1 @c Copyright (C) 1988-2019 Free Software Foundation, Inc.
2 @c This is part of the GCC manual.
3 @c For copying conditions, see the file gcc.texi.
10 @c man begin COPYRIGHT
11 Copyright @copyright{} 1988-2019 Free Software Foundation, Inc.
13 Permission is granted to copy, distribute and/or modify this document
14 under the terms of the GNU Free Documentation License, Version 1.3 or
15 any later version published by the Free Software Foundation; with the
16 Invariant Sections being ``GNU General Public License'' and ``Funding
17 Free Software'', the Front-Cover texts being (a) (see below), and with
18 the Back-Cover Texts being (b) (see below). A copy of the license is
19 included in the gfdl(7) man page.
21 (a) The FSF's Front-Cover Text is:
25 (b) The FSF's Back-Cover Text is:
27 You have freedom to copy and modify this GNU Manual, like GNU
28 software. Copies published by the Free Software Foundation raise
29 funds for GNU development.
31 @c Set file name and title for the man page.
33 @settitle GNU project C and C++ compiler
35 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
36 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
37 [@option{-W}@var{warn}@dots{}] [@option{-Wpedantic}]
38 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
39 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
40 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
41 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
43 Only the most useful options are listed here; see below for the
44 remainder. @command{g++} accepts mostly the same options as @command{gcc}.
47 gpl(7), gfdl(7), fsf-funding(7),
48 cpp(1), gcov(1), as(1), ld(1), gdb(1), dbx(1)
49 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
50 @file{ld}, @file{binutils} and @file{gdb}.
53 For instructions on reporting bugs, see
57 See the Info entry for @command{gcc}, or
58 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
59 for contributors to GCC@.
64 @chapter GCC Command Options
65 @cindex GCC command options
66 @cindex command options
67 @cindex options, GCC command
69 @c man begin DESCRIPTION
70 When you invoke GCC, it normally does preprocessing, compilation,
71 assembly and linking. The ``overall options'' allow you to stop this
72 process at an intermediate stage. For example, the @option{-c} option
73 says not to run the linker. Then the output consists of object files
74 output by the assembler.
75 @xref{Overall Options,,Options Controlling the Kind of Output}.
77 Other options are passed on to one or more stages of processing. Some options
78 control the preprocessor and others the compiler itself. Yet other
79 options control the assembler and linker; most of these are not
80 documented here, since you rarely need to use any of them.
82 @cindex C compilation options
83 Most of the command-line options that you can use with GCC are useful
84 for C programs; when an option is only useful with another language
85 (usually C++), the explanation says so explicitly. If the description
86 for a particular option does not mention a source language, you can use
87 that option with all supported languages.
89 @cindex cross compiling
90 @cindex specifying machine version
91 @cindex specifying compiler version and target machine
92 @cindex compiler version, specifying
93 @cindex target machine, specifying
94 The usual way to run GCC is to run the executable called @command{gcc}, or
95 @command{@var{machine}-gcc} when cross-compiling, or
96 @command{@var{machine}-gcc-@var{version}} to run a specific version of GCC.
97 When you compile C++ programs, you should invoke GCC as @command{g++}
98 instead. @xref{Invoking G++,,Compiling C++ Programs},
99 for information about the differences in behavior between @command{gcc}
100 and @code{g++} when compiling C++ programs.
102 @cindex grouping options
103 @cindex options, grouping
104 The @command{gcc} program accepts options and file names as operands. Many
105 options have multi-letter names; therefore multiple single-letter options
106 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
109 @cindex order of options
110 @cindex options, order
111 You can mix options and other arguments. For the most part, the order
112 you use doesn't matter. Order does matter when you use several
113 options of the same kind; for example, if you specify @option{-L} more
114 than once, the directories are searched in the order specified. Also,
115 the placement of the @option{-l} option is significant.
117 Many options have long names starting with @samp{-f} or with
118 @samp{-W}---for example,
119 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
120 these have both positive and negative forms; the negative form of
121 @option{-ffoo} is @option{-fno-foo}. This manual documents
122 only one of these two forms, whichever one is not the default.
124 Some options take one or more arguments typically separated either
125 by a space or by the equals sign (@samp{=}) from the option name.
126 Unless documented otherwise, an argument can be either numeric or
127 a string. Numeric arguments must typically be small unsigned decimal
128 or hexadecimal integers. Hexadecimal arguments must begin with
129 the @samp{0x} prefix. Arguments to options that specify a size
130 threshold of some sort may be arbitrarily large decimal or hexadecimal
131 integers followed by a byte size suffix designating a multiple of bytes
132 such as @code{kB} and @code{KiB} for kilobyte and kibibyte, respectively,
133 @code{MB} and @code{MiB} for megabyte and mebibyte, @code{GB} and
134 @code{GiB} for gigabyte and gigibyte, and so on. Such arguments are
135 designated by @var{byte-size} in the following text. Refer to the NIST,
136 IEC, and other relevant national and international standards for the full
137 listing and explanation of the binary and decimal byte size prefixes.
141 @xref{Option Index}, for an index to GCC's options.
144 * Option Summary:: Brief list of all options, without explanations.
145 * Overall Options:: Controlling the kind of output:
146 an executable, object files, assembler files,
147 or preprocessed source.
148 * Invoking G++:: Compiling C++ programs.
149 * C Dialect Options:: Controlling the variant of C language compiled.
150 * C++ Dialect Options:: Variations on C++.
151 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
153 * Diagnostic Message Formatting Options:: Controlling how diagnostics should
155 * Warning Options:: How picky should the compiler be?
156 * Debugging Options:: Producing debuggable code.
157 * Optimize Options:: How much optimization?
158 * Instrumentation Options:: Enabling profiling and extra run-time error checking.
159 * Preprocessor Options:: Controlling header files and macro definitions.
160 Also, getting dependency information for Make.
161 * Assembler Options:: Passing options to the assembler.
162 * Link Options:: Specifying libraries and so on.
163 * Directory Options:: Where to find header files and libraries.
164 Where to find the compiler executable files.
165 * Code Gen Options:: Specifying conventions for function calls, data layout
167 * Developer Options:: Printing GCC configuration info, statistics, and
169 * Submodel Options:: Target-specific options, such as compiling for a
170 specific processor variant.
171 * Spec Files:: How to pass switches to sub-processes.
172 * Environment Variables:: Env vars that affect GCC.
173 * Precompiled Headers:: Compiling a header once, and using it many times.
179 @section Option Summary
181 Here is a summary of all the options, grouped by type. Explanations are
182 in the following sections.
185 @item Overall Options
186 @xref{Overall Options,,Options Controlling the Kind of Output}.
187 @gccoptlist{-c -S -E -o @var{file} -x @var{language} @gol
188 -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help --version @gol
189 -pass-exit-codes -pipe -specs=@var{file} -wrapper @gol
190 @@@var{file} -ffile-prefix-map=@var{old}=@var{new} @gol
191 -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
192 -fdump-ada-spec@r{[}-slim@r{]} -fada-spec-parent=@var{unit} -fdump-go-spec=@var{file}}
194 @item C Language Options
195 @xref{C Dialect Options,,Options Controlling C Dialect}.
196 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
197 -fpermitted-flt-eval-methods=@var{standard} @gol
198 -aux-info @var{filename} -fallow-parameterless-variadic-functions @gol
199 -fno-asm -fno-builtin -fno-builtin-@var{function} -fgimple@gol
200 -fhosted -ffreestanding @gol
201 -fopenacc -fopenacc-dim=@var{geom} @gol
202 -fopenmp -fopenmp-simd @gol
203 -fms-extensions -fplan9-extensions -fsso-struct=@var{endianness} @gol
204 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
205 -fsigned-bitfields -fsigned-char @gol
206 -funsigned-bitfields -funsigned-char}
208 @item C++ Language Options
209 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
210 @gccoptlist{-fabi-version=@var{n} -fno-access-control @gol
211 -faligned-new=@var{n} -fargs-in-order=@var{n} -fchar8_t -fcheck-new @gol
212 -fconstexpr-depth=@var{n} -fconstexpr-loop-limit=@var{n} @gol
213 -fconstexpr-ops-limit=@var{n} -fno-elide-constructors @gol
214 -fno-enforce-eh-specs @gol
215 -fno-gnu-keywords @gol
216 -fno-implicit-templates @gol
217 -fno-implicit-inline-templates @gol
218 -fno-implement-inlines -fms-extensions @gol
219 -fnew-inheriting-ctors @gol
220 -fnew-ttp-matching @gol
221 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
222 -fno-optional-diags -fpermissive @gol
223 -fno-pretty-templates @gol
224 -frepo -fno-rtti -fsized-deallocation @gol
225 -ftemplate-backtrace-limit=@var{n} @gol
226 -ftemplate-depth=@var{n} @gol
227 -fno-threadsafe-statics -fuse-cxa-atexit @gol
228 -fno-weak -nostdinc++ @gol
229 -fvisibility-inlines-hidden @gol
230 -fvisibility-ms-compat @gol
231 -fext-numeric-literals @gol
232 -Wabi=@var{n} -Wabi-tag -Wconversion-null -Wctor-dtor-privacy @gol
233 -Wdelete-non-virtual-dtor -Wdeprecated-copy -Wdeprecated-copy-dtor @gol
234 -Wliteral-suffix @gol
235 -Wmultiple-inheritance -Wno-init-list-lifetime @gol
236 -Wnamespaces -Wnarrowing @gol
237 -Wpessimizing-move -Wredundant-move @gol
238 -Wnoexcept -Wnoexcept-type -Wclass-memaccess @gol
239 -Wnon-virtual-dtor -Wreorder -Wregister @gol
240 -Weffc++ -Wstrict-null-sentinel -Wtemplates @gol
241 -Wno-non-template-friend -Wold-style-cast @gol
242 -Woverloaded-virtual -Wno-pmf-conversions @gol
243 -Wno-class-conversion -Wno-terminate @gol
244 -Wsign-promo -Wvirtual-inheritance}
246 @item Objective-C and Objective-C++ Language Options
247 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
248 Objective-C and Objective-C++ Dialects}.
249 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
250 -fgnu-runtime -fnext-runtime @gol
251 -fno-nil-receivers @gol
252 -fobjc-abi-version=@var{n} @gol
253 -fobjc-call-cxx-cdtors @gol
254 -fobjc-direct-dispatch @gol
255 -fobjc-exceptions @gol
258 -fobjc-std=objc1 @gol
259 -fno-local-ivars @gol
260 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
261 -freplace-objc-classes @gol
264 -Wassign-intercept @gol
265 -Wno-protocol -Wselector @gol
266 -Wstrict-selector-match @gol
267 -Wundeclared-selector}
269 @item Diagnostic Message Formatting Options
270 @xref{Diagnostic Message Formatting Options,,Options to Control Diagnostic Messages Formatting}.
271 @gccoptlist{-fmessage-length=@var{n} @gol
272 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
273 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
274 -fdiagnostics-format=@r{[}text@r{|}json@r{]} @gol
275 -fno-diagnostics-show-option -fno-diagnostics-show-caret @gol
276 -fno-diagnostics-show-labels -fno-diagnostics-show-line-numbers @gol
277 -fdiagnostics-minimum-margin-width=@var{width} @gol
278 -fdiagnostics-parseable-fixits -fdiagnostics-generate-patch @gol
279 -fdiagnostics-show-template-tree -fno-elide-type @gol
282 @item Warning Options
283 @xref{Warning Options,,Options to Request or Suppress Warnings}.
284 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
285 -pedantic-errors @gol
286 -w -Wextra -Wall -Waddress -Waddress-of-packed-member @gol
287 -Waggregate-return -Waligned-new @gol
288 -Walloc-zero -Walloc-size-larger-than=@var{byte-size} @gol
289 -Walloca -Walloca-larger-than=@var{byte-size} @gol
290 -Wno-aggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
291 -Wno-attributes -Wattribute-alias=@var{n} @gol
292 -Wbool-compare -Wbool-operation @gol
293 -Wno-builtin-declaration-mismatch @gol
294 -Wno-builtin-macro-redefined -Wc90-c99-compat -Wc99-c11-compat @gol
295 -Wc++-compat -Wc++11-compat -Wc++14-compat -Wc++17-compat @gol
296 -Wcast-align -Wcast-align=strict -Wcast-function-type -Wcast-qual @gol
297 -Wchar-subscripts -Wcatch-value -Wcatch-value=@var{n} @gol
298 -Wclobbered -Wcomment -Wconditionally-supported @gol
299 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wdangling-else -Wdate-time @gol
300 -Wdelete-incomplete @gol
301 -Wno-attribute-warning @gol
302 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
303 -Wdisabled-optimization @gol
304 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
305 -Wno-div-by-zero -Wdouble-promotion @gol
306 -Wduplicated-branches -Wduplicated-cond @gol
307 -Wempty-body -Wenum-compare -Wno-endif-labels -Wexpansion-to-defined @gol
308 -Werror -Werror=* -Wextra-semi -Wfatal-errors @gol
309 -Wfloat-equal -Wformat -Wformat=2 @gol
310 -Wno-format-contains-nul -Wno-format-extra-args @gol
311 -Wformat-nonliteral -Wformat-overflow=@var{n} @gol
312 -Wformat-security -Wformat-signedness -Wformat-truncation=@var{n} @gol
313 -Wformat-y2k -Wframe-address @gol
314 -Wframe-larger-than=@var{byte-size} -Wno-free-nonheap-object @gol
315 -Wjump-misses-init @gol
316 -Whsa -Wif-not-aligned @gol
317 -Wignored-qualifiers -Wignored-attributes -Wincompatible-pointer-types @gol
318 -Wimplicit -Wimplicit-fallthrough -Wimplicit-fallthrough=@var{n} @gol
319 -Wimplicit-function-declaration -Wimplicit-int @gol
320 -Winit-self -Winline -Wno-int-conversion -Wint-in-bool-context @gol
321 -Wno-int-to-pointer-cast -Winvalid-memory-model -Wno-invalid-offsetof @gol
322 -Winvalid-pch -Wlarger-than=@var{byte-size} @gol
323 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
324 -Wmain -Wmaybe-uninitialized -Wmemset-elt-size -Wmemset-transposed-args @gol
325 -Wmisleading-indentation -Wmissing-attributes -Wmissing-braces @gol
326 -Wmissing-field-initializers -Wmissing-format-attribute @gol
327 -Wmissing-include-dirs -Wmissing-noreturn -Wmissing-profile @gol
328 -Wno-multichar -Wmultistatement-macros -Wnonnull -Wnonnull-compare @gol
329 -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
330 -Wnull-dereference -Wodr -Wno-overflow -Wopenmp-simd @gol
331 -Woverride-init-side-effects -Woverlength-strings @gol
332 -Wpacked -Wpacked-bitfield-compat -Wpacked-not-aligned -Wpadded @gol
333 -Wparentheses -Wno-pedantic-ms-format @gol
334 -Wplacement-new -Wplacement-new=@var{n} @gol
335 -Wpointer-arith -Wpointer-compare -Wno-pointer-to-int-cast @gol
336 -Wno-pragmas -Wno-prio-ctor-dtor -Wredundant-decls @gol
337 -Wrestrict -Wno-return-local-addr @gol
338 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
339 -Wshadow=global, -Wshadow=local, -Wshadow=compatible-local @gol
340 -Wshift-overflow -Wshift-overflow=@var{n} @gol
341 -Wshift-count-negative -Wshift-count-overflow -Wshift-negative-value @gol
342 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
343 -Wno-scalar-storage-order -Wsizeof-pointer-div @gol
344 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
345 -Wstack-protector -Wstack-usage=@var{byte-size} -Wstrict-aliasing @gol
346 -Wstrict-aliasing=n -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
347 -Wstringop-overflow=@var{n} -Wstringop-truncation -Wsubobject-linkage @gol
348 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{|}malloc@r{]} @gol
349 -Wsuggest-final-types @gol -Wsuggest-final-methods -Wsuggest-override @gol
350 -Wswitch -Wswitch-bool -Wswitch-default -Wswitch-enum @gol
351 -Wswitch-unreachable -Wsync-nand @gol
352 -Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs @gol
353 -Wtype-limits -Wundef @gol
354 -Wuninitialized -Wunknown-pragmas @gol
355 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
356 -Wunused-label -Wunused-local-typedefs -Wunused-macros @gol
357 -Wunused-parameter -Wno-unused-result @gol
358 -Wunused-value -Wunused-variable @gol
359 -Wunused-const-variable -Wunused-const-variable=@var{n} @gol
360 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
361 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
362 -Wvla -Wvla-larger-than=@var{byte-size} -Wvolatile-register-var @gol
364 -Wzero-as-null-pointer-constant}
366 @item C and Objective-C-only Warning Options
367 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
368 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
369 -Wold-style-declaration -Wold-style-definition @gol
370 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
371 -Wdeclaration-after-statement -Wpointer-sign}
373 @item Debugging Options
374 @xref{Debugging Options,,Options for Debugging Your Program}.
375 @gccoptlist{-g -g@var{level} -gdwarf -gdwarf-@var{version} @gol
376 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
377 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
378 -gas-loc-support -gno-as-loc-support @gol
379 -gas-locview-support -gno-as-locview-support @gol
380 -gcolumn-info -gno-column-info @gol
381 -gstatement-frontiers -gno-statement-frontiers @gol
382 -gvariable-location-views -gno-variable-location-views @gol
383 -ginternal-reset-location-views -gno-internal-reset-location-views @gol
384 -ginline-points -gno-inline-points @gol
385 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
386 -gsplit-dwarf -gdescribe-dies -gno-describe-dies @gol
387 -fdebug-prefix-map=@var{old}=@var{new} -fdebug-types-section @gol
388 -fno-eliminate-unused-debug-types @gol
389 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
390 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
391 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
392 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
393 -fvar-tracking -fvar-tracking-assignments}
395 @item Optimization Options
396 @xref{Optimize Options,,Options that Control Optimization}.
397 @gccoptlist{-faggressive-loop-optimizations @gol
398 -falign-functions[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
399 -falign-jumps[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
400 -falign-labels[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
401 -falign-loops[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
402 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
403 -fauto-inc-dec -fbranch-probabilities @gol
404 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
405 -fbtr-bb-exclusive -fcaller-saves @gol
406 -fcombine-stack-adjustments -fconserve-stack @gol
407 -fcompare-elim -fcprop-registers -fcrossjumping @gol
408 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
409 -fcx-limited-range @gol
410 -fdata-sections -fdce -fdelayed-branch @gol
411 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
412 -fdevirtualize-at-ltrans -fdse @gol
413 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
414 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
415 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
416 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
417 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
418 -fif-conversion2 -findirect-inlining @gol
419 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
420 -finline-small-functions -fipa-cp -fipa-cp-clone @gol
421 -fipa-bit-cp -fipa-vrp -fipa-pta -fipa-profile -fipa-pure-const @gol
422 -fipa-reference -fipa-reference-addressable @gol
423 -fipa-stack-alignment -fipa-icf -fira-algorithm=@var{algorithm} @gol
424 -flive-patching=@var{level} @gol
425 -fira-region=@var{region} -fira-hoist-pressure @gol
426 -fira-loop-pressure -fno-ira-share-save-slots @gol
427 -fno-ira-share-spill-slots @gol
428 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
429 -fivopts -fkeep-inline-functions -fkeep-static-functions @gol
430 -fkeep-static-consts -flimit-function-alignment -flive-range-shrinkage @gol
431 -floop-block -floop-interchange -floop-strip-mine @gol
432 -floop-unroll-and-jam -floop-nest-optimize @gol
433 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
434 -flto-partition=@var{alg} -fmerge-all-constants @gol
435 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
436 -fmove-loop-invariants -fno-branch-count-reg @gol
437 -fno-defer-pop -fno-fp-int-builtin-inexact -fno-function-cse @gol
438 -fno-guess-branch-probability -fno-inline -fno-math-errno -fno-peephole @gol
439 -fno-peephole2 -fno-printf-return-value -fno-sched-interblock @gol
440 -fno-sched-spec -fno-signed-zeros @gol
441 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
442 -fomit-frame-pointer -foptimize-sibling-calls @gol
443 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
444 -fprefetch-loop-arrays @gol
445 -fprofile-correction @gol
446 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
447 -fprofile-reorder-functions @gol
448 -freciprocal-math -free -frename-registers -freorder-blocks @gol
449 -freorder-blocks-algorithm=@var{algorithm} @gol
450 -freorder-blocks-and-partition -freorder-functions @gol
451 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
452 -frounding-math -fsave-optimization-record @gol
453 -fsched2-use-superblocks -fsched-pressure @gol
454 -fsched-spec-load -fsched-spec-load-dangerous @gol
455 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
456 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
457 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
458 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
459 -fschedule-fusion @gol
460 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
461 -fselective-scheduling -fselective-scheduling2 @gol
462 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
463 -fsemantic-interposition -fshrink-wrap -fshrink-wrap-separate @gol
464 -fsignaling-nans @gol
465 -fsingle-precision-constant -fsplit-ivs-in-unroller -fsplit-loops@gol
467 -fsplit-wide-types -fssa-backprop -fssa-phiopt @gol
468 -fstdarg-opt -fstore-merging -fstrict-aliasing @gol
469 -fthread-jumps -ftracer -ftree-bit-ccp @gol
470 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
471 -ftree-coalesce-vars -ftree-copy-prop -ftree-dce -ftree-dominator-opts @gol
472 -ftree-dse -ftree-forwprop -ftree-fre -fcode-hoisting @gol
473 -ftree-loop-if-convert -ftree-loop-im @gol
474 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
475 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
476 -ftree-loop-vectorize @gol
477 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
478 -ftree-reassoc -ftree-scev-cprop -ftree-sink -ftree-slsr -ftree-sra @gol
479 -ftree-switch-conversion -ftree-tail-merge @gol
480 -ftree-ter -ftree-vectorize -ftree-vrp -funconstrained-commons @gol
481 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
482 -funsafe-math-optimizations -funswitch-loops @gol
483 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
484 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
485 --param @var{name}=@var{value}
486 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
488 @item Program Instrumentation Options
489 @xref{Instrumentation Options,,Program Instrumentation Options}.
490 @gccoptlist{-p -pg -fprofile-arcs --coverage -ftest-coverage @gol
491 -fprofile-abs-path @gol
492 -fprofile-dir=@var{path} -fprofile-generate -fprofile-generate=@var{path} @gol
493 -fprofile-update=@var{method} -fprofile-filter-files=@var{regex} @gol
494 -fprofile-exclude-files=@var{regex} @gol
495 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
496 -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
497 -fsanitize-undefined-trap-on-error -fbounds-check @gol
498 -fcf-protection=@r{[}full@r{|}branch@r{|}return@r{|}none@r{]} @gol
499 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
500 -fstack-protector-explicit -fstack-check @gol
501 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
502 -fno-stack-limit -fsplit-stack @gol
503 -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} @gol
504 -fvtv-counts -fvtv-debug @gol
505 -finstrument-functions @gol
506 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
507 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}}
509 @item Preprocessor Options
510 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
511 @gccoptlist{-A@var{question}=@var{answer} @gol
512 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
513 -C -CC -D@var{macro}@r{[}=@var{defn}@r{]} @gol
514 -dD -dI -dM -dN -dU @gol
515 -fdebug-cpp -fdirectives-only -fdollars-in-identifiers @gol
516 -fexec-charset=@var{charset} -fextended-identifiers @gol
517 -finput-charset=@var{charset} -fmacro-prefix-map=@var{old}=@var{new} @gol
518 -fno-canonical-system-headers -fpch-deps -fpch-preprocess @gol
519 -fpreprocessed -ftabstop=@var{width} -ftrack-macro-expansion @gol
520 -fwide-exec-charset=@var{charset} -fworking-directory @gol
521 -H -imacros @var{file} -include @var{file} @gol
522 -M -MD -MF -MG -MM -MMD -MP -MQ -MT @gol
523 -no-integrated-cpp -P -pthread -remap @gol
524 -traditional -traditional-cpp -trigraphs @gol
525 -U@var{macro} -undef @gol
526 -Wp,@var{option} -Xpreprocessor @var{option}}
528 @item Assembler Options
529 @xref{Assembler Options,,Passing Options to the Assembler}.
530 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
533 @xref{Link Options,,Options for Linking}.
534 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
535 -nostartfiles -nodefaultlibs -nolibc -nostdlib @gol
536 -e @var{entry} --entry=@var{entry} @gol
537 -pie -pthread -r -rdynamic @gol
538 -s -static -static-pie -static-libgcc -static-libstdc++ @gol
539 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
540 -shared -shared-libgcc -symbolic @gol
541 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
542 -u @var{symbol} -z @var{keyword}}
544 @item Directory Options
545 @xref{Directory Options,,Options for Directory Search}.
546 @gccoptlist{-B@var{prefix} -I@var{dir} -I- @gol
547 -idirafter @var{dir} @gol
548 -imacros @var{file} -imultilib @var{dir} @gol
549 -iplugindir=@var{dir} -iprefix @var{file} @gol
550 -iquote @var{dir} -isysroot @var{dir} -isystem @var{dir} @gol
551 -iwithprefix @var{dir} -iwithprefixbefore @var{dir} @gol
552 -L@var{dir} -no-canonical-prefixes --no-sysroot-suffix @gol
553 -nostdinc -nostdinc++ --sysroot=@var{dir}}
555 @item Code Generation Options
556 @xref{Code Gen Options,,Options for Code Generation Conventions}.
557 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
558 -ffixed-@var{reg} -fexceptions @gol
559 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
560 -fasynchronous-unwind-tables @gol
562 -finhibit-size-directive -fno-common -fno-ident @gol
563 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol
564 -fno-jump-tables @gol
565 -frecord-gcc-switches @gol
566 -freg-struct-return -fshort-enums -fshort-wchar @gol
567 -fverbose-asm -fpack-struct[=@var{n}] @gol
568 -fleading-underscore -ftls-model=@var{model} @gol
569 -fstack-reuse=@var{reuse_level} @gol
570 -ftrampolines -ftrapv -fwrapv @gol
571 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
572 -fstrict-volatile-bitfields -fsync-libcalls}
574 @item Developer Options
575 @xref{Developer Options,,GCC Developer Options}.
576 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
577 -dumpfullversion -fchecking -fchecking=@var{n} -fdbg-cnt-list @gol
578 -fdbg-cnt=@var{counter-value-list} @gol
579 -fdisable-ipa-@var{pass_name} @gol
580 -fdisable-rtl-@var{pass_name} @gol
581 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
582 -fdisable-tree-@var{pass_name} @gol
583 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
584 -fdump-debug -fdump-earlydebug @gol
585 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
586 -fdump-final-insns@r{[}=@var{file}@r{]} @gol
587 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
589 -fdump-lang-@var{switch} @gol
590 -fdump-lang-@var{switch}-@var{options} @gol
591 -fdump-lang-@var{switch}-@var{options}=@var{filename} @gol
593 -fdump-rtl-@var{pass} -fdump-rtl-@var{pass}=@var{filename} @gol
594 -fdump-statistics @gol
596 -fdump-tree-@var{switch} @gol
597 -fdump-tree-@var{switch}-@var{options} @gol
598 -fdump-tree-@var{switch}-@var{options}=@var{filename} @gol
599 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
600 -fenable-@var{kind}-@var{pass} @gol
601 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
602 -fira-verbose=@var{n} @gol
603 -flto-report -flto-report-wpa -fmem-report-wpa @gol
604 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report @gol
605 -fopt-info -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
606 -fprofile-report @gol
607 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
608 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
609 -fstats -fstack-usage -ftime-report -ftime-report-details @gol
610 -fvar-tracking-assignments-toggle -gtoggle @gol
611 -print-file-name=@var{library} -print-libgcc-file-name @gol
612 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
613 -print-prog-name=@var{program} -print-search-dirs -Q @gol
614 -print-sysroot -print-sysroot-headers-suffix @gol
615 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
617 @item Machine-Dependent Options
618 @xref{Submodel Options,,Machine-Dependent Options}.
619 @c This list is ordered alphanumerically by subsection name.
620 @c Try and put the significant identifier (CPU or system) first,
621 @c so users have a clue at guessing where the ones they want will be.
623 @emph{AArch64 Options}
624 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
625 -mgeneral-regs-only @gol
626 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
627 -mstrict-align -mno-strict-align @gol
628 -momit-leaf-frame-pointer @gol
629 -mtls-dialect=desc -mtls-dialect=traditional @gol
630 -mtls-size=@var{size} @gol
631 -mfix-cortex-a53-835769 -mfix-cortex-a53-843419 @gol
632 -mlow-precision-recip-sqrt -mlow-precision-sqrt -mlow-precision-div @gol
633 -mpc-relative-literal-loads @gol
634 -msign-return-address=@var{scope} @gol
635 -mbranch-protection=@var{none}|@var{standard}|@var{pac-ret}[+@var{leaf}]|@var{bti} @gol
