div
15866 @opindex mlow-precision-div
15867 @opindex mno-low-precision-div
15868 Enable or disable the division approximation.
15869 This option only has an effect if @option{-ffast-math} or
15870 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
15871 precision of division results to about 16 bits for
15872 single precision and to 32 bits for double precision.
15874 @item -mtrack-speculation
15875 @itemx -mno-track-speculation
15876 Enable or disable generation of additional code to track speculative
15877 execution through conditional branches. The tracking state can then
15878 be used by the compiler when expanding calls to
15879 @code{__builtin_speculation_safe_copy} to permit a more efficient code
15880 sequence to be generated.
15882 @item -march=@var{name}
15884 Specify the name of the target architecture and, optionally, one or
15885 more feature modifiers. This option has the form
15886 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
15888 The permissible values for @var{arch} are @samp{armv8-a},
15889 @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a}, @samp{armv8.4-a},
15890 @samp{armv8.5-a} or @var{native}.
15892 The value @samp{armv8.5-a} implies @samp{armv8.4-a} and enables compiler
15893 support for the ARMv8.5-A architecture extensions.
15895 The value @samp{armv8.4-a} implies @samp{armv8.3-a} and enables compiler
15896 support for the ARMv8.4-A architecture extensions.
15898 The value @samp{armv8.3-a} implies @samp{armv8.2-a} and enables compiler
15899 support for the ARMv8.3-A architecture extensions.
15901 The value @samp{armv8.2-a} implies @samp{armv8.1-a} and enables compiler
15902 support for the ARMv8.2-A architecture extensions.
15904 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
15905 support for the ARMv8.1-A architecture extension. In particular, it
15906 enables the @samp{+crc}, @samp{+lse}, and @samp{+rdma} features.
15908 The value @samp{native} is available on native AArch64 GNU/Linux and
15909 causes the compiler to pick the architecture of the host system. This
15910 option has no effect if the compiler is unable to recognize the
15911 architecture of the host system,
15913 The permissible values for @var{feature} are listed in the sub-section
15914 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
15915 Feature Modifiers}. Where conflicting feature modifiers are
15916 specified, the right-most feature is used.
15918 GCC uses @var{name} to determine what kind of instructions it can emit
15919 when generating assembly code. If @option{-march} is specified
15920 without either of @option{-mtune} or @option{-mcpu} also being
15921 specified, the code is tuned to perform well across a range of target
15922 processors implementing the target architecture.
15924 @item -mtune=@var{name}
15926 Specify the name of the target processor for which GCC should tune the
15927 performance of the code. Permissible values for this option are:
15928 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
15929 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
15930 @samp{cortex-a76}, @samp{cortex-a76ae}, @samp{cortex-a77},
15931 @samp{cortex-a65}, @samp{cortex-a65ae}, @samp{cortex-a34},
15932 @samp{ares}, @samp{exynos-m1}, @samp{emag}, @samp{falkor},
15933 @samp{neoverse-e1},@samp{neoverse-n1},@samp{qdf24xx}, @samp{saphira},
15934 @samp{phecda}, @samp{xgene1}, @samp{vulcan}, @samp{octeontx},
15935 @samp{octeontx81}, @samp{octeontx83}, @samp{thunderx}, @samp{thunderxt88},
15936 @samp{thunderxt88p1}, @samp{thunderxt81}, @samp{tsv110},
15937 @samp{thunderxt83}, @samp{thunderx2t99},
15938 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15939 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53},
15940 @samp{cortex-a75.cortex-a55}, @samp{cortex-a76.cortex-a55}
15943 The values @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15944 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53},
15945 @samp{cortex-a75.cortex-a55}, @samp{cortex-a76.cortex-a55} specify that GCC
15946 should tune for a big.LITTLE system.
15948 Additionally on native AArch64 GNU/Linux systems the value
15949 @samp{native} tunes performance to the host system. This option has no effect
15950 if the compiler is unable to recognize the processor of the host system.
15952 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
15953 are specified, the code is tuned to perform well across a range
15954 of target processors.
15956 This option cannot be suffixed by feature modifiers.
15958 @item -mcpu=@var{name}
15960 Specify the name of the target processor, optionally suffixed by one
15961 or more feature modifiers. This option has the form
15962 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
15963 the permissible values for @var{cpu} are the same as those available
15964 for @option{-mtune}. The permissible values for @var{feature} are
15965 documented in the sub-section on
15966 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
15967 Feature Modifiers}. Where conflicting feature modifiers are
15968 specified, the right-most feature is used.
15970 GCC uses @var{name} to determine what kind of instructions it can emit when
15971 generating assembly code (as if by @option{-march}) and to determine
15972 the target processor for which to tune for performance (as if
15973 by @option{-mtune}). Where this option is used in conjunction
15974 with @option{-march} or @option{-mtune}, those options take precedence
15975 over the appropriate part of this option.
15977 @item -moverride=@var{string}
15979 Override tuning decisions made by the back-end in response to a
15980 @option{-mtune=} switch. The syntax, semantics, and accepted values
15981 for @var{string} in this option are not guaranteed to be consistent
15984 This option is only intended to be useful when developing GCC.
15986 @item -mverbose-cost-dump
15987 @opindex mverbose-cost-dump
15988 Enable verbose cost model dumping in the debug dump files. This option is
15989 provided for use in debugging the compiler.
15991 @item -mpc-relative-literal-loads
15992 @itemx -mno-pc-relative-literal-loads
15993 @opindex mpc-relative-literal-loads
15994 @opindex mno-pc-relative-literal-loads
15995 Enable or disable PC-relative literal loads. With this option literal pools are
15996 accessed using a single instruction and emitted after each function. This
15997 limits the maximum size of functions to 1MB. This is enabled by default for
15998 @option{-mcmodel=tiny}.
16000 @item -msign-return-address=@var{scope}
16001 @opindex msign-return-address
16002 Select the function scope on which return address signing will be applied.
16003 Permissible values are @samp{none}, which disables return address signing,
16004 @samp{non-leaf}, which enables pointer signing for functions which are not leaf
16005 functions, and @samp{all}, which enables pointer signing for all functions. The
16006 default value is @samp{none}. This option has been deprecated by
16007 -mbranch-protection.
16009 @item -mbranch-protection=@var{none}|@var{standard}|@var{pac-ret}[+@var{leaf}+@var{b-key}]|@var{bti}
16010 @opindex mbranch-protection
16011 Select the branch protection features to use.
16012 @samp{none} is the default and turns off all types of branch protection.
16013 @samp{standard} turns on all types of branch protection features. If a feature
16014 has additional tuning options, then @samp{standard} sets it to its standard
16016 @samp{pac-ret[+@var{leaf}]} turns on return address signing to its standard
16017 level: signing functions that save the return address to memory (non-leaf
16018 functions will practically always do this) using the a-key. The optional
16019 argument @samp{leaf} can be used to extend the signing to include leaf
16020 functions. The optional argument @samp{b-key} can be used to sign the functions
16021 with the B-key instead of the A-key.
16022 @samp{bti} turns on branch target identification mechanism.
16024 @item -msve-vector-bits=@var{bits}
16025 @opindex msve-vector-bits
16026 Specify the number of bits in an SVE vector register. This option only has
16027 an effect when SVE is enabled.
16029 GCC supports two forms of SVE code generation: ``vector-length
16030 agnostic'' output that works with any size of vector register and
16031 ``vector-length specific'' output that allows GCC to make assumptions
16032 about the vector length when it is useful for optimization reasons.
16033 The possible values of @samp{bits} are: @samp{scalable}, @samp{128},
16034 @samp{256}, @samp{512}, @samp{1024} and @samp{2048}.
16035 Specifying @samp{scalable} selects vector-length agnostic
16036 output. At present @samp{-msve-vector-bits=128} also generates vector-length
16037 agnostic output. All other values generate vector-length specific code.
16038 The behavior of these values may change in future releases and no value except
16039 @samp{scalable} should be relied on for producing code that is portable across
16040 different hardware SVE vector lengths.
16042 The default is @samp{-msve-vector-bits=scalable}, which produces
16043 vector-length agnostic code.
16046 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
16047 @anchor{aarch64-feature-modifiers}
16048 @cindex @option{-march} feature modifiers
16049 @cindex @option{-mcpu} feature modifiers
16050 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
16051 the following and their inverses @option{no@var{feature}}:
16055 Enable CRC extension. This is on by default for
16056 @option{-march=armv8.1-a}.
16058 Enable Crypto extension. This also enables Advanced SIMD and floating-point
16061 Enable floating-point instructions. This is on by default for all possible
16062 values for options @option{-march} and @option{-mcpu}.
16064 Enable Advanced SIMD instructions. This also enables floating-point
16065 instructions. This is on by default for all possible values for options
16066 @option{-march} and @option{-mcpu}.
16068 Enable Scalable Vector Extension instructions. This also enables Advanced
16069 SIMD and floating-point instructions.
16071 Enable Large System Extension instructions. This is on by default for
16072 @option{-march=armv8.1-a}.
16074 Enable Round Double Multiply Accumulate instructions. This is on by default
16075 for @option{-march=armv8.1-a}.
16077 Enable FP16 extension. This also enables floating-point instructions.
16079 Enable FP16 fmla extension. This also enables FP16 extensions and
16080 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.
16083 Enable the RcPc extension. This does not change code generation from GCC,
16084 but is passed on to the assembler, enabling inline asm statements to use
16085 instructions from the RcPc extension.
16087 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
16089 Enable the Armv8-a aes and pmull crypto extension. This also enables Advanced
16092 Enable the Armv8-a sha2 crypto extension. This also enables Advanced SIMD instructions.
16094 Enable the sha512 and sha3 crypto extension. This also enables Advanced SIMD
16095 instructions. Use of this option with architectures prior to Armv8.2-A is not supported.
16097 Enable the sm3 and sm4 crypto extension. This also enables Advanced SIMD instructions.
16098 Use of this option with architectures prior to Armv8.2-A is not supported.
16100 Enable the Statistical Profiling extension. This option is only to enable the
16101 extension at the assembler level and does not affect code generation.
16103 Enable the Armv8.5-a Random Number instructions. This option is only to
16104 enable the extension at the assembler level and does not affect code
16107 Enable the Armv8.5-a Memory Tagging Extensions. This option is only to
16108 enable the extension at the assembler level and does not affect code
16111 Enable the Armv8-a Speculation Barrier instruction. This option is only to
16112 enable the extension at the assembler level and does not affect code
16113 generation. This option is enabled by default for @option{-march=armv8.5-a}.
16115 Enable the Armv8-a Speculative Store Bypass Safe instruction. This option
16116 is only to enable the extension at the assembler level and does not affect code
16117 generation. This option is enabled by default for @option{-march=armv8.5-a}.
16119 Enable the Armv8-a Execution and Data Prediction Restriction instructions.
16120 This option is only to enable the extension at the assembler level and does
16121 not affect code generation. This option is enabled by default for
16123 Enable the Armv8-a Scalable Vector Extension 2. This also enables SVE
16126 Enable SVE2 bitperm instructions. This also enables SVE2 instructions.
16128 Enable SVE2 sm4 instructions. This also enables SVE2 instructions.
16130 Enable SVE2 aes instructions. This also enables SVE2 instructions.
16132 Enable SVE2 sha3 instructions. This also enables SVE2 instructions.
16133 @option{-march=armv8.5-a}.
16135 Enable the Transactional Memory Extension.
16139 Feature @option{crypto} implies @option{aes}, @option{sha2}, and @option{simd},
16140 which implies @option{fp}.
16141 Conversely, @option{nofp} implies @option{nosimd}, which implies
16142 @option{nocrypto}, @option{noaes} and @option{nosha2}.
16144 @node Adapteva Epiphany Options
16145 @subsection Adapteva Epiphany Options
16147 These @samp{-m} options are defined for Adapteva Epiphany:
16150 @item -mhalf-reg-file
16151 @opindex mhalf-reg-file
16152 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
16153 That allows code to run on hardware variants that lack these registers.
16155 @item -mprefer-short-insn-regs
16156 @opindex mprefer-short-insn-regs
16157 Preferentially allocate registers that allow short instruction generation.
16158 This can result in increased instruction count, so this may either reduce or
16159 increase overall code size.
16161 @item -mbranch-cost=@var{num}
16162 @opindex mbranch-cost
16163 Set the cost of branches to roughly @var{num} ``simple'' instructions.
16164 This cost is only a heuristic and is not guaranteed to produce
16165 consistent results across releases.
16169 Enable the generation of conditional moves.
16171 @item -mnops=@var{num}
16173 Emit @var{num} NOPs before every other generated instruction.
16175 @item -mno-soft-cmpsf
16176 @opindex mno-soft-cmpsf
16177 @opindex msoft-cmpsf
16178 For single-precision floating-point comparisons, emit an @code{fsub} instruction
16179 and test the flags. This is faster than a software comparison, but can
16180 get incorrect results in the presence of NaNs, or when two different small
16181 numbers are compared such that their difference is calculated as zero.
16182 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
16183 software comparisons.
16185 @item -mstack-offset=@var{num}
16186 @opindex mstack-offset
16187 Set the offset between the top of the stack and the stack pointer.
16188 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
16189 can be used by leaf functions without stack allocation.
16190 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
16191 Note also that this option changes the ABI; compiling a program with a
16192 different stack offset than the libraries have been compiled with
16193 generally does not work.
16194 This option can be useful if you want to evaluate if a different stack
16195 offset would give you better code, but to actually use a different stack
16196 offset to build working programs, it is recommended to configure the
16197 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
16199 @item -mno-round-nearest
16200 @opindex mno-round-nearest
16201 @opindex mround-nearest
16202 Make the scheduler assume that the rounding mode has been set to
16203 truncating. The default is @option{-mround-nearest}.
16206 @opindex mlong-calls
16207 If not otherwise specified by an attribute, assume all calls might be beyond
16208 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
16209 function address into a register before performing a (otherwise direct) call.
16210 This is the default.
16212 @item -mshort-calls
16213 @opindex short-calls
16214 If not otherwise specified by an attribute, assume all direct calls are
16215 in the range of the @code{b} / @code{bl} instructions, so use these instructions
16216 for direct calls. The default is @option{-mlong-calls}.
16220 Assume addresses can be loaded as 16-bit unsigned values. This does not
16221 apply to function addresses for which @option{-mlong-calls} semantics
16224 @item -mfp-mode=@var{mode}
16226 Set the prevailing mode of the floating-point unit.
16227 This determines the floating-point mode that is provided and expected
16228 at function call and return time. Making this mode match the mode you
16229 predominantly need at function start can make your programs smaller and
16230 faster by avoiding unnecessary mode switches.
16232 @var{mode} can be set to one the following values:
16236 Any mode at function entry is valid, and retained or restored when
16237 the function returns, and when it calls other functions.
16238 This mode is useful for compiling libraries or other compilation units
16239 you might want to incorporate into different programs with different
16240 prevailing FPU modes, and the convenience of being able to use a single
16241 object file outweighs the size and speed overhead for any extra
16242 mode switching that might be needed, compared with what would be needed
16243 with a more specific choice of prevailing FPU mode.
16246 This is the mode used for floating-point calculations with
16247 truncating (i.e.@: round towards zero) rounding mode. That includes
16248 conversion from floating point to integer.
16250 @item round-nearest
16251 This is the mode used for floating-point calculations with
16252 round-to-nearest-or-even rounding mode.
16255 This is the mode used to perform integer calculations in the FPU, e.g.@:
16256 integer multiply, or integer multiply-and-accumulate.
16259 The default is @option{-mfp-mode=caller}
16261 @item -mno-split-lohi
16262 @itemx -mno-postinc
16263 @itemx -mno-postmodify
16264 @opindex mno-split-lohi
16265 @opindex msplit-lohi
16266 @opindex mno-postinc
16268 @opindex mno-postmodify
16269 @opindex mpostmodify
16270 Code generation tweaks that disable, respectively, splitting of 32-bit
16271 loads, generation of post-increment addresses, and generation of
16272 post-modify addresses. The defaults are @option{msplit-lohi},
16273 @option{-mpost-inc}, and @option{-mpost-modify}.
16275 @item -mnovect-double
16276 @opindex mno-vect-double
16277 @opindex mvect-double
16278 Change the preferred SIMD mode to SImode. The default is
16279 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
16281 @item -max-vect-align=@var{num}
16282 @opindex max-vect-align
16283 The maximum alignment for SIMD vector mode types.
16284 @var{num} may be 4 or 8. The default is 8.
16285 Note that this is an ABI change, even though many library function
16286 interfaces are unaffected if they don't use SIMD vector modes
16287 in places that affect size and/or alignment of relevant types.
16289 @item -msplit-vecmove-early
16290 @opindex msplit-vecmove-early
16291 Split vector moves into single word moves before reload. In theory this
16292 can give better register allocation, but so far the reverse seems to be
16293 generally the case.
16295 @item -m1reg-@var{reg}
16297 Specify a register to hold the constant @minus{}1, which makes loading small negative
16298 constants and certain bitmasks faster.
16299 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
16300 which specify use of that register as a fixed register,
16301 and @samp{none}, which means that no register is used for this
16302 purpose. The default is @option{-m1reg-none}.
16306 @node AMD GCN Options
16307 @subsection AMD GCN Options
16308 @cindex AMD GCN Options
16310 These options are defined specifically for the AMD GCN port.
16314 @item -march=@var{gpu}
16316 @itemx -mtune=@var{gpu}
16318 Set architecture type or tuning for @var{gpu}. Supported values for @var{gpu}
16324 Compile for GCN3 Fiji devices (gfx803).
16327 Compile for GCN5 Vega 10 devices (gfx900).
16330 Compile for GCN5 Vega 20 devices (gfx906).
16334 @item -mstack-size=@var{bytes}
16335 @opindex mstack-size
16336 Specify how many @var{bytes} of stack space will be requested for each GPU
16337 thread (wave-front). Beware that there may be many threads and limited memory
16338 available. The size of the stack allocation may also have an impact on
16339 run-time performance. The default is 32KB when using OpenACC or OpenMP, and
16345 @subsection ARC Options
16346 @cindex ARC options
16348 The following options control the architecture variant for which code
16351 @c architecture variants
16354 @item -mbarrel-shifter
16355 @opindex mbarrel-shifter
16356 Generate instructions supported by barrel shifter. This is the default
16357 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
16360 @opindex mjli-alawys
16361 Force to call a function using jli_s instruction. This option is
16362 valid only for ARCv2 architecture.
16364 @item -mcpu=@var{cpu}
16366 Set architecture type, register usage, and instruction scheduling
16367 parameters for @var{cpu}. There are also shortcut alias options
16368 available for backward compatibility and convenience. Supported
16369 values for @var{cpu} are
16375 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
16379 Compile for ARC601. Alias: @option{-mARC601}.
16384 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
16385 This is the default when configured with @option{--with-cpu=arc700}@.
16388 Compile for ARC EM.
16391 Compile for ARC HS.
16394 Compile for ARC EM CPU with no hardware extensions.
16397 Compile for ARC EM4 CPU.
16400 Compile for ARC EM4 DMIPS CPU.
16403 Compile for ARC EM4 DMIPS CPU with the single-precision floating-point
16407 Compile for ARC EM4 DMIPS CPU with single-precision floating-point and
16408 double assist instructions.
16411 Compile for ARC HS CPU with no hardware extensions except the atomic
16415 Compile for ARC HS34 CPU.
16418 Compile for ARC HS38 CPU.
16421 Compile for ARC HS38 CPU with all hardware extensions on.
16424 Compile for ARC 600 CPU with @code{norm} instructions enabled.
16426 @item arc600_mul32x16
16427 Compile for ARC 600 CPU with @code{norm} and 32x16-bit multiply
16428 instructions enabled.
16431 Compile for ARC 600 CPU with @code{norm} and @code{mul64}-family
16432 instructions enabled.
16435 Compile for ARC 601 CPU with @code{norm} instructions enabled.
16437 @item arc601_mul32x16
16438 Compile for ARC 601 CPU with @code{norm} and 32x16-bit multiply
16439 instructions enabled.
16442 Compile for ARC 601 CPU with @code{norm} and @code{mul64}-family
16443 instructions enabled.
16446 Compile for ARC 700 on NPS400 chip.
16449 Compile for ARC EM minimalist configuration featuring reduced register
16456 @itemx -mdpfp-compact
16457 @opindex mdpfp-compact
16458 Generate double-precision FPX instructions, tuned for the compact
16462 @opindex mdpfp-fast
16463 Generate double-precision FPX instructions, tuned for the fast
16466 @item -mno-dpfp-lrsr
16467 @opindex mno-dpfp-lrsr
16468 Disable @code{lr} and @code{sr} instructions from using FPX extension
16473 Generate extended arithmetic instructions. Currently only
16474 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
16475 supported. This is always enabled for @option{-mcpu=ARC700}.
16480 Do not generate @code{mpy}-family instructions for ARC700. This option is
16485 Generate 32x16-bit multiply and multiply-accumulate instructions.
16489 Generate @code{mul64} and @code{mulu64} instructions.
16490 Only valid for @option{-mcpu=ARC600}.
16494 Generate @code{norm} instructions. This is the default if @option{-mcpu=ARC700}
16499 @itemx -mspfp-compact
16500 @opindex mspfp-compact
16501 Generate single-precision FPX instructions, tuned for the compact
16505 @opindex mspfp-fast
16506 Generate single-precision FPX instructions, tuned for the fast
16511 Enable generation of ARC SIMD instructions via target-specific
16512 builtins. Only valid for @option{-mcpu=ARC700}.
16515 @opindex msoft-float
16516 This option ignored; it is provided for compatibility purposes only.
16517 Software floating-point code is emitted by default, and this default
16518 can overridden by FPX options; @option{-mspfp}, @option{-mspfp-compact}, or
16519 @option{-mspfp-fast} for single precision, and @option{-mdpfp},
16520 @option{-mdpfp-compact}, or @option{-mdpfp-fast} for double precision.
16524 Generate @code{swap} instructions.
16528 This enables use of the locked load/store conditional extension to implement
16529 atomic memory built-in functions. Not available for ARC 6xx or ARC
16534 Enable @code{div} and @code{rem} instructions for ARCv2 cores.
16536 @item -mcode-density
16537 @opindex mcode-density
16538 Enable code density instructions for ARC EM.
16539 This option is on by default for ARC HS.
16543 Enable double load/store operations for ARC HS cores.
16545 @item -mtp-regno=@var{regno}
16547 Specify thread pointer register number.
16549 @item -mmpy-option=@var{multo}
16550 @opindex mmpy-option
16551 Compile ARCv2 code with a multiplier design option. You can specify
16552 the option using either a string or numeric value for @var{multo}.
16553 @samp{wlh1} is the default value. The recognized values are:
16558 No multiplier available.
16562 16x16 multiplier, fully pipelined.
16563 The following instructions are enabled: @code{mpyw} and @code{mpyuw}.
16567 32x32 multiplier, fully
16568 pipelined (1 stage). The following instructions are additionally
16569 enabled: @code{mpy}, @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16573 32x32 multiplier, fully pipelined
16574 (2 stages). The following instructions are additionally enabled: @code{mpy},
16575 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16579 Two 16x16 multipliers, blocking,
16580 sequential. The following instructions are additionally enabled: @code{mpy},
16581 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16585 One 16x16 multiplier, blocking,
16586 sequential. The following instructions are additionally enabled: @code{mpy},
16587 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16591 One 32x4 multiplier, blocking,
16592 sequential. The following instructions are additionally enabled: @code{mpy},
16593 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16597 ARC HS SIMD support.
16601 ARC HS SIMD support.
16605 ARC HS SIMD support.
16609 This option is only available for ARCv2 cores@.
16611 @item -mfpu=@var{fpu}
16613 Enables support for specific floating-point hardware extensions for ARCv2
16614 cores. Supported values for @var{fpu} are:
16619 Enables support for single-precision floating-point hardware
16623 Enables support for double-precision floating-point hardware
16624 extensions. The single-precision floating-point extension is also
16625 enabled. Not available for ARC EM@.
16628 Enables support for double-precision floating-point hardware
16629 extensions using double-precision assist instructions. The single-precision
16630 floating-point extension is also enabled. This option is
16631 only available for ARC EM@.
16634 Enables support for double-precision floating-point hardware
16635 extensions using double-precision assist instructions.
16636 The single-precision floating-point, square-root, and divide
16637 extensions are also enabled. This option is
16638 only available for ARC EM@.
16641 Enables support for double-precision floating-point hardware
16642 extensions using double-precision assist instructions.
16643 The single-precision floating-point and fused multiply and add
16644 hardware extensions are also enabled. This option is
16645 only available for ARC EM@.
16648 Enables support for double-precision floating-point hardware
16649 extensions using double-precision assist instructions.
16650 All single-precision floating-point hardware extensions are also
16651 enabled. This option is only available for ARC EM@.
16654 Enables support for single-precision floating-point, square-root and divide
16655 hardware extensions@.
16658 Enables support for double-precision floating-point, square-root and divide
16659 hardware extensions. This option
16660 includes option @samp{fpus_div}. Not available for ARC EM@.
16663 Enables support for single-precision floating-point and
16664 fused multiply and add hardware extensions@.
16667 Enables support for double-precision floating-point and
16668 fused multiply and add hardware extensions. This option
16669 includes option @samp{fpus_fma}. Not available for ARC EM@.
16672 Enables support for all single-precision floating-point hardware
16676 Enables support for all single- and double-precision floating-point
16677 hardware extensions. Not available for ARC EM@.
16681 @item -mirq-ctrl-saved=@var{register-range}, @var{blink}, @var{lp_count}
16682 @opindex mirq-ctrl-saved
16683 Specifies general-purposes registers that the processor automatically
16684 saves/restores on interrupt entry and exit. @var{register-range} is
16685 specified as two registers separated by a dash. The register range
16686 always starts with @code{r0}, the upper limit is @code{fp} register.
16687 @var{blink} and @var{lp_count} are optional. This option is only
16688 valid for ARC EM and ARC HS cores.
16690 @item -mrgf-banked-regs=@var{number}
16691 @opindex mrgf-banked-regs
16692 Specifies the number of registers replicated in second register bank
16693 on entry to fast interrupt. Fast interrupts are interrupts with the
16694 highest priority level P0. These interrupts save only PC and STATUS32
16695 registers to avoid memory transactions during interrupt entry and exit
16696 sequences. Use this option when you are using fast interrupts in an
16697 ARC V2 family processor. Permitted values are 4, 8, 16, and 32.
16699 @item -mlpc-width=@var{width}
16700 @opindex mlpc-width
16701 Specify the width of the @code{lp_count} register. Valid values for
16702 @var{width} are 8, 16, 20, 24, 28 and 32 bits. The default width is
16703 fixed to 32 bits. If the width is less than 32, the compiler does not
16704 attempt to transform loops in your program to use the zero-delay loop
16705 mechanism unless it is known that the @code{lp_count} register can
16706 hold the required loop-counter value. Depending on the width
16707 specified, the compiler and run-time library might continue to use the
16708 loop mechanism for various needs. This option defines macro
16709 @code{__ARC_LPC_WIDTH__} with the value of @var{width}.
16713 This option instructs the compiler to generate code for a 16-entry
16714 register file. This option defines the @code{__ARC_RF16__}
16715 preprocessor macro.
16717 @item -mbranch-index
16718 @opindex mbranch-index
16719 Enable use of @code{bi} or @code{bih} instructions to implement jump
16724 The following options are passed through to the assembler, and also
16725 define preprocessor macro symbols.
16727 @c Flags used by the assembler, but for which we define preprocessor
16728 @c macro symbols as well.
16731 @opindex mdsp-packa
16732 Passed down to the assembler to enable the DSP Pack A extensions.
16733 Also sets the preprocessor symbol @code{__Xdsp_packa}. This option is
16738 Passed down to the assembler to enable the dual Viterbi butterfly
16739 extension. Also sets the preprocessor symbol @code{__Xdvbf}. This
16740 option is deprecated.
16742 @c ARC700 4.10 extension instruction
16745 Passed down to the assembler to enable the locked load/store
16746 conditional extension. Also sets the preprocessor symbol
16751 Passed down to the assembler. Also sets the preprocessor symbol
16752 @code{__Xxmac_d16}. This option is deprecated.
16756 Passed down to the assembler. Also sets the preprocessor symbol
16757 @code{__Xxmac_24}. This option is deprecated.
16759 @c ARC700 4.10 extension instruction
16762 Passed down to the assembler to enable the 64-bit time-stamp counter
16763 extension instruction. Also sets the preprocessor symbol
16764 @code{__Xrtsc}. This option is deprecated.
16766 @c ARC700 4.10 extension instruction
16769 Passed down to the assembler to enable the swap byte ordering
16770 extension instruction. Also sets the preprocessor symbol
16774 @opindex mtelephony
16775 Passed down to the assembler to enable dual- and single-operand
16776 instructions for telephony. Also sets the preprocessor symbol
16777 @code{__Xtelephony}. This option is deprecated.
16781 Passed down to the assembler to enable the XY memory extension. Also
16782 sets the preprocessor symbol @code{__Xxy}.
16786 The following options control how the assembly code is annotated:
16788 @c Assembly annotation options
16792 Annotate assembler instructions with estimated addresses.
16794 @item -mannotate-align
16795 @opindex mannotate-align
16796 Explain what alignment considerations lead to the decision to make an
16797 instruction short or long.
16801 The following options are passed through to the linker:
16803 @c options passed through to the linker
16807 Passed through to the linker, to specify use of the @code{arclinux} emulation.
16808 This option is enabled by default in tool chains built for
16809 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
16810 when profiling is not requested.
16812 @item -marclinux_prof
16813 @opindex marclinux_prof
16814 Passed through to the linker, to specify use of the
16815 @code{arclinux_prof} emulation. This option is enabled by default in
16816 tool chains built for @w{@code{arc-linux-uclibc}} and
16817 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
16821 The following options control the semantics of generated code:
16823 @c semantically relevant code generation options
16826 @opindex mlong-calls
16827 Generate calls as register indirect calls, thus providing access
16828 to the full 32-bit address range.
16830 @item -mmedium-calls
16831 @opindex mmedium-calls
16832 Don't use less than 25-bit addressing range for calls, which is the
16833 offset available for an unconditional branch-and-link
16834 instruction. Conditional execution of function calls is suppressed, to
16835 allow use of the 25-bit range, rather than the 21-bit range with
16836 conditional branch-and-link. This is the default for tool chains built
16837 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
16841 Put definitions of externally-visible data in a small data section if
16842 that data is no bigger than @var{num} bytes. The default value of
16843 @var{num} is 4 for any ARC configuration, or 8 when we have double
16844 load/store operations.
16849 Do not generate sdata references. This is the default for tool chains
16850 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
16853 @item -mvolatile-cache
16854 @opindex mvolatile-cache
16855 Use ordinarily cached memory accesses for volatile references. This is the
16858 @item -mno-volatile-cache
16859 @opindex mno-volatile-cache
16860 @opindex mvolatile-cache
16861 Enable cache bypass for volatile references.
16865 The following options fine tune code generation:
16866 @c code generation tuning options
16869 @opindex malign-call
16870 Do alignment optimizations for call instructions.
16872 @item -mauto-modify-reg
16873 @opindex mauto-modify-reg
16874 Enable the use of pre/post modify with register displacement.
16876 @item -mbbit-peephole
16877 @opindex mbbit-peephole
16878 Enable bbit peephole2.
16882 This option disables a target-specific pass in @file{arc_reorg} to
16883 generate compare-and-branch (@code{br@var{cc}}) instructions.
16884 It has no effect on
16885 generation of these instructions driven by the combiner pass.
16887 @item -mcase-vector-pcrel
16888 @opindex mcase-vector-pcrel
16889 Use PC-relative switch case tables to enable case table shortening.
16890 This is the default for @option{-Os}.
16892 @item -mcompact-casesi
16893 @opindex mcompact-casesi
16894 Enable compact @code{casesi} pattern. This is the default for @option{-Os},
16895 and only available for ARCv1 cores. This option is deprecated.
16897 @item -mno-cond-exec
16898 @opindex mno-cond-exec
16899 Disable the ARCompact-specific pass to generate conditional
16900 execution instructions.
16902 Due to delay slot scheduling and interactions between operand numbers,
16903 literal sizes, instruction lengths, and the support for conditional execution,
16904 the target-independent pass to generate conditional execution is often lacking,
16905 so the ARC port has kept a special pass around that tries to find more
16906 conditional execution generation opportunities after register allocation,
16907 branch shortening, and delay slot scheduling have been done. This pass
16908 generally, but not always, improves performance and code size, at the cost of
16909 extra compilation time, which is why there is an option to switch it off.
16910 If you have a problem with call instructions exceeding their allowable
16911 offset range because they are conditionalized, you should consider using
16912 @option{-mmedium-calls} instead.
16914 @item -mearly-cbranchsi
16915 @opindex mearly-cbranchsi
16916 Enable pre-reload use of the @code{cbranchsi} pattern.
16918 @item -mexpand-adddi
16919 @opindex mexpand-adddi
16920 Expand @code{adddi3} and @code{subdi3} at RTL generation time into
16921 @code{add.f}, @code{adc} etc. This option is deprecated.
16923 @item -mindexed-loads
16924 @opindex mindexed-loads
16925 Enable the use of indexed loads. This can be problematic because some
16926 optimizers then assume that indexed stores exist, which is not
16931 Enable Local Register Allocation. This is still experimental for ARC,
16932 so by default the compiler uses standard reload
16933 (i.e.@: @option{-mno-lra}).
16935 @item -mlra-priority-none
16936 @opindex mlra-priority-none
16937 Don't indicate any priority for target registers.
16939 @item -mlra-priority-compact
16940 @opindex mlra-priority-compact
16941 Indicate target register priority for r0..r3 / r12..r15.
16943 @item -mlra-priority-noncompact
16944 @opindex mlra-priority-noncompact
16945 Reduce target register priority for r0..r3 / r12..r15.
16948 @opindex mmillicode
16949 When optimizing for size (using @option{-Os}), prologues and epilogues
16950 that have to save or restore a large number of registers are often
16951 shortened by using call to a special function in libgcc; this is
16952 referred to as a @emph{millicode} call. As these calls can pose
16953 performance issues, and/or cause linking issues when linking in a
16954 nonstandard way, this option is provided to turn on or off millicode
16957 @item -mcode-density-frame
16958 @opindex mcode-density-frame
16959 This option enable the compiler to emit @code{enter} and @code{leave}
16960 instructions. These instructions are only valid for CPUs with
16961 code-density feature.
16964 @opindex mmixed-code
16965 Tweak register allocation to help 16-bit instruction generation.
16966 This generally has the effect of decreasing the average instruction size
16967 while increasing the instruction count.
16971 Enable @samp{q} instruction alternatives.
16972 This is the default for @option{-Os}.
16976 Enable @samp{Rcq} constraint handling.
16977 Most short code generation depends on this.
16978 This is the default.
16982 Enable @samp{Rcw} constraint handling.
16983 Most ccfsm condexec mostly depends on this.
16984 This is the default.
16986 @item -msize-level=@var{level}
16987 @opindex msize-level
16988 Fine-tune size optimization with regards to instruction lengths and alignment.
16989 The recognized values for @var{level} are:
16992 No size optimization. This level is deprecated and treated like @samp{1}.
16995 Short instructions are used opportunistically.
16998 In addition, alignment of loops and of code after barriers are dropped.
17001 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
17005 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
17006 the behavior when this is not set is equivalent to level @samp{1}.
17008 @item -mtune=@var{cpu}
17010 Set instruction scheduling parameters for @var{cpu}, overriding any implied
17011 by @option{-mcpu=}.
17013 Supported values for @var{cpu} are
17017 Tune for ARC600 CPU.
17020 Tune for ARC601 CPU.
17023 Tune for ARC700 CPU with standard multiplier block.
17026 Tune for ARC700 CPU with XMAC block.
17029 Tune for ARC725D CPU.
17032 Tune for ARC750D CPU.
17036 @item -mmultcost=@var{num}
17038 Cost to assume for a multiply instruction, with @samp{4} being equal to a
17039 normal instruction.
17041 @item -munalign-prob-threshold=@var{probability}
17042 @opindex munalign-prob-threshold
17043 Set probability threshold for unaligning branches.
17044 When tuning for @samp{ARC700} and optimizing for speed, branches without
17045 filled delay slot are preferably emitted unaligned and long, unless
17046 profiling indicates that the probability for the branch to be taken
17047 is below @var{probability}. @xref{Cross-profiling}.
17048 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
17052 The following options are maintained for backward compatibility, but
17053 are now deprecated and will be removed in a future release:
17055 @c Deprecated options
17063 @opindex mbig-endian
17066 Compile code for big-endian targets. Use of these options is now
17067 deprecated. Big-endian code is supported by configuring GCC to build
17068 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets,
17069 for which big endian is the default.
17071 @item -mlittle-endian
17072 @opindex mlittle-endian
17075 Compile code for little-endian targets. Use of these options is now
17076 deprecated. Little-endian code is supported by configuring GCC to build
17077 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets,
17078 for which little endian is the default.
17080 @item -mbarrel_shifter
17081 @opindex mbarrel_shifter
17082 Replaced by @option{-mbarrel-shifter}.
17084 @item -mdpfp_compact
17085 @opindex mdpfp_compact
17086 Replaced by @option{-mdpfp-compact}.
17089 @opindex mdpfp_fast
17090 Replaced by @option{-mdpfp-fast}.
17093 @opindex mdsp_packa
17094 Replaced by @option{-mdsp-packa}.
17098 Replaced by @option{-mea}.
17102 Replaced by @option{-mmac-24}.
17106 Replaced by @option{-mmac-d16}.
17108 @item -mspfp_compact
17109 @opindex mspfp_compact
17110 Replaced by @option{-mspfp-compact}.
17113 @opindex mspfp_fast
17114 Replaced by @option{-mspfp-fast}.
17116 @item -mtune=@var{cpu}
17118 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
17119 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
17120 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively.
17122 @item -multcost=@var{num}
17124 Replaced by @option{-mmultcost}.
17129 @subsection ARM Options
17130 @cindex ARM options
17132 These @samp{-m} options are defined for the ARM port:
17135 @item -mabi=@var{name}
17137 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
17138 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
17141 @opindex mapcs-frame
17142 Generate a stack frame that is compliant with the ARM Procedure Call
17143 Standard for all functions, even if this is not strictly necessary for
17144 correct execution of the code. Specifying @option{-fomit-frame-pointer}
17145 with this option causes the stack frames not to be generated for
17146 leaf functions. The default is @option{-mno-apcs-frame}.
17147 This option is deprecated.
17151 This is a synonym for @option{-mapcs-frame} and is deprecated.
17154 @c not currently implemented
17155 @item -mapcs-stack-check
17156 @opindex mapcs-stack-check
17157 Generate code to check the amount of stack space available upon entry to
17158 every function (that actually uses some stack space). If there is
17159 insufficient space available then either the function
17160 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
17161 called, depending upon the amount of stack space required. The runtime
17162 system is required to provide these functions. The default is
17163 @option{-mno-apcs-stack-check}, since this produces smaller code.
17165 @c not currently implemented
17166 @item -mapcs-reentrant
17167 @opindex mapcs-reentrant
17168 Generate reentrant, position-independent code. The default is
17169 @option{-mno-apcs-reentrant}.
17172 @item -mthumb-interwork
17173 @opindex mthumb-interwork
17174 Generate code that supports calling between the ARM and Thumb
17175 instruction sets. Without this option, on pre-v5 architectures, the
17176 two instruction sets cannot be reliably used inside one program. The
17177 default is @option{-mno-thumb-interwork}, since slightly larger code
17178 is generated when @option{-mthumb-interwork} is specified. In AAPCS
17179 configurations this option is meaningless.
17181 @item -mno-sched-prolog
17182 @opindex mno-sched-prolog
17183 @opindex msched-prolog
17184 Prevent the reordering of instructions in the function prologue, or the
17185 merging of those instruction with the instructions in the function's
17186 body. This means that all functions start with a recognizable set
17187 of instructions (or in fact one of a choice from a small set of
17188 different function prologues), and this information can be used to
17189 locate the start of functions inside an executable piece of code. The
17190 default is @option{-msched-prolog}.
17192 @item -mfloat-abi=@var{name}
17193 @opindex mfloat-abi
17194 Specifies which floating-point ABI to use. Permissible values
17195 are: @samp{soft}, @samp{softfp} and @samp{hard}.
17197 Specifying @samp{soft} causes GCC to generate output containing
17198 library calls for floating-point operations.
17199 @samp{softfp} allows the generation of code using hardware floating-point
17200 instructions, but still uses the soft-float calling conventions.
17201 @samp{hard} allows generation of floating-point instructions
17202 and uses FPU-specific calling conventions.
17204 The default depends on the specific target configuration. Note that
17205 the hard-float and soft-float ABIs are not link-compatible; you must
17206 compile your entire program with the same ABI, and link with a
17207 compatible set of libraries.
17209 @item -mgeneral-regs-only
17210 @opindex mgeneral-regs-only
17211 Generate code which uses only the general-purpose registers. This will prevent
17212 the compiler from using floating-point and Advanced SIMD registers but will not
17213 impose any restrictions on the assembler.
17215 @item -mlittle-endian
17216 @opindex mlittle-endian
17217 Generate code for a processor running in little-endian mode. This is
17218 the default for all standard configurations.
17221 @opindex mbig-endian
17222 Generate code for a processor running in big-endian mode; the default is
17223 to compile code for a little-endian processor.
17228 When linking a big-endian image select between BE8 and BE32 formats.
17229 The option has no effect for little-endian images and is ignored. The
17230 default is dependent on the selected target architecture. For ARMv6
17231 and later architectures the default is BE8, for older architectures
17232 the default is BE32. BE32 format has been deprecated by ARM.
17234 @item -march=@var{name}@r{[}+extension@dots{}@r{]}
17236 This specifies the name of the target ARM architecture. GCC uses this
17237 name to determine what kind of instructions it can emit when generating
17238 assembly code. This option can be used in conjunction with or instead
17239 of the @option{-mcpu=} option.
17241 Permissible names are:
17243 @samp{armv5t}, @samp{armv5te},
17244 @samp{armv6}, @samp{armv6j}, @samp{armv6k}, @samp{armv6kz}, @samp{armv6t2},
17245 @samp{armv6z}, @samp{armv6zk},
17246 @samp{armv7}, @samp{armv7-a}, @samp{armv7ve},
17247 @samp{armv8-a}, @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a},
17252 @samp{armv6-m}, @samp{armv6s-m},
17253 @samp{armv7-m}, @samp{armv7e-m},
17254 @samp{armv8-m.base}, @samp{armv8-m.main},
17255 @samp{iwmmxt} and @samp{iwmmxt2}.
17257 Additionally, the following architectures, which lack support for the
17258 Thumb execution state, are recognized but support is deprecated: @samp{armv4}.
17260 Many of the architectures support extensions. These can be added by
17261 appending @samp{+@var{extension}} to the architecture name. Extension
17262 options are processed in order and capabilities accumulate. An extension
17263 will also enable any necessary base extensions
17264 upon which it depends. For example, the @samp{+crypto} extension
17265 will always enable the @samp{+simd} extension. The exception to the
17266 additive construction is for extensions that are prefixed with
17267 @samp{+no@dots{}}: these extensions disable the specified option and
17268 any other extensions that may depend on the presence of that
17271 For example, @samp{-march=armv7-a+simd+nofp+vfpv4} is equivalent to
17272 writing @samp{-march=armv7-a+vfpv4} since the @samp{+simd} option is
17273 entirely disabled by the @samp{+nofp} option that follows it.
17275 Most extension names are generically named, but have an effect that is
17276 dependent upon the architecture to which it is applied. For example,
17277 the @samp{+simd} option can be applied to both @samp{armv7-a} and
17278 @samp{armv8-a} architectures, but will enable the original ARMv7-A
17279 Advanced SIMD (Neon) extensions for @samp{armv7-a} and the ARMv8-A
17280 variant for @samp{armv8-a}.
17282 The table below lists the supported extensions for each architecture.
17283 Architectures not mentioned do not support any extensions.
17296 The VFPv2 floating-point instructions. The extension @samp{+vfpv2} can be
17297 used as an alias for this extension.
17300 Disable the floating-point instructions.
17304 The common subset of the ARMv7-A, ARMv7-R and ARMv7-M architectures.
17307 The VFPv3 floating-point instructions, with 16 double-precision
17308 registers. The extension @samp{+vfpv3-d16} can be used as an alias
17309 for this extension. Note that floating-point is not supported by the
17310 base ARMv7-M architecture, but is compatible with both the ARMv7-A and
17311 ARMv7-R architectures.
17314 Disable the floating-point instructions.
17320 The multiprocessing extension.
17323 The security extension.
17326 The VFPv3 floating-point instructions, with 16 double-precision
17327 registers. The extension @samp{+vfpv3-d16} can be used as an alias
17328 for this extension.
17331 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
17332 The extensions @samp{+neon} and @samp{+neon-vfpv3} can be used as aliases
17333 for this extension.
17336 The VFPv3 floating-point instructions, with 32 double-precision
17339 @item +vfpv3-d16-fp16
17340 The VFPv3 floating-point instructions, with 16 double-precision
17341 registers and the half-precision floating-point conversion operations.
17344 The VFPv3 floating-point instructions, with 32 double-precision
17345 registers and the half-precision floating-point conversion operations.
17348 The VFPv4 floating-point instructions, with 16 double-precision
17352 The VFPv4 floating-point instructions, with 32 double-precision
17356 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
17357 the half-precision floating-point conversion operations.
17360 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions.
17363 Disable the Advanced SIMD instructions (does not disable floating point).
17366 Disable the floating-point and Advanced SIMD instructions.
17370 The extended version of the ARMv7-A architecture with support for
17374 The VFPv4 floating-point instructions, with 16 double-precision registers.
17375 The extension @samp{+vfpv4-d16} can be used as an alias for this extension.
17378 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions. The
17379 extension @samp{+neon-vfpv4} can be used as an alias for this extension.
17382 The VFPv3 floating-point instructions, with 16 double-precision
17386 The VFPv3 floating-point instructions, with 32 double-precision
17389 @item +vfpv3-d16-fp16
17390 The VFPv3 floating-point instructions, with 16 double-precision
17391 registers and the half-precision floating-point conversion operations.
17394 The VFPv3 floating-point instructions, with 32 double-precision
17395 registers and the half-precision floating-point conversion operations.
17398 The VFPv4 floating-point instructions, with 16 double-precision
17402 The VFPv4 floating-point instructions, with 32 double-precision
17406 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
17407 The extension @samp{+neon-vfpv3} can be used as an alias for this extension.
17410 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
17411 the half-precision floating-point conversion operations.
17414 Disable the Advanced SIMD instructions (does not disable floating point).
17417 Disable the floating-point and Advanced SIMD instructions.
17423 The Cyclic Redundancy Check (CRC) instructions.
17425 The ARMv8-A Advanced SIMD and floating-point instructions.
17427 The cryptographic instructions.
17429 Disable the cryptographic instructions.
17431 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17433 Speculation Barrier Instruction.
17435 Execution and Data Prediction Restriction Instructions.
17441 The ARMv8.1-A Advanced SIMD and floating-point instructions.
17444 The cryptographic instructions. This also enables the Advanced SIMD and
17445 floating-point instructions.
17448 Disable the cryptographic instructions.
17451 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17454 Speculation Barrier Instruction.
17457 Execution and Data Prediction Restriction Instructions.
17464 The half-precision floating-point data processing instructions.
17465 This also enables the Advanced SIMD and floating-point instructions.
17468 The half-precision floating-point fmla extension. This also enables
17469 the half-precision floating-point extension and Advanced SIMD and
17470 floating-point instructions.
17473 The ARMv8.1-A Advanced SIMD and floating-point instructions.
17476 The cryptographic instructions. This also enables the Advanced SIMD and
17477 floating-point instructions.
17480 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
17483 Disable the cryptographic extension.
17486 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17489 Speculation Barrier Instruction.
17492 Execution and Data Prediction Restriction Instructions.
17498 The half-precision floating-point data processing instructions.
17499 This also enables the Advanced SIMD and floating-point instructions as well
17500 as the Dot Product extension and the half-precision floating-point fmla
17504 The ARMv8.3-A Advanced SIMD and floating-point instructions as well as the
17505 Dot Product extension.
17508 The cryptographic instructions. This also enables the Advanced SIMD and
17509 floating-point instructions as well as the Dot Product extension.
17512 Disable the cryptographic extension.
17515 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17518 Speculation Barrier Instruction.
17521 Execution and Data Prediction Restriction Instructions.
17527 The half-precision floating-point data processing instructions.
17528 This also enables the Advanced SIMD and floating-point instructions as well
17529 as the Dot Product extension and the half-precision floating-point fmla
17533 The ARMv8.3-A Advanced SIMD and floating-point instructions as well as the
17534 Dot Product extension.
17537 The cryptographic instructions. This also enables the Advanced SIMD and
17538 floating-point instructions as well as the Dot Product extension.
17541 Disable the cryptographic extension.
17544 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17550 The single-precision VFPv3 floating-point instructions. The extension
17551 @samp{+vfpv3xd} can be used as an alias for this extension.
17554 The VFPv3 floating-point instructions with 16 double-precision registers.
17555 The extension +vfpv3-d16 can be used as an alias for this extension.
17557 @item +vfpv3xd-d16-fp16
17558 The single-precision VFPv3 floating-point instructions with 16 double-precision
17559 registers and the half-precision floating-point conversion operations.
17561 @item +vfpv3-d16-fp16
17562 The VFPv3 floating-point instructions with 16 double-precision
17563 registers and the half-precision floating-point conversion operations.
17566 Disable the floating-point extension.
17569 The ARM-state integer division instructions.
17572 Disable the ARM-state integer division extension.
17578 The single-precision VFPv4 floating-point instructions.
17581 The single-precision FPv5 floating-point instructions.
17584 The single- and double-precision FPv5 floating-point instructions.
17587 Disable the floating-point extensions.
17593 The DSP instructions.
17596 Disable the DSP extension.
17599 The single-precision floating-point instructions.
17602 The single- and double-precision floating-point instructions.
17605 Disable the floating-point extension.
17611 The Cyclic Redundancy Check (CRC) instructions.
17613 The single-precision FPv5 floating-point instructions.
17615 The ARMv8-A Advanced SIMD and floating-point instructions.
17617 The cryptographic instructions.
17619 Disable the cryptographic instructions.
17621 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17626 @option{-march=native} causes the compiler to auto-detect the architecture
17627 of the build computer. At present, this feature is only supported on
17628 GNU/Linux, and not all architectures are recognized. If the auto-detect
17629 is unsuccessful the option has no effect.
17631 @item -mtune=@var{name}
17633 This option specifies the name of the target ARM processor for
17634 which GCC should tune the performance of the code.
17635 For some ARM implementations better performance can be obtained by using
17637 Permissible names are: @samp{arm7tdmi}, @samp{arm7tdmi-s}, @samp{arm710t},
17638 @samp{arm720t}, @samp{arm740t}, @samp{strongarm}, @samp{strongarm110},
17639 @samp{strongarm1100}, 0@samp{strongarm1110}, @samp{arm8}, @samp{arm810},
17640 @samp{arm9}, @samp{arm9e}, @samp{arm920}, @samp{arm920t}, @samp{arm922t},
17641 @samp{arm946e-s}, @samp{arm966e-s}, @samp{arm968e-s}, @samp{arm926ej-s},
17642 @samp{arm940t}, @samp{arm9tdmi}, @samp{arm10tdmi}, @samp{arm1020t},
17643 @samp{arm1026ej-s}, @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
17644 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
17645 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
17646 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
17647 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
17648 @samp{cortex-a32}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
17649 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
17650 @samp{cortex-a76}, @samp{cortex-a76ae}, @samp{cortex-a77},
17651 @samp{ares}, @samp{cortex-r4}, @samp{cortex-r4f},
17652 @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8}, @samp{cortex-r52},
17653 @samp{cortex-m0}, @samp{cortex-m0plus}, @samp{cortex-m1}, @samp{cortex-m3},
17654 @samp{cortex-m4}, @samp{cortex-m7}, @samp{cortex-m23}, @samp{cortex-m33},
17655 @samp{cortex-m35p},
17656 @samp{cortex-m1.small-multiply}, @samp{cortex-m0.small-multiply},
17657 @samp{cortex-m0plus.small-multiply}, @samp{exynos-m1}, @samp{marvell-pj4},
17658 @samp{neoverse-n1}, @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2},
17659 @samp{ep9312}, @samp{fa526}, @samp{fa626}, @samp{fa606te}, @samp{fa626te},
17660 @samp{fmp626}, @samp{fa726te}, @samp{xgene1}.
17662 Additionally, this option can specify that GCC should tune the performance
17663 of the code for a big.LITTLE system. Permissible names are:
17664 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
17665 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
17666 @samp{cortex-a72.cortex-a35}, @samp{cortex-a73.cortex-a53},
17667 @samp{cortex-a75.cortex-a55}, @samp{cortex-a76.cortex-a55}.
17669 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
17670 performance for a blend of processors within architecture @var{arch}.
17671 The aim is to generate code that run well on the current most popular
17672 processors, balancing between optimizations that benefit some CPUs in the
17673 range, and avoiding performance pitfalls of other CPUs. The effects of
17674 this option may change in future GCC versions as CPU models come and go.
17676 @option{-mtune} permits the same extension options as @option{-mcpu}, but
17677 the extension options do not affect the tuning of the generated code.
17679 @option{-mtune=native} causes the compiler to auto-detect the CPU
17680 of the build computer. At present, this feature is only supported on
17681 GNU/Linux, and not all architectures are recognized. If the auto-detect is
17682 unsuccessful the option has no effect.
17684 @item -mcpu=@var{name}@r{[}+extension@dots{}@r{]}
17686 This specifies the name of the target ARM processor. GCC uses this name
17687 to derive the name of the target ARM architecture (as if specified
17688 by @option{-march}) and the ARM processor type for which to tune for
17689 performance (as if specified by @option{-mtune}). Where this option
17690 is used in conjunction with @option{-march} or @option{-mtune},
17691 those options take precedence over the appropriate part of this option.
17693 Many of the supported CPUs implement optional architectural
17694 extensions. Where this is so the architectural extensions are
17695 normally enabled by default. If implementations that lack the
17696 extension exist, then the extension syntax can be used to disable
17697 those extensions that have been omitted. For floating-point and
17698 Advanced SIMD (Neon) instructions, the settings of the options
17699 @option{-mfloat-abi} and @option{-mfpu} must also be considered:
17700 floating-point and Advanced SIMD instructions will only be used if
17701 @option{-mfloat-abi} is not set to @samp{soft}; and any setting of
17702 @option{-mfpu} other than @samp{auto} will override the available
17703 floating-point and SIMD extension instructions.
17705 For example, @samp{cortex-a9} can be found in three major
17706 configurations: integer only, with just a floating-point unit or with
17707 floating-point and Advanced SIMD. The default is to enable all the
17708 instructions, but the extensions @samp{+nosimd} and @samp{+nofp} can
17709 be used to disable just the SIMD or both the SIMD and floating-point
17710 instructions respectively.
17712 Permissible names for this option are the same as those for
17715 The following extension options are common to the listed CPUs:
17719 Disable the DSP instructions on @samp{cortex-m33}, @samp{cortex-m35p}.
17722 Disables the floating-point instructions on @samp{arm9e},
17723 @samp{arm946e-s}, @samp{arm966e-s}, @samp{arm968e-s}, @samp{arm10e},
17724 @samp{arm1020e}, @samp{arm1022e}, @samp{arm926ej-s},
17725 @samp{arm1026ej-s}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8},
17726 @samp{cortex-m4}, @samp{cortex-m7}, @samp{cortex-m33} and @samp{cortex-m35p}.
17727 Disables the floating-point and SIMD instructions on
17728 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7},
17729 @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a12},
17730 @samp{cortex-a15}, @samp{cortex-a17}, @samp{cortex-a15.cortex-a7},
17731 @samp{cortex-a17.cortex-a7}, @samp{cortex-a32}, @samp{cortex-a35},
17732 @samp{cortex-a53} and @samp{cortex-a55}.
17735 Disables the double-precision component of the floating-point instructions
17736 on @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8}, @samp{cortex-r52} and
17740 Disables the SIMD (but not floating-point) instructions on
17741 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}
17742 and @samp{cortex-a9}.
17745 Enables the cryptographic instructions on @samp{cortex-a32},
17746 @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55}, @samp{cortex-a57},
17747 @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75}, @samp{exynos-m1},
17748 @samp{xgene1}, @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
17749 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53} and
17750 @samp{cortex-a75.cortex-a55}.
17753 Additionally the @samp{generic-armv7-a} pseudo target defaults to
17754 VFPv3 with 16 double-precision registers. It supports the following
17755 extension options: @samp{mp}, @samp{sec}, @samp{vfpv3-d16},
17756 @samp{vfpv3}, @samp{vfpv3-d16-fp16}, @samp{vfpv3-fp16},
17757 @samp{vfpv4-d16}, @samp{vfpv4}, @samp{neon}, @samp{neon-vfpv3},
17758 @samp{neon-fp16}, @samp{neon-vfpv4}. The meanings are the same as for
17759 the extensions to @option{-march=armv7-a}.
17761 @option{-mcpu=generic-@var{arch}} is also permissible, and is
17762 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
17763 See @option{-mtune} for more information.
17765 @option{-mcpu=native} causes the compiler to auto-detect the CPU
17766 of the build computer. At present, this feature is only supported on
17767 GNU/Linux, and not all architectures are recognized. If the auto-detect
17768 is unsuccessful the option has no effect.
17770 @item -mfpu=@var{name}
17772 This specifies what floating-point hardware (or hardware emulation) is
17773 available on the target. Permissible names are: @samp{auto}, @samp{vfpv2},
17775 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
17776 @samp{vfpv3xd-fp16}, @samp{neon-vfpv3}, @samp{neon-fp16}, @samp{vfpv4},
17777 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
17778 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
17779 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
17780 Note that @samp{neon} is an alias for @samp{neon-vfpv3} and @samp{vfp}
17781 is an alias for @samp{vfpv2}.
17783 The setting @samp{auto} is the default and is special. It causes the
17784 compiler to select the floating-point and Advanced SIMD instructions
17785 based on the settings of @option{-mcpu} and @option{-march}.
17787 If the selected floating-point hardware includes the NEON extension
17788 (e.g.@: @option{-mfpu=neon}), note that floating-point
17789 operations are not generated by GCC's auto-vectorization pass unless
17790 @option{-funsafe-math-optimizations} is also specified. This is
17791 because NEON hardware does not fully implement the IEEE 754 standard for
17792 floating-point arithmetic (in particular denormal values are treated as
17793 zero), so the use of NEON instructions may lead to a loss of precision.
17795 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}).
17797 @item -mfp16-format=@var{name}
17798 @opindex mfp16-format
17799 Specify the format of the @code{__fp16} half-precision floating-point type.
17800 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
17801 the default is @samp{none}, in which case the @code{__fp16} type is not
17802 defined. @xref{Half-Precision}, for more information.
17804 @item -mstructure-size-boundary=@var{n}
17805 @opindex mstructure-size-boundary
17806 The sizes of all structures and unions are rounded up to a multiple
17807 of the number of bits set by this option. Permissible values are 8, 32
17808 and 64. The default value varies for different toolchains. For the COFF
17809 targeted toolchain the default value is 8. A value of 64 is only allowed
17810 if the underlying ABI supports it.
17812 Specifying a larger number can produce faster, more efficient code, but
17813 can also increase the size of the program. Different values are potentially
17814 incompatible. Code compiled with one value cannot necessarily expect to
17815 work with code or libraries compiled with another value, if they exchange
17816 information using structures or unions.
17818 This option is deprecated.
17820 @item -mabort-on-noreturn
17821 @opindex mabort-on-noreturn
17822 Generate a call to the function @code{abort} at the end of a
17823 @code{noreturn} function. It is executed if the function tries to
17827 @itemx -mno-long-calls
17828 @opindex mlong-calls
17829 @opindex mno-long-calls
17830 Tells the compiler to perform function calls by first loading the
17831 address of the function into a register and then performing a subroutine
17832 call on this register. This switch is needed if the target function
17833 lies outside of the 64-megabyte addressing range of the offset-based
17834 version of subroutine call instruction.
17836 Even if this switch is enabled, not all function calls are turned
17837 into long calls. The heuristic is that static functions, functions
17838 that have the @code{short_call} attribute, functions that are inside
17839 the scope of a @code{#pragma no_long_calls} directive, and functions whose
17840 definitions have already been compiled within the current compilation
17841 unit are not turned into long calls. The exceptions to this rule are
17842 that weak function definitions, functions with the @code{long_call}
17843 attribute or the @code{section} attribute, and functions that are within
17844 the scope of a @code{#pragma long_calls} directive are always
17845 turned into long calls.
17847 This feature is not enabled by default. Specifying
17848 @option{-mno-long-calls} restores the default behavior, as does
17849 placing the function calls within the scope of a @code{#pragma
17850 long_calls_off} directive. Note these switches have no effect on how
17851 the compiler generates code to handle function calls via function
17854 @item -msingle-pic-base
17855 @opindex msingle-pic-base
17856 Treat the register used for PIC addressing as read-only, rather than
17857 loading it in the prologue for each function. The runtime system is
17858 responsible for initializing this register with an appropriate value
17859 before execution begins.
17861 @item -mpic-register=@var{reg}
17862 @opindex mpic-register
17863 Specify the register to be used for PIC addressing.
17864 For standard PIC base case, the default is any suitable register
17865 determined by compiler. For single PIC base case, the default is
17866 @samp{R9} if target is EABI based or stack-checking is enabled,
17867 otherwise the default is @samp{R10}.
17869 @item -mpic-data-is-text-relative
17870 @opindex mpic-data-is-text-relative
17871 Assume that the displacement between the text and data segments is fixed
17872 at static link time. This permits using PC-relative addressing
17873 operations to access data known to be in the data segment. For
17874 non-VxWorks RTP targets, this option is enabled by default. When
17875 disabled on such targets, it will enable @option{-msingle-pic-base} by
17878 @item -mpoke-function-name
17879 @opindex mpoke-function-name
17880 Write the name of each function into the text section, directly
17881 preceding the function prologue. The generated code is similar to this:
17885 .ascii "arm_poke_function_name", 0
17888 .word 0xff000000 + (t1 - t0)
17889 arm_poke_function_name
17891 stmfd sp!, @{fp, ip, lr, pc@}
17895 When performing a stack backtrace, code can inspect the value of
17896 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
17897 location @code{pc - 12} and the top 8 bits are set, then we know that
17898 there is a function name embedded immediately preceding this location
17899 and has length @code{((pc[-3]) & 0xff000000)}.
17906 Select between generating code that executes in ARM and Thumb
17907 states. The default for most configurations is to generate code
17908 that executes in ARM state, but the default can be changed by
17909 configuring GCC with the @option{--with-mode=}@var{state}
17912 You can also override the ARM and Thumb mode for each function
17913 by using the @code{target("thumb")} and @code{target("arm")} function attributes
17914 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
17917 @opindex mflip-thumb
17918 Switch ARM/Thumb modes on alternating functions.
17919 This option is provided for regression testing of mixed Thumb/ARM code
17920 generation, and is not intended for ordinary use in compiling code.
17923 @opindex mtpcs-frame
17924 Generate a stack frame that is compliant with the Thumb Procedure Call
17925 Standard for all non-leaf functions. (A leaf function is one that does
17926 not call any other functions.) The default is @option{-mno-tpcs-frame}.
17928 @item -mtpcs-leaf-frame
17929 @opindex mtpcs-leaf-frame
17930 Generate a stack frame that is compliant with the Thumb Procedure Call
17931 Standard for all leaf functions. (A leaf function is one that does
17932 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
17934 @item -mcallee-super-interworking
17935 @opindex mcallee-super-interworking
17936 Gives all externally visible functions in the file being compiled an ARM
17937 instruction set header which switches to Thumb mode before executing the
17938 rest of the function. This allows these functions to be called from
17939 non-interworking code. This option is not valid in AAPCS configurations
17940 because interworking is enabled by default.
17942 @item -mcaller-super-interworking
17943 @opindex mcaller-super-interworking
17944 Allows calls via function pointers (including virtual functions) to
17945 execute correctly regardless of whether the target code has been
17946 compiled for interworking or not. There is a small overhead in the cost
17947 of executing a function pointer if this option is enabled. This option
17948 is not valid in AAPCS configurations because interworking is enabled
17951 @item -mtp=@var{name}
17953 Specify the access model for the thread local storage pointer. The valid
17954 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
17955 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
17956 (supported in the arm6k architecture), and @samp{auto}, which uses the
17957 best available method for the selected processor. The default setting is
17960 @item -mtls-dialect=@var{dialect}
17961 @opindex mtls-dialect
17962 Specify the dialect to use for accessing thread local storage. Two
17963 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
17964 @samp{gnu} dialect selects the original GNU scheme for supporting
17965 local and global dynamic TLS models. The @samp{gnu2} dialect
17966 selects the GNU descriptor scheme, which provides better performance
17967 for shared libraries. The GNU descriptor scheme is compatible with
17968 the original scheme, but does require new assembler, linker and
17969 library support. Initial and local exec TLS models are unaffected by
17970 this option and always use the original scheme.
17972 @item -mword-relocations
17973 @opindex mword-relocations
17974 Only generate absolute relocations on word-sized values (i.e.@: R_ARM_ABS32).
17975 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
17976 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
17977 is specified. This option conflicts with @option{-mslow-flash-data}.
17979 @item -mfix-cortex-m3-ldrd
17980 @opindex mfix-cortex-m3-ldrd
17981 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
17982 with overlapping destination and base registers are used. This option avoids
17983 generating these instructions. This option is enabled by default when
17984 @option{-mcpu=cortex-m3} is specified.
17986 @item -munaligned-access
17987 @itemx -mno-unaligned-access
17988 @opindex munaligned-access
17989 @opindex mno-unaligned-access
17990 Enables (or disables) reading and writing of 16- and 32- bit values
17991 from addresses that are not 16- or 32- bit aligned. By default
17992 unaligned access is disabled for all pre-ARMv6, all ARMv6-M and for
17993 ARMv8-M Baseline architectures, and enabled for all other
17994 architectures. If unaligned access is not enabled then words in packed
17995 data structures are accessed a byte at a time.
17997 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
17998 generated object file to either true or false, depending upon the
17999 setting of this option. If unaligned access is enabled then the
18000 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
18003 @item -mneon-for-64bits
18004 @opindex mneon-for-64bits
18005 This option is deprecated and has no effect.
18007 @item -mslow-flash-data
18008 @opindex mslow-flash-data
18009 Assume loading data from flash is slower than fetching instruction.
18010 Therefore literal load is minimized for better performance.
18011 This option is only supported when compiling for ARMv7 M-profile and
18012 off by default. It conflicts with @option{-mword-relocations}.
18014 @item -masm-syntax-unified
18015 @opindex masm-syntax-unified
18016 Assume inline assembler is using unified asm syntax. The default is
18017 currently off which implies divided syntax. This option has no impact
18018 on Thumb2. However, this may change in future releases of GCC.
18019 Divided syntax should be considered deprecated.
18021 @item -mrestrict-it
18022 @opindex mrestrict-it
18023 Restricts generation of IT blocks to conform to the rules of ARMv8-A.
18024 IT blocks can only contain a single 16-bit instruction from a select
18025 set of instructions. This option is on by default for ARMv8-A Thumb mode.
18027 @item -mprint-tune-info
18028 @opindex mprint-tune-info
18029 Print CPU tuning information as comment in assembler file. This is
18030 an option used only for regression testing of the compiler and not
18031 intended for ordinary use in compiling code. This option is disabled
18034 @item -mverbose-cost-dump
18035 @opindex mverbose-cost-dump
18036 Enable verbose cost model dumping in the debug dump files. This option is
18037 provided for use in debugging the compiler.
18040 @opindex mpure-code
18041 Do not allow constant data to be placed in code sections.
18042 Additionally, when compiling for ELF object format give all text sections the
18043 ELF processor-specific section attribute @code{SHF_ARM_PURECODE}. This option
18044 is only available when generating non-pic code for M-profile targets with the
18049 Generate secure code as per the "ARMv8-M Security Extensions: Requirements on
18050 Development Tools Engineering Specification", which can be found on
18051 @url{http://infocenter.arm.com/help/topic/com.arm.doc.ecm0359818/ECM0359818_armv8m_security_extensions_reqs_on_dev_tools_1_0.pdf}.
18055 @subsection AVR Options
18056 @cindex AVR Options
18058 These options are defined for AVR implementations:
18061 @item -mmcu=@var{mcu}
18063 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
18065 The default for this option is@tie{}@samp{avr2}.
18067 GCC supports the following AVR devices and ISAs:
18069 @include avr-mmcu.texi
18074 Assume that all data in static storage can be accessed by LDS / STS
18075 instructions. This option has only an effect on reduced Tiny devices like
18076 ATtiny40. See also the @code{absdata}
18077 @ref{AVR Variable Attributes,variable attribute}.
18079 @item -maccumulate-args
18080 @opindex maccumulate-args
18081 Accumulate outgoing function arguments and acquire/release the needed
18082 stack space for outgoing function arguments once in function
18083 prologue/epilogue. Without this option, outgoing arguments are pushed
18084 before calling a function and popped afterwards.
18086 Popping the arguments after the function call can be expensive on
18087 AVR so that accumulating the stack space might lead to smaller
18088 executables because arguments need not be removed from the
18089 stack after such a function call.
18091 This option can lead to reduced code size for functions that perform
18092 several calls to functions that get their arguments on the stack like
18093 calls to printf-like functions.
18095 @item -mbranch-cost=@var{cost}
18096 @opindex mbranch-cost
18097 Set the branch costs for conditional branch instructions to
18098 @var{cost}. Reasonable values for @var{cost} are small, non-negative
18099 integers. The default branch cost is 0.
18101 @item -mcall-prologues
18102 @opindex mcall-prologues
18103 Functions prologues/epilogues are expanded as calls to appropriate
18104 subroutines. Code size is smaller.
18106 @item -mgas-isr-prologues
18107 @opindex mgas-isr-prologues
18108 Interrupt service routines (ISRs) may use the @code{__gcc_isr} pseudo
18109 instruction supported by GNU Binutils.
18110 If this option is on, the feature can still be disabled for individual
18111 ISRs by means of the @ref{AVR Function Attributes,,@code{no_gccisr}}
18112 function attribute. This feature is activated per default
18113 if optimization is on (but not with @option{-Og}, @pxref{Optimize Options}),
18114 and if GNU Binutils support @w{@uref{https://sourceware.org/PR21683,PR21683}}.
18118 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
18119 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
18120 and @code{long long} is 4 bytes. Please note that this option does not
18121 conform to the C standards, but it results in smaller code
18124 @item -mmain-is-OS_task
18125 @opindex mmain-is-OS_task
18126 Do not save registers in @code{main}. The effect is the same like
18127 attaching attribute @ref{AVR Function Attributes,,@code{OS_task}}
18128 to @code{main}. It is activated per default if optimization is on.
18130 @item -mn-flash=@var{num}
18132 Assume that the flash memory has a size of
18133 @var{num} times 64@tie{}KiB.
18135 @item -mno-interrupts
18136 @opindex mno-interrupts
18137 Generated code is not compatible with hardware interrupts.
18138 Code size is smaller.
18142 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
18143 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
18144 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
18145 the assembler's command line and the @option{--relax} option to the
18146 linker's command line.
18148 Jump relaxing is performed by the linker because jump offsets are not
18149 known before code is located. Therefore, the assembler code generated by the
18150 compiler is the same, but the instructions in the executable may
18151 differ from instructions in the assembler code.
18153 Relaxing must be turned on if linker stubs are needed, see the
18154 section on @code{EIND} and linker stubs below.
18158 Assume that the device supports the Read-Modify-Write
18159 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
18161 @item -mshort-calls
18162 @opindex mshort-calls
18164 Assume that @code{RJMP} and @code{RCALL} can target the whole
18167 This option is used internally for multilib selection. It is
18168 not an optimization option, and you don't need to set it by hand.
18172 Treat the stack pointer register as an 8-bit register,
18173 i.e.@: assume the high byte of the stack pointer is zero.
18174 In general, you don't need to set this option by hand.
18176 This option is used internally by the compiler to select and
18177 build multilibs for architectures @code{avr2} and @code{avr25}.
18178 These architectures mix devices with and without @code{SPH}.
18179 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
18180 the compiler driver adds or removes this option from the compiler
18181 proper's command line, because the compiler then knows if the device
18182 or architecture has an 8-bit stack pointer and thus no @code{SPH}
18187 Use address register @code{X} in a way proposed by the hardware. This means
18188 that @code{X} is only used in indirect, post-increment or
18189 pre-decrement addressing.
18191 Without this option, the @code{X} register may be used in the same way
18192 as @code{Y} or @code{Z} which then is emulated by additional
18194 For example, loading a value with @code{X+const} addressing with a
18195 small non-negative @code{const < 64} to a register @var{Rn} is
18199 adiw r26, const ; X += const
18200 ld @var{Rn}, X ; @var{Rn} = *X
18201 sbiw r26, const ; X -= const
18205 @opindex mtiny-stack
18206 Only change the lower 8@tie{}bits of the stack pointer.
18208 @item -mfract-convert-truncate
18209 @opindex mfract-convert-truncate
18210 Allow to use truncation instead of rounding towards zero for fractional fixed-point types.
18213 @opindex nodevicelib
18214 Don't link against AVR-LibC's device specific library @code{lib<mcu>.a}.
18216 @item -Waddr-space-convert
18217 @opindex Waddr-space-convert
18218 @opindex Wno-addr-space-convert
18219 Warn about conversions between address spaces in the case where the
18220 resulting address space is not contained in the incoming address space.
18222 @item -Wmisspelled-isr
18223 @opindex Wmisspelled-isr
18224 @opindex Wno-misspelled-isr
18225 Warn if the ISR is misspelled, i.e.@: without __vector prefix.
18226 Enabled by default.
18229 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
18230 @cindex @code{EIND}
18231 Pointers in the implementation are 16@tie{}bits wide.
18232 The address of a function or label is represented as word address so
18233 that indirect jumps and calls can target any code address in the
18234 range of 64@tie{}Ki words.
18236 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
18237 bytes of program memory space, there is a special function register called
18238 @code{EIND} that serves as most significant part of the target address
18239 when @code{EICALL} or @code{EIJMP} instructions are used.
18241 Indirect jumps and calls on these devices are handled as follows by
18242 the compiler and are subject to some limitations:
18247 The compiler never sets @code{EIND}.
18250 The compiler uses @code{EIND} implicitly in @code{EICALL}/@code{EIJMP}
18251 instructions or might read @code{EIND} directly in order to emulate an
18252 indirect call/jump by means of a @code{RET} instruction.
18255 The compiler assumes that @code{EIND} never changes during the startup
18256 code or during the application. In particular, @code{EIND} is not
18257 saved/restored in function or interrupt service routine
18261 For indirect calls to functions and computed goto, the linker
18262 generates @emph{stubs}. Stubs are jump pads sometimes also called
18263 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
18264 The stub contains a direct jump to the desired address.
18267 Linker relaxation must be turned on so that the linker generates
18268 the stubs correctly in all situations. See the compiler option
18269 @option{-mrelax} and the linker option @option{--relax}.
18270 There are corner cases where the linker is supposed to generate stubs
18271 but aborts without relaxation and without a helpful error message.
18274 The default linker script is arranged for code with @code{EIND = 0}.
18275 If code is supposed to work for a setup with @code{EIND != 0}, a custom
18276 linker script has to be used in order to place the sections whose
18277 name start with @code{.trampolines} into the segment where @code{EIND}
18281 The startup code from libgcc never sets @code{EIND}.
18282 Notice that startup code is a blend of code from libgcc and AVR-LibC.
18283 For the impact of AVR-LibC on @code{EIND}, see the
18284 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
18287 It is legitimate for user-specific startup code to set up @code{EIND}
18288 early, for example by means of initialization code located in
18289 section @code{.init3}. Such code runs prior to general startup code
18290 that initializes RAM and calls constructors, but after the bit
18291 of startup code from AVR-LibC that sets @code{EIND} to the segment
18292 where the vector table is located.
18294 #include <avr/io.h>
18297 __attribute__((section(".init3"),naked,used,no_instrument_function))
18298 init3_set_eind (void)
18300 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
18301 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
18306 The @code{__trampolines_start} symbol is defined in the linker script.
18309 Stubs are generated automatically by the linker if
18310 the following two conditions are met:
18313 @item The address of a label is taken by means of the @code{gs} modifier
18314 (short for @emph{generate stubs}) like so:
18316 LDI r24, lo8(gs(@var{func}))
18317 LDI r25, hi8(gs(@var{func}))
18319 @item The final location of that label is in a code segment
18320 @emph{outside} the segment where the stubs are located.
18324 The compiler emits such @code{gs} modifiers for code labels in the
18325 following situations:
18327 @item Taking address of a function or code label.
18328 @item Computed goto.
18329 @item If prologue-save function is used, see @option{-mcall-prologues}
18330 command-line option.
18331 @item Switch/case dispatch tables. If you do not want such dispatch
18332 tables you can specify the @option{-fno-jump-tables} command-line option.
18333 @item C and C++ constructors/destructors called during startup/shutdown.
18334 @item If the tools hit a @code{gs()} modifier explained above.
18338 Jumping to non-symbolic addresses like so is @emph{not} supported:
18343 /* Call function at word address 0x2 */
18344 return ((int(*)(void)) 0x2)();
18348 Instead, a stub has to be set up, i.e.@: the function has to be called
18349 through a symbol (@code{func_4} in the example):
18354 extern int func_4 (void);
18356 /* Call function at byte address 0x4 */
18361 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
18362 Alternatively, @code{func_4} can be defined in the linker script.
18365 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
18366 @cindex @code{RAMPD}
18367 @cindex @code{RAMPX}
18368 @cindex @code{RAMPY}
18369 @cindex @code{RAMPZ}
18370 Some AVR devices support memories larger than the 64@tie{}KiB range
18371 that can be accessed with 16-bit pointers. To access memory locations
18372 outside this 64@tie{}KiB range, the content of a @code{RAMP}
18373 register is used as high part of the address:
18374 The @code{X}, @code{Y}, @code{Z} address register is concatenated
18375 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
18376 register, respectively, to get a wide address. Similarly,
18377 @code{RAMPD} is used together with direct addressing.
18381 The startup code initializes the @code{RAMP} special function
18382 registers with zero.
18385 If a @ref{AVR Named Address Spaces,named address space} other than
18386 generic or @code{__flash} is used, then @code{RAMPZ} is set
18387 as needed before the operation.
18390 If the device supports RAM larger than 64@tie{}KiB and the compiler
18391 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
18392 is reset to zero after the operation.
18395 If the device comes with a specific @code{RAMP} register, the ISR
18396 prologue/epilogue saves/restores that SFR and initializes it with
18397 zero in case the ISR code might (implicitly) use it.
18400 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
18401 If you use inline assembler to read from locations outside the
18402 16-bit address range and change one of the @code{RAMP} registers,
18403 you must reset it to zero after the access.
18407 @subsubsection AVR Built-in Macros
18409 GCC defines several built-in macros so that the user code can test
18410 for the presence or absence of features. Almost any of the following
18411 built-in macros are deduced from device capabilities and thus
18412 triggered by the @option{-mmcu=} command-line option.
18414 For even more AVR-specific built-in macros see
18415 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
18420 Build-in macro that resolves to a decimal number that identifies the
18421 architecture and depends on the @option{-mmcu=@var{mcu}} option.
18422 Possible values are:
18424 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
18425 @code{4}, @code{5}, @code{51}, @code{6}
18427 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
18428 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
18433 @code{102}, @code{103}, @code{104},
18434 @code{105}, @code{106}, @code{107}
18436 for @var{mcu}=@code{avrtiny},
18437 @code{avrxmega2}, @code{avrxmega3}, @code{avrxmega4},
18438 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
18439 If @var{mcu} specifies a device, this built-in macro is set
18440 accordingly. For example, with @option{-mmcu=atmega8} the macro is
18441 defined to @code{4}.
18443 @item __AVR_@var{Device}__
18444 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
18445 the device's name. For example, @option{-mmcu=atmega8} defines the
18446 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
18447 @code{__AVR_ATtiny261A__}, etc.
18449 The built-in macros' names follow
18450 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
18451 the device name as from the AVR user manual. The difference between
18452 @var{Device} in the built-in macro and @var{device} in
18453 @option{-mmcu=@var{device}} is that the latter is always lowercase.
18455 If @var{device} is not a device but only a core architecture like
18456 @samp{avr51}, this macro is not defined.
18458 @item __AVR_DEVICE_NAME__
18459 Setting @option{-mmcu=@var{device}} defines this built-in macro to
18460 the device's name. For example, with @option{-mmcu=atmega8} the macro
18461 is defined to @code{atmega8}.
18463 If @var{device} is not a device but only a core architecture like
18464 @samp{avr51}, this macro is not defined.
18466 @item __AVR_XMEGA__
18467 The device / architecture belongs to the XMEGA family of devices.
18469 @item __AVR_HAVE_ELPM__
18470 The device has the @code{ELPM} instruction.
18472 @item __AVR_HAVE_ELPMX__
18473 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
18474 R@var{n},Z+} instructions.
18476 @item __AVR_HAVE_MOVW__
18477 The device has the @code{MOVW} instruction to perform 16-bit
18478 register-register moves.
18480 @item __AVR_HAVE_LPMX__
18481 The device has the @code{LPM R@var{n},Z} and
18482 @code{LPM R@var{n},Z+} instructions.
18484 @item __AVR_HAVE_MUL__
18485 The device has a hardware multiplier.
18487 @item __AVR_HAVE_JMP_CALL__
18488 The device has the @code{JMP} and @code{CALL} instructions.
18489 This is the case for devices with more than 8@tie{}KiB of program
18492 @item __AVR_HAVE_EIJMP_EICALL__
18493 @itemx __AVR_3_BYTE_PC__
18494 The device has the @code{EIJMP} and @code{EICALL} instructions.
18495 This is the case for devices with more than 128@tie{}KiB of program memory.
18496 This also means that the program counter
18497 (PC) is 3@tie{}bytes wide.
18499 @item __AVR_2_BYTE_PC__
18500 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
18501 with up to 128@tie{}KiB of program memory.
18503 @item __AVR_HAVE_8BIT_SP__
18504 @itemx __AVR_HAVE_16BIT_SP__
18505 The stack pointer (SP) register is treated as 8-bit respectively
18506 16-bit register by the compiler.
18507 The definition of these macros is affected by @option{-mtiny-stack}.
18509 @item __AVR_HAVE_SPH__
18511 The device has the SPH (high part of stack pointer) special function
18512 register or has an 8-bit stack pointer, respectively.
18513 The definition of these macros is affected by @option{-mmcu=} and
18514 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
18517 @item __AVR_HAVE_RAMPD__
18518 @itemx __AVR_HAVE_RAMPX__
18519 @itemx __AVR_HAVE_RAMPY__
18520 @itemx __AVR_HAVE_RAMPZ__
18521 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
18522 @code{RAMPZ} special function register, respectively.
18524 @item __NO_INTERRUPTS__
18525 This macro reflects the @option{-mno-interrupts} command-line option.
18527 @item __AVR_ERRATA_SKIP__
18528 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
18529 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
18530 instructions because of a hardware erratum. Skip instructions are
18531 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
18532 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
18535 @item __AVR_ISA_RMW__
18536 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
18538 @item __AVR_SFR_OFFSET__=@var{offset}
18539 Instructions that can address I/O special function registers directly
18540 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
18541 address as if addressed by an instruction to access RAM like @code{LD}
18542 or @code{STS}. This offset depends on the device architecture and has
18543 to be subtracted from the RAM address in order to get the
18544 respective I/O@tie{}address.
18546 @item __AVR_SHORT_CALLS__
18547 The @option{-mshort-calls} command line option is set.
18549 @item __AVR_PM_BASE_ADDRESS__=@var{addr}
18550 Some devices support reading from flash memory by means of @code{LD*}
18551 instructions. The flash memory is seen in the data address space
18552 at an offset of @code{__AVR_PM_BASE_ADDRESS__}. If this macro
18553 is not defined, this feature is not available. If defined,
18554 the address space is linear and there is no need to put
18555 @code{.rodata} into RAM. This is handled by the default linker
18556 description file, and is currently available for
18557 @code{avrtiny} and @code{avrxmega3}. Even more convenient,
18558 there is no need to use address spaces like @code{__flash} or
18559 features like attribute @code{progmem} and @code{pgm_read_*}.
18561 @item __WITH_AVRLIBC__
18562 The compiler is configured to be used together with AVR-Libc.
18563 See the @option{--with-avrlibc} configure option.
18567 @node Blackfin Options
18568 @subsection Blackfin Options
18569 @cindex Blackfin Options
18572 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
18574 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
18575 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
18576 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
18577 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
18578 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
18579 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
18580 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
18581 @samp{bf561}, @samp{bf592}.
18583 The optional @var{sirevision} specifies the silicon revision of the target
18584 Blackfin processor. Any workarounds available for the targeted silicon revision
18585 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
18586 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
18587 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
18588 hexadecimal digits representing the major and minor numbers in the silicon
18589 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
18590 is not defined. If @var{sirevision} is @samp{any}, the
18591 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
18592 If this optional @var{sirevision} is not used, GCC assumes the latest known
18593 silicon revision of the targeted Blackfin processor.
18595 GCC defines a preprocessor macro for the specified @var{cpu}.
18596 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
18597 provided by libgloss to be linked in if @option{-msim} is not given.
18599 Without this option, @samp{bf532} is used as the processor by default.
18601 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
18602 only the preprocessor macro is defined.
18606 Specifies that the program will be run on the simulator. This causes
18607 the simulator BSP provided by libgloss to be linked in. This option
18608 has effect only for @samp{bfin-elf} toolchain.
18609 Certain other options, such as @option{-mid-shared-library} and
18610 @option{-mfdpic}, imply @option{-msim}.
18612 @item -momit-leaf-frame-pointer
18613 @opindex momit-leaf-frame-pointer
18614 Don't keep the frame pointer in a register for leaf functions. This
18615 avoids the instructions to save, set up and restore frame pointers and
18616 makes an extra register available in leaf functions.
18618 @item -mspecld-anomaly
18619 @opindex mspecld-anomaly
18620 When enabled, the compiler ensures that the generated code does not
18621 contain speculative loads after jump instructions. If this option is used,
18622 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
18624 @item -mno-specld-anomaly
18625 @opindex mno-specld-anomaly
18626 @opindex mspecld-anomaly
18627 Don't generate extra code to prevent speculative loads from occurring.
18629 @item -mcsync-anomaly
18630 @opindex mcsync-anomaly
18631 When enabled, the compiler ensures that the generated code does not
18632 contain CSYNC or SSYNC instructions too soon after conditional branches.
18633 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
18635 @item -mno-csync-anomaly
18636 @opindex mno-csync-anomaly
18637 @opindex mcsync-anomaly
18638 Don't generate extra code to prevent CSYNC or SSYNC instructions from
18639 occurring too soon after a conditional branch.
18643 When enabled, the compiler is free to take advantage of the knowledge that
18644 the entire program fits into the low 64k of memory.
18647 @opindex mno-low64k
18648 Assume that the program is arbitrarily large. This is the default.
18650 @item -mstack-check-l1
18651 @opindex mstack-check-l1
18652 Do stack checking using information placed into L1 scratchpad memory by the
18655 @item -mid-shared-library
18656 @opindex mid-shared-library
18657 Generate code that supports shared libraries via the library ID method.
18658 This allows for execute in place and shared libraries in an environment
18659 without virtual memory management. This option implies @option{-fPIC}.
18660 With a @samp{bfin-elf} target, this option implies @option{-msim}.
18662 @item -mno-id-shared-library
18663 @opindex mno-id-shared-library
18664 @opindex mid-shared-library
18665 Generate code that doesn't assume ID-based shared libraries are being used.
18666 This is the default.
18668 @item -mleaf-id-shared-library
18669 @opindex mleaf-id-shared-library
18670 Generate code that supports shared libraries via the library ID method,
18671 but assumes that this library or executable won't link against any other
18672 ID shared libraries. That allows the compiler to use faster code for jumps
18675 @item -mno-leaf-id-shared-library
18676 @opindex mno-leaf-id-shared-library
18677 @opindex mleaf-id-shared-library
18678 Do not assume that the code being compiled won't link against any ID shared
18679 libraries. Slower code is generated for jump and call insns.
18681 @item -mshared-library-id=n
18682 @opindex mshared-library-id
18683 Specifies the identification number of the ID-based shared library being
18684 compiled. Specifying a value of 0 generates more compact code; specifying
18685 other values forces the allocation of that number to the current
18686 library but is no more space- or time-efficient than omitting this option.
18690 Generate code that allows the data segment to be located in a different
18691 area of memory from the text segment. This allows for execute in place in
18692 an environment without virtual memory management by eliminating relocations
18693 against the text section.
18695 @item -mno-sep-data
18696 @opindex mno-sep-data
18698 Generate code that assumes that the data segment follows the text segment.
18699 This is the default.
18702 @itemx -mno-long-calls
18703 @opindex mlong-calls
18704 @opindex mno-long-calls
18705 Tells the compiler to perform function calls by first loading the
18706 address of the function into a register and then performing a subroutine
18707 call on this register. This switch is needed if the target function
18708 lies outside of the 24-bit addressing range of the offset-based
18709 version of subroutine call instruction.
18711 This feature is not enabled by default. Specifying
18712 @option{-mno-long-calls} restores the default behavior. Note these
18713 switches have no effect on how the compiler generates code to handle
18714 function calls via function pointers.
18718 Link with the fast floating-point library. This library relaxes some of
18719 the IEEE floating-point standard's rules for checking inputs against
18720 Not-a-Number (NAN), in the interest of performance.
18723 @opindex minline-plt
18724 Enable inlining of PLT entries in function calls to functions that are
18725 not known to bind locally. It has no effect without @option{-mfdpic}.
18728 @opindex mmulticore
18729 Build a standalone application for multicore Blackfin processors.
18730 This option causes proper start files and link scripts supporting
18731 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
18732 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
18734 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
18735 selects the one-application-per-core programming model. Without
18736 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
18737 programming model is used. In this model, the main function of Core B
18738 should be named as @code{coreb_main}.
18740 If this option is not used, the single-core application programming
18745 Build a standalone application for Core A of BF561 when using
18746 the one-application-per-core programming model. Proper start files
18747 and link scripts are used to support Core A, and the macro
18748 @code{__BFIN_COREA} is defined.
18749 This option can only be used in conjunction with @option{-mmulticore}.
18753 Build a standalone application for Core B of BF561 when using
18754 the one-application-per-core programming model. Proper start files
18755 and link scripts are used to support Core B, and the macro
18756 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
18757 should be used instead of @code{main}.
18758 This option can only be used in conjunction with @option{-mmulticore}.
18762 Build a standalone application for SDRAM. Proper start files and
18763 link scripts are used to put the application into SDRAM, and the macro
18764 @code{__BFIN_SDRAM} is defined.
18765 The loader should initialize SDRAM before loading the application.
18769 Assume that ICPLBs are enabled at run time. This has an effect on certain
18770 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
18771 are enabled; for standalone applications the default is off.
18775 @subsection C6X Options
18776 @cindex C6X Options
18779 @item -march=@var{name}
18781 This specifies the name of the target architecture. GCC uses this
18782 name to determine what kind of instructions it can emit when generating
18783 assembly code. Permissible names are: @samp{c62x},
18784 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
18787 @opindex mbig-endian
18788 Generate code for a big-endian target.
18790 @item -mlittle-endian
18791 @opindex mlittle-endian
18792 Generate code for a little-endian target. This is the default.
18796 Choose startup files and linker script suitable for the simulator.
18798 @item -msdata=default
18799 @opindex msdata=default
18800 Put small global and static data in the @code{.neardata} section,
18801 which is pointed to by register @code{B14}. Put small uninitialized
18802 global and static data in the @code{.bss} section, which is adjacent
18803 to the @code{.neardata} section. Put small read-only data into the
18804 @code{.rodata} section. The corresponding sections used for large
18805 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
18808 @opindex msdata=all
18809 Put all data, not just small objects, into the sections reserved for
18810 small data, and use addressing relative to the @code{B14} register to
18814 @opindex msdata=none
18815 Make no use of the sections reserved for small data, and use absolute
18816 addresses to access all data. Put all initialized global and static
18817 data in the @code{.fardata} section, and all uninitialized data in the
18818 @code{.far} section. Put all constant data into the @code{.const}
18823 @subsection CRIS Options
18824 @cindex CRIS Options
18826 These options are defined specifically for the CRIS ports.
18829 @item -march=@var{architecture-type}
18830 @itemx -mcpu=@var{architecture-type}
18833 Generate code for the specified architecture. The choices for
18834 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
18835 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
18836 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
18839 @item -mtune=@var{architecture-type}
18841 Tune to @var{architecture-type} everything applicable about the generated
18842 code, except for the ABI and the set of available instructions. The
18843 choices for @var{architecture-type} are the same as for
18844 @option{-march=@var{architecture-type}}.
18846 @item -mmax-stack-frame=@var{n}
18847 @opindex mmax-stack-frame
18848 Warn when the stack frame of a function exceeds @var{n} bytes.
18854 The options @option{-metrax4} and @option{-metrax100} are synonyms for
18855 @option{-march=v3} and @option{-march=v8} respectively.
18857 @item -mmul-bug-workaround
18858 @itemx -mno-mul-bug-workaround
18859 @opindex mmul-bug-workaround
18860 @opindex mno-mul-bug-workaround
18861 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
18862 models where it applies. This option is active by default.
18866 Enable CRIS-specific verbose debug-related information in the assembly
18867 code. This option also has the effect of turning off the @samp{#NO_APP}
18868 formatted-code indicator to the assembler at the beginning of the
18873 Do not use condition-code results from previous instruction; always emit
18874 compare and test instructions before use of condition codes.
18876 @item -mno-side-effects
18877 @opindex mno-side-effects
18878 @opindex mside-effects
18879 Do not emit instructions with side effects in addressing modes other than
18882 @item -mstack-align
18883 @itemx -mno-stack-align
18884 @itemx -mdata-align
18885 @itemx -mno-data-align
18886 @itemx -mconst-align
18887 @itemx -mno-const-align
18888 @opindex mstack-align
18889 @opindex mno-stack-align
18890 @opindex mdata-align
18891 @opindex mno-data-align
18892 @opindex mconst-align
18893 @opindex mno-const-align
18894 These options (@samp{no-} options) arrange (eliminate arrangements) for the
18895 stack frame, individual data and constants to be aligned for the maximum
18896 single data access size for the chosen CPU model. The default is to
18897 arrange for 32-bit alignment. ABI details such as structure layout are
18898 not affected by these options.
18906 Similar to the stack- data- and const-align options above, these options
18907 arrange for stack frame, writable data and constants to all be 32-bit,
18908 16-bit or 8-bit aligned. The default is 32-bit alignment.
18910 @item -mno-prologue-epilogue
18911 @itemx -mprologue-epilogue
18912 @opindex mno-prologue-epilogue
18913 @opindex mprologue-epilogue
18914 With @option{-mno-prologue-epilogue}, the normal function prologue and
18915 epilogue which set up the stack frame are omitted and no return
18916 instructions or return sequences are generated in the code. Use this
18917 option only together with visual inspection of the compiled code: no
18918 warnings or errors are generated when call-saved registers must be saved,
18919 or storage for local variables needs to be allocated.
18923 @opindex mno-gotplt
18925 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
18926 instruction sequences that load addresses for functions from the PLT part
18927 of the GOT rather than (traditional on other architectures) calls to the
18928 PLT@. The default is @option{-mgotplt}.
18932 Legacy no-op option only recognized with the cris-axis-elf and
18933 cris-axis-linux-gnu targets.
18937 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
18941 This option, recognized for the cris-axis-elf, arranges
18942 to link with input-output functions from a simulator library. Code,
18943 initialized data and zero-initialized data are allocated consecutively.
18947 Like @option{-sim}, but pass linker options to locate initialized data at
18948 0x40000000 and zero-initialized data at 0x80000000.
18952 @subsection CR16 Options
18953 @cindex CR16 Options
18955 These options are defined specifically for the CR16 ports.
18961 Enable the use of multiply-accumulate instructions. Disabled by default.
18965 @opindex mcr16cplus
18967 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
18972 Links the library libsim.a which is in compatible with simulator. Applicable
18973 to ELF compiler only.
18977 Choose integer type as 32-bit wide.
18981 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
18983 @item -mdata-model=@var{model}
18984 @opindex mdata-model
18985 Choose a data model. The choices for @var{model} are @samp{near},
18986 @samp{far} or @samp{medium}. @samp{medium} is default.
18987 However, @samp{far} is not valid with @option{-mcr16c}, as the
18988 CR16C architecture does not support the far data model.
18991 @node C-SKY Options
18992 @subsection C-SKY Options
18993 @cindex C-SKY Options
18995 GCC supports these options when compiling for C-SKY V2 processors.
18999 @item -march=@var{arch}
19001 Specify the C-SKY target architecture. Valid values for @var{arch} are:
19002 @samp{ck801}, @samp{ck802}, @samp{ck803}, @samp{ck807}, and @samp{ck810}.
19003 The default is @samp{ck810}.
19005 @item -mcpu=@var{cpu}
19007 Specify the C-SKY target processor. Valid values for @var{cpu} are:
19008 @samp{ck801}, @samp{ck801t},
19009 @samp{ck802}, @samp{ck802t}, @samp{ck802j},
19010 @samp{ck803}, @samp{ck803h}, @samp{ck803t}, @samp{ck803ht},
19011 @samp{ck803f}, @samp{ck803fh}, @samp{ck803e}, @samp{ck803eh},
19012 @samp{ck803et}, @samp{ck803eht}, @samp{ck803ef}, @samp{ck803efh},
19013 @samp{ck803ft}, @samp{ck803eft}, @samp{ck803efht}, @samp{ck803r1},
19014 @samp{ck803hr1}, @samp{ck803tr1}, @samp{ck803htr1}, @samp{ck803fr1},
19015 @samp{ck803fhr1}, @samp{ck803er1}, @samp{ck803ehr1}, @samp{ck803etr1},
19016 @samp{ck803ehtr1}, @samp{ck803efr1}, @samp{ck803efhr1}, @samp{ck803ftr1},
19017 @samp{ck803eftr1}, @samp{ck803efhtr1},
19018 @samp{ck803s}, @samp{ck803st}, @samp{ck803se}, @samp{ck803sf},
19019 @samp{ck803sef}, @samp{ck803seft},
19020 @samp{ck807e}, @samp{ck807ef}, @samp{ck807}, @samp{ck807f},
19021 @samp{ck810e}, @samp{ck810et}, @samp{ck810ef}, @samp{ck810eft},
19022 @samp{ck810}, @samp{ck810v}, @samp{ck810f}, @samp{ck810t}, @samp{ck810fv},
19023 @samp{ck810tv}, @samp{ck810ft}, and @samp{ck810ftv}.
19026 @opindex mbig-endian
19029 @itemx -mlittle-endian
19030 @opindex mlittle-endian
19034 Select big- or little-endian code. The default is little-endian.
19037 @opindex mhard-float
19038 @itemx -msoft-float
19039 @opindex msoft-float
19041 Select hardware or software floating-point implementations.
19042 The default is soft float.
19044 @item -mdouble-float
19045 @itemx -mno-double-float
19046 @opindex mdouble-float
19047 When @option{-mhard-float} is in effect, enable generation of
19048 double-precision float instructions. This is the default except
19049 when compiling for CK803.
19054 When @option{-mhard-float} is in effect, enable generation of
19055 @code{frecipd}, @code{fsqrtd}, and @code{fdivd} instructions.
19056 This is the default except when compiling for CK803.
19058 @item -mfpu=@var{fpu}
19060 Select the floating-point processor. This option can only be used with
19061 @option{-mhard-float}.
19062 Values for @var{fpu} are
19063 @samp{fpv2_sf} (equivalent to @samp{-mno-double-float -mno-fdivdu}),
19064 @samp{fpv2} (@samp{-mdouble-float -mno-divdu}), and
19065 @samp{fpv2_divd} (@samp{-mdouble-float -mdivdu}).
19070 Enable the extended @code{lrw} instruction. This option defaults to on
19071 for CK801 and off otherwise.
19076 Enable interrupt stack instructions; the default is off.
19078 The @option{-mistack} option is required to handle the
19079 @code{interrupt} and @code{isr} function attributes
19080 (@pxref{C-SKY Function Attributes}).
19084 Enable multiprocessor instructions; the default is off.
19088 Enable coprocessor instructions; the default is off.
19092 Enable coprocessor instructions; the default is off.
19096 Enable C-SKY security instructions; the default is off.
19100 Enable C-SKY trust instructions; the default is off.
19108 Enable C-SKY DSP, Enhanced DSP, or Vector DSP instructions, respectively.
19109 All of these options default to off.
19114 Generate divide instructions. Default is off.
19119 Generate code for Smart Mode, using only registers numbered 0-7 to allow
19120 use of 16-bit instructions. This option is ignored for CK801 where this
19121 is the required behavior, and it defaults to on for CK802.
19122 For other targets, the default is off.
19124 @item -mhigh-registers
19125 @itemx -mno-high-registers
19126 @opindex mhigh-registers
19127 Generate code using the high registers numbered 16-31. This option
19128 is not supported on CK801, CK802, or CK803, and is enabled by default
19129 for other processors.
19134 Generate code using global anchor symbol addresses.
19137 @itemx -mno-pushpop
19139 Generate code using @code{push} and @code{pop} instructions. This option
19142 @item -mmultiple-stld
19144 @itemx -mno-multiple-stld
19146 @opindex mmultiple-stld
19147 Generate code using @code{stm} and @code{ldm} instructions. This option
19148 isn't supported on CK801 but is enabled by default on other processors.
19151 @itemx -mno-constpool
19152 @opindex mconstpool
19153 Create constant pools in the compiler instead of deferring it to the
19154 assembler. This option is the default and required for correct code
19155 generation on CK801 and CK802, and is optional on other processors.
19158 @item -mno-stack-size
19159 @opindex mstack-size
19160 Emit @code{.stack_size} directives for each function in the assembly
19161 output. This option defaults to off.
19166 Generate code for the C-SKY compiler runtime instead of libgcc. This
19167 option defaults to off.
19169 @item -mbranch-cost=@var{n}
19170 @opindex mbranch-cost=
19171 Set the branch costs to roughly @code{n} instructions. The default is 1.
19173 @item -msched-prolog
19174 @itemx -mno-sched-prolog
19175 @opindex msched-prolog
19176 Permit scheduling of function prologue and epilogue sequences. Using
19177 this option can result in code that is not compliant with the C-SKY V2 ABI
19178 prologue requirements and that cannot be debugged or backtraced.
19179 It is disabled by default.
19183 @node Darwin Options
19184 @subsection Darwin Options
19185 @cindex Darwin options
19187 These options are defined for all architectures running the Darwin operating
19190 FSF GCC on Darwin does not create ``fat'' object files; it creates
19191 an object file for the single architecture that GCC was built to
19192 target. Apple's GCC on Darwin does create ``fat'' files if multiple
19193 @option{-arch} options are used; it does so by running the compiler or
19194 linker multiple times and joining the results together with
19197 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
19198 @samp{i686}) is determined by the flags that specify the ISA
19199 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
19200 @option{-force_cpusubtype_ALL} option can be used to override this.
19202 The Darwin tools vary in their behavior when presented with an ISA
19203 mismatch. The assembler, @file{as}, only permits instructions to
19204 be used that are valid for the subtype of the file it is generating,
19205 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
19206 The linker for shared libraries, @file{/usr/bin/libtool}, fails
19207 and prints an error if asked to create a shared library with a less
19208 restrictive subtype than its input files (for instance, trying to put
19209 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
19210 for executables, @command{ld}, quietly gives the executable the most
19211 restrictive subtype of any of its input files.
19216 Add the framework directory @var{dir} to the head of the list of
19217 directories to be searched for header files. These directories are
19218 interleaved with those specified by @option{-I} options and are
19219 scanned in a left-to-right order.
19221 A framework directory is a directory with frameworks in it. A
19222 framework is a directory with a @file{Headers} and/or
19223 @file{PrivateHeaders} directory contained directly in it that ends
19224 in @file{.framework}. The name of a framework is the name of this
19225 directory excluding the @file{.framework}. Headers associated with
19226 the framework are found in one of those two directories, with
19227 @file{Headers} being searched first. A subframework is a framework
19228 directory that is in a framework's @file{Frameworks} directory.
19229 Includes of subframework headers can only appear in a header of a
19230 framework that contains the subframework, or in a sibling subframework
19231 header. Two subframeworks are siblings if they occur in the same
19232 framework. A subframework should not have the same name as a
19233 framework; a warning is issued if this is violated. Currently a
19234 subframework cannot have subframeworks; in the future, the mechanism
19235 may be extended to support this. The standard frameworks can be found
19236 in @file{/System/Library/Frameworks} and
19237 @file{/Library/Frameworks}. An example include looks like
19238 @code{#include <Framework/header.h>}, where @file{Framework} denotes
19239 the name of the framework and @file{header.h} is found in the
19240 @file{PrivateHeaders} or @file{Headers} directory.