636 -march=@var{name} -mcpu=@var{name} -mtune=@var{name} @gol
637 -moverride=@var{string} -mverbose-cost-dump @gol
638 -mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{sysreg} @gol
639 -mstack-protector-guard-offset=@var{offset} -mtrack-speculation }
641 @emph{Adapteva Epiphany Options}
642 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
643 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
644 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
645 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
646 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
647 -msplit-vecmove-early -m1reg-@var{reg}}
649 @emph{AMD GCN Options}
650 @gccoptlist{-march=@var{gpu} -mtune=@var{gpu} -mstack-size=@var{bytes}}
653 @gccoptlist{-mbarrel-shifter -mjli-always @gol
654 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
655 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
656 -mea -mno-mpy -mmul32x16 -mmul64 -matomic @gol
657 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
658 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
659 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
660 -mlong-calls -mmedium-calls -msdata -mirq-ctrl-saved @gol
661 -mrgf-banked-regs -mlpc-width=@var{width} -G @var{num} @gol
662 -mvolatile-cache -mtp-regno=@var{regno} @gol
663 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
664 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
665 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
666 -mlra-priority-compact mlra-priority-noncompact -mmillicode @gol
667 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
668 -mtune=@var{cpu} -mmultcost=@var{num} -mcode-density-frame @gol
669 -munalign-prob-threshold=@var{probability} -mmpy-option=@var{multo} @gol
670 -mdiv-rem -mcode-density -mll64 -mfpu=@var{fpu} -mrf16 -mbranch-index}
673 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
674 -mabi=@var{name} @gol
675 -mapcs-stack-check -mno-apcs-stack-check @gol
676 -mapcs-reentrant -mno-apcs-reentrant @gol
677 -msched-prolog -mno-sched-prolog @gol
678 -mlittle-endian -mbig-endian @gol
680 -mfloat-abi=@var{name} @gol
681 -mfp16-format=@var{name}
682 -mthumb-interwork -mno-thumb-interwork @gol
683 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
684 -mtune=@var{name} -mprint-tune-info @gol
685 -mstructure-size-boundary=@var{n} @gol
686 -mabort-on-noreturn @gol
687 -mlong-calls -mno-long-calls @gol
688 -msingle-pic-base -mno-single-pic-base @gol
689 -mpic-register=@var{reg} @gol
690 -mnop-fun-dllimport @gol
691 -mpoke-function-name @gol
692 -mthumb -marm -mflip-thumb @gol
693 -mtpcs-frame -mtpcs-leaf-frame @gol
694 -mcaller-super-interworking -mcallee-super-interworking @gol
695 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
696 -mword-relocations @gol
697 -mfix-cortex-m3-ldrd @gol
698 -munaligned-access @gol
699 -mneon-for-64bits @gol
700 -mslow-flash-data @gol
701 -masm-syntax-unified @gol
703 -mverbose-cost-dump @gol
708 @gccoptlist{-mmcu=@var{mcu} -mabsdata -maccumulate-args @gol
709 -mbranch-cost=@var{cost} @gol
710 -mcall-prologues -mgas-isr-prologues -mint8 @gol
711 -mn_flash=@var{size} -mno-interrupts @gol
712 -mmain-is-OS_task -mrelax -mrmw -mstrict-X -mtiny-stack @gol
713 -mfract-convert-truncate @gol
714 -mshort-calls -nodevicelib @gol
715 -Waddr-space-convert -Wmisspelled-isr}
717 @emph{Blackfin Options}
718 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
719 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
720 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
721 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
722 -mno-id-shared-library -mshared-library-id=@var{n} @gol
723 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
724 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
725 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
729 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
730 -msim -msdata=@var{sdata-type}}
733 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
734 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
735 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
736 -mstack-align -mdata-align -mconst-align @gol
737 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
738 -melf -maout -melinux -mlinux -sim -sim2 @gol
739 -mmul-bug-workaround -mno-mul-bug-workaround}
742 @gccoptlist{-mmac @gol
743 -mcr16cplus -mcr16c @gol
744 -msim -mint32 -mbit-ops
745 -mdata-model=@var{model}}
748 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} @gol
749 -mbig-endian -EB -mlittle-endian -EL @gol
750 -mhard-float -msoft-float -mfpu=@var{fpu} -mdouble-float -mfdivdu @gol
751 -melrw -mistack -mmp -mcp -mcache -msecurity -mtrust @gol
752 -mdsp -medsp -mvdsp @gol
753 -mdiv -msmart -mhigh-registers -manchor @gol
754 -mpushpop -mmultiple-stld -mconstpool -mstack-size -mccrt @gol
755 -mbranch-cost=@var{n} -mcse-cc -msched-prolog}
757 @emph{Darwin Options}
758 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
759 -arch_only -bind_at_load -bundle -bundle_loader @gol
760 -client_name -compatibility_version -current_version @gol
762 -dependency-file -dylib_file -dylinker_install_name @gol
763 -dynamic -dynamiclib -exported_symbols_list @gol
764 -filelist -flat_namespace -force_cpusubtype_ALL @gol
765 -force_flat_namespace -headerpad_max_install_names @gol
767 -image_base -init -install_name -keep_private_externs @gol
768 -multi_module -multiply_defined -multiply_defined_unused @gol
769 -noall_load -no_dead_strip_inits_and_terms @gol
770 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
771 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
772 -private_bundle -read_only_relocs -sectalign @gol
773 -sectobjectsymbols -whyload -seg1addr @gol
774 -sectcreate -sectobjectsymbols -sectorder @gol
775 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
776 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
777 -segprot -segs_read_only_addr -segs_read_write_addr @gol
778 -single_module -static -sub_library -sub_umbrella @gol
779 -twolevel_namespace -umbrella -undefined @gol
780 -unexported_symbols_list -weak_reference_mismatches @gol
781 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
782 -mkernel -mone-byte-bool}
784 @emph{DEC Alpha Options}
785 @gccoptlist{-mno-fp-regs -msoft-float @gol
786 -mieee -mieee-with-inexact -mieee-conformant @gol
787 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
788 -mtrap-precision=@var{mode} -mbuild-constants @gol
789 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
790 -mbwx -mmax -mfix -mcix @gol
791 -mfloat-vax -mfloat-ieee @gol
792 -mexplicit-relocs -msmall-data -mlarge-data @gol
793 -msmall-text -mlarge-text @gol
794 -mmemory-latency=@var{time}}
797 @gccoptlist{-msmall-model -mno-lsim}
800 @gccoptlist{-msim -mlra -mnodiv -mft32b -mcompress -mnopm}
803 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
804 -mhard-float -msoft-float @gol
805 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
806 -mdouble -mno-double @gol
807 -mmedia -mno-media -mmuladd -mno-muladd @gol
808 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
809 -mlinked-fp -mlong-calls -malign-labels @gol
810 -mlibrary-pic -macc-4 -macc-8 @gol
811 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
812 -moptimize-membar -mno-optimize-membar @gol
813 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
814 -mvliw-branch -mno-vliw-branch @gol
815 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
816 -mno-nested-cond-exec -mtomcat-stats @gol
820 @emph{GNU/Linux Options}
821 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
822 -tno-android-cc -tno-android-ld}
824 @emph{H8/300 Options}
825 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
828 @gccoptlist{-march=@var{architecture-type} @gol
829 -mcaller-copies -mdisable-fpregs -mdisable-indexing @gol
830 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
831 -mfixed-range=@var{register-range} @gol
832 -mjump-in-delay -mlinker-opt -mlong-calls @gol
833 -mlong-load-store -mno-disable-fpregs @gol
834 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
835 -mno-jump-in-delay -mno-long-load-store @gol
836 -mno-portable-runtime -mno-soft-float @gol
837 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
838 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
839 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
840 -munix=@var{unix-std} -nolibdld -static -threads}
843 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
844 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
845 -mconstant-gp -mauto-pic -mfused-madd @gol
846 -minline-float-divide-min-latency @gol
847 -minline-float-divide-max-throughput @gol
848 -mno-inline-float-divide @gol
849 -minline-int-divide-min-latency @gol
850 -minline-int-divide-max-throughput @gol
851 -mno-inline-int-divide @gol
852 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
853 -mno-inline-sqrt @gol
854 -mdwarf2-asm -mearly-stop-bits @gol
855 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
856 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
857 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
858 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
859 -msched-spec-ldc -msched-spec-control-ldc @gol
860 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
861 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
862 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
863 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
866 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
867 -msign-extend-enabled -muser-enabled}
869 @emph{M32R/D Options}
870 @gccoptlist{-m32r2 -m32rx -m32r @gol
872 -malign-loops -mno-align-loops @gol
873 -missue-rate=@var{number} @gol
874 -mbranch-cost=@var{number} @gol
875 -mmodel=@var{code-size-model-type} @gol
876 -msdata=@var{sdata-type} @gol
877 -mno-flush-func -mflush-func=@var{name} @gol
878 -mno-flush-trap -mflush-trap=@var{number} @gol
882 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
884 @emph{M680x0 Options}
885 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
886 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
887 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
888 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
889 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
890 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
891 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
892 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
893 -mxgot -mno-xgot -mlong-jump-table-offsets}
896 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
897 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
898 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
899 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
900 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
903 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
904 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
905 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
906 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
909 @emph{MicroBlaze Options}
910 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
911 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
912 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
913 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
914 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model} @gol
915 -mpic-data-is-text-relative}
918 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
919 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
920 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
921 -mips16 -mno-mips16 -mflip-mips16 @gol
922 -minterlink-compressed -mno-interlink-compressed @gol
923 -minterlink-mips16 -mno-interlink-mips16 @gol
924 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
925 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
926 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
927 -mno-float -msingle-float -mdouble-float @gol
928 -modd-spreg -mno-odd-spreg @gol
929 -mabs=@var{mode} -mnan=@var{encoding} @gol
930 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
933 -mvirt -mno-virt @gol
936 -mginv -mno-ginv @gol
937 -mmicromips -mno-micromips @gol
939 -mloongson-mmi -mno-loongson-mmi @gol
940 -mloongson-ext -mno-loongson-ext @gol
941 -mloongson-ext2 -mno-loongson-ext2 @gol
942 -mfpu=@var{fpu-type} @gol
943 -msmartmips -mno-smartmips @gol
944 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
945 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
946 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
947 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
948 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
949 -membedded-data -mno-embedded-data @gol
950 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
951 -mcode-readable=@var{setting} @gol
952 -msplit-addresses -mno-split-addresses @gol
953 -mexplicit-relocs -mno-explicit-relocs @gol
954 -mcheck-zero-division -mno-check-zero-division @gol
955 -mdivide-traps -mdivide-breaks @gol
956 -mload-store-pairs -mno-load-store-pairs @gol
957 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
958 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
959 -mfix-24k -mno-fix-24k @gol
960 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
961 -mfix-r5900 -mno-fix-r5900 @gol
962 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
963 -mfix-vr4120 -mno-fix-vr4120 @gol
964 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
965 -mflush-func=@var{func} -mno-flush-func @gol
966 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
967 -mcompact-branches=@var{policy} @gol
968 -mfp-exceptions -mno-fp-exceptions @gol
969 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
970 -mlxc1-sxc1 -mno-lxc1-sxc1 -mmadd4 -mno-madd4 @gol
971 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address @gol
972 -mframe-header-opt -mno-frame-header-opt}
975 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
976 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
977 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
978 -mno-base-addresses -msingle-exit -mno-single-exit}
980 @emph{MN10300 Options}
981 @gccoptlist{-mmult-bug -mno-mult-bug @gol
982 -mno-am33 -mam33 -mam33-2 -mam34 @gol
983 -mtune=@var{cpu-type} @gol
984 -mreturn-pointer-on-d0 @gol
985 -mno-crt0 -mrelax -mliw -msetlb}
988 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
990 @emph{MSP430 Options}
991 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
993 -mcode-region= -mdata-region= @gol
994 -msilicon-errata= -msilicon-errata-warn= @gol
998 @gccoptlist{-mbig-endian -mlittle-endian @gol
999 -mreduced-regs -mfull-regs @gol
1000 -mcmov -mno-cmov @gol
1001 -mext-perf -mno-ext-perf @gol
1002 -mext-perf2 -mno-ext-perf2 @gol
1003 -mext-string -mno-ext-string @gol
1004 -mv3push -mno-v3push @gol
1005 -m16bit -mno-16bit @gol
1006 -misr-vector-size=@var{num} @gol
1007 -mcache-block-size=@var{num} @gol
1008 -march=@var{arch} @gol
1009 -mcmodel=@var{code-model} @gol
1010 -mctor-dtor -mrelax}
1012 @emph{Nios II Options}
1013 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
1014 -mgprel-sec=@var{regexp} -mr0rel-sec=@var{regexp} @gol
1016 -mno-bypass-cache -mbypass-cache @gol
1017 -mno-cache-volatile -mcache-volatile @gol
1018 -mno-fast-sw-div -mfast-sw-div @gol
1019 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
1020 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
1021 -mcustom-fpu-cfg=@var{name} @gol
1022 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
1023 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
1025 @emph{Nvidia PTX Options}
1026 @gccoptlist{-m32 -m64 -mmainkernel -moptimize}
1028 @emph{OpenRISC Options}
1029 @gccoptlist{-mboard=@var{name} -mnewlib -mhard-mul -mhard-div @gol
1030 -msoft-mul -msoft-div @gol
1031 -mcmov -mror -msext -msfimm -mshftimm}
1033 @emph{PDP-11 Options}
1034 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
1035 -mint32 -mno-int16 -mint16 -mno-int32 @gol
1036 -msplit -munix-asm -mdec-asm -mgnu-asm -mlra}
1038 @emph{picoChip Options}
1039 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
1040 -msymbol-as-address -mno-inefficient-warnings}
1042 @emph{PowerPC Options}
1043 See RS/6000 and PowerPC Options.
1045 @emph{RISC-V Options}
1046 @gccoptlist{-mbranch-cost=@var{N-instruction} @gol
1048 -mabi=@var{ABI-string} @gol
1049 -mfdiv -mno-fdiv @gol
1051 -march=@var{ISA-string} @gol
1052 -mtune=@var{processor-string} @gol
1053 -mpreferred-stack-boundary=@var{num} @gol
1054 -msmall-data-limit=@var{N-bytes} @gol
1055 -msave-restore -mno-save-restore @gol
1056 -mstrict-align -mno-strict-align @gol
1057 -mcmodel=medlow -mcmodel=medany @gol
1058 -mexplicit-relocs -mno-explicit-relocs @gol
1059 -mrelax -mno-relax @gol
1060 -mriscv-attribute -mmo-riscv-attribute}
1063 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
1064 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
1065 -m64bit-doubles -m32bit-doubles -msave-mduc-in-interrupts}
1067 @emph{RS/6000 and PowerPC Options}
1068 @gccoptlist{-mcpu=@var{cpu-type} @gol
1069 -mtune=@var{cpu-type} @gol
1070 -mcmodel=@var{code-model} @gol
1072 -maltivec -mno-altivec @gol
1073 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
1074 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
1075 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
1076 -mfprnd -mno-fprnd @gol
1077 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
1078 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
1079 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
1080 -malign-power -malign-natural @gol
1081 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
1082 -mupdate -mno-update @gol
1083 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
1084 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
1085 -mstrict-align -mno-strict-align -mrelocatable @gol
1086 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
1087 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
1088 -mdynamic-no-pic -mswdiv -msingle-pic-base @gol
1089 -mprioritize-restricted-insns=@var{priority} @gol
1090 -msched-costly-dep=@var{dependence_type} @gol
1091 -minsert-sched-nops=@var{scheme} @gol
1092 -mcall-aixdesc -mcall-eabi -mcall-freebsd @gol
1093 -mcall-linux -mcall-netbsd -mcall-openbsd @gol
1094 -mcall-sysv -mcall-sysv-eabi -mcall-sysv-noeabi @gol
1095 -mtraceback=@var{traceback_type} @gol
1096 -maix-struct-return -msvr4-struct-return @gol
1097 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
1098 -mlongcall -mno-longcall -mpltseq -mno-pltseq @gol
1099 -mblock-move-inline-limit=@var{num} @gol
1100 -mblock-compare-inline-limit=@var{num} @gol
1101 -mblock-compare-inline-loop-limit=@var{num} @gol
1102 -mstring-compare-inline-limit=@var{num} @gol
1103 -misel -mno-isel @gol
1104 -mvrsave -mno-vrsave @gol
1105 -mmulhw -mno-mulhw @gol
1106 -mdlmzb -mno-dlmzb @gol
1107 -mprototype -mno-prototype @gol
1108 -msim -mmvme -mads -myellowknife -memb -msdata @gol
1109 -msdata=@var{opt} -mreadonly-in-sdata -mvxworks -G @var{num} @gol
1110 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
1111 -mno-recip-precision @gol
1112 -mveclibabi=@var{type} -mfriz -mno-friz @gol
1113 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
1114 -msave-toc-indirect -mno-save-toc-indirect @gol
1115 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
1116 -mcrypto -mno-crypto -mhtm -mno-htm @gol
1117 -mquad-memory -mno-quad-memory @gol
1118 -mquad-memory-atomic -mno-quad-memory-atomic @gol
1119 -mcompat-align-parm -mno-compat-align-parm @gol
1120 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol
1121 -mgnu-attribute -mno-gnu-attribute @gol
1122 -mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{reg} @gol
1123 -mstack-protector-guard-offset=@var{offset}}
1126 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
1128 -mbig-endian-data -mlittle-endian-data @gol
1131 -mas100-syntax -mno-as100-syntax@gol
1133 -mmax-constant-size=@gol
1136 -mallow-string-insns -mno-allow-string-insns@gol
1138 -mno-warn-multiple-fast-interrupts@gol
1139 -msave-acc-in-interrupts}
1141 @emph{S/390 and zSeries Options}
1142 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1143 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
1144 -mlong-double-64 -mlong-double-128 @gol
1145 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
1146 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
1147 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
1148 -mhtm -mvx -mzvector @gol
1149 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
1150 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
1151 -mhotpatch=@var{halfwords},@var{halfwords}}
1153 @emph{Score Options}
1154 @gccoptlist{-meb -mel @gol
1158 -mscore5 -mscore5u -mscore7 -mscore7d}
1161 @gccoptlist{-m1 -m2 -m2e @gol
1162 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
1164 -m4-nofpu -m4-single-only -m4-single -m4 @gol
1165 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
1166 -mb -ml -mdalign -mrelax @gol
1167 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
1168 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
1169 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1170 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1171 -maccumulate-outgoing-args @gol
1172 -matomic-model=@var{atomic-model} @gol
1173 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1174 -mcbranch-force-delay-slot @gol
1175 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1176 -mpretend-cmove -mtas}
1178 @emph{Solaris 2 Options}
1179 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1182 @emph{SPARC Options}
1183 @gccoptlist{-mcpu=@var{cpu-type} @gol
1184 -mtune=@var{cpu-type} @gol
1185 -mcmodel=@var{code-model} @gol
1186 -mmemory-model=@var{mem-model} @gol
1187 -m32 -m64 -mapp-regs -mno-app-regs @gol
1188 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1189 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1190 -mhard-quad-float -msoft-quad-float @gol
1191 -mstack-bias -mno-stack-bias @gol
1192 -mstd-struct-return -mno-std-struct-return @gol
1193 -munaligned-doubles -mno-unaligned-doubles @gol
1194 -muser-mode -mno-user-mode @gol
1195 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1196 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1197 -mvis4 -mno-vis4 -mvis4b -mno-vis4b @gol
1198 -mcbcond -mno-cbcond -mfmaf -mno-fmaf -mfsmuld -mno-fsmuld @gol
1199 -mpopc -mno-popc -msubxc -mno-subxc @gol
1200 -mfix-at697f -mfix-ut699 -mfix-ut700 -mfix-gr712rc @gol
1204 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1205 -msafe-dma -munsafe-dma @gol
1207 -msmall-mem -mlarge-mem -mstdmain @gol
1208 -mfixed-range=@var{register-range} @gol
1210 -maddress-space-conversion -mno-address-space-conversion @gol
1211 -mcache-size=@var{cache-size} @gol
1212 -matomic-updates -mno-atomic-updates}
1214 @emph{System V Options}
1215 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1217 @emph{TILE-Gx Options}
1218 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1219 -mcmodel=@var{code-model}}
1221 @emph{TILEPro Options}
1222 @gccoptlist{-mcpu=@var{cpu} -m32}
1225 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1226 -mprolog-function -mno-prolog-function -mspace @gol
1227 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1228 -mapp-regs -mno-app-regs @gol
1229 -mdisable-callt -mno-disable-callt @gol
1230 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1231 -mv850e -mv850 -mv850e3v5 @gol
1242 @gccoptlist{-mg -mgnu -munix}
1244 @emph{Visium Options}
1245 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1246 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1249 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1250 -mpointer-size=@var{size}}
1252 @emph{VxWorks Options}
1253 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1254 -Xbind-lazy -Xbind-now}
1257 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1258 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1259 -mfpmath=@var{unit} @gol
1260 -masm=@var{dialect} -mno-fancy-math-387 @gol
1261 -mno-fp-ret-in-387 -m80387 -mhard-float -msoft-float @gol
1262 -mno-wide-multiply -mrtd -malign-double @gol
1263 -mpreferred-stack-boundary=@var{num} @gol
1264 -mincoming-stack-boundary=@var{num} @gol
1265 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1266 -mrecip -mrecip=@var{opt} @gol
1267 -mvzeroupper -mprefer-avx128 -mprefer-vector-width=@var{opt} @gol
1268 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1269 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1270 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1271 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma -mpconfig -mwbnoinvd @gol
1272 -mptwrite -mprefetchwt1 -mclflushopt -mclwb -mxsavec -mxsaves @gol
1273 -msse4a -m3dnow -m3dnowa -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop @gol
1274 -madx -mlzcnt -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mhle -mlwp @gol
1275 -mmwaitx -mclzero -mpku -mthreads -mgfni -mvaes -mwaitpkg @gol
1276 -mshstk -mmanual-endbr -mforce-indirect-call -mavx512vbmi2 @gol
1277 -mvpclmulqdq -mavx512bitalg -mmovdiri -mmovdir64b -mavx512vpopcntdq @gol
1278 -mavx5124fmaps -mavx512vnni -mavx5124vnniw -mprfchw -mrdpid @gol
1280 -mcldemote -mms-bitfields -mno-align-stringops -minline-all-stringops @gol
1281 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1282 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1283 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1284 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1285 -mregparm=@var{num} -msseregparm @gol
1286 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1287 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1288 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1289 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1290 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1291 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1292 -minstrument-return=@var{type} -mfentry-name=@var{name} -mfentry-section=@var{name} @gol
1293 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1294 -malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol
1295 -mstack-protector-guard-reg=@var{reg} @gol
1296 -mstack-protector-guard-offset=@var{offset} @gol
1297 -mstack-protector-guard-symbol=@var{symbol} @gol
1298 -mgeneral-regs-only -mcall-ms2sysv-xlogues @gol
1299 -mindirect-branch=@var{choice} -mfunction-return=@var{choice} @gol
1300 -mindirect-branch-register}
1302 @emph{x86 Windows Options}
1303 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1304 -mnop-fun-dllimport -mthread @gol
1305 -municode -mwin32 -mwindows -fno-set-stack-executable}
1307 @emph{Xstormy16 Options}
1310 @emph{Xtensa Options}
1311 @gccoptlist{-mconst16 -mno-const16 @gol
1312 -mfused-madd -mno-fused-madd @gol
1314 -mserialize-volatile -mno-serialize-volatile @gol
1315 -mtext-section-literals -mno-text-section-literals @gol
1316 -mauto-litpools -mno-auto-litpools @gol
1317 -mtarget-align -mno-target-align @gol
1318 -mlongcalls -mno-longcalls}
1320 @emph{zSeries Options}
1321 See S/390 and zSeries Options.
1325 @node Overall Options
1326 @section Options Controlling the Kind of Output
1328 Compilation can involve up to four stages: preprocessing, compilation
1329 proper, assembly and linking, always in that order. GCC is capable of
1330 preprocessing and compiling several files either into several
1331 assembler input files, or into one assembler input file; then each
1332 assembler input file produces an object file, and linking combines all
1333 the object files (those newly compiled, and those specified as input)
1334 into an executable file.
1336 @cindex file name suffix
1337 For any given input file, the file name suffix determines what kind of
1338 compilation is done:
1342 C source code that must be preprocessed.
1345 C source code that should not be preprocessed.
1348 C++ source code that should not be preprocessed.
1351 Objective-C source code. Note that you must link with the @file{libobjc}
1352 library to make an Objective-C program work.
1355 Objective-C source code that should not be preprocessed.
1359 Objective-C++ source code. Note that you must link with the @file{libobjc}
1360 library to make an Objective-C++ program work. Note that @samp{.M} refers
1361 to a literal capital M@.
1363 @item @var{file}.mii
1364 Objective-C++ source code that should not be preprocessed.
1367 C, C++, Objective-C or Objective-C++ header file to be turned into a
1368 precompiled header (default), or C, C++ header file to be turned into an
1369 Ada spec (via the @option{-fdump-ada-spec} switch).
1372 @itemx @var{file}.cp
1373 @itemx @var{file}.cxx
1374 @itemx @var{file}.cpp
1375 @itemx @var{file}.CPP
1376 @itemx @var{file}.c++
1378 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1379 the last two letters must both be literally @samp{x}. Likewise,
1380 @samp{.C} refers to a literal capital C@.
1384 Objective-C++ source code that must be preprocessed.
1386 @item @var{file}.mii
1387 Objective-C++ source code that should not be preprocessed.
1391 @itemx @var{file}.hp
1392 @itemx @var{file}.hxx
1393 @itemx @var{file}.hpp
1394 @itemx @var{file}.HPP
1395 @itemx @var{file}.h++
1396 @itemx @var{file}.tcc
1397 C++ header file to be turned into a precompiled header or Ada spec.
1400 @itemx @var{file}.for
1401 @itemx @var{file}.ftn
1402 Fixed form Fortran source code that should not be preprocessed.
1405 @itemx @var{file}.FOR
1406 @itemx @var{file}.fpp
1407 @itemx @var{file}.FPP
1408 @itemx @var{file}.FTN
1409 Fixed form Fortran source code that must be preprocessed (with the traditional
1412 @item @var{file}.f90
1413 @itemx @var{file}.f95
1414 @itemx @var{file}.f03
1415 @itemx @var{file}.f08
1416 Free form Fortran source code that should not be preprocessed.
1418 @item @var{file}.F90
1419 @itemx @var{file}.F95
1420 @itemx @var{file}.F03
1421 @itemx @var{file}.F08
1422 Free form Fortran source code that must be preprocessed (with the
1423 traditional preprocessor).
1428 @item @var{file}.brig
1429 BRIG files (binary representation of HSAIL).
1438 D documentation code (Ddoc).
1440 @item @var{file}.ads
1441 Ada source code file that contains a library unit declaration (a
1442 declaration of a package, subprogram, or generic, or a generic
1443 instantiation), or a library unit renaming declaration (a package,
1444 generic, or subprogram renaming declaration). Such files are also
1447 @item @var{file}.adb
1448 Ada source code file containing a library unit body (a subprogram or
1449 package body). Such files are also called @dfn{bodies}.
1451 @c GCC also knows about some suffixes for languages not yet included:
1459 @itemx @var{file}.sx
1460 Assembler code that must be preprocessed.
1463 An object file to be fed straight into linking.
1464 Any file name with no recognized suffix is treated this way.
1468 You can specify the input language explicitly with the @option{-x} option:
1471 @item -x @var{language}
1472 Specify explicitly the @var{language} for the following input files
1473 (rather than letting the compiler choose a default based on the file
1474 name suffix). This option applies to all following input files until
1475 the next @option{-x} option. Possible values for @var{language} are:
1477 c c-header cpp-output
1478 c++ c++-header c++-cpp-output
1479 objective-c objective-c-header objective-c-cpp-output
1480 objective-c++ objective-c++-header objective-c++-cpp-output
1481 assembler assembler-with-cpp
1484 f77 f77-cpp-input f95 f95-cpp-input
1490 Turn off any specification of a language, so that subsequent files are
1491 handled according to their file name suffixes (as they are if @option{-x}
1492 has not been used at all).
1495 If you only want some of the stages of compilation, you can use
1496 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1497 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1498 @command{gcc} is to stop. Note that some combinations (for example,
1499 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1504 Compile or assemble the source files, but do not link. The linking
1505 stage simply is not done. The ultimate output is in the form of an
1506 object file for each source file.
1508 By default, the object file name for a source file is made by replacing
1509 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1511 Unrecognized input files, not requiring compilation or assembly, are
1516 Stop after the stage of compilation proper; do not assemble. The output
1517 is in the form of an assembler code file for each non-assembler input
1520 By default, the assembler file name for a source file is made by
1521 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1523 Input files that don't require compilation are ignored.
1527 Stop after the preprocessing stage; do not run the compiler proper. The
1528 output is in the form of preprocessed source code, which is sent to the
1531 Input files that don't require preprocessing are ignored.
1533 @cindex output file option
1536 Place output in file @var{file}. This applies to whatever
1537 sort of output is being produced, whether it be an executable file,
1538 an object file, an assembler file or preprocessed C code.
1540 If @option{-o} is not specified, the default is to put an executable
1541 file in @file{a.out}, the object file for
1542 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1543 assembler file in @file{@var{source}.s}, a precompiled header file in
1544 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1549 Print (on standard error output) the commands executed to run the stages
1550 of compilation. Also print the version number of the compiler driver
1551 program and of the preprocessor and the compiler proper.
1555 Like @option{-v} except the commands are not executed and arguments
1556 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1557 This is useful for shell scripts to capture the driver-generated command lines.
1561 Print (on the standard output) a description of the command-line options
1562 understood by @command{gcc}. If the @option{-v} option is also specified
1563 then @option{--help} is also passed on to the various processes
1564 invoked by @command{gcc}, so that they can display the command-line options
1565 they accept. If the @option{-Wextra} option has also been specified
1566 (prior to the @option{--help} option), then command-line options that
1567 have no documentation associated with them are also displayed.
1570 @opindex target-help
1571 Print (on the standard output) a description of target-specific command-line
1572 options for each tool. For some targets extra target-specific
1573 information may also be printed.
1575 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1576 Print (on the standard output) a description of the command-line
1577 options understood by the compiler that fit into all specified classes
1578 and qualifiers. These are the supported classes:
1581 @item @samp{optimizers}
1582 Display all of the optimization options supported by the
1585 @item @samp{warnings}
1586 Display all of the options controlling warning messages
1587 produced by the compiler.
1590 Display target-specific options. Unlike the
1591 @option{--target-help} option however, target-specific options of the
1592 linker and assembler are not displayed. This is because those
1593 tools do not currently support the extended @option{--help=} syntax.
1596 Display the values recognized by the @option{--param}
1599 @item @var{language}
1600 Display the options supported for @var{language}, where
1601 @var{language} is the name of one of the languages supported in this
1605 Display the options that are common to all languages.
1608 These are the supported qualifiers:
1611 @item @samp{undocumented}
1612 Display only those options that are undocumented.
1615 Display options taking an argument that appears after an equal
1616 sign in the same continuous piece of text, such as:
1617 @samp{--help=target}.
1619 @item @samp{separate}
1620 Display options taking an argument that appears as a separate word
1621 following the original option, such as: @samp{-o output-file}.
1624 Thus for example to display all the undocumented target-specific
1625 switches supported by the compiler, use:
1628 --help=target,undocumented
1631 The sense of a qualifier can be inverted by prefixing it with the
1632 @samp{^} character, so for example to display all binary warning
1633 options (i.e., ones that are either on or off and that do not take an
1634 argument) that have a description, use:
1637 --help=warnings,^joined,^undocumented
1640 The argument to @option{--help=} should not consist solely of inverted
1643 Combining several classes is possible, although this usually
1644 restricts the output so much that there is nothing to display. One
1645 case where it does work, however, is when one of the classes is
1646 @var{target}. For example, to display all the target-specific
1647 optimization options, use:
1650 --help=target,optimizers
1653 The @option{--help=} option can be repeated on the command line. Each
1654 successive use displays its requested class of options, skipping
1655 those that have already been displayed. If @option{--help} is also
1656 specified anywhere on the command line then this takes precedence
1657 over any @option{--help=} option.
1659 If the @option{-Q} option appears on the command line before the
1660 @option{--help=} option, then the descriptive text displayed by
1661 @option{--help=} is changed. Instead of describing the displayed
1662 options, an indication is given as to whether the option is enabled,
1663 disabled or set to a specific value (assuming that the compiler
1664 knows this at the point where the @option{--help=} option is used).
1666 Here is a truncated example from the ARM port of @command{gcc}:
1669 % gcc -Q -mabi=2 --help=target -c
1670 The following options are target specific:
1672 -mabort-on-noreturn [disabled]
1676 The output is sensitive to the effects of previous command-line
1677 options, so for example it is possible to find out which optimizations
1678 are enabled at @option{-O2} by using:
1681 -Q -O2 --help=optimizers
1684 Alternatively you can discover which binary optimizations are enabled
1685 by @option{-O3} by using:
1688 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1689 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1690 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1695 Display the version number and copyrights of the invoked GCC@.
1697 @item -pass-exit-codes
1698 @opindex pass-exit-codes
1699 Normally the @command{gcc} program exits with the code of 1 if any
1700 phase of the compiler returns a non-success return code. If you specify
1701 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1702 the numerically highest error produced by any phase returning an error
1703 indication. The C, C++, and Fortran front ends return 4 if an internal
1704 compiler error is encountered.
1708 Use pipes rather than temporary files for communication between the
1709 various stages of compilation. This fails to work on some systems where
1710 the assembler is unable to read from a pipe; but the GNU assembler has
1713 @item -specs=@var{file}
1715 Process @var{file} after the compiler reads in the standard @file{specs}
1716 file, in order to override the defaults which the @command{gcc} driver
1717 program uses when determining what switches to pass to @command{cc1},
1718 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
1719 @option{-specs=@var{file}} can be specified on the command line, and they
1720 are processed in order, from left to right. @xref{Spec Files}, for
1721 information about the format of the @var{file}.
1725 Invoke all subcommands under a wrapper program. The name of the
1726 wrapper program and its parameters are passed as a comma separated
1730 gcc -c t.c -wrapper gdb,--args
1734 This invokes all subprograms of @command{gcc} under
1735 @samp{gdb --args}, thus the invocation of @command{cc1} is
1736 @samp{gdb --args cc1 @dots{}}.
1738 @item -ffile-prefix-map=@var{old}=@var{new}
1739 @opindex ffile-prefix-map
1740 When compiling files residing in directory @file{@var{old}}, record
1741 any references to them in the result of the compilation as if the
1742 files resided in directory @file{@var{new}} instead. Specifying this
1743 option is equivalent to specifying all the individual
1744 @option{-f*-prefix-map} options. This can be used to make reproducible
1745 builds that are location independent. See also
1746 @option{-fmacro-prefix-map} and @option{-fdebug-prefix-map}.
1748 @item -fplugin=@var{name}.so
1750 Load the plugin code in file @var{name}.so, assumed to be a
1751 shared object to be dlopen'd by the compiler. The base name of
1752 the shared object file is used to identify the plugin for the
1753 purposes of argument parsing (See
1754 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1755 Each plugin should define the callback functions specified in the
1758 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1759 @opindex fplugin-arg
1760 Define an argument called @var{key} with a value of @var{value}
1761 for the plugin called @var{name}.
1763 @item -fdump-ada-spec@r{[}-slim@r{]}
1764 @opindex fdump-ada-spec
1765 For C and C++ source and include files, generate corresponding Ada specs.
1766 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1767 GNAT User's Guide}, which provides detailed documentation on this feature.
1769 @item -fada-spec-parent=@var{unit}
1770 @opindex fada-spec-parent
1771 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1772 Ada specs as child units of parent @var{unit}.
1774 @item -fdump-go-spec=@var{file}
1775 @opindex fdump-go-spec
1776 For input files in any language, generate corresponding Go
1777 declarations in @var{file}. This generates Go @code{const},
1778 @code{type}, @code{var}, and @code{func} declarations which may be a
1779 useful way to start writing a Go interface to code written in some
1782 @include @value{srcdir}/../libiberty/at-file.texi
1786 @section Compiling C++ Programs
1788 @cindex suffixes for C++ source
1789 @cindex C++ source file suffixes
1790 C++ source files conventionally use one of the suffixes @samp{.C},
1791 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1792 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1793 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1794 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1795 files with these names and compiles them as C++ programs even if you
1796 call the compiler the same way as for compiling C programs (usually
1797 with the name @command{gcc}).
1801 However, the use of @command{gcc} does not add the C++ library.
1802 @command{g++} is a program that calls GCC and automatically specifies linking
1803 against the C++ library. It treats @samp{.c},
1804 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1805 files unless @option{-x} is used. This program is also useful when
1806 precompiling a C header file with a @samp{.h} extension for use in C++
1807 compilations. On many systems, @command{g++} is also installed with
1808 the name @command{c++}.
1810 @cindex invoking @command{g++}
1811 When you compile C++ programs, you may specify many of the same
1812 command-line options that you use for compiling programs in any
1813 language; or command-line options meaningful for C and related
1814 languages; or options that are meaningful only for C++ programs.
1815 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1816 explanations of options for languages related to C@.
1817 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1818 explanations of options that are meaningful only for C++ programs.
1820 @node C Dialect Options
1821 @section Options Controlling C Dialect
1822 @cindex dialect options
1823 @cindex language dialect options
1824 @cindex options, dialect
1826 The following options control the dialect of C (or languages derived
1827 from C, such as C++, Objective-C and Objective-C++) that the compiler
1831 @cindex ANSI support
1835 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1836 equivalent to @option{-std=c++98}.
1838 This turns off certain features of GCC that are incompatible with ISO
1839 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1840 such as the @code{asm} and @code{typeof} keywords, and
1841 predefined macros such as @code{unix} and @code{vax} that identify the
1842 type of system you are using. It also enables the undesirable and
1843 rarely used ISO trigraph feature. For the C compiler,
1844 it disables recognition of C++ style @samp{//} comments as well as
1845 the @code{inline} keyword.
1847 The alternate keywords @code{__asm__}, @code{__extension__},
1848 @code{__inline__} and @code{__typeof__} continue to work despite
1849 @option{-ansi}. You would not want to use them in an ISO C program, of
1850 course, but it is useful to put them in header files that might be included
1851 in compilations done with @option{-ansi}. Alternate predefined macros
1852 such as @code{__unix__} and @code{__vax__} are also available, with or
1853 without @option{-ansi}.
1855 The @option{-ansi} option does not cause non-ISO programs to be
1856 rejected gratuitously. For that, @option{-Wpedantic} is required in
1857 addition to @option{-ansi}. @xref{Warning Options}.
1859 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1860 option is used. Some header files may notice this macro and refrain
1861 from declaring certain functions or defining certain macros that the
1862 ISO standard doesn't call for; this is to avoid interfering with any
1863 programs that might use these names for other things.
1865 Functions that are normally built in but do not have semantics
1866 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1867 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1868 built-in functions provided by GCC}, for details of the functions
1873 Determine the language standard. @xref{Standards,,Language Standards
1874 Supported by GCC}, for details of these standard versions. This option
1875 is currently only supported when compiling C or C++.
1877 The compiler can accept several base standards, such as @samp{c90} or
1878 @samp{c++98}, and GNU dialects of those standards, such as
1879 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1880 compiler accepts all programs following that standard plus those
1881 using GNU extensions that do not contradict it. For example,
1882 @option{-std=c90} turns off certain features of GCC that are
1883 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1884 keywords, but not other GNU extensions that do not have a meaning in
1885 ISO C90, such as omitting the middle term of a @code{?:}
1886 expression. On the other hand, when a GNU dialect of a standard is
1887 specified, all features supported by the compiler are enabled, even when
1888 those features change the meaning of the base standard. As a result, some
1889 strict-conforming programs may be rejected. The particular standard
1890 is used by @option{-Wpedantic} to identify which features are GNU
1891 extensions given that version of the standard. For example
1892 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1893 comments, while @option{-std=gnu99 -Wpedantic} does not.
1895 A value for this option must be provided; possible values are
1901 Support all ISO C90 programs (certain GNU extensions that conflict
1902 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1904 @item iso9899:199409
1905 ISO C90 as modified in amendment 1.
1911 ISO C99. This standard is substantially completely supported, modulo
1912 bugs and floating-point issues
1913 (mainly but not entirely relating to optional C99 features from
1914 Annexes F and G). See
1915 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1916 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1921 ISO C11, the 2011 revision of the ISO C standard. This standard is
1922 substantially completely supported, modulo bugs, floating-point issues
1923 (mainly but not entirely relating to optional C11 features from
1924 Annexes F and G) and the optional Annexes K (Bounds-checking
1925 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1931 ISO C17, the 2017 revision of the ISO C standard
1932 (published in 2018). This standard is
1933 same as C11 except for corrections of defects (all of which are also
1934 applied with @option{-std=c11}) and a new value of
1935 @code{__STDC_VERSION__}, and so is supported to the same extent as C11.
1938 The next version of the ISO C standard, still under development. The
1939 support for this version is experimental and incomplete.
1943 GNU dialect of ISO C90 (including some C99 features).
1947 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1951 GNU dialect of ISO C11.
1952 The name @samp{gnu1x} is deprecated.
1956 GNU dialect of ISO C17. This is the default for C code.
1959 The next version of the ISO C standard, still under development, plus
1960 GNU extensions. The support for this version is experimental and
1965 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1966 additional defect reports. Same as @option{-ansi} for C++ code.
1970 GNU dialect of @option{-std=c++98}.
1974 The 2011 ISO C++ standard plus amendments.
1975 The name @samp{c++0x} is deprecated.
1979 GNU dialect of @option{-std=c++11}.
1980 The name @samp{gnu++0x} is deprecated.
1984 The 2014 ISO C++ standard plus amendments.
1985 The name @samp{c++1y} is deprecated.
1989 GNU dialect of @option{-std=c++14}.
1990 This is the default for C++ code.
1991 The name @samp{gnu++1y} is deprecated.
1995 The 2017 ISO C++ standard plus amendments.
1996 The name @samp{c++1z} is deprecated.
2000 GNU dialect of @option{-std=c++17}.
2001 The name @samp{gnu++1z} is deprecated.
2004 The next revision of the ISO C++ standard, tentatively planned for
2005 2020. Support is highly experimental, and will almost certainly
2006 change in incompatible ways in future releases.
2009 GNU dialect of @option{-std=c++2a}. Support is highly experimental,
2010 and will almost certainly change in incompatible ways in future
2014 @item -fgnu89-inline
2015 @opindex fgnu89-inline
2016 The option @option{-fgnu89-inline} tells GCC to use the traditional
2017 GNU semantics for @code{inline} functions when in C99 mode.
2018 @xref{Inline,,An Inline Function is As Fast As a Macro}.
2019 Using this option is roughly equivalent to adding the
2020 @code{gnu_inline} function attribute to all inline functions
2021 (@pxref{Function Attributes}).
2023 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
2024 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
2025 specifies the default behavior).
2026 This option is not supported in @option{-std=c90} or
2027 @option{-std=gnu90} mode.
2029 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
2030 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
2031 in effect for @code{inline} functions. @xref{Common Predefined
2032 Macros,,,cpp,The C Preprocessor}.
2034 @item -fpermitted-flt-eval-methods=@var{style}
2035 @opindex fpermitted-flt-eval-methods
2036 @opindex fpermitted-flt-eval-methods=c11
2037 @opindex fpermitted-flt-eval-methods=ts-18661-3
2038 ISO/IEC TS 18661-3 defines new permissible values for
2039 @code{FLT_EVAL_METHOD} that indicate that operations and constants with
2040 a semantic type that is an interchange or extended format should be
2041 evaluated to the precision and range of that type. These new values are
2042 a superset of those permitted under C99/C11, which does not specify the
2043 meaning of other positive values of @code{FLT_EVAL_METHOD}. As such, code
2044 conforming to C11 may not have been written expecting the possibility of
2047 @option{-fpermitted-flt-eval-methods} specifies whether the compiler
2048 should allow only the values of @code{FLT_EVAL_METHOD} specified in C99/C11,
2049 or the extended set of values specified in ISO/IEC TS 18661-3.
2051 @var{style} is either @code{c11} or @code{ts-18661-3} as appropriate.
2053 The default when in a standards compliant mode (@option{-std=c11} or similar)
2054 is @option{-fpermitted-flt-eval-methods=c11}. The default when in a GNU
2055 dialect (@option{-std=gnu11} or similar) is
2056 @option{-fpermitted-flt-eval-methods=ts-18661-3}.
2058 @item -aux-info @var{filename}
2060 Output to the given filename prototyped declarations for all functions
2061 declared and/or defined in a translation unit, including those in header
2062 files. This option is silently ignored in any language other than C@.
2064 Besides declarations, the file indicates, in comments, the origin of
2065 each declaration (source file and line), whether the declaration was
2066 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
2067 @samp{O} for old, respectively, in the first character after the line
2068 number and the colon), and whether it came from a declaration or a
2069 definition (@samp{C} or @samp{F}, respectively, in the following
2070 character). In the case of function definitions, a K&R-style list of
2071 arguments followed by their declarations is also provided, inside
2072 comments, after the declaration.
2074 @item -fallow-parameterless-variadic-functions
2075 @opindex fallow-parameterless-variadic-functions
2076 Accept variadic functions without named parameters.
2078 Although it is possible to define such a function, this is not very
2079 useful as it is not possible to read the arguments. This is only
2080 supported for C as this construct is allowed by C++.
2085 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
2086 keyword, so that code can use these words as identifiers. You can use
2087 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
2088 instead. @option{-ansi} implies @option{-fno-asm}.
2090 In C++, this switch only affects the @code{typeof} keyword, since
2091 @code{asm} and @code{inline} are standard keywords. You may want to
2092 use the @option{-fno-gnu-keywords} flag instead, which has the same
2093 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
2094 switch only affects the @code{asm} and @code{typeof} keywords, since
2095 @code{inline} is a standard keyword in ISO C99.
2098 @itemx -fno-builtin-@var{function}
2099 @opindex fno-builtin
2101 @cindex built-in functions
2102 Don't recognize built-in functions that do not begin with
2103 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
2104 functions provided by GCC}, for details of the functions affected,
2105 including those which are not built-in functions when @option{-ansi} or
2106 @option{-std} options for strict ISO C conformance are used because they
2107 do not have an ISO standard meaning.
2109 GCC normally generates special code to handle certain built-in functions
2110 more efficiently; for instance, calls to @code{alloca} may become single
2111 instructions which adjust the stack directly, and calls to @code{memcpy}
2112 may become inline copy loops. The resulting code is often both smaller
2113 and faster, but since the function calls no longer appear as such, you
2114 cannot set a breakpoint on those calls, nor can you change the behavior
2115 of the functions by linking with a different library. In addition,
2116 when a function is recognized as a built-in function, GCC may use
2117 information about that function to warn about problems with calls to
2118 that function, or to generate more efficient code, even if the
2119 resulting code still contains calls to that function. For example,
2120 warnings are given with @option{-Wformat} for bad calls to
2121 @code{printf} when @code{printf} is built in and @code{strlen} is
2122 known not to modify global memory.
2124 With the @option{-fno-builtin-@var{function}} option
2125 only the built-in function @var{function} is
2126 disabled. @var{function} must not begin with @samp{__builtin_}. If a
2127 function is named that is not built-in in this version of GCC, this
2128 option is ignored. There is no corresponding
2129 @option{-fbuiltin-@var{function}} option; if you wish to enable
2130 built-in functions selectively when using @option{-fno-builtin} or
2131 @option{-ffreestanding}, you may define macros such as:
2134 #define abs(n) __builtin_abs ((n))
2135 #define strcpy(d, s) __builtin_strcpy ((d), (s))
2141 Enable parsing of function definitions marked with @code{__GIMPLE}.
2142 This is an experimental feature that allows unit testing of GIMPLE
2147 @cindex hosted environment
2149 Assert that compilation targets a hosted environment. This implies
2150 @option{-fbuiltin}. A hosted environment is one in which the
2151 entire standard library is available, and in which @code{main} has a return
2152 type of @code{int}. Examples are nearly everything except a kernel.
2153 This is equivalent to @option{-fno-freestanding}.
2155 @item -ffreestanding
2156 @opindex ffreestanding
2157 @cindex hosted environment
2159 Assert that compilation targets a freestanding environment. This
2160 implies @option{-fno-builtin}. A freestanding environment
2161 is one in which the standard library may not exist, and program startup may
2162 not necessarily be at @code{main}. The most obvious example is an OS kernel.
2163 This is equivalent to @option{-fno-hosted}.
2165 @xref{Standards,,Language Standards Supported by GCC}, for details of
2166 freestanding and hosted environments.
2170 @cindex OpenACC accelerator programming
2171 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
2172 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
2173 compiler generates accelerated code according to the OpenACC Application
2174 Programming Interface v2.0 @w{@uref{https://www.openacc.org}}. This option
2175 implies @option{-pthread}, and thus is only supported on targets that
2176 have support for @option{-pthread}.
2178 @item -fopenacc-dim=@var{geom}
2179 @opindex fopenacc-dim
2180 @cindex OpenACC accelerator programming
2181 Specify default compute dimensions for parallel offload regions that do
2182 not explicitly specify. The @var{geom} value is a triple of
2183 ':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A size
2184 can be omitted, to use a target-specific default value.
2188 @cindex OpenMP parallel
2189 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
2190 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
2191 compiler generates parallel code according to the OpenMP Application
2192 Program Interface v4.5 @w{@uref{https://www.openmp.org}}. This option
2193 implies @option{-pthread}, and thus is only supported on targets that
2194 have support for @option{-pthread}. @option{-fopenmp} implies
2195 @option{-fopenmp-simd}.
2198 @opindex fopenmp-simd
2201 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
2202 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
2207 When the option @option{-fgnu-tm} is specified, the compiler
2208 generates code for the Linux variant of Intel's current Transactional
2209 Memory ABI specification document (Revision 1.1, May 6 2009). This is
2210 an experimental feature whose interface may change in future versions
2211 of GCC, as the official specification changes. Please note that not
2212 all architectures are supported for this feature.