19242 @item -iframework@var{dir}
19243 @opindex iframework
19244 Like @option{-F} except the directory is a treated as a system
19245 directory. The main difference between this @option{-iframework} and
19246 @option{-F} is that with @option{-iframework} the compiler does not
19247 warn about constructs contained within header files found via
19248 @var{dir}. This option is valid only for the C family of languages.
19252 Emit debugging information for symbols that are used. For stabs
19253 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
19254 This is by default ON@.
19258 Emit debugging information for all symbols and types.
19260 @item -mmacosx-version-min=@var{version}
19261 The earliest version of MacOS X that this executable will run on
19262 is @var{version}. Typical values of @var{version} include @code{10.1},
19263 @code{10.2}, and @code{10.3.9}.
19265 If the compiler was built to use the system's headers by default,
19266 then the default for this option is the system version on which the
19267 compiler is running, otherwise the default is to make choices that
19268 are compatible with as many systems and code bases as possible.
19272 Enable kernel development mode. The @option{-mkernel} option sets
19273 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
19274 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
19275 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
19276 applicable. This mode also sets @option{-mno-altivec},
19277 @option{-msoft-float}, @option{-fno-builtin} and
19278 @option{-mlong-branch} for PowerPC targets.
19280 @item -mone-byte-bool
19281 @opindex mone-byte-bool
19282 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
19283 By default @code{sizeof(bool)} is @code{4} when compiling for
19284 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
19285 option has no effect on x86.
19287 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
19288 to generate code that is not binary compatible with code generated
19289 without that switch. Using this switch may require recompiling all
19290 other modules in a program, including system libraries. Use this
19291 switch to conform to a non-default data model.
19293 @item -mfix-and-continue
19294 @itemx -ffix-and-continue
19295 @itemx -findirect-data
19296 @opindex mfix-and-continue
19297 @opindex ffix-and-continue
19298 @opindex findirect-data
19299 Generate code suitable for fast turnaround development, such as to
19300 allow GDB to dynamically load @file{.o} files into already-running
19301 programs. @option{-findirect-data} and @option{-ffix-and-continue}
19302 are provided for backwards compatibility.
19306 Loads all members of static archive libraries.
19307 See man ld(1) for more information.
19309 @item -arch_errors_fatal
19310 @opindex arch_errors_fatal
19311 Cause the errors having to do with files that have the wrong architecture
19314 @item -bind_at_load
19315 @opindex bind_at_load
19316 Causes the output file to be marked such that the dynamic linker will
19317 bind all undefined references when the file is loaded or launched.
19321 Produce a Mach-o bundle format file.
19322 See man ld(1) for more information.
19324 @item -bundle_loader @var{executable}
19325 @opindex bundle_loader
19326 This option specifies the @var{executable} that will load the build
19327 output file being linked. See man ld(1) for more information.
19330 @opindex dynamiclib
19331 When passed this option, GCC produces a dynamic library instead of
19332 an executable when linking, using the Darwin @file{libtool} command.
19334 @item -force_cpusubtype_ALL
19335 @opindex force_cpusubtype_ALL
19336 This causes GCC's output file to have the @samp{ALL} subtype, instead of
19337 one controlled by the @option{-mcpu} or @option{-march} option.
19339 @item -allowable_client @var{client_name}
19340 @itemx -client_name
19341 @itemx -compatibility_version
19342 @itemx -current_version
19344 @itemx -dependency-file
19346 @itemx -dylinker_install_name
19348 @itemx -exported_symbols_list
19351 @itemx -flat_namespace
19352 @itemx -force_flat_namespace
19353 @itemx -headerpad_max_install_names
19356 @itemx -install_name
19357 @itemx -keep_private_externs
19358 @itemx -multi_module
19359 @itemx -multiply_defined
19360 @itemx -multiply_defined_unused
19363 @itemx -no_dead_strip_inits_and_terms
19364 @itemx -nofixprebinding
19365 @itemx -nomultidefs
19367 @itemx -noseglinkedit
19368 @itemx -pagezero_size
19370 @itemx -prebind_all_twolevel_modules
19371 @itemx -private_bundle
19373 @itemx -read_only_relocs
19375 @itemx -sectobjectsymbols
19379 @itemx -sectobjectsymbols
19382 @itemx -segs_read_only_addr
19384 @itemx -segs_read_write_addr
19385 @itemx -seg_addr_table
19386 @itemx -seg_addr_table_filename
19387 @itemx -seglinkedit
19389 @itemx -segs_read_only_addr
19390 @itemx -segs_read_write_addr
19391 @itemx -single_module
19393 @itemx -sub_library
19395 @itemx -sub_umbrella
19396 @itemx -twolevel_namespace
19399 @itemx -unexported_symbols_list
19400 @itemx -weak_reference_mismatches
19401 @itemx -whatsloaded
19402 @opindex allowable_client
19403 @opindex client_name
19404 @opindex compatibility_version
19405 @opindex current_version
19406 @opindex dead_strip
19407 @opindex dependency-file
19408 @opindex dylib_file
19409 @opindex dylinker_install_name
19411 @opindex exported_symbols_list
19413 @opindex flat_namespace
19414 @opindex force_flat_namespace
19415 @opindex headerpad_max_install_names
19416 @opindex image_base
19418 @opindex install_name
19419 @opindex keep_private_externs
19420 @opindex multi_module
19421 @opindex multiply_defined
19422 @opindex multiply_defined_unused
19423 @opindex noall_load
19424 @opindex no_dead_strip_inits_and_terms
19425 @opindex nofixprebinding
19426 @opindex nomultidefs
19428 @opindex noseglinkedit
19429 @opindex pagezero_size
19431 @opindex prebind_all_twolevel_modules
19432 @opindex private_bundle
19433 @opindex read_only_relocs
19435 @opindex sectobjectsymbols
19438 @opindex sectcreate
19439 @opindex sectobjectsymbols
19442 @opindex segs_read_only_addr
19443 @opindex segs_read_write_addr
19444 @opindex seg_addr_table
19445 @opindex seg_addr_table_filename
19446 @opindex seglinkedit
19448 @opindex segs_read_only_addr
19449 @opindex segs_read_write_addr
19450 @opindex single_module
19452 @opindex sub_library
19453 @opindex sub_umbrella
19454 @opindex twolevel_namespace
19457 @opindex unexported_symbols_list
19458 @opindex weak_reference_mismatches
19459 @opindex whatsloaded
19460 These options are passed to the Darwin linker. The Darwin linker man page
19461 describes them in detail.
19464 @node DEC Alpha Options
19465 @subsection DEC Alpha Options
19467 These @samp{-m} options are defined for the DEC Alpha implementations:
19470 @item -mno-soft-float
19471 @itemx -msoft-float
19472 @opindex mno-soft-float
19473 @opindex msoft-float
19474 Use (do not use) the hardware floating-point instructions for
19475 floating-point operations. When @option{-msoft-float} is specified,
19476 functions in @file{libgcc.a} are used to perform floating-point
19477 operations. Unless they are replaced by routines that emulate the
19478 floating-point operations, or compiled in such a way as to call such
19479 emulations routines, these routines issue floating-point
19480 operations. If you are compiling for an Alpha without floating-point
19481 operations, you must ensure that the library is built so as not to call
19484 Note that Alpha implementations without floating-point operations are
19485 required to have floating-point registers.
19488 @itemx -mno-fp-regs
19490 @opindex mno-fp-regs
19491 Generate code that uses (does not use) the floating-point register set.
19492 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
19493 register set is not used, floating-point operands are passed in integer
19494 registers as if they were integers and floating-point results are passed
19495 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
19496 so any function with a floating-point argument or return value called by code
19497 compiled with @option{-mno-fp-regs} must also be compiled with that
19500 A typical use of this option is building a kernel that does not use,
19501 and hence need not save and restore, any floating-point registers.
19505 The Alpha architecture implements floating-point hardware optimized for
19506 maximum performance. It is mostly compliant with the IEEE floating-point
19507 standard. However, for full compliance, software assistance is
19508 required. This option generates code fully IEEE-compliant code
19509 @emph{except} that the @var{inexact-flag} is not maintained (see below).
19510 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
19511 defined during compilation. The resulting code is less efficient but is
19512 able to correctly support denormalized numbers and exceptional IEEE
19513 values such as not-a-number and plus/minus infinity. Other Alpha
19514 compilers call this option @option{-ieee_with_no_inexact}.
19516 @item -mieee-with-inexact
19517 @opindex mieee-with-inexact
19518 This is like @option{-mieee} except the generated code also maintains
19519 the IEEE @var{inexact-flag}. Turning on this option causes the
19520 generated code to implement fully-compliant IEEE math. In addition to
19521 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
19522 macro. On some Alpha implementations the resulting code may execute
19523 significantly slower than the code generated by default. Since there is
19524 very little code that depends on the @var{inexact-flag}, you should
19525 normally not specify this option. Other Alpha compilers call this
19526 option @option{-ieee_with_inexact}.
19528 @item -mfp-trap-mode=@var{trap-mode}
19529 @opindex mfp-trap-mode
19530 This option controls what floating-point related traps are enabled.
19531 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
19532 The trap mode can be set to one of four values:
19536 This is the default (normal) setting. The only traps that are enabled
19537 are the ones that cannot be disabled in software (e.g., division by zero
19541 In addition to the traps enabled by @samp{n}, underflow traps are enabled
19545 Like @samp{u}, but the instructions are marked to be safe for software
19546 completion (see Alpha architecture manual for details).
19549 Like @samp{su}, but inexact traps are enabled as well.
19552 @item -mfp-rounding-mode=@var{rounding-mode}
19553 @opindex mfp-rounding-mode
19554 Selects the IEEE rounding mode. Other Alpha compilers call this option
19555 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
19560 Normal IEEE rounding mode. Floating-point numbers are rounded towards
19561 the nearest machine number or towards the even machine number in case
19565 Round towards minus infinity.
19568 Chopped rounding mode. Floating-point numbers are rounded towards zero.
19571 Dynamic rounding mode. A field in the floating-point control register
19572 (@var{fpcr}, see Alpha architecture reference manual) controls the
19573 rounding mode in effect. The C library initializes this register for
19574 rounding towards plus infinity. Thus, unless your program modifies the
19575 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
19578 @item -mtrap-precision=@var{trap-precision}
19579 @opindex mtrap-precision
19580 In the Alpha architecture, floating-point traps are imprecise. This
19581 means without software assistance it is impossible to recover from a
19582 floating trap and program execution normally needs to be terminated.
19583 GCC can generate code that can assist operating system trap handlers
19584 in determining the exact location that caused a floating-point trap.
19585 Depending on the requirements of an application, different levels of
19586 precisions can be selected:
19590 Program precision. This option is the default and means a trap handler
19591 can only identify which program caused a floating-point exception.
19594 Function precision. The trap handler can determine the function that
19595 caused a floating-point exception.
19598 Instruction precision. The trap handler can determine the exact
19599 instruction that caused a floating-point exception.
19602 Other Alpha compilers provide the equivalent options called
19603 @option{-scope_safe} and @option{-resumption_safe}.
19605 @item -mieee-conformant
19606 @opindex mieee-conformant
19607 This option marks the generated code as IEEE conformant. You must not
19608 use this option unless you also specify @option{-mtrap-precision=i} and either
19609 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
19610 is to emit the line @samp{.eflag 48} in the function prologue of the
19611 generated assembly file.
19613 @item -mbuild-constants
19614 @opindex mbuild-constants
19615 Normally GCC examines a 32- or 64-bit integer constant to
19616 see if it can construct it from smaller constants in two or three
19617 instructions. If it cannot, it outputs the constant as a literal and
19618 generates code to load it from the data segment at run time.
19620 Use this option to require GCC to construct @emph{all} integer constants
19621 using code, even if it takes more instructions (the maximum is six).
19623 You typically use this option to build a shared library dynamic
19624 loader. Itself a shared library, it must relocate itself in memory
19625 before it can find the variables and constants in its own data segment.
19643 Indicate whether GCC should generate code to use the optional BWX,
19644 CIX, FIX and MAX instruction sets. The default is to use the instruction
19645 sets supported by the CPU type specified via @option{-mcpu=} option or that
19646 of the CPU on which GCC was built if none is specified.
19649 @itemx -mfloat-ieee
19650 @opindex mfloat-vax
19651 @opindex mfloat-ieee
19652 Generate code that uses (does not use) VAX F and G floating-point
19653 arithmetic instead of IEEE single and double precision.
19655 @item -mexplicit-relocs
19656 @itemx -mno-explicit-relocs
19657 @opindex mexplicit-relocs
19658 @opindex mno-explicit-relocs
19659 Older Alpha assemblers provided no way to generate symbol relocations
19660 except via assembler macros. Use of these macros does not allow
19661 optimal instruction scheduling. GNU binutils as of version 2.12
19662 supports a new syntax that allows the compiler to explicitly mark
19663 which relocations should apply to which instructions. This option
19664 is mostly useful for debugging, as GCC detects the capabilities of
19665 the assembler when it is built and sets the default accordingly.
19668 @itemx -mlarge-data
19669 @opindex msmall-data
19670 @opindex mlarge-data
19671 When @option{-mexplicit-relocs} is in effect, static data is
19672 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
19673 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
19674 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
19675 16-bit relocations off of the @code{$gp} register. This limits the
19676 size of the small data area to 64KB, but allows the variables to be
19677 directly accessed via a single instruction.
19679 The default is @option{-mlarge-data}. With this option the data area
19680 is limited to just below 2GB@. Programs that require more than 2GB of
19681 data must use @code{malloc} or @code{mmap} to allocate the data in the
19682 heap instead of in the program's data segment.
19684 When generating code for shared libraries, @option{-fpic} implies
19685 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
19688 @itemx -mlarge-text
19689 @opindex msmall-text
19690 @opindex mlarge-text
19691 When @option{-msmall-text} is used, the compiler assumes that the
19692 code of the entire program (or shared library) fits in 4MB, and is
19693 thus reachable with a branch instruction. When @option{-msmall-data}
19694 is used, the compiler can assume that all local symbols share the
19695 same @code{$gp} value, and thus reduce the number of instructions
19696 required for a function call from 4 to 1.
19698 The default is @option{-mlarge-text}.
19700 @item -mcpu=@var{cpu_type}
19702 Set the instruction set and instruction scheduling parameters for
19703 machine type @var{cpu_type}. You can specify either the @samp{EV}
19704 style name or the corresponding chip number. GCC supports scheduling
19705 parameters for the EV4, EV5 and EV6 family of processors and
19706 chooses the default values for the instruction set from the processor
19707 you specify. If you do not specify a processor type, GCC defaults
19708 to the processor on which the compiler was built.
19710 Supported values for @var{cpu_type} are
19716 Schedules as an EV4 and has no instruction set extensions.
19720 Schedules as an EV5 and has no instruction set extensions.
19724 Schedules as an EV5 and supports the BWX extension.
19729 Schedules as an EV5 and supports the BWX and MAX extensions.
19733 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
19737 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
19740 Native toolchains also support the value @samp{native},
19741 which selects the best architecture option for the host processor.
19742 @option{-mcpu=native} has no effect if GCC does not recognize
19745 @item -mtune=@var{cpu_type}
19747 Set only the instruction scheduling parameters for machine type
19748 @var{cpu_type}. The instruction set is not changed.
19750 Native toolchains also support the value @samp{native},
19751 which selects the best architecture option for the host processor.
19752 @option{-mtune=native} has no effect if GCC does not recognize
19755 @item -mmemory-latency=@var{time}
19756 @opindex mmemory-latency
19757 Sets the latency the scheduler should assume for typical memory
19758 references as seen by the application. This number is highly
19759 dependent on the memory access patterns used by the application
19760 and the size of the external cache on the machine.
19762 Valid options for @var{time} are
19766 A decimal number representing clock cycles.
19772 The compiler contains estimates of the number of clock cycles for
19773 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
19774 (also called Dcache, Scache, and Bcache), as well as to main memory.
19775 Note that L3 is only valid for EV5.
19781 @subsection FR30 Options
19782 @cindex FR30 Options
19784 These options are defined specifically for the FR30 port.
19788 @item -msmall-model
19789 @opindex msmall-model
19790 Use the small address space model. This can produce smaller code, but
19791 it does assume that all symbolic values and addresses fit into a
19796 Assume that runtime support has been provided and so there is no need
19797 to include the simulator library (@file{libsim.a}) on the linker
19803 @subsection FT32 Options
19804 @cindex FT32 Options
19806 These options are defined specifically for the FT32 port.
19812 Specifies that the program will be run on the simulator. This causes
19813 an alternate runtime startup and library to be linked.
19814 You must not use this option when generating programs that will run on
19815 real hardware; you must provide your own runtime library for whatever
19816 I/O functions are needed.
19820 Enable Local Register Allocation. This is still experimental for FT32,
19821 so by default the compiler uses standard reload.
19825 Do not use div and mod instructions.
19829 Enable use of the extended instructions of the FT32B processor.
19833 Compress all code using the Ft32B code compression scheme.
19837 Do not generate code that reads program memory.
19842 @subsection FRV Options
19843 @cindex FRV Options
19849 Only use the first 32 general-purpose registers.
19854 Use all 64 general-purpose registers.
19859 Use only the first 32 floating-point registers.
19864 Use all 64 floating-point registers.
19867 @opindex mhard-float
19869 Use hardware instructions for floating-point operations.
19872 @opindex msoft-float
19874 Use library routines for floating-point operations.
19879 Dynamically allocate condition code registers.
19884 Do not try to dynamically allocate condition code registers, only
19885 use @code{icc0} and @code{fcc0}.
19890 Change ABI to use double word insns.
19896 Do not use double word instructions.
19901 Use floating-point double instructions.
19904 @opindex mno-double
19906 Do not use floating-point double instructions.
19911 Use media instructions.
19916 Do not use media instructions.
19921 Use multiply and add/subtract instructions.
19924 @opindex mno-muladd
19926 Do not use multiply and add/subtract instructions.
19931 Select the FDPIC ABI, which uses function descriptors to represent
19932 pointers to functions. Without any PIC/PIE-related options, it
19933 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
19934 assumes GOT entries and small data are within a 12-bit range from the
19935 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
19936 are computed with 32 bits.
19937 With a @samp{bfin-elf} target, this option implies @option{-msim}.
19940 @opindex minline-plt
19942 Enable inlining of PLT entries in function calls to functions that are
19943 not known to bind locally. It has no effect without @option{-mfdpic}.
19944 It's enabled by default if optimizing for speed and compiling for
19945 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
19946 optimization option such as @option{-O3} or above is present in the
19952 Assume a large TLS segment when generating thread-local code.
19957 Do not assume a large TLS segment when generating thread-local code.
19962 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
19963 that is known to be in read-only sections. It's enabled by default,
19964 except for @option{-fpic} or @option{-fpie}: even though it may help
19965 make the global offset table smaller, it trades 1 instruction for 4.
19966 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
19967 one of which may be shared by multiple symbols, and it avoids the need
19968 for a GOT entry for the referenced symbol, so it's more likely to be a
19969 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
19971 @item -multilib-library-pic
19972 @opindex multilib-library-pic
19974 Link with the (library, not FD) pic libraries. It's implied by
19975 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
19976 @option{-fpic} without @option{-mfdpic}. You should never have to use
19980 @opindex mlinked-fp
19982 Follow the EABI requirement of always creating a frame pointer whenever
19983 a stack frame is allocated. This option is enabled by default and can
19984 be disabled with @option{-mno-linked-fp}.
19987 @opindex mlong-calls
19989 Use indirect addressing to call functions outside the current
19990 compilation unit. This allows the functions to be placed anywhere
19991 within the 32-bit address space.
19993 @item -malign-labels
19994 @opindex malign-labels
19996 Try to align labels to an 8-byte boundary by inserting NOPs into the
19997 previous packet. This option only has an effect when VLIW packing
19998 is enabled. It doesn't create new packets; it merely adds NOPs to
20001 @item -mlibrary-pic
20002 @opindex mlibrary-pic
20004 Generate position-independent EABI code.
20009 Use only the first four media accumulator registers.
20014 Use all eight media accumulator registers.
20019 Pack VLIW instructions.
20024 Do not pack VLIW instructions.
20027 @opindex mno-eflags
20029 Do not mark ABI switches in e_flags.
20032 @opindex mcond-move
20034 Enable the use of conditional-move instructions (default).
20036 This switch is mainly for debugging the compiler and will likely be removed
20037 in a future version.
20039 @item -mno-cond-move
20040 @opindex mno-cond-move
20042 Disable the use of conditional-move instructions.
20044 This switch is mainly for debugging the compiler and will likely be removed
20045 in a future version.
20050 Enable the use of conditional set instructions (default).
20052 This switch is mainly for debugging the compiler and will likely be removed
20053 in a future version.
20058 Disable the use of conditional set instructions.
20060 This switch is mainly for debugging the compiler and will likely be removed
20061 in a future version.
20064 @opindex mcond-exec
20066 Enable the use of conditional execution (default).
20068 This switch is mainly for debugging the compiler and will likely be removed
20069 in a future version.
20071 @item -mno-cond-exec
20072 @opindex mno-cond-exec
20074 Disable the use of conditional execution.
20076 This switch is mainly for debugging the compiler and will likely be removed
20077 in a future version.
20079 @item -mvliw-branch
20080 @opindex mvliw-branch
20082 Run a pass to pack branches into VLIW instructions (default).
20084 This switch is mainly for debugging the compiler and will likely be removed
20085 in a future version.
20087 @item -mno-vliw-branch
20088 @opindex mno-vliw-branch
20090 Do not run a pass to pack branches into VLIW instructions.
20092 This switch is mainly for debugging the compiler and will likely be removed
20093 in a future version.
20095 @item -mmulti-cond-exec
20096 @opindex mmulti-cond-exec
20098 Enable optimization of @code{&&} and @code{||} in conditional execution
20101 This switch is mainly for debugging the compiler and will likely be removed
20102 in a future version.
20104 @item -mno-multi-cond-exec
20105 @opindex mno-multi-cond-exec
20107 Disable optimization of @code{&&} and @code{||} in conditional execution.
20109 This switch is mainly for debugging the compiler and will likely be removed
20110 in a future version.
20112 @item -mnested-cond-exec
20113 @opindex mnested-cond-exec
20115 Enable nested conditional execution optimizations (default).
20117 This switch is mainly for debugging the compiler and will likely be removed
20118 in a future version.
20120 @item -mno-nested-cond-exec
20121 @opindex mno-nested-cond-exec
20123 Disable nested conditional execution optimizations.
20125 This switch is mainly for debugging the compiler and will likely be removed
20126 in a future version.
20128 @item -moptimize-membar
20129 @opindex moptimize-membar
20131 This switch removes redundant @code{membar} instructions from the
20132 compiler-generated code. It is enabled by default.
20134 @item -mno-optimize-membar
20135 @opindex mno-optimize-membar
20136 @opindex moptimize-membar
20138 This switch disables the automatic removal of redundant @code{membar}
20139 instructions from the generated code.
20141 @item -mtomcat-stats
20142 @opindex mtomcat-stats
20144 Cause gas to print out tomcat statistics.
20146 @item -mcpu=@var{cpu}
20149 Select the processor type for which to generate code. Possible values are
20150 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
20151 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
20155 @node GNU/Linux Options
20156 @subsection GNU/Linux Options
20158 These @samp{-m} options are defined for GNU/Linux targets:
20163 Use the GNU C library. This is the default except
20164 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
20165 @samp{*-*-linux-*android*} targets.
20169 Use uClibc C library. This is the default on
20170 @samp{*-*-linux-*uclibc*} targets.
20174 Use the musl C library. This is the default on
20175 @samp{*-*-linux-*musl*} targets.
20179 Use Bionic C library. This is the default on
20180 @samp{*-*-linux-*android*} targets.
20184 Compile code compatible with Android platform. This is the default on
20185 @samp{*-*-linux-*android*} targets.
20187 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
20188 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
20189 this option makes the GCC driver pass Android-specific options to the linker.
20190 Finally, this option causes the preprocessor macro @code{__ANDROID__}
20193 @item -tno-android-cc
20194 @opindex tno-android-cc
20195 Disable compilation effects of @option{-mandroid}, i.e., do not enable
20196 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
20197 @option{-fno-rtti} by default.
20199 @item -tno-android-ld
20200 @opindex tno-android-ld
20201 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
20202 linking options to the linker.
20206 @node H8/300 Options
20207 @subsection H8/300 Options
20209 These @samp{-m} options are defined for the H8/300 implementations:
20214 Shorten some address references at link time, when possible; uses the
20215 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
20216 ld, Using ld}, for a fuller description.
20220 Generate code for the H8/300H@.
20224 Generate code for the H8S@.
20228 Generate code for the H8S and H8/300H in the normal mode. This switch
20229 must be used either with @option{-mh} or @option{-ms}.
20233 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
20237 Extended registers are stored on stack before execution of function
20238 with monitor attribute. Default option is @option{-mexr}.
20239 This option is valid only for H8S targets.
20244 Extended registers are not stored on stack before execution of function
20245 with monitor attribute. Default option is @option{-mno-exr}.
20246 This option is valid only for H8S targets.
20250 Make @code{int} data 32 bits by default.
20253 @opindex malign-300
20254 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
20255 The default for the H8/300H and H8S is to align longs and floats on
20257 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
20258 This option has no effect on the H8/300.
20262 @subsection HPPA Options
20263 @cindex HPPA Options
20265 These @samp{-m} options are defined for the HPPA family of computers:
20268 @item -march=@var{architecture-type}
20270 Generate code for the specified architecture. The choices for
20271 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
20272 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
20273 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
20274 architecture option for your machine. Code compiled for lower numbered
20275 architectures runs on higher numbered architectures, but not the
20278 @item -mpa-risc-1-0
20279 @itemx -mpa-risc-1-1
20280 @itemx -mpa-risc-2-0
20281 @opindex mpa-risc-1-0
20282 @opindex mpa-risc-1-1
20283 @opindex mpa-risc-2-0
20284 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
20286 @item -mcaller-copies
20287 @opindex mcaller-copies
20288 The caller copies function arguments passed by hidden reference. This
20289 option should be used with care as it is not compatible with the default
20290 32-bit runtime. However, only aggregates larger than eight bytes are
20291 passed by hidden reference and the option provides better compatibility
20294 @item -mjump-in-delay
20295 @opindex mjump-in-delay
20296 This option is ignored and provided for compatibility purposes only.
20298 @item -mdisable-fpregs
20299 @opindex mdisable-fpregs
20300 Prevent floating-point registers from being used in any manner. This is
20301 necessary for compiling kernels that perform lazy context switching of
20302 floating-point registers. If you use this option and attempt to perform
20303 floating-point operations, the compiler aborts.
20305 @item -mdisable-indexing
20306 @opindex mdisable-indexing
20307 Prevent the compiler from using indexing address modes. This avoids some
20308 rather obscure problems when compiling MIG generated code under MACH@.
20310 @item -mno-space-regs
20311 @opindex mno-space-regs
20312 @opindex mspace-regs
20313 Generate code that assumes the target has no space registers. This allows
20314 GCC to generate faster indirect calls and use unscaled index address modes.
20316 Such code is suitable for level 0 PA systems and kernels.
20318 @item -mfast-indirect-calls
20319 @opindex mfast-indirect-calls
20320 Generate code that assumes calls never cross space boundaries. This
20321 allows GCC to emit code that performs faster indirect calls.
20323 This option does not work in the presence of shared libraries or nested
20326 @item -mfixed-range=@var{register-range}
20327 @opindex mfixed-range
20328 Generate code treating the given register range as fixed registers.
20329 A fixed register is one that the register allocator cannot use. This is
20330 useful when compiling kernel code. A register range is specified as
20331 two registers separated by a dash. Multiple register ranges can be
20332 specified separated by a comma.
20334 @item -mlong-load-store
20335 @opindex mlong-load-store
20336 Generate 3-instruction load and store sequences as sometimes required by
20337 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
20340 @item -mportable-runtime
20341 @opindex mportable-runtime
20342 Use the portable calling conventions proposed by HP for ELF systems.
20346 Enable the use of assembler directives only GAS understands.
20348 @item -mschedule=@var{cpu-type}
20350 Schedule code according to the constraints for the machine type
20351 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
20352 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
20353 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
20354 proper scheduling option for your machine. The default scheduling is
20358 @opindex mlinker-opt
20359 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
20360 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
20361 linkers in which they give bogus error messages when linking some programs.
20364 @opindex msoft-float
20365 Generate output containing library calls for floating point.
20366 @strong{Warning:} the requisite libraries are not available for all HPPA
20367 targets. Normally the facilities of the machine's usual C compiler are
20368 used, but this cannot be done directly in cross-compilation. You must make
20369 your own arrangements to provide suitable library functions for
20372 @option{-msoft-float} changes the calling convention in the output file;
20373 therefore, it is only useful if you compile @emph{all} of a program with
20374 this option. In particular, you need to compile @file{libgcc.a}, the
20375 library that comes with GCC, with @option{-msoft-float} in order for
20380 Generate the predefine, @code{_SIO}, for server IO@. The default is
20381 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
20382 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
20383 options are available under HP-UX and HI-UX@.
20387 Use options specific to GNU @command{ld}.
20388 This passes @option{-shared} to @command{ld} when
20389 building a shared library. It is the default when GCC is configured,
20390 explicitly or implicitly, with the GNU linker. This option does not
20391 affect which @command{ld} is called; it only changes what parameters
20392 are passed to that @command{ld}.
20393 The @command{ld} that is called is determined by the
20394 @option{--with-ld} configure option, GCC's program search path, and
20395 finally by the user's @env{PATH}. The linker used by GCC can be printed
20396 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
20397 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
20401 Use options specific to HP @command{ld}.
20402 This passes @option{-b} to @command{ld} when building
20403 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
20404 links. It is the default when GCC is configured, explicitly or
20405 implicitly, with the HP linker. This option does not affect
20406 which @command{ld} is called; it only changes what parameters are passed to that
20408 The @command{ld} that is called is determined by the @option{--with-ld}
20409 configure option, GCC's program search path, and finally by the user's
20410 @env{PATH}. The linker used by GCC can be printed using @samp{which
20411 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
20412 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
20415 @opindex mno-long-calls
20416 @opindex mlong-calls
20417 Generate code that uses long call sequences. This ensures that a call
20418 is always able to reach linker generated stubs. The default is to generate
20419 long calls only when the distance from the call site to the beginning
20420 of the function or translation unit, as the case may be, exceeds a
20421 predefined limit set by the branch type being used. The limits for
20422 normal calls are 7,600,000 and 240,000 bytes, respectively for the
20423 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
20426 Distances are measured from the beginning of functions when using the
20427 @option{-ffunction-sections} option, or when using the @option{-mgas}
20428 and @option{-mno-portable-runtime} options together under HP-UX with
20431 It is normally not desirable to use this option as it degrades
20432 performance. However, it may be useful in large applications,
20433 particularly when partial linking is used to build the application.
20435 The types of long calls used depends on the capabilities of the
20436 assembler and linker, and the type of code being generated. The
20437 impact on systems that support long absolute calls, and long pic
20438 symbol-difference or pc-relative calls should be relatively small.
20439 However, an indirect call is used on 32-bit ELF systems in pic code
20440 and it is quite long.
20442 @item -munix=@var{unix-std}
20444 Generate compiler predefines and select a startfile for the specified
20445 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
20446 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
20447 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
20448 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
20449 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
20452 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
20453 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
20454 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
20455 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
20456 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
20457 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
20459 It is @emph{important} to note that this option changes the interfaces
20460 for various library routines. It also affects the operational behavior
20461 of the C library. Thus, @emph{extreme} care is needed in using this
20464 Library code that is intended to operate with more than one UNIX
20465 standard must test, set and restore the variable @code{__xpg4_extended_mask}
20466 as appropriate. Most GNU software doesn't provide this capability.
20470 Suppress the generation of link options to search libdld.sl when the
20471 @option{-static} option is specified on HP-UX 10 and later.
20475 The HP-UX implementation of setlocale in libc has a dependency on
20476 libdld.sl. There isn't an archive version of libdld.sl. Thus,
20477 when the @option{-static} option is specified, special link options
20478 are needed to resolve this dependency.
20480 On HP-UX 10 and later, the GCC driver adds the necessary options to
20481 link with libdld.sl when the @option{-static} option is specified.
20482 This causes the resulting binary to be dynamic. On the 64-bit port,
20483 the linkers generate dynamic binaries by default in any case. The
20484 @option{-nolibdld} option can be used to prevent the GCC driver from
20485 adding these link options.
20489 Add support for multithreading with the @dfn{dce thread} library
20490 under HP-UX@. This option sets flags for both the preprocessor and
20494 @node IA-64 Options
20495 @subsection IA-64 Options
20496 @cindex IA-64 Options
20498 These are the @samp{-m} options defined for the Intel IA-64 architecture.
20502 @opindex mbig-endian
20503 Generate code for a big-endian target. This is the default for HP-UX@.
20505 @item -mlittle-endian
20506 @opindex mlittle-endian
20507 Generate code for a little-endian target. This is the default for AIX5
20513 @opindex mno-gnu-as
20514 Generate (or don't) code for the GNU assembler. This is the default.
20515 @c Also, this is the default if the configure option @option{--with-gnu-as}
20521 @opindex mno-gnu-ld
20522 Generate (or don't) code for the GNU linker. This is the default.
20523 @c Also, this is the default if the configure option @option{--with-gnu-ld}
20528 Generate code that does not use a global pointer register. The result
20529 is not position independent code, and violates the IA-64 ABI@.
20531 @item -mvolatile-asm-stop
20532 @itemx -mno-volatile-asm-stop
20533 @opindex mvolatile-asm-stop
20534 @opindex mno-volatile-asm-stop
20535 Generate (or don't) a stop bit immediately before and after volatile asm
20538 @item -mregister-names
20539 @itemx -mno-register-names
20540 @opindex mregister-names
20541 @opindex mno-register-names
20542 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
20543 the stacked registers. This may make assembler output more readable.
20549 Disable (or enable) optimizations that use the small data section. This may
20550 be useful for working around optimizer bugs.
20552 @item -mconstant-gp
20553 @opindex mconstant-gp
20554 Generate code that uses a single constant global pointer value. This is
20555 useful when compiling kernel code.
20559 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
20560 This is useful when compiling firmware code.
20562 @item -minline-float-divide-min-latency
20563 @opindex minline-float-divide-min-latency
20564 Generate code for inline divides of floating-point values
20565 using the minimum latency algorithm.
20567 @item -minline-float-divide-max-throughput
20568 @opindex minline-float-divide-max-throughput
20569 Generate code for inline divides of floating-point values
20570 using the maximum throughput algorithm.
20572 @item -mno-inline-float-divide
20573 @opindex mno-inline-float-divide
20574 Do not generate inline code for divides of floating-point values.
20576 @item -minline-int-divide-min-latency
20577 @opindex minline-int-divide-min-latency
20578 Generate code for inline divides of integer values
20579 using the minimum latency algorithm.
20581 @item -minline-int-divide-max-throughput
20582 @opindex minline-int-divide-max-throughput
20583 Generate code for inline divides of integer values
20584 using the maximum throughput algorithm.
20586 @item -mno-inline-int-divide
20587 @opindex mno-inline-int-divide
20588 @opindex minline-int-divide
20589 Do not generate inline code for divides of integer values.
20591 @item -minline-sqrt-min-latency
20592 @opindex minline-sqrt-min-latency
20593 Generate code for inline square roots
20594 using the minimum latency algorithm.
20596 @item -minline-sqrt-max-throughput
20597 @opindex minline-sqrt-max-throughput
20598 Generate code for inline square roots
20599 using the maximum throughput algorithm.
20601 @item -mno-inline-sqrt
20602 @opindex mno-inline-sqrt
20603 Do not generate inline code for @code{sqrt}.
20606 @itemx -mno-fused-madd
20607 @opindex mfused-madd
20608 @opindex mno-fused-madd
20609 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
20610 instructions. The default is to use these instructions.
20612 @item -mno-dwarf2-asm
20613 @itemx -mdwarf2-asm
20614 @opindex mno-dwarf2-asm
20615 @opindex mdwarf2-asm
20616 Don't (or do) generate assembler code for the DWARF line number debugging
20617 info. This may be useful when not using the GNU assembler.
20619 @item -mearly-stop-bits
20620 @itemx -mno-early-stop-bits
20621 @opindex mearly-stop-bits
20622 @opindex mno-early-stop-bits
20623 Allow stop bits to be placed earlier than immediately preceding the
20624 instruction that triggered the stop bit. This can improve instruction
20625 scheduling, but does not always do so.
20627 @item -mfixed-range=@var{register-range}
20628 @opindex mfixed-range
20629 Generate code treating the given register range as fixed registers.
20630 A fixed register is one that the register allocator cannot use. This is
20631 useful when compiling kernel code. A register range is specified as
20632 two registers separated by a dash. Multiple register ranges can be
20633 specified separated by a comma.
20635 @item -mtls-size=@var{tls-size}
20637 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
20640 @item -mtune=@var{cpu-type}
20642 Tune the instruction scheduling for a particular CPU, Valid values are
20643 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
20644 and @samp{mckinley}.
20650 Generate code for a 32-bit or 64-bit environment.
20651 The 32-bit environment sets int, long and pointer to 32 bits.
20652 The 64-bit environment sets int to 32 bits and long and pointer
20653 to 64 bits. These are HP-UX specific flags.
20655 @item -mno-sched-br-data-spec
20656 @itemx -msched-br-data-spec
20657 @opindex mno-sched-br-data-spec
20658 @opindex msched-br-data-spec
20659 (Dis/En)able data speculative scheduling before reload.
20660 This results in generation of @code{ld.a} instructions and
20661 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
20662 The default setting is disabled.
20664 @item -msched-ar-data-spec
20665 @itemx -mno-sched-ar-data-spec
20666 @opindex msched-ar-data-spec
20667 @opindex mno-sched-ar-data-spec
20668 (En/Dis)able data speculative scheduling after reload.
20669 This results in generation of @code{ld.a} instructions and
20670 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
20671 The default setting is enabled.
20673 @item -mno-sched-control-spec
20674 @itemx -msched-control-spec
20675 @opindex mno-sched-control-spec
20676 @opindex msched-control-spec
20677 (Dis/En)able control speculative scheduling. This feature is
20678 available only during region scheduling (i.e.@: before reload).
20679 This results in generation of the @code{ld.s} instructions and
20680 the corresponding check instructions @code{chk.s}.
20681 The default setting is disabled.
20683 @item -msched-br-in-data-spec
20684 @itemx -mno-sched-br-in-data-spec
20685 @opindex msched-br-in-data-spec
20686 @opindex mno-sched-br-in-data-spec
20687 (En/Dis)able speculative scheduling of the instructions that
20688 are dependent on the data speculative loads before reload.
20689 This is effective only with @option{-msched-br-data-spec} enabled.
20690 The default setting is enabled.
20692 @item -msched-ar-in-data-spec
20693 @itemx -mno-sched-ar-in-data-spec
20694 @opindex msched-ar-in-data-spec
20695 @opindex mno-sched-ar-in-data-spec
20696 (En/Dis)able speculative scheduling of the instructions that
20697 are dependent on the data speculative loads after reload.
20698 This is effective only with @option{-msched-ar-data-spec} enabled.
20699 The default setting is enabled.
20701 @item -msched-in-control-spec
20702 @itemx -mno-sched-in-control-spec
20703 @opindex msched-in-control-spec
20704 @opindex mno-sched-in-control-spec
20705 (En/Dis)able speculative scheduling of the instructions that
20706 are dependent on the control speculative loads.
20707 This is effective only with @option{-msched-control-spec} enabled.
20708 The default setting is enabled.
20710 @item -mno-sched-prefer-non-data-spec-insns
20711 @itemx -msched-prefer-non-data-spec-insns
20712 @opindex mno-sched-prefer-non-data-spec-insns
20713 @opindex msched-prefer-non-data-spec-insns
20714 If enabled, data-speculative instructions are chosen for schedule
20715 only if there are no other choices at the moment. This makes
20716 the use of the data speculation much more conservative.
20717 The default setting is disabled.
20719 @item -mno-sched-prefer-non-control-spec-insns
20720 @itemx -msched-prefer-non-control-spec-insns
20721 @opindex mno-sched-prefer-non-control-spec-insns
20722 @opindex msched-prefer-non-control-spec-insns
20723 If enabled, control-speculative instructions are chosen for schedule
20724 only if there are no other choices at the moment. This makes
20725 the use of the control speculation much more conservative.
20726 The default setting is disabled.
20728 @item -mno-sched-count-spec-in-critical-path
20729 @itemx -msched-count-spec-in-critical-path
20730 @opindex mno-sched-count-spec-in-critical-path
20731 @opindex msched-count-spec-in-critical-path
20732 If enabled, speculative dependencies are considered during
20733 computation of the instructions priorities. This makes the use of the
20734 speculation a bit more conservative.
20735 The default setting is disabled.
20737 @item -msched-spec-ldc
20738 @opindex msched-spec-ldc
20739 Use a simple data speculation check. This option is on by default.
20741 @item -msched-control-spec-ldc
20742 @opindex msched-spec-ldc
20743 Use a simple check for control speculation. This option is on by default.
20745 @item -msched-stop-bits-after-every-cycle
20746 @opindex msched-stop-bits-after-every-cycle
20747 Place a stop bit after every cycle when scheduling. This option is on
20750 @item -msched-fp-mem-deps-zero-cost
20751 @opindex msched-fp-mem-deps-zero-cost
20752 Assume that floating-point stores and loads are not likely to cause a conflict
20753 when placed into the same instruction group. This option is disabled by
20756 @item -msel-sched-dont-check-control-spec
20757 @opindex msel-sched-dont-check-control-spec
20758 Generate checks for control speculation in selective scheduling.
20759 This flag is disabled by default.
20761 @item -msched-max-memory-insns=@var{max-insns}
20762 @opindex msched-max-memory-insns
20763 Limit on the number of memory insns per instruction group, giving lower
20764 priority to subsequent memory insns attempting to schedule in the same
20765 instruction group. Frequently useful to prevent cache bank conflicts.
20766 The default value is 1.
20768 @item -msched-max-memory-insns-hard-limit
20769 @opindex msched-max-memory-insns-hard-limit
20770 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
20771 disallowing more than that number in an instruction group.
20772 Otherwise, the limit is ``soft'', meaning that non-memory operations
20773 are preferred when the limit is reached, but memory operations may still
20779 @subsection LM32 Options
20780 @cindex LM32 options
20782 These @option{-m} options are defined for the LatticeMico32 architecture:
20785 @item -mbarrel-shift-enabled
20786 @opindex mbarrel-shift-enabled
20787 Enable barrel-shift instructions.
20789 @item -mdivide-enabled
20790 @opindex mdivide-enabled
20791 Enable divide and modulus instructions.
20793 @item -mmultiply-enabled
20794 @opindex multiply-enabled
20795 Enable multiply instructions.
20797 @item -msign-extend-enabled
20798 @opindex msign-extend-enabled
20799 Enable sign extend instructions.
20801 @item -muser-enabled
20802 @opindex muser-enabled
20803 Enable user-defined instructions.
20808 @subsection M32C Options
20809 @cindex M32C options
20812 @item -mcpu=@var{name}
20814 Select the CPU for which code is generated. @var{name} may be one of
20815 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
20816 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
20817 the M32C/80 series.
20821 Specifies that the program will be run on the simulator. This causes
20822 an alternate runtime library to be linked in which supports, for
20823 example, file I/O@. You must not use this option when generating
20824 programs that will run on real hardware; you must provide your own
20825 runtime library for whatever I/O functions are needed.
20827 @item -memregs=@var{number}
20829 Specifies the number of memory-based pseudo-registers GCC uses
20830 during code generation. These pseudo-registers are used like real
20831 registers, so there is a tradeoff between GCC's ability to fit the
20832 code into available registers, and the performance penalty of using
20833 memory instead of registers. Note that all modules in a program must
20834 be compiled with the same value for this option. Because of that, you
20835 must not use this option with GCC's default runtime libraries.
20839 @node M32R/D Options
20840 @subsection M32R/D Options
20841 @cindex M32R/D options
20843 These @option{-m} options are defined for Renesas M32R/D architectures:
20848 Generate code for the M32R/2@.
20852 Generate code for the M32R/X@.
20856 Generate code for the M32R@. This is the default.
20858 @item -mmodel=small
20859 @opindex mmodel=small
20860 Assume all objects live in the lower 16MB of memory (so that their addresses
20861 can be loaded with the @code{ld24} instruction), and assume all subroutines
20862 are reachable with the @code{bl} instruction.
20863 This is the default.
20865 The addressability of a particular object can be set with the
20866 @code{model} attribute.
20868 @item -mmodel=medium
20869 @opindex mmodel=medium
20870 Assume objects may be anywhere in the 32-bit address space (the compiler
20871 generates @code{seth/add3} instructions to load their addresses), and
20872 assume all subroutines are reachable with the @code{bl} instruction.
20874 @item -mmodel=large
20875 @opindex mmodel=large
20876 Assume objects may be anywhere in the 32-bit address space (the compiler
20877 generates @code{seth/add3} instructions to load their addresses), and
20878 assume subroutines may not be reachable with the @code{bl} instruction
20879 (the compiler generates the much slower @code{seth/add3/jl}
20880 instruction sequence).
20883 @opindex msdata=none
20884 Disable use of the small data area. Variables are put into
20885 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
20886 @code{section} attribute has been specified).
20887 This is the default.
20889 The small data area consists of sections @code{.sdata} and @code{.sbss}.
20890 Objects may be explicitly put in the small data area with the
20891 @code{section} attribute using one of these sections.
20893 @item -msdata=sdata
20894 @opindex msdata=sdata
20895 Put small global and static data in the small data area, but do not
20896 generate special code to reference them.
20899 @opindex msdata=use
20900 Put small global and static data in the small data area, and generate
20901 special instructions to reference them.
20905 @cindex smaller data references
20906 Put global and static objects less than or equal to @var{num} bytes
20907 into the small data or BSS sections instead of the normal data or BSS
20908 sections. The default value of @var{num} is 8.
20909 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
20910 for this option to have any effect.
20912 All modules should be compiled with the same @option{-G @var{num}} value.
20913 Compiling with different values of @var{num} may or may not work; if it
20914 doesn't the linker gives an error message---incorrect code is not
20919 Makes the M32R-specific code in the compiler display some statistics
20920 that might help in debugging programs.
20922 @item -malign-loops
20923 @opindex malign-loops
20924 Align all loops to a 32-byte boundary.
20926 @item -mno-align-loops
20927 @opindex mno-align-loops
20928 Do not enforce a 32-byte alignment for loops. This is the default.
20930 @item -missue-rate=@var{number}
20931 @opindex missue-rate=@var{number}
20932 Issue @var{number} instructions per cycle. @var{number} can only be 1
20935 @item -mbranch-cost=@var{number}
20936 @opindex mbranch-cost=@var{number}
20937 @var{number} can only be 1 or 2. If it is 1 then branches are
20938 preferred over conditional code, if it is 2, then the opposite applies.
20940 @item -mflush-trap=@var{number}
20941 @opindex mflush-trap=@var{number}
20942 Specifies the trap number to use to flush the cache. The default is
20943 12. Valid numbers are between 0 and 15 inclusive.
20945 @item -mno-flush-trap
20946 @opindex mno-flush-trap
20947 Specifies that the cache cannot be flushed by using a trap.
20949 @item -mflush-func=@var{name}
20950 @opindex mflush-func=@var{name}
20951 Specifies the name of the operating system function to call to flush
20952 the cache. The default is @samp{_flush_cache}, but a function call
20953 is only used if a trap is not available.
20955 @item -mno-flush-func
20956 @opindex mno-flush-func
20957 Indicates that there is no OS function for flushing the cache.
20961 @node M680x0 Options
20962 @subsection M680x0 Options
20963 @cindex M680x0 options
20965 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
20966 The default settings depend on which architecture was selected when
20967 the compiler was configured; the defaults for the most common choices
20971 @item -march=@var{arch}
20973 Generate code for a specific M680x0 or ColdFire instruction set
20974 architecture. Permissible values of @var{arch} for M680x0
20975 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
20976 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
20977 architectures are selected according to Freescale's ISA classification
20978 and the permissible values are: @samp{isaa}, @samp{isaaplus},
20979 @samp{isab} and @samp{isac}.
20981 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
20982 code for a ColdFire target. The @var{arch} in this macro is one of the
20983 @option{-march} arguments given above.
20985 When used together, @option{-march} and @option{-mtune} select code
20986 that runs on a family of similar processors but that is optimized
20987 for a particular microarchitecture.
20989 @item -mcpu=@var{cpu}
20991 Generate code for a specific M680x0 or ColdFire processor.
20992 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
20993 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
20994 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
20995 below, which also classifies the CPUs into families:
20997 @multitable @columnfractions 0.20 0.80
20998 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
20999 @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}
21000 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
21001 @item @samp{5206e} @tab @samp{5206e}
21002 @item @samp{5208} @tab @samp{5207} @samp{5208}
21003 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
21004 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
21005 @item @samp{5216} @tab @samp{5214} @samp{5216}
21006 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
21007 @item @samp{5225} @tab @samp{5224} @samp{5225}
21008 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
21009 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
21010 @item @samp{5249} @tab @samp{5249}
21011 @item @samp{5250} @tab @samp{5250}
21012 @item @samp{5271} @tab @samp{5270} @samp{5271}
21013 @item @samp{5272} @tab @samp{5272}
21014 @item @samp{5275} @tab @samp{5274} @samp{5275}
21015 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
21016 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
21017 @item @samp{5307} @tab @samp{5307}
21018 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
21019 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
21020 @item @samp{5407} @tab @samp{5407}
21021 @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}
21024 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
21025 @var{arch} is compatible with @var{cpu}. Other combinations of
21026 @option{-mcpu} and @option{-march} are rejected.
21028 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
21029 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
21030 where the value of @var{family} is given by the table above.
21032 @item -mtune=@var{tune}
21034 Tune the code for a particular microarchitecture within the
21035 constraints set by @option{-march} and @option{-mcpu}.
21036 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
21037 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
21038 and @samp{cpu32}. The ColdFire microarchitectures
21039 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
21041 You can also use @option{-mtune=68020-40} for code that needs
21042 to run relatively well on 68020, 68030 and 68040 targets.
21043 @option{-mtune=68020-60} is similar but includes 68060 targets
21044 as well. These two options select the same tuning decisions as
21045 @option{-m68020-40} and @option{-m68020-60} respectively.
21047 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
21048 when tuning for 680x0 architecture @var{arch}. It also defines
21049 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
21050 option is used. If GCC is tuning for a range of architectures,
21051 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
21052 it defines the macros for every architecture in the range.
21054 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
21055 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
21056 of the arguments given above.
21062 Generate output for a 68000. This is the default
21063 when the compiler is configured for 68000-based systems.
21064 It is equivalent to @option{-march=68000}.
21066 Use this option for microcontrollers with a 68000 or EC000 core,
21067 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
21071 Generate output for a 68010. This is the default
21072 when the compiler is configured for 68010-based systems.
21073 It is equivalent to @option{-march=68010}.
21079 Generate output for a 68020. This is the default
21080 when the compiler is configured for 68020-based systems.
21081 It is equivalent to @option{-march=68020}.
21085 Generate output for a 68030. This is the default when the compiler is
21086 configured for 68030-based systems. It is equivalent to
21087 @option{-march=68030}.
21091 Generate output for a 68040. This is the default when the compiler is
21092 configured for 68040-based systems. It is equivalent to
21093 @option{-march=68040}.
21095 This option inhibits the use of 68881/68882 instructions that have to be
21096 emulated by software on the 68040. Use this option if your 68040 does not
21097 have code to emulate those instructions.
21101 Generate output for a 68060. This is the default when the compiler is
21102 configured for 68060-based systems. It is equivalent to
21103 @option{-march=68060}.
21105 This option inhibits the use of 68020 and 68881/68882 instructions that
21106 have to be emulated by software on the 68060. Use this option if your 68060
21107 does not have code to emulate those instructions.
21111 Generate output for a CPU32. This is the default
21112 when the compiler is configured for CPU32-based systems.
21113 It is equivalent to @option{-march=cpu32}.
21115 Use this option for microcontrollers with a
21116 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
21117 68336, 68340, 68341, 68349 and 68360.
21121 Generate output for a 520X ColdFire CPU@. This is the default
21122 when the compiler is configured for 520X-based systems.
21123 It is equivalent to @option{-mcpu=5206}, and is now deprecated
21124 in favor of that option.
21126 Use this option for microcontroller with a 5200 core, including
21127 the MCF5202, MCF5203, MCF5204 and MCF5206.
21131 Generate output for a 5206e ColdFire CPU@. The option is now
21132 deprecated in favor of the equivalent @option{-mcpu=5206e}.
21136 Generate output for a member of the ColdFire 528X family.
21137 The option is now deprecated in favor of the equivalent
21138 @option{-mcpu=528x}.
21142 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
21143 in favor of the equivalent @option{-mcpu=5307}.
21147 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
21148 in favor of the equivalent @option{-mcpu=5407}.
21152 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
21153 This includes use of hardware floating-point instructions.
21154 The option is equivalent to @option{-mcpu=547x}, and is now
21155 deprecated in favor of that option.
21159 Generate output for a 68040, without using any of the new instructions.
21160 This results in code that can run relatively efficiently on either a
21161 68020/68881 or a 68030 or a 68040. The generated code does use the
21162 68881 instructions that are emulated on the 68040.
21164 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
21168 Generate output for a 68060, without using any of the new instructions.
21169 This results in code that can run relatively efficiently on either a
21170 68020/68881 or a 68030 or a 68040. The generated code does use the
21171 68881 instructions that are emulated on the 68060.
21173 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
21177 @opindex mhard-float
21179 Generate floating-point instructions. This is the default for 68020
21180 and above, and for ColdFire devices that have an FPU@. It defines the
21181 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
21182 on ColdFire targets.
21185 @opindex msoft-float
21186 Do not generate floating-point instructions; use library calls instead.
21187 This is the default for 68000, 68010, and 68832 targets. It is also
21188 the default for ColdFire devices that have no FPU.
21194 Generate (do not generate) ColdFire hardware divide and remainder
21195 instructions. If @option{-march} is used without @option{-mcpu},
21196 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
21197 architectures. Otherwise, the default is taken from the target CPU
21198 (either the default CPU, or the one specified by @option{-mcpu}). For
21199 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
21200 @option{-mcpu=5206e}.
21202 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
21206 Consider type @code{int} to be 16 bits wide, like @code{short int}.
21207 Additionally, parameters passed on the stack are also aligned to a
21208 16-bit boundary even on targets whose API mandates promotion to 32-bit.
21212 Do not consider type @code{int} to be 16 bits wide. This is the default.
21215 @itemx -mno-bitfield
21216 @opindex mnobitfield
21217 @opindex mno-bitfield
21218 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
21219 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
21223 Do use the bit-field instructions. The @option{-m68020} option implies
21224 @option{-mbitfield}. This is the default if you use a configuration
21225 designed for a 68020.
21229 Use a different function-calling convention, in which functions
21230 that take a fixed number of arguments return with the @code{rtd}
21231 instruction, which pops their arguments while returning. This
21232 saves one instruction in the caller since there is no need to pop
21233 the arguments there.
21235 This calling convention is incompatible with the one normally
21236 used on Unix, so you cannot use it if you need to call libraries
21237 compiled with the Unix compiler.
21239 Also, you must provide function prototypes for all functions that
21240 take variable numbers of arguments (including @code{printf});
21241 otherwise incorrect code is generated for calls to those
21244 In addition, seriously incorrect code results if you call a
21245 function with too many arguments. (Normally, extra arguments are
21246 harmlessly ignored.)
21248 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
21249 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
21251 The default is @option{-mno-rtd}.
21254 @itemx -mno-align-int
21255 @opindex malign-int
21256 @opindex mno-align-int
21257 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
21258 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
21259 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
21260 Aligning variables on 32-bit boundaries produces code that runs somewhat
21261 faster on processors with 32-bit busses at the expense of more memory.
21263 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
21264 aligns structures containing the above types differently than
21265 most published application binary interface specifications for the m68k.
21269 Use the pc-relative addressing mode of the 68000 directly, instead of
21270 using a global offset table. At present, this option implies @option{-fpic},
21271 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
21272 not presently supported with @option{-mpcrel}, though this could be supported for
21273 68020 and higher processors.
21275 @item -mno-strict-align
21276 @itemx -mstrict-align
21277 @opindex mno-strict-align
21278 @opindex mstrict-align
21279 Do not (do) assume that unaligned memory references are handled by
21283 Generate code that allows the data segment to be located in a different
21284 area of memory from the text segment. This allows for execute-in-place in
21285 an environment without virtual memory management. This option implies
21288 @item -mno-sep-data
21289 Generate code that assumes that the data segment follows the text segment.
21290 This is the default.
21292 @item -mid-shared-library
21293 Generate code that supports shared libraries via the library ID method.
21294 This allows for execute-in-place and shared libraries in an environment
21295 without virtual memory management. This option implies @option{-fPIC}.
21297 @item -mno-id-shared-library
21298 Generate code that doesn't assume ID-based shared libraries are being used.
21299 This is the default.
21301 @item -mshared-library-id=n
21302 Specifies the identification number of the ID-based shared library being
21303 compiled. Specifying a value of 0 generates more compact code; specifying
21304 other values forces the allocation of that number to the current
21305 library, but is no more space- or time-efficient than omitting this option.
21311 When generating position-independent code for ColdFire, generate code
21312 that works if the GOT has more than 8192 entries. This code is
21313 larger and slower than code generated without this option. On M680x0
21314 processors, this option is not needed; @option{-fPIC} suffices.
21316 GCC normally uses a single instruction to load values from the GOT@.
21317 While this is relatively efficient, it only works if the GOT
21318 is smaller than about 64k. Anything larger causes the linker
21319 to report an error such as:
21321 @cindex relocation truncated to fit (ColdFire)
21323 relocation truncated to fit: R_68K_GOT16O foobar
21326 If this happens, you should recompile your code with @option{-mxgot}.
21327 It should then work with very large GOTs. However, code generated with
21328 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
21329 the value of a global symbol.
21331 Note that some linkers, including newer versions of the GNU linker,
21332 can create multiple GOTs and sort GOT entries. If you have such a linker,
21333 you should only need to use @option{-mxgot} when compiling a single
21334 object file that accesses more than 8192 GOT entries. Very few do.
21336 These options have no effect unless GCC is generating
21337 position-independent code.
21339 @item -mlong-jump-table-offsets
21340 @opindex mlong-jump-table-offsets
21341 Use 32-bit offsets in @code{switch} tables. The default is to use
21346 @node MCore Options
21347 @subsection MCore Options
21348 @cindex MCore options
21350 These are the @samp{-m} options defined for the Motorola M*Core
21356 @itemx -mno-hardlit
21358 @opindex mno-hardlit
21359 Inline constants into the code stream if it can be done in two
21360 instructions or less.
21366 Use the divide instruction. (Enabled by default).
21368 @item -mrelax-immediate
21369 @itemx -mno-relax-immediate
21370 @opindex mrelax-immediate
21371 @opindex mno-relax-immediate
21372 Allow arbitrary-sized immediates in bit operations.
21374 @item -mwide-bitfields
21375 @itemx -mno-wide-bitfields
21376 @opindex mwide-bitfields
21377 @opindex mno-wide-bitfields
21378 Always treat bit-fields as @code{int}-sized.
21380 @item -m4byte-functions
21381 @itemx -mno-4byte-functions
21382 @opindex m4byte-functions
21383 @opindex mno-4byte-functions
21384 Force all functions to be aligned to a 4-byte boundary.
21386 @item -mcallgraph-data
21387 @itemx -mno-callgraph-data
21388 @opindex mcallgraph-data
21389 @opindex mno-callgraph-data
21390 Emit callgraph information.
21393 @itemx -mno-slow-bytes
21394 @opindex mslow-bytes
21395 @opindex mno-slow-bytes
21396 Prefer word access when reading byte quantities.
21398 @item -mlittle-endian
21399 @itemx -mbig-endian
21400 @opindex mlittle-endian
21401 @opindex mbig-endian
21402 Generate code for a little-endian target.
21408 Generate code for the 210 processor.
21412 Assume that runtime support has been provided and so omit the
21413 simulator library (@file{libsim.a)} from the linker command line.
21415 @item -mstack-increment=@var{size}
21416 @opindex mstack-increment
21417 Set the maximum amount for a single stack increment operation. Large
21418 values can increase the speed of programs that contain functions
21419 that need a large amount of stack space, but they can also trigger a
21420 segmentation fault if the stack is extended too much. The default
21426 @subsection MeP Options
21427 @cindex MeP options
21433 Enables the @code{abs} instruction, which is the absolute difference
21434 between two registers.
21438 Enables all the optional instructions---average, multiply, divide, bit
21439 operations, leading zero, absolute difference, min/max, clip, and
21445 Enables the @code{ave} instruction, which computes the average of two
21448 @item -mbased=@var{n}
21450 Variables of size @var{n} bytes or smaller are placed in the
21451 @code{.based} section by default. Based variables use the @code{$tp}
21452 register as a base register, and there is a 128-byte limit to the
21453 @code{.based} section.
21457 Enables the bit operation instructions---bit test (@code{btstm}), set
21458 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
21459 test-and-set (@code{tas}).
21461 @item -mc=@var{name}
21463 Selects which section constant data is placed in. @var{name} may
21464 be @samp{tiny}, @samp{near}, or @samp{far}.
21468 Enables the @code{clip} instruction. Note that @option{-mclip} is not
21469 useful unless you also provide @option{-mminmax}.
21471 @item -mconfig=@var{name}
21473 Selects one of the built-in core configurations. Each MeP chip has
21474 one or more modules in it; each module has a core CPU and a variety of
21475 coprocessors, optional instructions, and peripherals. The
21476 @code{MeP-Integrator} tool, not part of GCC, provides these
21477 configurations through this option; using this option is the same as
21478 using all the corresponding command-line options. The default
21479 configuration is @samp{default}.
21483 Enables the coprocessor instructions. By default, this is a 32-bit
21484 coprocessor. Note that the coprocessor is normally enabled via the
21485 @option{-mconfig=} option.
21489 Enables the 32-bit coprocessor's instructions.
21493 Enables the 64-bit coprocessor's instructions.
21497 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
21501 Causes constant variables to be placed in the @code{.near} section.
21505 Enables the @code{div} and @code{divu} instructions.
21509 Generate big-endian code.
21513 Generate little-endian code.
21515 @item -mio-volatile
21516 @opindex mio-volatile
21517 Tells the compiler that any variable marked with the @code{io}
21518 attribute is to be considered volatile.
21522 Causes variables to be assigned to the @code{.far} section by default.
21526 Enables the @code{leadz} (leading zero) instruction.
21530 Causes variables to be assigned to the @code{.near} section by default.
21534 Enables the @code{min} and @code{max} instructions.
21538 Enables the multiplication and multiply-accumulate instructions.
21542 Disables all the optional instructions enabled by @option{-mall-opts}.
21546 Enables the @code{repeat} and @code{erepeat} instructions, used for
21547 low-overhead looping.
21551 Causes all variables to default to the @code{.tiny} section. Note
21552 that there is a 65536-byte limit to this section. Accesses to these
21553 variables use the @code{%gp} base register.
21557 Enables the saturation instructions. Note that the compiler does not
21558 currently generate these itself, but this option is included for
21559 compatibility with other tools, like @code{as}.
21563 Link the SDRAM-based runtime instead of the default ROM-based runtime.
21567 Link the simulator run-time libraries.
21571 Link the simulator runtime libraries, excluding built-in support
21572 for reset and exception vectors and tables.
21576 Causes all functions to default to the @code{.far} section. Without
21577 this option, functions default to the @code{.near} section.
21579 @item -mtiny=@var{n}
21581 Variables that are @var{n} bytes or smaller are allocated to the
21582 @code{.tiny} section. These variables use the @code{$gp} base
21583 register. The default for this option is 4, but note that there's a
21584 65536-byte limit to the @code{.tiny} section.
21588 @node MicroBlaze Options
21589 @subsection MicroBlaze Options
21590 @cindex MicroBlaze Options
21595 @opindex msoft-float
21596 Use software emulation for floating point (default).
21599 @opindex mhard-float
21600 Use hardware floating-point instructions.
21604 Do not optimize block moves, use @code{memcpy}.
21606 @item -mno-clearbss
21607 @opindex mno-clearbss
21608 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
21610 @item -mcpu=@var{cpu-type}
21612 Use features of, and schedule code for, the given CPU.
21613 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
21614 where @var{X} is a major version, @var{YY} is the minor version, and
21615 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
21616 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v6.00.a}.
21618 @item -mxl-soft-mul
21619 @opindex mxl-soft-mul
21620 Use software multiply emulation (default).
21622 @item -mxl-soft-div
21623 @opindex mxl-soft-div
21624 Use software emulation for divides (default).
21626 @item -mxl-barrel-shift
21627 @opindex mxl-barrel-shift
21628 Use the hardware barrel shifter.
21630 @item -mxl-pattern-compare
21631 @opindex mxl-pattern-compare
21632 Use pattern compare instructions.
21634 @item -msmall-divides
21635 @opindex msmall-divides
21636 Use table lookup optimization for small signed integer divisions.
21638 @item -mxl-stack-check
21639 @opindex mxl-stack-check
21640 This option is deprecated. Use @option{-fstack-check} instead.
21643 @opindex mxl-gp-opt
21644 Use GP-relative @code{.sdata}/@code{.sbss} sections.
21646 @item -mxl-multiply-high
21647 @opindex mxl-multiply-high
21648 Use multiply high instructions for high part of 32x32 multiply.
21650 @item -mxl-float-convert
21651 @opindex mxl-float-convert
21652 Use hardware floating-point conversion instructions.
21654 @item -mxl-float-sqrt
21655 @opindex mxl-float-sqrt
21656 Use hardware floating-point square root instruction.
21659 @opindex mbig-endian
21660 Generate code for a big-endian target.
21662 @item -mlittle-endian
21663 @opindex mlittle-endian
21664 Generate code for a little-endian target.
21667 @opindex mxl-reorder
21668 Use reorder instructions (swap and byte reversed load/store).
21670 @item -mxl-mode-@var{app-model}
21671 Select application model @var{app-model}. Valid models are
21674 normal executable (default), uses startup code @file{crt0.o}.
21676 @item -mpic-data-is-text-relative
21677 @opindex mpic-data-is-text-relative
21678 Assume that the displacement between the text and data segments is fixed
21679 at static link time. This allows data to be referenced by offset from start of
21680 text address instead of GOT since PC-relative addressing is not supported.
21683 for use with Xilinx Microprocessor Debugger (XMD) based
21684 software intrusive debug agent called xmdstub. This uses startup file
21685 @file{crt1.o} and sets the start address of the program to 0x800.
21688 for applications that are loaded using a bootloader.
21689 This model uses startup file @file{crt2.o} which does not contain a processor
21690 reset vector handler. This is suitable for transferring control on a
21691 processor reset to the bootloader rather than the application.
21694 for applications that do not require any of the
21695 MicroBlaze vectors. This option may be useful for applications running
21696 within a monitoring application. This model uses @file{crt3.o} as a startup file.
21699 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
21700 @option{-mxl-mode-@var{app-model}}.
21705 @subsection MIPS Options
21706 @cindex MIPS options
21712 Generate big-endian code.
21716 Generate little-endian code. This is the default for @samp{mips*el-*-*}
21719 @item -march=@var{arch}
21721 Generate code that runs on @var{arch}, which can be the name of a
21722 generic MIPS ISA, or the name of a particular processor.
21724 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
21725 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
21726 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
21727 @samp{mips64r5} and @samp{mips64r6}.
21728 The processor names are:
21729 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
21730 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
21731 @samp{5kc}, @samp{5kf},
21733 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
21734 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
21735 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
21736 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
21737 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
21738 @samp{i6400}, @samp{i6500},
21740 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a}, @samp{gs464},
21741 @samp{gs464e}, @samp{gs264e},
21743 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
21744 @samp{m5100}, @samp{m5101},
21745 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
21747 @samp{p5600}, @samp{p6600},
21748 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
21749 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r5900},
21750 @samp{r6000}, @samp{r8000},
21751 @samp{rm7000}, @samp{rm9000},
21752 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
21755 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
21756 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
21757 @samp{xlr} and @samp{xlp}.
21758 The special value @samp{from-abi} selects the
21759 most compatible architecture for the selected ABI (that is,
21760 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
21762 The native Linux/GNU toolchain also supports the value @samp{native},
21763 which selects the best architecture option for the host processor.
21764 @option{-march=native} has no effect if GCC does not recognize
21767 In processor names, a final @samp{000} can be abbreviated as @samp{k}
21768 (for example, @option{-march=r2k}). Prefixes are optional, and
21769 @samp{vr} may be written @samp{r}.
21771 Names of the form @samp{@var{n}f2_1} refer to processors with
21772 FPUs clocked at half the rate of the core, names of the form
21773 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
21774 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
21775 processors with FPUs clocked a ratio of 3:2 with respect to the core.
21776 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
21777 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
21778 accepted as synonyms for @samp{@var{n}f1_1}.
21780 GCC defines two macros based on the value of this option. The first
21781 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
21782 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
21783 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
21784 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
21785 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
21787 Note that the @code{_MIPS_ARCH} macro uses the processor names given
21788 above. In other words, it has the full prefix and does not
21789 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
21790 the macro names the resolved architecture (either @code{"mips1"} or
21791 @code{"mips3"}). It names the default architecture when no
21792 @option{-march} option is given.
21794 @item -mtune=@var{arch}
21796 Optimize for @var{arch}. Among other things, this option controls
21797 the way instructions are scheduled, and the perceived cost of arithmetic
21798 operations. The list of @var{arch} values is the same as for
21801 When this option is not used, GCC optimizes for the processor
21802 specified by @option{-march}. By using @option{-march} and
21803 @option{-mtune} together, it is possible to generate code that
21804 runs on a family of processors, but optimize the code for one
21805 particular member of that family.
21807 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
21808 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
21809 @option{-march} ones described above.
21813 Equivalent to @option{-march=mips1}.
21817 Equivalent to @option{-march=mips2}.
21821 Equivalent to @option{-march=mips3}.
21825 Equivalent to @option{-march=mips4}.
21829 Equivalent to @option{-march=mips32}.
21833 Equivalent to @option{-march=mips32r3}.
21837 Equivalent to @option{-march=mips32r5}.
21841 Equivalent to @option{-march=mips32r6}.
21845 Equivalent to @option{-march=mips64}.
21849 Equivalent to @option{-march=mips64r2}.
21853 Equivalent to @option{-march=mips64r3}.
21857 Equivalent to @option{-march=mips64r5}.
21861 Equivalent to @option{-march=mips64r6}.
21866 @opindex mno-mips16
21867 Generate (do not generate) MIPS16 code. If GCC is targeting a
21868 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
21870 MIPS16 code generation can also be controlled on a per-function basis
21871 by means of @code{mips16} and @code{nomips16} attributes.
21872 @xref{Function Attributes}, for more information.
21874 @item -mflip-mips16
21875 @opindex mflip-mips16
21876 Generate MIPS16 code on alternating functions. This option is provided
21877 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
21878 not intended for ordinary use in compiling user code.
21880 @item -minterlink-compressed
21881 @itemx -mno-interlink-compressed
21882 @opindex minterlink-compressed
21883 @opindex mno-interlink-compressed
21884 Require (do not require) that code using the standard (uncompressed) MIPS ISA
21885 be link-compatible with MIPS16 and microMIPS code, and vice versa.
21887 For example, code using the standard ISA encoding cannot jump directly
21888 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
21889 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
21890 knows that the target of the jump is not compressed.
21892 @item -minterlink-mips16
21893 @itemx -mno-interlink-mips16
21894 @opindex minterlink-mips16
21895 @opindex mno-interlink-mips16
21896 Aliases of @option{-minterlink-compressed} and
21897 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
21898 and are retained for backwards compatibility.
21910 Generate code for the given ABI@.
21912 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
21913 generates 64-bit code when you select a 64-bit architecture, but you
21914 can use @option{-mgp32} to get 32-bit code instead.
21916 For information about the O64 ABI, see
21917 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
21919 GCC supports a variant of the o32 ABI in which floating-point registers
21920 are 64 rather than 32 bits wide. You can select this combination with
21921 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
21922 and @code{mfhc1} instructions and is therefore only supported for
21923 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
21925 The register assignments for arguments and return values remain the
21926 same, but each scalar value is passed in a single 64-bit register
21927 rather than a pair of 32-bit registers. For example, scalar
21928 floating-point values are returned in @samp{$f0} only, not a
21929 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
21930 remains the same in that the even-numbered double-precision registers
21933 Two additional variants of the o32 ABI are supported to enable
21934 a transition from 32-bit to 64-bit registers. These are FPXX
21935 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
21936 The FPXX extension mandates that all code must execute correctly
21937 when run using 32-bit or 64-bit registers. The code can be interlinked
21938 with either FP32 or FP64, but not both.
21939 The FP64A extension is similar to the FP64 extension but forbids the
21940 use of odd-numbered single-precision registers. This can be used
21941 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
21942 processors and allows both FP32 and FP64A code to interlink and
21943 run in the same process without changing FPU modes.
21946 @itemx -mno-abicalls
21948 @opindex mno-abicalls
21949 Generate (do not generate) code that is suitable for SVR4-style
21950 dynamic objects. @option{-mabicalls} is the default for SVR4-based
21955 Generate (do not generate) code that is fully position-independent,
21956 and that can therefore be linked into shared libraries. This option
21957 only affects @option{-mabicalls}.
21959 All @option{-mabicalls} code has traditionally been position-independent,
21960 regardless of options like @option{-fPIC} and @option{-fpic}. However,
21961 as an extension, the GNU toolchain allows executables to use absolute
21962 accesses for locally-binding symbols. It can also use shorter GP
21963 initialization sequences and generate direct calls to locally-defined
21964 functions. This mode is selected by @option{-mno-shared}.
21966 @option{-mno-shared} depends on binutils 2.16 or higher and generates
21967 objects that can only be linked by the GNU linker. However, the option
21968 does not affect the ABI of the final executable; it only affects the ABI
21969 of relocatable objects. Using @option{-mno-shared} generally makes
21970 executables both smaller and quicker.
21972 @option{-mshared} is the default.
21978 Assume (do not assume) that the static and dynamic linkers
21979 support PLTs and copy relocations. This option only affects
21980 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
21981 has no effect without @option{-msym32}.
21983 You can make @option{-mplt} the default by configuring
21984 GCC with @option{--with-mips-plt}. The default is
21985 @option{-mno-plt} otherwise.
21991 Lift (do not lift) the usual restrictions on the size of the global
21994 GCC normally uses a single instruction to load values from the GOT@.
21995 While this is relatively efficient, it only works if the GOT
21996 is smaller than about 64k. Anything larger causes the linker
21997 to report an error such as:
21999 @cindex relocation truncated to fit (MIPS)
22001 relocation truncated to fit: R_MIPS_GOT16 foobar
22004 If this happens, you should recompile your code with @option{-mxgot}.
22005 This works with very large GOTs, although the code is also
22006 less efficient, since it takes three instructions to fetch the
22007 value of a global symbol.
22009 Note that some linkers can create multiple GOTs. If you have such a
22010 linker, you should only need to use @option{-mxgot} when a single object
22011 file accesses more than 64k's worth of GOT entries. Very few do.
22013 These options have no effect unless GCC is generating position
22018 Assume that general-purpose registers are 32 bits wide.
22022 Assume that general-purpose registers are 64 bits wide.
22026 Assume that floating-point registers are 32 bits wide.
22030 Assume that floating-point registers are 64 bits wide.
22034 Do not assume the width of floating-point registers.
22037 @opindex mhard-float
22038 Use floating-point coprocessor instructions.
22041 @opindex msoft-float
22042 Do not use floating-point coprocessor instructions. Implement
22043 floating-point calculations using library calls instead.
22047 Equivalent to @option{-msoft-float}, but additionally asserts that the
22048 program being compiled does not perform any floating-point operations.
22049 This option is presently supported only by some bare-metal MIPS
22050 configurations, where it may select a special set of libraries
22051 that lack all floating-point support (including, for example, the
22052 floating-point @code{printf} formats).
22053 If code compiled with @option{-mno-float} accidentally contains
22054 floating-point operations, it is likely to suffer a link-time
22055 or run-time failure.
22057 @item -msingle-float
22058 @opindex msingle-float
22059 Assume that the floating-point coprocessor only supports single-precision
22062 @item -mdouble-float
22063 @opindex mdouble-float
22064 Assume that the floating-point coprocessor supports double-precision
22065 operations. This is the default.
22068 @itemx -mno-odd-spreg
22069 @opindex modd-spreg
22070 @opindex mno-odd-spreg
22071 Enable the use of odd-numbered single-precision floating-point registers
22072 for the o32 ABI. This is the default for processors that are known to
22073 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
22077 @itemx -mabs=legacy
22079 @opindex mabs=legacy
22080 These options control the treatment of the special not-a-number (NaN)
22081 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
22082 @code{neg.@i{fmt}} machine instructions.
22084 By default or when @option{-mabs=legacy} is used the legacy
22085 treatment is selected. In this case these instructions are considered
22086 arithmetic and avoided where correct operation is required and the
22087 input operand might be a NaN. A longer sequence of instructions that
22088 manipulate the sign bit of floating-point datum manually is used
22089 instead unless the @option{-ffinite-math-only} option has also been
22092 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
22093 this case these instructions are considered non-arithmetic and therefore
22094 operating correctly in all cases, including in particular where the
22095 input operand is a NaN. These instructions are therefore always used
22096 for the respective operations.
22099 @itemx -mnan=legacy
22101 @opindex mnan=legacy
22102 These options control the encoding of the special not-a-number (NaN)
22103 IEEE 754 floating-point data.
22105 The @option{-mnan=legacy} option selects the legacy encoding. In this
22106 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
22107 significand field being 0, whereas signaling NaNs (sNaNs) are denoted
22108 by the first bit of their trailing significand field being 1.
22110 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
22111 this case qNaNs are denoted by the first bit of their trailing
22112 significand field being 1, whereas sNaNs are denoted by the first bit of
22113 their trailing significand field being 0.
22115 The default is @option{-mnan=legacy} unless GCC has been configured with
22116 @option{--with-nan=2008}.
22122 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
22123 implement atomic memory built-in functions. When neither option is
22124 specified, GCC uses the instructions if the target architecture
22127 @option{-mllsc} is useful if the runtime environment can emulate the
22128 instructions and @option{-mno-llsc} can be useful when compiling for
22129 nonstandard ISAs. You can make either option the default by
22130 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
22131 respectively. @option{--with-llsc} is the default for some
22132 configurations; see the installation documentation for details.
22138 Use (do not use) revision 1 of the MIPS DSP ASE@.
22139 @xref{MIPS DSP Built-in Functions}. This option defines the
22140 preprocessor macro @code{__mips_dsp}. It also defines
22141 @code{__mips_dsp_rev} to 1.
22147 Use (do not use) revision 2 of the MIPS DSP ASE@.
22148 @xref{MIPS DSP Built-in Functions}. This option defines the
22149 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
22150 It also defines @code{__mips_dsp_rev} to 2.
22153 @itemx -mno-smartmips
22154 @opindex msmartmips
22155 @opindex mno-smartmips
22156 Use (do not use) the MIPS SmartMIPS ASE.
22158 @item -mpaired-single
22159 @itemx -mno-paired-single
22160 @opindex mpaired-single
22161 @opindex mno-paired-single
22162 Use (do not use) paired-single floating-point instructions.
22163 @xref{MIPS Paired-Single Support}. This option requires
22164 hardware floating-point support to be enabled.
22170 Use (do not use) MIPS Digital Media Extension instructions.
22171 This option can only be used when generating 64-bit code and requires
22172 hardware floating-point support to be enabled.
22177 @opindex mno-mips3d
22178 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
22179 The option @option{-mips3d} implies @option{-mpaired-single}.
22182 @itemx -mno-micromips
22183 @opindex mmicromips
22184 @opindex mno-mmicromips
22185 Generate (do not generate) microMIPS code.
22187 MicroMIPS code generation can also be controlled on a per-function basis
22188 by means of @code{micromips} and @code{nomicromips} attributes.
22189 @xref{Function Attributes}, for more information.
22195 Use (do not use) MT Multithreading instructions.
22201 Use (do not use) the MIPS MCU ASE instructions.
22207 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
22213 Use (do not use) the MIPS Virtualization (VZ) instructions.
22219 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
22225 Use (do not use) the MIPS Cyclic Redundancy Check (CRC) instructions.
22231 Use (do not use) the MIPS Global INValidate (GINV) instructions.
22233 @item -mloongson-mmi
22234 @itemx -mno-loongson-mmi
22235 @opindex mloongson-mmi
22236 @opindex mno-loongson-mmi
22237 Use (do not use) the MIPS Loongson MultiMedia extensions Instructions (MMI).
22239 @item -mloongson-ext
22240 @itemx -mno-loongson-ext
22241 @opindex mloongson-ext
22242 @opindex mno-loongson-ext
22243 Use (do not use) the MIPS Loongson EXTensions (EXT) instructions.
22245 @item -mloongson-ext2
22246 @itemx -mno-loongson-ext2
22247 @opindex mloongson-ext2
22248 @opindex mno-loongson-ext2
22249 Use (do not use) the MIPS Loongson EXTensions r2 (EXT2) instructions.
22253 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
22254 an explanation of the default and the way that the pointer size is
22259 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
22261 The default size of @code{int}s, @code{long}s and pointers depends on
22262 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
22263 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
22264 32-bit @code{long}s. Pointers are the same size as @code{long}s,
22265 or the same size as integer registers, whichever is smaller.
22271 Assume (do not assume) that all symbols have 32-bit values, regardless
22272 of the selected ABI@. This option is useful in combination with
22273 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
22274 to generate shorter and faster references to symbolic addresses.
22278 Put definitions of externally-visible data in a small data section
22279 if that data is no bigger than @var{num} bytes. GCC can then generate
22280 more efficient accesses to the data; see @option{-mgpopt} for details.
22282 The default @option{-G} option depends on the configuration.
22284 @item -mlocal-sdata
22285 @itemx -mno-local-sdata
22286 @opindex mlocal-sdata
22287 @opindex mno-local-sdata
22288 Extend (do not extend) the @option{-G} behavior to local data too,
22289 such as to static variables in C@. @option{-mlocal-sdata} is the
22290 default for all configurations.
22292 If the linker complains that an application is using too much small data,
22293 you might want to try rebuilding the less performance-critical parts with
22294 @option{-mno-local-sdata}. You might also want to build large
22295 libraries with @option{-mno-local-sdata}, so that the libraries leave
22296 more room for the main program.
22298 @item -mextern-sdata
22299 @itemx -mno-extern-sdata
22300 @opindex mextern-sdata
22301 @opindex mno-extern-sdata
22302 Assume (do not assume) that externally-defined data is in
22303 a small data section if the size of that data is within the @option{-G} limit.
22304 @option{-mextern-sdata} is the default for all configurations.
22306 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
22307 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
22308 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
22309 is placed in a small data section. If @var{Var} is defined by another
22310 module, you must either compile that module with a high-enough
22311 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
22312 definition. If @var{Var} is common, you must link the application
22313 with a high-enough @option{-G} setting.
22315 The easiest way of satisfying these restrictions is to compile
22316 and link every module with the same @option{-G} option. However,
22317 you may wish to build a library that supports several different
22318 small data limits. You can do this by compiling the library with
22319 the highest supported @option{-G} setting and additionally using
22320 @option{-mno-extern-sdata} to stop the library from making assumptions
22321 about externally-defined data.
22327 Use (do not use) GP-relative accesses for symbols that are known to be
22328 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
22329 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
22332 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
22333 might not hold the value of @code{_gp}. For example, if the code is
22334 part of a library that might be used in a boot monitor, programs that
22335 call boot monitor routines pass an unknown value in @code{$gp}.
22336 (In such situations, the boot monitor itself is usually compiled
22337 with @option{-G0}.)
22339 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
22340 @option{-mno-extern-sdata}.
22342 @item -membedded-data
22343 @itemx -mno-embedded-data
22344 @opindex membedded-data
22345 @opindex mno-embedded-data
22346 Allocate variables to the read-only data section first if possible, then
22347 next in the small data section if possible, otherwise in data. This gives
22348 slightly slower code than the default, but reduces the amount of RAM required
22349 when executing, and thus may be preferred for some embedded systems.
22351 @item -muninit-const-in-rodata
22352 @itemx -mno-uninit-const-in-rodata
22353 @opindex muninit-const-in-rodata
22354 @opindex mno-uninit-const-in-rodata
22355 Put uninitialized @code{const} variables in the read-only data section.
22356 This option is only meaningful in conjunction with @option{-membedded-data}.
22358 @item -mcode-readable=@var{setting}
22359 @opindex mcode-readable
22360 Specify whether GCC may generate code that reads from executable sections.
22361 There are three possible settings:
22364 @item -mcode-readable=yes
22365 Instructions may freely access executable sections. This is the
22368 @item -mcode-readable=pcrel
22369 MIPS16 PC-relative load instructions can access executable sections,
22370 but other instructions must not do so. This option is useful on 4KSc
22371 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
22372 It is also useful on processors that can be configured to have a dual
22373 instruction/data SRAM interface and that, like the M4K, automatically
22374 redirect PC-relative loads to the instruction RAM.
22376 @item -mcode-readable=no
22377 Instructions must not access executable sections. This option can be
22378 useful on targets that are configured to have a dual instruction/data
22379 SRAM interface but that (unlike the M4K) do not automatically redirect
22380 PC-relative loads to the instruction RAM.
22383 @item -msplit-addresses
22384 @itemx -mno-split-addresses
22385 @opindex msplit-addresses
22386 @opindex mno-split-addresses
22387 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
22388 relocation operators. This option has been superseded by
22389 @option{-mexplicit-relocs} but is retained for backwards compatibility.
22391 @item -mexplicit-relocs
22392 @itemx -mno-explicit-relocs
22393 @opindex mexplicit-relocs
22394 @opindex mno-explicit-relocs
22395 Use (do not use) assembler relocation operators when dealing with symbolic
22396 addresses. The alternative, selected by @option{-mno-explicit-relocs},
22397 is to use assembler macros instead.
22399 @option{-mexplicit-relocs} is the default if GCC was configured
22400 to use an assembler that supports relocation operators.
22402 @item -mcheck-zero-division
22403 @itemx -mno-check-zero-division
22404 @opindex mcheck-zero-division
22405 @opindex mno-check-zero-division
22406 Trap (do not trap) on integer division by zero.
22408 The default is @option{-mcheck-zero-division}.
22410 @item -mdivide-traps
22411 @itemx -mdivide-breaks
22412 @opindex mdivide-traps
22413 @opindex mdivide-breaks
22414 MIPS systems check for division by zero by generating either a
22415 conditional trap or a break instruction. Using traps results in
22416 smaller code, but is only supported on MIPS II and later. Also, some
22417 versions of the Linux kernel have a bug that prevents trap from
22418 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
22419 allow conditional traps on architectures that support them and
22420 @option{-mdivide-breaks} to force the use of breaks.
22422 The default is usually @option{-mdivide-traps}, but this can be
22423 overridden at configure time using @option{--with-divide=breaks}.
22424 Divide-by-zero checks can be completely disabled using
22425 @option{-mno-check-zero-division}.
22427 @item -mload-store-pairs
22428 @itemx -mno-load-store-pairs
22429 @opindex mload-store-pairs
22430 @opindex mno-load-store-pairs
22431 Enable (disable) an optimization that pairs consecutive load or store
22432 instructions to enable load/store bonding. This option is enabled by
22433 default but only takes effect when the selected architecture is known
22434 to support bonding.
22439 @opindex mno-memcpy
22440 Force (do not force) the use of @code{memcpy} for non-trivial block
22441 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
22442 most constant-sized copies.
22445 @itemx -mno-long-calls
22446 @opindex mlong-calls
22447 @opindex mno-long-calls
22448 Disable (do not disable) use of the @code{jal} instruction. Calling
22449 functions using @code{jal} is more efficient but requires the caller
22450 and callee to be in the same 256 megabyte segment.
22452 This option has no effect on abicalls code. The default is
22453 @option{-mno-long-calls}.
22459 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
22460 instructions, as provided by the R4650 ISA@.
22466 Enable (disable) use of the @code{madd} and @code{msub} integer
22467 instructions. The default is @option{-mimadd} on architectures
22468 that support @code{madd} and @code{msub} except for the 74k
22469 architecture where it was found to generate slower code.
22472 @itemx -mno-fused-madd
22473 @opindex mfused-madd
22474 @opindex mno-fused-madd
22475 Enable (disable) use of the floating-point multiply-accumulate
22476 instructions, when they are available. The default is
22477 @option{-mfused-madd}.
22479 On the R8000 CPU when multiply-accumulate instructions are used,
22480 the intermediate product is calculated to infinite precision
22481 and is not subject to the FCSR Flush to Zero bit. This may be
22482 undesirable in some circumstances. On other processors the result
22483 is numerically identical to the equivalent computation using
22484 separate multiply, add, subtract and negate instructions.
22488 Tell the MIPS assembler to not run its preprocessor over user
22489 assembler files (with a @samp{.s} suffix) when assembling them.
22492 @itemx -mno-fix-24k
22494 @opindex mno-fix-24k
22495 Work around the 24K E48 (lost data on stores during refill) errata.
22496 The workarounds are implemented by the assembler rather than by GCC@.
22499 @itemx -mno-fix-r4000
22500 @opindex mfix-r4000
22501 @opindex mno-fix-r4000
22502 Work around certain R4000 CPU errata:
22505 A double-word or a variable shift may give an incorrect result if executed
22506 immediately after starting an integer division.
22508 A double-word or a variable shift may give an incorrect result if executed
22509 while an integer multiplication is in progress.
22511 An integer division may give an incorrect result if started in a delay slot
22512 of a taken branch or a jump.
22516 @itemx -mno-fix-r4400
22517 @opindex mfix-r4400
22518 @opindex mno-fix-r4400
22519 Work around certain R4400 CPU errata:
22522 A double-word or a variable shift may give an incorrect result if executed
22523 immediately after starting an integer division.
22527 @itemx -mno-fix-r10000
22528 @opindex mfix-r10000
22529 @opindex mno-fix-r10000
22530 Work around certain R10000 errata:
22533 @code{ll}/@code{sc} sequences may not behave atomically on revisions
22534 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
22537 This option can only be used if the target architecture supports
22538 branch-likely instructions. @option{-mfix-r10000} is the default when
22539 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
22543 @itemx -mno-fix-r5900
22544 @opindex mfix-r5900
22545 Do not attempt to schedule the preceding instruction into the delay slot
22546 of a branch instruction placed at the end of a short loop of six
22547 instructions or fewer and always schedule a @code{nop} instruction there
22548 instead. The short loop bug under certain conditions causes loops to
22549 execute only once or twice, due to a hardware bug in the R5900 chip. The
22550 workaround is implemented by the assembler rather than by GCC@.
22553 @itemx -mno-fix-rm7000
22554 @opindex mfix-rm7000
22555 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
22556 workarounds are implemented by the assembler rather than by GCC@.
22559 @itemx -mno-fix-vr4120
22560 @opindex mfix-vr4120
22561 Work around certain VR4120 errata:
22564 @code{dmultu} does not always produce the correct result.
22566 @code{div} and @code{ddiv} do not always produce the correct result if one
22567 of the operands is negative.
22569 The workarounds for the division errata rely on special functions in
22570 @file{libgcc.a}. At present, these functions are only provided by
22571 the @code{mips64vr*-elf} configurations.
22573 Other VR4120 errata require a NOP to be inserted between certain pairs of
22574 instructions. These errata are handled by the assembler, not by GCC itself.
22577 @opindex mfix-vr4130
22578 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
22579 workarounds are implemented by the assembler rather than by GCC,
22580 although GCC avoids using @code{mflo} and @code{mfhi} if the
22581 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
22582 instructions are available instead.
22585 @itemx -mno-fix-sb1
22587 Work around certain SB-1 CPU core errata.
22588 (This flag currently works around the SB-1 revision 2
22589 ``F1'' and ``F2'' floating-point errata.)
22591 @item -mr10k-cache-barrier=@var{setting}
22592 @opindex mr10k-cache-barrier
22593 Specify whether GCC should insert cache barriers to avoid the
22594 side effects of speculation on R10K processors.
22596 In common with many processors, the R10K tries to predict the outcome
22597 of a conditional branch and speculatively executes instructions from
22598 the ``taken'' branch. It later aborts these instructions if the
22599 predicted outcome is wrong. However, on the R10K, even aborted
22600 instructions can have side effects.
22602 This problem only affects kernel stores and, depending on the system,
22603 kernel loads. As an example, a speculatively-executed store may load
22604 the target memory into cache and mark the cache line as dirty, even if
22605 the store itself is later aborted. If a DMA operation writes to the
22606 same area of memory before the ``dirty'' line is flushed, the cached
22607 data overwrites the DMA-ed data. See the R10K processor manual
22608 for a full description, including other potential problems.
22610 One workaround is to insert cache barrier instructions before every memory
22611 access that might be speculatively executed and that might have side
22612 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
22613 controls GCC's implementation of this workaround. It assumes that
22614 aborted accesses to any byte in the following regions does not have
22619 the memory occupied by the current function's stack frame;
22622 the memory occupied by an incoming stack argument;
22625 the memory occupied by an object with a link-time-constant address.
22628 It is the kernel's responsibility to ensure that speculative
22629 accesses to these regions are indeed safe.
22631 If the input program contains a function declaration such as:
22637 then the implementation of @code{foo} must allow @code{j foo} and
22638 @code{jal foo} to be executed speculatively. GCC honors this
22639 restriction for functions it compiles itself. It expects non-GCC
22640 functions (such as hand-written assembly code) to do the same.
22642 The option has three forms:
22645 @item -mr10k-cache-barrier=load-store
22646 Insert a cache barrier before a load or store that might be
22647 speculatively executed and that might have side effects even
22650 @item -mr10k-cache-barrier=store
22651 Insert a cache barrier before a store that might be speculatively
22652 executed and that might have side effects even if aborted.
22654 @item -mr10k-cache-barrier=none
22655 Disable the insertion of cache barriers. This is the default setting.
22658 @item -mflush-func=@var{func}
22659 @itemx -mno-flush-func
22660 @opindex mflush-func
22661 Specifies the function to call to flush the I and D caches, or to not
22662 call any such function. If called, the function must take the same
22663 arguments as the common @code{_flush_func}, that is, the address of the
22664 memory range for which the cache is being flushed, the size of the
22665 memory range, and the number 3 (to flush both caches). The default
22666 depends on the target GCC was configured for, but commonly is either
22667 @code{_flush_func} or @code{__cpu_flush}.
22669 @item mbranch-cost=@var{num}
22670 @opindex mbranch-cost
22671 Set the cost of branches to roughly @var{num} ``simple'' instructions.
22672 This cost is only a heuristic and is not guaranteed to produce
22673 consistent results across releases. A zero cost redundantly selects
22674 the default, which is based on the @option{-mtune} setting.
22676 @item -mbranch-likely
22677 @itemx -mno-branch-likely
22678 @opindex mbranch-likely
22679 @opindex mno-branch-likely
22680 Enable or disable use of Branch Likely instructions, regardless of the
22681 default for the selected architecture. By default, Branch Likely
22682 instructions may be generated if they are supported by the selected
22683 architecture. An exception is for the MIPS32 and MIPS64 architectures
22684 and processors that implement those architectures; for those, Branch
22685 Likely instructions are not be generated by default because the MIPS32
22686 and MIPS64 architectures specifically deprecate their use.
22688 @item -mcompact-branches=never
22689 @itemx -mcompact-branches=optimal
22690 @itemx -mcompact-branches=always
22691 @opindex mcompact-branches=never
22692 @opindex mcompact-branches=optimal
22693 @opindex mcompact-branches=always
22694 These options control which form of branches will be generated. The
22695 default is @option{-mcompact-branches=optimal}.
22697 The @option{-mcompact-branches=never} option ensures that compact branch
22698 instructions will never be generated.
22700 The @option{-mcompact-branches=always} option ensures that a compact
22701 branch instruction will be generated if available. If a compact branch
22702 instruction is not available, a delay slot form of the branch will be
22705 This option is supported from MIPS Release 6 onwards.
22707 The @option{-mcompact-branches=optimal} option will cause a delay slot
22708 branch to be used if one is available in the current ISA and the delay
22709 slot is successfully filled. If the delay slot is not filled, a compact
22710 branch will be chosen if one is available.
22712 @item -mfp-exceptions
22713 @itemx -mno-fp-exceptions
22714 @opindex mfp-exceptions
22715 Specifies whether FP exceptions are enabled. This affects how
22716 FP instructions are scheduled for some processors.
22717 The default is that FP exceptions are
22720 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
22721 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
22724 @item -mvr4130-align
22725 @itemx -mno-vr4130-align
22726 @opindex mvr4130-align
22727 The VR4130 pipeline is two-way superscalar, but can only issue two
22728 instructions together if the first one is 8-byte aligned. When this
22729 option is enabled, GCC aligns pairs of instructions that it
22730 thinks should execute in parallel.
22732 This option only has an effect when optimizing for the VR4130.
22733 It normally makes code faster, but at the expense of making it bigger.
22734 It is enabled by default at optimization level @option{-O3}.
22739 Enable (disable) generation of @code{synci} instructions on
22740 architectures that support it. The @code{synci} instructions (if
22741 enabled) are generated when @code{__builtin___clear_cache} is
22744 This option defaults to @option{-mno-synci}, but the default can be
22745 overridden by configuring GCC with @option{--with-synci}.
22747 When compiling code for single processor systems, it is generally safe
22748 to use @code{synci}. However, on many multi-core (SMP) systems, it
22749 does not invalidate the instruction caches on all cores and may lead
22750 to undefined behavior.
22752 @item -mrelax-pic-calls
22753 @itemx -mno-relax-pic-calls
22754 @opindex mrelax-pic-calls
22755 Try to turn PIC calls that are normally dispatched via register
22756 @code{$25} into direct calls. This is only possible if the linker can
22757 resolve the destination at link time and if the destination is within
22758 range for a direct call.
22760 @option{-mrelax-pic-calls} is the default if GCC was configured to use
22761 an assembler and a linker that support the @code{.reloc} assembly
22762 directive and @option{-mexplicit-relocs} is in effect. With
22763 @option{-mno-explicit-relocs}, this optimization can be performed by the
22764 assembler and the linker alone without help from the compiler.
22766 @item -mmcount-ra-address
22767 @itemx -mno-mcount-ra-address
22768 @opindex mmcount-ra-address
22769 @opindex mno-mcount-ra-address
22770 Emit (do not emit) code that allows @code{_mcount} to modify the
22771 calling function's return address. When enabled, this option extends
22772 the usual @code{_mcount} interface with a new @var{ra-address}
22773 parameter, which has type @code{intptr_t *} and is passed in register
22774 @code{$12}. @code{_mcount} can then modify the return address by
22775 doing both of the following:
22778 Returning the new address in register @code{$31}.
22780 Storing the new address in @code{*@var{ra-address}},
22781 if @var{ra-address} is nonnull.
22784 The default is @option{-mno-mcount-ra-address}.
22786 @item -mframe-header-opt
22787 @itemx -mno-frame-header-opt
22788 @opindex mframe-header-opt
22789 Enable (disable) frame header optimization in the o32 ABI. When using the
22790 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
22791 function to write out register arguments. When enabled, this optimization
22792 will suppress the allocation of the frame header if it can be determined that
22795 This optimization is off by default at all optimization levels.
22798 @itemx -mno-lxc1-sxc1
22799 @opindex mlxc1-sxc1
22800 When applicable, enable (disable) the generation of @code{lwxc1},
22801 @code{swxc1}, @code{ldxc1}, @code{sdxc1} instructions. Enabled by default.
22806 When applicable, enable (disable) the generation of 4-operand @code{madd.s},
22807 @code{madd.d} and related instructions. Enabled by default.
22812 @subsection MMIX Options
22813 @cindex MMIX Options
22815 These options are defined for the MMIX:
22819 @itemx -mno-libfuncs
22821 @opindex mno-libfuncs
22822 Specify that intrinsic library functions are being compiled, passing all
22823 values in registers, no matter the size.
22826 @itemx -mno-epsilon
22828 @opindex mno-epsilon
22829 Generate floating-point comparison instructions that compare with respect
22830 to the @code{rE} epsilon register.
22832 @item -mabi=mmixware
22834 @opindex mabi=mmixware
22836 Generate code that passes function parameters and return values that (in
22837 the called function) are seen as registers @code{$0} and up, as opposed to
22838 the GNU ABI which uses global registers @code{$231} and up.
22840 @item -mzero-extend
22841 @itemx -mno-zero-extend
22842 @opindex mzero-extend
22843 @opindex mno-zero-extend
22844 When reading data from memory in sizes shorter than 64 bits, use (do not
22845 use) zero-extending load instructions by default, rather than
22846 sign-extending ones.
22849 @itemx -mno-knuthdiv
22851 @opindex mno-knuthdiv
22852 Make the result of a division yielding a remainder have the same sign as
22853 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
22854 remainder follows the sign of the dividend. Both methods are
22855 arithmetically valid, the latter being almost exclusively used.
22857 @item -mtoplevel-symbols
22858 @itemx -mno-toplevel-symbols
22859 @opindex mtoplevel-symbols
22860 @opindex mno-toplevel-symbols
22861 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
22862 code can be used with the @code{PREFIX} assembly directive.
22866 Generate an executable in the ELF format, rather than the default
22867 @samp{mmo} format used by the @command{mmix} simulator.
22869 @item -mbranch-predict
22870 @itemx -mno-branch-predict
22871 @opindex mbranch-predict
22872 @opindex mno-branch-predict
22873 Use (do not use) the probable-branch instructions, when static branch
22874 prediction indicates a probable branch.
22876 @item -mbase-addresses
22877 @itemx -mno-base-addresses
22878 @opindex mbase-addresses
22879 @opindex mno-base-addresses
22880 Generate (do not generate) code that uses @emph{base addresses}. Using a
22881 base address automatically generates a request (handled by the assembler
22882 and the linker) for a constant to be set up in a global register. The
22883 register is used for one or more base address requests within the range 0
22884 to 255 from the value held in the register. The generally leads to short
22885 and fast code, but the number of different data items that can be
22886 addressed is limited. This means that a program that uses lots of static
22887 data may require @option{-mno-base-addresses}.
22889 @item -msingle-exit
22890 @itemx -mno-single-exit
22891 @opindex msingle-exit
22892 @opindex mno-single-exit
22893 Force (do not force) generated code to have a single exit point in each
22897 @node MN10300 Options
22898 @subsection MN10300 Options
22899 @cindex MN10300 options
22901 These @option{-m} options are defined for Matsushita MN10300 architectures:
22906 Generate code to avoid bugs in the multiply instructions for the MN10300
22907 processors. This is the default.
22909 @item -mno-mult-bug
22910 @opindex mno-mult-bug
22911 Do not generate code to avoid bugs in the multiply instructions for the
22912 MN10300 processors.
22916 Generate code using features specific to the AM33 processor.
22920 Do not generate code using features specific to the AM33 processor. This
22925 Generate code using features specific to the AM33/2.0 processor.
22929 Generate code using features specific to the AM34 processor.
22931 @item -mtune=@var{cpu-type}
22933 Use the timing characteristics of the indicated CPU type when
22934 scheduling instructions. This does not change the targeted processor
22935 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
22936 @samp{am33-2} or @samp{am34}.
22938 @item -mreturn-pointer-on-d0
22939 @opindex mreturn-pointer-on-d0
22940 When generating a function that returns a pointer, return the pointer
22941 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
22942 only in @code{a0}, and attempts to call such functions without a prototype
22943 result in errors. Note that this option is on by default; use
22944 @option{-mno-return-pointer-on-d0} to disable it.
22948 Do not link in the C run-time initialization object file.
22952 Indicate to the linker that it should perform a relaxation optimization pass
22953 to shorten branches, calls and absolute memory addresses. This option only
22954 has an effect when used on the command line for the final link step.
22956 This option makes symbolic debugging impossible.
22960 Allow the compiler to generate @emph{Long Instruction Word}
22961 instructions if the target is the @samp{AM33} or later. This is the
22962 default. This option defines the preprocessor macro @code{__LIW__}.
22966 Do not allow the compiler to generate @emph{Long Instruction Word}
22967 instructions. This option defines the preprocessor macro
22972 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
22973 instructions if the target is the @samp{AM33} or later. This is the
22974 default. This option defines the preprocessor macro @code{__SETLB__}.
22978 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
22979 instructions. This option defines the preprocessor macro
22980 @code{__NO_SETLB__}.
22984 @node Moxie Options
22985 @subsection Moxie Options
22986 @cindex Moxie Options
22992 Generate big-endian code. This is the default for @samp{moxie-*-*}
22997 Generate little-endian code.
23001 Generate mul.x and umul.x instructions. This is the default for
23002 @samp{moxiebox-*-*} configurations.