2214 For more information on GCC's support for transactional memory,
2215 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
2216 Transactional Memory Library}.
2218 Note that the transactional memory feature is not supported with
2219 non-call exceptions (@option{-fnon-call-exceptions}).
2221 @item -fms-extensions
2222 @opindex fms-extensions
2223 Accept some non-standard constructs used in Microsoft header files.
2225 In C++ code, this allows member names in structures to be similar
2226 to previous types declarations.
2235 Some cases of unnamed fields in structures and unions are only
2236 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2237 fields within structs/unions}, for details.
2239 Note that this option is off for all targets but x86
2240 targets using ms-abi.
2242 @item -fplan9-extensions
2243 @opindex fplan9-extensions
2244 Accept some non-standard constructs used in Plan 9 code.
2246 This enables @option{-fms-extensions}, permits passing pointers to
2247 structures with anonymous fields to functions that expect pointers to
2248 elements of the type of the field, and permits referring to anonymous
2249 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2250 struct/union fields within structs/unions}, for details. This is only
2251 supported for C, not C++.
2253 @item -fcond-mismatch
2254 @opindex fcond-mismatch
2255 Allow conditional expressions with mismatched types in the second and
2256 third arguments. The value of such an expression is void. This option
2257 is not supported for C++.
2259 @item -flax-vector-conversions
2260 @opindex flax-vector-conversions
2261 Allow implicit conversions between vectors with differing numbers of
2262 elements and/or incompatible element types. This option should not be
2265 @item -funsigned-char
2266 @opindex funsigned-char
2267 Let the type @code{char} be unsigned, like @code{unsigned char}.
2269 Each kind of machine has a default for what @code{char} should
2270 be. It is either like @code{unsigned char} by default or like
2271 @code{signed char} by default.
2273 Ideally, a portable program should always use @code{signed char} or
2274 @code{unsigned char} when it depends on the signedness of an object.
2275 But many programs have been written to use plain @code{char} and
2276 expect it to be signed, or expect it to be unsigned, depending on the
2277 machines they were written for. This option, and its inverse, let you
2278 make such a program work with the opposite default.
2280 The type @code{char} is always a distinct type from each of
2281 @code{signed char} or @code{unsigned char}, even though its behavior
2282 is always just like one of those two.
2285 @opindex fsigned-char
2286 Let the type @code{char} be signed, like @code{signed char}.
2288 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2289 the negative form of @option{-funsigned-char}. Likewise, the option
2290 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2292 @item -fsigned-bitfields
2293 @itemx -funsigned-bitfields
2294 @itemx -fno-signed-bitfields
2295 @itemx -fno-unsigned-bitfields
2296 @opindex fsigned-bitfields
2297 @opindex funsigned-bitfields
2298 @opindex fno-signed-bitfields
2299 @opindex fno-unsigned-bitfields
2300 These options control whether a bit-field is signed or unsigned, when the
2301 declaration does not use either @code{signed} or @code{unsigned}. By
2302 default, such a bit-field is signed, because this is consistent: the
2303 basic integer types such as @code{int} are signed types.
2305 @item -fsso-struct=@var{endianness}
2306 @opindex fsso-struct
2307 Set the default scalar storage order of structures and unions to the
2308 specified endianness. The accepted values are @samp{big-endian},
2309 @samp{little-endian} and @samp{native} for the native endianness of
2310 the target (the default). This option is not supported for C++.
2312 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2313 code that is not binary compatible with code generated without it if the
2314 specified endianness is not the native endianness of the target.
2317 @node C++ Dialect Options
2318 @section Options Controlling C++ Dialect
2320 @cindex compiler options, C++
2321 @cindex C++ options, command-line
2322 @cindex options, C++
2323 This section describes the command-line options that are only meaningful
2324 for C++ programs. You can also use most of the GNU compiler options
2325 regardless of what language your program is in. For example, you
2326 might compile a file @file{firstClass.C} like this:
2329 g++ -g -fstrict-enums -O -c firstClass.C
2333 In this example, only @option{-fstrict-enums} is an option meant
2334 only for C++ programs; you can use the other options with any
2335 language supported by GCC@.
2337 Some options for compiling C programs, such as @option{-std}, are also
2338 relevant for C++ programs.
2339 @xref{C Dialect Options,,Options Controlling C Dialect}.
2341 Here is a list of options that are @emph{only} for compiling C++ programs:
2345 @item -fabi-version=@var{n}
2346 @opindex fabi-version
2347 Use version @var{n} of the C++ ABI@. The default is version 0.
2349 Version 0 refers to the version conforming most closely to
2350 the C++ ABI specification. Therefore, the ABI obtained using version 0
2351 will change in different versions of G++ as ABI bugs are fixed.
2353 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2355 Version 2 is the version of the C++ ABI that first appeared in G++
2356 3.4, and was the default through G++ 4.9.
2358 Version 3 corrects an error in mangling a constant address as a
2361 Version 4, which first appeared in G++ 4.5, implements a standard
2362 mangling for vector types.
2364 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2365 attribute const/volatile on function pointer types, decltype of a
2366 plain decl, and use of a function parameter in the declaration of
2369 Version 6, which first appeared in G++ 4.7, corrects the promotion
2370 behavior of C++11 scoped enums and the mangling of template argument
2371 packs, const/static_cast, prefix ++ and --, and a class scope function
2372 used as a template argument.
2374 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2375 builtin type and corrects the mangling of lambdas in default argument
2378 Version 8, which first appeared in G++ 4.9, corrects the substitution
2379 behavior of function types with function-cv-qualifiers.
2381 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2384 Version 10, which first appeared in G++ 6.1, adds mangling of
2385 attributes that affect type identity, such as ia32 calling convention
2386 attributes (e.g.@: @samp{stdcall}).
2388 Version 11, which first appeared in G++ 7, corrects the mangling of
2389 sizeof... expressions and operator names. For multiple entities with
2390 the same name within a function, that are declared in different scopes,
2391 the mangling now changes starting with the twelfth occurrence. It also
2392 implies @option{-fnew-inheriting-ctors}.
2394 Version 12, which first appeared in G++ 8, corrects the calling
2395 conventions for empty classes on the x86_64 target and for classes
2396 with only deleted copy/move constructors. It accidentally changes the
2397 calling convention for classes with a deleted copy constructor and a
2398 trivial move constructor.
2400 Version 13, which first appeared in G++ 8.2, fixes the accidental
2401 change in version 12.
2403 See also @option{-Wabi}.
2405 @item -fabi-compat-version=@var{n}
2406 @opindex fabi-compat-version
2407 On targets that support strong aliases, G++
2408 works around mangling changes by creating an alias with the correct
2409 mangled name when defining a symbol with an incorrect mangled name.
2410 This switch specifies which ABI version to use for the alias.
2412 With @option{-fabi-version=0} (the default), this defaults to 11 (GCC 7
2413 compatibility). If another ABI version is explicitly selected, this
2414 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2415 use @option{-fabi-compat-version=2}.
2417 If this option is not provided but @option{-Wabi=@var{n}} is, that
2418 version is used for compatibility aliases. If this option is provided
2419 along with @option{-Wabi} (without the version), the version from this
2420 option is used for the warning.
2422 @item -fno-access-control
2423 @opindex fno-access-control
2424 @opindex faccess-control
2425 Turn off all access checking. This switch is mainly useful for working
2426 around bugs in the access control code.
2429 @opindex faligned-new
2430 Enable support for C++17 @code{new} of types that require more
2431 alignment than @code{void* ::operator new(std::size_t)} provides. A
2432 numeric argument such as @code{-faligned-new=32} can be used to
2433 specify how much alignment (in bytes) is provided by that function,
2434 but few users will need to override the default of
2435 @code{alignof(std::max_align_t)}.
2437 This flag is enabled by default for @option{-std=c++17}.
2442 @opindex fno-char8_t
2443 Enable support for @code{char8_t} as adopted for C++2a. This includes
2444 the addition of a new @code{char8_t} fundamental type, changes to the
2445 types of UTF-8 string and character literals, new signatures for
2446 user-defined literals, associated standard library updates, and new
2447 @code{__cpp_char8_t} and @code{__cpp_lib_char8_t} feature test macros.
2449 This option enables functions to be overloaded for ordinary and UTF-8
2453 int f(const char *); // #1
2454 int f(const char8_t *); // #2
2455 int v1 = f("text"); // Calls #1
2456 int v2 = f(u8"text"); // Calls #2
2460 and introduces new signatures for user-defined literals:
2463 int operator""_udl1(char8_t);
2464 int v3 = u8'x'_udl1;
2465 int operator""_udl2(const char8_t*, std::size_t);
2466 int v4 = u8"text"_udl2;
2467 template<typename T, T...> int operator""_udl3();
2468 int v5 = u8"text"_udl3;
2472 The change to the types of UTF-8 string and character literals introduces
2473 incompatibilities with ISO C++11 and later standards. For example, the
2474 following code is well-formed under ISO C++11, but is ill-formed when
2475 @option{-fchar8_t} is specified.
2478 char ca[] = u8"xx"; // error: char-array initialized from wide
2480 const char *cp = u8"xx";// error: invalid conversion from
2481 // `const char8_t*' to `const char*'
2483 auto v = f(u8"xx"); // error: invalid conversion from
2484 // `const char8_t*' to `const char*'
2485 std::string s@{u8"xx"@}; // error: no matching function for call to
2486 // `std::basic_string<char>::basic_string()'
2487 using namespace std::literals;
2488 s = u8"xx"s; // error: conversion from
2489 // `basic_string<char8_t>' to non-scalar
2490 // type `basic_string<char>' requested
2495 Check that the pointer returned by @code{operator new} is non-null
2496 before attempting to modify the storage allocated. This check is
2497 normally unnecessary because the C++ standard specifies that
2498 @code{operator new} only returns @code{0} if it is declared
2499 @code{throw()}, in which case the compiler always checks the
2500 return value even without this option. In all other cases, when
2501 @code{operator new} has a non-empty exception specification, memory
2502 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2503 @samp{new (nothrow)}.
2507 Enable support for the C++ Extensions for Concepts Technical
2508 Specification, ISO 19217 (2015), which allows code like
2511 template <class T> concept bool Addable = requires (T t) @{ t + t; @};
2512 template <Addable T> T add (T a, T b) @{ return a + b; @}
2515 @item -fconstexpr-depth=@var{n}
2516 @opindex fconstexpr-depth
2517 Set the maximum nested evaluation depth for C++11 constexpr functions
2518 to @var{n}. A limit is needed to detect endless recursion during
2519 constant expression evaluation. The minimum specified by the standard
2522 @item -fconstexpr-loop-limit=@var{n}
2523 @opindex fconstexpr-loop-limit
2524 Set the maximum number of iterations for a loop in C++14 constexpr functions
2525 to @var{n}. A limit is needed to detect infinite loops during
2526 constant expression evaluation. The default is 262144 (1<<18).
2528 @item -fconstexpr-ops-limit=@var{n}
2529 @opindex fconstexpr-ops-limit
2530 Set the maximum number of operations during a single constexpr evaluation.
2531 Even when number of iterations of a single loop is limited with the above limit,
2532 if there are several nested loops and each of them has many iterations but still
2533 smaller than the above limit, or if in a body of some loop or even outside
2534 of a loop too many expressions need to be evaluated, the resulting constexpr
2535 evaluation might take too long.
2536 The default is 33554432 (1<<25).
2538 @item -fdeduce-init-list
2539 @opindex fdeduce-init-list
2540 Enable deduction of a template type parameter as
2541 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2544 template <class T> auto forward(T t) -> decltype (realfn (t))
2551 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2555 This deduction was implemented as a possible extension to the
2556 originally proposed semantics for the C++11 standard, but was not part
2557 of the final standard, so it is disabled by default. This option is
2558 deprecated, and may be removed in a future version of G++.
2560 @item -fno-elide-constructors
2561 @opindex fno-elide-constructors
2562 @opindex felide-constructors
2563 The C++ standard allows an implementation to omit creating a temporary
2564 that is only used to initialize another object of the same type.
2565 Specifying this option disables that optimization, and forces G++ to
2566 call the copy constructor in all cases. This option also causes G++
2567 to call trivial member functions which otherwise would be expanded inline.
2569 In C++17, the compiler is required to omit these temporaries, but this
2570 option still affects trivial member functions.
2572 @item -fno-enforce-eh-specs
2573 @opindex fno-enforce-eh-specs
2574 @opindex fenforce-eh-specs
2575 Don't generate code to check for violation of exception specifications
2576 at run time. This option violates the C++ standard, but may be useful
2577 for reducing code size in production builds, much like defining
2578 @code{NDEBUG}. This does not give user code permission to throw
2579 exceptions in violation of the exception specifications; the compiler
2580 still optimizes based on the specifications, so throwing an
2581 unexpected exception results in undefined behavior at run time.
2583 @item -fextern-tls-init
2584 @itemx -fno-extern-tls-init
2585 @opindex fextern-tls-init
2586 @opindex fno-extern-tls-init
2587 The C++11 and OpenMP standards allow @code{thread_local} and
2588 @code{threadprivate} variables to have dynamic (runtime)
2589 initialization. To support this, any use of such a variable goes
2590 through a wrapper function that performs any necessary initialization.
2591 When the use and definition of the variable are in the same
2592 translation unit, this overhead can be optimized away, but when the
2593 use is in a different translation unit there is significant overhead
2594 even if the variable doesn't actually need dynamic initialization. If
2595 the programmer can be sure that no use of the variable in a
2596 non-defining TU needs to trigger dynamic initialization (either
2597 because the variable is statically initialized, or a use of the
2598 variable in the defining TU will be executed before any uses in
2599 another TU), they can avoid this overhead with the
2600 @option{-fno-extern-tls-init} option.
2602 On targets that support symbol aliases, the default is
2603 @option{-fextern-tls-init}. On targets that do not support symbol
2604 aliases, the default is @option{-fno-extern-tls-init}.
2606 @item -fno-gnu-keywords
2607 @opindex fno-gnu-keywords
2608 @opindex fgnu-keywords
2609 Do not recognize @code{typeof} as a keyword, so that code can use this
2610 word as an identifier. You can use the keyword @code{__typeof__} instead.
2611 This option is implied by the strict ISO C++ dialects: @option{-ansi},
2612 @option{-std=c++98}, @option{-std=c++11}, etc.
2614 @item -fno-implicit-templates
2615 @opindex fno-implicit-templates
2616 @opindex fimplicit-templates
2617 Never emit code for non-inline templates that are instantiated
2618 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2619 If you use this option, you must take care to structure your code to
2620 include all the necessary explicit instantiations to avoid getting
2621 undefined symbols at link time.
2622 @xref{Template Instantiation}, for more information.
2624 @item -fno-implicit-inline-templates
2625 @opindex fno-implicit-inline-templates
2626 @opindex fimplicit-inline-templates
2627 Don't emit code for implicit instantiations of inline templates, either.
2628 The default is to handle inlines differently so that compiles with and
2629 without optimization need the same set of explicit instantiations.
2631 @item -fno-implement-inlines
2632 @opindex fno-implement-inlines
2633 @opindex fimplement-inlines
2634 To save space, do not emit out-of-line copies of inline functions
2635 controlled by @code{#pragma implementation}. This causes linker
2636 errors if these functions are not inlined everywhere they are called.
2638 @item -fms-extensions
2639 @opindex fms-extensions
2640 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2641 int and getting a pointer to member function via non-standard syntax.
2643 @item -fnew-inheriting-ctors
2644 @opindex fnew-inheriting-ctors
2645 Enable the P0136 adjustment to the semantics of C++11 constructor
2646 inheritance. This is part of C++17 but also considered to be a Defect
2647 Report against C++11 and C++14. This flag is enabled by default
2648 unless @option{-fabi-version=10} or lower is specified.
2650 @item -fnew-ttp-matching
2651 @opindex fnew-ttp-matching
2652 Enable the P0522 resolution to Core issue 150, template template
2653 parameters and default arguments: this allows a template with default
2654 template arguments as an argument for a template template parameter
2655 with fewer template parameters. This flag is enabled by default for
2656 @option{-std=c++17}.
2658 @item -fno-nonansi-builtins
2659 @opindex fno-nonansi-builtins
2660 @opindex fnonansi-builtins
2661 Disable built-in declarations of functions that are not mandated by
2662 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2663 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2666 @opindex fnothrow-opt
2667 Treat a @code{throw()} exception specification as if it were a
2668 @code{noexcept} specification to reduce or eliminate the text size
2669 overhead relative to a function with no exception specification. If
2670 the function has local variables of types with non-trivial
2671 destructors, the exception specification actually makes the
2672 function smaller because the EH cleanups for those variables can be
2673 optimized away. The semantic effect is that an exception thrown out of
2674 a function with such an exception specification results in a call
2675 to @code{terminate} rather than @code{unexpected}.
2677 @item -fno-operator-names
2678 @opindex fno-operator-names
2679 @opindex foperator-names
2680 Do not treat the operator name keywords @code{and}, @code{bitand},
2681 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2682 synonyms as keywords.
2684 @item -fno-optional-diags
2685 @opindex fno-optional-diags
2686 @opindex foptional-diags
2687 Disable diagnostics that the standard says a compiler does not need to
2688 issue. Currently, the only such diagnostic issued by G++ is the one for
2689 a name having multiple meanings within a class.
2692 @opindex fpermissive
2693 Downgrade some diagnostics about nonconformant code from errors to
2694 warnings. Thus, using @option{-fpermissive} allows some
2695 nonconforming code to compile.
2697 @item -fno-pretty-templates
2698 @opindex fno-pretty-templates
2699 @opindex fpretty-templates
2700 When an error message refers to a specialization of a function
2701 template, the compiler normally prints the signature of the
2702 template followed by the template arguments and any typedefs or
2703 typenames in the signature (e.g.@: @code{void f(T) [with T = int]}
2704 rather than @code{void f(int)}) so that it's clear which template is
2705 involved. When an error message refers to a specialization of a class
2706 template, the compiler omits any template arguments that match
2707 the default template arguments for that template. If either of these
2708 behaviors make it harder to understand the error message rather than
2709 easier, you can use @option{-fno-pretty-templates} to disable them.
2713 Enable automatic template instantiation at link time. This option also
2714 implies @option{-fno-implicit-templates}. @xref{Template
2715 Instantiation}, for more information.
2720 Disable generation of information about every class with virtual
2721 functions for use by the C++ run-time type identification features
2722 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2723 of the language, you can save some space by using this flag. Note that
2724 exception handling uses the same information, but G++ generates it as
2725 needed. The @code{dynamic_cast} operator can still be used for casts that
2726 do not require run-time type information, i.e.@: casts to @code{void *} or to
2727 unambiguous base classes.
2729 Mixing code compiled with @option{-frtti} with that compiled with
2730 @option{-fno-rtti} may not work. For example, programs may
2731 fail to link if a class compiled with @option{-fno-rtti} is used as a base
2732 for a class compiled with @option{-frtti}.
2734 @item -fsized-deallocation
2735 @opindex fsized-deallocation
2736 Enable the built-in global declarations
2738 void operator delete (void *, std::size_t) noexcept;
2739 void operator delete[] (void *, std::size_t) noexcept;
2741 as introduced in C++14. This is useful for user-defined replacement
2742 deallocation functions that, for example, use the size of the object
2743 to make deallocation faster. Enabled by default under
2744 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2745 warns about places that might want to add a definition.
2747 @item -fstrict-enums
2748 @opindex fstrict-enums
2749 Allow the compiler to optimize using the assumption that a value of
2750 enumerated type can only be one of the values of the enumeration (as
2751 defined in the C++ standard; basically, a value that can be
2752 represented in the minimum number of bits needed to represent all the
2753 enumerators). This assumption may not be valid if the program uses a
2754 cast to convert an arbitrary integer value to the enumerated type.
2756 @item -fstrong-eval-order
2757 @opindex fstrong-eval-order
2758 Evaluate member access, array subscripting, and shift expressions in
2759 left-to-right order, and evaluate assignment in right-to-left order,
2760 as adopted for C++17. Enabled by default with @option{-std=c++17}.
2761 @option{-fstrong-eval-order=some} enables just the ordering of member
2762 access and shift expressions, and is the default without
2763 @option{-std=c++17}.
2765 @item -ftemplate-backtrace-limit=@var{n}
2766 @opindex ftemplate-backtrace-limit
2767 Set the maximum number of template instantiation notes for a single
2768 warning or error to @var{n}. The default value is 10.
2770 @item -ftemplate-depth=@var{n}
2771 @opindex ftemplate-depth
2772 Set the maximum instantiation depth for template classes to @var{n}.
2773 A limit on the template instantiation depth is needed to detect
2774 endless recursions during template class instantiation. ANSI/ISO C++
2775 conforming programs must not rely on a maximum depth greater than 17
2776 (changed to 1024 in C++11). The default value is 900, as the compiler
2777 can run out of stack space before hitting 1024 in some situations.
2779 @item -fno-threadsafe-statics
2780 @opindex fno-threadsafe-statics
2781 @opindex fthreadsafe-statics
2782 Do not emit the extra code to use the routines specified in the C++
2783 ABI for thread-safe initialization of local statics. You can use this
2784 option to reduce code size slightly in code that doesn't need to be
2787 @item -fuse-cxa-atexit
2788 @opindex fuse-cxa-atexit
2789 Register destructors for objects with static storage duration with the
2790 @code{__cxa_atexit} function rather than the @code{atexit} function.
2791 This option is required for fully standards-compliant handling of static
2792 destructors, but only works if your C library supports
2793 @code{__cxa_atexit}.
2795 @item -fno-use-cxa-get-exception-ptr
2796 @opindex fno-use-cxa-get-exception-ptr
2797 @opindex fuse-cxa-get-exception-ptr
2798 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2799 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2800 if the runtime routine is not available.
2802 @item -fvisibility-inlines-hidden
2803 @opindex fvisibility-inlines-hidden
2804 This switch declares that the user does not attempt to compare
2805 pointers to inline functions or methods where the addresses of the two functions
2806 are taken in different shared objects.
2808 The effect of this is that GCC may, effectively, mark inline methods with
2809 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2810 appear in the export table of a DSO and do not require a PLT indirection
2811 when used within the DSO@. Enabling this option can have a dramatic effect
2812 on load and link times of a DSO as it massively reduces the size of the
2813 dynamic export table when the library makes heavy use of templates.
2815 The behavior of this switch is not quite the same as marking the
2816 methods as hidden directly, because it does not affect static variables
2817 local to the function or cause the compiler to deduce that
2818 the function is defined in only one shared object.
2820 You may mark a method as having a visibility explicitly to negate the
2821 effect of the switch for that method. For example, if you do want to
2822 compare pointers to a particular inline method, you might mark it as
2823 having default visibility. Marking the enclosing class with explicit
2824 visibility has no effect.
2826 Explicitly instantiated inline methods are unaffected by this option
2827 as their linkage might otherwise cross a shared library boundary.
2828 @xref{Template Instantiation}.
2830 @item -fvisibility-ms-compat
2831 @opindex fvisibility-ms-compat
2832 This flag attempts to use visibility settings to make GCC's C++
2833 linkage model compatible with that of Microsoft Visual Studio.
2835 The flag makes these changes to GCC's linkage model:
2839 It sets the default visibility to @code{hidden}, like
2840 @option{-fvisibility=hidden}.
2843 Types, but not their members, are not hidden by default.
2846 The One Definition Rule is relaxed for types without explicit
2847 visibility specifications that are defined in more than one
2848 shared object: those declarations are permitted if they are
2849 permitted when this option is not used.
2852 In new code it is better to use @option{-fvisibility=hidden} and
2853 export those classes that are intended to be externally visible.
2854 Unfortunately it is possible for code to rely, perhaps accidentally,
2855 on the Visual Studio behavior.
2857 Among the consequences of these changes are that static data members
2858 of the same type with the same name but defined in different shared
2859 objects are different, so changing one does not change the other;
2860 and that pointers to function members defined in different shared
2861 objects may not compare equal. When this flag is given, it is a
2862 violation of the ODR to define types with the same name differently.
2867 Do not use weak symbol support, even if it is provided by the linker.
2868 By default, G++ uses weak symbols if they are available. This
2869 option exists only for testing, and should not be used by end-users;
2870 it results in inferior code and has no benefits. This option may
2871 be removed in a future release of G++.
2875 Do not search for header files in the standard directories specific to
2876 C++, but do still search the other standard directories. (This option
2877 is used when building the C++ library.)
2880 In addition, these optimization, warning, and code generation options
2881 have meanings only for C++ programs:
2884 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2887 Warn when G++ it generates code that is probably not compatible with
2888 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2889 ABI with each major release, normally @option{-Wabi} will warn only if
2890 there is a check added later in a release series for an ABI issue
2891 discovered since the initial release. @option{-Wabi} will warn about
2892 more things if an older ABI version is selected (with
2893 @option{-fabi-version=@var{n}}).
2895 @option{-Wabi} can also be used with an explicit version number to
2896 warn about compatibility with a particular @option{-fabi-version}
2897 level, e.g.@: @option{-Wabi=2} to warn about changes relative to
2898 @option{-fabi-version=2}.
2900 If an explicit version number is provided and
2901 @option{-fabi-compat-version} is not specified, the version number
2902 from this option is used for compatibility aliases. If no explicit
2903 version number is provided with this option, but
2904 @option{-fabi-compat-version} is specified, that version number is
2905 used for ABI warnings.
2907 Although an effort has been made to warn about
2908 all such cases, there are probably some cases that are not warned about,
2909 even though G++ is generating incompatible code. There may also be
2910 cases where warnings are emitted even though the code that is generated
2913 You should rewrite your code to avoid these warnings if you are
2914 concerned about the fact that code generated by G++ may not be binary
2915 compatible with code generated by other compilers.
2917 Known incompatibilities in @option{-fabi-version=2} (which was the
2918 default from GCC 3.4 to 4.9) include:
2923 A template with a non-type template parameter of reference type was
2924 mangled incorrectly:
2927 template <int &> struct S @{@};
2931 This was fixed in @option{-fabi-version=3}.
2934 SIMD vector types declared using @code{__attribute ((vector_size))} were
2935 mangled in a non-standard way that does not allow for overloading of
2936 functions taking vectors of different sizes.
2938 The mangling was changed in @option{-fabi-version=4}.
2941 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2942 qualifiers, and @code{decltype} of a plain declaration was folded away.
2944 These mangling issues were fixed in @option{-fabi-version=5}.
2947 Scoped enumerators passed as arguments to a variadic function are
2948 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2949 On most targets this does not actually affect the parameter passing
2950 ABI, as there is no way to pass an argument smaller than @code{int}.
2952 Also, the ABI changed the mangling of template argument packs,
2953 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2954 a class scope function used as a template argument.
2956 These issues were corrected in @option{-fabi-version=6}.
2959 Lambdas in default argument scope were mangled incorrectly, and the
2960 ABI changed the mangling of @code{nullptr_t}.
2962 These issues were corrected in @option{-fabi-version=7}.
2965 When mangling a function type with function-cv-qualifiers, the
2966 un-qualified function type was incorrectly treated as a substitution
2969 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2972 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2973 unaligned accesses. Note that this did not affect the ABI of a
2974 function with a @code{nullptr_t} parameter, as parameters have a
2977 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2980 Target-specific attributes that affect the identity of a type, such as
2981 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2982 did not affect the mangled name, leading to name collisions when
2983 function pointers were used as template arguments.
2985 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2989 It also warns about psABI-related changes. The known psABI changes at this
2995 For SysV/x86-64, unions with @code{long double} members are
2996 passed in memory as specified in psABI. For example:
3006 @code{union U} is always passed in memory.
3010 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
3013 Warn when a type with an ABI tag is used in a context that does not
3014 have that ABI tag. See @ref{C++ Attributes} for more information
3017 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
3018 @opindex Wctor-dtor-privacy
3019 @opindex Wno-ctor-dtor-privacy
3020 Warn when a class seems unusable because all the constructors or
3021 destructors in that class are private, and it has neither friends nor
3022 public static member functions. Also warn if there are no non-private
3023 methods, and there's at least one private member function that isn't
3024 a constructor or destructor.
3026 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
3027 @opindex Wdelete-non-virtual-dtor
3028 @opindex Wno-delete-non-virtual-dtor
3029 Warn when @code{delete} is used to destroy an instance of a class that
3030 has virtual functions and non-virtual destructor. It is unsafe to delete
3031 an instance of a derived class through a pointer to a base class if the
3032 base class does not have a virtual destructor. This warning is enabled
3035 @item -Wdeprecated-copy @r{(C++ and Objective-C++ only)}
3036 @opindex Wdeprecated-copy
3037 @opindex Wno-deprecated-copy
3038 Warn that the implicit declaration of a copy constructor or copy
3039 assignment operator is deprecated if the class has a user-provided
3040 copy constructor or copy assignment operator, in C++11 and up. This
3041 warning is enabled by @option{-Wextra}. With
3042 @option{-Wdeprecated-copy-dtor}, also deprecate if the class has a
3043 user-provided destructor.
3045 @item -Wno-init-list-lifetime @r{(C++ and Objective-C++ only)}
3046 @opindex Winit-list-lifetime
3047 @opindex Wno-init-list-lifetime
3048 Do not warn about uses of @code{std::initializer_list} that are likely
3049 to result in dangling pointers. Since the underlying array for an
3050 @code{initializer_list} is handled like a normal C++ temporary object,
3051 it is easy to inadvertently keep a pointer to the array past the end
3052 of the array's lifetime. For example:
3056 If a function returns a temporary @code{initializer_list}, or a local
3057 @code{initializer_list} variable, the array's lifetime ends at the end
3058 of the return statement, so the value returned has a dangling pointer.
3061 If a new-expression creates an @code{initializer_list}, the array only
3062 lives until the end of the enclosing full-expression, so the
3063 @code{initializer_list} in the heap has a dangling pointer.
3066 When an @code{initializer_list} variable is assigned from a
3067 brace-enclosed initializer list, the temporary array created for the
3068 right side of the assignment only lives until the end of the
3069 full-expression, so at the next statement the @code{initializer_list}
3070 variable has a dangling pointer.
3073 // li's initial underlying array lives as long as li
3074 std::initializer_list<int> li = @{ 1,2,3 @};
3075 // assignment changes li to point to a temporary array
3077 // now the temporary is gone and li has a dangling pointer
3078 int i = li.begin()[0] // undefined behavior
3082 When a list constructor stores the @code{begin} pointer from the
3083 @code{initializer_list} argument, this doesn't extend the lifetime of
3084 the array, so if a class variable is constructed from a temporary
3085 @code{initializer_list}, the pointer is left dangling by the end of
3086 the variable declaration statement.
3090 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
3091 @opindex Wliteral-suffix
3092 @opindex Wno-literal-suffix
3093 Warn when a string or character literal is followed by a ud-suffix which does
3094 not begin with an underscore. As a conforming extension, GCC treats such
3095 suffixes as separate preprocessing tokens in order to maintain backwards
3096 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
3100 #define __STDC_FORMAT_MACROS
3101 #include <inttypes.h>
3106 printf("My int64: %" PRId64"\n", i64);
3110 In this case, @code{PRId64} is treated as a separate preprocessing token.
3112 Additionally, warn when a user-defined literal operator is declared with
3113 a literal suffix identifier that doesn't begin with an underscore. Literal
3114 suffix identifiers that don't begin with an underscore are reserved for
3115 future standardization.
3117 This warning is enabled by default.
3119 @item -Wlto-type-mismatch
3120 @opindex Wlto-type-mismatch
3121 @opindex Wno-lto-type-mismatch
3123 During the link-time optimization warn about type mismatches in
3124 global declarations from different compilation units.
3125 Requires @option{-flto} to be enabled. Enabled by default.
3127 @item -Wno-narrowing @r{(C++ and Objective-C++ only)}
3129 @opindex Wno-narrowing
3130 For C++11 and later standards, narrowing conversions are diagnosed by default,
3131 as required by the standard. A narrowing conversion from a constant produces
3132 an error, and a narrowing conversion from a non-constant produces a warning,
3133 but @option{-Wno-narrowing} suppresses the diagnostic.
3134 Note that this does not affect the meaning of well-formed code;
3135 narrowing conversions are still considered ill-formed in SFINAE contexts.
3137 With @option{-Wnarrowing} in C++98, warn when a narrowing
3138 conversion prohibited by C++11 occurs within
3142 int i = @{ 2.2 @}; // error: narrowing from double to int
3145 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
3147 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
3149 @opindex Wno-noexcept
3150 Warn when a noexcept-expression evaluates to false because of a call
3151 to a function that does not have a non-throwing exception
3152 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
3153 the compiler to never throw an exception.
3155 @item -Wnoexcept-type @r{(C++ and Objective-C++ only)}
3156 @opindex Wnoexcept-type
3157 @opindex Wno-noexcept-type
3158 Warn if the C++17 feature making @code{noexcept} part of a function
3159 type changes the mangled name of a symbol relative to C++14. Enabled
3160 by @option{-Wabi} and @option{-Wc++17-compat}.
3165 template <class T> void f(T t) @{ t(); @};
3167 void h() @{ f(g); @}
3171 In C++14, @code{f} calls @code{f<void(*)()>}, but in
3172 C++17 it calls @code{f<void(*)()noexcept>}.
3174 @item -Wclass-memaccess @r{(C++ and Objective-C++ only)}
3175 @opindex Wclass-memaccess
3176 @opindex Wno-class-memaccess
3177 Warn when the destination of a call to a raw memory function such as
3178 @code{memset} or @code{memcpy} is an object of class type, and when writing
3179 into such an object might bypass the class non-trivial or deleted constructor
3180 or copy assignment, violate const-correctness or encapsulation, or corrupt
3181 virtual table pointers. Modifying the representation of such objects may
3182 violate invariants maintained by member functions of the class. For example,
3183 the call to @code{memset} below is undefined because it modifies a non-trivial
3184 class object and is, therefore, diagnosed. The safe way to either initialize
3185 or clear the storage of objects of such types is by using the appropriate
3186 constructor or assignment operator, if one is available.
3188 std::string str = "abc";
3189 memset (&str, 0, sizeof str);
3191 The @option{-Wclass-memaccess} option is enabled by @option{-Wall}.
3192 Explicitly casting the pointer to the class object to @code{void *} or
3193 to a type that can be safely accessed by the raw memory function suppresses
3196 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
3197 @opindex Wnon-virtual-dtor
3198 @opindex Wno-non-virtual-dtor
3199 Warn when a class has virtual functions and an accessible non-virtual
3200 destructor itself or in an accessible polymorphic base class, in which
3201 case it is possible but unsafe to delete an instance of a derived
3202 class through a pointer to the class itself or base class. This
3203 warning is automatically enabled if @option{-Weffc++} is specified.
3205 @item -Wregister @r{(C++ and Objective-C++ only)}
3207 @opindex Wno-register
3208 Warn on uses of the @code{register} storage class specifier, except
3209 when it is part of the GNU @ref{Explicit Register Variables} extension.
3210 The use of the @code{register} keyword as storage class specifier has
3211 been deprecated in C++11 and removed in C++17.
3212 Enabled by default with @option{-std=c++17}.
3214 @item -Wreorder @r{(C++ and Objective-C++ only)}
3216 @opindex Wno-reorder
3217 @cindex reordering, warning
3218 @cindex warning for reordering of member initializers
3219 Warn when the order of member initializers given in the code does not
3220 match the order in which they must be executed. For instance:
3226 A(): j (0), i (1) @{ @}
3231 The compiler rearranges the member initializers for @code{i}
3232 and @code{j} to match the declaration order of the members, emitting
3233 a warning to that effect. This warning is enabled by @option{-Wall}.
3235 @item -Wno-pessimizing-move @r{(C++ and Objective-C++ only)}
3236 @opindex Wpessimizing-move
3237 @opindex Wno-pessimizing-move
3238 This warning warns when a call to @code{std::move} prevents copy
3239 elision. A typical scenario when copy elision can occur is when returning in
3240 a function with a class return type, when the expression being returned is the
3241 name of a non-volatile automatic object, and is not a function parameter, and
3242 has the same type as the function return type.
3252 return std::move (t);
3256 But in this example, the @code{std::move} call prevents copy elision.
3258 This warning is enabled by @option{-Wall}.
3260 @item -Wno-redundant-move @r{(C++ and Objective-C++ only)}
3261 @opindex Wredundant-move
3262 @opindex Wno-redundant-move
3263 This warning warns about redundant calls to @code{std::move}; that is, when
3264 a move operation would have been performed even without the @code{std::move}
3265 call. This happens because the compiler is forced to treat the object as if
3266 it were an rvalue in certain situations such as returning a local variable,
3267 where copy elision isn't applicable. Consider:
3276 return std::move (t);
3280 Here, the @code{std::move} call is redundant. Because G++ implements Core
3281 Issue 1579, another example is:
3284 struct T @{ // convertible to U
3294 return std::move (t);
3297 In this example, copy elision isn't applicable because the type of the
3298 expression being returned and the function return type differ, yet G++
3299 treats the return value as if it were designated by an rvalue.
3301 This warning is enabled by @option{-Wextra}.
3303 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
3304 @opindex fext-numeric-literals
3305 @opindex fno-ext-numeric-literals
3306 Accept imaginary, fixed-point, or machine-defined
3307 literal number suffixes as GNU extensions.
3308 When this option is turned off these suffixes are treated
3309 as C++11 user-defined literal numeric suffixes.
3310 This is on by default for all pre-C++11 dialects and all GNU dialects:
3311 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
3312 @option{-std=gnu++14}.
3313 This option is off by default
3314 for ISO C++11 onwards (@option{-std=c++11}, ...).
3317 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
3320 @item -Weffc++ @r{(C++ and Objective-C++ only)}
3323 Warn about violations of the following style guidelines from Scott Meyers'
3324 @cite{Effective C++} series of books:
3328 Define a copy constructor and an assignment operator for classes
3329 with dynamically-allocated memory.
3332 Prefer initialization to assignment in constructors.
3335 Have @code{operator=} return a reference to @code{*this}.
3338 Don't try to return a reference when you must return an object.
3341 Distinguish between prefix and postfix forms of increment and
3342 decrement operators.
3345 Never overload @code{&&}, @code{||}, or @code{,}.
3349 This option also enables @option{-Wnon-virtual-dtor}, which is also
3350 one of the effective C++ recommendations. However, the check is
3351 extended to warn about the lack of virtual destructor in accessible
3352 non-polymorphic bases classes too.
3354 When selecting this option, be aware that the standard library
3355 headers do not obey all of these guidelines; use @samp{grep -v}
3356 to filter out those warnings.
3358 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
3359 @opindex Wstrict-null-sentinel
3360 @opindex Wno-strict-null-sentinel
3361 Warn about the use of an uncasted @code{NULL} as sentinel. When
3362 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
3363 to @code{__null}. Although it is a null pointer constant rather than a
3364 null pointer, it is guaranteed to be of the same size as a pointer.
3365 But this use is not portable across different compilers.
3367 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
3368 @opindex Wno-non-template-friend
3369 @opindex Wnon-template-friend
3370 Disable warnings when non-template friend functions are declared
3371 within a template. In very old versions of GCC that predate implementation
3372 of the ISO standard, declarations such as
3373 @samp{friend int foo(int)}, where the name of the friend is an unqualified-id,
3374 could be interpreted as a particular specialization of a template
3375 function; the warning exists to diagnose compatibility problems,
3376 and is enabled by default.