23006 Do not link in the C run-time initialization object file.
23010 @node MSP430 Options
23011 @subsection MSP430 Options
23012 @cindex MSP430 Options
23014 These options are defined for the MSP430:
23020 Force assembly output to always use hex constants. Normally such
23021 constants are signed decimals, but this option is available for
23022 testsuite and/or aesthetic purposes.
23026 Select the MCU to target. This is used to create a C preprocessor
23027 symbol based upon the MCU name, converted to upper case and pre- and
23028 post-fixed with @samp{__}. This in turn is used by the
23029 @file{msp430.h} header file to select an MCU-specific supplementary
23032 The option also sets the ISA to use. If the MCU name is one that is
23033 known to only support the 430 ISA then that is selected, otherwise the
23034 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
23035 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
23036 name selects the 430X ISA.
23038 In addition an MCU-specific linker script is added to the linker
23039 command line. The script's name is the name of the MCU with
23040 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
23041 command line defines the C preprocessor symbol @code{__XXX__} and
23042 cause the linker to search for a script called @file{xxx.ld}.
23044 The ISA and hardware multiply supported for the different MCUs is hard-coded
23045 into GCC. However, an external @samp{devices.csv} file can be used to
23046 extend device support beyond those that have been hard-coded.
23048 GCC searches for the @samp{devices.csv} file on the paths specified
23049 with the @code{-I} and @code{-L} options.
23052 @itemx -mno-warn-mcu
23054 @opindex mno-warn-mcu
23055 This option enables or disables warnings about conflicts between the
23056 MCU name specified by the @option{-mmcu} option and the ISA set by the
23057 @option{-mcpu} option and/or the hardware multiply support set by the
23058 @option{-mhwmult} option. It also toggles warnings about unrecognized
23059 MCU names. This option is on by default.
23063 Specifies the ISA to use. Accepted values are @samp{msp430},
23064 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
23065 @option{-mmcu=} option should be used to select the ISA.
23069 Link to the simulator runtime libraries and linker script. Overrides
23070 any scripts that would be selected by the @option{-mmcu=} option.
23074 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
23078 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
23082 This option is passed to the assembler and linker, and allows the
23083 linker to perform certain optimizations that cannot be done until
23088 Describes the type of hardware multiply supported by the target.
23089 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
23090 for the original 16-bit-only multiply supported by early MCUs.
23091 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
23092 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
23093 A value of @samp{auto} can also be given. This tells GCC to deduce
23094 the hardware multiply support based upon the MCU name provided by the
23095 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
23096 the MCU name is not recognized then no hardware multiply support is
23097 assumed. @code{auto} is the default setting.
23099 Hardware multiplies are normally performed by calling a library
23100 routine. This saves space in the generated code. When compiling at
23101 @option{-O3} or higher however the hardware multiplier is invoked
23102 inline. This makes for bigger, but faster code.
23104 The hardware multiply routines disable interrupts whilst running and
23105 restore the previous interrupt state when they finish. This makes
23106 them safe to use inside interrupt handlers as well as in normal code.
23110 Enable the use of a minimum runtime environment - no static
23111 initializers or constructors. This is intended for memory-constrained
23112 devices. The compiler includes special symbols in some objects
23113 that tell the linker and runtime which code fragments are required.
23115 @item -mcode-region=
23116 @itemx -mdata-region=
23117 @opindex mcode-region
23118 @opindex mdata-region
23119 These options tell the compiler where to place functions and data that
23120 do not have one of the @code{lower}, @code{upper}, @code{either} or
23121 @code{section} attributes. Possible values are @code{lower},
23122 @code{upper}, @code{either} or @code{any}. The first three behave
23123 like the corresponding attribute. The fourth possible value -
23124 @code{any} - is the default. It leaves placement entirely up to the
23125 linker script and how it assigns the standard sections
23126 (@code{.text}, @code{.data}, etc) to the memory regions.
23128 @item -msilicon-errata=
23129 @opindex msilicon-errata
23130 This option passes on a request to assembler to enable the fixes for
23131 the named silicon errata.
23133 @item -msilicon-errata-warn=
23134 @opindex msilicon-errata-warn
23135 This option passes on a request to the assembler to enable warning
23136 messages when a silicon errata might need to be applied.
23138 @item -mwarn-devices-csv
23139 @itemx -mno-warn-devices-csv
23140 @opindex mwarn-devices-csv
23141 @opindex mno-warn-devices-csv
23142 Warn if @samp{devices.csv} is not found or there are problem parsing it
23147 @node NDS32 Options
23148 @subsection NDS32 Options
23149 @cindex NDS32 Options
23151 These options are defined for NDS32 implementations:
23156 @opindex mbig-endian
23157 Generate code in big-endian mode.
23159 @item -mlittle-endian
23160 @opindex mlittle-endian
23161 Generate code in little-endian mode.
23163 @item -mreduced-regs
23164 @opindex mreduced-regs
23165 Use reduced-set registers for register allocation.
23168 @opindex mfull-regs
23169 Use full-set registers for register allocation.
23173 Generate conditional move instructions.
23177 Do not generate conditional move instructions.
23181 Generate performance extension instructions.
23183 @item -mno-ext-perf
23184 @opindex mno-ext-perf
23185 Do not generate performance extension instructions.
23188 @opindex mext-perf2
23189 Generate performance extension 2 instructions.
23191 @item -mno-ext-perf2
23192 @opindex mno-ext-perf2
23193 Do not generate performance extension 2 instructions.
23196 @opindex mext-string
23197 Generate string extension instructions.
23199 @item -mno-ext-string
23200 @opindex mno-ext-string
23201 Do not generate string extension instructions.
23205 Generate v3 push25/pop25 instructions.
23208 @opindex mno-v3push
23209 Do not generate v3 push25/pop25 instructions.
23213 Generate 16-bit instructions.
23216 @opindex mno-16-bit
23217 Do not generate 16-bit instructions.
23219 @item -misr-vector-size=@var{num}
23220 @opindex misr-vector-size
23221 Specify the size of each interrupt vector, which must be 4 or 16.
23223 @item -mcache-block-size=@var{num}
23224 @opindex mcache-block-size
23225 Specify the size of each cache block,
23226 which must be a power of 2 between 4 and 512.
23228 @item -march=@var{arch}
23230 Specify the name of the target architecture.
23232 @item -mcmodel=@var{code-model}
23234 Set the code model to one of
23237 All the data and read-only data segments must be within 512KB addressing space.
23238 The text segment must be within 16MB addressing space.
23239 @item @samp{medium}
23240 The data segment must be within 512KB while the read-only data segment can be
23241 within 4GB addressing space. The text segment should be still within 16MB
23244 All the text and data segments can be within 4GB addressing space.
23248 @opindex mctor-dtor
23249 Enable constructor/destructor feature.
23253 Guide linker to relax instructions.
23257 @node Nios II Options
23258 @subsection Nios II Options
23259 @cindex Nios II options
23260 @cindex Altera Nios II options
23262 These are the options defined for the Altera Nios II processor.
23268 @cindex smaller data references
23269 Put global and static objects less than or equal to @var{num} bytes
23270 into the small data or BSS sections instead of the normal data or BSS
23271 sections. The default value of @var{num} is 8.
23273 @item -mgpopt=@var{option}
23278 Generate (do not generate) GP-relative accesses. The following
23279 @var{option} names are recognized:
23284 Do not generate GP-relative accesses.
23287 Generate GP-relative accesses for small data objects that are not
23288 external, weak, or uninitialized common symbols.
23289 Also use GP-relative addressing for objects that
23290 have been explicitly placed in a small data section via a @code{section}
23294 As for @samp{local}, but also generate GP-relative accesses for
23295 small data objects that are external, weak, or common. If you use this option,
23296 you must ensure that all parts of your program (including libraries) are
23297 compiled with the same @option{-G} setting.
23300 Generate GP-relative accesses for all data objects in the program. If you
23301 use this option, the entire data and BSS segments
23302 of your program must fit in 64K of memory and you must use an appropriate
23303 linker script to allocate them within the addressable range of the
23307 Generate GP-relative addresses for function pointers as well as data
23308 pointers. If you use this option, the entire text, data, and BSS segments
23309 of your program must fit in 64K of memory and you must use an appropriate
23310 linker script to allocate them within the addressable range of the
23315 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
23316 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
23318 The default is @option{-mgpopt} except when @option{-fpic} or
23319 @option{-fPIC} is specified to generate position-independent code.
23320 Note that the Nios II ABI does not permit GP-relative accesses from
23323 You may need to specify @option{-mno-gpopt} explicitly when building
23324 programs that include large amounts of small data, including large
23325 GOT data sections. In this case, the 16-bit offset for GP-relative
23326 addressing may not be large enough to allow access to the entire
23327 small data section.
23329 @item -mgprel-sec=@var{regexp}
23330 @opindex mgprel-sec
23331 This option specifies additional section names that can be accessed via
23332 GP-relative addressing. It is most useful in conjunction with
23333 @code{section} attributes on variable declarations
23334 (@pxref{Common Variable Attributes}) and a custom linker script.
23335 The @var{regexp} is a POSIX Extended Regular Expression.
23337 This option does not affect the behavior of the @option{-G} option, and
23338 the specified sections are in addition to the standard @code{.sdata}
23339 and @code{.sbss} small-data sections that are recognized by @option{-mgpopt}.
23341 @item -mr0rel-sec=@var{regexp}
23342 @opindex mr0rel-sec
23343 This option specifies names of sections that can be accessed via a
23344 16-bit offset from @code{r0}; that is, in the low 32K or high 32K
23345 of the 32-bit address space. It is most useful in conjunction with
23346 @code{section} attributes on variable declarations
23347 (@pxref{Common Variable Attributes}) and a custom linker script.
23348 The @var{regexp} is a POSIX Extended Regular Expression.
23350 In contrast to the use of GP-relative addressing for small data,
23351 zero-based addressing is never generated by default and there are no
23352 conventional section names used in standard linker scripts for sections
23353 in the low or high areas of memory.
23359 Generate little-endian (default) or big-endian (experimental) code,
23362 @item -march=@var{arch}
23364 This specifies the name of the target Nios II architecture. GCC uses this
23365 name to determine what kind of instructions it can emit when generating
23366 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
23368 The preprocessor macro @code{__nios2_arch__} is available to programs,
23369 with value 1 or 2, indicating the targeted ISA level.
23371 @item -mbypass-cache
23372 @itemx -mno-bypass-cache
23373 @opindex mno-bypass-cache
23374 @opindex mbypass-cache
23375 Force all load and store instructions to always bypass cache by
23376 using I/O variants of the instructions. The default is not to
23379 @item -mno-cache-volatile
23380 @itemx -mcache-volatile
23381 @opindex mcache-volatile
23382 @opindex mno-cache-volatile
23383 Volatile memory access bypass the cache using the I/O variants of
23384 the load and store instructions. The default is not to bypass the cache.
23386 @item -mno-fast-sw-div
23387 @itemx -mfast-sw-div
23388 @opindex mno-fast-sw-div
23389 @opindex mfast-sw-div
23390 Do not use table-based fast divide for small numbers. The default
23391 is to use the fast divide at @option{-O3} and above.
23395 @itemx -mno-hw-mulx
23399 @opindex mno-hw-mul
23401 @opindex mno-hw-mulx
23403 @opindex mno-hw-div
23405 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
23406 instructions by the compiler. The default is to emit @code{mul}
23407 and not emit @code{div} and @code{mulx}.
23413 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
23414 CDX (code density) instructions. Enabling these instructions also
23415 requires @option{-march=r2}. Since these instructions are optional
23416 extensions to the R2 architecture, the default is not to emit them.
23418 @item -mcustom-@var{insn}=@var{N}
23419 @itemx -mno-custom-@var{insn}
23420 @opindex mcustom-@var{insn}
23421 @opindex mno-custom-@var{insn}
23422 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
23423 custom instruction with encoding @var{N} when generating code that uses
23424 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
23425 instruction 253 for single-precision floating-point add operations instead
23426 of the default behavior of using a library call.
23428 The following values of @var{insn} are supported. Except as otherwise
23429 noted, floating-point operations are expected to be implemented with
23430 normal IEEE 754 semantics and correspond directly to the C operators or the
23431 equivalent GCC built-in functions (@pxref{Other Builtins}).
23433 Single-precision floating point:
23436 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
23437 Binary arithmetic operations.
23443 Unary absolute value.
23445 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
23446 Comparison operations.
23448 @item @samp{fmins}, @samp{fmaxs}
23449 Floating-point minimum and maximum. These instructions are only
23450 generated if @option{-ffinite-math-only} is specified.
23452 @item @samp{fsqrts}
23453 Unary square root operation.
23455 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
23456 Floating-point trigonometric and exponential functions. These instructions
23457 are only generated if @option{-funsafe-math-optimizations} is also specified.
23461 Double-precision floating point:
23464 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
23465 Binary arithmetic operations.
23471 Unary absolute value.
23473 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
23474 Comparison operations.
23476 @item @samp{fmind}, @samp{fmaxd}
23477 Double-precision minimum and maximum. These instructions are only
23478 generated if @option{-ffinite-math-only} is specified.
23480 @item @samp{fsqrtd}
23481 Unary square root operation.
23483 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
23484 Double-precision trigonometric and exponential functions. These instructions
23485 are only generated if @option{-funsafe-math-optimizations} is also specified.
23491 @item @samp{fextsd}
23492 Conversion from single precision to double precision.
23494 @item @samp{ftruncds}
23495 Conversion from double precision to single precision.
23497 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
23498 Conversion from floating point to signed or unsigned integer types, with
23499 truncation towards zero.
23502 Conversion from single-precision floating point to signed integer,
23503 rounding to the nearest integer and ties away from zero.
23504 This corresponds to the @code{__builtin_lroundf} function when
23505 @option{-fno-math-errno} is used.
23507 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
23508 Conversion from signed or unsigned integer types to floating-point types.
23512 In addition, all of the following transfer instructions for internal
23513 registers X and Y must be provided to use any of the double-precision
23514 floating-point instructions. Custom instructions taking two
23515 double-precision source operands expect the first operand in the
23516 64-bit register X. The other operand (or only operand of a unary
23517 operation) is given to the custom arithmetic instruction with the
23518 least significant half in source register @var{src1} and the most
23519 significant half in @var{src2}. A custom instruction that returns a
23520 double-precision result returns the most significant 32 bits in the
23521 destination register and the other half in 32-bit register Y.
23522 GCC automatically generates the necessary code sequences to write
23523 register X and/or read register Y when double-precision floating-point
23524 instructions are used.
23529 Write @var{src1} into the least significant half of X and @var{src2} into
23530 the most significant half of X.
23533 Write @var{src1} into Y.
23535 @item @samp{frdxhi}, @samp{frdxlo}
23536 Read the most or least (respectively) significant half of X and store it in
23540 Read the value of Y and store it into @var{dest}.
23543 Note that you can gain more local control over generation of Nios II custom
23544 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
23545 and @code{target("no-custom-@var{insn}")} function attributes
23546 (@pxref{Function Attributes})
23547 or pragmas (@pxref{Function Specific Option Pragmas}).
23549 @item -mcustom-fpu-cfg=@var{name}
23550 @opindex mcustom-fpu-cfg
23552 This option enables a predefined, named set of custom instruction encodings
23553 (see @option{-mcustom-@var{insn}} above).
23554 Currently, the following sets are defined:
23556 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
23557 @gccoptlist{-mcustom-fmuls=252 @gol
23558 -mcustom-fadds=253 @gol
23559 -mcustom-fsubs=254 @gol
23560 -fsingle-precision-constant}
23562 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
23563 @gccoptlist{-mcustom-fmuls=252 @gol
23564 -mcustom-fadds=253 @gol
23565 -mcustom-fsubs=254 @gol
23566 -mcustom-fdivs=255 @gol
23567 -fsingle-precision-constant}
23569 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
23570 @gccoptlist{-mcustom-floatus=243 @gol
23571 -mcustom-fixsi=244 @gol
23572 -mcustom-floatis=245 @gol
23573 -mcustom-fcmpgts=246 @gol
23574 -mcustom-fcmples=249 @gol
23575 -mcustom-fcmpeqs=250 @gol
23576 -mcustom-fcmpnes=251 @gol
23577 -mcustom-fmuls=252 @gol
23578 -mcustom-fadds=253 @gol
23579 -mcustom-fsubs=254 @gol
23580 -mcustom-fdivs=255 @gol
23581 -fsingle-precision-constant}
23583 Custom instruction assignments given by individual
23584 @option{-mcustom-@var{insn}=} options override those given by
23585 @option{-mcustom-fpu-cfg=}, regardless of the
23586 order of the options on the command line.
23588 Note that you can gain more local control over selection of a FPU
23589 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
23590 function attribute (@pxref{Function Attributes})
23591 or pragma (@pxref{Function Specific Option Pragmas}).
23595 These additional @samp{-m} options are available for the Altera Nios II
23596 ELF (bare-metal) target:
23602 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
23603 startup and termination code, and is typically used in conjunction with
23604 @option{-msys-crt0=} to specify the location of the alternate startup code
23605 provided by the HAL BSP.
23609 Link with a limited version of the C library, @option{-lsmallc}, rather than
23612 @item -msys-crt0=@var{startfile}
23614 @var{startfile} is the file name of the startfile (crt0) to use
23615 when linking. This option is only useful in conjunction with @option{-mhal}.
23617 @item -msys-lib=@var{systemlib}
23619 @var{systemlib} is the library name of the library that provides
23620 low-level system calls required by the C library,
23621 e.g.@: @code{read} and @code{write}.
23622 This option is typically used to link with a library provided by a HAL BSP.
23626 @node Nvidia PTX Options
23627 @subsection Nvidia PTX Options
23628 @cindex Nvidia PTX options
23629 @cindex nvptx options
23631 These options are defined for Nvidia PTX:
23639 Generate code for 32-bit or 64-bit ABI.
23641 @item -misa=@var{ISA-string}
23643 Generate code for given the specified PTX ISA (e.g.@: @samp{sm_35}). ISA
23644 strings must be lower-case. Valid ISA strings include @samp{sm_30} and
23645 @samp{sm_35}. The default ISA is sm_30.
23648 @opindex mmainkernel
23649 Link in code for a __main kernel. This is for stand-alone instead of
23650 offloading execution.
23654 Apply partitioned execution optimizations. This is the default when any
23655 level of optimization is selected.
23658 @opindex msoft-stack
23659 Generate code that does not use @code{.local} memory
23660 directly for stack storage. Instead, a per-warp stack pointer is
23661 maintained explicitly. This enables variable-length stack allocation (with
23662 variable-length arrays or @code{alloca}), and when global memory is used for
23663 underlying storage, makes it possible to access automatic variables from other
23664 threads, or with atomic instructions. This code generation variant is used
23665 for OpenMP offloading, but the option is exposed on its own for the purpose
23666 of testing the compiler; to generate code suitable for linking into programs
23667 using OpenMP offloading, use option @option{-mgomp}.
23669 @item -muniform-simt
23670 @opindex muniform-simt
23671 Switch to code generation variant that allows to execute all threads in each
23672 warp, while maintaining memory state and side effects as if only one thread
23673 in each warp was active outside of OpenMP SIMD regions. All atomic operations
23674 and calls to runtime (malloc, free, vprintf) are conditionally executed (iff
23675 current lane index equals the master lane index), and the register being
23676 assigned is copied via a shuffle instruction from the master lane. Outside of
23677 SIMD regions lane 0 is the master; inside, each thread sees itself as the
23678 master. Shared memory array @code{int __nvptx_uni[]} stores all-zeros or
23679 all-ones bitmasks for each warp, indicating current mode (0 outside of SIMD
23680 regions). Each thread can bitwise-and the bitmask at position @code{tid.y}
23681 with current lane index to compute the master lane index.
23685 Generate code for use in OpenMP offloading: enables @option{-msoft-stack} and
23686 @option{-muniform-simt} options, and selects corresponding multilib variant.
23690 @node OpenRISC Options
23691 @subsection OpenRISC Options
23692 @cindex OpenRISC Options
23694 These options are defined for OpenRISC:
23698 @item -mboard=@var{name}
23700 Configure a board specific runtime. This will be passed to the linker for
23701 newlib board library linking. The default is @code{or1ksim}.
23705 This option is ignored; it is for compatibility purposes only. This used to
23706 select linker and preprocessor options for use with newlib.
23712 Select software or hardware divide (@code{l.div}, @code{l.divu}) instructions.
23713 This default is hardware divide.
23719 Select software or hardware multiply (@code{l.mul}, @code{l.muli}) instructions.
23720 This default is hardware multiply.
23723 @itemx -mhard-float
23724 @opindex msoft-float
23725 @opindex mhard-float
23726 Select software or hardware for floating point operations.
23727 The default is software.
23729 @item -mdouble-float
23730 @opindex mdouble-float
23731 When @option{-mhard-float} is selected, enables generation of double-precision
23732 floating point instructions. By default functions from @file{libgcc} are used
23733 to perform double-precision floating point operations.
23735 @item -munordered-float
23736 @opindex munordered-float
23737 When @option{-mhard-float} is selected, enables generation of unordered
23738 floating point compare and set flag (@code{lf.sfun*}) instructions. By default
23739 functions from @file{libgcc} are used to perform unordered floating point
23740 compare and set flag operations.
23744 Enable generation of conditional move (@code{l.cmov}) instructions. By
23745 default the equivalent will be generated using using set and branch.
23749 Enable generation of rotate right (@code{l.ror}) instructions. By default
23750 functions from @file{libgcc} are used to perform rotate right operations.
23754 Enable generation of rotate right with immediate (@code{l.rori}) instructions.
23755 By default functions from @file{libgcc} are used to perform rotate right with
23756 immediate operations.
23760 Enable generation of sign extension (@code{l.ext*}) instructions. By default
23761 memory loads are used to perform sign extension.
23765 Enable generation of compare and set flag with immediate (@code{l.sf*i})
23766 instructions. By default extra instructions will be generated to store the
23767 immediate to a register first.
23771 Enable generation of shift with immediate (@code{l.srai}, @code{l.srli},
23772 @code{l.slli}) instructions. By default extra instructions will be generated
23773 to store the immediate to a register first.
23778 @node PDP-11 Options
23779 @subsection PDP-11 Options
23780 @cindex PDP-11 Options
23782 These options are defined for the PDP-11:
23787 Use hardware FPP floating point. This is the default. (FIS floating
23788 point on the PDP-11/40 is not supported.) Implies -m45.
23791 @opindex msoft-float
23792 Do not use hardware floating point.
23796 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
23800 Return floating-point results in memory. This is the default.
23804 Generate code for a PDP-11/40. Implies -msoft-float -mno-split.
23808 Generate code for a PDP-11/45. This is the default.
23812 Generate code for a PDP-11/10. Implies -msoft-float -mno-split.
23818 Use 16-bit @code{int}. This is the default.
23824 Use 32-bit @code{int}.
23828 Target has split instruction and data space. Implies -m45.
23832 Use Unix assembler syntax.
23836 Use DEC assembler syntax.
23840 Use GNU assembler syntax. This is the default.
23844 Use the new LRA register allocator. By default, the old ``reload''
23848 @node picoChip Options
23849 @subsection picoChip Options
23850 @cindex picoChip options
23852 These @samp{-m} options are defined for picoChip implementations:
23856 @item -mae=@var{ae_type}
23858 Set the instruction set, register set, and instruction scheduling
23859 parameters for array element type @var{ae_type}. Supported values
23860 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
23862 @option{-mae=ANY} selects a completely generic AE type. Code
23863 generated with this option runs on any of the other AE types. The
23864 code is not as efficient as it would be if compiled for a specific
23865 AE type, and some types of operation (e.g., multiplication) do not
23866 work properly on all types of AE.
23868 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
23869 for compiled code, and is the default.
23871 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
23872 option may suffer from poor performance of byte (char) manipulation,
23873 since the DSP AE does not provide hardware support for byte load/stores.
23875 @item -msymbol-as-address
23876 Enable the compiler to directly use a symbol name as an address in a
23877 load/store instruction, without first loading it into a
23878 register. Typically, the use of this option generates larger
23879 programs, which run faster than when the option isn't used. However, the
23880 results vary from program to program, so it is left as a user option,
23881 rather than being permanently enabled.
23883 @item -mno-inefficient-warnings
23884 Disables warnings about the generation of inefficient code. These
23885 warnings can be generated, for example, when compiling code that
23886 performs byte-level memory operations on the MAC AE type. The MAC AE has
23887 no hardware support for byte-level memory operations, so all byte
23888 load/stores must be synthesized from word load/store operations. This is
23889 inefficient and a warning is generated to indicate
23890 that you should rewrite the code to avoid byte operations, or to target
23891 an AE type that has the necessary hardware support. This option disables
23896 @node PowerPC Options
23897 @subsection PowerPC Options
23898 @cindex PowerPC options
23900 These are listed under @xref{RS/6000 and PowerPC Options}.
23903 @subsection PRU Options
23904 @cindex PRU Options
23906 These command-line options are defined for PRU target:
23911 Link with a minimum runtime environment, with no support for static
23912 initializers and constructors. Using this option can significantly reduce
23913 the size of the final ELF binary. Beware that the compiler could still
23914 generate code with static initializers and constructors. It is up to the
23915 programmer to ensure that the source program will not use those features.
23917 @item -mmcu=@var{mcu}
23919 Specify the PRU MCU variant to use. Check Newlib for the exact list of
23924 Make GCC pass the @option{--no-relax} command-line option to the linker
23925 instead of the @option{--relax} option.
23929 Allow (or do not allow) GCC to use the LOOP instruction.
23931 @item -mabi=@var{variant}
23933 Specify the ABI variant to output code for. @option{-mabi=ti} selects the
23934 unmodified TI ABI while @option{-mabi=gnu} selects a GNU variant that copes
23935 more naturally with certain GCC assumptions. These are the differences:
23938 @item Function Pointer Size
23939 TI ABI specifies that function (code) pointers are 16-bit, whereas GNU
23940 supports only 32-bit data and code pointers.
23942 @item Optional Return Value Pointer
23943 Function return values larger than 64 bits are passed by using a hidden
23944 pointer as the first argument of the function. TI ABI, though, mandates that
23945 the pointer can be NULL in case the caller is not using the returned value.
23946 GNU always passes and expects a valid return value pointer.
23950 The current @option{-mabi=ti} implementation simply raises a compile error
23951 when any of the above code constructs is detected. As a consequence
23952 the standard C library cannot be built and it is omitted when linking with
23955 Relaxation is a GNU feature and for safety reasons is disabled when using
23956 @option{-mabi=ti}. The TI toolchain does not emit relocations for QBBx
23957 instructions, so the GNU linker cannot adjust them when shortening adjacent
23958 LDI32 pseudo instructions.
23962 @node RISC-V Options
23963 @subsection RISC-V Options
23964 @cindex RISC-V Options
23966 These command-line options are defined for RISC-V targets:
23969 @item -mbranch-cost=@var{n}
23970 @opindex mbranch-cost
23971 Set the cost of branches to roughly @var{n} instructions.
23976 When generating PIC code, do or don't allow the use of PLTs. Ignored for
23977 non-PIC. The default is @option{-mplt}.
23979 @item -mabi=@var{ABI-string}
23981 Specify integer and floating-point calling convention. @var{ABI-string}
23982 contains two parts: the size of integer types and the registers used for
23983 floating-point types. For example @samp{-march=rv64ifd -mabi=lp64d} means that
23984 @samp{long} and pointers are 64-bit (implicitly defining @samp{int} to be
23985 32-bit), and that floating-point values up to 64 bits wide are passed in F
23986 registers. Contrast this with @samp{-march=rv64ifd -mabi=lp64f}, which still
23987 allows the compiler to generate code that uses the F and D extensions but only
23988 allows floating-point values up to 32 bits long to be passed in registers; or
23989 @samp{-march=rv64ifd -mabi=lp64}, in which no floating-point arguments will be
23990 passed in registers.
23992 The default for this argument is system dependent, users who want a specific
23993 calling convention should specify one explicitly. The valid calling
23994 conventions are: @samp{ilp32}, @samp{ilp32f}, @samp{ilp32d}, @samp{lp64},
23995 @samp{lp64f}, and @samp{lp64d}. Some calling conventions are impossible to
23996 implement on some ISAs: for example, @samp{-march=rv32if -mabi=ilp32d} is
23997 invalid because the ABI requires 64-bit values be passed in F registers, but F
23998 registers are only 32 bits wide. There is also the @samp{ilp32e} ABI that can
23999 only be used with the @samp{rv32e} architecture. This ABI is not well
24000 specified at present, and is subject to change.
24005 Do or don't use hardware floating-point divide and square root instructions.
24006 This requires the F or D extensions for floating-point registers. The default
24007 is to use them if the specified architecture has these instructions.
24012 Do or don't use hardware instructions for integer division. This requires the
24013 M extension. The default is to use them if the specified architecture has
24014 these instructions.
24016 @item -march=@var{ISA-string}
24018 Generate code for given RISC-V ISA (e.g.@: @samp{rv64im}). ISA strings must be
24019 lower-case. Examples include @samp{rv64i}, @samp{rv32g}, @samp{rv32e}, and
24022 @item -mtune=@var{processor-string}
24024 Optimize the output for the given processor, specified by microarchitecture
24025 name. Permissible values for this option are: @samp{rocket},
24026 @samp{sifive-3-series}, @samp{sifive-5-series}, @samp{sifive-7-series},
24029 When @option{-mtune=} is not specified, the default is @samp{rocket}.
24031 The @samp{size} choice is not intended for use by end-users. This is used
24032 when @option{-Os} is specified. It overrides the instruction cost info
24033 provided by @option{-mtune=}, but does not override the pipeline info. This
24034 helps reduce code size while still giving good performance.
24036 @item -mpreferred-stack-boundary=@var{num}
24037 @opindex mpreferred-stack-boundary
24038 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
24039 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
24040 the default is 4 (16 bytes or 128-bits).
24042 @strong{Warning:} If you use this switch, then you must build all modules with
24043 the same value, including any libraries. This includes the system libraries
24044 and startup modules.
24046 @item -msmall-data-limit=@var{n}
24047 @opindex msmall-data-limit
24048 Put global and static data smaller than @var{n} bytes into a special section
24051 @item -msave-restore
24052 @itemx -mno-save-restore
24053 @opindex msave-restore
24054 Do or don't use smaller but slower prologue and epilogue code that uses
24055 library function calls. The default is to use fast inline prologues and
24058 @item -mstrict-align
24059 @itemx -mno-strict-align
24060 @opindex mstrict-align
24061 Do not or do generate unaligned memory accesses. The default is set depending
24062 on whether the processor we are optimizing for supports fast unaligned access
24065 @item -mcmodel=medlow
24066 @opindex mcmodel=medlow
24067 Generate code for the medium-low code model. The program and its statically
24068 defined symbols must lie within a single 2 GiB address range and must lie
24069 between absolute addresses @minus{}2 GiB and +2 GiB. Programs can be
24070 statically or dynamically linked. This is the default code model.
24072 @item -mcmodel=medany
24073 @opindex mcmodel=medany
24074 Generate code for the medium-any code model. The program and its statically
24075 defined symbols must be within any single 2 GiB address range. Programs can be
24076 statically or dynamically linked.
24078 @item -mexplicit-relocs
24079 @itemx -mno-exlicit-relocs
24080 Use or do not use assembler relocation operators when dealing with symbolic
24081 addresses. The alternative is to use assembler macros instead, which may
24082 limit optimization.
24086 Take advantage of linker relaxations to reduce the number of instructions
24087 required to materialize symbol addresses. The default is to take advantage of
24088 linker relaxations.
24090 @item -memit-attribute
24091 @itemx -mno-emit-attribute
24092 Emit (do not emit) RISC-V attribute to record extra information into ELF
24093 objects. This feature requires at least binutils 2.32.
24095 @item -malign-data=@var{type}
24096 @opindex malign-data
24097 Control how GCC aligns variables and constants of array, structure, or union
24098 types. Supported values for @var{type} are @samp{xlen} which uses x register
24099 width as the alignment value, and @samp{natural} which uses natural alignment.
24100 @samp{xlen} is the default.
24104 @subsection RL78 Options
24105 @cindex RL78 Options
24111 Links in additional target libraries to support operation within a
24120 Specifies the type of hardware multiplication and division support to
24121 be used. The simplest is @code{none}, which uses software for both
24122 multiplication and division. This is the default. The @code{g13}
24123 value is for the hardware multiply/divide peripheral found on the
24124 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
24125 the multiplication and division instructions supported by the RL78/G14
24126 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
24127 the value @code{mg10} is an alias for @code{none}.
24129 In addition a C preprocessor macro is defined, based upon the setting
24130 of this option. Possible values are: @code{__RL78_MUL_NONE__},
24131 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
24138 Specifies the RL78 core to target. The default is the G14 core, also
24139 known as an S3 core or just RL78. The G13 or S2 core does not have
24140 multiply or divide instructions, instead it uses a hardware peripheral
24141 for these operations. The G10 or S1 core does not have register
24142 banks, so it uses a different calling convention.
24144 If this option is set it also selects the type of hardware multiply
24145 support to use, unless this is overridden by an explicit
24146 @option{-mmul=none} option on the command line. Thus specifying
24147 @option{-mcpu=g13} enables the use of the G13 hardware multiply
24148 peripheral and specifying @option{-mcpu=g10} disables the use of
24149 hardware multiplications altogether.
24151 Note, although the RL78/G14 core is the default target, specifying
24152 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
24153 change the behavior of the toolchain since it also enables G14
24154 hardware multiply support. If these options are not specified on the
24155 command line then software multiplication routines will be used even
24156 though the code targets the RL78 core. This is for backwards
24157 compatibility with older toolchains which did not have hardware
24158 multiply and divide support.
24160 In addition a C preprocessor macro is defined, based upon the setting
24161 of this option. Possible values are: @code{__RL78_G10__},
24162 @code{__RL78_G13__} or @code{__RL78_G14__}.
24172 These are aliases for the corresponding @option{-mcpu=} option. They
24173 are provided for backwards compatibility.
24177 Allow the compiler to use all of the available registers. By default
24178 registers @code{r24..r31} are reserved for use in interrupt handlers.
24179 With this option enabled these registers can be used in ordinary
24182 @item -m64bit-doubles
24183 @itemx -m32bit-doubles
24184 @opindex m64bit-doubles
24185 @opindex m32bit-doubles
24186 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
24187 or 32 bits (@option{-m32bit-doubles}) in size. The default is
24188 @option{-m32bit-doubles}.
24190 @item -msave-mduc-in-interrupts
24191 @itemx -mno-save-mduc-in-interrupts
24192 @opindex msave-mduc-in-interrupts
24193 @opindex mno-save-mduc-in-interrupts
24194 Specifies that interrupt handler functions should preserve the
24195 MDUC registers. This is only necessary if normal code might use
24196 the MDUC registers, for example because it performs multiplication
24197 and division operations. The default is to ignore the MDUC registers
24198 as this makes the interrupt handlers faster. The target option -mg13
24199 needs to be passed for this to work as this feature is only available
24200 on the G13 target (S2 core). The MDUC registers will only be saved
24201 if the interrupt handler performs a multiplication or division
24202 operation or it calls another function.
24206 @node RS/6000 and PowerPC Options
24207 @subsection IBM RS/6000 and PowerPC Options
24208 @cindex RS/6000 and PowerPC Options
24209 @cindex IBM RS/6000 and PowerPC Options
24211 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
24213 @item -mpowerpc-gpopt
24214 @itemx -mno-powerpc-gpopt
24215 @itemx -mpowerpc-gfxopt
24216 @itemx -mno-powerpc-gfxopt
24219 @itemx -mno-powerpc64
24223 @itemx -mno-popcntb
24225 @itemx -mno-popcntd
24232 @itemx -mno-hard-dfp
24233 @opindex mpowerpc-gpopt
24234 @opindex mno-powerpc-gpopt
24235 @opindex mpowerpc-gfxopt
24236 @opindex mno-powerpc-gfxopt
24237 @opindex mpowerpc64
24238 @opindex mno-powerpc64
24242 @opindex mno-popcntb
24244 @opindex mno-popcntd
24250 @opindex mno-hard-dfp
24251 You use these options to specify which instructions are available on the
24252 processor you are using. The default value of these options is
24253 determined when configuring GCC@. Specifying the
24254 @option{-mcpu=@var{cpu_type}} overrides the specification of these
24255 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
24256 rather than the options listed above.
24258 Specifying @option{-mpowerpc-gpopt} allows
24259 GCC to use the optional PowerPC architecture instructions in the
24260 General Purpose group, including floating-point square root. Specifying
24261 @option{-mpowerpc-gfxopt} allows GCC to
24262 use the optional PowerPC architecture instructions in the Graphics
24263 group, including floating-point select.
24265 The @option{-mmfcrf} option allows GCC to generate the move from
24266 condition register field instruction implemented on the POWER4
24267 processor and other processors that support the PowerPC V2.01
24269 The @option{-mpopcntb} option allows GCC to generate the popcount and
24270 double-precision FP reciprocal estimate instruction implemented on the
24271 POWER5 processor and other processors that support the PowerPC V2.02
24273 The @option{-mpopcntd} option allows GCC to generate the popcount
24274 instruction implemented on the POWER7 processor and other processors
24275 that support the PowerPC V2.06 architecture.
24276 The @option{-mfprnd} option allows GCC to generate the FP round to
24277 integer instructions implemented on the POWER5+ processor and other
24278 processors that support the PowerPC V2.03 architecture.
24279 The @option{-mcmpb} option allows GCC to generate the compare bytes
24280 instruction implemented on the POWER6 processor and other processors
24281 that support the PowerPC V2.05 architecture.
24282 The @option{-mhard-dfp} option allows GCC to generate the decimal
24283 floating-point instructions implemented on some POWER processors.
24285 The @option{-mpowerpc64} option allows GCC to generate the additional
24286 64-bit instructions that are found in the full PowerPC64 architecture
24287 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
24288 @option{-mno-powerpc64}.
24290 @item -mcpu=@var{cpu_type}
24292 Set architecture type, register usage, and
24293 instruction scheduling parameters for machine type @var{cpu_type}.
24294 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
24295 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
24296 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
24297 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
24298 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
24299 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
24300 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
24301 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
24302 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
24303 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
24304 @samp{power9}, @samp{future}, @samp{powerpc}, @samp{powerpc64},
24305 @samp{powerpc64le}, @samp{rs64}, and @samp{native}.
24307 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
24308 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
24309 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
24310 architecture machine types, with an appropriate, generic processor
24311 model assumed for scheduling purposes.
24313 Specifying @samp{native} as cpu type detects and selects the
24314 architecture option that corresponds to the host processor of the
24315 system performing the compilation.
24316 @option{-mcpu=native} has no effect if GCC does not recognize the
24319 The other options specify a specific processor. Code generated under
24320 those options runs best on that processor, and may not run at all on
24323 The @option{-mcpu} options automatically enable or disable the
24326 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
24327 -mpopcntb -mpopcntd -mpowerpc64 @gol
24328 -mpowerpc-gpopt -mpowerpc-gfxopt @gol
24329 -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
24330 -mcrypto -mhtm -mpower8-fusion -mpower8-vector @gol
24331 -mquad-memory -mquad-memory-atomic -mfloat128 -mfloat128-hardware}
24333 The particular options set for any particular CPU varies between
24334 compiler versions, depending on what setting seems to produce optimal
24335 code for that CPU; it doesn't necessarily reflect the actual hardware's
24336 capabilities. If you wish to set an individual option to a particular
24337 value, you may specify it after the @option{-mcpu} option, like
24338 @option{-mcpu=970 -mno-altivec}.
24340 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
24341 not enabled or disabled by the @option{-mcpu} option at present because
24342 AIX does not have full support for these options. You may still
24343 enable or disable them individually if you're sure it'll work in your
24346 @item -mtune=@var{cpu_type}
24348 Set the instruction scheduling parameters for machine type
24349 @var{cpu_type}, but do not set the architecture type or register usage,
24350 as @option{-mcpu=@var{cpu_type}} does. The same
24351 values for @var{cpu_type} are used for @option{-mtune} as for
24352 @option{-mcpu}. If both are specified, the code generated uses the
24353 architecture and registers set by @option{-mcpu}, but the
24354 scheduling parameters set by @option{-mtune}.
24356 @item -mcmodel=small
24357 @opindex mcmodel=small
24358 Generate PowerPC64 code for the small model: The TOC is limited to
24361 @item -mcmodel=medium
24362 @opindex mcmodel=medium
24363 Generate PowerPC64 code for the medium model: The TOC and other static
24364 data may be up to a total of 4G in size. This is the default for 64-bit
24367 @item -mcmodel=large
24368 @opindex mcmodel=large
24369 Generate PowerPC64 code for the large model: The TOC may be up to 4G
24370 in size. Other data and code is only limited by the 64-bit address
24374 @itemx -mno-altivec
24376 @opindex mno-altivec
24377 Generate code that uses (does not use) AltiVec instructions, and also
24378 enable the use of built-in functions that allow more direct access to
24379 the AltiVec instruction set. You may also need to set
24380 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
24383 When @option{-maltivec} is used, the element order for AltiVec intrinsics
24384 such as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
24385 match array element order corresponding to the endianness of the
24386 target. That is, element zero identifies the leftmost element in a
24387 vector register when targeting a big-endian platform, and identifies
24388 the rightmost element in a vector register when targeting a
24389 little-endian platform.
24394 @opindex mno-vrsave
24395 Generate VRSAVE instructions when generating AltiVec code.
24398 @opindex msecure-plt
24399 Generate code that allows @command{ld} and @command{ld.so}
24400 to build executables and shared
24401 libraries with non-executable @code{.plt} and @code{.got} sections.
24403 32-bit SYSV ABI option.
24407 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
24409 requires @code{.plt} and @code{.got}
24410 sections that are both writable and executable.
24411 This is a PowerPC 32-bit SYSV ABI option.
24417 This switch enables or disables the generation of ISEL instructions.
24423 Generate code that uses (does not use) vector/scalar (VSX)
24424 instructions, and also enable the use of built-in functions that allow
24425 more direct access to the VSX instruction set.
24430 @opindex mno-crypto
24431 Enable the use (disable) of the built-in functions that allow direct
24432 access to the cryptographic instructions that were added in version
24433 2.07 of the PowerPC ISA.
24439 Enable (disable) the use of the built-in functions that allow direct
24440 access to the Hardware Transactional Memory (HTM) instructions that
24441 were added in version 2.07 of the PowerPC ISA.
24443 @item -mpower8-fusion
24444 @itemx -mno-power8-fusion
24445 @opindex mpower8-fusion
24446 @opindex mno-power8-fusion
24447 Generate code that keeps (does not keeps) some integer operations
24448 adjacent so that the instructions can be fused together on power8 and
24451 @item -mpower8-vector
24452 @itemx -mno-power8-vector
24453 @opindex mpower8-vector
24454 @opindex mno-power8-vector
24455 Generate code that uses (does not use) the vector and scalar
24456 instructions that were added in version 2.07 of the PowerPC ISA. Also
24457 enable the use of built-in functions that allow more direct access to
24458 the vector instructions.
24460 @item -mquad-memory
24461 @itemx -mno-quad-memory
24462 @opindex mquad-memory
24463 @opindex mno-quad-memory
24464 Generate code that uses (does not use) the non-atomic quad word memory
24465 instructions. The @option{-mquad-memory} option requires use of
24468 @item -mquad-memory-atomic
24469 @itemx -mno-quad-memory-atomic
24470 @opindex mquad-memory-atomic
24471 @opindex mno-quad-memory-atomic
24472 Generate code that uses (does not use) the atomic quad word memory
24473 instructions. The @option{-mquad-memory-atomic} option requires use of
24477 @itemx -mno-float128
24479 @opindex mno-float128
24480 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
24481 and use either software emulation for IEEE 128-bit floating point or
24482 hardware instructions.
24484 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7},
24485 @option{-mcpu=power8}), or @option{-mcpu=power9} must be enabled to
24486 use the IEEE 128-bit floating point support. The IEEE 128-bit
24487 floating point support only works on PowerPC Linux systems.
24489 The default for @option{-mfloat128} is enabled on PowerPC Linux
24490 systems using the VSX instruction set, and disabled on other systems.
24492 If you use the ISA 3.0 instruction set (@option{-mpower9-vector} or
24493 @option{-mcpu=power9}) on a 64-bit system, the IEEE 128-bit floating
24494 point support will also enable the generation of ISA 3.0 IEEE 128-bit
24495 floating point instructions. Otherwise, if you do not specify to
24496 generate ISA 3.0 instructions or you are targeting a 32-bit big endian
24497 system, IEEE 128-bit floating point will be done with software
24500 @item -mfloat128-hardware
24501 @itemx -mno-float128-hardware
24502 @opindex mfloat128-hardware
24503 @opindex mno-float128-hardware
24504 Enable/disable using ISA 3.0 hardware instructions to support the
24505 @var{__float128} data type.
24507 The default for @option{-mfloat128-hardware} is enabled on PowerPC
24508 Linux systems using the ISA 3.0 instruction set, and disabled on other
24515 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
24516 targets (including GNU/Linux). The 32-bit environment sets int, long
24517 and pointer to 32 bits and generates code that runs on any PowerPC
24518 variant. The 64-bit environment sets int to 32 bits and long and
24519 pointer to 64 bits, and generates code for PowerPC64, as for
24520 @option{-mpowerpc64}.
24523 @itemx -mno-fp-in-toc
24524 @itemx -mno-sum-in-toc
24525 @itemx -mminimal-toc
24527 @opindex mno-fp-in-toc
24528 @opindex mno-sum-in-toc
24529 @opindex mminimal-toc
24530 Modify generation of the TOC (Table Of Contents), which is created for
24531 every executable file. The @option{-mfull-toc} option is selected by
24532 default. In that case, GCC allocates at least one TOC entry for
24533 each unique non-automatic variable reference in your program. GCC
24534 also places floating-point constants in the TOC@. However, only
24535 16,384 entries are available in the TOC@.
24537 If you receive a linker error message that saying you have overflowed
24538 the available TOC space, you can reduce the amount of TOC space used
24539 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
24540 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
24541 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
24542 generate code to calculate the sum of an address and a constant at
24543 run time instead of putting that sum into the TOC@. You may specify one
24544 or both of these options. Each causes GCC to produce very slightly
24545 slower and larger code at the expense of conserving TOC space.
24547 If you still run out of space in the TOC even when you specify both of
24548 these options, specify @option{-mminimal-toc} instead. This option causes
24549 GCC to make only one TOC entry for every file. When you specify this
24550 option, GCC produces code that is slower and larger but which
24551 uses extremely little TOC space. You may wish to use this option
24552 only on files that contain less frequently-executed code.
24558 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
24559 @code{long} type, and the infrastructure needed to support them.
24560 Specifying @option{-maix64} implies @option{-mpowerpc64},
24561 while @option{-maix32} disables the 64-bit ABI and
24562 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
24565 @itemx -mno-xl-compat
24566 @opindex mxl-compat
24567 @opindex mno-xl-compat
24568 Produce code that conforms more closely to IBM XL compiler semantics
24569 when using AIX-compatible ABI@. Pass floating-point arguments to
24570 prototyped functions beyond the register save area (RSA) on the stack
24571 in addition to argument FPRs. Do not assume that most significant
24572 double in 128-bit long double value is properly rounded when comparing
24573 values and converting to double. Use XL symbol names for long double
24576 The AIX calling convention was extended but not initially documented to
24577 handle an obscure K&R C case of calling a function that takes the
24578 address of its arguments with fewer arguments than declared. IBM XL
24579 compilers access floating-point arguments that do not fit in the
24580 RSA from the stack when a subroutine is compiled without
24581 optimization. Because always storing floating-point arguments on the
24582 stack is inefficient and rarely needed, this option is not enabled by
24583 default and only is necessary when calling subroutines compiled by IBM
24584 XL compilers without optimization.
24588 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
24589 application written to use message passing with special startup code to
24590 enable the application to run. The system must have PE installed in the
24591 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
24592 must be overridden with the @option{-specs=} option to specify the
24593 appropriate directory location. The Parallel Environment does not
24594 support threads, so the @option{-mpe} option and the @option{-pthread}
24595 option are incompatible.
24597 @item -malign-natural
24598 @itemx -malign-power
24599 @opindex malign-natural
24600 @opindex malign-power
24601 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
24602 @option{-malign-natural} overrides the ABI-defined alignment of larger
24603 types, such as floating-point doubles, on their natural size-based boundary.
24604 The option @option{-malign-power} instructs GCC to follow the ABI-specified
24605 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
24607 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
24611 @itemx -mhard-float
24612 @opindex msoft-float
24613 @opindex mhard-float
24614 Generate code that does not use (uses) the floating-point register set.
24615 Software floating-point emulation is provided if you use the
24616 @option{-msoft-float} option, and pass the option to GCC when linking.
24619 @itemx -mno-multiple
24621 @opindex mno-multiple
24622 Generate code that uses (does not use) the load multiple word
24623 instructions and the store multiple word instructions. These
24624 instructions are generated by default on POWER systems, and not
24625 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
24626 PowerPC systems, since those instructions do not work when the
24627 processor is in little-endian mode. The exceptions are PPC740 and
24628 PPC750 which permit these instructions in little-endian mode.
24633 @opindex mno-update
24634 Generate code that uses (does not use) the load or store instructions
24635 that update the base register to the address of the calculated memory
24636 location. These instructions are generated by default. If you use
24637 @option{-mno-update}, there is a small window between the time that the
24638 stack pointer is updated and the address of the previous frame is
24639 stored, which means code that walks the stack frame across interrupts or
24640 signals may get corrupted data.
24642 @item -mavoid-indexed-addresses
24643 @itemx -mno-avoid-indexed-addresses
24644 @opindex mavoid-indexed-addresses
24645 @opindex mno-avoid-indexed-addresses
24646 Generate code that tries to avoid (not avoid) the use of indexed load
24647 or store instructions. These instructions can incur a performance
24648 penalty on Power6 processors in certain situations, such as when
24649 stepping through large arrays that cross a 16M boundary. This option
24650 is enabled by default when targeting Power6 and disabled otherwise.
24653 @itemx -mno-fused-madd
24654 @opindex mfused-madd
24655 @opindex mno-fused-madd
24656 Generate code that uses (does not use) the floating-point multiply and
24657 accumulate instructions. These instructions are generated by default
24658 if hardware floating point is used. The machine-dependent
24659 @option{-mfused-madd} option is now mapped to the machine-independent
24660 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
24661 mapped to @option{-ffp-contract=off}.
24667 Generate code that uses (does not use) the half-word multiply and
24668 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
24669 These instructions are generated by default when targeting those
24676 Generate code that uses (does not use) the string-search @samp{dlmzb}
24677 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
24678 generated by default when targeting those processors.
24680 @item -mno-bit-align
24682 @opindex mno-bit-align
24683 @opindex mbit-align
24684 On System V.4 and embedded PowerPC systems do not (do) force structures
24685 and unions that contain bit-fields to be aligned to the base type of the
24688 For example, by default a structure containing nothing but 8
24689 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
24690 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
24691 the structure is aligned to a 1-byte boundary and is 1 byte in
24694 @item -mno-strict-align
24695 @itemx -mstrict-align
24696 @opindex mno-strict-align
24697 @opindex mstrict-align
24698 On System V.4 and embedded PowerPC systems do not (do) assume that
24699 unaligned memory references are handled by the system.
24701 @item -mrelocatable
24702 @itemx -mno-relocatable
24703 @opindex mrelocatable
24704 @opindex mno-relocatable
24705 Generate code that allows (does not allow) a static executable to be
24706 relocated to a different address at run time. A simple embedded
24707 PowerPC system loader should relocate the entire contents of
24708 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
24709 a table of 32-bit addresses generated by this option. For this to
24710 work, all objects linked together must be compiled with
24711 @option{-mrelocatable} or @option{-mrelocatable-lib}.
24712 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
24714 @item -mrelocatable-lib
24715 @itemx -mno-relocatable-lib
24716 @opindex mrelocatable-lib
24717 @opindex mno-relocatable-lib
24718 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
24719 @code{.fixup} section to allow static executables to be relocated at
24720 run time, but @option{-mrelocatable-lib} does not use the smaller stack
24721 alignment of @option{-mrelocatable}. Objects compiled with
24722 @option{-mrelocatable-lib} may be linked with objects compiled with
24723 any combination of the @option{-mrelocatable} options.
24729 On System V.4 and embedded PowerPC systems do not (do) assume that
24730 register 2 contains a pointer to a global area pointing to the addresses
24731 used in the program.
24734 @itemx -mlittle-endian
24736 @opindex mlittle-endian
24737 On System V.4 and embedded PowerPC systems compile code for the
24738 processor in little-endian mode. The @option{-mlittle-endian} option is
24739 the same as @option{-mlittle}.
24742 @itemx -mbig-endian
24744 @opindex mbig-endian
24745 On System V.4 and embedded PowerPC systems compile code for the
24746 processor in big-endian mode. The @option{-mbig-endian} option is
24747 the same as @option{-mbig}.
24749 @item -mdynamic-no-pic
24750 @opindex mdynamic-no-pic
24751 On Darwin and Mac OS X systems, compile code so that it is not
24752 relocatable, but that its external references are relocatable. The
24753 resulting code is suitable for applications, but not shared
24756 @item -msingle-pic-base
24757 @opindex msingle-pic-base
24758 Treat the register used for PIC addressing as read-only, rather than
24759 loading it in the prologue for each function. The runtime system is
24760 responsible for initializing this register with an appropriate value
24761 before execution begins.
24763 @item -mprioritize-restricted-insns=@var{priority}
24764 @opindex mprioritize-restricted-insns
24765 This option controls the priority that is assigned to
24766 dispatch-slot restricted instructions during the second scheduling
24767 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
24768 or @samp{2} to assign no, highest, or second-highest (respectively)
24769 priority to dispatch-slot restricted
24772 @item -msched-costly-dep=@var{dependence_type}
24773 @opindex msched-costly-dep
24774 This option controls which dependences are considered costly
24775 by the target during instruction scheduling. The argument
24776 @var{dependence_type} takes one of the following values:
24780 No dependence is costly.
24783 All dependences are costly.
24785 @item @samp{true_store_to_load}
24786 A true dependence from store to load is costly.
24788 @item @samp{store_to_load}
24789 Any dependence from store to load is costly.
24792 Any dependence for which the latency is greater than or equal to
24793 @var{number} is costly.
24796 @item -minsert-sched-nops=@var{scheme}
24797 @opindex minsert-sched-nops
24798 This option controls which NOP insertion scheme is used during
24799 the second scheduling pass. The argument @var{scheme} takes one of the
24807 Pad with NOPs any dispatch group that has vacant issue slots,
24808 according to the scheduler's grouping.
24810 @item @samp{regroup_exact}
24811 Insert NOPs to force costly dependent insns into
24812 separate groups. Insert exactly as many NOPs as needed to force an insn
24813 to a new group, according to the estimated processor grouping.
24816 Insert NOPs to force costly dependent insns into
24817 separate groups. Insert @var{number} NOPs to force an insn to a new group.
24821 @opindex mcall-sysv
24822 On System V.4 and embedded PowerPC systems compile code using calling
24823 conventions that adhere to the March 1995 draft of the System V
24824 Application Binary Interface, PowerPC processor supplement. This is the
24825 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
24827 @item -mcall-sysv-eabi
24829 @opindex mcall-sysv-eabi
24830 @opindex mcall-eabi
24831 Specify both @option{-mcall-sysv} and @option{-meabi} options.
24833 @item -mcall-sysv-noeabi
24834 @opindex mcall-sysv-noeabi
24835 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
24837 @item -mcall-aixdesc
24839 On System V.4 and embedded PowerPC systems compile code for the AIX
24843 @opindex mcall-linux
24844 On System V.4 and embedded PowerPC systems compile code for the
24845 Linux-based GNU system.
24847 @item -mcall-freebsd
24848 @opindex mcall-freebsd
24849 On System V.4 and embedded PowerPC systems compile code for the
24850 FreeBSD operating system.
24852 @item -mcall-netbsd
24853 @opindex mcall-netbsd
24854 On System V.4 and embedded PowerPC systems compile code for the
24855 NetBSD operating system.
24857 @item -mcall-openbsd
24858 @opindex mcall-netbsd
24859 On System V.4 and embedded PowerPC systems compile code for the
24860 OpenBSD operating system.
24862 @item -mtraceback=@var{traceback_type}
24863 @opindex mtraceback
24864 Select the type of traceback table. Valid values for @var{traceback_type}
24865 are @samp{full}, @samp{part}, and @samp{no}.
24867 @item -maix-struct-return
24868 @opindex maix-struct-return
24869 Return all structures in memory (as specified by the AIX ABI)@.
24871 @item -msvr4-struct-return
24872 @opindex msvr4-struct-return
24873 Return structures smaller than 8 bytes in registers (as specified by the
24876 @item -mabi=@var{abi-type}
24878 Extend the current ABI with a particular extension, or remove such extension.
24879 Valid values are @samp{altivec}, @samp{no-altivec},
24880 @samp{ibmlongdouble}, @samp{ieeelongdouble},
24881 @samp{elfv1}, @samp{elfv2}@.
24883 @item -mabi=ibmlongdouble
24884 @opindex mabi=ibmlongdouble
24885 Change the current ABI to use IBM extended-precision long double.
24886 This is not likely to work if your system defaults to using IEEE
24887 extended-precision long double. If you change the long double type
24888 from IEEE extended-precision, the compiler will issue a warning unless
24889 you use the @option{-Wno-psabi} option. Requires @option{-mlong-double-128}
24892 @item -mabi=ieeelongdouble
24893 @opindex mabi=ieeelongdouble
24894 Change the current ABI to use IEEE extended-precision long double.
24895 This is not likely to work if your system defaults to using IBM
24896 extended-precision long double. If you change the long double type
24897 from IBM extended-precision, the compiler will issue a warning unless
24898 you use the @option{-Wno-psabi} option. Requires @option{-mlong-double-128}
24902 @opindex mabi=elfv1
24903 Change the current ABI to use the ELFv1 ABI.
24904 This is the default ABI for big-endian PowerPC 64-bit Linux.
24905 Overriding the default ABI requires special system support and is
24906 likely to fail in spectacular ways.
24909 @opindex mabi=elfv2
24910 Change the current ABI to use the ELFv2 ABI.
24911 This is the default ABI for little-endian PowerPC 64-bit Linux.
24912 Overriding the default ABI requires special system support and is
24913 likely to fail in spectacular ways.
24915 @item -mgnu-attribute
24916 @itemx -mno-gnu-attribute
24917 @opindex mgnu-attribute
24918 @opindex mno-gnu-attribute
24919 Emit .gnu_attribute assembly directives to set tag/value pairs in a
24920 .gnu.attributes section that specify ABI variations in function
24921 parameters or return values.
24924 @itemx -mno-prototype
24925 @opindex mprototype
24926 @opindex mno-prototype
24927 On System V.4 and embedded PowerPC systems assume that all calls to
24928 variable argument functions are properly prototyped. Otherwise, the
24929 compiler must insert an instruction before every non-prototyped call to
24930 set or clear bit 6 of the condition code register (@code{CR}) to
24931 indicate whether floating-point values are passed in the floating-point
24932 registers in case the function takes variable arguments. With
24933 @option{-mprototype}, only calls to prototyped variable argument functions
24934 set or clear the bit.
24938 On embedded PowerPC systems, assume that the startup module is called
24939 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
24940 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
24945 On embedded PowerPC systems, assume that the startup module is called
24946 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
24951 On embedded PowerPC systems, assume that the startup module is called
24952 @file{crt0.o} and the standard C libraries are @file{libads.a} and
24955 @item -myellowknife
24956 @opindex myellowknife
24957 On embedded PowerPC systems, assume that the startup module is called
24958 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
24963 On System V.4 and embedded PowerPC systems, specify that you are
24964 compiling for a VxWorks system.
24968 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
24969 header to indicate that @samp{eabi} extended relocations are used.
24975 On System V.4 and embedded PowerPC systems do (do not) adhere to the
24976 Embedded Applications Binary Interface (EABI), which is a set of
24977 modifications to the System V.4 specifications. Selecting @option{-meabi}
24978 means that the stack is aligned to an 8-byte boundary, a function
24979 @code{__eabi} is called from @code{main} to set up the EABI
24980 environment, and the @option{-msdata} option can use both @code{r2} and
24981 @code{r13} to point to two separate small data areas. Selecting
24982 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
24983 no EABI initialization function is called from @code{main}, and the
24984 @option{-msdata} option only uses @code{r13} to point to a single
24985 small data area. The @option{-meabi} option is on by default if you
24986 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
24989 @opindex msdata=eabi
24990 On System V.4 and embedded PowerPC systems, put small initialized
24991 @code{const} global and static data in the @code{.sdata2} section, which
24992 is pointed to by register @code{r2}. Put small initialized
24993 non-@code{const} global and static data in the @code{.sdata} section,
24994 which is pointed to by register @code{r13}. Put small uninitialized
24995 global and static data in the @code{.sbss} section, which is adjacent to
24996 the @code{.sdata} section. The @option{-msdata=eabi} option is
24997 incompatible with the @option{-mrelocatable} option. The
24998 @option{-msdata=eabi} option also sets the @option{-memb} option.
25001 @opindex msdata=sysv
25002 On System V.4 and embedded PowerPC systems, put small global and static
25003 data in the @code{.sdata} section, which is pointed to by register
25004 @code{r13}. Put small uninitialized global and static data in the
25005 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
25006 The @option{-msdata=sysv} option is incompatible with the
25007 @option{-mrelocatable} option.
25009 @item -msdata=default
25011 @opindex msdata=default
25013 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
25014 compile code the same as @option{-msdata=eabi}, otherwise compile code the
25015 same as @option{-msdata=sysv}.
25018 @opindex msdata=data
25019 On System V.4 and embedded PowerPC systems, put small global
25020 data in the @code{.sdata} section. Put small uninitialized global
25021 data in the @code{.sbss} section. Do not use register @code{r13}
25022 to address small data however. This is the default behavior unless
25023 other @option{-msdata} options are used.
25027 @opindex msdata=none
25029 On embedded PowerPC systems, put all initialized global and static data
25030 in the @code{.data} section, and all uninitialized data in the
25031 @code{.bss} section.
25033 @item -mreadonly-in-sdata
25034 @opindex mreadonly-in-sdata
25035 @opindex mno-readonly-in-sdata
25036 Put read-only objects in the @code{.sdata} section as well. This is the
25039 @item -mblock-move-inline-limit=@var{num}
25040 @opindex mblock-move-inline-limit
25041 Inline all block moves (such as calls to @code{memcpy} or structure
25042 copies) less than or equal to @var{num} bytes. The minimum value for
25043 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
25044 targets. The default value is target-specific.
25046 @item -mblock-compare-inline-limit=@var{num}
25047 @opindex mblock-compare-inline-limit
25048 Generate non-looping inline code for all block compares (such as calls
25049 to @code{memcmp} or structure compares) less than or equal to @var{num}
25050 bytes. If @var{num} is 0, all inline expansion (non-loop and loop) of
25051 block compare is disabled. The default value is target-specific.
25053 @item -mblock-compare-inline-loop-limit=@var{num}
25054 @opindex mblock-compare-inline-loop-limit
25055 Generate an inline expansion using loop code for all block compares that
25056 are less than or equal to @var{num} bytes, but greater than the limit
25057 for non-loop inline block compare expansion. If the block length is not
25058 constant, at most @var{num} bytes will be compared before @code{memcmp}
25059 is called to compare the remainder of the block. The default value is
25062 @item -mstring-compare-inline-limit=@var{num}
25063 @opindex mstring-compare-inline-limit
25064 Compare at most @var{num} string bytes with inline code.
25065 If the difference or end of string is not found at the
25066 end of the inline compare a call to @code{strcmp} or @code{strncmp} will
25067 take care of the rest of the comparison. The default is 64 bytes.
25071 @cindex smaller data references (PowerPC)
25072 @cindex .sdata/.sdata2 references (PowerPC)
25073 On embedded PowerPC systems, put global and static items less than or
25074 equal to @var{num} bytes into the small data or BSS sections instead of
25075 the normal data or BSS section. By default, @var{num} is 8. The
25076 @option{-G @var{num}} switch is also passed to the linker.
25077 All modules should be compiled with the same @option{-G @var{num}} value.
25080 @itemx -mno-regnames
25082 @opindex mno-regnames
25083 On System V.4 and embedded PowerPC systems do (do not) emit register
25084 names in the assembly language output using symbolic forms.
25087 @itemx -mno-longcall
25089 @opindex mno-longcall
25090 By default assume that all calls are far away so that a longer and more
25091 expensive calling sequence is required. This is required for calls
25092 farther than 32 megabytes (33,554,432 bytes) from the current location.
25093 A short call is generated if the compiler knows
25094 the call cannot be that far away. This setting can be overridden by
25095 the @code{shortcall} function attribute, or by @code{#pragma
25098 Some linkers are capable of detecting out-of-range calls and generating
25099 glue code on the fly. On these systems, long calls are unnecessary and
25100 generate slower code. As of this writing, the AIX linker can do this,
25101 as can the GNU linker for PowerPC/64. It is planned to add this feature
25102 to the GNU linker for 32-bit PowerPC systems as well.
25104 On PowerPC64 ELFv2 and 32-bit PowerPC systems with newer GNU linkers,
25105 GCC can generate long calls using an inline PLT call sequence (see
25106 @option{-mpltseq}). PowerPC with @option{-mbss-plt} and PowerPC64
25107 ELFv1 (big-endian) do not support inline PLT calls.
25109 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
25110 callee, L42}, plus a @dfn{branch island} (glue code). The two target
25111 addresses represent the callee and the branch island. The
25112 Darwin/PPC linker prefers the first address and generates a @code{bl
25113 callee} if the PPC @code{bl} instruction reaches the callee directly;
25114 otherwise, the linker generates @code{bl L42} to call the branch
25115 island. The branch island is appended to the body of the
25116 calling function; it computes the full 32-bit address of the callee
25119 On Mach-O (Darwin) systems, this option directs the compiler emit to
25120 the glue for every direct call, and the Darwin linker decides whether
25121 to use or discard it.
25123 In the future, GCC may ignore all longcall specifications
25124 when the linker is known to generate glue.
25129 @opindex mno-pltseq
25130 Implement (do not implement) -fno-plt and long calls using an inline
25131 PLT call sequence that supports lazy linking and long calls to
25132 functions in dlopen'd shared libraries. Inline PLT calls are only
25133 supported on PowerPC64 ELFv2 and 32-bit PowerPC systems with newer GNU
25134 linkers, and are enabled by default if the support is detected when
25135 configuring GCC, and, in the case of 32-bit PowerPC, if GCC is
25136 configured with @option{--enable-secureplt}. @option{-mpltseq} code
25137 and @option{-mbss-plt} 32-bit PowerPC relocatable objects may not be
25140 @item -mtls-markers
25141 @itemx -mno-tls-markers
25142 @opindex mtls-markers
25143 @opindex mno-tls-markers
25144 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
25145 specifying the function argument. The relocation allows the linker to
25146 reliably associate function call with argument setup instructions for
25147 TLS optimization, which in turn allows GCC to better schedule the
25153 This option enables use of the reciprocal estimate and
25154 reciprocal square root estimate instructions with additional
25155 Newton-Raphson steps to increase precision instead of doing a divide or
25156 square root and divide for floating-point arguments. You should use
25157 the @option{-ffast-math} option when using @option{-mrecip} (or at
25158 least @option{-funsafe-math-optimizations},
25159 @option{-ffinite-math-only}, @option{-freciprocal-math} and
25160 @option{-fno-trapping-math}). Note that while the throughput of the
25161 sequence is generally higher than the throughput of the non-reciprocal
25162 instruction, the precision of the sequence can be decreased by up to 2
25163 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
25166 @item -mrecip=@var{opt}
25167 @opindex mrecip=opt
25168 This option controls which reciprocal estimate instructions
25169 may be used. @var{opt} is a comma-separated list of options, which may
25170 be preceded by a @code{!} to invert the option:
25175 Enable all estimate instructions.
25178 Enable the default instructions, equivalent to @option{-mrecip}.
25181 Disable all estimate instructions, equivalent to @option{-mno-recip}.
25184 Enable the reciprocal approximation instructions for both
25185 single and double precision.
25188 Enable the single-precision reciprocal approximation instructions.
25191 Enable the double-precision reciprocal approximation instructions.
25194 Enable the reciprocal square root approximation instructions for both
25195 single and double precision.
25198 Enable the single-precision reciprocal square root approximation instructions.
25201 Enable the double-precision reciprocal square root approximation instructions.
25205 So, for example, @option{-mrecip=all,!rsqrtd} enables
25206 all of the reciprocal estimate instructions, except for the
25207 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
25208 which handle the double-precision reciprocal square root calculations.
25210 @item -mrecip-precision
25211 @itemx -mno-recip-precision
25212 @opindex mrecip-precision
25213 Assume (do not assume) that the reciprocal estimate instructions
25214 provide higher-precision estimates than is mandated by the PowerPC
25215 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
25216 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
25217 The double-precision square root estimate instructions are not generated by
25218 default on low-precision machines, since they do not provide an
25219 estimate that converges after three steps.
25221 @item -mveclibabi=@var{type}
25222 @opindex mveclibabi
25223 Specifies the ABI type to use for vectorizing intrinsics using an
25224 external library. The only type supported at present is @samp{mass},
25225 which specifies to use IBM's Mathematical Acceleration Subsystem
25226 (MASS) libraries for vectorizing intrinsics using external libraries.
25227 GCC currently emits calls to @code{acosd2}, @code{acosf4},
25228 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
25229 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
25230 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
25231 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
25232 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
25233 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
25234 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
25235 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
25236 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
25237 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
25238 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
25239 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
25240 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
25241 for power7. Both @option{-ftree-vectorize} and
25242 @option{-funsafe-math-optimizations} must also be enabled. The MASS
25243 libraries must be specified at link time.
25248 Generate (do not generate) the @code{friz} instruction when the
25249 @option{-funsafe-math-optimizations} option is used to optimize
25250 rounding of floating-point values to 64-bit integer and back to floating
25251 point. The @code{friz} instruction does not return the same value if
25252 the floating-point number is too large to fit in an integer.
25254 @item -mpointers-to-nested-functions
25255 @itemx -mno-pointers-to-nested-functions
25256 @opindex mpointers-to-nested-functions
25257 Generate (do not generate) code to load up the static chain register
25258 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
25259 systems where a function pointer points to a 3-word descriptor giving
25260 the function address, TOC value to be loaded in register @code{r2}, and
25261 static chain value to be loaded in register @code{r11}. The
25262 @option{-mpointers-to-nested-functions} is on by default. You cannot
25263 call through pointers to nested functions or pointers
25264 to functions compiled in other languages that use the static chain if
25265 you use @option{-mno-pointers-to-nested-functions}.
25267 @item -msave-toc-indirect
25268 @itemx -mno-save-toc-indirect
25269 @opindex msave-toc-indirect
25270 Generate (do not generate) code to save the TOC value in the reserved
25271 stack location in the function prologue if the function calls through
25272 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
25273 saved in the prologue, it is saved just before the call through the
25274 pointer. The @option{-mno-save-toc-indirect} option is the default.
25276 @item -mcompat-align-parm
25277 @itemx -mno-compat-align-parm
25278 @opindex mcompat-align-parm
25279 Generate (do not generate) code to pass structure parameters with a
25280 maximum alignment of 64 bits, for compatibility with older versions
25283 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
25284 structure parameter on a 128-bit boundary when that structure contained
25285 a member requiring 128-bit alignment. This is corrected in more
25286 recent versions of GCC. This option may be used to generate code
25287 that is compatible with functions compiled with older versions of
25290 The @option{-mno-compat-align-parm} option is the default.
25292 @item -mstack-protector-guard=@var{guard}
25293 @itemx -mstack-protector-guard-reg=@var{reg}
25294 @itemx -mstack-protector-guard-offset=@var{offset}
25295 @itemx -mstack-protector-guard-symbol=@var{symbol}
25296 @opindex mstack-protector-guard
25297 @opindex mstack-protector-guard-reg
25298 @opindex mstack-protector-guard-offset
25299 @opindex mstack-protector-guard-symbol
25300 Generate stack protection code using canary at @var{guard}. Supported
25301 locations are @samp{global} for global canary or @samp{tls} for per-thread
25302 canary in the TLS block (the default with GNU libc version 2.4 or later).
25304 With the latter choice the options
25305 @option{-mstack-protector-guard-reg=@var{reg}} and
25306 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
25307 which register to use as base register for reading the canary, and from what
25308 offset from that base register. The default for those is as specified in the
25309 relevant ABI. @option{-mstack-protector-guard-symbol=@var{symbol}} overrides
25310 the offset with a symbol reference to a canary in the TLS block.
25316 Generate (do not generate) pc-relative addressing when the option
25317 @option{-mcpu=future} is used.
25321 @subsection RX Options
25324 These command-line options are defined for RX targets:
25327 @item -m64bit-doubles
25328 @itemx -m32bit-doubles
25329 @opindex m64bit-doubles
25330 @opindex m32bit-doubles
25331 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
25332 or 32 bits (@option{-m32bit-doubles}) in size. The default is
25333 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
25334 works on 32-bit values, which is why the default is
25335 @option{-m32bit-doubles}.
25341 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
25342 floating-point hardware. The default is enabled for the RX600
25343 series and disabled for the RX200 series.
25345 Floating-point instructions are only generated for 32-bit floating-point
25346 values, however, so the FPU hardware is not used for doubles if the
25347 @option{-m64bit-doubles} option is used.
25349 @emph{Note} If the @option{-fpu} option is enabled then
25350 @option{-funsafe-math-optimizations} is also enabled automatically.
25351 This is because the RX FPU instructions are themselves unsafe.
25353 @item -mcpu=@var{name}
25355 Selects the type of RX CPU to be targeted. Currently three types are
25356 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
25357 the specific @samp{RX610} CPU. The default is @samp{RX600}.
25359 The only difference between @samp{RX600} and @samp{RX610} is that the
25360 @samp{RX610} does not support the @code{MVTIPL} instruction.
25362 The @samp{RX200} series does not have a hardware floating-point unit
25363 and so @option{-nofpu} is enabled by default when this type is
25366 @item -mbig-endian-data
25367 @itemx -mlittle-endian-data
25368 @opindex mbig-endian-data
25369 @opindex mlittle-endian-data
25370 Store data (but not code) in the big-endian format. The default is
25371 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
25374 @item -msmall-data-limit=@var{N}
25375 @opindex msmall-data-limit
25376 Specifies the maximum size in bytes of global and static variables
25377 which can be placed into the small data area. Using the small data
25378 area can lead to smaller and faster code, but the size of area is
25379 limited and it is up to the programmer to ensure that the area does
25380 not overflow. Also when the small data area is used one of the RX's
25381 registers (usually @code{r13}) is reserved for use pointing to this
25382 area, so it is no longer available for use by the compiler. This
25383 could result in slower and/or larger code if variables are pushed onto
25384 the stack instead of being held in this register.
25386 Note, common variables (variables that have not been initialized) and
25387 constants are not placed into the small data area as they are assigned
25388 to other sections in the output executable.
25390 The default value is zero, which disables this feature. Note, this
25391 feature is not enabled by default with higher optimization levels
25392 (@option{-O2} etc) because of the potentially detrimental effects of
25393 reserving a register. It is up to the programmer to experiment and
25394 discover whether this feature is of benefit to their program. See the
25395 description of the @option{-mpid} option for a description of how the
25396 actual register to hold the small data area pointer is chosen.
25402 Use the simulator runtime. The default is to use the libgloss
25403 board-specific runtime.
25405 @item -mas100-syntax
25406 @itemx -mno-as100-syntax
25407 @opindex mas100-syntax
25408 @opindex mno-as100-syntax
25409 When generating assembler output use a syntax that is compatible with
25410 Renesas's AS100 assembler. This syntax can also be handled by the GAS
25411 assembler, but it has some restrictions so it is not generated by default.
25413 @item -mmax-constant-size=@var{N}
25414 @opindex mmax-constant-size
25415 Specifies the maximum size, in bytes, of a constant that can be used as
25416 an operand in a RX instruction. Although the RX instruction set does
25417 allow constants of up to 4 bytes in length to be used in instructions,
25418 a longer value equates to a longer instruction. Thus in some
25419 circumstances it can be beneficial to restrict the size of constants
25420 that are used in instructions. Constants that are too big are instead
25421 placed into a constant pool and referenced via register indirection.
25423 The value @var{N} can be between 0 and 4. A value of 0 (the default)
25424 or 4 means that constants of any size are allowed.
25428 Enable linker relaxation. Linker relaxation is a process whereby the
25429 linker attempts to reduce the size of a program by finding shorter
25430 versions of various instructions. Disabled by default.
25432 @item -mint-register=@var{N}
25433 @opindex mint-register
25434 Specify the number of registers to reserve for fast interrupt handler
25435 functions. The value @var{N} can be between 0 and 4. A value of 1
25436 means that register @code{r13} is reserved for the exclusive use
25437 of fast interrupt handlers. A value of 2 reserves @code{r13} and
25438 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
25439 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
25440 A value of 0, the default, does not reserve any registers.
25442 @item -msave-acc-in-interrupts
25443 @opindex msave-acc-in-interrupts
25444 Specifies that interrupt handler functions should preserve the
25445 accumulator register. This is only necessary if normal code might use
25446 the accumulator register, for example because it performs 64-bit
25447 multiplications. The default is to ignore the accumulator as this
25448 makes the interrupt handlers faster.
25454 Enables the generation of position independent data. When enabled any
25455 access to constant data is done via an offset from a base address
25456 held in a register. This allows the location of constant data to be
25457 determined at run time without requiring the executable to be
25458 relocated, which is a benefit to embedded applications with tight
25459 memory constraints. Data that can be modified is not affected by this
25462 Note, using this feature reserves a register, usually @code{r13}, for
25463 the constant data base address. This can result in slower and/or
25464 larger code, especially in complicated functions.
25466 The actual register chosen to hold the constant data base address
25467 depends upon whether the @option{-msmall-data-limit} and/or the
25468 @option{-mint-register} command-line options are enabled. Starting
25469 with register @code{r13} and proceeding downwards, registers are
25470 allocated first to satisfy the requirements of @option{-mint-register},
25471 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
25472 is possible for the small data area register to be @code{r8} if both
25473 @option{-mint-register=4} and @option{-mpid} are specified on the
25476 By default this feature is not enabled. The default can be restored
25477 via the @option{-mno-pid} command-line option.
25479 @item -mno-warn-multiple-fast-interrupts
25480 @itemx -mwarn-multiple-fast-interrupts
25481 @opindex mno-warn-multiple-fast-interrupts
25482 @opindex mwarn-multiple-fast-interrupts
25483 Prevents GCC from issuing a warning message if it finds more than one
25484 fast interrupt handler when it is compiling a file. The default is to
25485 issue a warning for each extra fast interrupt handler found, as the RX
25486 only supports one such interrupt.
25488 @item -mallow-string-insns
25489 @itemx -mno-allow-string-insns
25490 @opindex mallow-string-insns
25491 @opindex mno-allow-string-insns
25492 Enables or disables the use of the string manipulation instructions
25493 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
25494 @code{SWHILE} and also the @code{RMPA} instruction. These
25495 instructions may prefetch data, which is not safe to do if accessing
25496 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
25497 for more information).
25499 The default is to allow these instructions, but it is not possible for
25500 GCC to reliably detect all circumstances where a string instruction
25501 might be used to access an I/O register, so their use cannot be
25502 disabled automatically. Instead it is reliant upon the programmer to
25503 use the @option{-mno-allow-string-insns} option if their program
25504 accesses I/O space.
25506 When the instructions are enabled GCC defines the C preprocessor
25507 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
25508 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
25514 Use only (or not only) @code{JSR} instructions to access functions.
25515 This option can be used when code size exceeds the range of @code{BSR}
25516 instructions. Note that @option{-mno-jsr} does not mean to not use
25517 @code{JSR} but instead means that any type of branch may be used.
25520 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
25521 has special significance to the RX port when used with the
25522 @code{interrupt} function attribute. This attribute indicates a
25523 function intended to process fast interrupts. GCC ensures
25524 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
25525 and/or @code{r13} and only provided that the normal use of the
25526 corresponding registers have been restricted via the
25527 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
25530 @node S/390 and zSeries Options
25531 @subsection S/390 and zSeries Options
25532 @cindex S/390 and zSeries Options
25534 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
25538 @itemx -msoft-float
25539 @opindex mhard-float
25540 @opindex msoft-float
25541 Use (do not use) the hardware floating-point instructions and registers
25542 for floating-point operations. When @option{-msoft-float} is specified,
25543 functions in @file{libgcc.a} are used to perform floating-point
25544 operations. When @option{-mhard-float} is specified, the compiler
25545 generates IEEE floating-point instructions. This is the default.
25548 @itemx -mno-hard-dfp
25550 @opindex mno-hard-dfp
25551 Use (do not use) the hardware decimal-floating-point instructions for
25552 decimal-floating-point operations. When @option{-mno-hard-dfp} is
25553 specified, functions in @file{libgcc.a} are used to perform
25554 decimal-floating-point operations. When @option{-mhard-dfp} is
25555 specified, the compiler generates decimal-floating-point hardware
25556 instructions. This is the default for @option{-march=z9-ec} or higher.
25558 @item -mlong-double-64
25559 @itemx -mlong-double-128
25560 @opindex mlong-double-64
25561 @opindex mlong-double-128
25562 These switches control the size of @code{long double} type. A size
25563 of 64 bits makes the @code{long double} type equivalent to the @code{double}
25564 type. This is the default.
25567 @itemx -mno-backchain
25568 @opindex mbackchain
25569 @opindex mno-backchain
25570 Store (do not store) the address of the caller's frame as backchain pointer
25571 into the callee's stack frame.
25572 A backchain may be needed to allow debugging using tools that do not understand
25573 DWARF call frame information.
25574 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
25575 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
25576 the backchain is placed into the topmost word of the 96/160 byte register
25579 In general, code compiled with @option{-mbackchain} is call-compatible with
25580 code compiled with @option{-mmo-backchain}; however, use of the backchain
25581 for debugging purposes usually requires that the whole binary is built with
25582 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
25583 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
25584 to build a linux kernel use @option{-msoft-float}.
25586 The default is to not maintain the backchain.
25588 @item -mpacked-stack
25589 @itemx -mno-packed-stack
25590 @opindex mpacked-stack
25591 @opindex mno-packed-stack
25592 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
25593 specified, the compiler uses the all fields of the 96/160 byte register save
25594 area only for their default purpose; unused fields still take up stack space.
25595 When @option{-mpacked-stack} is specified, register save slots are densely
25596 packed at the top of the register save area; unused space is reused for other
25597 purposes, allowing for more efficient use of the available stack space.
25598 However, when @option{-mbackchain} is also in effect, the topmost word of
25599 the save area is always used to store the backchain, and the return address
25600 register is always saved two words below the backchain.
25602 As long as the stack frame backchain is not used, code generated with
25603 @option{-mpacked-stack} is call-compatible with code generated with
25604 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
25605 S/390 or zSeries generated code that uses the stack frame backchain at run
25606 time, not just for debugging purposes. Such code is not call-compatible
25607 with code compiled with @option{-mpacked-stack}. Also, note that the
25608 combination of @option{-mbackchain},
25609 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
25610 to build a linux kernel use @option{-msoft-float}.
25612 The default is to not use the packed stack layout.
25615 @itemx -mno-small-exec
25616 @opindex msmall-exec
25617 @opindex mno-small-exec
25618 Generate (or do not generate) code using the @code{bras} instruction
25619 to do subroutine calls.
25620 This only works reliably if the total executable size does not
25621 exceed 64k. The default is to use the @code{basr} instruction instead,
25622 which does not have this limitation.
25628 When @option{-m31} is specified, generate code compliant to the
25629 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
25630 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
25631 particular to generate 64-bit instructions. For the @samp{s390}
25632 targets, the default is @option{-m31}, while the @samp{s390x}
25633 targets default to @option{-m64}.
25639 When @option{-mzarch} is specified, generate code using the
25640 instructions available on z/Architecture.
25641 When @option{-mesa} is specified, generate code using the
25642 instructions available on ESA/390. Note that @option{-mesa} is
25643 not possible with @option{-m64}.
25644 When generating code compliant to the GNU/Linux for S/390 ABI,
25645 the default is @option{-mesa}. When generating code compliant
25646 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
25652 The @option{-mhtm} option enables a set of builtins making use of
25653 instructions available with the transactional execution facility
25654 introduced with the IBM zEnterprise EC12 machine generation
25655 @ref{S/390 System z Built-in Functions}.
25656 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
25662 When @option{-mvx} is specified, generate code using the instructions
25663 available with the vector extension facility introduced with the IBM
25664 z13 machine generation.
25665 This option changes the ABI for some vector type values with regard to
25666 alignment and calling conventions. In case vector type values are
25667 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
25668 command will be added to mark the resulting binary with the ABI used.
25669 @option{-mvx} is enabled by default when using @option{-march=z13}.
25672 @itemx -mno-zvector
25674 @opindex mno-zvector
25675 The @option{-mzvector} option enables vector language extensions and
25676 builtins using instructions available with the vector extension
25677 facility introduced with the IBM z13 machine generation.
25678 This option adds support for @samp{vector} to be used as a keyword to
25679 define vector type variables and arguments. @samp{vector} is only
25680 available when GNU extensions are enabled. It will not be expanded
25681 when requesting strict standard compliance e.g.@: with @option{-std=c99}.
25682 In addition to the GCC low-level builtins @option{-mzvector} enables
25683 a set of builtins added for compatibility with AltiVec-style
25684 implementations like Power and Cell. In order to make use of these
25685 builtins the header file @file{vecintrin.h} needs to be included.
25686 @option{-mzvector} is disabled by default.
25692 Generate (or do not generate) code using the @code{mvcle} instruction
25693 to perform block moves. When @option{-mno-mvcle} is specified,
25694 use a @code{mvc} loop instead. This is the default unless optimizing for
25701 Print (or do not print) additional debug information when compiling.
25702 The default is to not print debug information.
25704 @item -march=@var{cpu-type}
25706 Generate code that runs on @var{cpu-type}, which is the name of a
25707 system representing a certain processor type. Possible values for
25708 @var{cpu-type} are @samp{z900}/@samp{arch5}, @samp{z990}/@samp{arch6},
25709 @samp{z9-109}, @samp{z9-ec}/@samp{arch7}, @samp{z10}/@samp{arch8},
25710 @samp{z196}/@samp{arch9}, @samp{zEC12}, @samp{z13}/@samp{arch11},
25711 @samp{z14}/@samp{arch12}, and @samp{native}.
25713 The default is @option{-march=z900}.
25715 Specifying @samp{native} as cpu type can be used to select the best
25716 architecture option for the host processor.
25717 @option{-march=native} has no effect if GCC does not recognize the
25720 @item -mtune=@var{cpu-type}
25722 Tune to @var{cpu-type} everything applicable about the generated code,
25723 except for the ABI and the set of available instructions.
25724 The list of @var{cpu-type} values is the same as for @option{-march}.
25725 The default is the value used for @option{-march}.
25728 @itemx -mno-tpf-trace
25729 @opindex mtpf-trace
25730 @opindex mno-tpf-trace
25731 Generate code that adds (does not add) in TPF OS specific branches to trace
25732 routines in the operating system. This option is off by default, even
25733 when compiling for the TPF OS@.
25736 @itemx -mno-fused-madd
25737 @opindex mfused-madd
25738 @opindex mno-fused-madd
25739 Generate code that uses (does not use) the floating-point multiply and
25740 accumulate instructions. These instructions are generated by default if
25741 hardware floating point is used.
25743 @item -mwarn-framesize=@var{framesize}
25744 @opindex mwarn-framesize
25745 Emit a warning if the current function exceeds the given frame size. Because
25746 this is a compile-time check it doesn't need to be a real problem when the program
25747 runs. It is intended to identify functions that most probably cause
25748 a stack overflow. It is useful to be used in an environment with limited stack
25749 size e.g.@: the linux kernel.
25751 @item -mwarn-dynamicstack
25752 @opindex mwarn-dynamicstack
25753 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
25754 arrays. This is generally a bad idea with a limited stack size.
25756 @item -mstack-guard=@var{stack-guard}
25757 @itemx -mstack-size=@var{stack-size}
25758 @opindex mstack-guard
25759 @opindex mstack-size
25760 If these options are provided the S/390 back end emits additional instructions in
25761 the function prologue that trigger a trap if the stack size is @var{stack-guard}
25762 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
25763 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
25764 the frame size of the compiled function is chosen.
25765 These options are intended to be used to help debugging stack overflow problems.
25766 The additionally emitted code causes only little overhead and hence can also be
25767 used in production-like systems without greater performance degradation. The given
25768 values have to be exact powers of 2 and @var{stack-size} has to be greater than
25769 @var{stack-guard} without exceeding 64k.
25770 In order to be efficient the extra code makes the assumption that the stack starts
25771 at an address aligned to the value given by @var{stack-size}.
25772 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
25774 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
25776 If the hotpatch option is enabled, a ``hot-patching'' function
25777 prologue is generated for all functions in the compilation unit.
25778 The funtion label is prepended with the given number of two-byte
25779 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
25780 the label, 2 * @var{post-halfwords} bytes are appended, using the
25781 largest NOP like instructions the architecture allows (maximum
25784 If both arguments are zero, hotpatching is disabled.
25786 This option can be overridden for individual functions with the
25787 @code{hotpatch} attribute.
25790 @node Score Options
25791 @subsection Score Options
25792 @cindex Score Options
25794 These options are defined for Score implementations:
25799 Compile code for big-endian mode. This is the default.
25803 Compile code for little-endian mode.
25807 Disable generation of @code{bcnz} instructions.
25811 Enable generation of unaligned load and store instructions.
25815 Enable the use of multiply-accumulate instructions. Disabled by default.
25819 Specify the SCORE5 as the target architecture.
25823 Specify the SCORE5U of the target architecture.
25827 Specify the SCORE7 as the target architecture. This is the default.
25831 Specify the SCORE7D as the target architecture.
25835 @subsection SH Options
25837 These @samp{-m} options are defined for the SH implementations:
25842 Generate code for the SH1.
25846 Generate code for the SH2.
25849 Generate code for the SH2e.
25853 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
25854 that the floating-point unit is not used.
25856 @item -m2a-single-only
25857 @opindex m2a-single-only
25858 Generate code for the SH2a-FPU, in such a way that no double-precision
25859 floating-point operations are used.
25862 @opindex m2a-single
25863 Generate code for the SH2a-FPU assuming the floating-point unit is in
25864 single-precision mode by default.
25868 Generate code for the SH2a-FPU assuming the floating-point unit is in
25869 double-precision mode by default.
25873 Generate code for the SH3.
25877 Generate code for the SH3e.
25881 Generate code for the SH4 without a floating-point unit.
25883 @item -m4-single-only
25884 @opindex m4-single-only
25885 Generate code for the SH4 with a floating-point unit that only
25886 supports single-precision arithmetic.
25890 Generate code for the SH4 assuming the floating-point unit is in
25891 single-precision mode by default.
25895 Generate code for the SH4.
25899 Generate code for SH4-100.
25901 @item -m4-100-nofpu
25902 @opindex m4-100-nofpu
25903 Generate code for SH4-100 in such a way that the
25904 floating-point unit is not used.
25906 @item -m4-100-single
25907 @opindex m4-100-single
25908 Generate code for SH4-100 assuming the floating-point unit is in
25909 single-precision mode by default.
25911 @item -m4-100-single-only
25912 @opindex m4-100-single-only
25913 Generate code for SH4-100 in such a way that no double-precision
25914 floating-point operations are used.
25918 Generate code for SH4-200.
25920 @item -m4-200-nofpu
25921 @opindex m4-200-nofpu
25922 Generate code for SH4-200 without in such a way that the
25923 floating-point unit is not used.
25925 @item -m4-200-single
25926 @opindex m4-200-single
25927 Generate code for SH4-200 assuming the floating-point unit is in
25928 single-precision mode by default.
25930 @item -m4-200-single-only
25931 @opindex m4-200-single-only
25932 Generate code for SH4-200 in such a way that no double-precision
25933 floating-point operations are used.
25937 Generate code for SH4-300.
25939 @item -m4-300-nofpu
25940 @opindex m4-300-nofpu
25941 Generate code for SH4-300 without in such a way that the
25942 floating-point unit is not used.
25944 @item -m4-300-single
25945 @opindex m4-300-single
25946 Generate code for SH4-300 in such a way that no double-precision
25947 floating-point operations are used.
25949 @item -m4-300-single-only
25950 @opindex m4-300-single-only
25951 Generate code for SH4-300 in such a way that no double-precision
25952 floating-point operations are used.
25956 Generate code for SH4-340 (no MMU, no FPU).
25960 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
25965 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
25966 floating-point unit is not used.
25968 @item -m4a-single-only
25969 @opindex m4a-single-only
25970 Generate code for the SH4a, in such a way that no double-precision
25971 floating-point operations are used.
25974 @opindex m4a-single
25975 Generate code for the SH4a assuming the floating-point unit is in
25976 single-precision mode by default.
25980 Generate code for the SH4a.
25984 Same as @option{-m4a-nofpu}, except that it implicitly passes
25985 @option{-dsp} to the assembler. GCC doesn't generate any DSP
25986 instructions at the moment.
25990 Compile code for the processor in big-endian mode.
25994 Compile code for the processor in little-endian mode.
25998 Align doubles at 64-bit boundaries. Note that this changes the calling
25999 conventions, and thus some functions from the standard C library do
26000 not work unless you recompile it first with @option{-mdalign}.
26004 Shorten some address references at link time, when possible; uses the
26005 linker option @option{-relax}.
26009 Use 32-bit offsets in @code{switch} tables. The default is to use
26014 Enable the use of bit manipulation instructions on SH2A.
26018 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
26019 alignment constraints.
26023 Comply with the calling conventions defined by Renesas.
26026 @opindex mno-renesas
26027 Comply with the calling conventions defined for GCC before the Renesas
26028 conventions were available. This option is the default for all
26029 targets of the SH toolchain.
26032 @opindex mnomacsave
26033 Mark the @code{MAC} register as call-clobbered, even if
26034 @option{-mrenesas} is given.
26040 Control the IEEE compliance of floating-point comparisons, which affects the
26041 handling of cases where the result of a comparison is unordered. By default
26042 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
26043 enabled @option{-mno-ieee} is implicitly set, which results in faster
26044 floating-point greater-equal and less-equal comparisons. The implicit settings
26045 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
26047 @item -minline-ic_invalidate
26048 @opindex minline-ic_invalidate
26049 Inline code to invalidate instruction cache entries after setting up
26050 nested function trampolines.
26051 This option has no effect if @option{-musermode} is in effect and the selected
26052 code generation option (e.g.@: @option{-m4}) does not allow the use of the @code{icbi}
26054 If the selected code generation option does not allow the use of the @code{icbi}
26055 instruction, and @option{-musermode} is not in effect, the inlined code
26056 manipulates the instruction cache address array directly with an associative
26057 write. This not only requires privileged mode at run time, but it also
26058 fails if the cache line had been mapped via the TLB and has become unmapped.
26062 Dump instruction size and location in the assembly code.
26065 @opindex mpadstruct
26066 This option is deprecated. It pads structures to multiple of 4 bytes,
26067 which is incompatible with the SH ABI@.
26069 @item -matomic-model=@var{model}
26070 @opindex matomic-model=@var{model}
26071 Sets the model of atomic operations and additional parameters as a comma
26072 separated list. For details on the atomic built-in functions see
26073 @ref{__atomic Builtins}. The following models and parameters are supported:
26078 Disable compiler generated atomic sequences and emit library calls for atomic
26079 operations. This is the default if the target is not @code{sh*-*-linux*}.
26082 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
26083 built-in functions. The generated atomic sequences require additional support
26084 from the interrupt/exception handling code of the system and are only suitable
26085 for SH3* and SH4* single-core systems. This option is enabled by default when
26086 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
26087 this option also partially utilizes the hardware atomic instructions
26088 @code{movli.l} and @code{movco.l} to create more efficient code, unless
26089 @samp{strict} is specified.
26092 Generate software atomic sequences that use a variable in the thread control
26093 block. This is a variation of the gUSA sequences which can also be used on
26094 SH1* and SH2* targets. The generated atomic sequences require additional
26095 support from the interrupt/exception handling code of the system and are only
26096 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
26097 parameter has to be specified as well.
26100 Generate software atomic sequences that temporarily disable interrupts by
26101 setting @code{SR.IMASK = 1111}. This model works only when the program runs
26102 in privileged mode and is only suitable for single-core systems. Additional
26103 support from the interrupt/exception handling code of the system is not
26104 required. This model is enabled by default when the target is
26105 @code{sh*-*-linux*} and SH1* or SH2*.
26108 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
26109 instructions only. This is only available on SH4A and is suitable for
26110 multi-core systems. Since the hardware instructions support only 32 bit atomic
26111 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
26112 Code compiled with this option is also compatible with other software
26113 atomic model interrupt/exception handling systems if executed on an SH4A
26114 system. Additional support from the interrupt/exception handling code of the
26115 system is not required for this model.
26118 This parameter specifies the offset in bytes of the variable in the thread
26119 control block structure that should be used by the generated atomic sequences
26120 when the @samp{soft-tcb} model has been selected. For other models this
26121 parameter is ignored. The specified value must be an integer multiple of four
26122 and in the range 0-1020.
26125 This parameter prevents mixed usage of multiple atomic models, even if they
26126 are compatible, and makes the compiler generate atomic sequences of the
26127 specified model only.
26133 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
26134 Notice that depending on the particular hardware and software configuration
26135 this can degrade overall performance due to the operand cache line flushes
26136 that are implied by the @code{tas.b} instruction. On multi-core SH4A
26137 processors the @code{tas.b} instruction must be used with caution since it
26138 can result in data corruption for certain cache configurations.
26141 @opindex mprefergot
26142 When generating position-independent code, emit function calls using
26143 the Global Offset Table instead of the Procedure Linkage Table.
26146 @itemx -mno-usermode
26148 @opindex mno-usermode
26149 Don't allow (allow) the compiler generating privileged mode code. Specifying
26150 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
26151 inlined code would not work in user mode. @option{-musermode} is the default
26152 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
26153 @option{-musermode} has no effect, since there is no user mode.
26155 @item -multcost=@var{number}
26156 @opindex multcost=@var{number}
26157 Set the cost to assume for a multiply insn.
26159 @item -mdiv=@var{strategy}
26160 @opindex mdiv=@var{strategy}
26161 Set the division strategy to be used for integer division operations.
26162 @var{strategy} can be one of:
26167 Calls a library function that uses the single-step division instruction
26168 @code{div1} to perform the operation. Division by zero calculates an
26169 unspecified result and does not trap. This is the default except for SH4,
26170 SH2A and SHcompact.
26173 Calls a library function that performs the operation in double precision
26174 floating point. Division by zero causes a floating-point exception. This is
26175 the default for SHcompact with FPU. Specifying this for targets that do not
26176 have a double precision FPU defaults to @code{call-div1}.
26179 Calls a library function that uses a lookup table for small divisors and
26180 the @code{div1} instruction with case distinction for larger divisors. Division
26181 by zero calculates an unspecified result and does not trap. This is the default
26182 for SH4. Specifying this for targets that do not have dynamic shift
26183 instructions defaults to @code{call-div1}.
26187 When a division strategy has not been specified the default strategy is
26188 selected based on the current target. For SH2A the default strategy is to
26189 use the @code{divs} and @code{divu} instructions instead of library function
26192 @item -maccumulate-outgoing-args
26193 @opindex maccumulate-outgoing-args
26194 Reserve space once for outgoing arguments in the function prologue rather
26195 than around each call. Generally beneficial for performance and size. Also
26196 needed for unwinding to avoid changing the stack frame around conditional code.
26198 @item -mdivsi3_libfunc=@var{name}
26199 @opindex mdivsi3_libfunc=@var{name}
26200 Set the name of the library function used for 32-bit signed division to
26202 This only affects the name used in the @samp{call} division strategies, and
26203 the compiler still expects the same sets of input/output/clobbered registers as
26204 if this option were not present.
26206 @item -mfixed-range=@var{register-range}
26207 @opindex mfixed-range
26208 Generate code treating the given register range as fixed registers.
26209 A fixed register is one that the register allocator cannot use. This is
26210 useful when compiling kernel code. A register range is specified as
26211 two registers separated by a dash. Multiple register ranges can be
26212 specified separated by a comma.
26214 @item -mbranch-cost=@var{num}
26215 @opindex mbranch-cost=@var{num}
26216 Assume @var{num} to be the cost for a branch instruction. Higher numbers
26217 make the compiler try to generate more branch-free code if possible.
26218 If not specified the value is selected depending on the processor type that
26219 is being compiled for.
26222 @itemx -mno-zdcbranch
26223 @opindex mzdcbranch
26224 @opindex mno-zdcbranch
26225 Assume (do not assume) that zero displacement conditional branch instructions
26226 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
26227 compiler prefers zero displacement branch code sequences. This is
26228 enabled by default when generating code for SH4 and SH4A. It can be explicitly
26229 disabled by specifying @option{-mno-zdcbranch}.
26231 @item -mcbranch-force-delay-slot
26232 @opindex mcbranch-force-delay-slot
26233 Force the usage of delay slots for conditional branches, which stuffs the delay
26234 slot with a @code{nop} if a suitable instruction cannot be found. By default
26235 this option is disabled. It can be enabled to work around hardware bugs as
26236 found in the original SH7055.
26239 @itemx -mno-fused-madd
26240 @opindex mfused-madd
26241 @opindex mno-fused-madd
26242 Generate code that uses (does not use) the floating-point multiply and
26243 accumulate instructions. These instructions are generated by default
26244 if hardware floating point is used. The machine-dependent
26245 @option{-mfused-madd} option is now mapped to the machine-independent
26246 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
26247 mapped to @option{-ffp-contract=off}.
26253 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
26254 and cosine approximations. The option @option{-mfsca} must be used in
26255 combination with @option{-funsafe-math-optimizations}. It is enabled by default
26256 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
26257 approximations even if @option{-funsafe-math-optimizations} is in effect.
26263 Allow or disallow the compiler to emit the @code{fsrra} instruction for
26264 reciprocal square root approximations. The option @option{-mfsrra} must be used
26265 in combination with @option{-funsafe-math-optimizations} and
26266 @option{-ffinite-math-only}. It is enabled by default when generating code for
26267 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
26268 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
26271 @item -mpretend-cmove
26272 @opindex mpretend-cmove
26273 Prefer zero-displacement conditional branches for conditional move instruction
26274 patterns. This can result in faster code on the SH4 processor.
26278 Generate code using the FDPIC ABI.
26282 @node Solaris 2 Options
26283 @subsection Solaris 2 Options
26284 @cindex Solaris 2 options
26286 These @samp{-m} options are supported on Solaris 2:
26289 @item -mclear-hwcap
26290 @opindex mclear-hwcap
26291 @option{-mclear-hwcap} tells the compiler to remove the hardware
26292 capabilities generated by the Solaris assembler. This is only necessary
26293 when object files use ISA extensions not supported by the current
26294 machine, but check at runtime whether or not to use them.
26296 @item -mimpure-text
26297 @opindex mimpure-text
26298 @option{-mimpure-text}, used in addition to @option{-shared}, tells
26299 the compiler to not pass @option{-z text} to the linker when linking a
26300 shared object. Using this option, you can link position-dependent
26301 code into a shared object.
26303 @option{-mimpure-text} suppresses the ``relocations remain against
26304 allocatable but non-writable sections'' linker error message.
26305 However, the necessary relocations trigger copy-on-write, and the
26306 shared object is not actually shared across processes. Instead of
26307 using @option{-mimpure-text}, you should compile all source code with
26308 @option{-fpic} or @option{-fPIC}.
26312 These switches are supported in addition to the above on Solaris 2:
26317 This is a synonym for @option{-pthread}.
26320 @node SPARC Options
26321 @subsection SPARC Options
26322 @cindex SPARC options
26324 These @samp{-m} options are supported on the SPARC:
26327 @item -mno-app-regs
26329 @opindex mno-app-regs
26331 Specify @option{-mapp-regs} to generate output using the global registers
26332 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
26333 global register 1, each global register 2 through 4 is then treated as an
26334 allocable register that is clobbered by function calls. This is the default.
26336 To be fully SVR4 ABI-compliant at the cost of some performance loss,
26337 specify @option{-mno-app-regs}. You should compile libraries and system
26338 software with this option.
26344 With @option{-mflat}, the compiler does not generate save/restore instructions
26345 and uses a ``flat'' or single register window model. This model is compatible
26346 with the regular register window model. The local registers and the input
26347 registers (0--5) are still treated as ``call-saved'' registers and are
26348 saved on the stack as needed.
26350 With @option{-mno-flat} (the default), the compiler generates save/restore
26351 instructions (except for leaf functions). This is the normal operating mode.