3378 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
3379 @opindex Wold-style-cast
3380 @opindex Wno-old-style-cast
3381 Warn if an old-style (C-style) cast to a non-void type is used within
3382 a C++ program. The new-style casts (@code{dynamic_cast},
3383 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
3384 less vulnerable to unintended effects and much easier to search for.
3386 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
3387 @opindex Woverloaded-virtual
3388 @opindex Wno-overloaded-virtual
3389 @cindex overloaded virtual function, warning
3390 @cindex warning for overloaded virtual function
3391 Warn when a function declaration hides virtual functions from a
3392 base class. For example, in:
3399 struct B: public A @{
3404 the @code{A} class version of @code{f} is hidden in @code{B}, and code
3415 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
3416 @opindex Wno-pmf-conversions
3417 @opindex Wpmf-conversions
3418 Disable the diagnostic for converting a bound pointer to member function
3421 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
3422 @opindex Wsign-promo
3423 @opindex Wno-sign-promo
3424 Warn when overload resolution chooses a promotion from unsigned or
3425 enumerated type to a signed type, over a conversion to an unsigned type of
3426 the same size. Previous versions of G++ tried to preserve
3427 unsignedness, but the standard mandates the current behavior.
3429 @item -Wtemplates @r{(C++ and Objective-C++ only)}
3431 @opindex Wno-templates
3432 Warn when a primary template declaration is encountered. Some coding
3433 rules disallow templates, and this may be used to enforce that rule.
3434 The warning is inactive inside a system header file, such as the STL, so
3435 one can still use the STL. One may also instantiate or specialize
3438 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
3439 @opindex Wmultiple-inheritance
3440 @opindex Wno-multiple-inheritance
3441 Warn when a class is defined with multiple direct base classes. Some
3442 coding rules disallow multiple inheritance, and this may be used to
3443 enforce that rule. The warning is inactive inside a system header file,
3444 such as the STL, so one can still use the STL. One may also define
3445 classes that indirectly use multiple inheritance.
3447 @item -Wvirtual-inheritance
3448 @opindex Wvirtual-inheritance
3449 @opindex Wno-virtual-inheritance
3450 Warn when a class is defined with a virtual direct base class. Some
3451 coding rules disallow multiple inheritance, and this may be used to
3452 enforce that rule. The warning is inactive inside a system header file,
3453 such as the STL, so one can still use the STL. One may also define
3454 classes that indirectly use virtual inheritance.
3457 @opindex Wnamespaces
3458 @opindex Wno-namespaces
3459 Warn when a namespace definition is opened. Some coding rules disallow
3460 namespaces, and this may be used to enforce that rule. The warning is
3461 inactive inside a system header file, such as the STL, so one can still
3462 use the STL. One may also use using directives and qualified names.
3464 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
3466 @opindex Wno-terminate
3467 Disable the warning about a throw-expression that will immediately
3468 result in a call to @code{terminate}.
3470 @item -Wno-class-conversion @r{(C++ and Objective-C++ only)}
3471 @opindex Wno-class-conversion
3472 @opindex Wclass-conversion
3473 Disable the warning about the case when a conversion function converts an
3474 object to the same type, to a base class of that type, or to void; such
3475 a conversion function will never be called.
3478 @node Objective-C and Objective-C++ Dialect Options
3479 @section Options Controlling Objective-C and Objective-C++ Dialects
3481 @cindex compiler options, Objective-C and Objective-C++
3482 @cindex Objective-C and Objective-C++ options, command-line
3483 @cindex options, Objective-C and Objective-C++
3484 (NOTE: This manual does not describe the Objective-C and Objective-C++
3485 languages themselves. @xref{Standards,,Language Standards
3486 Supported by GCC}, for references.)
3488 This section describes the command-line options that are only meaningful
3489 for Objective-C and Objective-C++ programs. You can also use most of
3490 the language-independent GNU compiler options.
3491 For example, you might compile a file @file{some_class.m} like this:
3494 gcc -g -fgnu-runtime -O -c some_class.m
3498 In this example, @option{-fgnu-runtime} is an option meant only for
3499 Objective-C and Objective-C++ programs; you can use the other options with
3500 any language supported by GCC@.
3502 Note that since Objective-C is an extension of the C language, Objective-C
3503 compilations may also use options specific to the C front-end (e.g.,
3504 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3505 C++-specific options (e.g., @option{-Wabi}).
3507 Here is a list of options that are @emph{only} for compiling Objective-C
3508 and Objective-C++ programs:
3511 @item -fconstant-string-class=@var{class-name}
3512 @opindex fconstant-string-class
3513 Use @var{class-name} as the name of the class to instantiate for each
3514 literal string specified with the syntax @code{@@"@dots{}"}. The default
3515 class name is @code{NXConstantString} if the GNU runtime is being used, and
3516 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3517 @option{-fconstant-cfstrings} option, if also present, overrides the
3518 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3519 to be laid out as constant CoreFoundation strings.
3522 @opindex fgnu-runtime
3523 Generate object code compatible with the standard GNU Objective-C
3524 runtime. This is the default for most types of systems.
3526 @item -fnext-runtime
3527 @opindex fnext-runtime
3528 Generate output compatible with the NeXT runtime. This is the default
3529 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3530 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3533 @item -fno-nil-receivers
3534 @opindex fno-nil-receivers
3535 @opindex fnil-receivers
3536 Assume that all Objective-C message dispatches (@code{[receiver
3537 message:arg]}) in this translation unit ensure that the receiver is
3538 not @code{nil}. This allows for more efficient entry points in the
3539 runtime to be used. This option is only available in conjunction with
3540 the NeXT runtime and ABI version 0 or 1.
3542 @item -fobjc-abi-version=@var{n}
3543 @opindex fobjc-abi-version
3544 Use version @var{n} of the Objective-C ABI for the selected runtime.
3545 This option is currently supported only for the NeXT runtime. In that
3546 case, Version 0 is the traditional (32-bit) ABI without support for
3547 properties and other Objective-C 2.0 additions. Version 1 is the
3548 traditional (32-bit) ABI with support for properties and other
3549 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3550 nothing is specified, the default is Version 0 on 32-bit target
3551 machines, and Version 2 on 64-bit target machines.
3553 @item -fobjc-call-cxx-cdtors
3554 @opindex fobjc-call-cxx-cdtors
3555 For each Objective-C class, check if any of its instance variables is a
3556 C++ object with a non-trivial default constructor. If so, synthesize a
3557 special @code{- (id) .cxx_construct} instance method which runs
3558 non-trivial default constructors on any such instance variables, in order,
3559 and then return @code{self}. Similarly, check if any instance variable
3560 is a C++ object with a non-trivial destructor, and if so, synthesize a
3561 special @code{- (void) .cxx_destruct} method which runs
3562 all such default destructors, in reverse order.
3564 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3565 methods thusly generated only operate on instance variables
3566 declared in the current Objective-C class, and not those inherited
3567 from superclasses. It is the responsibility of the Objective-C
3568 runtime to invoke all such methods in an object's inheritance
3569 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3570 by the runtime immediately after a new object instance is allocated;
3571 the @code{- (void) .cxx_destruct} methods are invoked immediately
3572 before the runtime deallocates an object instance.
3574 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3575 support for invoking the @code{- (id) .cxx_construct} and
3576 @code{- (void) .cxx_destruct} methods.
3578 @item -fobjc-direct-dispatch
3579 @opindex fobjc-direct-dispatch
3580 Allow fast jumps to the message dispatcher. On Darwin this is
3581 accomplished via the comm page.
3583 @item -fobjc-exceptions
3584 @opindex fobjc-exceptions
3585 Enable syntactic support for structured exception handling in
3586 Objective-C, similar to what is offered by C++. This option
3587 is required to use the Objective-C keywords @code{@@try},
3588 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3589 @code{@@synchronized}. This option is available with both the GNU
3590 runtime and the NeXT runtime (but not available in conjunction with
3591 the NeXT runtime on Mac OS X 10.2 and earlier).
3595 Enable garbage collection (GC) in Objective-C and Objective-C++
3596 programs. This option is only available with the NeXT runtime; the
3597 GNU runtime has a different garbage collection implementation that
3598 does not require special compiler flags.
3600 @item -fobjc-nilcheck
3601 @opindex fobjc-nilcheck
3602 For the NeXT runtime with version 2 of the ABI, check for a nil
3603 receiver in method invocations before doing the actual method call.
3604 This is the default and can be disabled using
3605 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3606 checked for nil in this way no matter what this flag is set to.
3607 Currently this flag does nothing when the GNU runtime, or an older
3608 version of the NeXT runtime ABI, is used.
3610 @item -fobjc-std=objc1
3612 Conform to the language syntax of Objective-C 1.0, the language
3613 recognized by GCC 4.0. This only affects the Objective-C additions to
3614 the C/C++ language; it does not affect conformance to C/C++ standards,
3615 which is controlled by the separate C/C++ dialect option flags. When
3616 this option is used with the Objective-C or Objective-C++ compiler,
3617 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3618 This is useful if you need to make sure that your Objective-C code can
3619 be compiled with older versions of GCC@.
3621 @item -freplace-objc-classes
3622 @opindex freplace-objc-classes
3623 Emit a special marker instructing @command{ld(1)} not to statically link in
3624 the resulting object file, and allow @command{dyld(1)} to load it in at
3625 run time instead. This is used in conjunction with the Fix-and-Continue
3626 debugging mode, where the object file in question may be recompiled and
3627 dynamically reloaded in the course of program execution, without the need
3628 to restart the program itself. Currently, Fix-and-Continue functionality
3629 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3634 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3635 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3636 compile time) with static class references that get initialized at load time,
3637 which improves run-time performance. Specifying the @option{-fzero-link} flag
3638 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3639 to be retained. This is useful in Zero-Link debugging mode, since it allows
3640 for individual class implementations to be modified during program execution.
3641 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3642 regardless of command-line options.
3644 @item -fno-local-ivars
3645 @opindex fno-local-ivars
3646 @opindex flocal-ivars
3647 By default instance variables in Objective-C can be accessed as if
3648 they were local variables from within the methods of the class they're
3649 declared in. This can lead to shadowing between instance variables
3650 and other variables declared either locally inside a class method or
3651 globally with the same name. Specifying the @option{-fno-local-ivars}
3652 flag disables this behavior thus avoiding variable shadowing issues.
3654 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3655 @opindex fivar-visibility
3656 Set the default instance variable visibility to the specified option
3657 so that instance variables declared outside the scope of any access
3658 modifier directives default to the specified visibility.
3662 Dump interface declarations for all classes seen in the source file to a
3663 file named @file{@var{sourcename}.decl}.
3665 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3666 @opindex Wassign-intercept
3667 @opindex Wno-assign-intercept
3668 Warn whenever an Objective-C assignment is being intercepted by the
3671 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3672 @opindex Wno-protocol
3674 If a class is declared to implement a protocol, a warning is issued for
3675 every method in the protocol that is not implemented by the class. The
3676 default behavior is to issue a warning for every method not explicitly
3677 implemented in the class, even if a method implementation is inherited
3678 from the superclass. If you use the @option{-Wno-protocol} option, then
3679 methods inherited from the superclass are considered to be implemented,
3680 and no warning is issued for them.
3682 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3684 @opindex Wno-selector
3685 Warn if multiple methods of different types for the same selector are
3686 found during compilation. The check is performed on the list of methods
3687 in the final stage of compilation. Additionally, a check is performed
3688 for each selector appearing in a @code{@@selector(@dots{})}
3689 expression, and a corresponding method for that selector has been found
3690 during compilation. Because these checks scan the method table only at
3691 the end of compilation, these warnings are not produced if the final
3692 stage of compilation is not reached, for example because an error is
3693 found during compilation, or because the @option{-fsyntax-only} option is
3696 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3697 @opindex Wstrict-selector-match
3698 @opindex Wno-strict-selector-match
3699 Warn if multiple methods with differing argument and/or return types are
3700 found for a given selector when attempting to send a message using this
3701 selector to a receiver of type @code{id} or @code{Class}. When this flag
3702 is off (which is the default behavior), the compiler omits such warnings
3703 if any differences found are confined to types that share the same size
3706 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3707 @opindex Wundeclared-selector
3708 @opindex Wno-undeclared-selector
3709 Warn if a @code{@@selector(@dots{})} expression referring to an
3710 undeclared selector is found. A selector is considered undeclared if no
3711 method with that name has been declared before the
3712 @code{@@selector(@dots{})} expression, either explicitly in an
3713 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3714 an @code{@@implementation} section. This option always performs its
3715 checks as soon as a @code{@@selector(@dots{})} expression is found,
3716 while @option{-Wselector} only performs its checks in the final stage of
3717 compilation. This also enforces the coding style convention
3718 that methods and selectors must be declared before being used.
3720 @item -print-objc-runtime-info
3721 @opindex print-objc-runtime-info
3722 Generate C header describing the largest structure that is passed by
3727 @node Diagnostic Message Formatting Options
3728 @section Options to Control Diagnostic Messages Formatting
3729 @cindex options to control diagnostics formatting
3730 @cindex diagnostic messages
3731 @cindex message formatting
3733 Traditionally, diagnostic messages have been formatted irrespective of
3734 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3735 options described below
3736 to control the formatting algorithm for diagnostic messages,
3737 e.g.@: how many characters per line, how often source location
3738 information should be reported. Note that some language front ends may not
3739 honor these options.
3742 @item -fmessage-length=@var{n}
3743 @opindex fmessage-length
3744 Try to format error messages so that they fit on lines of about
3745 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3746 done; each error message appears on a single line. This is the
3747 default for all front ends.
3749 Note - this option also affects the display of the @samp{#error} and
3750 @samp{#warning} pre-processor directives, and the @samp{deprecated}
3751 function/type/variable attribute. It does not however affect the
3752 @samp{pragma GCC warning} and @samp{pragma GCC error} pragmas.
3754 @item -fdiagnostics-show-location=once
3755 @opindex fdiagnostics-show-location
3756 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3757 reporter to emit source location information @emph{once}; that is, in
3758 case the message is too long to fit on a single physical line and has to
3759 be wrapped, the source location won't be emitted (as prefix) again,
3760 over and over, in subsequent continuation lines. This is the default
3763 @item -fdiagnostics-show-location=every-line
3764 Only meaningful in line-wrapping mode. Instructs the diagnostic
3765 messages reporter to emit the same source location information (as
3766 prefix) for physical lines that result from the process of breaking
3767 a message which is too long to fit on a single line.
3769 @item -fdiagnostics-color[=@var{WHEN}]
3770 @itemx -fno-diagnostics-color
3771 @opindex fdiagnostics-color
3772 @cindex highlight, color
3773 @vindex GCC_COLORS @r{environment variable}
3774 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3775 or @samp{auto}. The default depends on how the compiler has been configured,
3776 it can be any of the above @var{WHEN} options or also @samp{never}
3777 if @env{GCC_COLORS} environment variable isn't present in the environment,
3778 and @samp{auto} otherwise.
3779 @samp{auto} means to use color only when the standard error is a terminal.
3780 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3781 aliases for @option{-fdiagnostics-color=always} and
3782 @option{-fdiagnostics-color=never}, respectively.
3784 The colors are defined by the environment variable @env{GCC_COLORS}.
3785 Its value is a colon-separated list of capabilities and Select Graphic
3786 Rendition (SGR) substrings. SGR commands are interpreted by the
3787 terminal or terminal emulator. (See the section in the documentation
3788 of your text terminal for permitted values and their meanings as
3789 character attributes.) These substring values are integers in decimal
3790 representation and can be concatenated with semicolons.
3791 Common values to concatenate include
3793 @samp{4} for underline,
3795 @samp{7} for inverse,
3796 @samp{39} for default foreground color,
3797 @samp{30} to @samp{37} for foreground colors,
3798 @samp{90} to @samp{97} for 16-color mode foreground colors,
3799 @samp{38;5;0} to @samp{38;5;255}
3800 for 88-color and 256-color modes foreground colors,
3801 @samp{49} for default background color,
3802 @samp{40} to @samp{47} for background colors,
3803 @samp{100} to @samp{107} for 16-color mode background colors,
3804 and @samp{48;5;0} to @samp{48;5;255}
3805 for 88-color and 256-color modes background colors.
3807 The default @env{GCC_COLORS} is
3809 error=01;31:warning=01;35:note=01;36:range1=32:range2=34:locus=01:\
3810 quote=01:fixit-insert=32:fixit-delete=31:\
3811 diff-filename=01:diff-hunk=32:diff-delete=31:diff-insert=32:\
3815 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3816 @samp{01;36} is bold cyan, @samp{32} is green, @samp{34} is blue,
3817 @samp{01} is bold, and @samp{31} is red.
3818 Setting @env{GCC_COLORS} to the empty string disables colors.
3819 Supported capabilities are as follows.
3823 @vindex error GCC_COLORS @r{capability}
3824 SGR substring for error: markers.
3827 @vindex warning GCC_COLORS @r{capability}
3828 SGR substring for warning: markers.
3831 @vindex note GCC_COLORS @r{capability}
3832 SGR substring for note: markers.
3835 @vindex range1 GCC_COLORS @r{capability}
3836 SGR substring for first additional range.
3839 @vindex range2 GCC_COLORS @r{capability}
3840 SGR substring for second additional range.
3843 @vindex locus GCC_COLORS @r{capability}
3844 SGR substring for location information, @samp{file:line} or
3845 @samp{file:line:column} etc.
3848 @vindex quote GCC_COLORS @r{capability}
3849 SGR substring for information printed within quotes.
3852 @vindex fixit-insert GCC_COLORS @r{capability}
3853 SGR substring for fix-it hints suggesting text to
3854 be inserted or replaced.
3857 @vindex fixit-delete GCC_COLORS @r{capability}
3858 SGR substring for fix-it hints suggesting text to
3861 @item diff-filename=
3862 @vindex diff-filename GCC_COLORS @r{capability}
3863 SGR substring for filename headers within generated patches.
3866 @vindex diff-hunk GCC_COLORS @r{capability}
3867 SGR substring for the starts of hunks within generated patches.
3870 @vindex diff-delete GCC_COLORS @r{capability}
3871 SGR substring for deleted lines within generated patches.
3874 @vindex diff-insert GCC_COLORS @r{capability}
3875 SGR substring for inserted lines within generated patches.
3878 @vindex type-diff GCC_COLORS @r{capability}
3879 SGR substring for highlighting mismatching types within template
3880 arguments in the C++ frontend.
3883 @item -fno-diagnostics-show-option
3884 @opindex fno-diagnostics-show-option
3885 @opindex fdiagnostics-show-option
3886 By default, each diagnostic emitted includes text indicating the
3887 command-line option that directly controls the diagnostic (if such an
3888 option is known to the diagnostic machinery). Specifying the
3889 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3891 @item -fno-diagnostics-show-caret
3892 @opindex fno-diagnostics-show-caret
3893 @opindex fdiagnostics-show-caret
3894 By default, each diagnostic emitted includes the original source line
3895 and a caret @samp{^} indicating the column. This option suppresses this
3896 information. The source line is truncated to @var{n} characters, if
3897 the @option{-fmessage-length=n} option is given. When the output is done
3898 to the terminal, the width is limited to the width given by the
3899 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3901 @item -fno-diagnostics-show-labels
3902 @opindex fno-diagnostics-show-labels
3903 @opindex fdiagnostics-show-labels
3904 By default, when printing source code (via @option{-fdiagnostics-show-caret}),
3905 diagnostics can label ranges of source code with pertinent information, such
3906 as the types of expressions:
3909 printf ("foo %s bar", long_i + long_j);
3915 This option suppresses the printing of these labels (in the example above,
3916 the vertical bars and the ``char *'' and ``long int'' text).
3918 @item -fno-diagnostics-show-line-numbers
3919 @opindex fno-diagnostics-show-line-numbers
3920 @opindex fdiagnostics-show-line-numbers
3921 By default, when printing source code (via @option{-fdiagnostics-show-caret}),
3922 a left margin is printed, showing line numbers. This option suppresses this
3925 @item -fdiagnostics-minimum-margin-width=@var{width}
3926 @opindex fdiagnostics-minimum-margin-width
3927 This option controls the minimum width of the left margin printed by
3928 @option{-fdiagnostics-show-line-numbers}. It defaults to 6.
3930 @item -fdiagnostics-parseable-fixits
3931 @opindex fdiagnostics-parseable-fixits
3932 Emit fix-it hints in a machine-parseable format, suitable for consumption
3933 by IDEs. For each fix-it, a line will be printed after the relevant
3934 diagnostic, starting with the string ``fix-it:''. For example:
3937 fix-it:"test.c":@{45:3-45:21@}:"gtk_widget_show_all"
3940 The location is expressed as a half-open range, expressed as a count of
3941 bytes, starting at byte 1 for the initial column. In the above example,
3942 bytes 3 through 20 of line 45 of ``test.c'' are to be replaced with the
3946 00000000011111111112222222222
3947 12345678901234567890123456789
3948 gtk_widget_showall (dlg);
3953 The filename and replacement string escape backslash as ``\\", tab as ``\t'',
3954 newline as ``\n'', double quotes as ``\"'', non-printable characters as octal
3955 (e.g. vertical tab as ``\013'').
3957 An empty replacement string indicates that the given range is to be removed.
3958 An empty range (e.g. ``45:3-45:3'') indicates that the string is to
3959 be inserted at the given position.
3961 @item -fdiagnostics-generate-patch
3962 @opindex fdiagnostics-generate-patch
3963 Print fix-it hints to stderr in unified diff format, after any diagnostics
3964 are printed. For example:
3971 void show_cb(GtkDialog *dlg)
3973 - gtk_widget_showall(dlg);
3974 + gtk_widget_show_all(dlg);
3979 The diff may or may not be colorized, following the same rules
3980 as for diagnostics (see @option{-fdiagnostics-color}).
3982 @item -fdiagnostics-show-template-tree
3983 @opindex fdiagnostics-show-template-tree
3985 In the C++ frontend, when printing diagnostics showing mismatching
3986 template types, such as:
3989 could not convert 'std::map<int, std::vector<double> >()'
3990 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
3993 the @option{-fdiagnostics-show-template-tree} flag enables printing a
3994 tree-like structure showing the common and differing parts of the types,
4004 The parts that differ are highlighted with color (``double'' and
4005 ``float'' in this case).
4007 @item -fno-elide-type
4008 @opindex fno-elide-type
4009 @opindex felide-type
4010 By default when the C++ frontend prints diagnostics showing mismatching
4011 template types, common parts of the types are printed as ``[...]'' to
4012 simplify the error message. For example:
4015 could not convert 'std::map<int, std::vector<double> >()'
4016 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
4019 Specifying the @option{-fno-elide-type} flag suppresses that behavior.
4020 This flag also affects the output of the
4021 @option{-fdiagnostics-show-template-tree} flag.
4023 @item -fno-show-column
4024 @opindex fno-show-column
4025 @opindex fshow-column
4026 Do not print column numbers in diagnostics. This may be necessary if
4027 diagnostics are being scanned by a program that does not understand the
4028 column numbers, such as @command{dejagnu}.
4030 @item -fdiagnostics-format=@var{FORMAT}
4031 @opindex fdiagnostics-format
4032 Select a different format for printing diagnostics.
4033 @var{FORMAT} is @samp{text} or @samp{json}.
4034 The default is @samp{text}.
4036 The @samp{json} format consists of a top-level JSON array containing JSON
4037 objects representing the diagnostics.
4039 The JSON is emitted as one line, without formatting; the examples below
4040 have been formatted for clarity.
4042 Diagnostics can have child diagnostics. For example, this error and note:
4045 misleading-indentation.c:15:3: warning: this 'if' clause does not
4046 guard... [-Wmisleading-indentation]
4049 misleading-indentation.c:17:5: note: ...this statement, but the latter
4050 is misleadingly indented as if it were guarded by the 'if'
4056 might be printed in JSON form (after formatting) like this:
4066 "file": "misleading-indentation.c",
4071 "file": "misleading-indentation.c",
4076 "message": "this \u2018if\u2019 clause does not guard...",
4077 "option": "-Wmisleading-indentation",
4085 "file": "misleading-indentation.c",
4090 "message": "...this statement, but the latter is @dots{}"
4099 where the @code{note} is a child of the @code{warning}.
4101 A diagnostic has a @code{kind}. If this is @code{warning}, then there is
4102 an @code{option} key describing the command-line option controlling the
4105 A diagnostic can contain zero or more locations. Each location has up
4106 to three positions within it: a @code{caret} position and optional
4107 @code{start} and @code{finish} positions. A location can also have
4108 an optional @code{label} string. For example, this error:
4111 bad-binary-ops.c:64:23: error: invalid operands to binary + (have 'S' @{aka
4112 'struct s'@} and 'T' @{aka 'struct t'@})
4113 64 | return callee_4a () + callee_4b ();
4114 | ~~~~~~~~~~~~ ^ ~~~~~~~~~~~~
4116 | | T @{aka struct t@}
4117 | S @{aka struct s@}
4121 has three locations. Its primary location is at the ``+'' token at column
4122 23. It has two secondary locations, describing the left and right-hand sides
4123 of the expression, which have labels. It might be printed in JSON form as:
4132 "column": 23, "file": "bad-binary-ops.c", "line": 64
4137 "column": 10, "file": "bad-binary-ops.c", "line": 64
4140 "column": 21, "file": "bad-binary-ops.c", "line": 64
4142 "label": "S @{aka struct s@}"
4146 "column": 25, "file": "bad-binary-ops.c", "line": 64
4149 "column": 36, "file": "bad-binary-ops.c", "line": 64
4151 "label": "T @{aka struct t@}"
4154 "message": "invalid operands to binary + @dots{}"
4158 If a diagnostic contains fix-it hints, it has a @code{fixits} array,
4159 consisting of half-open intervals, similar to the output of
4160 @option{-fdiagnostics-parseable-fixits}. For example, this diagnostic
4161 with a replacement fix-it hint:
4164 demo.c:8:15: error: 'struct s' has no member named 'colour'; did you
4166 8 | return ptr->colour;
4172 might be printed in JSON form as:
4207 "message": "\u2018struct s\u2019 has no member named @dots{}"
4212 where the fix-it hint suggests replacing the text from @code{start} up
4213 to but not including @code{next} with @code{string}'s value. Deletions
4214 are expressed via an empty value for @code{string}, insertions by
4215 having @code{start} equal @code{next}.
4219 @node Warning Options
4220 @section Options to Request or Suppress Warnings
4221 @cindex options to control warnings
4222 @cindex warning messages
4223 @cindex messages, warning
4224 @cindex suppressing warnings
4226 Warnings are diagnostic messages that report constructions that
4227 are not inherently erroneous but that are risky or suggest there
4228 may have been an error.
4230 The following language-independent options do not enable specific
4231 warnings but control the kinds of diagnostics produced by GCC@.
4234 @cindex syntax checking
4236 @opindex fsyntax-only
4237 Check the code for syntax errors, but don't do anything beyond that.
4239 @item -fmax-errors=@var{n}
4240 @opindex fmax-errors
4241 Limits the maximum number of error messages to @var{n}, at which point
4242 GCC bails out rather than attempting to continue processing the source
4243 code. If @var{n} is 0 (the default), there is no limit on the number
4244 of error messages produced. If @option{-Wfatal-errors} is also
4245 specified, then @option{-Wfatal-errors} takes precedence over this
4250 Inhibit all warning messages.
4255 Make all warnings into errors.
4260 Make the specified warning into an error. The specifier for a warning
4261 is appended; for example @option{-Werror=switch} turns the warnings
4262 controlled by @option{-Wswitch} into errors. This switch takes a
4263 negative form, to be used to negate @option{-Werror} for specific
4264 warnings; for example @option{-Wno-error=switch} makes
4265 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
4268 The warning message for each controllable warning includes the
4269 option that controls the warning. That option can then be used with
4270 @option{-Werror=} and @option{-Wno-error=} as described above.
4271 (Printing of the option in the warning message can be disabled using the
4272 @option{-fno-diagnostics-show-option} flag.)
4274 Note that specifying @option{-Werror=}@var{foo} automatically implies
4275 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
4278 @item -Wfatal-errors
4279 @opindex Wfatal-errors
4280 @opindex Wno-fatal-errors
4281 This option causes the compiler to abort compilation on the first error
4282 occurred rather than trying to keep going and printing further error
4287 You can request many specific warnings with options beginning with
4288 @samp{-W}, for example @option{-Wimplicit} to request warnings on
4289 implicit declarations. Each of these specific warning options also
4290 has a negative form beginning @samp{-Wno-} to turn off warnings; for
4291 example, @option{-Wno-implicit}. This manual lists only one of the
4292 two forms, whichever is not the default. For further
4293 language-specific options also refer to @ref{C++ Dialect Options} and
4294 @ref{Objective-C and Objective-C++ Dialect Options}.
4296 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
4297 options, such as @option{-Wunused}, which may turn on further options,
4298 such as @option{-Wunused-value}. The combined effect of positive and
4299 negative forms is that more specific options have priority over less
4300 specific ones, independently of their position in the command-line. For
4301 options of the same specificity, the last one takes effect. Options
4302 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
4303 as if they appeared at the end of the command-line.
4305 When an unrecognized warning option is requested (e.g.,
4306 @option{-Wunknown-warning}), GCC emits a diagnostic stating
4307 that the option is not recognized. However, if the @option{-Wno-} form
4308 is used, the behavior is slightly different: no diagnostic is
4309 produced for @option{-Wno-unknown-warning} unless other diagnostics
4310 are being produced. This allows the use of new @option{-Wno-} options
4311 with old compilers, but if something goes wrong, the compiler
4312 warns that an unrecognized option is present.
4319 @opindex Wno-pedantic
4320 Issue all the warnings demanded by strict ISO C and ISO C++;
4321 reject all programs that use forbidden extensions, and some other
4322 programs that do not follow ISO C and ISO C++. For ISO C, follows the
4323 version of the ISO C standard specified by any @option{-std} option used.
4325 Valid ISO C and ISO C++ programs should compile properly with or without
4326 this option (though a rare few require @option{-ansi} or a
4327 @option{-std} option specifying the required version of ISO C)@. However,
4328 without this option, certain GNU extensions and traditional C and C++
4329 features are supported as well. With this option, they are rejected.
4331 @option{-Wpedantic} does not cause warning messages for use of the
4332 alternate keywords whose names begin and end with @samp{__}. Pedantic
4333 warnings are also disabled in the expression that follows
4334 @code{__extension__}. However, only system header files should use
4335 these escape routes; application programs should avoid them.
4336 @xref{Alternate Keywords}.
4338 Some users try to use @option{-Wpedantic} to check programs for strict ISO
4339 C conformance. They soon find that it does not do quite what they want:
4340 it finds some non-ISO practices, but not all---only those for which
4341 ISO C @emph{requires} a diagnostic, and some others for which
4342 diagnostics have been added.
4344 A feature to report any failure to conform to ISO C might be useful in
4345 some instances, but would require considerable additional work and would
4346 be quite different from @option{-Wpedantic}. We don't have plans to
4347 support such a feature in the near future.
4349 Where the standard specified with @option{-std} represents a GNU
4350 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
4351 corresponding @dfn{base standard}, the version of ISO C on which the GNU
4352 extended dialect is based. Warnings from @option{-Wpedantic} are given
4353 where they are required by the base standard. (It does not make sense
4354 for such warnings to be given only for features not in the specified GNU
4355 C dialect, since by definition the GNU dialects of C include all
4356 features the compiler supports with the given option, and there would be
4357 nothing to warn about.)
4359 @item -pedantic-errors
4360 @opindex pedantic-errors
4361 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
4362 requires a diagnostic, in some cases where there is undefined behavior
4363 at compile-time and in some other cases that do not prevent compilation
4364 of programs that are valid according to the standard. This is not
4365 equivalent to @option{-Werror=pedantic}, since there are errors enabled
4366 by this option and not enabled by the latter and vice versa.
4371 This enables all the warnings about constructions that some users
4372 consider questionable, and that are easy to avoid (or modify to
4373 prevent the warning), even in conjunction with macros. This also
4374 enables some language-specific warnings described in @ref{C++ Dialect
4375 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
4377 @option{-Wall} turns on the following warning flags:
4379 @gccoptlist{-Waddress @gol
4380 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
4382 -Wbool-operation @gol
4383 -Wc++11-compat -Wc++14-compat @gol
4384 -Wcatch-value @r{(C++ and Objective-C++ only)} @gol
4385 -Wchar-subscripts @gol
4387 -Wduplicate-decl-specifier @r{(C and Objective-C only)} @gol
4388 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
4390 -Wint-in-bool-context @gol
4391 -Wimplicit @r{(C and Objective-C only)} @gol
4392 -Wimplicit-int @r{(C and Objective-C only)} @gol
4393 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
4394 -Winit-self @r{(only for C++)} @gol
4395 -Wlogical-not-parentheses @gol
4396 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
4397 -Wmaybe-uninitialized @gol
4398 -Wmemset-elt-size @gol
4399 -Wmemset-transposed-args @gol
4400 -Wmisleading-indentation @r{(only for C/C++)} @gol
4401 -Wmissing-attributes @gol
4402 -Wmissing-braces @r{(only for C/ObjC)} @gol
4403 -Wmultistatement-macros @gol
4404 -Wnarrowing @r{(only for C++)} @gol
4406 -Wnonnull-compare @gol
4409 -Wpessimizing-move @r{(only for C++)} @gol
4414 -Wsequence-point @gol
4415 -Wsign-compare @r{(only in C++)} @gol
4416 -Wsizeof-pointer-div @gol
4417 -Wsizeof-pointer-memaccess @gol
4418 -Wstrict-aliasing @gol
4419 -Wstrict-overflow=1 @gol
4421 -Wtautological-compare @gol
4423 -Wuninitialized @gol
4424 -Wunknown-pragmas @gol
4425 -Wunused-function @gol
4428 -Wunused-variable @gol
4429 -Wvolatile-register-var}
4431 Note that some warning flags are not implied by @option{-Wall}. Some of
4432 them warn about constructions that users generally do not consider
4433 questionable, but which occasionally you might wish to check for;
4434 others warn about constructions that are necessary or hard to avoid in
4435 some cases, and there is no simple way to modify the code to suppress
4436 the warning. Some of them are enabled by @option{-Wextra} but many of
4437 them must be enabled individually.
4443 This enables some extra warning flags that are not enabled by
4444 @option{-Wall}. (This option used to be called @option{-W}. The older
4445 name is still supported, but the newer name is more descriptive.)
4447 @gccoptlist{-Wclobbered @gol
4448 -Wcast-function-type @gol
4449 -Wdeprecated-copy @r{(C++ only)} @gol
4451 -Wignored-qualifiers @gol
4452 -Wimplicit-fallthrough=3 @gol
4453 -Wmissing-field-initializers @gol
4454 -Wmissing-parameter-type @r{(C only)} @gol
4455 -Wold-style-declaration @r{(C only)} @gol
4456 -Woverride-init @gol
4457 -Wsign-compare @r{(C only)} @gol
4458 -Wredundant-move @r{(only for C++)} @gol
4460 -Wuninitialized @gol
4461 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
4462 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
4463 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)}}
4466 The option @option{-Wextra} also prints warning messages for the
4472 A pointer is compared against integer zero with @code{<}, @code{<=},
4473 @code{>}, or @code{>=}.
4476 (C++ only) An enumerator and a non-enumerator both appear in a
4477 conditional expression.
4480 (C++ only) Ambiguous virtual bases.
4483 (C++ only) Subscripting an array that has been declared @code{register}.
4486 (C++ only) Taking the address of a variable that has been declared
4490 (C++ only) A base class is not initialized in the copy constructor
4495 @item -Wchar-subscripts
4496 @opindex Wchar-subscripts
4497 @opindex Wno-char-subscripts
4498 Warn if an array subscript has type @code{char}. This is a common cause
4499 of error, as programmers often forget that this type is signed on some
4501 This warning is enabled by @option{-Wall}.
4503 @item -Wno-coverage-mismatch
4504 @opindex Wno-coverage-mismatch
4505 @opindex Wcoverage-mismatch
4506 Warn if feedback profiles do not match when using the
4507 @option{-fprofile-use} option.
4508 If a source file is changed between compiling with @option{-fprofile-generate}
4509 and with @option{-fprofile-use}, the files with the profile feedback can fail
4510 to match the source file and GCC cannot use the profile feedback
4511 information. By default, this warning is enabled and is treated as an
4512 error. @option{-Wno-coverage-mismatch} can be used to disable the
4513 warning or @option{-Wno-error=coverage-mismatch} can be used to
4514 disable the error. Disabling the error for this warning can result in
4515 poorly optimized code and is useful only in the
4516 case of very minor changes such as bug fixes to an existing code-base.
4517 Completely disabling the warning is not recommended.
4520 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
4522 Suppress warning messages emitted by @code{#warning} directives.
4524 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
4525 @opindex Wdouble-promotion
4526 @opindex Wno-double-promotion
4527 Give a warning when a value of type @code{float} is implicitly
4528 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
4529 floating-point unit implement @code{float} in hardware, but emulate
4530 @code{double} in software. On such a machine, doing computations
4531 using @code{double} values is much more expensive because of the
4532 overhead required for software emulation.
4534 It is easy to accidentally do computations with @code{double} because
4535 floating-point literals are implicitly of type @code{double}. For
4539 float area(float radius)
4541 return 3.14159 * radius * radius;
4545 the compiler performs the entire computation with @code{double}
4546 because the floating-point literal is a @code{double}.
4548 @item -Wduplicate-decl-specifier @r{(C and Objective-C only)}
4549 @opindex Wduplicate-decl-specifier
4550 @opindex Wno-duplicate-decl-specifier
4551 Warn if a declaration has duplicate @code{const}, @code{volatile},
4552 @code{restrict} or @code{_Atomic} specifier. This warning is enabled by
4556 @itemx -Wformat=@var{n}
4559 @opindex ffreestanding
4560 @opindex fno-builtin
4562 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
4563 the arguments supplied have types appropriate to the format string
4564 specified, and that the conversions specified in the format string make
4565 sense. This includes standard functions, and others specified by format
4566 attributes (@pxref{Function Attributes}), in the @code{printf},
4567 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
4568 not in the C standard) families (or other target-specific families).
4569 Which functions are checked without format attributes having been
4570 specified depends on the standard version selected, and such checks of
4571 functions without the attribute specified are disabled by
4572 @option{-ffreestanding} or @option{-fno-builtin}.
4574 The formats are checked against the format features supported by GNU
4575 libc version 2.2. These include all ISO C90 and C99 features, as well
4576 as features from the Single Unix Specification and some BSD and GNU
4577 extensions. Other library implementations may not support all these
4578 features; GCC does not support warning about features that go beyond a
4579 particular library's limitations. However, if @option{-Wpedantic} is used
4580 with @option{-Wformat}, warnings are given about format features not
4581 in the selected standard version (but not for @code{strfmon} formats,
4582 since those are not in any version of the C standard). @xref{C Dialect
4583 Options,,Options Controlling C Dialect}.
4590 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
4591 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
4592 @option{-Wformat} also checks for null format arguments for several
4593 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
4594 aspects of this level of format checking can be disabled by the
4595 options: @option{-Wno-format-contains-nul},
4596 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
4597 @option{-Wformat} is enabled by @option{-Wall}.
4599 @item -Wno-format-contains-nul
4600 @opindex Wno-format-contains-nul
4601 @opindex Wformat-contains-nul
4602 If @option{-Wformat} is specified, do not warn about format strings that
4605 @item -Wno-format-extra-args
4606 @opindex Wno-format-extra-args
4607 @opindex Wformat-extra-args
4608 If @option{-Wformat} is specified, do not warn about excess arguments to a
4609 @code{printf} or @code{scanf} format function. The C standard specifies
4610 that such arguments are ignored.