26354 @itemx -mhard-float
26356 @opindex mhard-float
26357 Generate output containing floating-point instructions. This is the
26361 @itemx -msoft-float
26363 @opindex msoft-float
26364 Generate output containing library calls for floating point.
26365 @strong{Warning:} the requisite libraries are not available for all SPARC
26366 targets. Normally the facilities of the machine's usual C compiler are
26367 used, but this cannot be done directly in cross-compilation. You must make
26368 your own arrangements to provide suitable library functions for
26369 cross-compilation. The embedded targets @samp{sparc-*-aout} and
26370 @samp{sparclite-*-*} do provide software floating-point support.
26372 @option{-msoft-float} changes the calling convention in the output file;
26373 therefore, it is only useful if you compile @emph{all} of a program with
26374 this option. In particular, you need to compile @file{libgcc.a}, the
26375 library that comes with GCC, with @option{-msoft-float} in order for
26378 @item -mhard-quad-float
26379 @opindex mhard-quad-float
26380 Generate output containing quad-word (long double) floating-point
26383 @item -msoft-quad-float
26384 @opindex msoft-quad-float
26385 Generate output containing library calls for quad-word (long double)
26386 floating-point instructions. The functions called are those specified
26387 in the SPARC ABI@. This is the default.
26389 As of this writing, there are no SPARC implementations that have hardware
26390 support for the quad-word floating-point instructions. They all invoke
26391 a trap handler for one of these instructions, and then the trap handler
26392 emulates the effect of the instruction. Because of the trap handler overhead,
26393 this is much slower than calling the ABI library routines. Thus the
26394 @option{-msoft-quad-float} option is the default.
26396 @item -mno-unaligned-doubles
26397 @itemx -munaligned-doubles
26398 @opindex mno-unaligned-doubles
26399 @opindex munaligned-doubles
26400 Assume that doubles have 8-byte alignment. This is the default.
26402 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
26403 alignment only if they are contained in another type, or if they have an
26404 absolute address. Otherwise, it assumes they have 4-byte alignment.
26405 Specifying this option avoids some rare compatibility problems with code
26406 generated by other compilers. It is not the default because it results
26407 in a performance loss, especially for floating-point code.
26410 @itemx -mno-user-mode
26411 @opindex muser-mode
26412 @opindex mno-user-mode
26413 Do not generate code that can only run in supervisor mode. This is relevant
26414 only for the @code{casa} instruction emitted for the LEON3 processor. This
26417 @item -mfaster-structs
26418 @itemx -mno-faster-structs
26419 @opindex mfaster-structs
26420 @opindex mno-faster-structs
26421 With @option{-mfaster-structs}, the compiler assumes that structures
26422 should have 8-byte alignment. This enables the use of pairs of
26423 @code{ldd} and @code{std} instructions for copies in structure
26424 assignment, in place of twice as many @code{ld} and @code{st} pairs.
26425 However, the use of this changed alignment directly violates the SPARC
26426 ABI@. Thus, it's intended only for use on targets where the developer
26427 acknowledges that their resulting code is not directly in line with
26428 the rules of the ABI@.
26430 @item -mstd-struct-return
26431 @itemx -mno-std-struct-return
26432 @opindex mstd-struct-return
26433 @opindex mno-std-struct-return
26434 With @option{-mstd-struct-return}, the compiler generates checking code
26435 in functions returning structures or unions to detect size mismatches
26436 between the two sides of function calls, as per the 32-bit ABI@.
26438 The default is @option{-mno-std-struct-return}. This option has no effect
26445 Enable Local Register Allocation. This is the default for SPARC since GCC 7
26446 so @option{-mno-lra} needs to be passed to get old Reload.
26448 @item -mcpu=@var{cpu_type}
26450 Set the instruction set, register set, and instruction scheduling parameters
26451 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
26452 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
26453 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
26454 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
26455 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
26456 @samp{niagara3}, @samp{niagara4}, @samp{niagara7} and @samp{m8}.
26458 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
26459 which selects the best architecture option for the host processor.
26460 @option{-mcpu=native} has no effect if GCC does not recognize
26463 Default instruction scheduling parameters are used for values that select
26464 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
26465 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
26467 Here is a list of each supported architecture and their supported
26475 supersparc, hypersparc, leon, leon3
26478 f930, f934, sparclite86x
26484 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4,
26488 By default (unless configured otherwise), GCC generates code for the V7
26489 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
26490 additionally optimizes it for the Cypress CY7C602 chip, as used in the
26491 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
26492 SPARCStation 1, 2, IPX etc.
26494 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
26495 architecture. The only difference from V7 code is that the compiler emits
26496 the integer multiply and integer divide instructions which exist in SPARC-V8
26497 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
26498 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
26501 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
26502 the SPARC architecture. This adds the integer multiply, integer divide step
26503 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
26504 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
26505 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
26506 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
26507 MB86934 chip, which is the more recent SPARClite with FPU@.
26509 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
26510 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
26511 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
26512 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
26513 optimizes it for the TEMIC SPARClet chip.
26515 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
26516 architecture. This adds 64-bit integer and floating-point move instructions,
26517 3 additional floating-point condition code registers and conditional move
26518 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
26519 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
26520 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
26521 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
26522 @option{-mcpu=niagara}, the compiler additionally optimizes it for
26523 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
26524 additionally optimizes it for Sun UltraSPARC T2 chips. With
26525 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
26526 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
26527 additionally optimizes it for Sun UltraSPARC T4 chips. With
26528 @option{-mcpu=niagara7}, the compiler additionally optimizes it for
26529 Oracle SPARC M7 chips. With @option{-mcpu=m8}, the compiler
26530 additionally optimizes it for Oracle M8 chips.
26532 @item -mtune=@var{cpu_type}
26534 Set the instruction scheduling parameters for machine type
26535 @var{cpu_type}, but do not set the instruction set or register set that the
26536 option @option{-mcpu=@var{cpu_type}} does.
26538 The same values for @option{-mcpu=@var{cpu_type}} can be used for
26539 @option{-mtune=@var{cpu_type}}, but the only useful values are those
26540 that select a particular CPU implementation. Those are
26541 @samp{cypress}, @samp{supersparc}, @samp{hypersparc}, @samp{leon},
26542 @samp{leon3}, @samp{leon3v7}, @samp{f930}, @samp{f934},
26543 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
26544 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
26545 @samp{niagara4}, @samp{niagara7} and @samp{m8}. With native Solaris
26546 and GNU/Linux toolchains, @samp{native} can also be used.
26551 @opindex mno-v8plus
26552 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
26553 difference from the V8 ABI is that the global and out registers are
26554 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
26555 mode for all SPARC-V9 processors.
26561 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
26562 Visual Instruction Set extensions. The default is @option{-mno-vis}.
26568 With @option{-mvis2}, GCC generates code that takes advantage of
26569 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
26570 default is @option{-mvis2} when targeting a cpu that supports such
26571 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
26572 also sets @option{-mvis}.
26578 With @option{-mvis3}, GCC generates code that takes advantage of
26579 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
26580 default is @option{-mvis3} when targeting a cpu that supports such
26581 instructions, such as niagara-3 and later. Setting @option{-mvis3}
26582 also sets @option{-mvis2} and @option{-mvis}.
26588 With @option{-mvis4}, GCC generates code that takes advantage of
26589 version 4.0 of the UltraSPARC Visual Instruction Set extensions. The
26590 default is @option{-mvis4} when targeting a cpu that supports such
26591 instructions, such as niagara-7 and later. Setting @option{-mvis4}
26592 also sets @option{-mvis3}, @option{-mvis2} and @option{-mvis}.
26598 With @option{-mvis4b}, GCC generates code that takes advantage of
26599 version 4.0 of the UltraSPARC Visual Instruction Set extensions, plus
26600 the additional VIS instructions introduced in the Oracle SPARC
26601 Architecture 2017. The default is @option{-mvis4b} when targeting a
26602 cpu that supports such instructions, such as m8 and later. Setting
26603 @option{-mvis4b} also sets @option{-mvis4}, @option{-mvis3},
26604 @option{-mvis2} and @option{-mvis}.
26609 @opindex mno-cbcond
26610 With @option{-mcbcond}, GCC generates code that takes advantage of the UltraSPARC
26611 Compare-and-Branch-on-Condition instructions. The default is @option{-mcbcond}
26612 when targeting a CPU that supports such instructions, such as Niagara-4 and
26619 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
26620 Fused Multiply-Add Floating-point instructions. The default is @option{-mfmaf}
26621 when targeting a CPU that supports such instructions, such as Niagara-3 and
26627 @opindex mno-fsmuld
26628 With @option{-mfsmuld}, GCC generates code that takes advantage of the
26629 Floating-point Multiply Single to Double (FsMULd) instruction. The default is
26630 @option{-mfsmuld} when targeting a CPU supporting the architecture versions V8
26631 or V9 with FPU except @option{-mcpu=leon}.
26637 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
26638 Population Count instruction. The default is @option{-mpopc}
26639 when targeting a CPU that supports such an instruction, such as Niagara-2 and
26646 With @option{-msubxc}, GCC generates code that takes advantage of the UltraSPARC
26647 Subtract-Extended-with-Carry instruction. The default is @option{-msubxc}
26648 when targeting a CPU that supports such an instruction, such as Niagara-7 and
26652 @opindex mfix-at697f
26653 Enable the documented workaround for the single erratum of the Atmel AT697F
26654 processor (which corresponds to erratum #13 of the AT697E processor).
26657 @opindex mfix-ut699
26658 Enable the documented workarounds for the floating-point errata and the data
26659 cache nullify errata of the UT699 processor.
26662 @opindex mfix-ut700
26663 Enable the documented workaround for the back-to-back store errata of
26664 the UT699E/UT700 processor.
26666 @item -mfix-gr712rc
26667 @opindex mfix-gr712rc
26668 Enable the documented workaround for the back-to-back store errata of
26669 the GR712RC processor.
26672 These @samp{-m} options are supported in addition to the above
26673 on SPARC-V9 processors in 64-bit environments:
26680 Generate code for a 32-bit or 64-bit environment.
26681 The 32-bit environment sets int, long and pointer to 32 bits.
26682 The 64-bit environment sets int to 32 bits and long and pointer
26685 @item -mcmodel=@var{which}
26687 Set the code model to one of
26691 The Medium/Low code model: 64-bit addresses, programs
26692 must be linked in the low 32 bits of memory. Programs can be statically
26693 or dynamically linked.
26696 The Medium/Middle code model: 64-bit addresses, programs
26697 must be linked in the low 44 bits of memory, the text and data segments must
26698 be less than 2GB in size and the data segment must be located within 2GB of
26702 The Medium/Anywhere code model: 64-bit addresses, programs
26703 may be linked anywhere in memory, the text and data segments must be less
26704 than 2GB in size and the data segment must be located within 2GB of the
26708 The Medium/Anywhere code model for embedded systems:
26709 64-bit addresses, the text and data segments must be less than 2GB in
26710 size, both starting anywhere in memory (determined at link time). The
26711 global register %g4 points to the base of the data segment. Programs
26712 are statically linked and PIC is not supported.
26715 @item -mmemory-model=@var{mem-model}
26716 @opindex mmemory-model
26717 Set the memory model in force on the processor to one of
26721 The default memory model for the processor and operating system.
26724 Relaxed Memory Order
26727 Partial Store Order
26733 Sequential Consistency
26736 These memory models are formally defined in Appendix D of the SPARC-V9
26737 architecture manual, as set in the processor's @code{PSTATE.MM} field.
26740 @itemx -mno-stack-bias
26741 @opindex mstack-bias
26742 @opindex mno-stack-bias
26743 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
26744 frame pointer if present, are offset by @minus{}2047 which must be added back
26745 when making stack frame references. This is the default in 64-bit mode.
26746 Otherwise, assume no such offset is present.
26749 @node System V Options
26750 @subsection Options for System V
26752 These additional options are available on System V Release 4 for
26753 compatibility with other compilers on those systems:
26758 Create a shared object.
26759 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
26763 Identify the versions of each tool used by the compiler, in a
26764 @code{.ident} assembler directive in the output.
26768 Refrain from adding @code{.ident} directives to the output file (this is
26771 @item -YP,@var{dirs}
26773 Search the directories @var{dirs}, and no others, for libraries
26774 specified with @option{-l}.
26776 @item -Ym,@var{dir}
26778 Look in the directory @var{dir} to find the M4 preprocessor.
26779 The assembler uses this option.
26780 @c This is supposed to go with a -Yd for predefined M4 macro files, but
26781 @c the generic assembler that comes with Solaris takes just -Ym.
26784 @node TILE-Gx Options
26785 @subsection TILE-Gx Options
26786 @cindex TILE-Gx options
26788 These @samp{-m} options are supported on the TILE-Gx:
26791 @item -mcmodel=small
26792 @opindex mcmodel=small
26793 Generate code for the small model. The distance for direct calls is
26794 limited to 500M in either direction. PC-relative addresses are 32
26795 bits. Absolute addresses support the full address range.
26797 @item -mcmodel=large
26798 @opindex mcmodel=large
26799 Generate code for the large model. There is no limitation on call
26800 distance, pc-relative addresses, or absolute addresses.
26802 @item -mcpu=@var{name}
26804 Selects the type of CPU to be targeted. Currently the only supported
26805 type is @samp{tilegx}.
26811 Generate code for a 32-bit or 64-bit environment. The 32-bit
26812 environment sets int, long, and pointer to 32 bits. The 64-bit
26813 environment sets int to 32 bits and long and pointer to 64 bits.
26816 @itemx -mlittle-endian
26817 @opindex mbig-endian
26818 @opindex mlittle-endian
26819 Generate code in big/little endian mode, respectively.
26822 @node TILEPro Options
26823 @subsection TILEPro Options
26824 @cindex TILEPro options
26826 These @samp{-m} options are supported on the TILEPro:
26829 @item -mcpu=@var{name}
26831 Selects the type of CPU to be targeted. Currently the only supported
26832 type is @samp{tilepro}.
26836 Generate code for a 32-bit environment, which sets int, long, and
26837 pointer to 32 bits. This is the only supported behavior so the flag
26838 is essentially ignored.
26842 @subsection V850 Options
26843 @cindex V850 Options
26845 These @samp{-m} options are defined for V850 implementations:
26849 @itemx -mno-long-calls
26850 @opindex mlong-calls
26851 @opindex mno-long-calls
26852 Treat all calls as being far away (near). If calls are assumed to be
26853 far away, the compiler always loads the function's address into a
26854 register, and calls indirect through the pointer.
26860 Do not optimize (do optimize) basic blocks that use the same index
26861 pointer 4 or more times to copy pointer into the @code{ep} register, and
26862 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
26863 option is on by default if you optimize.
26865 @item -mno-prolog-function
26866 @itemx -mprolog-function
26867 @opindex mno-prolog-function
26868 @opindex mprolog-function
26869 Do not use (do use) external functions to save and restore registers
26870 at the prologue and epilogue of a function. The external functions
26871 are slower, but use less code space if more than one function saves
26872 the same number of registers. The @option{-mprolog-function} option
26873 is on by default if you optimize.
26877 Try to make the code as small as possible. At present, this just turns
26878 on the @option{-mep} and @option{-mprolog-function} options.
26880 @item -mtda=@var{n}
26882 Put static or global variables whose size is @var{n} bytes or less into
26883 the tiny data area that register @code{ep} points to. The tiny data
26884 area can hold up to 256 bytes in total (128 bytes for byte references).
26886 @item -msda=@var{n}
26888 Put static or global variables whose size is @var{n} bytes or less into
26889 the small data area that register @code{gp} points to. The small data
26890 area can hold up to 64 kilobytes.
26892 @item -mzda=@var{n}
26894 Put static or global variables whose size is @var{n} bytes or less into
26895 the first 32 kilobytes of memory.
26899 Specify that the target processor is the V850.
26903 Specify that the target processor is the V850E3V5. The preprocessor
26904 constant @code{__v850e3v5__} is defined if this option is used.
26908 Specify that the target processor is the V850E3V5. This is an alias for
26909 the @option{-mv850e3v5} option.
26913 Specify that the target processor is the V850E2V3. The preprocessor
26914 constant @code{__v850e2v3__} is defined if this option is used.
26918 Specify that the target processor is the V850E2. The preprocessor
26919 constant @code{__v850e2__} is defined if this option is used.
26923 Specify that the target processor is the V850E1. The preprocessor
26924 constants @code{__v850e1__} and @code{__v850e__} are defined if
26925 this option is used.
26929 Specify that the target processor is the V850ES. This is an alias for
26930 the @option{-mv850e1} option.
26934 Specify that the target processor is the V850E@. The preprocessor
26935 constant @code{__v850e__} is defined if this option is used.
26937 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
26938 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
26939 are defined then a default target processor is chosen and the
26940 relevant @samp{__v850*__} preprocessor constant is defined.
26942 The preprocessor constants @code{__v850} and @code{__v851__} are always
26943 defined, regardless of which processor variant is the target.
26945 @item -mdisable-callt
26946 @itemx -mno-disable-callt
26947 @opindex mdisable-callt
26948 @opindex mno-disable-callt
26949 This option suppresses generation of the @code{CALLT} instruction for the
26950 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
26953 This option is enabled by default when the RH850 ABI is
26954 in use (see @option{-mrh850-abi}), and disabled by default when the
26955 GCC ABI is in use. If @code{CALLT} instructions are being generated
26956 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
26962 Pass on (or do not pass on) the @option{-mrelax} command-line option
26966 @itemx -mno-long-jumps
26967 @opindex mlong-jumps
26968 @opindex mno-long-jumps
26969 Disable (or re-enable) the generation of PC-relative jump instructions.
26972 @itemx -mhard-float
26973 @opindex msoft-float
26974 @opindex mhard-float
26975 Disable (or re-enable) the generation of hardware floating point
26976 instructions. This option is only significant when the target
26977 architecture is @samp{V850E2V3} or higher. If hardware floating point
26978 instructions are being generated then the C preprocessor symbol
26979 @code{__FPU_OK__} is defined, otherwise the symbol
26980 @code{__NO_FPU__} is defined.
26984 Enables the use of the e3v5 LOOP instruction. The use of this
26985 instruction is not enabled by default when the e3v5 architecture is
26986 selected because its use is still experimental.
26990 @opindex mrh850-abi
26992 Enables support for the RH850 version of the V850 ABI. This is the
26993 default. With this version of the ABI the following rules apply:
26997 Integer sized structures and unions are returned via a memory pointer
26998 rather than a register.
27001 Large structures and unions (more than 8 bytes in size) are passed by
27005 Functions are aligned to 16-bit boundaries.
27008 The @option{-m8byte-align} command-line option is supported.
27011 The @option{-mdisable-callt} command-line option is enabled by
27012 default. The @option{-mno-disable-callt} command-line option is not
27016 When this version of the ABI is enabled the C preprocessor symbol
27017 @code{__V850_RH850_ABI__} is defined.
27021 Enables support for the old GCC version of the V850 ABI. With this
27022 version of the ABI the following rules apply:
27026 Integer sized structures and unions are returned in register @code{r10}.
27029 Large structures and unions (more than 8 bytes in size) are passed by
27033 Functions are aligned to 32-bit boundaries, unless optimizing for
27037 The @option{-m8byte-align} command-line option is not supported.
27040 The @option{-mdisable-callt} command-line option is supported but not
27041 enabled by default.
27044 When this version of the ABI is enabled the C preprocessor symbol
27045 @code{__V850_GCC_ABI__} is defined.
27047 @item -m8byte-align
27048 @itemx -mno-8byte-align
27049 @opindex m8byte-align
27050 @opindex mno-8byte-align
27051 Enables support for @code{double} and @code{long long} types to be
27052 aligned on 8-byte boundaries. The default is to restrict the
27053 alignment of all objects to at most 4-bytes. When
27054 @option{-m8byte-align} is in effect the C preprocessor symbol
27055 @code{__V850_8BYTE_ALIGN__} is defined.
27058 @opindex mbig-switch
27059 Generate code suitable for big switch tables. Use this option only if
27060 the assembler/linker complain about out of range branches within a switch
27065 This option causes r2 and r5 to be used in the code generated by
27066 the compiler. This setting is the default.
27068 @item -mno-app-regs
27069 @opindex mno-app-regs
27070 This option causes r2 and r5 to be treated as fixed registers.
27075 @subsection VAX Options
27076 @cindex VAX options
27078 These @samp{-m} options are defined for the VAX:
27083 Do not output certain jump instructions (@code{aobleq} and so on)
27084 that the Unix assembler for the VAX cannot handle across long
27089 Do output those jump instructions, on the assumption that the
27090 GNU assembler is being used.
27094 Output code for G-format floating-point numbers instead of D-format.
27097 @node Visium Options
27098 @subsection Visium Options
27099 @cindex Visium options
27105 A program which performs file I/O and is destined to run on an MCM target
27106 should be linked with this option. It causes the libraries libc.a and
27107 libdebug.a to be linked. The program should be run on the target under
27108 the control of the GDB remote debugging stub.
27112 A program which performs file I/O and is destined to run on the simulator
27113 should be linked with option. This causes libraries libc.a and libsim.a to
27117 @itemx -mhard-float
27119 @opindex mhard-float
27120 Generate code containing floating-point instructions. This is the
27124 @itemx -msoft-float
27126 @opindex msoft-float
27127 Generate code containing library calls for floating-point.
27129 @option{-msoft-float} changes the calling convention in the output file;
27130 therefore, it is only useful if you compile @emph{all} of a program with
27131 this option. In particular, you need to compile @file{libgcc.a}, the
27132 library that comes with GCC, with @option{-msoft-float} in order for
27135 @item -mcpu=@var{cpu_type}
27137 Set the instruction set, register set, and instruction scheduling parameters
27138 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
27139 @samp{mcm}, @samp{gr5} and @samp{gr6}.
27141 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
27143 By default (unless configured otherwise), GCC generates code for the GR5
27144 variant of the Visium architecture.
27146 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
27147 architecture. The only difference from GR5 code is that the compiler will
27148 generate block move instructions.
27150 @item -mtune=@var{cpu_type}
27152 Set the instruction scheduling parameters for machine type @var{cpu_type},
27153 but do not set the instruction set or register set that the option
27154 @option{-mcpu=@var{cpu_type}} would.
27158 Generate code for the supervisor mode, where there are no restrictions on
27159 the access to general registers. This is the default.
27162 @opindex muser-mode
27163 Generate code for the user mode, where the access to some general registers
27164 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
27165 mode; on the GR6, only registers r29 to r31 are affected.
27169 @subsection VMS Options
27171 These @samp{-m} options are defined for the VMS implementations:
27174 @item -mvms-return-codes
27175 @opindex mvms-return-codes
27176 Return VMS condition codes from @code{main}. The default is to return POSIX-style
27177 condition (e.g.@: error) codes.
27179 @item -mdebug-main=@var{prefix}
27180 @opindex mdebug-main=@var{prefix}
27181 Flag the first routine whose name starts with @var{prefix} as the main
27182 routine for the debugger.
27186 Default to 64-bit memory allocation routines.
27188 @item -mpointer-size=@var{size}
27189 @opindex mpointer-size=@var{size}
27190 Set the default size of pointers. Possible options for @var{size} are
27191 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
27192 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
27193 The later option disables @code{pragma pointer_size}.
27196 @node VxWorks Options
27197 @subsection VxWorks Options
27198 @cindex VxWorks Options
27200 The options in this section are defined for all VxWorks targets.
27201 Options specific to the target hardware are listed with the other
27202 options for that target.
27207 GCC can generate code for both VxWorks kernels and real time processes
27208 (RTPs). This option switches from the former to the latter. It also
27209 defines the preprocessor macro @code{__RTP__}.
27212 @opindex non-static
27213 Link an RTP executable against shared libraries rather than static
27214 libraries. The options @option{-static} and @option{-shared} can
27215 also be used for RTPs (@pxref{Link Options}); @option{-static}
27222 These options are passed down to the linker. They are defined for
27223 compatibility with Diab.
27226 @opindex Xbind-lazy
27227 Enable lazy binding of function calls. This option is equivalent to
27228 @option{-Wl,-z,now} and is defined for compatibility with Diab.
27232 Disable lazy binding of function calls. This option is the default and
27233 is defined for compatibility with Diab.
27237 @subsection x86 Options
27238 @cindex x86 Options
27240 These @samp{-m} options are defined for the x86 family of computers.
27244 @item -march=@var{cpu-type}
27246 Generate instructions for the machine type @var{cpu-type}. In contrast to
27247 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
27248 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
27249 to generate code that may not run at all on processors other than the one
27250 indicated. Specifying @option{-march=@var{cpu-type}} implies
27251 @option{-mtune=@var{cpu-type}}.
27253 The choices for @var{cpu-type} are:
27257 This selects the CPU to generate code for at compilation time by determining
27258 the processor type of the compiling machine. Using @option{-march=native}
27259 enables all instruction subsets supported by the local machine (hence
27260 the result might not run on different machines). Using @option{-mtune=native}
27261 produces code optimized for the local machine under the constraints
27262 of the selected instruction set.
27265 A generic CPU with 64-bit extensions.
27268 Original Intel i386 CPU@.
27271 Intel i486 CPU@. (No scheduling is implemented for this chip.)
27275 Intel Pentium CPU with no MMX support.
27278 Intel Lakemont MCU, based on Intel Pentium CPU.
27281 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
27284 Intel Pentium Pro CPU@.
27287 When used with @option{-march}, the Pentium Pro
27288 instruction set is used, so the code runs on all i686 family chips.
27289 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
27292 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
27297 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
27301 Intel Pentium M; low-power version of Intel Pentium III CPU
27302 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
27306 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
27309 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
27313 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
27314 SSE2 and SSE3 instruction set support.
27317 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
27318 instruction set support.
27321 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27322 SSE4.1, SSE4.2 and POPCNT instruction set support.
27325 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27326 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
27329 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27330 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
27333 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27334 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
27335 instruction set support.
27338 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27339 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27340 BMI, BMI2 and F16C instruction set support.
27343 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27344 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27345 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
27348 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27349 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27350 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
27351 XSAVES instruction set support.
27354 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
27355 instruction set support.
27358 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27359 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
27362 Intel Goldmont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27363 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL, RDRND, XSAVE, XSAVEOPT and FSGSBASE
27364 instruction set support.
27366 @item goldmont-plus
27367 Intel Goldmont Plus CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27368 SSSE3, SSE4.1, SSE4.2, POPCNT, AES, PCLMUL, RDRND, XSAVE, XSAVEOPT, FSGSBASE,
27369 PTWRITE, RDPID, SGX and UMIP instruction set support.
27372 Intel Tremont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27373 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL, RDRND, XSAVE, XSAVEOPT, FSGSBASE, PTWRITE,
27374 RDPID, SGX, UMIP, GFNI-SSE, CLWB and ENCLV instruction set support.
27377 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27378 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27379 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
27380 AVX512CD instruction set support.
27383 Intel Knights Mill CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27384 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27385 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER, AVX512CD,
27386 AVX5124VNNIW, AVX5124FMAPS and AVX512VPOPCNTDQ instruction set support.
27388 @item skylake-avx512
27389 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27390 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27391 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
27392 CLWB, AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
27395 Intel Cannonlake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
27396 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
27397 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
27398 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
27399 AVX512IFMA, SHA and UMIP instruction set support.
27401 @item icelake-client
27402 Intel Icelake Client CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
27403 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
27404 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
27405 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
27406 AVX512IFMA, SHA, CLWB, UMIP, RDPID, GFNI, AVX512VBMI2, AVX512VPOPCNTDQ,
27407 AVX512BITALG, AVX512VNNI, VPCLMULQDQ, VAES instruction set support.
27409 @item icelake-server
27410 Intel Icelake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
27411 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
27412 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
27413 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
27414 AVX512IFMA, SHA, CLWB, UMIP, RDPID, GFNI, AVX512VBMI2, AVX512VPOPCNTDQ,
27415 AVX512BITALG, AVX512VNNI, VPCLMULQDQ, VAES, PCONFIG and WBNOINVD instruction
27419 Intel Cascadelake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27420 SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI,
27421 BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F, CLWB,
27422 AVX512VL, AVX512BW, AVX512DQ, AVX512CD and AVX512VNNI instruction set support.
27425 Intel cooperlake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27426 SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI,
27427 BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F, CLWB,
27428 AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VNNI and AVX512BF16 instruction
27432 Intel Tigerlake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27433 SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI,
27434 BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
27435 AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI, AVX512IFMA, SHA, CLWB, UMIP,
27436 RDPID, GFNI, AVX512VBMI2, AVX512VPOPCNTDQ, AVX512BITALG, AVX512VNNI, VPCLMULQDQ,
27437 VAES, PCONFIG, WBNOINVD, MOVDIRI, MOVDIR64B and AVX512VP2INTERSECT instruction
27441 AMD K6 CPU with MMX instruction set support.
27445 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
27448 @itemx athlon-tbird
27449 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
27455 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
27456 instruction set support.
27462 Processors based on the AMD K8 core with x86-64 instruction set support,
27463 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
27464 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
27465 instruction set extensions.)
27468 @itemx opteron-sse3
27469 @itemx athlon64-sse3
27470 Improved versions of AMD K8 cores with SSE3 instruction set support.
27474 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
27475 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
27476 instruction set extensions.)
27479 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
27480 supersets FMA4, AVX, XOP, LWP, AES, PCLMUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
27481 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
27483 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
27484 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCLMUL, CX16, MMX,
27485 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
27488 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
27489 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
27490 PCLMUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
27491 64-bit instruction set extensions.
27493 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
27494 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
27495 AES, PCLMUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
27496 SSE4.2, ABM and 64-bit instruction set extensions.
27499 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
27500 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
27501 SHA, CLZERO, AES, PCLMUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
27502 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
27503 instruction set extensions.
27505 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
27506 supersets BMI, BMI2, ,CLWB, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED,
27507 MWAITX, SHA, CLZERO, AES, PCLMUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A,
27508 SSSE3, SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
27509 instruction set extensions.)
27513 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
27514 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
27515 instruction set extensions.)
27518 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
27519 includes MOVBE, F16C, BMI, AVX, PCLMUL, AES, SSE4.2, SSE4.1, CX16, ABM,
27520 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
27523 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
27527 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
27528 instruction set support.
27531 VIA C3 CPU with MMX and 3DNow!@: instruction set support.
27532 (No scheduling is implemented for this chip.)
27535 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
27536 (No scheduling is implemented for this chip.)
27539 VIA C7 (Esther) CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
27540 (No scheduling is implemented for this chip.)
27543 VIA Eden Samuel 2 CPU with MMX and 3DNow!@: instruction set support.
27544 (No scheduling is implemented for this chip.)
27547 VIA Eden Nehemiah CPU with MMX and SSE instruction set support.
27548 (No scheduling is implemented for this chip.)
27551 VIA Eden Esther CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
27552 (No scheduling is implemented for this chip.)
27555 VIA Eden X2 CPU with x86-64, MMX, SSE, SSE2 and SSE3 instruction set support.
27556 (No scheduling is implemented for this chip.)
27559 VIA Eden X4 CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2,
27560 AVX and AVX2 instruction set support.
27561 (No scheduling is implemented for this chip.)
27564 Generic VIA Nano CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
27565 instruction set support.
27566 (No scheduling is implemented for this chip.)
27569 VIA Nano 1xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
27570 instruction set support.
27571 (No scheduling is implemented for this chip.)
27574 VIA Nano 2xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
27575 instruction set support.
27576 (No scheduling is implemented for this chip.)
27579 VIA Nano 3xxx CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
27580 instruction set support.
27581 (No scheduling is implemented for this chip.)
27584 VIA Nano Dual Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
27585 instruction set support.
27586 (No scheduling is implemented for this chip.)
27589 VIA Nano Quad Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
27590 instruction set support.
27591 (No scheduling is implemented for this chip.)
27594 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
27597 @item -mtune=@var{cpu-type}
27599 Tune to @var{cpu-type} everything applicable about the generated code, except
27600 for the ABI and the set of available instructions.
27601 While picking a specific @var{cpu-type} schedules things appropriately
27602 for that particular chip, the compiler does not generate any code that
27603 cannot run on the default machine type unless you use a
27604 @option{-march=@var{cpu-type}} option.
27605 For example, if GCC is configured for i686-pc-linux-gnu
27606 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
27607 but still runs on i686 machines.
27609 The choices for @var{cpu-type} are the same as for @option{-march}.
27610 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
27614 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
27615 If you know the CPU on which your code will run, then you should use
27616 the corresponding @option{-mtune} or @option{-march} option instead of
27617 @option{-mtune=generic}. But, if you do not know exactly what CPU users
27618 of your application will have, then you should use this option.
27620 As new processors are deployed in the marketplace, the behavior of this
27621 option will change. Therefore, if you upgrade to a newer version of
27622 GCC, code generation controlled by this option will change to reflect
27624 that are most common at the time that version of GCC is released.
27626 There is no @option{-march=generic} option because @option{-march}
27627 indicates the instruction set the compiler can use, and there is no
27628 generic instruction set applicable to all processors. In contrast,
27629 @option{-mtune} indicates the processor (or, in this case, collection of
27630 processors) for which the code is optimized.
27633 Produce code optimized for the most current Intel processors, which are
27634 Haswell and Silvermont for this version of GCC. If you know the CPU
27635 on which your code will run, then you should use the corresponding
27636 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
27637 But, if you want your application performs better on both Haswell and
27638 Silvermont, then you should use this option.
27640 As new Intel processors are deployed in the marketplace, the behavior of
27641 this option will change. Therefore, if you upgrade to a newer version of
27642 GCC, code generation controlled by this option will change to reflect
27643 the most current Intel processors at the time that version of GCC is
27646 There is no @option{-march=intel} option because @option{-march} indicates
27647 the instruction set the compiler can use, and there is no common
27648 instruction set applicable to all processors. In contrast,
27649 @option{-mtune} indicates the processor (or, in this case, collection of
27650 processors) for which the code is optimized.
27653 @item -mcpu=@var{cpu-type}
27655 A deprecated synonym for @option{-mtune}.
27657 @item -mfpmath=@var{unit}
27659 Generate floating-point arithmetic for selected unit @var{unit}. The choices
27660 for @var{unit} are:
27664 Use the standard 387 floating-point coprocessor present on the majority of chips and
27665 emulated otherwise. Code compiled with this option runs almost everywhere.
27666 The temporary results are computed in 80-bit precision instead of the precision
27667 specified by the type, resulting in slightly different results compared to most
27668 of other chips. See @option{-ffloat-store} for more detailed description.
27670 This is the default choice for non-Darwin x86-32 targets.
27673 Use scalar floating-point instructions present in the SSE instruction set.
27674 This instruction set is supported by Pentium III and newer chips,
27675 and in the AMD line
27676 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
27677 instruction set supports only single-precision arithmetic, thus the double and
27678 extended-precision arithmetic are still done using 387. A later version, present
27679 only in Pentium 4 and AMD x86-64 chips, supports double-precision
27682 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
27683 or @option{-msse2} switches to enable SSE extensions and make this option
27684 effective. For the x86-64 compiler, these extensions are enabled by default.
27686 The resulting code should be considerably faster in the majority of cases and avoid
27687 the numerical instability problems of 387 code, but may break some existing
27688 code that expects temporaries to be 80 bits.
27690 This is the default choice for the x86-64 compiler, Darwin x86-32 targets,
27691 and the default choice for x86-32 targets with the SSE2 instruction set
27692 when @option{-ffast-math} is enabled.
27697 Attempt to utilize both instruction sets at once. This effectively doubles the
27698 amount of available registers, and on chips with separate execution units for
27699 387 and SSE the execution resources too. Use this option with care, as it is
27700 still experimental, because the GCC register allocator does not model separate
27701 functional units well, resulting in unstable performance.
27704 @item -masm=@var{dialect}
27705 @opindex masm=@var{dialect}
27706 Output assembly instructions using selected @var{dialect}. Also affects
27707 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
27708 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
27709 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
27710 not support @samp{intel}.
27713 @itemx -mno-ieee-fp
27715 @opindex mno-ieee-fp
27716 Control whether or not the compiler uses IEEE floating-point
27717 comparisons. These correctly handle the case where the result of a
27718 comparison is unordered.
27721 @itemx -mhard-float
27723 @opindex mhard-float
27724 Generate output containing 80387 instructions for floating point.
27727 @itemx -msoft-float
27729 @opindex msoft-float
27730 Generate output containing library calls for floating point.
27732 @strong{Warning:} the requisite libraries are not part of GCC@.
27733 Normally the facilities of the machine's usual C compiler are used, but
27734 this cannot be done directly in cross-compilation. You must make your
27735 own arrangements to provide suitable library functions for
27738 On machines where a function returns floating-point results in the 80387
27739 register stack, some floating-point opcodes may be emitted even if
27740 @option{-msoft-float} is used.
27742 @item -mno-fp-ret-in-387
27743 @opindex mno-fp-ret-in-387
27744 @opindex mfp-ret-in-387
27745 Do not use the FPU registers for return values of functions.
27747 The usual calling convention has functions return values of types
27748 @code{float} and @code{double} in an FPU register, even if there
27749 is no FPU@. The idea is that the operating system should emulate
27752 The option @option{-mno-fp-ret-in-387} causes such values to be returned
27753 in ordinary CPU registers instead.
27755 @item -mno-fancy-math-387
27756 @opindex mno-fancy-math-387
27757 @opindex mfancy-math-387
27758 Some 387 emulators do not support the @code{sin}, @code{cos} and
27759 @code{sqrt} instructions for the 387. Specify this option to avoid
27760 generating those instructions.
27761 This option is overridden when @option{-march}
27762 indicates that the target CPU always has an FPU and so the
27763 instruction does not need emulation. These
27764 instructions are not generated unless you also use the
27765 @option{-funsafe-math-optimizations} switch.
27767 @item -malign-double
27768 @itemx -mno-align-double
27769 @opindex malign-double
27770 @opindex mno-align-double
27771 Control whether GCC aligns @code{double}, @code{long double}, and
27772 @code{long long} variables on a two-word boundary or a one-word
27773 boundary. Aligning @code{double} variables on a two-word boundary
27774 produces code that runs somewhat faster on a Pentium at the
27775 expense of more memory.
27777 On x86-64, @option{-malign-double} is enabled by default.
27779 @strong{Warning:} if you use the @option{-malign-double} switch,
27780 structures containing the above types are aligned differently than
27781 the published application binary interface specifications for the x86-32
27782 and are not binary compatible with structures in code compiled
27783 without that switch.
27785 @item -m96bit-long-double
27786 @itemx -m128bit-long-double
27787 @opindex m96bit-long-double
27788 @opindex m128bit-long-double
27789 These switches control the size of @code{long double} type. The x86-32
27790 application binary interface specifies the size to be 96 bits,
27791 so @option{-m96bit-long-double} is the default in 32-bit mode.
27793 Modern architectures (Pentium and newer) prefer @code{long double}
27794 to be aligned to an 8- or 16-byte boundary. In arrays or structures
27795 conforming to the ABI, this is not possible. So specifying
27796 @option{-m128bit-long-double} aligns @code{long double}
27797 to a 16-byte boundary by padding the @code{long double} with an additional
27800 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
27801 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
27803 Notice that neither of these options enable any extra precision over the x87
27804 standard of 80 bits for a @code{long double}.
27806 @strong{Warning:} if you override the default value for your target ABI, this
27807 changes the size of
27808 structures and arrays containing @code{long double} variables,
27809 as well as modifying the function calling convention for functions taking
27810 @code{long double}. Hence they are not binary-compatible
27811 with code compiled without that switch.
27813 @item -mlong-double-64
27814 @itemx -mlong-double-80
27815 @itemx -mlong-double-128
27816 @opindex mlong-double-64
27817 @opindex mlong-double-80
27818 @opindex mlong-double-128
27819 These switches control the size of @code{long double} type. A size
27820 of 64 bits makes the @code{long double} type equivalent to the @code{double}
27821 type. This is the default for 32-bit Bionic C library. A size
27822 of 128 bits makes the @code{long double} type equivalent to the
27823 @code{__float128} type. This is the default for 64-bit Bionic C library.
27825 @strong{Warning:} if you override the default value for your target ABI, this
27826 changes the size of
27827 structures and arrays containing @code{long double} variables,
27828 as well as modifying the function calling convention for functions taking
27829 @code{long double}. Hence they are not binary-compatible
27830 with code compiled without that switch.
27832 @item -malign-data=@var{type}
27833 @opindex malign-data
27834 Control how GCC aligns variables. Supported values for @var{type} are
27835 @samp{compat} uses increased alignment value compatible uses GCC 4.8
27836 and earlier, @samp{abi} uses alignment value as specified by the
27837 psABI, and @samp{cacheline} uses increased alignment value to match
27838 the cache line size. @samp{compat} is the default.
27840 @item -mlarge-data-threshold=@var{threshold}
27841 @opindex mlarge-data-threshold
27842 When @option{-mcmodel=medium} is specified, data objects larger than
27843 @var{threshold} are placed in the large data section. This value must be the
27844 same across all objects linked into the binary, and defaults to 65535.
27848 Use a different function-calling convention, in which functions that
27849 take a fixed number of arguments return with the @code{ret @var{num}}
27850 instruction, which pops their arguments while returning. This saves one
27851 instruction in the caller since there is no need to pop the arguments
27854 You can specify that an individual function is called with this calling
27855 sequence with the function attribute @code{stdcall}. You can also
27856 override the @option{-mrtd} option by using the function attribute
27857 @code{cdecl}. @xref{Function Attributes}.
27859 @strong{Warning:} this calling convention is incompatible with the one
27860 normally used on Unix, so you cannot use it if you need to call
27861 libraries compiled with the Unix compiler.
27863 Also, you must provide function prototypes for all functions that
27864 take variable numbers of arguments (including @code{printf});
27865 otherwise incorrect code is generated for calls to those
27868 In addition, seriously incorrect code results if you call a
27869 function with too many arguments. (Normally, extra arguments are
27870 harmlessly ignored.)
27872 @item -mregparm=@var{num}
27874 Control how many registers are used to pass integer arguments. By
27875 default, no registers are used to pass arguments, and at most 3
27876 registers can be used. You can control this behavior for a specific
27877 function by using the function attribute @code{regparm}.
27878 @xref{Function Attributes}.
27880 @strong{Warning:} if you use this switch, and
27881 @var{num} is nonzero, then you must build all modules with the same
27882 value, including any libraries. This includes the system libraries and
27886 @opindex msseregparm
27887 Use SSE register passing conventions for float and double arguments
27888 and return values. You can control this behavior for a specific
27889 function by using the function attribute @code{sseregparm}.
27890 @xref{Function Attributes}.
27892 @strong{Warning:} if you use this switch then you must build all
27893 modules with the same value, including any libraries. This includes
27894 the system libraries and startup modules.
27896 @item -mvect8-ret-in-mem
27897 @opindex mvect8-ret-in-mem
27898 Return 8-byte vectors in memory instead of MMX registers. This is the
27899 default on VxWorks to match the ABI of the Sun Studio compilers until
27900 version 12. @emph{Only} use this option if you need to remain
27901 compatible with existing code produced by those previous compiler
27902 versions or older versions of GCC@.
27911 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
27912 is specified, the significands of results of floating-point operations are
27913 rounded to 24 bits (single precision); @option{-mpc64} rounds the
27914 significands of results of floating-point operations to 53 bits (double
27915 precision) and @option{-mpc80} rounds the significands of results of
27916 floating-point operations to 64 bits (extended double precision), which is
27917 the default. When this option is used, floating-point operations in higher
27918 precisions are not available to the programmer without setting the FPU
27919 control word explicitly.
27921 Setting the rounding of floating-point operations to less than the default
27922 80 bits can speed some programs by 2% or more. Note that some mathematical
27923 libraries assume that extended-precision (80-bit) floating-point operations
27924 are enabled by default; routines in such libraries could suffer significant
27925 loss of accuracy, typically through so-called ``catastrophic cancellation'',
27926 when this option is used to set the precision to less than extended precision.
27928 @item -mstackrealign
27929 @opindex mstackrealign
27930 Realign the stack at entry. On the x86, the @option{-mstackrealign}
27931 option generates an alternate prologue and epilogue that realigns the
27932 run-time stack if necessary. This supports mixing legacy codes that keep
27933 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
27934 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
27935 applicable to individual functions.
27937 @item -mpreferred-stack-boundary=@var{num}
27938 @opindex mpreferred-stack-boundary
27939 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
27940 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
27941 the default is 4 (16 bytes or 128 bits).
27943 @strong{Warning:} When generating code for the x86-64 architecture with
27944 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
27945 used to keep the stack boundary aligned to 8 byte boundary. Since
27946 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
27947 intended to be used in controlled environment where stack space is
27948 important limitation. This option leads to wrong code when functions
27949 compiled with 16 byte stack alignment (such as functions from a standard
27950 library) are called with misaligned stack. In this case, SSE
27951 instructions may lead to misaligned memory access traps. In addition,
27952 variable arguments are handled incorrectly for 16 byte aligned
27953 objects (including x87 long double and __int128), leading to wrong
27954 results. You must build all modules with
27955 @option{-mpreferred-stack-boundary=3}, including any libraries. This
27956 includes the system libraries and startup modules.
27958 @item -mincoming-stack-boundary=@var{num}
27959 @opindex mincoming-stack-boundary
27960 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
27961 boundary. If @option{-mincoming-stack-boundary} is not specified,
27962 the one specified by @option{-mpreferred-stack-boundary} is used.
27964 On Pentium and Pentium Pro, @code{double} and @code{long double} values
27965 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
27966 suffer significant run time performance penalties. On Pentium III, the
27967 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
27968 properly if it is not 16-byte aligned.
27970 To ensure proper alignment of this values on the stack, the stack boundary
27971 must be as aligned as that required by any value stored on the stack.
27972 Further, every function must be generated such that it keeps the stack
27973 aligned. Thus calling a function compiled with a higher preferred
27974 stack boundary from a function compiled with a lower preferred stack
27975 boundary most likely misaligns the stack. It is recommended that
27976 libraries that use callbacks always use the default setting.
27978 This extra alignment does consume extra stack space, and generally
27979 increases code size. Code that is sensitive to stack space usage, such
27980 as embedded systems and operating system kernels, may want to reduce the
27981 preferred alignment to @option{-mpreferred-stack-boundary=2}.
28038 @itemx -mavx512ifma
28039 @opindex mavx512ifma
28041 @itemx -mavx512vbmi
28042 @opindex mavx512vbmi
28053 @itemx -mclflushopt
28054 @opindex mclflushopt
28089 @itemx -mprefetchwt1
28090 @opindex mprefetchwt1
28161 @itemx -mavx512vbmi2
28162 @opindex mavx512vbmi2
28164 @itemx -mavx512bf16
28165 @opindex mavx512bf16
28176 @itemx -mvpclmulqdq
28177 @opindex mvpclmulqdq
28179 @itemx -mavx512bitalg
28180 @opindex mavx512bitalg
28186 @opindex mmovdir64b
28191 @itemx -mavx512vpopcntdq
28192 @opindex mavx512vpopcntdq
28194 @itemx -mavx512vp2intersect
28195 @opindex mavx512vp2intersect
28197 @itemx -mavx5124fmaps
28198 @opindex mavx5124fmaps
28200 @itemx -mavx512vnni
28201 @opindex mavx512vnni
28203 @itemx -mavx5124vnniw
28204 @opindex mavx5124vnniw
28208 These switches enable the use of instructions in the MMX, SSE,
28209 SSE2, SSE3, SSSE3, SSE4, SSE4A, SSE4.1, SSE4.2, AVX, AVX2, AVX512F, AVX512PF,
28210 AVX512ER, AVX512CD, AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA, AVX512VBMI, SHA,
28211 AES, PCLMUL, CLFLUSHOPT, CLWB, FSGSBASE, PTWRITE, RDRND, F16C, FMA, PCONFIG,
28212 WBNOINVD, FMA4, PREFETCHW, RDPID, PREFETCHWT1, RDSEED, SGX, XOP, LWP,
28213 3DNow!@:, enhanced 3DNow!@:, POPCNT, ABM, ADX, BMI, BMI2, LZCNT, FXSR, XSAVE,
28214 XSAVEOPT, XSAVEC, XSAVES, RTM, HLE, TBM, MWAITX, CLZERO, PKU, AVX512VBMI2,
28215 GFNI, VAES, WAITPKG, VPCLMULQDQ, AVX512BITALG, MOVDIRI, MOVDIR64B, AVX512BF16,
28216 ENQCMD, AVX512VPOPCNTDQ, AVX5124FMAPS, AVX512VNNI, AVX5124VNNIW, or CLDEMOTE
28217 extended instruction sets. Each has a corresponding @option{-mno-} option to
28218 disable use of these instructions.
28220 These extensions are also available as built-in functions: see
28221 @ref{x86 Built-in Functions}, for details of the functions enabled and
28222 disabled by these switches.
28224 To generate SSE/SSE2 instructions automatically from floating-point
28225 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
28227 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
28228 generates new AVX instructions or AVX equivalence for all SSEx instructions
28231 These options enable GCC to use these extended instructions in
28232 generated code, even without @option{-mfpmath=sse}. Applications that
28233 perform run-time CPU detection must compile separate files for each
28234 supported architecture, using the appropriate flags. In particular,
28235 the file containing the CPU detection code should be compiled without
28238 @item -mdump-tune-features
28239 @opindex mdump-tune-features
28240 This option instructs GCC to dump the names of the x86 performance
28241 tuning features and default settings. The names can be used in
28242 @option{-mtune-ctrl=@var{feature-list}}.
28244 @item -mtune-ctrl=@var{feature-list}
28245 @opindex mtune-ctrl=@var{feature-list}
28246 This option is used to do fine grain control of x86 code generation features.
28247 @var{feature-list} is a comma separated list of @var{feature} names. See also
28248 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
28249 on if it is not preceded with @samp{^}, otherwise, it is turned off.
28250 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
28251 developers. Using it may lead to code paths not covered by testing and can
28252 potentially result in compiler ICEs or runtime errors.
28255 @opindex mno-default
28256 This option instructs GCC to turn off all tunable features. See also
28257 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
28261 This option instructs GCC to emit a @code{cld} instruction in the prologue
28262 of functions that use string instructions. String instructions depend on
28263 the DF flag to select between autoincrement or autodecrement mode. While the
28264 ABI specifies the DF flag to be cleared on function entry, some operating
28265 systems violate this specification by not clearing the DF flag in their
28266 exception dispatchers. The exception handler can be invoked with the DF flag
28267 set, which leads to wrong direction mode when string instructions are used.
28268 This option can be enabled by default on 32-bit x86 targets by configuring
28269 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
28270 instructions can be suppressed with the @option{-mno-cld} compiler option
28274 @opindex mvzeroupper
28275 This option instructs GCC to emit a @code{vzeroupper} instruction
28276 before a transfer of control flow out of the function to minimize
28277 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
28280 @item -mprefer-avx128
28281 @opindex mprefer-avx128
28282 This option instructs GCC to use 128-bit AVX instructions instead of
28283 256-bit AVX instructions in the auto-vectorizer.
28285 @item -mprefer-vector-width=@var{opt}
28286 @opindex mprefer-vector-width
28287 This option instructs GCC to use @var{opt}-bit vector width in instructions
28288 instead of default on the selected platform.
28292 No extra limitations applied to GCC other than defined by the selected platform.
28295 Prefer 128-bit vector width for instructions.
28298 Prefer 256-bit vector width for instructions.
28301 Prefer 512-bit vector width for instructions.
28306 This option enables GCC to generate @code{CMPXCHG16B} instructions in 64-bit
28307 code to implement compare-and-exchange operations on 16-byte aligned 128-bit
28308 objects. This is useful for atomic updates of data structures exceeding one
28309 machine word in size. The compiler uses this instruction to implement
28310 @ref{__sync Builtins}. However, for @ref{__atomic Builtins} operating on
28311 128-bit integers, a library call is always used.
28315 This option enables generation of @code{SAHF} instructions in 64-bit code.
28316 Early Intel Pentium 4 CPUs with Intel 64 support,
28317 prior to the introduction of Pentium 4 G1 step in December 2005,
28318 lacked the @code{LAHF} and @code{SAHF} instructions
28319 which are supported by AMD64.
28320 These are load and store instructions, respectively, for certain status flags.
28321 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
28322 @code{drem}, and @code{remainder} built-in functions;
28323 see @ref{Other Builtins} for details.
28327 This option enables use of the @code{movbe} instruction to implement
28328 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
28332 The @option{-mshstk} option enables shadow stack built-in functions
28333 from x86 Control-flow Enforcement Technology (CET).
28337 This option enables built-in functions @code{__builtin_ia32_crc32qi},
28338 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
28339 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
28343 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
28344 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
28345 with an additional Newton-Raphson step
28346 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
28347 (and their vectorized
28348 variants) for single-precision floating-point arguments. These instructions
28349 are generated only when @option{-funsafe-math-optimizations} is enabled
28350 together with @option{-ffinite-math-only} and @option{-fno-trapping-math}.