4612 Where the unused arguments lie between used arguments that are
4613 specified with @samp{$} operand number specifications, normally
4614 warnings are still given, since the implementation could not know what
4615 type to pass to @code{va_arg} to skip the unused arguments. However,
4616 in the case of @code{scanf} formats, this option suppresses the
4617 warning if the unused arguments are all pointers, since the Single
4618 Unix Specification says that such unused arguments are allowed.
4620 @item -Wformat-overflow
4621 @itemx -Wformat-overflow=@var{level}
4622 @opindex Wformat-overflow
4623 @opindex Wno-format-overflow
4624 Warn about calls to formatted input/output functions such as @code{sprintf}
4625 and @code{vsprintf} that might overflow the destination buffer. When the
4626 exact number of bytes written by a format directive cannot be determined
4627 at compile-time it is estimated based on heuristics that depend on the
4628 @var{level} argument and on optimization. While enabling optimization
4629 will in most cases improve the accuracy of the warning, it may also
4630 result in false positives.
4633 @item -Wformat-overflow
4634 @itemx -Wformat-overflow=1
4635 @opindex Wformat-overflow
4636 @opindex Wno-format-overflow
4637 Level @var{1} of @option{-Wformat-overflow} enabled by @option{-Wformat}
4638 employs a conservative approach that warns only about calls that most
4639 likely overflow the buffer. At this level, numeric arguments to format
4640 directives with unknown values are assumed to have the value of one, and
4641 strings of unknown length to be empty. Numeric arguments that are known
4642 to be bounded to a subrange of their type, or string arguments whose output
4643 is bounded either by their directive's precision or by a finite set of
4644 string literals, are assumed to take on the value within the range that
4645 results in the most bytes on output. For example, the call to @code{sprintf}
4646 below is diagnosed because even with both @var{a} and @var{b} equal to zero,
4647 the terminating NUL character (@code{'\0'}) appended by the function
4648 to the destination buffer will be written past its end. Increasing
4649 the size of the buffer by a single byte is sufficient to avoid the
4650 warning, though it may not be sufficient to avoid the overflow.
4653 void f (int a, int b)
4656 sprintf (buf, "a = %i, b = %i\n", a, b);
4660 @item -Wformat-overflow=2
4661 Level @var{2} warns also about calls that might overflow the destination
4662 buffer given an argument of sufficient length or magnitude. At level
4663 @var{2}, unknown numeric arguments are assumed to have the minimum
4664 representable value for signed types with a precision greater than 1, and
4665 the maximum representable value otherwise. Unknown string arguments whose
4666 length cannot be assumed to be bounded either by the directive's precision,
4667 or by a finite set of string literals they may evaluate to, or the character
4668 array they may point to, are assumed to be 1 character long.
4670 At level @var{2}, the call in the example above is again diagnosed, but
4671 this time because with @var{a} equal to a 32-bit @code{INT_MIN} the first
4672 @code{%i} directive will write some of its digits beyond the end of
4673 the destination buffer. To make the call safe regardless of the values
4674 of the two variables, the size of the destination buffer must be increased
4675 to at least 34 bytes. GCC includes the minimum size of the buffer in
4676 an informational note following the warning.
4678 An alternative to increasing the size of the destination buffer is to
4679 constrain the range of formatted values. The maximum length of string
4680 arguments can be bounded by specifying the precision in the format
4681 directive. When numeric arguments of format directives can be assumed
4682 to be bounded by less than the precision of their type, choosing
4683 an appropriate length modifier to the format specifier will reduce
4684 the required buffer size. For example, if @var{a} and @var{b} in the
4685 example above can be assumed to be within the precision of
4686 the @code{short int} type then using either the @code{%hi} format
4687 directive or casting the argument to @code{short} reduces the maximum
4688 required size of the buffer to 24 bytes.
4691 void f (int a, int b)
4694 sprintf (buf, "a = %hi, b = %i\n", a, (short)b);
4699 @item -Wno-format-zero-length
4700 @opindex Wno-format-zero-length
4701 @opindex Wformat-zero-length
4702 If @option{-Wformat} is specified, do not warn about zero-length formats.
4703 The C standard specifies that zero-length formats are allowed.
4708 Enable @option{-Wformat} plus additional format checks. Currently
4709 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
4712 @item -Wformat-nonliteral
4713 @opindex Wformat-nonliteral
4714 @opindex Wno-format-nonliteral
4715 If @option{-Wformat} is specified, also warn if the format string is not a
4716 string literal and so cannot be checked, unless the format function
4717 takes its format arguments as a @code{va_list}.
4719 @item -Wformat-security
4720 @opindex Wformat-security
4721 @opindex Wno-format-security
4722 If @option{-Wformat} is specified, also warn about uses of format
4723 functions that represent possible security problems. At present, this
4724 warns about calls to @code{printf} and @code{scanf} functions where the
4725 format string is not a string literal and there are no format arguments,
4726 as in @code{printf (foo);}. This may be a security hole if the format
4727 string came from untrusted input and contains @samp{%n}. (This is
4728 currently a subset of what @option{-Wformat-nonliteral} warns about, but
4729 in future warnings may be added to @option{-Wformat-security} that are not
4730 included in @option{-Wformat-nonliteral}.)
4732 @item -Wformat-signedness
4733 @opindex Wformat-signedness
4734 @opindex Wno-format-signedness
4735 If @option{-Wformat} is specified, also warn if the format string
4736 requires an unsigned argument and the argument is signed and vice versa.
4738 @item -Wformat-truncation
4739 @itemx -Wformat-truncation=@var{level}
4740 @opindex Wformat-truncation
4741 @opindex Wno-format-truncation
4742 Warn about calls to formatted input/output functions such as @code{snprintf}
4743 and @code{vsnprintf} that might result in output truncation. When the exact
4744 number of bytes written by a format directive cannot be determined at
4745 compile-time it is estimated based on heuristics that depend on
4746 the @var{level} argument and on optimization. While enabling optimization
4747 will in most cases improve the accuracy of the warning, it may also result
4748 in false positives. Except as noted otherwise, the option uses the same
4749 logic @option{-Wformat-overflow}.
4752 @item -Wformat-truncation
4753 @itemx -Wformat-truncation=1
4754 @opindex Wformat-truncation
4755 @opindex Wno-format-truncation
4756 Level @var{1} of @option{-Wformat-truncation} enabled by @option{-Wformat}
4757 employs a conservative approach that warns only about calls to bounded
4758 functions whose return value is unused and that will most likely result
4759 in output truncation.
4761 @item -Wformat-truncation=2
4762 Level @var{2} warns also about calls to bounded functions whose return
4763 value is used and that might result in truncation given an argument of
4764 sufficient length or magnitude.
4768 @opindex Wformat-y2k
4769 @opindex Wno-format-y2k
4770 If @option{-Wformat} is specified, also warn about @code{strftime}
4771 formats that may yield only a two-digit year.
4776 @opindex Wno-nonnull
4777 Warn about passing a null pointer for arguments marked as
4778 requiring a non-null value by the @code{nonnull} function attribute.
4780 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
4781 can be disabled with the @option{-Wno-nonnull} option.
4783 @item -Wnonnull-compare
4784 @opindex Wnonnull-compare
4785 @opindex Wno-nonnull-compare
4786 Warn when comparing an argument marked with the @code{nonnull}
4787 function attribute against null inside the function.
4789 @option{-Wnonnull-compare} is included in @option{-Wall}. It
4790 can be disabled with the @option{-Wno-nonnull-compare} option.
4792 @item -Wnull-dereference
4793 @opindex Wnull-dereference
4794 @opindex Wno-null-dereference
4795 Warn if the compiler detects paths that trigger erroneous or
4796 undefined behavior due to dereferencing a null pointer. This option
4797 is only active when @option{-fdelete-null-pointer-checks} is active,
4798 which is enabled by optimizations in most targets. The precision of
4799 the warnings depends on the optimization options used.
4801 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
4803 @opindex Wno-init-self
4804 Warn about uninitialized variables that are initialized with themselves.
4805 Note this option can only be used with the @option{-Wuninitialized} option.
4807 For example, GCC warns about @code{i} being uninitialized in the
4808 following snippet only when @option{-Winit-self} has been specified:
4819 This warning is enabled by @option{-Wall} in C++.
4821 @item -Wimplicit-int @r{(C and Objective-C only)}
4822 @opindex Wimplicit-int
4823 @opindex Wno-implicit-int
4824 Warn when a declaration does not specify a type.
4825 This warning is enabled by @option{-Wall}.
4827 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
4828 @opindex Wimplicit-function-declaration
4829 @opindex Wno-implicit-function-declaration
4830 Give a warning whenever a function is used before being declared. In
4831 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
4832 enabled by default and it is made into an error by
4833 @option{-pedantic-errors}. This warning is also enabled by
4836 @item -Wimplicit @r{(C and Objective-C only)}
4838 @opindex Wno-implicit
4839 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
4840 This warning is enabled by @option{-Wall}.
4842 @item -Wimplicit-fallthrough
4843 @opindex Wimplicit-fallthrough
4844 @opindex Wno-implicit-fallthrough
4845 @option{-Wimplicit-fallthrough} is the same as @option{-Wimplicit-fallthrough=3}
4846 and @option{-Wno-implicit-fallthrough} is the same as
4847 @option{-Wimplicit-fallthrough=0}.
4849 @item -Wimplicit-fallthrough=@var{n}
4850 @opindex Wimplicit-fallthrough=
4851 Warn when a switch case falls through. For example:
4869 This warning does not warn when the last statement of a case cannot
4870 fall through, e.g. when there is a return statement or a call to function
4871 declared with the noreturn attribute. @option{-Wimplicit-fallthrough=}
4872 also takes into account control flow statements, such as ifs, and only
4873 warns when appropriate. E.g.@:
4883 @} else if (i < 1) @{
4893 Since there are occasions where a switch case fall through is desirable,
4894 GCC provides an attribute, @code{__attribute__ ((fallthrough))}, that is
4895 to be used along with a null statement to suppress this warning that
4896 would normally occur:
4904 __attribute__ ((fallthrough));
4911 C++17 provides a standard way to suppress the @option{-Wimplicit-fallthrough}
4912 warning using @code{[[fallthrough]];} instead of the GNU attribute. In C++11
4913 or C++14 users can use @code{[[gnu::fallthrough]];}, which is a GNU extension.
4914 Instead of these attributes, it is also possible to add a fallthrough comment
4915 to silence the warning. The whole body of the C or C++ style comment should
4916 match the given regular expressions listed below. The option argument @var{n}
4917 specifies what kind of comments are accepted:
4921 @item @option{-Wimplicit-fallthrough=0} disables the warning altogether.
4923 @item @option{-Wimplicit-fallthrough=1} matches @code{.*} regular
4924 expression, any comment is used as fallthrough comment.
4926 @item @option{-Wimplicit-fallthrough=2} case insensitively matches
4927 @code{.*falls?[ \t-]*thr(ough|u).*} regular expression.
4929 @item @option{-Wimplicit-fallthrough=3} case sensitively matches one of the
4930 following regular expressions:
4934 @item @code{-fallthrough}
4936 @item @code{@@fallthrough@@}
4938 @item @code{lint -fallthrough[ \t]*}
4940 @item @code{[ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?@*FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)?}
4942 @item @code{[ \t.!]*(Else,? |Intentional(ly)? )?@*Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4944 @item @code{[ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?@*fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4948 @item @option{-Wimplicit-fallthrough=4} case sensitively matches one of the
4949 following regular expressions:
4953 @item @code{-fallthrough}
4955 @item @code{@@fallthrough@@}
4957 @item @code{lint -fallthrough[ \t]*}
4959 @item @code{[ \t]*FALLTHR(OUGH|U)[ \t]*}
4963 @item @option{-Wimplicit-fallthrough=5} doesn't recognize any comments as
4964 fallthrough comments, only attributes disable the warning.
4968 The comment needs to be followed after optional whitespace and other comments
4969 by @code{case} or @code{default} keywords or by a user label that precedes some
4970 @code{case} or @code{default} label.
4985 The @option{-Wimplicit-fallthrough=3} warning is enabled by @option{-Wextra}.
4987 @item -Wif-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
4988 @opindex Wif-not-aligned
4989 @opindex Wno-if-not-aligned
4990 Control if warning triggered by the @code{warn_if_not_aligned} attribute
4991 should be issued. This is enabled by default.
4992 Use @option{-Wno-if-not-aligned} to disable it.
4994 @item -Wignored-qualifiers @r{(C and C++ only)}
4995 @opindex Wignored-qualifiers
4996 @opindex Wno-ignored-qualifiers
4997 Warn if the return type of a function has a type qualifier
4998 such as @code{const}. For ISO C such a type qualifier has no effect,
4999 since the value returned by a function is not an lvalue.
5000 For C++, the warning is only emitted for scalar types or @code{void}.
5001 ISO C prohibits qualified @code{void} return types on function
5002 definitions, so such return types always receive a warning
5003 even without this option.
5005 This warning is also enabled by @option{-Wextra}.
5007 @item -Wignored-attributes @r{(C and C++ only)}
5008 @opindex Wignored-attributes
5009 @opindex Wno-ignored-attributes
5010 Warn when an attribute is ignored. This is different from the
5011 @option{-Wattributes} option in that it warns whenever the compiler decides
5012 to drop an attribute, not that the attribute is either unknown, used in a
5013 wrong place, etc. This warning is enabled by default.
5018 Warn if the type of @code{main} is suspicious. @code{main} should be
5019 a function with external linkage, returning int, taking either zero
5020 arguments, two, or three arguments of appropriate types. This warning
5021 is enabled by default in C++ and is enabled by either @option{-Wall}
5022 or @option{-Wpedantic}.
5024 @item -Wmisleading-indentation @r{(C and C++ only)}
5025 @opindex Wmisleading-indentation
5026 @opindex Wno-misleading-indentation
5027 Warn when the indentation of the code does not reflect the block structure.
5028 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
5029 @code{for} clauses with a guarded statement that does not use braces,
5030 followed by an unguarded statement with the same indentation.
5032 In the following example, the call to ``bar'' is misleadingly indented as
5033 if it were guarded by the ``if'' conditional.
5036 if (some_condition ())
5038 bar (); /* Gotcha: this is not guarded by the "if". */
5041 In the case of mixed tabs and spaces, the warning uses the
5042 @option{-ftabstop=} option to determine if the statements line up
5045 The warning is not issued for code involving multiline preprocessor logic
5046 such as the following example.
5051 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
5057 The warning is not issued after a @code{#line} directive, since this
5058 typically indicates autogenerated code, and no assumptions can be made
5059 about the layout of the file that the directive references.
5061 This warning is enabled by @option{-Wall} in C and C++.
5063 @item -Wmissing-attributes
5064 @opindex Wmissing-attributes
5065 @opindex Wno-missing-attributes
5066 Warn when a declaration of a function is missing one or more attributes
5067 that a related function is declared with and whose absence may adversely
5068 affect the correctness or efficiency of generated code. For example,
5069 the warning is issued for declarations of aliases that use attributes
5070 to specify less restrictive requirements than those of their targets.
5071 This typically represents a potential optimization opportunity.
5072 By contrast, the @option{-Wattribute-alias=2} option controls warnings
5073 issued when the alias is more restrictive than the target, which could
5074 lead to incorrect code generation.
5075 Attributes considered include @code{alloc_align}, @code{alloc_size},
5076 @code{cold}, @code{const}, @code{hot}, @code{leaf}, @code{malloc},
5077 @code{nonnull}, @code{noreturn}, @code{nothrow}, @code{pure},
5078 @code{returns_nonnull}, and @code{returns_twice}.
5080 In C++, the warning is issued when an explicit specialization of a primary
5081 template declared with attribute @code{alloc_align}, @code{alloc_size},
5082 @code{assume_aligned}, @code{format}, @code{format_arg}, @code{malloc},
5083 or @code{nonnull} is declared without it. Attributes @code{deprecated},
5084 @code{error}, and @code{warning} suppress the warning.
5085 (@pxref{Function Attributes}).
5087 You can use the @code{copy} attribute to apply the same
5088 set of attributes to a declaration as that on another declaration without
5089 explicitly enumerating the attributes. This attribute can be applied
5090 to declarations of functions (@pxref{Common Function Attributes}),
5091 variables (@pxref{Common Variable Attributes}), or types
5092 (@pxref{Common Type Attributes}).
5094 @option{-Wmissing-attributes} is enabled by @option{-Wall}.
5096 For example, since the declaration of the primary function template
5097 below makes use of both attribute @code{malloc} and @code{alloc_size}
5098 the declaration of the explicit specialization of the template is
5099 diagnosed because it is missing one of the attributes.
5103 T* __attribute__ ((malloc, alloc_size (1)))
5107 void* __attribute__ ((malloc)) // missing alloc_size
5108 allocate<void> (size_t);
5111 @item -Wmissing-braces
5112 @opindex Wmissing-braces
5113 @opindex Wno-missing-braces
5114 Warn if an aggregate or union initializer is not fully bracketed. In
5115 the following example, the initializer for @code{a} is not fully
5116 bracketed, but that for @code{b} is fully bracketed. This warning is
5117 enabled by @option{-Wall} in C.
5120 int a[2][2] = @{ 0, 1, 2, 3 @};
5121 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
5124 This warning is enabled by @option{-Wall}.
5126 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
5127 @opindex Wmissing-include-dirs
5128 @opindex Wno-missing-include-dirs
5129 Warn if a user-supplied include directory does not exist.
5131 @item -Wmissing-profile
5132 @opindex Wmissing-profile
5133 @opindex Wno-missing-profile
5134 Warn if feedback profiles are missing when using the
5135 @option{-fprofile-use} option.
5136 This option diagnoses those cases where a new function or a new file is added
5137 to the user code between compiling with @option{-fprofile-generate} and with
5138 @option{-fprofile-use}, without regenerating the profiles. In these cases, the
5139 profile feedback data files do not contain any profile feedback information for
5140 the newly added function or file respectively. Also, in the case when profile
5141 count data (.gcda) files are removed, GCC cannot use any profile feedback
5142 information. In all these cases, warnings are issued to inform the user that a
5143 profile generation step is due. @option{-Wno-missing-profile} can be used to
5144 disable the warning. Ignoring the warning can result in poorly optimized code.
5145 Completely disabling the warning is not recommended and should be done only
5146 when non-existent profile data is justified.
5148 @item -Wmultistatement-macros
5149 @opindex Wmultistatement-macros
5150 @opindex Wno-multistatement-macros
5151 Warn about unsafe multiple statement macros that appear to be guarded
5152 by a clause such as @code{if}, @code{else}, @code{for}, @code{switch}, or
5153 @code{while}, in which only the first statement is actually guarded after
5154 the macro is expanded.
5159 #define DOIT x++; y++
5164 will increment @code{y} unconditionally, not just when @code{c} holds.
5165 The can usually be fixed by wrapping the macro in a do-while loop:
5167 #define DOIT do @{ x++; y++; @} while (0)
5172 This warning is enabled by @option{-Wall} in C and C++.
5175 @opindex Wparentheses
5176 @opindex Wno-parentheses
5177 Warn if parentheses are omitted in certain contexts, such
5178 as when there is an assignment in a context where a truth value
5179 is expected, or when operators are nested whose precedence people
5180 often get confused about.
5182 Also warn if a comparison like @code{x<=y<=z} appears; this is
5183 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
5184 interpretation from that of ordinary mathematical notation.
5186 Also warn for dangerous uses of the GNU extension to
5187 @code{?:} with omitted middle operand. When the condition
5188 in the @code{?}: operator is a boolean expression, the omitted value is
5189 always 1. Often programmers expect it to be a value computed
5190 inside the conditional expression instead.
5192 For C++ this also warns for some cases of unnecessary parentheses in
5193 declarations, which can indicate an attempt at a function call instead
5197 // Declares a local variable called mymutex.
5198 std::unique_lock<std::mutex> (mymutex);
5199 // User meant std::unique_lock<std::mutex> lock (mymutex);
5203 This warning is enabled by @option{-Wall}.
5205 @item -Wsequence-point
5206 @opindex Wsequence-point
5207 @opindex Wno-sequence-point
5208 Warn about code that may have undefined semantics because of violations
5209 of sequence point rules in the C and C++ standards.
5211 The C and C++ standards define the order in which expressions in a C/C++
5212 program are evaluated in terms of @dfn{sequence points}, which represent
5213 a partial ordering between the execution of parts of the program: those
5214 executed before the sequence point, and those executed after it. These
5215 occur after the evaluation of a full expression (one which is not part
5216 of a larger expression), after the evaluation of the first operand of a
5217 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
5218 function is called (but after the evaluation of its arguments and the
5219 expression denoting the called function), and in certain other places.
5220 Other than as expressed by the sequence point rules, the order of
5221 evaluation of subexpressions of an expression is not specified. All
5222 these rules describe only a partial order rather than a total order,
5223 since, for example, if two functions are called within one expression
5224 with no sequence point between them, the order in which the functions
5225 are called is not specified. However, the standards committee have
5226 ruled that function calls do not overlap.
5228 It is not specified when between sequence points modifications to the
5229 values of objects take effect. Programs whose behavior depends on this
5230 have undefined behavior; the C and C++ standards specify that ``Between
5231 the previous and next sequence point an object shall have its stored
5232 value modified at most once by the evaluation of an expression.
5233 Furthermore, the prior value shall be read only to determine the value
5234 to be stored.''. If a program breaks these rules, the results on any
5235 particular implementation are entirely unpredictable.
5237 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
5238 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
5239 diagnosed by this option, and it may give an occasional false positive
5240 result, but in general it has been found fairly effective at detecting
5241 this sort of problem in programs.
5243 The C++17 standard will define the order of evaluation of operands in
5244 more cases: in particular it requires that the right-hand side of an
5245 assignment be evaluated before the left-hand side, so the above
5246 examples are no longer undefined. But this warning will still warn
5247 about them, to help people avoid writing code that is undefined in C
5248 and earlier revisions of C++.
5250 The standard is worded confusingly, therefore there is some debate
5251 over the precise meaning of the sequence point rules in subtle cases.
5252 Links to discussions of the problem, including proposed formal
5253 definitions, may be found on the GCC readings page, at
5254 @uref{http://gcc.gnu.org/@/readings.html}.
5256 This warning is enabled by @option{-Wall} for C and C++.
5258 @item -Wno-return-local-addr
5259 @opindex Wno-return-local-addr
5260 @opindex Wreturn-local-addr
5261 Do not warn about returning a pointer (or in C++, a reference) to a
5262 variable that goes out of scope after the function returns.
5265 @opindex Wreturn-type
5266 @opindex Wno-return-type
5267 Warn whenever a function is defined with a return type that defaults
5268 to @code{int}. Also warn about any @code{return} statement with no
5269 return value in a function whose return type is not @code{void}
5270 (falling off the end of the function body is considered returning
5273 For C only, warn about a @code{return} statement with an expression in a
5274 function whose return type is @code{void}, unless the expression type is
5275 also @code{void}. As a GNU extension, the latter case is accepted
5276 without a warning unless @option{-Wpedantic} is used. Attempting
5277 to use the return value of a non-@code{void} function other than @code{main}
5278 that flows off the end by reaching the closing curly brace that terminates
5279 the function is undefined.
5281 Unlike in C, in C++, flowing off the end of a non-@code{void} function other
5282 than @code{main} results in undefined behavior even when the value of
5283 the function is not used.
5285 This warning is enabled by default in C++ and by @option{-Wall} otherwise.
5287 @item -Wshift-count-negative
5288 @opindex Wshift-count-negative
5289 @opindex Wno-shift-count-negative
5290 Warn if shift count is negative. This warning is enabled by default.
5292 @item -Wshift-count-overflow
5293 @opindex Wshift-count-overflow
5294 @opindex Wno-shift-count-overflow
5295 Warn if shift count >= width of type. This warning is enabled by default.
5297 @item -Wshift-negative-value
5298 @opindex Wshift-negative-value
5299 @opindex Wno-shift-negative-value
5300 Warn if left shifting a negative value. This warning is enabled by
5301 @option{-Wextra} in C99 and C++11 modes (and newer).
5303 @item -Wshift-overflow
5304 @itemx -Wshift-overflow=@var{n}
5305 @opindex Wshift-overflow
5306 @opindex Wno-shift-overflow
5307 Warn about left shift overflows. This warning is enabled by
5308 default in C99 and C++11 modes (and newer).
5311 @item -Wshift-overflow=1
5312 This is the warning level of @option{-Wshift-overflow} and is enabled
5313 by default in C99 and C++11 modes (and newer). This warning level does
5314 not warn about left-shifting 1 into the sign bit. (However, in C, such
5315 an overflow is still rejected in contexts where an integer constant expression
5316 is required.) No warning is emitted in C++2A mode (and newer), as signed left
5319 @item -Wshift-overflow=2
5320 This warning level also warns about left-shifting 1 into the sign bit,
5321 unless C++14 mode (or newer) is active.
5327 Warn whenever a @code{switch} statement has an index of enumerated type
5328 and lacks a @code{case} for one or more of the named codes of that
5329 enumeration. (The presence of a @code{default} label prevents this
5330 warning.) @code{case} labels outside the enumeration range also
5331 provoke warnings when this option is used (even if there is a
5332 @code{default} label).
5333 This warning is enabled by @option{-Wall}.
5335 @item -Wswitch-default
5336 @opindex Wswitch-default
5337 @opindex Wno-switch-default
5338 Warn whenever a @code{switch} statement does not have a @code{default}
5342 @opindex Wswitch-enum
5343 @opindex Wno-switch-enum
5344 Warn whenever a @code{switch} statement has an index of enumerated type
5345 and lacks a @code{case} for one or more of the named codes of that
5346 enumeration. @code{case} labels outside the enumeration range also
5347 provoke warnings when this option is used. The only difference
5348 between @option{-Wswitch} and this option is that this option gives a
5349 warning about an omitted enumeration code even if there is a
5350 @code{default} label.
5353 @opindex Wswitch-bool
5354 @opindex Wno-switch-bool
5355 Warn whenever a @code{switch} statement has an index of boolean type
5356 and the case values are outside the range of a boolean type.
5357 It is possible to suppress this warning by casting the controlling
5358 expression to a type other than @code{bool}. For example:
5361 switch ((int) (a == 4))
5367 This warning is enabled by default for C and C++ programs.
5369 @item -Wswitch-unreachable
5370 @opindex Wswitch-unreachable
5371 @opindex Wno-switch-unreachable
5372 Warn whenever a @code{switch} statement contains statements between the
5373 controlling expression and the first case label, which will never be
5374 executed. For example:
5386 @option{-Wswitch-unreachable} does not warn if the statement between the
5387 controlling expression and the first case label is just a declaration:
5400 This warning is enabled by default for C and C++ programs.
5402 @item -Wsync-nand @r{(C and C++ only)}
5404 @opindex Wno-sync-nand
5405 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
5406 built-in functions are used. These functions changed semantics in GCC 4.4.
5408 @item -Wunused-but-set-parameter
5409 @opindex Wunused-but-set-parameter
5410 @opindex Wno-unused-but-set-parameter
5411 Warn whenever a function parameter is assigned to, but otherwise unused
5412 (aside from its declaration).
5414 To suppress this warning use the @code{unused} attribute
5415 (@pxref{Variable Attributes}).
5417 This warning is also enabled by @option{-Wunused} together with
5420 @item -Wunused-but-set-variable
5421 @opindex Wunused-but-set-variable
5422 @opindex Wno-unused-but-set-variable
5423 Warn whenever a local variable is assigned to, but otherwise unused
5424 (aside from its declaration).
5425 This warning is enabled by @option{-Wall}.
5427 To suppress this warning use the @code{unused} attribute
5428 (@pxref{Variable Attributes}).
5430 This warning is also enabled by @option{-Wunused}, which is enabled
5433 @item -Wunused-function
5434 @opindex Wunused-function
5435 @opindex Wno-unused-function
5436 Warn whenever a static function is declared but not defined or a
5437 non-inline static function is unused.
5438 This warning is enabled by @option{-Wall}.
5440 @item -Wunused-label
5441 @opindex Wunused-label
5442 @opindex Wno-unused-label
5443 Warn whenever a label is declared but not used.
5444 This warning is enabled by @option{-Wall}.
5446 To suppress this warning use the @code{unused} attribute
5447 (@pxref{Variable Attributes}).
5449 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
5450 @opindex Wunused-local-typedefs
5451 @opindex Wno-unused-local-typedefs
5452 Warn when a typedef locally defined in a function is not used.
5453 This warning is enabled by @option{-Wall}.
5455 @item -Wunused-parameter
5456 @opindex Wunused-parameter
5457 @opindex Wno-unused-parameter
5458 Warn whenever a function parameter is unused aside from its declaration.
5460 To suppress this warning use the @code{unused} attribute
5461 (@pxref{Variable Attributes}).
5463 @item -Wno-unused-result
5464 @opindex Wunused-result
5465 @opindex Wno-unused-result
5466 Do not warn if a caller of a function marked with attribute
5467 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
5468 its return value. The default is @option{-Wunused-result}.
5470 @item -Wunused-variable
5471 @opindex Wunused-variable
5472 @opindex Wno-unused-variable
5473 Warn whenever a local or static variable is unused aside from its
5474 declaration. This option implies @option{-Wunused-const-variable=1} for C,
5475 but not for C++. This warning is enabled by @option{-Wall}.
5477 To suppress this warning use the @code{unused} attribute
5478 (@pxref{Variable Attributes}).
5480 @item -Wunused-const-variable
5481 @itemx -Wunused-const-variable=@var{n}
5482 @opindex Wunused-const-variable
5483 @opindex Wno-unused-const-variable
5484 Warn whenever a constant static variable is unused aside from its declaration.
5485 @option{-Wunused-const-variable=1} is enabled by @option{-Wunused-variable}
5486 for C, but not for C++. In C this declares variable storage, but in C++ this
5487 is not an error since const variables take the place of @code{#define}s.
5489 To suppress this warning use the @code{unused} attribute
5490 (@pxref{Variable Attributes}).
5493 @item -Wunused-const-variable=1
5494 This is the warning level that is enabled by @option{-Wunused-variable} for
5495 C. It warns only about unused static const variables defined in the main
5496 compilation unit, but not about static const variables declared in any
5499 @item -Wunused-const-variable=2
5500 This warning level also warns for unused constant static variables in
5501 headers (excluding system headers). This is the warning level of
5502 @option{-Wunused-const-variable} and must be explicitly requested since
5503 in C++ this isn't an error and in C it might be harder to clean up all
5507 @item -Wunused-value
5508 @opindex Wunused-value
5509 @opindex Wno-unused-value
5510 Warn whenever a statement computes a result that is explicitly not
5511 used. To suppress this warning cast the unused expression to
5512 @code{void}. This includes an expression-statement or the left-hand
5513 side of a comma expression that contains no side effects. For example,
5514 an expression such as @code{x[i,j]} causes a warning, while
5515 @code{x[(void)i,j]} does not.
5517 This warning is enabled by @option{-Wall}.
5522 All the above @option{-Wunused} options combined.
5524 In order to get a warning about an unused function parameter, you must
5525 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
5526 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
5528 @item -Wuninitialized
5529 @opindex Wuninitialized
5530 @opindex Wno-uninitialized
5531 Warn if an automatic variable is used without first being initialized
5532 or if a variable may be clobbered by a @code{setjmp} call. In C++,
5533 warn if a non-static reference or non-static @code{const} member
5534 appears in a class without constructors.
5536 If you want to warn about code that uses the uninitialized value of the
5537 variable in its own initializer, use the @option{-Winit-self} option.
5539 These warnings occur for individual uninitialized or clobbered
5540 elements of structure, union or array variables as well as for
5541 variables that are uninitialized or clobbered as a whole. They do
5542 not occur for variables or elements declared @code{volatile}. Because
5543 these warnings depend on optimization, the exact variables or elements
5544 for which there are warnings depends on the precise optimization
5545 options and version of GCC used.
5547 Note that there may be no warning about a variable that is used only
5548 to compute a value that itself is never used, because such
5549 computations may be deleted by data flow analysis before the warnings
5552 @item -Winvalid-memory-model
5553 @opindex Winvalid-memory-model
5554 @opindex Wno-invalid-memory-model
5555 Warn for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins},
5556 and the C11 atomic generic functions with a memory consistency argument
5557 that is either invalid for the operation or outside the range of values
5558 of the @code{memory_order} enumeration. For example, since the
5559 @code{__atomic_store} and @code{__atomic_store_n} built-ins are only
5560 defined for the relaxed, release, and sequentially consistent memory
5561 orders the following code is diagnosed:
5566 __atomic_store_n (i, 0, memory_order_consume);
5570 @option{-Winvalid-memory-model} is enabled by default.
5572 @item -Wmaybe-uninitialized
5573 @opindex Wmaybe-uninitialized
5574 @opindex Wno-maybe-uninitialized
5575 For an automatic (i.e.@: local) variable, if there exists a path from the
5576 function entry to a use of the variable that is initialized, but there exist
5577 some other paths for which the variable is not initialized, the compiler
5578 emits a warning if it cannot prove the uninitialized paths are not
5579 executed at run time.
5581 These warnings are only possible in optimizing compilation, because otherwise
5582 GCC does not keep track of the state of variables.
5584 These warnings are made optional because GCC may not be able to determine when
5585 the code is correct in spite of appearing to have an error. Here is one
5586 example of how this can happen:
5606 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
5607 always initialized, but GCC doesn't know this. To suppress the
5608 warning, you need to provide a default case with assert(0) or
5611 @cindex @code{longjmp} warnings
5612 This option also warns when a non-volatile automatic variable might be
5613 changed by a call to @code{longjmp}.
5614 The compiler sees only the calls to @code{setjmp}. It cannot know
5615 where @code{longjmp} will be called; in fact, a signal handler could
5616 call it at any point in the code. As a result, you may get a warning
5617 even when there is in fact no problem because @code{longjmp} cannot
5618 in fact be called at the place that would cause a problem.
5620 Some spurious warnings can be avoided if you declare all the functions
5621 you use that never return as @code{noreturn}. @xref{Function
5624 This warning is enabled by @option{-Wall} or @option{-Wextra}.
5626 @item -Wunknown-pragmas
5627 @opindex Wunknown-pragmas
5628 @opindex Wno-unknown-pragmas
5629 @cindex warning for unknown pragmas
5630 @cindex unknown pragmas, warning
5631 @cindex pragmas, warning of unknown
5632 Warn when a @code{#pragma} directive is encountered that is not understood by
5633 GCC@. If this command-line option is used, warnings are even issued
5634 for unknown pragmas in system header files. This is not the case if
5635 the warnings are only enabled by the @option{-Wall} command-line option.
5638 @opindex Wno-pragmas
5640 Do not warn about misuses of pragmas, such as incorrect parameters,
5641 invalid syntax, or conflicts between pragmas. See also
5642 @option{-Wunknown-pragmas}.
5644 @item -Wno-prio-ctor-dtor
5645 @opindex Wno-prio-ctor-dtor
5646 @opindex Wprio-ctor-dtor
5647 Do not warn if a priority from 0 to 100 is used for constructor or destructor.
5648 The use of constructor and destructor attributes allow you to assign a
5649 priority to the constructor/destructor to control its order of execution
5650 before @code{main} is called or after it returns. The priority values must be
5651 greater than 100 as the compiler reserves priority values between 0--100 for
5654 @item -Wstrict-aliasing
5655 @opindex Wstrict-aliasing
5656 @opindex Wno-strict-aliasing
5657 This option is only active when @option{-fstrict-aliasing} is active.
5658 It warns about code that might break the strict aliasing rules that the
5659 compiler is using for optimization. The warning does not catch all
5660 cases, but does attempt to catch the more common pitfalls. It is
5661 included in @option{-Wall}.
5662 It is equivalent to @option{-Wstrict-aliasing=3}
5664 @item -Wstrict-aliasing=n
5665 @opindex Wstrict-aliasing=n
5666 This option is only active when @option{-fstrict-aliasing} is active.
5667 It warns about code that might break the strict aliasing rules that the
5668 compiler is using for optimization.
5669 Higher levels correspond to higher accuracy (fewer false positives).
5670 Higher levels also correspond to more effort, similar to the way @option{-O}
5672 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
5674 Level 1: Most aggressive, quick, least accurate.
5675 Possibly useful when higher levels
5676 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
5677 false negatives. However, it has many false positives.
5678 Warns for all pointer conversions between possibly incompatible types,
5679 even if never dereferenced. Runs in the front end only.
5681 Level 2: Aggressive, quick, not too precise.
5682 May still have many false positives (not as many as level 1 though),
5683 and few false negatives (but possibly more than level 1).
5684 Unlike level 1, it only warns when an address is taken. Warns about
5685 incomplete types. Runs in the front end only.
5687 Level 3 (default for @option{-Wstrict-aliasing}):
5688 Should have very few false positives and few false
5689 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
5690 Takes care of the common pun+dereference pattern in the front end:
5691 @code{*(int*)&some_float}.
5692 If optimization is enabled, it also runs in the back end, where it deals
5693 with multiple statement cases using flow-sensitive points-to information.
5694 Only warns when the converted pointer is dereferenced.
5695 Does not warn about incomplete types.
5697 @item -Wstrict-overflow
5698 @itemx -Wstrict-overflow=@var{n}
5699 @opindex Wstrict-overflow
5700 @opindex Wno-strict-overflow
5701 This option is only active when signed overflow is undefined.
5702 It warns about cases where the compiler optimizes based on the
5703 assumption that signed overflow does not occur. Note that it does not
5704 warn about all cases where the code might overflow: it only warns
5705 about cases where the compiler implements some optimization. Thus
5706 this warning depends on the optimization level.
5708 An optimization that assumes that signed overflow does not occur is
5709 perfectly safe if the values of the variables involved are such that
5710 overflow never does, in fact, occur. Therefore this warning can
5711 easily give a false positive: a warning about code that is not
5712 actually a problem. To help focus on important issues, several
5713 warning levels are defined. No warnings are issued for the use of
5714 undefined signed overflow when estimating how many iterations a loop
5715 requires, in particular when determining whether a loop will be
5719 @item -Wstrict-overflow=1
5720 Warn about cases that are both questionable and easy to avoid. For
5721 example the compiler simplifies
5722 @code{x + 1 > x} to @code{1}. This level of
5723 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
5724 are not, and must be explicitly requested.
5726 @item -Wstrict-overflow=2
5727 Also warn about other cases where a comparison is simplified to a
5728 constant. For example: @code{abs (x) >= 0}. This can only be
5729 simplified when signed integer overflow is undefined, because
5730 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
5731 zero. @option{-Wstrict-overflow} (with no level) is the same as
5732 @option{-Wstrict-overflow=2}.
5734 @item -Wstrict-overflow=3
5735 Also warn about other cases where a comparison is simplified. For
5736 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
5738 @item -Wstrict-overflow=4
5739 Also warn about other simplifications not covered by the above cases.
5740 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
5742 @item -Wstrict-overflow=5
5743 Also warn about cases where the compiler reduces the magnitude of a
5744 constant involved in a comparison. For example: @code{x + 2 > y} is
5745 simplified to @code{x + 1 >= y}. This is reported only at the
5746 highest warning level because this simplification applies to many
5747 comparisons, so this warning level gives a very large number of
5751 @item -Wstringop-overflow
5752 @itemx -Wstringop-overflow=@var{type}
5753 @opindex Wstringop-overflow
5754 @opindex Wno-stringop-overflow
5755 Warn for calls to string manipulation functions such as @code{memcpy} and
5756 @code{strcpy} that are determined to overflow the destination buffer. The
5757 optional argument is one greater than the type of Object Size Checking to
5758 perform to determine the size of the destination. @xref{Object Size Checking}.
5759 The argument is meaningful only for functions that operate on character arrays
5760 but not for raw memory functions like @code{memcpy} which always make use
5761 of Object Size type-0. The option also warns for calls that specify a size
5762 in excess of the largest possible object or at most @code{SIZE_MAX / 2} bytes.
5763 The option produces the best results with optimization enabled but can detect
5764 a small subset of simple buffer overflows even without optimization in
5765 calls to the GCC built-in functions like @code{__builtin_memcpy} that
5766 correspond to the standard functions. In any case, the option warns about
5767 just a subset of buffer overflows detected by the corresponding overflow
5768 checking built-ins. For example, the option will issue a warning for
5769 the @code{strcpy} call below because it copies at least 5 characters
5770 (the string @code{"blue"} including the terminating NUL) into the buffer
5774 enum Color @{ blue, purple, yellow @};
5775 const char* f (enum Color clr)
5777 static char buf [4];
5781 case blue: str = "blue"; break;
5782 case purple: str = "purple"; break;
5783 case yellow: str = "yellow"; break;
5786 return strcpy (buf, str); // warning here
5790 Option @option{-Wstringop-overflow=2} is enabled by default.
5793 @item -Wstringop-overflow
5794 @itemx -Wstringop-overflow=1
5795 @opindex Wstringop-overflow
5796 @opindex Wno-stringop-overflow
5797 The @option{-Wstringop-overflow=1} option uses type-zero Object Size Checking
5798 to determine the sizes of destination objects. This is the default setting
5799 of the option. At this setting the option will not warn for writes past
5800 the end of subobjects of larger objects accessed by pointers unless the
5801 size of the largest surrounding object is known. When the destination may
5802 be one of several objects it is assumed to be the largest one of them. On
5803 Linux systems, when optimization is enabled at this setting the option warns
5804 for the same code as when the @code{_FORTIFY_SOURCE} macro is defined to
5807 @item -Wstringop-overflow=2
5808 The @option{-Wstringop-overflow=2} option uses type-one Object Size Checking
5809 to determine the sizes of destination objects. At this setting the option
5810 will warn about overflows when writing to members of the largest complete
5811 objects whose exact size is known. It will, however, not warn for excessive
5812 writes to the same members of unknown objects referenced by pointers since
5813 they may point to arrays containing unknown numbers of elements.
5815 @item -Wstringop-overflow=3
5816 The @option{-Wstringop-overflow=3} option uses type-two Object Size Checking
5817 to determine the sizes of destination objects. At this setting the option
5818 warns about overflowing the smallest object or data member. This is the
5819 most restrictive setting of the option that may result in warnings for safe
5822 @item -Wstringop-overflow=4
5823 The @option{-Wstringop-overflow=4} option uses type-three Object Size Checking
5824 to determine the sizes of destination objects. At this setting the option
5825 will warn about overflowing any data members, and when the destination is
5826 one of several objects it uses the size of the largest of them to decide
5827 whether to issue a warning. Similarly to @option{-Wstringop-overflow=3} this
5828 setting of the option may result in warnings for benign code.
5831 @item -Wstringop-truncation
5832 @opindex Wstringop-truncation
5833 @opindex Wno-stringop-truncation
5834 Warn for calls to bounded string manipulation functions such as @code{strncat},
5835 @code{strncpy}, and @code{stpncpy} that may either truncate the copied string
5836 or leave the destination unchanged.
5838 In the following example, the call to @code{strncat} specifies a bound that
5839 is less than the length of the source string. As a result, the copy of
5840 the source will be truncated and so the call is diagnosed. To avoid the
5841 warning use @code{bufsize - strlen (buf) - 1)} as the bound.
5844 void append (char *buf, size_t bufsize)
5846 strncat (buf, ".txt", 3);
5850 As another example, the following call to @code{strncpy} results in copying
5851 to @code{d} just the characters preceding the terminating NUL, without
5852 appending the NUL to the end. Assuming the result of @code{strncpy} is
5853 necessarily a NUL-terminated string is a common mistake, and so the call
5854 is diagnosed. To avoid the warning when the result is not expected to be
5855 NUL-terminated, call @code{memcpy} instead.
5858 void copy (char *d, const char *s)
5860 strncpy (d, s, strlen (s));
5864 In the following example, the call to @code{strncpy} specifies the size
5865 of the destination buffer as the bound. If the length of the source
5866 string is equal to or greater than this size the result of the copy will
5867 not be NUL-terminated. Therefore, the call is also diagnosed. To avoid
5868 the warning, specify @code{sizeof buf - 1} as the bound and set the last
5869 element of the buffer to @code{NUL}.
5872 void copy (const char *s)
5875 strncpy (buf, s, sizeof buf);
5880 In situations where a character array is intended to store a sequence
5881 of bytes with no terminating @code{NUL} such an array may be annotated
5882 with attribute @code{nonstring} to avoid this warning. Such arrays,
5883 however, are not suitable arguments to functions that expect
5884 @code{NUL}-terminated strings. To help detect accidental misuses of
5885 such arrays GCC issues warnings unless it can prove that the use is
5886 safe. @xref{Common Variable Attributes}.
5888 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{|}cold@r{|}malloc@r{]}
5889 @opindex Wsuggest-attribute=
5890 @opindex Wno-suggest-attribute=
5891 Warn for cases where adding an attribute may be beneficial. The
5892 attributes currently supported are listed below.
5895 @item -Wsuggest-attribute=pure
5896 @itemx -Wsuggest-attribute=const
5897 @itemx -Wsuggest-attribute=noreturn
5898 @itemx -Wmissing-noreturn
5899 @itemx -Wsuggest-attribute=malloc
5900 @opindex Wsuggest-attribute=pure
5901 @opindex Wno-suggest-attribute=pure
5902 @opindex Wsuggest-attribute=const
5903 @opindex Wno-suggest-attribute=const
5904 @opindex Wsuggest-attribute=noreturn
5905 @opindex Wno-suggest-attribute=noreturn
5906 @opindex Wmissing-noreturn
5907 @opindex Wno-missing-noreturn
5908 @opindex Wsuggest-attribute=malloc
5909 @opindex Wno-suggest-attribute=malloc
5911 Warn about functions that might be candidates for attributes
5912 @code{pure}, @code{const} or @code{noreturn} or @code{malloc}. The compiler
5913 only warns for functions visible in other compilation units or (in the case of
5914 @code{pure} and @code{const}) if it cannot prove that the function returns
5915 normally. A function returns normally if it doesn't contain an infinite loop or
5916 return abnormally by throwing, calling @code{abort} or trapping. This analysis
5917 requires option @option{-fipa-pure-const}, which is enabled by default at
5918 @option{-O} and higher. Higher optimization levels improve the accuracy
5921 @item -Wsuggest-attribute=format
5922 @itemx -Wmissing-format-attribute
5923 @opindex Wsuggest-attribute=format
5924 @opindex Wmissing-format-attribute
5925 @opindex Wno-suggest-attribute=format
5926 @opindex Wno-missing-format-attribute
5930 Warn about function pointers that might be candidates for @code{format}
5931 attributes. Note these are only possible candidates, not absolute ones.
5932 GCC guesses that function pointers with @code{format} attributes that
5933 are used in assignment, initialization, parameter passing or return
5934 statements should have a corresponding @code{format} attribute in the
5935 resulting type. I.e.@: the left-hand side of the assignment or
5936 initialization, the type of the parameter variable, or the return type
5937 of the containing function respectively should also have a @code{format}
5938 attribute to avoid the warning.
5940 GCC also warns about function definitions that might be
5941 candidates for @code{format} attributes. Again, these are only
5942 possible candidates. GCC guesses that @code{format} attributes
5943 might be appropriate for any function that calls a function like
5944 @code{vprintf} or @code{vscanf}, but this might not always be the
5945 case, and some functions for which @code{format} attributes are
5946 appropriate may not be detected.
5948 @item -Wsuggest-attribute=cold
5949 @opindex Wsuggest-attribute=cold
5950 @opindex Wno-suggest-attribute=cold
5952 Warn about functions that might be candidates for @code{cold} attribute. This
5953 is based on static detection and generally will only warn about functions which
5954 always leads to a call to another @code{cold} function such as wrappers of
5955 C++ @code{throw} or fatal error reporting functions leading to @code{abort}.
5958 @item -Wsuggest-final-types
5959 @opindex Wno-suggest-final-types
5960 @opindex Wsuggest-final-types
5961 Warn about types with virtual methods where code quality would be improved
5962 if the type were declared with the C++11 @code{final} specifier,
5964 declared in an anonymous namespace. This allows GCC to more aggressively
5965 devirtualize the polymorphic calls. This warning is more effective with link
5966 time optimization, where the information about the class hierarchy graph is
5969 @item -Wsuggest-final-methods
5970 @opindex Wno-suggest-final-methods
5971 @opindex Wsuggest-final-methods
5972 Warn about virtual methods where code quality would be improved if the method
5973 were declared with the C++11 @code{final} specifier,
5974 or, if possible, its type were
5975 declared in an anonymous namespace or with the @code{final} specifier.
5977 more effective with link-time optimization, where the information about the
5978 class hierarchy graph is more complete. It is recommended to first consider
5979 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
5982 @item -Wsuggest-override
5983 Warn about overriding virtual functions that are not marked with the override
5987 @opindex Wno-alloc-zero
5988 @opindex Walloc-zero
5989 Warn about calls to allocation functions decorated with attribute
5990 @code{alloc_size} that specify zero bytes, including those to the built-in
5991 forms of the functions @code{aligned_alloc}, @code{alloca}, @code{calloc},
5992 @code{malloc}, and @code{realloc}. Because the behavior of these functions
5993 when called with a zero size differs among implementations (and in the case
5994 of @code{realloc} has been deprecated) relying on it may result in subtle
5995 portability bugs and should be avoided.
5997 @item -Walloc-size-larger-than=@var{byte-size}
5998 @opindex Walloc-size-larger-than=
5999 @opindex Wno-alloc-size-larger-than
6000 Warn about calls to functions decorated with attribute @code{alloc_size}
6001 that attempt to allocate objects larger than the specified number of bytes,
6002 or where the result of the size computation in an integer type with infinite
6003 precision would exceed the value of @samp{PTRDIFF_MAX} on the target.
6004 @option{-Walloc-size-larger-than=}@samp{PTRDIFF_MAX} is enabled by default.
6005 Warnings controlled by the option can be disabled either by specifying
6006 @var{byte-size} of @samp{SIZE_MAX} or more or by
6007 @option{-Wno-alloc-size-larger-than}.
6008 @xref{Function Attributes}.
6010 @item -Wno-alloc-size-larger-than
6011 @opindex Wno-alloc-size-larger-than
6012 Disable @option{-Walloc-size-larger-than=} warnings. The option is
6013 equivalent to @option{-Walloc-size-larger-than=}@samp{SIZE_MAX} or
6019 This option warns on all uses of @code{alloca} in the source.
6021 @item -Walloca-larger-than=@var{byte-size}
6022 @opindex Walloca-larger-than=
6023 @opindex Wno-alloca-larger-than
6024 This option warns on calls to @code{alloca} with an integer argument whose
6025 value is either zero, or that is not bounded by a controlling predicate
6026 that limits its value to at most @var{byte-size}. It also warns for calls
6027 to @code{alloca} where the bound value is unknown. Arguments of non-integer
6028 types are considered unbounded even if they appear to be constrained to
6031 For example, a bounded case of @code{alloca} could be:
6034 void func (size_t n)
6045 In the above example, passing @code{-Walloca-larger-than=1000} would not
6046 issue a warning because the call to @code{alloca} is known to be at most
6047 1000 bytes. However, if @code{-Walloca-larger-than=500} were passed,
6048 the compiler would emit a warning.
6050 Unbounded uses, on the other hand, are uses of @code{alloca} with no
6051 controlling predicate constraining its integer argument. For example:
6056 void *p = alloca (n);
6061 If @code{-Walloca-larger-than=500} were passed, the above would trigger
6062 a warning, but this time because of the lack of bounds checking.
6064 Note, that even seemingly correct code involving signed integers could
6068 void func (signed int n)
6078 In the above example, @var{n} could be negative, causing a larger than
6079 expected argument to be implicitly cast into the @code{alloca} call.
6081 This option also warns when @code{alloca} is used in a loop.
6083 @option{-Walloca-larger-than=}@samp{PTRDIFF_MAX} is enabled by default
6084 but is usually only effective when @option{-ftree-vrp} is active (default
6085 for @option{-O2} and above).
6087 See also @option{-Wvla-larger-than=}@samp{byte-size}.
6089 @item -Wno-alloca-larger-than
6090 @opindex Wno-alloca-larger-than
6091 Disable @option{-Walloca-larger-than=} warnings. The option is
6092 equivalent to @option{-Walloca-larger-than=}@samp{SIZE_MAX} or larger.
6094 @item -Warray-bounds
6095 @itemx -Warray-bounds=@var{n}
6096 @opindex Wno-array-bounds
6097 @opindex Warray-bounds
6098 This option is only active when @option{-ftree-vrp} is active
6099 (default for @option{-O2} and above). It warns about subscripts to arrays
6100 that are always out of bounds. This warning is enabled by @option{-Wall}.
6103 @item -Warray-bounds=1
6104 This is the warning level of @option{-Warray-bounds} and is enabled
6105 by @option{-Wall}; higher levels are not, and must be explicitly requested.
6107 @item -Warray-bounds=2
6108 This warning level also warns about out of bounds access for
6109 arrays at the end of a struct and for arrays accessed through
6110 pointers. This warning level may give a larger number of
6111 false positives and is deactivated by default.
6114 @item -Wattribute-alias=@var{n}
6115 @itemx -Wno-attribute-alias
6116 @opindex -Wattribute-alias
6117 @opindex -Wno-attribute-alias
6118 Warn about declarations using the @code{alias} and similar attributes whose
6119 target is incompatible with the type of the alias.
6120 @xref{Function Attributes,,Declaring Attributes of Functions}.
6123 @item -Wattribute-alias=1
6124 The default warning level of the @option{-Wattribute-alias} option diagnoses
6125 incompatibilities between the type of the alias declaration and that of its
6126 target. Such incompatibilities are typically indicative of bugs.
6128 @item -Wattribute-alias=2
6130 At this level @option{-Wattribute-alias} also diagnoses cases where
6131 the attributes of the alias declaration are more restrictive than the
6132 attributes applied to its target. These mismatches can potentially
6133 result in incorrect code generation. In other cases they may be
6134 benign and could be resolved simply by adding the missing attribute to
6135 the target. For comparison, see the @option{-Wmissing-attributes}
6136 option, which controls diagnostics when the alias declaration is less
6137 restrictive than the target, rather than more restrictive.
6139 Attributes considered include @code{alloc_align}, @code{alloc_size},
6140 @code{cold}, @code{const}, @code{hot}, @code{leaf}, @code{malloc},
6141 @code{nonnull}, @code{noreturn}, @code{nothrow}, @code{pure},
6142 @code{returns_nonnull}, and @code{returns_twice}.
6145 @option{-Wattribute-alias} is equivalent to @option{-Wattribute-alias=1}.
6146 This is the default. You can disable these warnings with either
6147 @option{-Wno-attribute-alias} or @option{-Wattribute-alias=0}.
6149 @item -Wbool-compare
6150 @opindex Wno-bool-compare
6151 @opindex Wbool-compare
6152 Warn about boolean expression compared with an integer value different from
6153 @code{true}/@code{false}. For instance, the following comparison is
6158 if ((n > 1) == 2) @{ @dots{} @}
6160 This warning is enabled by @option{-Wall}.
6162 @item -Wbool-operation
6163 @opindex Wno-bool-operation
6164 @opindex Wbool-operation
6165 Warn about suspicious operations on expressions of a boolean type. For
6166 instance, bitwise negation of a boolean is very likely a bug in the program.
6167 For C, this warning also warns about incrementing or decrementing a boolean,
6168 which rarely makes sense. (In C++, decrementing a boolean is always invalid.
6169 Incrementing a boolean is invalid in C++17, and deprecated otherwise.)
6171 This warning is enabled by @option{-Wall}.
6173 @item -Wduplicated-branches
6174 @opindex Wno-duplicated-branches
6175 @opindex Wduplicated-branches
6176 Warn when an if-else has identical branches. This warning detects cases like
6183 It doesn't warn when both branches contain just a null statement. This warning
6184 also warn for conditional operators:
6186 int i = x ? *p : *p;
6189 @item -Wduplicated-cond
6190 @opindex Wno-duplicated-cond
6191 @opindex Wduplicated-cond
6192 Warn about duplicated conditions in an if-else-if chain. For instance,
6193 warn for the following code:
6195 if (p->q != NULL) @{ @dots{} @}
6196 else if (p->q != NULL) @{ @dots{} @}
6199 @item -Wframe-address
6200 @opindex Wno-frame-address
6201 @opindex Wframe-address
6202 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
6203 is called with an argument greater than 0. Such calls may return indeterminate
6204 values or crash the program. The warning is included in @option{-Wall}.
6206 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
6207 @opindex Wno-discarded-qualifiers
6208 @opindex Wdiscarded-qualifiers
6209 Do not warn if type qualifiers on pointers are being discarded.
6210 Typically, the compiler warns if a @code{const char *} variable is
6211 passed to a function that takes a @code{char *} parameter. This option
6212 can be used to suppress such a warning.
6214 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
6215 @opindex Wno-discarded-array-qualifiers
6216 @opindex Wdiscarded-array-qualifiers
6217 Do not warn if type qualifiers on arrays which are pointer targets
6218 are being discarded. Typically, the compiler warns if a
6219 @code{const int (*)[]} variable is passed to a function that
6220 takes a @code{int (*)[]} parameter. This option can be used to
6221 suppress such a warning.
6223 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
6224 @opindex Wno-incompatible-pointer-types
6225 @opindex Wincompatible-pointer-types
6226 Do not warn when there is a conversion between pointers that have incompatible
6227 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
6228 which warns for pointer argument passing or assignment with different
6231 @item -Wno-int-conversion @r{(C and Objective-C only)}
6232 @opindex Wno-int-conversion
6233 @opindex Wint-conversion
6234 Do not warn about incompatible integer to pointer and pointer to integer
6235 conversions. This warning is about implicit conversions; for explicit
6236 conversions the warnings @option{-Wno-int-to-pointer-cast} and
6237 @option{-Wno-pointer-to-int-cast} may be used.
6239 @item -Wno-div-by-zero
6240 @opindex Wno-div-by-zero
6241 @opindex Wdiv-by-zero
6242 Do not warn about compile-time integer division by zero. Floating-point
6243 division by zero is not warned about, as it can be a legitimate way of
6244 obtaining infinities and NaNs.
6246 @item -Wsystem-headers
6247 @opindex Wsystem-headers
6248 @opindex Wno-system-headers
6249 @cindex warnings from system headers
6250 @cindex system headers, warnings from
6251 Print warning messages for constructs found in system header files.
6252 Warnings from system headers are normally suppressed, on the assumption
6253 that they usually do not indicate real problems and would only make the
6254 compiler output harder to read. Using this command-line option tells
6255 GCC to emit warnings from system headers as if they occurred in user
6256 code. However, note that using @option{-Wall} in conjunction with this
6257 option does @emph{not} warn about unknown pragmas in system
6258 headers---for that, @option{-Wunknown-pragmas} must also be used.
6260 @item -Wtautological-compare
6261 @opindex Wtautological-compare
6262 @opindex Wno-tautological-compare
6263 Warn if a self-comparison always evaluates to true or false. This
6264 warning detects various mistakes such as:
6268 if (i > i) @{ @dots{} @}
6271 This warning also warns about bitwise comparisons that always evaluate
6272 to true or false, for instance:
6274 if ((a & 16) == 10) @{ @dots{} @}
6276 will always be false.
6278 This warning is enabled by @option{-Wall}.
6281 @opindex Wtrampolines
6282 @opindex Wno-trampolines
6283 Warn about trampolines generated for pointers to nested functions.
6284 A trampoline is a small piece of data or code that is created at run
6285 time on the stack when the address of a nested function is taken, and is
6286 used to call the nested function indirectly. For some targets, it is
6287 made up of data only and thus requires no special treatment. But, for
6288 most targets, it is made up of code and thus requires the stack to be
6289 made executable in order for the program to work properly.
6292 @opindex Wfloat-equal
6293 @opindex Wno-float-equal
6294 Warn if floating-point values are used in equality comparisons.
6296 The idea behind this is that sometimes it is convenient (for the
6297 programmer) to consider floating-point values as approximations to
6298 infinitely precise real numbers. If you are doing this, then you need
6299 to compute (by analyzing the code, or in some other way) the maximum or
6300 likely maximum error that the computation introduces, and allow for it
6301 when performing comparisons (and when producing output, but that's a
6302 different problem). In particular, instead of testing for equality, you
6303 should check to see whether the two values have ranges that overlap; and
6304 this is done with the relational operators, so equality comparisons are
6307 @item -Wtraditional @r{(C and Objective-C only)}
6308 @opindex Wtraditional
6309 @opindex Wno-traditional
6310 Warn about certain constructs that behave differently in traditional and
6311 ISO C@. Also warn about ISO C constructs that have no traditional C
6312 equivalent, and/or problematic constructs that should be avoided.
6316 Macro parameters that appear within string literals in the macro body.
6317 In traditional C macro replacement takes place within string literals,
6318 but in ISO C it does not.
6321 In traditional C, some preprocessor directives did not exist.
6322 Traditional preprocessors only considered a line to be a directive
6323 if the @samp{#} appeared in column 1 on the line. Therefore
6324 @option{-Wtraditional} warns about directives that traditional C
6325 understands but ignores because the @samp{#} does not appear as the
6326 first character on the line. It also suggests you hide directives like
6327 @code{#pragma} not understood by traditional C by indenting them. Some
6328 traditional implementations do not recognize @code{#elif}, so this option
6329 suggests avoiding it altogether.
6332 A function-like macro that appears without arguments.
6335 The unary plus operator.
6338 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
6339 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
6340 constants.) Note, these suffixes appear in macros defined in the system
6341 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
6342 Use of these macros in user code might normally lead to spurious
6343 warnings, however GCC's integrated preprocessor has enough context to
6344 avoid warning in these cases.
6347 A function declared external in one block and then used after the end of
6351 A @code{switch} statement has an operand of type @code{long}.
6354 A non-@code{static} function declaration follows a @code{static} one.
6355 This construct is not accepted by some traditional C compilers.
6358 The ISO type of an integer constant has a different width or
6359 signedness from its traditional type. This warning is only issued if
6360 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
6361 typically represent bit patterns, are not warned about.
6364 Usage of ISO string concatenation is detected.
6367 Initialization of automatic aggregates.
6370 Identifier conflicts with labels. Traditional C lacks a separate
6371 namespace for labels.
6374 Initialization of unions. If the initializer is zero, the warning is
6375 omitted. This is done under the assumption that the zero initializer in
6376 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
6377 initializer warnings and relies on default initialization to zero in the
6381 Conversions by prototypes between fixed/floating-point values and vice
6382 versa. The absence of these prototypes when compiling with traditional
6383 C causes serious problems. This is a subset of the possible
6384 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
6387 Use of ISO C style function definitions. This warning intentionally is
6388 @emph{not} issued for prototype declarations or variadic functions
6389 because these ISO C features appear in your code when using
6390 libiberty's traditional C compatibility macros, @code{PARAMS} and
6391 @code{VPARAMS}. This warning is also bypassed for nested functions
6392 because that feature is already a GCC extension and thus not relevant to
6393 traditional C compatibility.
6396 @item -Wtraditional-conversion @r{(C and Objective-C only)}
6397 @opindex Wtraditional-conversion
6398 @opindex Wno-traditional-conversion
6399 Warn if a prototype causes a type conversion that is different from what
6400 would happen to the same argument in the absence of a prototype. This
6401 includes conversions of fixed point to floating and vice versa, and
6402 conversions changing the width or signedness of a fixed-point argument
6403 except when the same as the default promotion.
6405 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
6406 @opindex Wdeclaration-after-statement
6407 @opindex Wno-declaration-after-statement
6408 Warn when a declaration is found after a statement in a block. This
6409 construct, known from C++, was introduced with ISO C99 and is by default
6410 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
6415 Warn whenever a local variable or type declaration shadows another
6416 variable, parameter, type, class member (in C++), or instance variable
6417 (in Objective-C) or whenever a built-in function is shadowed. Note
6418 that in C++, the compiler warns if a local variable shadows an
6419 explicit typedef, but not if it shadows a struct/class/enum.
6420 Same as @option{-Wshadow=global}.
6422 @item -Wno-shadow-ivar @r{(Objective-C only)}
6423 @opindex Wno-shadow-ivar
6424 @opindex Wshadow-ivar
6425 Do not warn whenever a local variable shadows an instance variable in an
6428 @item -Wshadow=global
6429 @opindex Wshadow=local
6430 The default for @option{-Wshadow}. Warns for any (global) shadowing.
6432 @item -Wshadow=local
6433 @opindex Wshadow=local
6434 Warn when a local variable shadows another local variable or parameter.
6435 This warning is enabled by @option{-Wshadow=global}.
6437 @item -Wshadow=compatible-local
6438 @opindex Wshadow=compatible-local
6439 Warn when a local variable shadows another local variable or parameter
6440 whose type is compatible with that of the shadowing variable. In C++,
6441 type compatibility here means the type of the shadowing variable can be
6442 converted to that of the shadowed variable. The creation of this flag
6443 (in addition to @option{-Wshadow=local}) is based on the idea that when
6444 a local variable shadows another one of incompatible type, it is most
6445 likely intentional, not a bug or typo, as shown in the following example:
6449 for (SomeIterator i = SomeObj.begin(); i != SomeObj.end(); ++i)
6451 for (int i = 0; i < N; ++i)
6460 Since the two variable @code{i} in the example above have incompatible types,
6461 enabling only @option{-Wshadow=compatible-local} will not emit a warning.
6462 Because their types are incompatible, if a programmer accidentally uses one
6463 in place of the other, type checking will catch that and emit an error or
6464 warning. So not warning (about shadowing) in this case will not lead to
6465 undetected bugs. Use of this flag instead of @option{-Wshadow=local} can
6466 possibly reduce the number of warnings triggered by intentional shadowing.
6468 This warning is enabled by @option{-Wshadow=local}.
6470 @item -Wlarger-than=@var{byte-size}
6471 @opindex Wlarger-than=
6472 @opindex Wlarger-than-@var{byte-size}
6473 Warn whenever an object is defined whose size exceeds @var{byte-size}.
6474 @option{-Wlarger-than=}@samp{PTRDIFF_MAX} is enabled by default.
6475 Warnings controlled by the option can be disabled either by specifying
6476 @var{byte-size} of @samp{SIZE_MAX} or more or by
6477 @option{-Wno-larger-than}.
6479 @item -Wno-larger-than
6480 @opindex Wno-larger-than
6481 Disable @option{-Wlarger-than=} warnings. The option is equivalent
6482 to @option{-Wlarger-than=}@samp{SIZE_MAX} or larger.
6484 @item -Wframe-larger-than=@var{byte-size}
6485 @opindex Wframe-larger-than=
6486 @opindex Wno-frame-larger-than
6487 Warn if the size of a function frame exceeds @var{byte-size}.
6488 The computation done to determine the stack frame size is approximate
6489 and not conservative.
6490 The actual requirements may be somewhat greater than @var{byte-size}
6491 even if you do not get a warning. In addition, any space allocated
6492 via @code{alloca}, variable-length arrays, or related constructs
6493 is not included by the compiler when determining
6494 whether or not to issue a warning.
6495 @option{-Wframe-larger-than=}@samp{PTRDIFF_MAX} is enabled by default.
6496 Warnings controlled by the option can be disabled either by specifying
6497 @var{byte-size} of @samp{SIZE_MAX} or more or by
6498 @option{-Wno-frame-larger-than}.
6500 @item -Wno-frame-larger-than
6501 @opindex Wno-frame-larger-than
6502 Disable @option{-Wframe-larger-than=} warnings. The option is equivalent
6503 to @option{-Wframe-larger-than=}@samp{SIZE_MAX} or larger.
6505 @item -Wno-free-nonheap-object
6506 @opindex Wno-free-nonheap-object
6507 @opindex Wfree-nonheap-object
6508 Do not warn when attempting to free an object that was not allocated
6511 @item -Wstack-usage=@var{byte-size}
6512 @opindex Wstack-usage
6513 @opindex Wno-stack-usage
6514 Warn if the stack usage of a function might exceed @var{byte-size}.
6515 The computation done to determine the stack usage is conservative.
6516 Any space allocated via @code{alloca}, variable-length arrays, or related
6517 constructs is included by the compiler when determining whether or not to
6520 The message is in keeping with the output of @option{-fstack-usage}.
6524 If the stack usage is fully static but exceeds the specified amount, it's:
6527 warning: stack usage is 1120 bytes
6530 If the stack usage is (partly) dynamic but bounded, it's:
6533 warning: stack usage might be 1648 bytes
6536 If the stack usage is (partly) dynamic and not bounded, it's:
6539 warning: stack usage might be unbounded
6543 @option{-Wstack-usage=}@samp{PTRDIFF_MAX} is enabled by default.
6544 Warnings controlled by the option can be disabled either by specifying
6545 @var{byte-size} of @samp{SIZE_MAX} or more or by
6546 @option{-Wno-stack-usage}.
6548 @item -Wno-stack-usage
6549 @opindex Wno-stack-usage
6550 Disable @option{-Wstack-usage=} warnings. The option is equivalent
6551 to @option{-Wstack-usage=}@samp{SIZE_MAX} or larger.
6553 @item -Wunsafe-loop-optimizations
6554 @opindex Wunsafe-loop-optimizations
6555 @opindex Wno-unsafe-loop-optimizations
6556 Warn if the loop cannot be optimized because the compiler cannot
6557 assume anything on the bounds of the loop indices. With
6558 @option{-funsafe-loop-optimizations} warn if the compiler makes
6561 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
6562 @opindex Wno-pedantic-ms-format
6563 @opindex Wpedantic-ms-format
6564 When used in combination with @option{-Wformat}
6565 and @option{-pedantic} without GNU extensions, this option
6566 disables the warnings about non-ISO @code{printf} / @code{scanf} format
6567 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
6568 which depend on the MS runtime.
6571 @opindex Waligned-new
6572 @opindex Wno-aligned-new
6573 Warn about a new-expression of a type that requires greater alignment
6574 than the @code{alignof(std::max_align_t)} but uses an allocation
6575 function without an explicit alignment parameter. This option is
6576 enabled by @option{-Wall}.
6578 Normally this only warns about global allocation functions, but
6579 @option{-Waligned-new=all} also warns about class member allocation
6582 @item -Wplacement-new
6583 @itemx -Wplacement-new=@var{n}
6584 @opindex Wplacement-new
6585 @opindex Wno-placement-new
6586 Warn about placement new expressions with undefined behavior, such as
6587 constructing an object in a buffer that is smaller than the type of
6588 the object. For example, the placement new expression below is diagnosed
6589 because it attempts to construct an array of 64 integers in a buffer only
6595 This warning is enabled by default.
6598 @item -Wplacement-new=1
6599 This is the default warning level of @option{-Wplacement-new}. At this
6600 level the warning is not issued for some strictly undefined constructs that
6601 GCC allows as extensions for compatibility with legacy code. For example,
6602 the following @code{new} expression is not diagnosed at this level even
6603 though it has undefined behavior according to the C++ standard because
6604 it writes past the end of the one-element array.
6606 struct S @{ int n, a[1]; @};
6607 S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]);
6608 new (s->a)int [32]();
6611 @item -Wplacement-new=2
6612 At this level, in addition to diagnosing all the same constructs as at level
6613 1, a diagnostic is also issued for placement new expressions that construct
6614 an object in the last member of structure whose type is an array of a single
6615 element and whose size is less than the size of the object being constructed.
6616 While the previous example would be diagnosed, the following construct makes
6617 use of the flexible member array extension to avoid the warning at level 2.
6619 struct S @{ int n, a[]; @};
6620 S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]);
6621 new (s->a)int [32]();
6626 @item -Wpointer-arith
6627 @opindex Wpointer-arith
6628 @opindex Wno-pointer-arith
6629 Warn about anything that depends on the ``size of'' a function type or
6630 of @code{void}. GNU C assigns these types a size of 1, for
6631 convenience in calculations with @code{void *} pointers and pointers
6632 to functions. In C++, warn also when an arithmetic operation involves
6633 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
6635 @item -Wpointer-compare
6636 @opindex Wpointer-compare
6637 @opindex Wno-pointer-compare
6638 Warn if a pointer is compared with a zero character constant. This usually
6639 means that the pointer was meant to be dereferenced. For example:
6642 const char *p = foo ();
6647 Note that the code above is invalid in C++11.
6649 This warning is enabled by default.
6652 @opindex Wtype-limits
6653 @opindex Wno-type-limits
6654 Warn if a comparison is always true or always false due to the limited
6655 range of the data type, but do not warn for constant expressions. For
6656 example, warn if an unsigned variable is compared against zero with
6657 @code{<} or @code{>=}. This warning is also enabled by
6660 @item -Wabsolute-value @r{(C and Objective-C only)}
6661 @opindex Wabsolute-value
6662 @opindex Wno-absolute-value
6663 Warn for calls to standard functions that compute the absolute value
6664 of an argument when a more appropriate standard function is available.
6665 For example, calling @code{abs(3.14)} triggers the warning because the
6666 appropriate function to call to compute the absolute value of a double
6667 argument is @code{fabs}. The option also triggers warnings when the
6668 argument in a call to such a function has an unsigned type. This
6669 warning can be suppressed with an explicit type cast and it is also
6670 enabled by @option{-Wextra}.
6672 @include cppwarnopts.texi
6674 @item -Wbad-function-cast @r{(C and Objective-C only)}
6675 @opindex Wbad-function-cast
6676 @opindex Wno-bad-function-cast
6677 Warn when a function call is cast to a non-matching type.
6678 For example, warn if a call to a function returning an integer type
6679 is cast to a pointer type.
6681 @item -Wc90-c99-compat @r{(C and Objective-C only)}
6682 @opindex Wc90-c99-compat
6683 @opindex Wno-c90-c99-compat
6684 Warn about features not present in ISO C90, but present in ISO C99.
6685 For instance, warn about use of variable length arrays, @code{long long}
6686 type, @code{bool} type, compound literals, designated initializers, and so
6687 on. This option is independent of the standards mode. Warnings are disabled
6688 in the expression that follows @code{__extension__}.
6690 @item -Wc99-c11-compat @r{(C and Objective-C only)}
6691 @opindex Wc99-c11-compat
6692 @opindex Wno-c99-c11-compat
6693 Warn about features not present in ISO C99, but present in ISO C11.
6694 For instance, warn about use of anonymous structures and unions,
6695 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
6696 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
6697 and so on. This option is independent of the standards mode. Warnings are
6698 disabled in the expression that follows @code{__extension__}.
6700 @item -Wc++-compat @r{(C and Objective-C only)}
6701 @opindex Wc++-compat
6702 @opindex Wno-c++-compat
6703 Warn about ISO C constructs that are outside of the common subset of
6704 ISO C and ISO C++, e.g.@: request for implicit conversion from
6705 @code{void *} to a pointer to non-@code{void} type.
6707 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
6708 @opindex Wc++11-compat
6709 @opindex Wno-c++11-compat
6710 Warn about C++ constructs whose meaning differs between ISO C++ 1998
6711 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
6712 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
6713 enabled by @option{-Wall}.
6715 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
6716 @opindex Wc++14-compat
6717 @opindex Wno-c++14-compat
6718 Warn about C++ constructs whose meaning differs between ISO C++ 2011
6719 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
6721 @item -Wc++17-compat @r{(C++ and Objective-C++ only)}
6722 @opindex Wc++17-compat
6723 @opindex Wno-c++17-compat
6724 Warn about C++ constructs whose meaning differs between ISO C++ 2014
6725 and ISO C++ 2017. This warning is enabled by @option{-Wall}.
6729 @opindex Wno-cast-qual
6730 Warn whenever a pointer is cast so as to remove a type qualifier from
6731 the target type. For example, warn if a @code{const char *} is cast
6732 to an ordinary @code{char *}.
6734 Also warn when making a cast that introduces a type qualifier in an
6735 unsafe way. For example, casting @code{char **} to @code{const char **}
6736 is unsafe, as in this example:
6739 /* p is char ** value. */
6740 const char **q = (const char **) p;
6741 /* Assignment of readonly string to const char * is OK. */
6743 /* Now char** pointer points to read-only memory. */
6748 @opindex Wcast-align
6749 @opindex Wno-cast-align
6750 Warn whenever a pointer is cast such that the required alignment of the
6751 target is increased. For example, warn if a @code{char *} is cast to
6752 an @code{int *} on machines where integers can only be accessed at
6753 two- or four-byte boundaries.
6755 @item -Wcast-align=strict
6756 @opindex Wcast-align=strict
6757 Warn whenever a pointer is cast such that the required alignment of the
6758 target is increased. For example, warn if a @code{char *} is cast to
6759 an @code{int *} regardless of the target machine.
6761 @item -Wcast-function-type
6762 @opindex Wcast-function-type
6763 @opindex Wno-cast-function-type
6764 Warn when a function pointer is cast to an incompatible function pointer.
6765 In a cast involving function types with a variable argument list only
6766 the types of initial arguments that are provided are considered.
6767 Any parameter of pointer-type matches any other pointer-type. Any benign
6768 differences in integral types are ignored, like @code{int} vs.@: @code{long}
6769 on ILP32 targets. Likewise type qualifiers are ignored. The function
6770 type @code{void (*) (void)} is special and matches everything, which can
6771 be used to suppress this warning.
6772 In a cast involving pointer to member types this warning warns whenever
6773 the type cast is changing the pointer to member type.
6774 This warning is enabled by @option{-Wextra}.
6776 @item -Wwrite-strings
6777 @opindex Wwrite-strings
6778 @opindex Wno-write-strings
6779 When compiling C, give string constants the type @code{const
6780 char[@var{length}]} so that copying the address of one into a
6781 non-@code{const} @code{char *} pointer produces a warning. These
6782 warnings help you find at compile time code that can try to write
6783 into a string constant, but only if you have been very careful about
6784 using @code{const} in declarations and prototypes. Otherwise, it is
6785 just a nuisance. This is why we did not make @option{-Wall} request
6788 When compiling C++, warn about the deprecated conversion from string
6789 literals to @code{char *}. This warning is enabled by default for C++
6793 @itemx -Wcatch-value=@var{n} @r{(C++ and Objective-C++ only)}
6794 @opindex Wcatch-value
6795 @opindex Wno-catch-value
6796 Warn about catch handlers that do not catch via reference.
6797 With @option{-Wcatch-value=1} (or @option{-Wcatch-value} for short)
6798 warn about polymorphic class types that are caught by value.
6799 With @option{-Wcatch-value=2} warn about all class types that are caught
6800 by value. With @option{-Wcatch-value=3} warn about all types that are
6801 not caught by reference. @option{-Wcatch-value} is enabled by @option{-Wall}.
6805 @opindex Wno-clobbered
6806 Warn for variables that might be changed by @code{longjmp} or
6807 @code{vfork}. This warning is also enabled by @option{-Wextra}.