28351 Note that while the throughput of the sequence is higher than the throughput
28352 of the non-reciprocal instruction, the precision of the sequence can be
28353 decreased by up to 2 ulp (i.e.@: the inverse of 1.0 equals 0.99999994).
28355 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
28356 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
28357 combination), and doesn't need @option{-mrecip}.
28359 Also note that GCC emits the above sequence with additional Newton-Raphson step
28360 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
28361 already with @option{-ffast-math} (or the above option combination), and
28362 doesn't need @option{-mrecip}.
28364 @item -mrecip=@var{opt}
28365 @opindex mrecip=opt
28366 This option controls which reciprocal estimate instructions
28367 may be used. @var{opt} is a comma-separated list of options, which may
28368 be preceded by a @samp{!} to invert the option:
28372 Enable all estimate instructions.
28375 Enable the default instructions, equivalent to @option{-mrecip}.
28378 Disable all estimate instructions, equivalent to @option{-mno-recip}.
28381 Enable the approximation for scalar division.
28384 Enable the approximation for vectorized division.
28387 Enable the approximation for scalar square root.
28390 Enable the approximation for vectorized square root.
28393 So, for example, @option{-mrecip=all,!sqrt} enables
28394 all of the reciprocal approximations, except for square root.
28396 @item -mveclibabi=@var{type}
28397 @opindex mveclibabi
28398 Specifies the ABI type to use for vectorizing intrinsics using an
28399 external library. Supported values for @var{type} are @samp{svml}
28400 for the Intel short
28401 vector math library and @samp{acml} for the AMD math core library.
28402 To use this option, both @option{-ftree-vectorize} and
28403 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
28404 ABI-compatible library must be specified at link time.
28406 GCC currently emits calls to @code{vmldExp2},
28407 @code{vmldLn2}, @code{vmldLog102}, @code{vmldPow2},
28408 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
28409 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
28410 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
28411 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4},
28412 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
28413 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
28414 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
28415 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
28416 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
28417 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
28418 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
28419 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
28420 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
28421 when @option{-mveclibabi=acml} is used.
28423 @item -mabi=@var{name}
28425 Generate code for the specified calling convention. Permissible values
28426 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
28427 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
28428 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
28429 You can control this behavior for specific functions by
28430 using the function attributes @code{ms_abi} and @code{sysv_abi}.
28431 @xref{Function Attributes}.
28433 @item -mforce-indirect-call
28434 @opindex mforce-indirect-call
28435 Force all calls to functions to be indirect. This is useful
28436 when using Intel Processor Trace where it generates more precise timing
28437 information for function calls.
28439 @item -mmanual-endbr
28440 @opindex mmanual-endbr
28441 Insert ENDBR instruction at function entry only via the @code{cf_check}
28442 function attribute. This is useful when used with the option
28443 @option{-fcf-protection=branch} to control ENDBR insertion at the
28446 @item -mcall-ms2sysv-xlogues
28447 @opindex mcall-ms2sysv-xlogues
28448 @opindex mno-call-ms2sysv-xlogues
28449 Due to differences in 64-bit ABIs, any Microsoft ABI function that calls a
28450 System V ABI function must consider RSI, RDI and XMM6-15 as clobbered. By
28451 default, the code for saving and restoring these registers is emitted inline,
28452 resulting in fairly lengthy prologues and epilogues. Using
28453 @option{-mcall-ms2sysv-xlogues} emits prologues and epilogues that
28454 use stubs in the static portion of libgcc to perform these saves and restores,
28455 thus reducing function size at the cost of a few extra instructions.
28457 @item -mtls-dialect=@var{type}
28458 @opindex mtls-dialect
28459 Generate code to access thread-local storage using the @samp{gnu} or
28460 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
28461 @samp{gnu2} is more efficient, but it may add compile- and run-time
28462 requirements that cannot be satisfied on all systems.
28465 @itemx -mno-push-args
28466 @opindex mpush-args
28467 @opindex mno-push-args
28468 Use PUSH operations to store outgoing parameters. This method is shorter
28469 and usually equally fast as method using SUB/MOV operations and is enabled
28470 by default. In some cases disabling it may improve performance because of
28471 improved scheduling and reduced dependencies.
28473 @item -maccumulate-outgoing-args
28474 @opindex maccumulate-outgoing-args
28475 If enabled, the maximum amount of space required for outgoing arguments is
28476 computed in the function prologue. This is faster on most modern CPUs
28477 because of reduced dependencies, improved scheduling and reduced stack usage
28478 when the preferred stack boundary is not equal to 2. The drawback is a notable
28479 increase in code size. This switch implies @option{-mno-push-args}.
28483 Support thread-safe exception handling on MinGW. Programs that rely
28484 on thread-safe exception handling must compile and link all code with the
28485 @option{-mthreads} option. When compiling, @option{-mthreads} defines
28486 @option{-D_MT}; when linking, it links in a special thread helper library
28487 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
28489 @item -mms-bitfields
28490 @itemx -mno-ms-bitfields
28491 @opindex mms-bitfields
28492 @opindex mno-ms-bitfields
28494 Enable/disable bit-field layout compatible with the native Microsoft
28497 If @code{packed} is used on a structure, or if bit-fields are used,
28498 it may be that the Microsoft ABI lays out the structure differently
28499 than the way GCC normally does. Particularly when moving packed
28500 data between functions compiled with GCC and the native Microsoft compiler
28501 (either via function call or as data in a file), it may be necessary to access
28504 This option is enabled by default for Microsoft Windows
28505 targets. This behavior can also be controlled locally by use of variable
28506 or type attributes. For more information, see @ref{x86 Variable Attributes}
28507 and @ref{x86 Type Attributes}.
28509 The Microsoft structure layout algorithm is fairly simple with the exception
28510 of the bit-field packing.
28511 The padding and alignment of members of structures and whether a bit-field
28512 can straddle a storage-unit boundary are determine by these rules:
28515 @item Structure members are stored sequentially in the order in which they are
28516 declared: the first member has the lowest memory address and the last member
28519 @item Every data object has an alignment requirement. The alignment requirement
28520 for all data except structures, unions, and arrays is either the size of the
28521 object or the current packing size (specified with either the
28522 @code{aligned} attribute or the @code{pack} pragma),
28523 whichever is less. For structures, unions, and arrays,
28524 the alignment requirement is the largest alignment requirement of its members.
28525 Every object is allocated an offset so that:
28528 offset % alignment_requirement == 0
28531 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
28532 unit if the integral types are the same size and if the next bit-field fits
28533 into the current allocation unit without crossing the boundary imposed by the
28534 common alignment requirements of the bit-fields.
28537 MSVC interprets zero-length bit-fields in the following ways:
28540 @item If a zero-length bit-field is inserted between two bit-fields that
28541 are normally coalesced, the bit-fields are not coalesced.
28548 unsigned long bf_1 : 12;
28550 unsigned long bf_2 : 12;
28555 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
28556 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
28558 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
28559 alignment of the zero-length bit-field is greater than the member that follows it,
28560 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
28581 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
28582 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
28583 bit-field does not affect the alignment of @code{bar} or, as a result, the size
28586 Taking this into account, it is important to note the following:
28589 @item If a zero-length bit-field follows a normal bit-field, the type of the
28590 zero-length bit-field may affect the alignment of the structure as whole. For
28591 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
28592 normal bit-field, and is of type short.
28594 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
28595 still affect the alignment of the structure:
28606 Here, @code{t4} takes up 4 bytes.
28609 @item Zero-length bit-fields following non-bit-field members are ignored:
28621 Here, @code{t5} takes up 2 bytes.
28625 @item -mno-align-stringops
28626 @opindex mno-align-stringops
28627 @opindex malign-stringops
28628 Do not align the destination of inlined string operations. This switch reduces
28629 code size and improves performance in case the destination is already aligned,
28630 but GCC doesn't know about it.
28632 @item -minline-all-stringops
28633 @opindex minline-all-stringops
28634 By default GCC inlines string operations only when the destination is
28635 known to be aligned to least a 4-byte boundary.
28636 This enables more inlining and increases code
28637 size, but may improve performance of code that depends on fast
28638 @code{memcpy} and @code{memset} for short lengths.
28639 The option enables inline expansion of @code{strlen} for all
28640 pointer alignments.
28642 @item -minline-stringops-dynamically
28643 @opindex minline-stringops-dynamically
28644 For string operations of unknown size, use run-time checks with
28645 inline code for small blocks and a library call for large blocks.
28647 @item -mstringop-strategy=@var{alg}
28648 @opindex mstringop-strategy=@var{alg}
28649 Override the internal decision heuristic for the particular algorithm to use
28650 for inlining string operations. The allowed values for @var{alg} are:
28656 Expand using i386 @code{rep} prefix of the specified size.
28660 @itemx unrolled_loop
28661 Expand into an inline loop.
28664 Always use a library call.
28667 @item -mmemcpy-strategy=@var{strategy}
28668 @opindex mmemcpy-strategy=@var{strategy}
28669 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
28670 should be inlined and what inline algorithm to use when the expected size
28671 of the copy operation is known. @var{strategy}
28672 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
28673 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
28674 the max byte size with which inline algorithm @var{alg} is allowed. For the last
28675 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
28676 in the list must be specified in increasing order. The minimal byte size for
28677 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
28680 @item -mmemset-strategy=@var{strategy}
28681 @opindex mmemset-strategy=@var{strategy}
28682 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
28683 @code{__builtin_memset} expansion.
28685 @item -momit-leaf-frame-pointer
28686 @opindex momit-leaf-frame-pointer
28687 Don't keep the frame pointer in a register for leaf functions. This
28688 avoids the instructions to save, set up, and restore frame pointers and
28689 makes an extra register available in leaf functions. The option
28690 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
28691 which might make debugging harder.
28693 @item -mtls-direct-seg-refs
28694 @itemx -mno-tls-direct-seg-refs
28695 @opindex mtls-direct-seg-refs
28696 Controls whether TLS variables may be accessed with offsets from the
28697 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
28698 or whether the thread base pointer must be added. Whether or not this
28699 is valid depends on the operating system, and whether it maps the
28700 segment to cover the entire TLS area.
28702 For systems that use the GNU C Library, the default is on.
28705 @itemx -mno-sse2avx
28707 Specify that the assembler should encode SSE instructions with VEX
28708 prefix. The option @option{-mavx} turns this on by default.
28713 If profiling is active (@option{-pg}), put the profiling
28714 counter call before the prologue.
28715 Note: On x86 architectures the attribute @code{ms_hook_prologue}
28716 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
28718 @item -mrecord-mcount
28719 @itemx -mno-record-mcount
28720 @opindex mrecord-mcount
28721 If profiling is active (@option{-pg}), generate a __mcount_loc section
28722 that contains pointers to each profiling call. This is useful for
28723 automatically patching and out calls.
28726 @itemx -mno-nop-mcount
28727 @opindex mnop-mcount
28728 If profiling is active (@option{-pg}), generate the calls to
28729 the profiling functions as NOPs. This is useful when they
28730 should be patched in later dynamically. This is likely only
28731 useful together with @option{-mrecord-mcount}.
28733 @item -minstrument-return=@var{type}
28734 @opindex minstrument-return
28735 Instrument function exit in -pg -mfentry instrumented functions with
28736 call to specified function. This only instruments true returns ending
28737 with ret, but not sibling calls ending with jump. Valid types
28738 are @var{none} to not instrument, @var{call} to generate a call to __return__,
28739 or @var{nop5} to generate a 5 byte nop.
28741 @item -mrecord-return
28742 @itemx -mno-record-return
28743 @opindex mrecord-return
28744 Generate a __return_loc section pointing to all return instrumentation code.
28746 @item -mfentry-name=@var{name}
28747 @opindex mfentry-name
28748 Set name of __fentry__ symbol called at function entry for -pg -mfentry functions.
28750 @item -mfentry-section=@var{name}
28751 @opindex mfentry-section
28752 Set name of section to record -mrecord-mcount calls (default __mcount_loc).
28754 @item -mskip-rax-setup
28755 @itemx -mno-skip-rax-setup
28756 @opindex mskip-rax-setup
28757 When generating code for the x86-64 architecture with SSE extensions
28758 disabled, @option{-mskip-rax-setup} can be used to skip setting up RAX
28759 register when there are no variable arguments passed in vector registers.
28761 @strong{Warning:} Since RAX register is used to avoid unnecessarily
28762 saving vector registers on stack when passing variable arguments, the
28763 impacts of this option are callees may waste some stack space,
28764 misbehave or jump to a random location. GCC 4.4 or newer don't have
28765 those issues, regardless the RAX register value.
28768 @itemx -mno-8bit-idiv
28769 @opindex m8bit-idiv
28770 On some processors, like Intel Atom, 8-bit unsigned integer divide is
28771 much faster than 32-bit/64-bit integer divide. This option generates a
28772 run-time check. If both dividend and divisor are within range of 0
28773 to 255, 8-bit unsigned integer divide is used instead of
28774 32-bit/64-bit integer divide.
28776 @item -mavx256-split-unaligned-load
28777 @itemx -mavx256-split-unaligned-store
28778 @opindex mavx256-split-unaligned-load
28779 @opindex mavx256-split-unaligned-store
28780 Split 32-byte AVX unaligned load and store.
28782 @item -mstack-protector-guard=@var{guard}
28783 @itemx -mstack-protector-guard-reg=@var{reg}
28784 @itemx -mstack-protector-guard-offset=@var{offset}
28785 @opindex mstack-protector-guard
28786 @opindex mstack-protector-guard-reg
28787 @opindex mstack-protector-guard-offset
28788 Generate stack protection code using canary at @var{guard}. Supported
28789 locations are @samp{global} for global canary or @samp{tls} for per-thread
28790 canary in the TLS block (the default). This option has effect only when
28791 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
28793 With the latter choice the options
28794 @option{-mstack-protector-guard-reg=@var{reg}} and
28795 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
28796 which segment register (@code{%fs} or @code{%gs}) to use as base register
28797 for reading the canary, and from what offset from that base register.
28798 The default for those is as specified in the relevant ABI.
28800 @item -mgeneral-regs-only
28801 @opindex mgeneral-regs-only
28802 Generate code that uses only the general-purpose registers. This
28803 prevents the compiler from using floating-point, vector, mask and bound
28806 @item -mindirect-branch=@var{choice}
28807 @opindex mindirect-branch
28808 Convert indirect call and jump with @var{choice}. The default is
28809 @samp{keep}, which keeps indirect call and jump unmodified.
28810 @samp{thunk} converts indirect call and jump to call and return thunk.
28811 @samp{thunk-inline} converts indirect call and jump to inlined call
28812 and return thunk. @samp{thunk-extern} converts indirect call and jump
28813 to external call and return thunk provided in a separate object file.
28814 You can control this behavior for a specific function by using the
28815 function attribute @code{indirect_branch}. @xref{Function Attributes}.
28817 Note that @option{-mcmodel=large} is incompatible with
28818 @option{-mindirect-branch=thunk} and
28819 @option{-mindirect-branch=thunk-extern} since the thunk function may
28820 not be reachable in the large code model.
28822 Note that @option{-mindirect-branch=thunk-extern} is incompatible with
28823 @option{-fcf-protection=branch} since the external thunk cannot be modified
28824 to disable control-flow check.
28826 @item -mfunction-return=@var{choice}
28827 @opindex mfunction-return
28828 Convert function return with @var{choice}. The default is @samp{keep},
28829 which keeps function return unmodified. @samp{thunk} converts function
28830 return to call and return thunk. @samp{thunk-inline} converts function
28831 return to inlined call and return thunk. @samp{thunk-extern} converts
28832 function return to external call and return thunk provided in a separate
28833 object file. You can control this behavior for a specific function by
28834 using the function attribute @code{function_return}.
28835 @xref{Function Attributes}.
28837 Note that @option{-mcmodel=large} is incompatible with
28838 @option{-mfunction-return=thunk} and
28839 @option{-mfunction-return=thunk-extern} since the thunk function may
28840 not be reachable in the large code model.
28843 @item -mindirect-branch-register
28844 @opindex mindirect-branch-register
28845 Force indirect call and jump via register.
28849 These @samp{-m} switches are supported in addition to the above
28850 on x86-64 processors in 64-bit environments.
28863 Generate code for a 16-bit, 32-bit or 64-bit environment.
28864 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
28866 generates code that runs on any i386 system.
28868 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
28869 types to 64 bits, and generates code for the x86-64 architecture.
28870 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
28871 and @option{-mdynamic-no-pic} options.
28873 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
28875 generates code for the x86-64 architecture.
28877 The @option{-m16} option is the same as @option{-m32}, except for that
28878 it outputs the @code{.code16gcc} assembly directive at the beginning of
28879 the assembly output so that the binary can run in 16-bit mode.
28881 The @option{-miamcu} option generates code which conforms to Intel MCU
28882 psABI. It requires the @option{-m32} option to be turned on.
28884 @item -mno-red-zone
28885 @opindex mno-red-zone
28887 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
28888 by the x86-64 ABI; it is a 128-byte area beyond the location of the
28889 stack pointer that is not modified by signal or interrupt handlers
28890 and therefore can be used for temporary data without adjusting the stack
28891 pointer. The flag @option{-mno-red-zone} disables this red zone.
28893 @item -mcmodel=small
28894 @opindex mcmodel=small
28895 Generate code for the small code model: the program and its symbols must
28896 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
28897 Programs can be statically or dynamically linked. This is the default
28900 @item -mcmodel=kernel
28901 @opindex mcmodel=kernel
28902 Generate code for the kernel code model. The kernel runs in the
28903 negative 2 GB of the address space.
28904 This model has to be used for Linux kernel code.
28906 @item -mcmodel=medium
28907 @opindex mcmodel=medium
28908 Generate code for the medium model: the program is linked in the lower 2
28909 GB of the address space. Small symbols are also placed there. Symbols
28910 with sizes larger than @option{-mlarge-data-threshold} are put into
28911 large data or BSS sections and can be located above 2GB. Programs can
28912 be statically or dynamically linked.
28914 @item -mcmodel=large
28915 @opindex mcmodel=large
28916 Generate code for the large model. This model makes no assumptions
28917 about addresses and sizes of sections.
28919 @item -maddress-mode=long
28920 @opindex maddress-mode=long
28921 Generate code for long address mode. This is only supported for 64-bit
28922 and x32 environments. It is the default address mode for 64-bit
28925 @item -maddress-mode=short
28926 @opindex maddress-mode=short
28927 Generate code for short address mode. This is only supported for 32-bit
28928 and x32 environments. It is the default address mode for 32-bit and
28932 @node x86 Windows Options
28933 @subsection x86 Windows Options
28934 @cindex x86 Windows Options
28935 @cindex Windows Options for x86
28937 These additional options are available for Microsoft Windows targets:
28943 specifies that a console application is to be generated, by
28944 instructing the linker to set the PE header subsystem type
28945 required for console applications.
28946 This option is available for Cygwin and MinGW targets and is
28947 enabled by default on those targets.
28951 This option is available for Cygwin and MinGW targets. It
28952 specifies that a DLL---a dynamic link library---is to be
28953 generated, enabling the selection of the required runtime
28954 startup object and entry point.
28956 @item -mnop-fun-dllimport
28957 @opindex mnop-fun-dllimport
28958 This option is available for Cygwin and MinGW targets. It
28959 specifies that the @code{dllimport} attribute should be ignored.
28963 This option is available for MinGW targets. It specifies
28964 that MinGW-specific thread support is to be used.
28968 This option is available for MinGW-w64 targets. It causes
28969 the @code{UNICODE} preprocessor macro to be predefined, and
28970 chooses Unicode-capable runtime startup code.
28974 This option is available for Cygwin and MinGW targets. It
28975 specifies that the typical Microsoft Windows predefined macros are to
28976 be set in the pre-processor, but does not influence the choice
28977 of runtime library/startup code.
28981 This option is available for Cygwin and MinGW targets. It
28982 specifies that a GUI application is to be generated by
28983 instructing the linker to set the PE header subsystem type
28986 @item -fno-set-stack-executable
28987 @opindex fno-set-stack-executable
28988 @opindex fset-stack-executable
28989 This option is available for MinGW targets. It specifies that
28990 the executable flag for the stack used by nested functions isn't
28991 set. This is necessary for binaries running in kernel mode of
28992 Microsoft Windows, as there the User32 API, which is used to set executable
28993 privileges, isn't available.
28995 @item -fwritable-relocated-rdata
28996 @opindex fno-writable-relocated-rdata
28997 @opindex fwritable-relocated-rdata
28998 This option is available for MinGW and Cygwin targets. It specifies
28999 that relocated-data in read-only section is put into the @code{.data}
29000 section. This is a necessary for older runtimes not supporting
29001 modification of @code{.rdata} sections for pseudo-relocation.
29003 @item -mpe-aligned-commons
29004 @opindex mpe-aligned-commons
29005 This option is available for Cygwin and MinGW targets. It
29006 specifies that the GNU extension to the PE file format that
29007 permits the correct alignment of COMMON variables should be
29008 used when generating code. It is enabled by default if
29009 GCC detects that the target assembler found during configuration
29010 supports the feature.
29013 See also under @ref{x86 Options} for standard options.
29015 @node Xstormy16 Options
29016 @subsection Xstormy16 Options
29017 @cindex Xstormy16 Options
29019 These options are defined for Xstormy16:
29024 Choose startup files and linker script suitable for the simulator.
29027 @node Xtensa Options
29028 @subsection Xtensa Options
29029 @cindex Xtensa Options
29031 These options are supported for Xtensa targets:
29035 @itemx -mno-const16
29037 @opindex mno-const16
29038 Enable or disable use of @code{CONST16} instructions for loading
29039 constant values. The @code{CONST16} instruction is currently not a
29040 standard option from Tensilica. When enabled, @code{CONST16}
29041 instructions are always used in place of the standard @code{L32R}
29042 instructions. The use of @code{CONST16} is enabled by default only if
29043 the @code{L32R} instruction is not available.
29046 @itemx -mno-fused-madd
29047 @opindex mfused-madd
29048 @opindex mno-fused-madd
29049 Enable or disable use of fused multiply/add and multiply/subtract
29050 instructions in the floating-point option. This has no effect if the
29051 floating-point option is not also enabled. Disabling fused multiply/add
29052 and multiply/subtract instructions forces the compiler to use separate
29053 instructions for the multiply and add/subtract operations. This may be
29054 desirable in some cases where strict IEEE 754-compliant results are
29055 required: the fused multiply add/subtract instructions do not round the
29056 intermediate result, thereby producing results with @emph{more} bits of
29057 precision than specified by the IEEE standard. Disabling fused multiply
29058 add/subtract instructions also ensures that the program output is not
29059 sensitive to the compiler's ability to combine multiply and add/subtract
29062 @item -mserialize-volatile
29063 @itemx -mno-serialize-volatile
29064 @opindex mserialize-volatile
29065 @opindex mno-serialize-volatile
29066 When this option is enabled, GCC inserts @code{MEMW} instructions before
29067 @code{volatile} memory references to guarantee sequential consistency.
29068 The default is @option{-mserialize-volatile}. Use
29069 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
29071 @item -mforce-no-pic
29072 @opindex mforce-no-pic
29073 For targets, like GNU/Linux, where all user-mode Xtensa code must be
29074 position-independent code (PIC), this option disables PIC for compiling
29077 @item -mtext-section-literals
29078 @itemx -mno-text-section-literals
29079 @opindex mtext-section-literals
29080 @opindex mno-text-section-literals
29081 These options control the treatment of literal pools. The default is
29082 @option{-mno-text-section-literals}, which places literals in a separate
29083 section in the output file. This allows the literal pool to be placed
29084 in a data RAM/ROM, and it also allows the linker to combine literal
29085 pools from separate object files to remove redundant literals and
29086 improve code size. With @option{-mtext-section-literals}, the literals
29087 are interspersed in the text section in order to keep them as close as
29088 possible to their references. This may be necessary for large assembly
29089 files. Literals for each function are placed right before that function.
29091 @item -mauto-litpools
29092 @itemx -mno-auto-litpools
29093 @opindex mauto-litpools
29094 @opindex mno-auto-litpools
29095 These options control the treatment of literal pools. The default is
29096 @option{-mno-auto-litpools}, which places literals in a separate
29097 section in the output file unless @option{-mtext-section-literals} is
29098 used. With @option{-mauto-litpools} the literals are interspersed in
29099 the text section by the assembler. Compiler does not produce explicit
29100 @code{.literal} directives and loads literals into registers with
29101 @code{MOVI} instructions instead of @code{L32R} to let the assembler
29102 do relaxation and place literals as necessary. This option allows
29103 assembler to create several literal pools per function and assemble
29104 very big functions, which may not be possible with
29105 @option{-mtext-section-literals}.
29107 @item -mtarget-align
29108 @itemx -mno-target-align
29109 @opindex mtarget-align
29110 @opindex mno-target-align
29111 When this option is enabled, GCC instructs the assembler to
29112 automatically align instructions to reduce branch penalties at the
29113 expense of some code density. The assembler attempts to widen density
29114 instructions to align branch targets and the instructions following call
29115 instructions. If there are not enough preceding safe density
29116 instructions to align a target, no widening is performed. The
29117 default is @option{-mtarget-align}. These options do not affect the
29118 treatment of auto-aligned instructions like @code{LOOP}, which the
29119 assembler always aligns, either by widening density instructions or
29120 by inserting NOP instructions.
29123 @itemx -mno-longcalls
29124 @opindex mlongcalls
29125 @opindex mno-longcalls
29126 When this option is enabled, GCC instructs the assembler to translate
29127 direct calls to indirect calls unless it can determine that the target
29128 of a direct call is in the range allowed by the call instruction. This
29129 translation typically occurs for calls to functions in other source
29130 files. Specifically, the assembler translates a direct @code{CALL}
29131 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
29132 The default is @option{-mno-longcalls}. This option should be used in
29133 programs where the call target can potentially be out of range. This
29134 option is implemented in the assembler, not the compiler, so the
29135 assembly code generated by GCC still shows direct call
29136 instructions---look at the disassembled object code to see the actual
29137 instructions. Note that the assembler uses an indirect call for
29138 every cross-file call, not just those that really are out of range.
29141 @node zSeries Options
29142 @subsection zSeries Options
29143 @cindex zSeries options
29145 These are listed under @xref{S/390 and zSeries Options}.
29151 @section Specifying Subprocesses and the Switches to Pass to Them
29154 @command{gcc} is a driver program. It performs its job by invoking a
29155 sequence of other programs to do the work of compiling, assembling and
29156 linking. GCC interprets its command-line parameters and uses these to
29157 deduce which programs it should invoke, and which command-line options
29158 it ought to place on their command lines. This behavior is controlled
29159 by @dfn{spec strings}. In most cases there is one spec string for each
29160 program that GCC can invoke, but a few programs have multiple spec
29161 strings to control their behavior. The spec strings built into GCC can
29162 be overridden by using the @option{-specs=} command-line switch to specify
29165 @dfn{Spec files} are plain-text files that are used to construct spec
29166 strings. They consist of a sequence of directives separated by blank
29167 lines. The type of directive is determined by the first non-whitespace
29168 character on the line, which can be one of the following:
29171 @item %@var{command}
29172 Issues a @var{command} to the spec file processor. The commands that can
29176 @item %include <@var{file}>
29177 @cindex @code{%include}
29178 Search for @var{file} and insert its text at the current point in the
29181 @item %include_noerr <@var{file}>
29182 @cindex @code{%include_noerr}
29183 Just like @samp{%include}, but do not generate an error message if the include
29184 file cannot be found.
29186 @item %rename @var{old_name} @var{new_name}
29187 @cindex @code{%rename}
29188 Rename the spec string @var{old_name} to @var{new_name}.
29192 @item *[@var{spec_name}]:
29193 This tells the compiler to create, override or delete the named spec
29194 string. All lines after this directive up to the next directive or
29195 blank line are considered to be the text for the spec string. If this
29196 results in an empty string then the spec is deleted. (Or, if the
29197 spec did not exist, then nothing happens.) Otherwise, if the spec
29198 does not currently exist a new spec is created. If the spec does
29199 exist then its contents are overridden by the text of this
29200 directive, unless the first character of that text is the @samp{+}
29201 character, in which case the text is appended to the spec.
29203 @item [@var{suffix}]:
29204 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
29205 and up to the next directive or blank line are considered to make up the
29206 spec string for the indicated suffix. When the compiler encounters an
29207 input file with the named suffix, it processes the spec string in
29208 order to work out how to compile that file. For example:
29212 z-compile -input %i
29215 This says that any input file whose name ends in @samp{.ZZ} should be
29216 passed to the program @samp{z-compile}, which should be invoked with the
29217 command-line switch @option{-input} and with the result of performing the
29218 @samp{%i} substitution. (See below.)
29220 As an alternative to providing a spec string, the text following a
29221 suffix directive can be one of the following:
29224 @item @@@var{language}
29225 This says that the suffix is an alias for a known @var{language}. This is
29226 similar to using the @option{-x} command-line switch to GCC to specify a
29227 language explicitly. For example:
29234 Says that .ZZ files are, in fact, C++ source files.
29237 This causes an error messages saying:
29240 @var{name} compiler not installed on this system.
29244 GCC already has an extensive list of suffixes built into it.
29245 This directive adds an entry to the end of the list of suffixes, but
29246 since the list is searched from the end backwards, it is effectively
29247 possible to override earlier entries using this technique.
29251 GCC has the following spec strings built into it. Spec files can
29252 override these strings or create their own. Note that individual
29253 targets can also add their own spec strings to this list.
29256 asm Options to pass to the assembler
29257 asm_final Options to pass to the assembler post-processor
29258 cpp Options to pass to the C preprocessor
29259 cc1 Options to pass to the C compiler
29260 cc1plus Options to pass to the C++ compiler
29261 endfile Object files to include at the end of the link
29262 link Options to pass to the linker
29263 lib Libraries to include on the command line to the linker
29264 libgcc Decides which GCC support library to pass to the linker
29265 linker Sets the name of the linker
29266 predefines Defines to be passed to the C preprocessor
29267 signed_char Defines to pass to CPP to say whether @code{char} is signed
29269 startfile Object files to include at the start of the link
29272 Here is a small example of a spec file:
29275 %rename lib old_lib
29278 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
29281 This example renames the spec called @samp{lib} to @samp{old_lib} and
29282 then overrides the previous definition of @samp{lib} with a new one.
29283 The new definition adds in some extra command-line options before
29284 including the text of the old definition.
29286 @dfn{Spec strings} are a list of command-line options to be passed to their
29287 corresponding program. In addition, the spec strings can contain
29288 @samp{%}-prefixed sequences to substitute variable text or to
29289 conditionally insert text into the command line. Using these constructs
29290 it is possible to generate quite complex command lines.
29292 Here is a table of all defined @samp{%}-sequences for spec
29293 strings. Note that spaces are not generated automatically around the
29294 results of expanding these sequences. Therefore you can concatenate them
29295 together or combine them with constant text in a single argument.
29299 Substitute one @samp{%} into the program name or argument.
29302 Substitute the name of the input file being processed.
29305 Substitute the basename of the input file being processed.
29306 This is the substring up to (and not including) the last period
29307 and not including the directory.
29310 This is the same as @samp{%b}, but include the file suffix (text after
29314 Marks the argument containing or following the @samp{%d} as a
29315 temporary file name, so that that file is deleted if GCC exits
29316 successfully. Unlike @samp{%g}, this contributes no text to the
29319 @item %g@var{suffix}
29320 Substitute a file name that has suffix @var{suffix} and is chosen
29321 once per compilation, and mark the argument in the same way as
29322 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
29323 name is now chosen in a way that is hard to predict even when previously
29324 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
29325 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
29326 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
29327 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
29328 was simply substituted with a file name chosen once per compilation,
29329 without regard to any appended suffix (which was therefore treated
29330 just like ordinary text), making such attacks more likely to succeed.
29332 @item %u@var{suffix}
29333 Like @samp{%g}, but generates a new temporary file name
29334 each time it appears instead of once per compilation.
29336 @item %U@var{suffix}
29337 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
29338 new one if there is no such last file name. In the absence of any
29339 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
29340 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
29341 involves the generation of two distinct file names, one
29342 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
29343 simply substituted with a file name chosen for the previous @samp{%u},
29344 without regard to any appended suffix.
29346 @item %j@var{suffix}
29347 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
29348 writable, and if @option{-save-temps} is not used;
29349 otherwise, substitute the name
29350 of a temporary file, just like @samp{%u}. This temporary file is not
29351 meant for communication between processes, but rather as a junk
29352 disposal mechanism.
29354 @item %|@var{suffix}
29355 @itemx %m@var{suffix}
29356 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
29357 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
29358 all. These are the two most common ways to instruct a program that it
29359 should read from standard input or write to standard output. If you
29360 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
29361 construct: see for example @file{gcc/fortran/lang-specs.h}.
29363 @item %.@var{SUFFIX}
29364 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
29365 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
29366 terminated by the next space or %.
29369 Marks the argument containing or following the @samp{%w} as the
29370 designated output file of this compilation. This puts the argument
29371 into the sequence of arguments that @samp{%o} substitutes.
29374 Substitutes the names of all the output files, with spaces
29375 automatically placed around them. You should write spaces
29376 around the @samp{%o} as well or the results are undefined.
29377 @samp{%o} is for use in the specs for running the linker.
29378 Input files whose names have no recognized suffix are not compiled
29379 at all, but they are included among the output files, so they are
29383 Substitutes the suffix for object files. Note that this is
29384 handled specially when it immediately follows @samp{%g, %u, or %U},
29385 because of the need for those to form complete file names. The
29386 handling is such that @samp{%O} is treated exactly as if it had already
29387 been substituted, except that @samp{%g, %u, and %U} do not currently
29388 support additional @var{suffix} characters following @samp{%O} as they do
29389 following, for example, @samp{.o}.
29392 Substitutes the standard macro predefinitions for the
29393 current target machine. Use this when running @command{cpp}.
29396 Like @samp{%p}, but puts @samp{__} before and after the name of each
29397 predefined macro, except for macros that start with @samp{__} or with
29398 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
29402 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
29403 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
29404 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
29405 and @option{-imultilib} as necessary.
29408 Current argument is the name of a library or startup file of some sort.
29409 Search for that file in a standard list of directories and substitute
29410 the full name found. The current working directory is included in the
29411 list of directories scanned.
29414 Current argument is the name of a linker script. Search for that file
29415 in the current list of directories to scan for libraries. If the file
29416 is located insert a @option{--script} option into the command line
29417 followed by the full path name found. If the file is not found then
29418 generate an error message. Note: the current working directory is not
29422 Print @var{str} as an error message. @var{str} is terminated by a newline.
29423 Use this when inconsistent options are detected.
29425 @item %(@var{name})
29426 Substitute the contents of spec string @var{name} at this point.
29428 @item %x@{@var{option}@}
29429 Accumulate an option for @samp{%X}.
29432 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
29436 Output the accumulated assembler options specified by @option{-Wa}.
29439 Output the accumulated preprocessor options specified by @option{-Wp}.
29442 Process the @code{asm} spec. This is used to compute the
29443 switches to be passed to the assembler.
29446 Process the @code{asm_final} spec. This is a spec string for
29447 passing switches to an assembler post-processor, if such a program is
29451 Process the @code{link} spec. This is the spec for computing the
29452 command line passed to the linker. Typically it makes use of the
29453 @samp{%L %G %S %D and %E} sequences.
29456 Dump out a @option{-L} option for each directory that GCC believes might
29457 contain startup files. If the target supports multilibs then the
29458 current multilib directory is prepended to each of these paths.
29461 Process the @code{lib} spec. This is a spec string for deciding which
29462 libraries are included on the command line to the linker.
29465 Process the @code{libgcc} spec. This is a spec string for deciding
29466 which GCC support library is included on the command line to the linker.
29469 Process the @code{startfile} spec. This is a spec for deciding which
29470 object files are the first ones passed to the linker. Typically
29471 this might be a file named @file{crt0.o}.
29474 Process the @code{endfile} spec. This is a spec string that specifies
29475 the last object files that are passed to the linker.
29478 Process the @code{cpp} spec. This is used to construct the arguments
29479 to be passed to the C preprocessor.
29482 Process the @code{cc1} spec. This is used to construct the options to be
29483 passed to the actual C compiler (@command{cc1}).
29486 Process the @code{cc1plus} spec. This is used to construct the options to be
29487 passed to the actual C++ compiler (@command{cc1plus}).
29490 Substitute the variable part of a matched option. See below.
29491 Note that each comma in the substituted string is replaced by
29495 Remove all occurrences of @code{-S} from the command line. Note---this
29496 command is position dependent. @samp{%} commands in the spec string
29497 before this one see @code{-S}, @samp{%} commands in the spec string
29498 after this one do not.
29500 @item %:@var{function}(@var{args})
29501 Call the named function @var{function}, passing it @var{args}.
29502 @var{args} is first processed as a nested spec string, then split
29503 into an argument vector in the usual fashion. The function returns
29504 a string which is processed as if it had appeared literally as part
29505 of the current spec.
29507 The following built-in spec functions are provided:
29510 @item @code{getenv}
29511 The @code{getenv} spec function takes two arguments: an environment
29512 variable name and a string. If the environment variable is not
29513 defined, a fatal error is issued. Otherwise, the return value is the
29514 value of the environment variable concatenated with the string. For
29515 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
29518 %:getenv(TOPDIR /include)
29521 expands to @file{/path/to/top/include}.
29523 @item @code{if-exists}
29524 The @code{if-exists} spec function takes one argument, an absolute
29525 pathname to a file. If the file exists, @code{if-exists} returns the
29526 pathname. Here is a small example of its usage:
29530 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
29533 @item @code{if-exists-else}
29534 The @code{if-exists-else} spec function is similar to the @code{if-exists}
29535 spec function, except that it takes two arguments. The first argument is
29536 an absolute pathname to a file. If the file exists, @code{if-exists-else}
29537 returns the pathname. If it does not exist, it returns the second argument.
29538 This way, @code{if-exists-else} can be used to select one file or another,
29539 based on the existence of the first. Here is a small example of its usage:
29543 crt0%O%s %:if-exists(crti%O%s) \
29544 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
29547 @item @code{replace-outfile}
29548 The @code{replace-outfile} spec function takes two arguments. It looks for the
29549 first argument in the outfiles array and replaces it with the second argument. Here
29550 is a small example of its usage:
29553 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
29556 @item @code{remove-outfile}
29557 The @code{remove-outfile} spec function takes one argument. It looks for the
29558 first argument in the outfiles array and removes it. Here is a small example
29562 %:remove-outfile(-lm)
29565 @item @code{pass-through-libs}
29566 The @code{pass-through-libs} spec function takes any number of arguments. It
29567 finds any @option{-l} options and any non-options ending in @file{.a} (which it
29568 assumes are the names of linker input library archive files) and returns a
29569 result containing all the found arguments each prepended by
29570 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
29571 intended to be passed to the LTO linker plugin.
29574 %:pass-through-libs(%G %L %G)
29577 @item @code{print-asm-header}
29578 The @code{print-asm-header} function takes no arguments and simply
29579 prints a banner like:
29585 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
29588 It is used to separate compiler options from assembler options
29589 in the @option{--target-help} output.
29593 Substitutes the @code{-S} switch, if that switch is given to GCC@.
29594 If that switch is not specified, this substitutes nothing. Note that
29595 the leading dash is omitted when specifying this option, and it is
29596 automatically inserted if the substitution is performed. Thus the spec
29597 string @samp{%@{foo@}} matches the command-line option @option{-foo}
29598 and outputs the command-line option @option{-foo}.
29601 Like %@{@code{S}@} but mark last argument supplied within as a file to be
29602 deleted on failure.
29605 Substitutes all the switches specified to GCC whose names start
29606 with @code{-S}, but which also take an argument. This is used for
29607 switches like @option{-o}, @option{-D}, @option{-I}, etc.
29608 GCC considers @option{-o foo} as being
29609 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
29610 text, including the space. Thus two arguments are generated.
29613 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
29614 (the order of @code{S} and @code{T} in the spec is not significant).
29615 There can be any number of ampersand-separated variables; for each the
29616 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
29619 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
29622 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
29625 Substitutes @code{X} if one or more switches whose names start with
29626 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
29627 once, no matter how many such switches appeared. However, if @code{%*}
29628 appears somewhere in @code{X}, then @code{X} is substituted once
29629 for each matching switch, with the @code{%*} replaced by the part of
29630 that switch matching the @code{*}.
29632 If @code{%*} appears as the last part of a spec sequence then a space
29633 is added after the end of the last substitution. If there is more
29634 text in the sequence, however, then a space is not generated. This
29635 allows the @code{%*} substitution to be used as part of a larger
29636 string. For example, a spec string like this:
29639 %@{mcu=*:--script=%*/memory.ld@}
29643 when matching an option like @option{-mcu=newchip} produces:
29646 --script=newchip/memory.ld
29650 Substitutes @code{X}, if processing a file with suffix @code{S}.
29653 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
29656 Substitutes @code{X}, if processing a file for language @code{S}.
29659 Substitutes @code{X}, if not processing a file for language @code{S}.
29662 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
29663 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
29664 @code{*} sequences as well, although they have a stronger binding than
29665 the @samp{|}. If @code{%*} appears in @code{X}, all of the
29666 alternatives must be starred, and only the first matching alternative
29669 For example, a spec string like this:
29672 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
29676 outputs the following command-line options from the following input
29677 command-line options:
29682 -d fred.c -foo -baz -boggle
29683 -d jim.d -bar -baz -boggle
29686 @item %@{S:X; T:Y; :D@}
29688 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
29689 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
29690 be as many clauses as you need. This may be combined with @code{.},
29691 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
29696 The switch matching text @code{S} in a @samp{%@{S@}}, @samp{%@{S:X@}}
29697 or similar construct can use a backslash to ignore the special meaning
29698 of the character following it, thus allowing literal matching of a
29699 character that is otherwise specially treated. For example,
29700 @samp{%@{std=iso9899\:1999:X@}} substitutes @code{X} if the
29701 @option{-std=iso9899:1999} option is given.
29703 The conditional text @code{X} in a @samp{%@{S:X@}} or similar
29704 construct may contain other nested @samp{%} constructs or spaces, or
29705 even newlines. They are processed as usual, as described above.
29706 Trailing white space in @code{X} is ignored. White space may also
29707 appear anywhere on the left side of the colon in these constructs,
29708 except between @code{.} or @code{*} and the corresponding word.
29710 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
29711 handled specifically in these constructs. If another value of
29712 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
29713 @option{-W} switch is found later in the command line, the earlier
29714 switch value is ignored, except with @{@code{S}*@} where @code{S} is
29715 just one letter, which passes all matching options.
29717 The character @samp{|} at the beginning of the predicate text is used to
29718 indicate that a command should be piped to the following command, but
29719 only if @option{-pipe} is specified.
29721 It is built into GCC which switches take arguments and which do not.
29722 (You might think it would be useful to generalize this to allow each
29723 compiler's spec to say which switches take arguments. But this cannot
29724 be done in a consistent fashion. GCC cannot even decide which input
29725 files have been specified without knowing which switches take arguments,
29726 and it must know which input files to compile in order to tell which
29729 GCC also knows implicitly that arguments starting in @option{-l} are to be
29730 treated as compiler output files, and passed to the linker in their
29731 proper position among the other output files.
29733 @node Environment Variables
29734 @section Environment Variables Affecting GCC
29735 @cindex environment variables
29737 @c man begin ENVIRONMENT
29738 This section describes several environment variables that affect how GCC
29739 operates. Some of them work by specifying directories or prefixes to use
29740 when searching for various kinds of files. Some are used to specify other
29741 aspects of the compilation environment.
29743 Note that you can also specify places to search using options such as
29744 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
29745 take precedence over places specified using environment variables, which
29746 in turn take precedence over those specified by the configuration of GCC@.
29747 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
29748 GNU Compiler Collection (GCC) Internals}.
29753 @c @itemx LC_COLLATE
29755 @c @itemx LC_MONETARY
29756 @c @itemx LC_NUMERIC
29761 @c @findex LC_COLLATE
29762 @findex LC_MESSAGES
29763 @c @findex LC_MONETARY
29764 @c @findex LC_NUMERIC
29768 These environment variables control the way that GCC uses
29769 localization information which allows GCC to work with different
29770 national conventions. GCC inspects the locale categories
29771 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
29772 so. These locale categories can be set to any value supported by your
29773 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
29774 Kingdom encoded in UTF-8.
29776 The @env{LC_CTYPE} environment variable specifies character
29777 classification. GCC uses it to determine the character boundaries in
29778 a string; this is needed for some multibyte encodings that contain quote
29779 and escape characters that are otherwise interpreted as a string
29782 The @env{LC_MESSAGES} environment variable specifies the language to
29783 use in diagnostic messages.
29785 If the @env{LC_ALL} environment variable is set, it overrides the value
29786 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
29787 and @env{LC_MESSAGES} default to the value of the @env{LANG}
29788 environment variable. If none of these variables are set, GCC
29789 defaults to traditional C English behavior.
29793 If @env{TMPDIR} is set, it specifies the directory to use for temporary
29794 files. GCC uses temporary files to hold the output of one stage of
29795 compilation which is to be used as input to the next stage: for example,
29796 the output of the preprocessor, which is the input to the compiler
29799 @item GCC_COMPARE_DEBUG
29800 @findex GCC_COMPARE_DEBUG
29801 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
29802 @option{-fcompare-debug} to the compiler driver. See the documentation
29803 of this option for more details.
29805 @item GCC_EXEC_PREFIX
29806 @findex GCC_EXEC_PREFIX
29807 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
29808 names of the subprograms executed by the compiler. No slash is added
29809 when this prefix is combined with the name of a subprogram, but you can
29810 specify a prefix that ends with a slash if you wish.
29812 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
29813 an appropriate prefix to use based on the pathname it is invoked with.
29815 If GCC cannot find the subprogram using the specified prefix, it
29816 tries looking in the usual places for the subprogram.
29818 The default value of @env{GCC_EXEC_PREFIX} is
29819 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
29820 the installed compiler. In many cases @var{prefix} is the value
29821 of @code{prefix} when you ran the @file{configure} script.
29823 Other prefixes specified with @option{-B} take precedence over this prefix.
29825 This prefix is also used for finding files such as @file{crt0.o} that are
29828 In addition, the prefix is used in an unusual way in finding the
29829 directories to search for header files. For each of the standard
29830 directories whose name normally begins with @samp{/usr/local/lib/gcc}
29831 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
29832 replacing that beginning with the specified prefix to produce an
29833 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
29834 @file{foo/bar} just before it searches the standard directory
29835 @file{/usr/local/lib/bar}.
29836 If a standard directory begins with the configured
29837 @var{prefix} then the value of @var{prefix} is replaced by
29838 @env{GCC_EXEC_PREFIX} when looking for header files.
29840 @item COMPILER_PATH
29841 @findex COMPILER_PATH
29842 The value of @env{COMPILER_PATH} is a colon-separated list of
29843 directories, much like @env{PATH}. GCC tries the directories thus
29844 specified when searching for subprograms, if it cannot find the
29845 subprograms using @env{GCC_EXEC_PREFIX}.
29848 @findex LIBRARY_PATH
29849 The value of @env{LIBRARY_PATH} is a colon-separated list of
29850 directories, much like @env{PATH}. When configured as a native compiler,
29851 GCC tries the directories thus specified when searching for special
29852 linker files, if it cannot find them using @env{GCC_EXEC_PREFIX}. Linking
29853 using GCC also uses these directories when searching for ordinary
29854 libraries for the @option{-l} option (but directories specified with
29855 @option{-L} come first).
29859 @cindex locale definition
29860 This variable is used to pass locale information to the compiler. One way in
29861 which this information is used is to determine the character set to be used
29862 when character literals, string literals and comments are parsed in C and C++.
29863 When the compiler is configured to allow multibyte characters,
29864 the following values for @env{LANG} are recognized:
29868 Recognize JIS characters.
29870 Recognize SJIS characters.
29872 Recognize EUCJP characters.
29875 If @env{LANG} is not defined, or if it has some other value, then the
29876 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
29877 recognize and translate multibyte characters.
29881 Some additional environment variables affect the behavior of the
29884 @include cppenv.texi
29888 @node Precompiled Headers
29889 @section Using Precompiled Headers
29890 @cindex precompiled headers
29891 @cindex speed of compilation
29893 Often large projects have many header files that are included in every
29894 source file. The time the compiler takes to process these header files
29895 over and over again can account for nearly all of the time required to
29896 build the project. To make builds faster, GCC allows you to
29897 @dfn{precompile} a header file.
29899 To create a precompiled header file, simply compile it as you would any
29900 other file, if necessary using the @option{-x} option to make the driver
29901 treat it as a C or C++ header file. You may want to use a
29902 tool like @command{make} to keep the precompiled header up-to-date when
29903 the headers it contains change.
29905 A precompiled header file is searched for when @code{#include} is
29906 seen in the compilation. As it searches for the included file
29907 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
29908 compiler looks for a precompiled header in each directory just before it
29909 looks for the include file in that directory. The name searched for is
29910 the name specified in the @code{#include} with @samp{.gch} appended. If
29911 the precompiled header file cannot be used, it is ignored.
29913 For instance, if you have @code{#include "all.h"}, and you have
29914 @file{all.h.gch} in the same directory as @file{all.h}, then the
29915 precompiled header file is used if possible, and the original
29916 header is used otherwise.
29918 Alternatively, you might decide to put the precompiled header file in a
29919 directory and use @option{-I} to ensure that directory is searched
29920 before (or instead of) the directory containing the original header.
29921 Then, if you want to check that the precompiled header file is always
29922 used, you can put a file of the same name as the original header in this
29923 directory containing an @code{#error} command.
29925 This also works with @option{-include}. So yet another way to use
29926 precompiled headers, good for projects not designed with precompiled
29927 header files in mind, is to simply take most of the header files used by
29928 a project, include them from another header file, precompile that header
29929 file, and @option{-include} the precompiled header. If the header files
29930 have guards against multiple inclusion, they are skipped because
29931 they've already been included (in the precompiled header).
29933 If you need to precompile the same header file for different
29934 languages, targets, or compiler options, you can instead make a
29935 @emph{directory} named like @file{all.h.gch}, and put each precompiled
29936 header in the directory, perhaps using @option{-o}. It doesn't matter
29937 what you call the files in the directory; every precompiled header in
29938 the directory is considered. The first precompiled header
29939 encountered in the directory that is valid for this compilation is
29940 used; they're searched in no particular order.
29942 There are many other possibilities, limited only by your imagination,
29943 good sense, and the constraints of your build system.
29945 A precompiled header file can be used only when these conditions apply:
29949 Only one precompiled header can be used in a particular compilation.
29952 A precompiled header cannot be used once the first C token is seen. You
29953 can have preprocessor directives before a precompiled header; you cannot
29954 include a precompiled header from inside another header.
29957 The precompiled header file must be produced for the same language as
29958 the current compilation. You cannot use a C precompiled header for a C++
29962 The precompiled header file must have been produced by the same compiler
29963 binary as the current compilation is using.
29966 Any macros defined before the precompiled header is included must
29967 either be defined in the same way as when the precompiled header was
29968 generated, or must not affect the precompiled header, which usually
29969 means that they don't appear in the precompiled header at all.
29971 The @option{-D} option is one way to define a macro before a
29972 precompiled header is included; using a @code{#define} can also do it.
29973 There are also some options that define macros implicitly, like
29974 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
29977 @item If debugging information is output when using the precompiled
29978 header, using @option{-g} or similar, the same kind of debugging information
29979 must have been output when building the precompiled header. However,
29980 a precompiled header built using @option{-g} can be used in a compilation
29981 when no debugging information is being output.
29983 @item The same @option{-m} options must generally be used when building
29984 and using the precompiled header. @xref{Submodel Options},
29985 for any cases where this rule is relaxed.
29987 @item Each of the following options must be the same when building and using
29988 the precompiled header:
29990 @gccoptlist{-fexceptions}
29993 Some other command-line options starting with @option{-f},
29994 @option{-p}, or @option{-O} must be defined in the same way as when
29995 the precompiled header was generated. At present, it's not clear
29996 which options are safe to change and which are not; the safest choice
29997 is to use exactly the same options when generating and using the
29998 precompiled header. The following are known to be safe:
30000 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
30001 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
30002 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
30007 For all of these except the last, the compiler automatically
30008 ignores the precompiled header if the conditions aren't met. If you
30009 find an option combination that doesn't work and doesn't cause the
30010 precompiled header to be ignored, please consider filing a bug report,
30013 If you do use differing options when generating and using the
30014 precompiled header, the actual behavior is a mixture of the
30015 behavior for the options. For instance, if you use @option{-g} to
30016 generate the precompiled header but not when using it, you may or may
30017 not get debugging information for routines in the precompiled header.