6809 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
6810 @opindex Wconditionally-supported
6811 @opindex Wno-conditionally-supported
6812 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
6815 @opindex Wconversion
6816 @opindex Wno-conversion
6817 Warn for implicit conversions that may alter a value. This includes
6818 conversions between real and integer, like @code{abs (x)} when
6819 @code{x} is @code{double}; conversions between signed and unsigned,
6820 like @code{unsigned ui = -1}; and conversions to smaller types, like
6821 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
6822 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
6823 changed by the conversion like in @code{abs (2.0)}. Warnings about
6824 conversions between signed and unsigned integers can be disabled by
6825 using @option{-Wno-sign-conversion}.
6827 For C++, also warn for confusing overload resolution for user-defined
6828 conversions; and conversions that never use a type conversion
6829 operator: conversions to @code{void}, the same type, a base class or a
6830 reference to them. Warnings about conversions between signed and
6831 unsigned integers are disabled by default in C++ unless
6832 @option{-Wsign-conversion} is explicitly enabled.
6834 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
6835 @opindex Wconversion-null
6836 @opindex Wno-conversion-null
6837 Do not warn for conversions between @code{NULL} and non-pointer
6838 types. @option{-Wconversion-null} is enabled by default.
6840 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
6841 @opindex Wzero-as-null-pointer-constant
6842 @opindex Wno-zero-as-null-pointer-constant
6843 Warn when a literal @samp{0} is used as null pointer constant. This can
6844 be useful to facilitate the conversion to @code{nullptr} in C++11.
6846 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
6847 @opindex Wsubobject-linkage
6848 @opindex Wno-subobject-linkage
6849 Warn if a class type has a base or a field whose type uses the anonymous
6850 namespace or depends on a type with no linkage. If a type A depends on
6851 a type B with no or internal linkage, defining it in multiple
6852 translation units would be an ODR violation because the meaning of B
6853 is different in each translation unit. If A only appears in a single
6854 translation unit, the best way to silence the warning is to give it
6855 internal linkage by putting it in an anonymous namespace as well. The
6856 compiler doesn't give this warning for types defined in the main .C
6857 file, as those are unlikely to have multiple definitions.
6858 @option{-Wsubobject-linkage} is enabled by default.
6860 @item -Wdangling-else
6861 @opindex Wdangling-else
6862 @opindex Wno-dangling-else
6863 Warn about constructions where there may be confusion to which
6864 @code{if} statement an @code{else} branch belongs. Here is an example of
6879 In C/C++, every @code{else} branch belongs to the innermost possible
6880 @code{if} statement, which in this example is @code{if (b)}. This is
6881 often not what the programmer expected, as illustrated in the above
6882 example by indentation the programmer chose. When there is the
6883 potential for this confusion, GCC issues a warning when this flag
6884 is specified. To eliminate the warning, add explicit braces around
6885 the innermost @code{if} statement so there is no way the @code{else}
6886 can belong to the enclosing @code{if}. The resulting code
6903 This warning is enabled by @option{-Wparentheses}.
6907 @opindex Wno-date-time
6908 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
6909 are encountered as they might prevent bit-wise-identical reproducible
6912 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
6913 @opindex Wdelete-incomplete
6914 @opindex Wno-delete-incomplete
6915 Warn when deleting a pointer to incomplete type, which may cause
6916 undefined behavior at runtime. This warning is enabled by default.
6918 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
6919 @opindex Wuseless-cast
6920 @opindex Wno-useless-cast
6921 Warn when an expression is casted to its own type.
6924 @opindex Wempty-body
6925 @opindex Wno-empty-body
6926 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
6927 while} statement. This warning is also enabled by @option{-Wextra}.
6929 @item -Wenum-compare
6930 @opindex Wenum-compare
6931 @opindex Wno-enum-compare
6932 Warn about a comparison between values of different enumerated types.
6933 In C++ enumerated type mismatches in conditional expressions are also
6934 diagnosed and the warning is enabled by default. In C this warning is
6935 enabled by @option{-Wall}.
6937 @item -Wextra-semi @r{(C++, Objective-C++ only)}
6938 @opindex Wextra-semi
6939 @opindex Wno-extra-semi
6940 Warn about redundant semicolon after in-class function definition.
6942 @item -Wjump-misses-init @r{(C, Objective-C only)}
6943 @opindex Wjump-misses-init
6944 @opindex Wno-jump-misses-init
6945 Warn if a @code{goto} statement or a @code{switch} statement jumps
6946 forward across the initialization of a variable, or jumps backward to a
6947 label after the variable has been initialized. This only warns about
6948 variables that are initialized when they are declared. This warning is
6949 only supported for C and Objective-C; in C++ this sort of branch is an
6952 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
6953 can be disabled with the @option{-Wno-jump-misses-init} option.
6955 @item -Wsign-compare
6956 @opindex Wsign-compare
6957 @opindex Wno-sign-compare
6958 @cindex warning for comparison of signed and unsigned values
6959 @cindex comparison of signed and unsigned values, warning
6960 @cindex signed and unsigned values, comparison warning
6961 Warn when a comparison between signed and unsigned values could produce
6962 an incorrect result when the signed value is converted to unsigned.
6963 In C++, this warning is also enabled by @option{-Wall}. In C, it is
6964 also enabled by @option{-Wextra}.
6966 @item -Wsign-conversion
6967 @opindex Wsign-conversion
6968 @opindex Wno-sign-conversion
6969 Warn for implicit conversions that may change the sign of an integer
6970 value, like assigning a signed integer expression to an unsigned
6971 integer variable. An explicit cast silences the warning. In C, this
6972 option is enabled also by @option{-Wconversion}.
6974 @item -Wfloat-conversion
6975 @opindex Wfloat-conversion
6976 @opindex Wno-float-conversion
6977 Warn for implicit conversions that reduce the precision of a real value.
6978 This includes conversions from real to integer, and from higher precision
6979 real to lower precision real values. This option is also enabled by
6980 @option{-Wconversion}.
6982 @item -Wno-scalar-storage-order
6983 @opindex Wno-scalar-storage-order
6984 @opindex Wscalar-storage-order
6985 Do not warn on suspicious constructs involving reverse scalar storage order.
6987 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
6988 @opindex Wsized-deallocation
6989 @opindex Wno-sized-deallocation
6990 Warn about a definition of an unsized deallocation function
6992 void operator delete (void *) noexcept;
6993 void operator delete[] (void *) noexcept;
6995 without a definition of the corresponding sized deallocation function
6997 void operator delete (void *, std::size_t) noexcept;
6998 void operator delete[] (void *, std::size_t) noexcept;
7000 or vice versa. Enabled by @option{-Wextra} along with
7001 @option{-fsized-deallocation}.
7003 @item -Wsizeof-pointer-div
7004 @opindex Wsizeof-pointer-div
7005 @opindex Wno-sizeof-pointer-div
7006 Warn for suspicious divisions of two sizeof expressions that divide
7007 the pointer size by the element size, which is the usual way to compute
7008 the array size but won't work out correctly with pointers. This warning
7009 warns e.g.@: about @code{sizeof (ptr) / sizeof (ptr[0])} if @code{ptr} is
7010 not an array, but a pointer. This warning is enabled by @option{-Wall}.
7012 @item -Wsizeof-pointer-memaccess
7013 @opindex Wsizeof-pointer-memaccess
7014 @opindex Wno-sizeof-pointer-memaccess
7015 Warn for suspicious length parameters to certain string and memory built-in
7016 functions if the argument uses @code{sizeof}. This warning triggers for
7017 example for @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not
7018 an array, but a pointer, and suggests a possible fix, or about
7019 @code{memcpy (&foo, ptr, sizeof (&foo));}. @option{-Wsizeof-pointer-memaccess}
7020 also warns about calls to bounded string copy functions like @code{strncat}
7021 or @code{strncpy} that specify as the bound a @code{sizeof} expression of
7022 the source array. For example, in the following function the call to
7023 @code{strncat} specifies the size of the source string as the bound. That
7024 is almost certainly a mistake and so the call is diagnosed.
7026 void make_file (const char *name)
7028 char path[PATH_MAX];
7029 strncpy (path, name, sizeof path - 1);
7030 strncat (path, ".text", sizeof ".text");
7035 The @option{-Wsizeof-pointer-memaccess} option is enabled by @option{-Wall}.
7037 @item -Wsizeof-array-argument
7038 @opindex Wsizeof-array-argument
7039 @opindex Wno-sizeof-array-argument
7040 Warn when the @code{sizeof} operator is applied to a parameter that is
7041 declared as an array in a function definition. This warning is enabled by
7042 default for C and C++ programs.
7044 @item -Wmemset-elt-size
7045 @opindex Wmemset-elt-size
7046 @opindex Wno-memset-elt-size
7047 Warn for suspicious calls to the @code{memset} built-in function, if the
7048 first argument references an array, and the third argument is a number
7049 equal to the number of elements, but not equal to the size of the array
7050 in memory. This indicates that the user has omitted a multiplication by
7051 the element size. This warning is enabled by @option{-Wall}.
7053 @item -Wmemset-transposed-args
7054 @opindex Wmemset-transposed-args
7055 @opindex Wno-memset-transposed-args
7056 Warn for suspicious calls to the @code{memset} built-in function where
7057 the second argument is not zero and the third argument is zero. For
7058 example, the call @code{memset (buf, sizeof buf, 0)} is diagnosed because
7059 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostic
7060 is only emitted if the third argument is a literal zero. Otherwise, if
7061 it is an expression that is folded to zero, or a cast of zero to some
7062 type, it is far less likely that the arguments have been mistakenly
7063 transposed and no warning is emitted. This warning is enabled
7068 @opindex Wno-address
7069 Warn about suspicious uses of memory addresses. These include using
7070 the address of a function in a conditional expression, such as
7071 @code{void func(void); if (func)}, and comparisons against the memory
7072 address of a string literal, such as @code{if (x == "abc")}. Such
7073 uses typically indicate a programmer error: the address of a function
7074 always evaluates to true, so their use in a conditional usually
7075 indicate that the programmer forgot the parentheses in a function
7076 call; and comparisons against string literals result in unspecified
7077 behavior and are not portable in C, so they usually indicate that the
7078 programmer intended to use @code{strcmp}. This warning is enabled by
7081 @item -Waddress-of-packed-member
7082 @opindex Waddress-of-packed-member
7083 @opindex Wno-address-of-packed-member
7084 Warn when the address of packed member of struct or union is taken,
7085 which usually results in an unaligned pointer value. This is
7089 @opindex Wlogical-op
7090 @opindex Wno-logical-op
7091 Warn about suspicious uses of logical operators in expressions.
7092 This includes using logical operators in contexts where a
7093 bit-wise operator is likely to be expected. Also warns when
7094 the operands of a logical operator are the same:
7097 if (a < 0 && a < 0) @{ @dots{} @}
7100 @item -Wlogical-not-parentheses
7101 @opindex Wlogical-not-parentheses
7102 @opindex Wno-logical-not-parentheses
7103 Warn about logical not used on the left hand side operand of a comparison.
7104 This option does not warn if the right operand is considered to be a boolean
7105 expression. Its purpose is to detect suspicious code like the following:
7109 if (!a > 1) @{ @dots{} @}
7112 It is possible to suppress the warning by wrapping the LHS into
7115 if ((!a) > 1) @{ @dots{} @}
7118 This warning is enabled by @option{-Wall}.
7120 @item -Waggregate-return
7121 @opindex Waggregate-return
7122 @opindex Wno-aggregate-return
7123 Warn if any functions that return structures or unions are defined or
7124 called. (In languages where you can return an array, this also elicits
7127 @item -Wno-aggressive-loop-optimizations
7128 @opindex Wno-aggressive-loop-optimizations
7129 @opindex Waggressive-loop-optimizations
7130 Warn if in a loop with constant number of iterations the compiler detects
7131 undefined behavior in some statement during one or more of the iterations.
7133 @item -Wno-attributes
7134 @opindex Wno-attributes
7135 @opindex Wattributes
7136 Do not warn if an unexpected @code{__attribute__} is used, such as
7137 unrecognized attributes, function attributes applied to variables,
7138 etc. This does not stop errors for incorrect use of supported
7141 @item -Wno-builtin-declaration-mismatch
7142 @opindex Wno-builtin-declaration-mismatch
7143 @opindex Wbuiltin-declaration-mismatch
7144 Warn if a built-in function is declared with an incompatible signature
7145 or as a non-function, or when a built-in function declared with a type
7146 that does not include a prototype is called with arguments whose promoted
7147 types do not match those expected by the function. When @option{-Wextra}
7148 is specified, also warn when a built-in function that takes arguments is
7149 declared without a prototype. The @option{-Wno-builtin-declaration-mismatch}
7150 warning is enabled by default. To avoid the warning include the appropriate
7151 header to bring the prototypes of built-in functions into scope.
7153 For example, the call to @code{memset} below is diagnosed by the warning
7154 because the function expects a value of type @code{size_t} as its argument
7155 but the type of @code{32} is @code{int}. With @option{-Wextra},
7156 the declaration of the function is diagnosed as well.
7158 extern void* memset ();
7161 memset (d, '\0', 32);
7165 @item -Wno-builtin-macro-redefined
7166 @opindex Wno-builtin-macro-redefined
7167 @opindex Wbuiltin-macro-redefined
7168 Do not warn if certain built-in macros are redefined. This suppresses
7169 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
7170 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
7172 @item -Wstrict-prototypes @r{(C and Objective-C only)}
7173 @opindex Wstrict-prototypes
7174 @opindex Wno-strict-prototypes
7175 Warn if a function is declared or defined without specifying the
7176 argument types. (An old-style function definition is permitted without
7177 a warning if preceded by a declaration that specifies the argument
7180 @item -Wold-style-declaration @r{(C and Objective-C only)}
7181 @opindex Wold-style-declaration
7182 @opindex Wno-old-style-declaration
7183 Warn for obsolescent usages, according to the C Standard, in a
7184 declaration. For example, warn if storage-class specifiers like
7185 @code{static} are not the first things in a declaration. This warning
7186 is also enabled by @option{-Wextra}.
7188 @item -Wold-style-definition @r{(C and Objective-C only)}
7189 @opindex Wold-style-definition
7190 @opindex Wno-old-style-definition
7191 Warn if an old-style function definition is used. A warning is given
7192 even if there is a previous prototype.
7194 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
7195 @opindex Wmissing-parameter-type
7196 @opindex Wno-missing-parameter-type
7197 A function parameter is declared without a type specifier in K&R-style
7204 This warning is also enabled by @option{-Wextra}.
7206 @item -Wmissing-prototypes @r{(C and Objective-C only)}
7207 @opindex Wmissing-prototypes
7208 @opindex Wno-missing-prototypes
7209 Warn if a global function is defined without a previous prototype
7210 declaration. This warning is issued even if the definition itself
7211 provides a prototype. Use this option to detect global functions
7212 that do not have a matching prototype declaration in a header file.
7213 This option is not valid for C++ because all function declarations
7214 provide prototypes and a non-matching declaration declares an
7215 overload rather than conflict with an earlier declaration.
7216 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
7218 @item -Wmissing-declarations
7219 @opindex Wmissing-declarations
7220 @opindex Wno-missing-declarations
7221 Warn if a global function is defined without a previous declaration.
7222 Do so even if the definition itself provides a prototype.
7223 Use this option to detect global functions that are not declared in
7224 header files. In C, no warnings are issued for functions with previous
7225 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
7226 missing prototypes. In C++, no warnings are issued for function templates,
7227 or for inline functions, or for functions in anonymous namespaces.
7229 @item -Wmissing-field-initializers
7230 @opindex Wmissing-field-initializers
7231 @opindex Wno-missing-field-initializers
7235 Warn if a structure's initializer has some fields missing. For
7236 example, the following code causes such a warning, because
7237 @code{x.h} is implicitly zero:
7240 struct s @{ int f, g, h; @};
7241 struct s x = @{ 3, 4 @};
7244 This option does not warn about designated initializers, so the following
7245 modification does not trigger a warning:
7248 struct s @{ int f, g, h; @};
7249 struct s x = @{ .f = 3, .g = 4 @};
7252 In C this option does not warn about the universal zero initializer
7256 struct s @{ int f, g, h; @};
7257 struct s x = @{ 0 @};
7260 Likewise, in C++ this option does not warn about the empty @{ @}
7261 initializer, for example:
7264 struct s @{ int f, g, h; @};
7268 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
7269 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
7271 @item -Wno-multichar
7272 @opindex Wno-multichar
7274 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
7275 Usually they indicate a typo in the user's code, as they have
7276 implementation-defined values, and should not be used in portable code.
7278 @item -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]}
7279 @opindex Wnormalized=
7280 @opindex Wnormalized
7281 @opindex Wno-normalized
7284 @cindex character set, input normalization
7285 In ISO C and ISO C++, two identifiers are different if they are
7286 different sequences of characters. However, sometimes when characters
7287 outside the basic ASCII character set are used, you can have two
7288 different character sequences that look the same. To avoid confusion,
7289 the ISO 10646 standard sets out some @dfn{normalization rules} which
7290 when applied ensure that two sequences that look the same are turned into
7291 the same sequence. GCC can warn you if you are using identifiers that
7292 have not been normalized; this option controls that warning.
7294 There are four levels of warning supported by GCC@. The default is
7295 @option{-Wnormalized=nfc}, which warns about any identifier that is
7296 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
7297 recommended form for most uses. It is equivalent to
7298 @option{-Wnormalized}.
7300 Unfortunately, there are some characters allowed in identifiers by
7301 ISO C and ISO C++ that, when turned into NFC, are not allowed in
7302 identifiers. That is, there's no way to use these symbols in portable
7303 ISO C or C++ and have all your identifiers in NFC@.
7304 @option{-Wnormalized=id} suppresses the warning for these characters.
7305 It is hoped that future versions of the standards involved will correct
7306 this, which is why this option is not the default.
7308 You can switch the warning off for all characters by writing
7309 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
7310 only do this if you are using some other normalization scheme (like
7311 ``D''), because otherwise you can easily create bugs that are
7312 literally impossible to see.
7314 Some characters in ISO 10646 have distinct meanings but look identical
7315 in some fonts or display methodologies, especially once formatting has
7316 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
7317 LETTER N'', displays just like a regular @code{n} that has been
7318 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
7319 normalization scheme to convert all these into a standard form as
7320 well, and GCC warns if your code is not in NFKC if you use
7321 @option{-Wnormalized=nfkc}. This warning is comparable to warning
7322 about every identifier that contains the letter O because it might be
7323 confused with the digit 0, and so is not the default, but may be
7324 useful as a local coding convention if the programming environment
7325 cannot be fixed to display these characters distinctly.
7327 @item -Wno-attribute-warning
7328 @opindex Wno-attribute-warning
7329 @opindex Wattribute-warning
7330 Do not warn about usage of functions (@pxref{Function Attributes})
7331 declared with @code{warning} attribute. By default, this warning is
7332 enabled. @option{-Wno-attribute-warning} can be used to disable the
7333 warning or @option{-Wno-error=attribute-warning} can be used to
7334 disable the error when compiled with @option{-Werror} flag.
7336 @item -Wno-deprecated
7337 @opindex Wno-deprecated
7338 @opindex Wdeprecated
7339 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
7341 @item -Wno-deprecated-declarations
7342 @opindex Wno-deprecated-declarations
7343 @opindex Wdeprecated-declarations
7344 Do not warn about uses of functions (@pxref{Function Attributes}),
7345 variables (@pxref{Variable Attributes}), and types (@pxref{Type
7346 Attributes}) marked as deprecated by using the @code{deprecated}
7350 @opindex Wno-overflow
7352 Do not warn about compile-time overflow in constant expressions.
7357 Warn about One Definition Rule violations during link-time optimization.
7358 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
7361 @opindex Wopenmp-simd
7362 @opindex Wno-openmp-simd
7363 Warn if the vectorizer cost model overrides the OpenMP
7364 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
7365 option can be used to relax the cost model.
7367 @item -Woverride-init @r{(C and Objective-C only)}
7368 @opindex Woverride-init
7369 @opindex Wno-override-init
7373 Warn if an initialized field without side effects is overridden when
7374 using designated initializers (@pxref{Designated Inits, , Designated
7377 This warning is included in @option{-Wextra}. To get other
7378 @option{-Wextra} warnings without this one, use @option{-Wextra
7379 -Wno-override-init}.
7381 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
7382 @opindex Woverride-init-side-effects
7383 @opindex Wno-override-init-side-effects
7384 Warn if an initialized field with side effects is overridden when
7385 using designated initializers (@pxref{Designated Inits, , Designated
7386 Initializers}). This warning is enabled by default.
7391 Warn if a structure is given the packed attribute, but the packed
7392 attribute has no effect on the layout or size of the structure.
7393 Such structures may be mis-aligned for little benefit. For
7394 instance, in this code, the variable @code{f.x} in @code{struct bar}
7395 is misaligned even though @code{struct bar} does not itself
7396 have the packed attribute:
7403 @} __attribute__((packed));
7411 @item -Wpacked-bitfield-compat
7412 @opindex Wpacked-bitfield-compat
7413 @opindex Wno-packed-bitfield-compat
7414 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
7415 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
7416 the change can lead to differences in the structure layout. GCC
7417 informs you when the offset of such a field has changed in GCC 4.4.
7418 For example there is no longer a 4-bit padding between field @code{a}
7419 and @code{b} in this structure:
7426 @} __attribute__ ((packed));
7429 This warning is enabled by default. Use
7430 @option{-Wno-packed-bitfield-compat} to disable this warning.
7432 @item -Wpacked-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
7433 @opindex Wpacked-not-aligned
7434 @opindex Wno-packed-not-aligned
7435 Warn if a structure field with explicitly specified alignment in a
7436 packed struct or union is misaligned. For example, a warning will
7437 be issued on @code{struct S}, like, @code{warning: alignment 1 of
7438 'struct S' is less than 8}, in this code:
7442 struct __attribute__ ((aligned (8))) S8 @{ char a[8]; @};
7443 struct __attribute__ ((packed)) S @{
7449 This warning is enabled by @option{-Wall}.
7454 Warn if padding is included in a structure, either to align an element
7455 of the structure or to align the whole structure. Sometimes when this
7456 happens it is possible to rearrange the fields of the structure to
7457 reduce the padding and so make the structure smaller.
7459 @item -Wredundant-decls
7460 @opindex Wredundant-decls
7461 @opindex Wno-redundant-decls
7462 Warn if anything is declared more than once in the same scope, even in
7463 cases where multiple declaration is valid and changes nothing.
7467 @opindex Wno-restrict
7468 Warn when an object referenced by a @code{restrict}-qualified parameter
7469 (or, in C++, a @code{__restrict}-qualified parameter) is aliased by another
7470 argument, or when copies between such objects overlap. For example,
7471 the call to the @code{strcpy} function below attempts to truncate the string
7472 by replacing its initial characters with the last four. However, because
7473 the call writes the terminating NUL into @code{a[4]}, the copies overlap and
7474 the call is diagnosed.
7479 char a[] = "abcd1234";
7484 The @option{-Wrestrict} option detects some instances of simple overlap
7485 even without optimization but works best at @option{-O2} and above. It
7486 is included in @option{-Wall}.
7488 @item -Wnested-externs @r{(C and Objective-C only)}
7489 @opindex Wnested-externs
7490 @opindex Wno-nested-externs
7491 Warn if an @code{extern} declaration is encountered within a function.
7493 @item -Wno-inherited-variadic-ctor
7494 @opindex Winherited-variadic-ctor
7495 @opindex Wno-inherited-variadic-ctor
7496 Suppress warnings about use of C++11 inheriting constructors when the
7497 base class inherited from has a C variadic constructor; the warning is
7498 on by default because the ellipsis is not inherited.
7503 Warn if a function that is declared as inline cannot be inlined.
7504 Even with this option, the compiler does not warn about failures to
7505 inline functions declared in system headers.
7507 The compiler uses a variety of heuristics to determine whether or not
7508 to inline a function. For example, the compiler takes into account
7509 the size of the function being inlined and the amount of inlining
7510 that has already been done in the current function. Therefore,
7511 seemingly insignificant changes in the source program can cause the
7512 warnings produced by @option{-Winline} to appear or disappear.
7514 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
7515 @opindex Wno-invalid-offsetof
7516 @opindex Winvalid-offsetof
7517 Suppress warnings from applying the @code{offsetof} macro to a non-POD
7518 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
7519 to a non-standard-layout type is undefined. In existing C++ implementations,
7520 however, @code{offsetof} typically gives meaningful results.
7521 This flag is for users who are aware that they are
7522 writing nonportable code and who have deliberately chosen to ignore the
7525 The restrictions on @code{offsetof} may be relaxed in a future version
7526 of the C++ standard.
7528 @item -Wint-in-bool-context
7529 @opindex Wint-in-bool-context
7530 @opindex Wno-int-in-bool-context
7531 Warn for suspicious use of integer values where boolean values are expected,
7532 such as conditional expressions (?:) using non-boolean integer constants in
7533 boolean context, like @code{if (a <= b ? 2 : 3)}. Or left shifting of signed
7534 integers in boolean context, like @code{for (a = 0; 1 << a; a++);}. Likewise
7535 for all kinds of multiplications regardless of the data type.
7536 This warning is enabled by @option{-Wall}.
7538 @item -Wno-int-to-pointer-cast
7539 @opindex Wno-int-to-pointer-cast
7540 @opindex Wint-to-pointer-cast
7541 Suppress warnings from casts to pointer type of an integer of a
7542 different size. In C++, casting to a pointer type of smaller size is
7543 an error. @option{Wint-to-pointer-cast} is enabled by default.
7546 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
7547 @opindex Wno-pointer-to-int-cast
7548 @opindex Wpointer-to-int-cast
7549 Suppress warnings from casts from a pointer to an integer type of a
7553 @opindex Winvalid-pch
7554 @opindex Wno-invalid-pch
7555 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
7556 the search path but cannot be used.
7560 @opindex Wno-long-long
7561 Warn if @code{long long} type is used. This is enabled by either
7562 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
7563 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
7565 @item -Wvariadic-macros
7566 @opindex Wvariadic-macros
7567 @opindex Wno-variadic-macros
7568 Warn if variadic macros are used in ISO C90 mode, or if the GNU
7569 alternate syntax is used in ISO C99 mode. This is enabled by either
7570 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
7571 messages, use @option{-Wno-variadic-macros}.
7575 @opindex Wno-varargs
7576 Warn upon questionable usage of the macros used to handle variable
7577 arguments like @code{va_start}. This is default. To inhibit the
7578 warning messages, use @option{-Wno-varargs}.
7580 @item -Wvector-operation-performance
7581 @opindex Wvector-operation-performance
7582 @opindex Wno-vector-operation-performance
7583 Warn if vector operation is not implemented via SIMD capabilities of the
7584 architecture. Mainly useful for the performance tuning.
7585 Vector operation can be implemented @code{piecewise}, which means that the
7586 scalar operation is performed on every vector element;
7587 @code{in parallel}, which means that the vector operation is implemented
7588 using scalars of wider type, which normally is more performance efficient;
7589 and @code{as a single scalar}, which means that vector fits into a
7592 @item -Wno-virtual-move-assign
7593 @opindex Wvirtual-move-assign
7594 @opindex Wno-virtual-move-assign
7595 Suppress warnings about inheriting from a virtual base with a
7596 non-trivial C++11 move assignment operator. This is dangerous because
7597 if the virtual base is reachable along more than one path, it is
7598 moved multiple times, which can mean both objects end up in the
7599 moved-from state. If the move assignment operator is written to avoid
7600 moving from a moved-from object, this warning can be disabled.
7605 Warn if a variable-length array is used in the code.
7606 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
7607 the variable-length array.
7609 @item -Wvla-larger-than=@var{byte-size}
7610 @opindex Wvla-larger-than=
7611 @opindex Wno-vla-larger-than
7612 If this option is used, the compiler will warn for declarations of
7613 variable-length arrays whose size is either unbounded, or bounded
7614 by an argument that allows the array size to exceed @var{byte-size}
7615 bytes. This is similar to how @option{-Walloca-larger-than=}@var{byte-size}
7616 works, but with variable-length arrays.
7618 Note that GCC may optimize small variable-length arrays of a known
7619 value into plain arrays, so this warning may not get triggered for
7622 @option{-Wvla-larger-than=}@samp{PTRDIFF_MAX} is enabled by default but
7623 is typically only effective when @option{-ftree-vrp} is active (default
7624 for @option{-O2} and above).
7626 See also @option{-Walloca-larger-than=@var{byte-size}}.
7628 @item -Wno-vla-larger-than
7629 @opindex Wno-vla-larger-than
7630 Disable @option{-Wvla-larger-than=} warnings. The option is equivalent
7631 to @option{-Wvla-larger-than=}@samp{SIZE_MAX} or larger.
7633 @item -Wvolatile-register-var
7634 @opindex Wvolatile-register-var
7635 @opindex Wno-volatile-register-var
7636 Warn if a register variable is declared volatile. The volatile
7637 modifier does not inhibit all optimizations that may eliminate reads
7638 and/or writes to register variables. This warning is enabled by
7641 @item -Wdisabled-optimization
7642 @opindex Wdisabled-optimization
7643 @opindex Wno-disabled-optimization
7644 Warn if a requested optimization pass is disabled. This warning does
7645 not generally indicate that there is anything wrong with your code; it
7646 merely indicates that GCC's optimizers are unable to handle the code
7647 effectively. Often, the problem is that your code is too big or too
7648 complex; GCC refuses to optimize programs when the optimization
7649 itself is likely to take inordinate amounts of time.
7651 @item -Wpointer-sign @r{(C and Objective-C only)}
7652 @opindex Wpointer-sign
7653 @opindex Wno-pointer-sign
7654 Warn for pointer argument passing or assignment with different signedness.
7655 This option is only supported for C and Objective-C@. It is implied by
7656 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
7657 @option{-Wno-pointer-sign}.
7659 @item -Wstack-protector
7660 @opindex Wstack-protector
7661 @opindex Wno-stack-protector
7662 This option is only active when @option{-fstack-protector} is active. It
7663 warns about functions that are not protected against stack smashing.
7665 @item -Woverlength-strings
7666 @opindex Woverlength-strings
7667 @opindex Wno-overlength-strings
7668 Warn about string constants that are longer than the ``minimum
7669 maximum'' length specified in the C standard. Modern compilers
7670 generally allow string constants that are much longer than the
7671 standard's minimum limit, but very portable programs should avoid
7672 using longer strings.
7674 The limit applies @emph{after} string constant concatenation, and does
7675 not count the trailing NUL@. In C90, the limit was 509 characters; in
7676 C99, it was raised to 4095. C++98 does not specify a normative
7677 minimum maximum, so we do not diagnose overlength strings in C++@.
7679 This option is implied by @option{-Wpedantic}, and can be disabled with
7680 @option{-Wno-overlength-strings}.
7682 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
7683 @opindex Wunsuffixed-float-constants
7684 @opindex Wno-unsuffixed-float-constants
7686 Issue a warning for any floating constant that does not have
7687 a suffix. When used together with @option{-Wsystem-headers} it
7688 warns about such constants in system header files. This can be useful
7689 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
7690 from the decimal floating-point extension to C99.
7692 @item -Wno-designated-init @r{(C and Objective-C only)}
7693 Suppress warnings when a positional initializer is used to initialize
7694 a structure that has been marked with the @code{designated_init}
7698 Issue a warning when HSAIL cannot be emitted for the compiled function or
7703 @node Debugging Options
7704 @section Options for Debugging Your Program
7705 @cindex options, debugging
7706 @cindex debugging information options
7708 To tell GCC to emit extra information for use by a debugger, in almost
7709 all cases you need only to add @option{-g} to your other options.
7711 GCC allows you to use @option{-g} with
7712 @option{-O}. The shortcuts taken by optimized code may occasionally
7713 be surprising: some variables you declared may not exist
7714 at all; flow of control may briefly move where you did not expect it;
7715 some statements may not be executed because they compute constant
7716 results or their values are already at hand; some statements may
7717 execute in different places because they have been moved out of loops.
7718 Nevertheless it is possible to debug optimized output. This makes
7719 it reasonable to use the optimizer for programs that might have bugs.
7721 If you are not using some other optimization option, consider
7722 using @option{-Og} (@pxref{Optimize Options}) with @option{-g}.
7723 With no @option{-O} option at all, some compiler passes that collect
7724 information useful for debugging do not run at all, so that
7725 @option{-Og} may result in a better debugging experience.
7730 Produce debugging information in the operating system's native format
7731 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
7734 On most systems that use stabs format, @option{-g} enables use of extra
7735 debugging information that only GDB can use; this extra information
7736 makes debugging work better in GDB but probably makes other debuggers
7738 refuse to read the program. If you want to control for certain whether
7739 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
7740 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
7744 Produce debugging information for use by GDB@. This means to use the
7745 most expressive format available (DWARF, stabs, or the native format
7746 if neither of those are supported), including GDB extensions if at all
7750 @itemx -gdwarf-@var{version}
7752 Produce debugging information in DWARF format (if that is supported).
7753 The value of @var{version} may be either 2, 3, 4 or 5; the default version
7754 for most targets is 4. DWARF Version 5 is only experimental.
7756 Note that with DWARF Version 2, some ports require and always
7757 use some non-conflicting DWARF 3 extensions in the unwind tables.
7759 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
7760 for maximum benefit.
7762 GCC no longer supports DWARF Version 1, which is substantially
7763 different than Version 2 and later. For historical reasons, some
7764 other DWARF-related options such as
7765 @option{-fno-dwarf2-cfi-asm}) retain a reference to DWARF Version 2
7766 in their names, but apply to all currently-supported versions of DWARF.
7770 Produce debugging information in stabs format (if that is supported),
7771 without GDB extensions. This is the format used by DBX on most BSD
7772 systems. On MIPS, Alpha and System V Release 4 systems this option
7773 produces stabs debugging output that is not understood by DBX@.
7774 On System V Release 4 systems this option requires the GNU assembler.
7778 Produce debugging information in stabs format (if that is supported),
7779 using GNU extensions understood only by the GNU debugger (GDB)@. The
7780 use of these extensions is likely to make other debuggers crash or
7781 refuse to read the program.
7785 Produce debugging information in XCOFF format (if that is supported).
7786 This is the format used by the DBX debugger on IBM RS/6000 systems.
7790 Produce debugging information in XCOFF format (if that is supported),
7791 using GNU extensions understood only by the GNU debugger (GDB)@. The
7792 use of these extensions is likely to make other debuggers crash or
7793 refuse to read the program, and may cause assemblers other than the GNU
7794 assembler (GAS) to fail with an error.
7798 Produce debugging information in Alpha/VMS debug format (if that is
7799 supported). This is the format used by DEBUG on Alpha/VMS systems.
7802 @itemx -ggdb@var{level}
7803 @itemx -gstabs@var{level}
7804 @itemx -gxcoff@var{level}
7805 @itemx -gvms@var{level}
7806 Request debugging information and also use @var{level} to specify how
7807 much information. The default level is 2.
7809 Level 0 produces no debug information at all. Thus, @option{-g0} negates
7812 Level 1 produces minimal information, enough for making backtraces in
7813 parts of the program that you don't plan to debug. This includes
7814 descriptions of functions and external variables, and line number
7815 tables, but no information about local variables.
7817 Level 3 includes extra information, such as all the macro definitions
7818 present in the program. Some debuggers support macro expansion when
7819 you use @option{-g3}.
7821 If you use multiple @option{-g} options, with or without level numbers,
7822 the last such option is the one that is effective.
7824 @option{-gdwarf} does not accept a concatenated debug level, to avoid
7825 confusion with @option{-gdwarf-@var{level}}.
7826 Instead use an additional @option{-g@var{level}} option to change the
7827 debug level for DWARF.
7829 @item -feliminate-unused-debug-symbols
7830 @opindex feliminate-unused-debug-symbols
7831 Produce debugging information in stabs format (if that is supported),
7832 for only symbols that are actually used.
7834 @item -femit-class-debug-always
7835 @opindex femit-class-debug-always
7836 Instead of emitting debugging information for a C++ class in only one
7837 object file, emit it in all object files using the class. This option
7838 should be used only with debuggers that are unable to handle the way GCC
7839 normally emits debugging information for classes because using this
7840 option increases the size of debugging information by as much as a
7843 @item -fno-merge-debug-strings
7844 @opindex fmerge-debug-strings
7845 @opindex fno-merge-debug-strings
7846 Direct the linker to not merge together strings in the debugging
7847 information that are identical in different object files. Merging is
7848 not supported by all assemblers or linkers. Merging decreases the size
7849 of the debug information in the output file at the cost of increasing
7850 link processing time. Merging is enabled by default.
7852 @item -fdebug-prefix-map=@var{old}=@var{new}
7853 @opindex fdebug-prefix-map
7854 When compiling files residing in directory @file{@var{old}}, record
7855 debugging information describing them as if the files resided in
7856 directory @file{@var{new}} instead. This can be used to replace a
7857 build-time path with an install-time path in the debug info. It can
7858 also be used to change an absolute path to a relative path by using
7859 @file{.} for @var{new}. This can give more reproducible builds, which
7860 are location independent, but may require an extra command to tell GDB
7861 where to find the source files. See also @option{-ffile-prefix-map}.
7863 @item -fvar-tracking
7864 @opindex fvar-tracking
7865 Run variable tracking pass. It computes where variables are stored at each
7866 position in code. Better debugging information is then generated
7867 (if the debugging information format supports this information).
7869 It is enabled by default when compiling with optimization (@option{-Os},
7870 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7871 the debug info format supports it.
7873 @item -fvar-tracking-assignments
7874 @opindex fvar-tracking-assignments
7875 @opindex fno-var-tracking-assignments
7876 Annotate assignments to user variables early in the compilation and
7877 attempt to carry the annotations over throughout the compilation all the
7878 way to the end, in an attempt to improve debug information while
7879 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7881 It can be enabled even if var-tracking is disabled, in which case
7882 annotations are created and maintained, but discarded at the end.
7883 By default, this flag is enabled together with @option{-fvar-tracking},
7884 except when selective scheduling is enabled.
7887 @opindex gsplit-dwarf
7888 Separate as much DWARF debugging information as possible into a
7889 separate output file with the extension @file{.dwo}. This option allows
7890 the build system to avoid linking files with debug information. To
7891 be useful, this option requires a debugger capable of reading @file{.dwo}
7894 @item -gdescribe-dies
7895 @opindex gdescribe-dies
7896 Add description attributes to some DWARF DIEs that have no name attribute,
7897 such as artificial variables, external references and call site
7902 Generate DWARF @code{.debug_pubnames} and @code{.debug_pubtypes} sections.
7904 @item -ggnu-pubnames
7905 @opindex ggnu-pubnames
7906 Generate @code{.debug_pubnames} and @code{.debug_pubtypes} sections in a format
7907 suitable for conversion into a GDB@ index. This option is only useful
7908 with a linker that can produce GDB@ index version 7.
7910 @item -fdebug-types-section
7911 @opindex fdebug-types-section
7912 @opindex fno-debug-types-section
7913 When using DWARF Version 4 or higher, type DIEs can be put into
7914 their own @code{.debug_types} section instead of making them part of the
7915 @code{.debug_info} section. It is more efficient to put them in a separate
7916 comdat section since the linker can then remove duplicates.
7917 But not all DWARF consumers support @code{.debug_types} sections yet
7918 and on some objects @code{.debug_types} produces larger instead of smaller
7919 debugging information.
7921 @item -grecord-gcc-switches
7922 @itemx -gno-record-gcc-switches
7923 @opindex grecord-gcc-switches
7924 @opindex gno-record-gcc-switches
7925 This switch causes the command-line options used to invoke the
7926 compiler that may affect code generation to be appended to the
7927 DW_AT_producer attribute in DWARF debugging information. The options
7928 are concatenated with spaces separating them from each other and from
7929 the compiler version.
7930 It is enabled by default.
7931 See also @option{-frecord-gcc-switches} for another
7932 way of storing compiler options into the object file.
7934 @item -gstrict-dwarf
7935 @opindex gstrict-dwarf
7936 Disallow using extensions of later DWARF standard version than selected
7937 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
7938 DWARF extensions from later standard versions is allowed.
7940 @item -gno-strict-dwarf
7941 @opindex gno-strict-dwarf
7942 Allow using extensions of later DWARF standard version than selected with
7943 @option{-gdwarf-@var{version}}.
7945 @item -gas-loc-support
7946 @opindex gas-loc-support
7947 Inform the compiler that the assembler supports @code{.loc} directives.
7948 It may then use them for the assembler to generate DWARF2+ line number
7951 This is generally desirable, because assembler-generated line-number
7952 tables are a lot more compact than those the compiler can generate
7955 This option will be enabled by default if, at GCC configure time, the
7956 assembler was found to support such directives.
7958 @item -gno-as-loc-support
7959 @opindex gno-as-loc-support
7960 Force GCC to generate DWARF2+ line number tables internally, if DWARF2+
7961 line number tables are to be generated.
7963 @item gas-locview-support
7964 @opindex gas-locview-support
7965 Inform the compiler that the assembler supports @code{view} assignment
7966 and reset assertion checking in @code{.loc} directives.
7968 This option will be enabled by default if, at GCC configure time, the
7969 assembler was found to support them.
7971 @item gno-as-locview-support
7972 Force GCC to assign view numbers internally, if
7973 @option{-gvariable-location-views} are explicitly requested.
7976 @itemx -gno-column-info
7977 @opindex gcolumn-info
7978 @opindex gno-column-info
7979 Emit location column information into DWARF debugging information, rather
7980 than just file and line.
7981 This option is enabled by default.
7983 @item -gstatement-frontiers
7984 @itemx -gno-statement-frontiers
7985 @opindex gstatement-frontiers
7986 @opindex gno-statement-frontiers
7987 This option causes GCC to create markers in the internal representation
7988 at the beginning of statements, and to keep them roughly in place
7989 throughout compilation, using them to guide the output of @code{is_stmt}
7990 markers in the line number table. This is enabled by default when
7991 compiling with optimization (@option{-Os}, @option{-O}, @option{-O2},
7992 @dots{}), and outputting DWARF 2 debug information at the normal level.
7994 @item -gvariable-location-views
7995 @itemx -gvariable-location-views=incompat5
7996 @itemx -gno-variable-location-views
7997 @opindex gvariable-location-views
7998 @opindex gvariable-location-views=incompat5
7999 @opindex gno-variable-location-views
8000 Augment variable location lists with progressive view numbers implied
8001 from the line number table. This enables debug information consumers to
8002 inspect state at certain points of the program, even if no instructions
8003 associated with the corresponding source locations are present at that
8004 point. If the assembler lacks support for view numbers in line number
8005 tables, this will cause the compiler to emit the line number table,
8006 which generally makes them somewhat less compact. The augmented line
8007 number tables and location lists are fully backward-compatible, so they
8008 can be consumed by debug information consumers that are not aware of
8009 these augmentations, but they won't derive any benefit from them either.
8011 This is enabled by default when outputting DWARF 2 debug information at
8012 the normal level, as long as there is assembler support,
8013 @option{-fvar-tracking-assignments} is enabled and
8014 @option{-gstrict-dwarf} is not. When assembler support is not
8015 available, this may still be enabled, but it will force GCC to output
8016 internal line number tables, and if
8017 @option{-ginternal-reset-location-views} is not enabled, that will most
8018 certainly lead to silently mismatching location views.
8020 There is a proposed representation for view numbers that is not backward
8021 compatible with the location list format introduced in DWARF 5, that can
8022 be enabled with @option{-gvariable-location-views=incompat5}. This
8023 option may be removed in the future, is only provided as a reference
8024 implementation of the proposed representation. Debug information
8025 consumers are not expected to support this extended format, and they
8026 would be rendered unable to decode location lists using it.
8028 @item -ginternal-reset-location-views
8029 @itemx -gnointernal-reset-location-views
8030 @opindex ginternal-reset-location-views
8031 @opindex gno-internal-reset-location-views
8032 Attempt to determine location views that can be omitted from location
8033 view lists. This requires the compiler to have very accurate insn
8034 length estimates, which isn't always the case, and it may cause
8035 incorrect view lists to be generated silently when using an assembler
8036 that does not support location view lists. The GNU assembler will flag
8037 any such error as a @code{view number mismatch}. This is only enabled
8038 on ports that define a reliable estimation function.
8040 @item -ginline-points
8041 @itemx -gno-inline-points
8042 @opindex ginline-points
8043 @opindex gno-inline-points
8044 Generate extended debug information for inlined functions. Location
8045 view tracking markers are inserted at inlined entry points, so that
8046 address and view numbers can be computed and output in debug
8047 information. This can be enabled independently of location views, in
8048 which case the view numbers won't be output, but it can only be enabled
8049 along with statement frontiers, and it is only enabled by default if
8050 location views are enabled.
8052 @item -gz@r{[}=@var{type}@r{]}
8054 Produce compressed debug sections in DWARF format, if that is supported.
8055 If @var{type} is not given, the default type depends on the capabilities
8056 of the assembler and linker used. @var{type} may be one of
8057 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
8058 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
8059 compression in traditional GNU format). If the linker doesn't support
8060 writing compressed debug sections, the option is rejected. Otherwise,
8061 if the assembler does not support them, @option{-gz} is silently ignored
8062 when producing object files.
8064 @item -femit-struct-debug-baseonly
8065 @opindex femit-struct-debug-baseonly
8066 Emit debug information for struct-like types
8067 only when the base name of the compilation source file
8068 matches the base name of file in which the struct is defined.
8070 This option substantially reduces the size of debugging information,
8071 but at significant potential loss in type information to the debugger.
8072 See @option{-femit-struct-debug-reduced} for a less aggressive option.
8073 See @option{-femit-struct-debug-detailed} for more detailed control.
8075 This option works only with DWARF debug output.
8077 @item -femit-struct-debug-reduced
8078 @opindex femit-struct-debug-reduced
8079 Emit debug information for struct-like types
8080 only when the base name of the compilation source file
8081 matches the base name of file in which the type is defined,
8082 unless the struct is a template or defined in a system header.
8084 This option significantly reduces the size of debugging information,
8085 with some potential loss in type information to the debugger.
8086 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
8087 See @option{-femit-struct-debug-detailed} for more detailed control.
8089 This option works only with DWARF debug output.
8091 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
8092 @opindex femit-struct-debug-detailed
8093 Specify the struct-like types
8094 for which the compiler generates debug information.
8095 The intent is to reduce duplicate struct debug information
8096 between different object files within the same program.
8098 This option is a detailed version of
8099 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
8100 which serves for most needs.
8102 A specification has the syntax@*
8103 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
8105 The optional first word limits the specification to
8106 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
8107 A struct type is used directly when it is the type of a variable, member.
8108 Indirect uses arise through pointers to structs.
8109 That is, when use of an incomplete struct is valid, the use is indirect.
8111 @samp{struct one direct; struct two * indirect;}.
8113 The optional second word limits the specification to
8114 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
8115 Generic structs are a bit complicated to explain.
8116 For C++, these are non-explicit specializations of template classes,
8117 or non-template classes within the above.
8118 Other programming languages have generics,
8119 but @option{-femit-struct-debug-detailed} does not yet implement them.
8121 The third word specifies the source files for those
8122 structs for which the compiler should emit debug information.
8123 The values @samp{none} and @samp{any} have the normal meaning.
8124 The value @samp{base} means that
8125 the base of name of the file in which the type declaration appears
8126 must match the base of the name of the main compilation file.
8127 In practice, this means that when compiling @file{foo.c}, debug information
8128 is generated for types declared in that file and @file{foo.h},
8129 but not other header files.
8130 The value @samp{sys} means those types satisfying @samp{base}
8131 or declared in system or compiler headers.
8133 You may need to experiment to determine the best settings for your application.
8135 The default is @option{-femit-struct-debug-detailed=all}.
8137 This option works only with DWARF debug output.
8139 @item -fno-dwarf2-cfi-asm
8140 @opindex fdwarf2-cfi-asm
8141 @opindex fno-dwarf2-cfi-asm
8142 Emit DWARF unwind info as compiler generated @code{.eh_frame} section
8143 instead of using GAS @code{.cfi_*} directives.
8145 @item -fno-eliminate-unused-debug-types
8146 @opindex feliminate-unused-debug-types
8147 @opindex fno-eliminate-unused-debug-types
8148 Normally, when producing DWARF output, GCC avoids producing debug symbol
8149 output for types that are nowhere used in the source file being compiled.
8150 Sometimes it is useful to have GCC emit debugging
8151 information for all types declared in a compilation
8152 unit, regardless of whether or not they are actually used
8153 in that compilation unit, for example
8154 if, in the debugger, you want to cast a value to a type that is
8155 not actually used in your program (but is declared). More often,
8156 however, this results in a significant amount of wasted space.
8159 @node Optimize Options
8160 @section Options That Control Optimization
8161 @cindex optimize options
8162 @cindex options, optimization
8164 These options control various sorts of optimizations.
8166 Without any optimization option, the compiler's goal is to reduce the
8167 cost of compilation and to make debugging produce the expected
8168 results. Statements are independent: if you stop the program with a
8169 breakpoint between statements, you can then assign a new value to any
8170 variable or change the program counter to any other statement in the
8171 function and get exactly the results you expect from the source
8174 Turning on optimization flags makes the compiler attempt to improve
8175 the performance and/or code size at the expense of compilation time
8176 and possibly the ability to debug the program.
8178 The compiler performs optimization based on the knowledge it has of the
8179 program. Compiling multiple files at once to a single output file mode allows
8180 the compiler to use information gained from all of the files when compiling
8183 Not all optimizations are controlled directly by a flag. Only
8184 optimizations that have a flag are listed in this section.
8186 Most optimizations are completely disabled at @option{-O0} or if an
8187 @option{-O} level is not set on the command line, even if individual
8188 optimization flags are specified. Similarly, @option{-Og} suppresses
8189 many optimization passes.
8191 Depending on the target and how GCC was configured, a slightly different
8192 set of optimizations may be enabled at each @option{-O} level than
8193 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
8194 to find out the exact set of optimizations that are enabled at each level.
8195 @xref{Overall Options}, for examples.
8202 Optimize. Optimizing compilation takes somewhat more time, and a lot
8203 more memory for a large function.
8205 With @option{-O}, the compiler tries to reduce code size and execution
8206 time, without performing any optimizations that take a great deal of
8209 @c Note that in addition to the default_options_table list in opts.c,
8210 @c several optimization flags default to true but control optimization
8211 @c passes that are explicitly disabled at -O0.
8213 @option{-O} turns on the following optimization flags:
8215 @c Please keep the following list alphabetized.
8216 @gccoptlist{-fauto-inc-dec @gol
8217 -fbranch-count-reg @gol
8218 -fcombine-stack-adjustments @gol
8220 -fcprop-registers @gol
8223 -fdelayed-branch @gol
8225 -fforward-propagate @gol
8226 -fguess-branch-probability @gol
8227 -fif-conversion @gol
8228 -fif-conversion2 @gol
8229 -finline-functions-called-once @gol
8231 -fipa-pure-const @gol
8232 -fipa-reference @gol
8233 -fipa-reference-addressable @gol
8234 -fmerge-constants @gol
8235 -fmove-loop-invariants @gol
8236 -fomit-frame-pointer @gol
8237 -freorder-blocks @gol
8239 -fshrink-wrap-separate @gol
8240 -fsplit-wide-types @gol
8246 -ftree-coalesce-vars @gol
8247 -ftree-copy-prop @gol
8249 -ftree-dominator-opts @gol
8251 -ftree-forwprop @gol
8255 -ftree-scev-cprop @gol
8264 Optimize even more. GCC performs nearly all supported optimizations
8265 that do not involve a space-speed tradeoff.
8266 As compared to @option{-O}, this option increases both compilation time
8267 and the performance of the generated code.
8269 @option{-O2} turns on all optimization flags specified by @option{-O}. It
8270 also turns on the following optimization flags:
8272 @c Please keep the following list alphabetized!
8273 @gccoptlist{-falign-functions -falign-jumps @gol
8274 -falign-labels -falign-loops @gol
8276 -fcode-hoisting @gol
8278 -fcse-follow-jumps -fcse-skip-blocks @gol
8279 -fdelete-null-pointer-checks @gol
8280 -fdevirtualize -fdevirtualize-speculatively @gol
8281 -fexpensive-optimizations @gol
8282 -fgcse -fgcse-lm @gol
8283 -fhoist-adjacent-loads @gol
8284 -finline-small-functions @gol
8285 -findirect-inlining @gol
8286 -fipa-bit-cp -fipa-cp -fipa-icf @gol
8287 -fipa-ra -fipa-sra -fipa-vrp @gol
8288 -fisolate-erroneous-paths-dereference @gol
8290 -foptimize-sibling-calls @gol
8291 -foptimize-strlen @gol
8292 -fpartial-inlining @gol
8294 -freorder-blocks-algorithm=stc @gol
8295 -freorder-blocks-and-partition -freorder-functions @gol
8296 -frerun-cse-after-loop @gol
8297 -fschedule-insns -fschedule-insns2 @gol
8298 -fsched-interblock -fsched-spec @gol
8299 -fstore-merging @gol
8300 -fstrict-aliasing @gol
8302 -ftree-builtin-call-dce @gol
8304 -ftree-switch-conversion -ftree-tail-merge @gol
8307 Please note the warning under @option{-fgcse} about
8308 invoking @option{-O2} on programs that use computed gotos.
8312 Optimize yet more. @option{-O3} turns on all optimizations specified
8313 by @option{-O2} and also turns on the following optimization flags:
8315 @c Please keep the following list alphabetized!
8316 @gccoptlist{-fgcse-after-reload @gol
8317 -finline-functions @gol
8319 -floop-interchange @gol
8320 -floop-unroll-and-jam @gol
8322 -fpredictive-commoning @gol
8324 -ftree-loop-distribute-patterns @gol
8325 -ftree-loop-distribution @gol
8326 -ftree-loop-vectorize @gol
8327 -ftree-partial-pre @gol
8328 -ftree-slp-vectorize @gol
8329 -funswitch-loops @gol
8330 -fvect-cost-model @gol
8331 -fversion-loops-for-strides}
8335 Reduce compilation time and make debugging produce the expected
8336 results. This is the default.
8340 Optimize for size. @option{-Os} enables all @option{-O2} optimizations
8341 except those that often increase code size:
8343 @gccoptlist{-falign-functions -falign-jumps @gol
8344 -falign-labels -falign-loops @gol
8345 -fprefetch-loop-arrays -freorder-blocks-algorithm=stc}
8347 It also enables @option{-finline-functions}, causes the compiler to tune for
8348 code size rather than execution speed, and performs further optimizations
8349 designed to reduce code size.
8353 Disregard strict standards compliance. @option{-Ofast} enables all
8354 @option{-O3} optimizations. It also enables optimizations that are not
8355 valid for all standard-compliant programs.
8356 It turns on @option{-ffast-math} and the Fortran-specific
8357 @option{-fstack-arrays}, unless @option{-fmax-stack-var-size} is
8358 specified, and @option{-fno-protect-parens}.
8362 Optimize debugging experience. @option{-Og} should be the optimization
8363 level of choice for the standard edit-compile-debug cycle, offering
8364 a reasonable level of optimization while maintaining fast compilation
8365 and a good debugging experience. It is a better choice than @option{-O0}
8366 for producing debuggable code because some compiler passes
8367 that collect debug information are disabled at @option{-O0}.
8369 Like @option{-O0}, @option{-Og} completely disables a number of
8370 optimization passes so that individual options controlling them have
8371 no effect. Otherwise @option{-Og} enables all @option{-O1}
8372 optimization flags except for those that may interfere with debugging:
8374 @gccoptlist{-fbranch-count-reg -fdelayed-branch @gol
8375 -fif-conversion -fif-conversion2 @gol
8376 -finline-functions-called-once @gol
8377 -fmove-loop-invariants -fssa-phiopt @gol
8378 -ftree-bit-ccp -ftree-pta -ftree-sra}
8382 If you use multiple @option{-O} options, with or without level numbers,
8383 the last such option is the one that is effective.
8385 Options of the form @option{-f@var{flag}} specify machine-independent
8386 flags. Most flags have both positive and negative forms; the negative
8387 form of @option{-ffoo} is @option{-fno-foo}. In the table
8388 below, only one of the forms is listed---the one you typically
8389 use. You can figure out the other form by either removing @samp{no-}
8392 The following options control specific optimizations. They are either
8393 activated by @option{-O} options or are related to ones that are. You
8394 can use the following flags in the rare cases when ``fine-tuning'' of
8395 optimizations to be performed is desired.
8398 @item -fno-defer-pop
8399 @opindex fno-defer-pop
8401 For machines that must pop arguments after a function call, always pop
8402 the arguments as soon as each function returns.
8403 At levels @option{-O1} and higher, @option{-fdefer-pop} is the default;
8404 this allows the compiler to let arguments accumulate on the stack for several
8405 function calls and pop them all at once.
8407 @item -fforward-propagate
8408 @opindex fforward-propagate
8409 Perform a forward propagation pass on RTL@. The pass tries to combine two
8410 instructions and checks if the result can be simplified. If loop unrolling
8411 is active, two passes are performed and the second is scheduled after
8414 This option is enabled by default at optimization levels @option{-O},
8415 @option{-O2}, @option{-O3}, @option{-Os}.
8417 @item -ffp-contract=@var{style}
8418 @opindex ffp-contract
8419 @option{-ffp-contract=off} disables floating-point expression contraction.
8420 @option{-ffp-contract=fast} enables floating-point expression contraction
8421 such as forming of fused multiply-add operations if the target has
8422 native support for them.
8423 @option{-ffp-contract=on} enables floating-point expression contraction
8424 if allowed by the language standard. This is currently not implemented
8425 and treated equal to @option{-ffp-contract=off}.
8427 The default is @option{-ffp-contract=fast}.
8429 @item -fomit-frame-pointer
8430 @opindex fomit-frame-pointer
8431 Omit the frame pointer in functions that don't need one. This avoids the
8432 instructions to save, set up and restore the frame pointer; on many targets
8433 it also makes an extra register available.
8435 On some targets this flag has no effect because the standard calling sequence
8436 always uses a frame pointer, so it cannot be omitted.
8438 Note that @option{-fno-omit-frame-pointer} doesn't guarantee the frame pointer
8439 is used in all functions. Several targets always omit the frame pointer in
8442 Enabled by default at @option{-O} and higher.
8444 @item -foptimize-sibling-calls
8445 @opindex foptimize-sibling-calls
8446 Optimize sibling and tail recursive calls.
8448 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8450 @item -foptimize-strlen
8451 @opindex foptimize-strlen
8452 Optimize various standard C string functions (e.g.@: @code{strlen},
8453 @code{strchr} or @code{strcpy}) and
8454 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
8456 Enabled at levels @option{-O2}, @option{-O3}.
8461 Do not expand any functions inline apart from those marked with
8462 the @code{always_inline} attribute. This is the default when not
8465 Single functions can be exempted from inlining by marking them
8466 with the @code{noinline} attribute.
8468 @item -finline-small-functions
8469 @opindex finline-small-functions
8470 Integrate functions into their callers when their body is smaller than expected
8471 function call code (so overall size of program gets smaller). The compiler
8472 heuristically decides which functions are simple enough to be worth integrating
8473 in this way. This inlining applies to all functions, even those not declared
8476 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8478 @item -findirect-inlining
8479 @opindex findirect-inlining
8480 Inline also indirect calls that are discovered to be known at compile
8481 time thanks to previous inlining. This option has any effect only
8482 when inlining itself is turned on by the @option{-finline-functions}
8483 or @option{-finline-small-functions} options.
8485 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8487 @item -finline-functions
8488 @opindex finline-functions
8489 Consider all functions for inlining, even if they are not declared inline.
8490 The compiler heuristically decides which functions are worth integrating
8493 If all calls to a given function are integrated, and the function is
8494 declared @code{static}, then the function is normally not output as
8495 assembler code in its own right.
8497 Enabled at levels @option{-O3}, @option{-Os}. Also enabled
8498 by @option{-fprofile-use} and @option{-fauto-profile}.
8500 @item -finline-functions-called-once
8501 @opindex finline-functions-called-once
8502 Consider all @code{static} functions called once for inlining into their
8503 caller even if they are not marked @code{inline}. If a call to a given
8504 function is integrated, then the function is not output as assembler code
8507 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os},
8508 but not @option{-Og}.
8510 @item -fearly-inlining
8511 @opindex fearly-inlining
8512 Inline functions marked by @code{always_inline} and functions whose body seems
8513 smaller than the function call overhead early before doing
8514 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
8515 makes profiling significantly cheaper and usually inlining faster on programs
8516 having large chains of nested wrapper functions.
8522 Perform interprocedural scalar replacement of aggregates, removal of
8523 unused parameters and replacement of parameters passed by reference
8524 by parameters passed by value.
8526 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
8528 @item -finline-limit=@var{n}
8529 @opindex finline-limit
8530 By default, GCC limits the size of functions that can be inlined. This flag
8531 allows coarse control of this limit. @var{n} is the size of functions that
8532 can be inlined in number of pseudo instructions.
8534 Inlining is actually controlled by a number of parameters, which may be
8535 specified individually by using @option{--param @var{name}=@var{value}}.
8536 The @option{-finline-limit=@var{n}} option sets some of these parameters
8540 @item max-inline-insns-single
8541 is set to @var{n}/2.
8542 @item max-inline-insns-auto
8543 is set to @var{n}/2.
8546 See below for a documentation of the individual
8547 parameters controlling inlining and for the defaults of these parameters.
8549 @emph{Note:} there may be no value to @option{-finline-limit} that results
8550 in default behavior.
8552 @emph{Note:} pseudo instruction represents, in this particular context, an
8553 abstract measurement of function's size. In no way does it represent a count
8554 of assembly instructions and as such its exact meaning might change from one
8555 release to an another.
8557 @item -fno-keep-inline-dllexport
8558 @opindex fno-keep-inline-dllexport
8559 @opindex fkeep-inline-dllexport
8560 This is a more fine-grained version of @option{-fkeep-inline-functions},
8561 which applies only to functions that are declared using the @code{dllexport}
8562 attribute or declspec. @xref{Function Attributes,,Declaring Attributes of
8565 @item -fkeep-inline-functions
8566 @opindex fkeep-inline-functions
8567 In C, emit @code{static} functions that are declared @code{inline}
8568 into the object file, even if the function has been inlined into all
8569 of its callers. This switch does not affect functions using the
8570 @code{extern inline} extension in GNU C90@. In C++, emit any and all
8571 inline functions into the object file.
8573 @item -fkeep-static-functions
8574 @opindex fkeep-static-functions
8575 Emit @code{static} functions into the object file, even if the function
8578 @item -fkeep-static-consts
8579 @opindex fkeep-static-consts
8580 Emit variables declared @code{static const} when optimization isn't turned
8581 on, even if the variables aren't referenced.
8583 GCC enables this option by default. If you want to force the compiler to
8584 check if a variable is referenced, regardless of whether or not
8585 optimization is turned on, use the @option{-fno-keep-static-consts} option.
8587 @item -fmerge-constants
8588 @opindex fmerge-constants
8589 Attempt to merge identical constants (string constants and floating-point
8590 constants) across compilation units.
8592 This option is the default for optimized compilation if the assembler and
8593 linker support it. Use @option{-fno-merge-constants} to inhibit this
8596 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8598 @item -fmerge-all-constants
8599 @opindex fmerge-all-constants
8600 Attempt to merge identical constants and identical variables.
8602 This option implies @option{-fmerge-constants}. In addition to
8603 @option{-fmerge-constants} this considers e.g.@: even constant initialized
8604 arrays or initialized constant variables with integral or floating-point
8605 types. Languages like C or C++ require each variable, including multiple
8606 instances of the same variable in recursive calls, to have distinct locations,
8607 so using this option results in non-conforming
8610 @item -fmodulo-sched
8611 @opindex fmodulo-sched
8612 Perform swing modulo scheduling immediately before the first scheduling
8613 pass. This pass looks at innermost loops and reorders their
8614 instructions by overlapping different iterations.
8616 @item -fmodulo-sched-allow-regmoves
8617 @opindex fmodulo-sched-allow-regmoves
8618 Perform more aggressive SMS-based modulo scheduling with register moves
8619 allowed. By setting this flag certain anti-dependences edges are
8620 deleted, which triggers the generation of reg-moves based on the
8621 life-range analysis. This option is effective only with
8622 @option{-fmodulo-sched} enabled.
8624 @item -fno-branch-count-reg
8625 @opindex fno-branch-count-reg
8626 @opindex fbranch-count-reg
8627 Disable the optimization pass that scans for opportunities to use
8628 ``decrement and branch'' instructions on a count register instead of
8629 instruction sequences that decrement a register, compare it against zero, and
8630 then branch based upon the result. This option is only meaningful on
8631 architectures that support such instructions, which include x86, PowerPC,
8632 IA-64 and S/390. Note that the @option{-fno-branch-count-reg} option
8633 doesn't remove the decrement and branch instructions from the generated
8634 instruction stream introduced by other optimization passes.
8636 The default is @option{-fbranch-count-reg} at @option{-O1} and higher,
8637 except for @option{-Og}.
8639 @item -fno-function-cse
8640 @opindex fno-function-cse
8641 @opindex ffunction-cse
8642 Do not put function addresses in registers; make each instruction that
8643 calls a constant function contain the function's address explicitly.
8645 This option results in less efficient code, but some strange hacks
8646 that alter the assembler output may be confused by the optimizations
8647 performed when this option is not used.
8649 The default is @option{-ffunction-cse}
8651 @item -fno-zero-initialized-in-bss
8652 @opindex fno-zero-initialized-in-bss
8653 @opindex fzero-initialized-in-bss
8654 If the target supports a BSS section, GCC by default puts variables that
8655 are initialized to zero into BSS@. This can save space in the resulting
8658 This option turns off this behavior because some programs explicitly
8659 rely on variables going to the data section---e.g., so that the
8660 resulting executable can find the beginning of that section and/or make
8661 assumptions based on that.
8663 The default is @option{-fzero-initialized-in-bss}.
8665 @item -fthread-jumps
8666 @opindex fthread-jumps
8667 Perform optimizations that check to see if a jump branches to a
8668 location where another comparison subsumed by the first is found. If
8669 so, the first branch is redirected to either the destination of the
8670 second branch or a point immediately following it, depending on whether
8671 the condition is known to be true or false.
8673 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8675 @item -fsplit-wide-types
8676 @opindex fsplit-wide-types
8677 When using a type that occupies multiple registers, such as @code{long
8678 long} on a 32-bit system, split the registers apart and allocate them
8679 independently. This normally generates better code for those types,
8680 but may make debugging more difficult.
8682 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
8685 @item -fcse-follow-jumps
8686 @opindex fcse-follow-jumps
8687 In common subexpression elimination (CSE), scan through jump instructions
8688 when the target of the jump is not reached by any other path. For
8689 example, when CSE encounters an @code{if} statement with an
8690 @code{else} clause, CSE follows the jump when the condition
8693 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8695 @item -fcse-skip-blocks
8696 @opindex fcse-skip-blocks
8697 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
8698 follow jumps that conditionally skip over blocks. When CSE
8699 encounters a simple @code{if} statement with no else clause,
8700 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
8701 body of the @code{if}.
8703 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8705 @item -frerun-cse-after-loop
8706 @opindex frerun-cse-after-loop
8707 Re-run common subexpression elimination after loop optimizations are
8710 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8714 Perform a global common subexpression elimination pass.
8715 This pass also performs global constant and copy propagation.
8717 @emph{Note:} When compiling a program using computed gotos, a GCC
8718 extension, you may get better run-time performance if you disable
8719 the global common subexpression elimination pass by adding
8720 @option{-fno-gcse} to the command line.
8722 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8726 When @option{-fgcse-lm} is enabled, global common subexpression elimination
8727 attempts to move loads that are only killed by stores into themselves. This
8728 allows a loop containing a load/store sequence to be changed to a load outside
8729 the loop, and a copy/store within the loop.
8731 Enabled by default when @option{-fgcse} is enabled.
8735 When @option{-fgcse-sm} is enabled, a store motion pass is run after
8736 global common subexpression elimination. This pass attempts to move
8737 stores out of loops. When used in conjunction with @option{-fgcse-lm},
8738 loops containing a load/store sequence can be changed to a load before
8739 the loop and a store after the loop.
8741 Not enabled at any optimization level.
8745 When @option{-fgcse-las} is enabled, the global common subexpression
8746 elimination pass eliminates redundant loads that come after stores to the
8747 same memory location (both partial and full redundancies).
8749 Not enabled at any optimization level.
8751 @item -fgcse-after-reload
8752 @opindex fgcse-after-reload
8753 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
8754 pass is performed after reload. The purpose of this pass is to clean up
8757 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
8759 @item -faggressive-loop-optimizations
8760 @opindex faggressive-loop-optimizations
8761 This option tells the loop optimizer to use language constraints to
8762 derive bounds for the number of iterations of a loop. This assumes that
8763 loop code does not invoke undefined behavior by for example causing signed
8764 integer overflows or out-of-bound array accesses. The bounds for the
8765 number of iterations of a loop are used to guide loop unrolling and peeling
8766 and loop exit test optimizations.
8767 This option is enabled by default.
8769 @item -funconstrained-commons
8770 @opindex funconstrained-commons
8771 This option tells the compiler that variables declared in common blocks
8772 (e.g.@: Fortran) may later be overridden with longer trailing arrays. This
8773 prevents certain optimizations that depend on knowing the array bounds.
8775 @item -fcrossjumping
8776 @opindex fcrossjumping
8777 Perform cross-jumping transformation.
8778 This transformation unifies equivalent code and saves code size. The
8779 resulting code may or may not perform better than without cross-jumping.
8781 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8783 @item -fauto-inc-dec
8784 @opindex fauto-inc-dec
8785 Combine increments or decrements of addresses with memory accesses.
8786 This pass is always skipped on architectures that do not have
8787 instructions to support this. Enabled by default at @option{-O} and
8788 higher on architectures that support this.
8792 Perform dead code elimination (DCE) on RTL@.
8793 Enabled by default at @option{-O} and higher.
8797 Perform dead store elimination (DSE) on RTL@.
8798 Enabled by default at @option{-O} and higher.
8800 @item -fif-conversion
8801 @opindex fif-conversion
8802 Attempt to transform conditional jumps into branch-less equivalents. This
8803 includes use of conditional moves, min, max, set flags and abs instructions, and
8804 some tricks doable by standard arithmetics. The use of conditional execution
8805 on chips where it is available is controlled by @option{-fif-conversion2}.
8807 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}, but
8808 not with @option{-Og}.
8810 @item -fif-conversion2
8811 @opindex fif-conversion2
8812 Use conditional execution (where available) to transform conditional jumps into
8813 branch-less equivalents.
8815 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}, but
8816 not with @option{-Og}.
8818 @item -fdeclone-ctor-dtor
8819 @opindex fdeclone-ctor-dtor
8820 The C++ ABI requires multiple entry points for constructors and
8821 destructors: one for a base subobject, one for a complete object, and
8822 one for a virtual destructor that calls operator delete afterwards.
8823 For a hierarchy with virtual bases, the base and complete variants are
8824 clones, which means two copies of the function. With this option, the
8825 base and complete variants are changed to be thunks that call a common
8828 Enabled by @option{-Os}.
8830 @item -fdelete-null-pointer-checks
8831 @opindex fdelete-null-pointer-checks
8832 Assume that programs cannot safely dereference null pointers, and that
8833 no code or data element resides at address zero.
8834 This option enables simple constant
8835 folding optimizations at all optimization levels. In addition, other
8836 optimization passes in GCC use this flag to control global dataflow
8837 analyses that eliminate useless checks for null pointers; these assume
8838 that a memory access to address zero always results in a trap, so
8839 that if a pointer is checked after it has already been dereferenced,
8842 Note however that in some environments this assumption is not true.
8843 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
8844 for programs that depend on that behavior.
8846 This option is enabled by default on most targets. On Nios II ELF, it
8847 defaults to off. On AVR, CR16, and MSP430, this option is completely disabled.
8849 Passes that use the dataflow information
8850 are enabled independently at different optimization levels.
8852 @item -fdevirtualize
8853 @opindex fdevirtualize
8854 Attempt to convert calls to virtual functions to direct calls. This
8855 is done both within a procedure and interprocedurally as part of
8856 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
8857 propagation (@option{-fipa-cp}).
8858 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8860 @item -fdevirtualize-speculatively
8861 @opindex fdevirtualize-speculatively
8862 Attempt to convert calls to virtual functions to speculative direct calls.
8863 Based on the analysis of the type inheritance graph, determine for a given call
8864 the set of likely targets. If the set is small, preferably of size 1, change
8865 the call into a conditional deciding between direct and indirect calls. The
8866 speculative calls enable more optimizations, such as inlining. When they seem
8867 useless after further optimization, they are converted back into original form.
8869 @item -fdevirtualize-at-ltrans
8870 @opindex fdevirtualize-at-ltrans
8871 Stream extra information needed for aggressive devirtualization when running
8872 the link-time optimizer in local transformation mode.
8873 This option enables more devirtualization but
8874 significantly increases the size of streamed data. For this reason it is
8875 disabled by default.
8877 @item -fexpensive-optimizations
8878 @opindex fexpensive-optimizations
8879 Perform a number of minor optimizations that are relatively expensive.
8881 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8885 Attempt to remove redundant extension instructions. This is especially
8886 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
8887 registers after writing to their lower 32-bit half.
8889 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
8890 @option{-O3}, @option{-Os}.
8892 @item -fno-lifetime-dse
8893 @opindex fno-lifetime-dse
8894 @opindex flifetime-dse
8895 In C++ the value of an object is only affected by changes within its
8896 lifetime: when the constructor begins, the object has an indeterminate
8897 value, and any changes during the lifetime of the object are dead when
8898 the object is destroyed. Normally dead store elimination will take
8899 advantage of this; if your code relies on the value of the object
8900 storage persisting beyond the lifetime of the object, you can use this
8901 flag to disable this optimization. To preserve stores before the
8902 constructor starts (e.g.@: because your operator new clears the object
8903 storage) but still treat the object as dead after the destructor you,
8904 can use @option{-flifetime-dse=1}. The default behavior can be
8905 explicitly selected with @option{-flifetime-dse=2}.
8906 @option{-flifetime-dse=0} is equivalent to @option{-fno-lifetime-dse}.
8908 @item -flive-range-shrinkage
8909 @opindex flive-range-shrinkage
8910 Attempt to decrease register pressure through register live range
8911 shrinkage. This is helpful for fast processors with small or moderate
8914 @item -fira-algorithm=@var{algorithm}
8915 @opindex fira-algorithm
8916 Use the specified coloring algorithm for the integrated register
8917 allocator. The @var{algorithm} argument can be @samp{priority}, which
8918 specifies Chow's priority coloring, or @samp{CB}, which specifies
8919 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
8920 for all architectures, but for those targets that do support it, it is
8921 the default because it generates better code.
8923 @item -fira-region=@var{region}
8924 @opindex fira-region
8925 Use specified regions for the integrated register allocator. The
8926 @var{region} argument should be one of the following:
8931 Use all loops as register allocation regions.
8932 This can give the best results for machines with a small and/or
8933 irregular register set.
8936 Use all loops except for loops with small register pressure
8937 as the regions. This value usually gives
8938 the best results in most cases and for most architectures,
8939 and is enabled by default when compiling with optimization for speed
8940 (@option{-O}, @option{-O2}, @dots{}).
8943 Use all functions as a single region.
8944 This typically results in the smallest code size, and is enabled by default for
8945 @option{-Os} or @option{-O0}.
8949 @item -fira-hoist-pressure
8950 @opindex fira-hoist-pressure
8951 Use IRA to evaluate register pressure in the code hoisting pass for
8952 decisions to hoist expressions. This option usually results in smaller
8953 code, but it can slow the compiler down.
8955 This option is enabled at level @option{-Os} for all targets.
8957 @item -fira-loop-pressure
8958 @opindex fira-loop-pressure
8959 Use IRA to evaluate register pressure in loops for decisions to move
8960 loop invariants. This option usually results in generation
8961 of faster and smaller code on machines with large register files (>= 32
8962 registers), but it can slow the compiler down.
8964 This option is enabled at level @option{-O3} for some targets.
8966 @item -fno-ira-share-save-slots
8967 @opindex fno-ira-share-save-slots
8968 @opindex fira-share-save-slots
8969 Disable sharing of stack slots used for saving call-used hard
8970 registers living through a call. Each hard register gets a
8971 separate stack slot, and as a result function stack frames are
8974 @item -fno-ira-share-spill-slots
8975 @opindex fno-ira-share-spill-slots
8976 @opindex fira-share-spill-slots
8977 Disable sharing of stack slots allocated for pseudo-registers. Each
8978 pseudo-register that does not get a hard register gets a separate
8979 stack slot, and as a result function stack frames are larger.
8983 Enable CFG-sensitive rematerialization in LRA. Instead of loading
8984 values of spilled pseudos, LRA tries to rematerialize (recalculate)
8985 values if it is profitable.
8987 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8989 @item -fdelayed-branch
8990 @opindex fdelayed-branch
8991 If supported for the target machine, attempt to reorder instructions
8992 to exploit instruction slots available after delayed branch
8995 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os},
8996 but not at @option{-Og}.
8998 @item -fschedule-insns
8999 @opindex fschedule-insns
9000 If supported for the target machine, attempt to reorder instructions to
9001 eliminate execution stalls due to required data being unavailable. This
9002 helps machines that have slow floating point or memory load instructions
9003 by allowing other instructions to be issued until the result of the load
9004 or floating-point instruction is required.
9006 Enabled at levels @option{-O2}, @option{-O3}.
9008 @item -fschedule-insns2
9009 @opindex fschedule-insns2
9010 Similar to @option{-fschedule-insns}, but requests an additional pass of
9011 instruction scheduling after register allocation has been done. This is
9012 especially useful on machines with a relatively small number of
9013 registers and where memory load instructions take more than one cycle.
9015 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9017 @item -fno-sched-interblock
9018 @opindex fno-sched-interblock
9019 @opindex fsched-interblock
9020 Disable instruction scheduling across basic blocks, which
9021 is normally enabled when scheduling before register allocation, i.e.@:
9022 with @option{-fschedule-insns} or at @option{-O2} or higher.
9024 @item -fno-sched-spec
9025 @opindex fno-sched-spec
9026 @opindex fsched-spec
9027 Disable speculative motion of non-load instructions, which
9028 is normally enabled when scheduling before register allocation, i.e.@:
9029 with @option{-fschedule-insns} or at @option{-O2} or higher.
9031 @item -fsched-pressure
9032 @opindex fsched-pressure
9033 Enable register pressure sensitive insn scheduling before register
9034 allocation. This only makes sense when scheduling before register
9035 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
9036 @option{-O2} or higher. Usage of this option can improve the
9037 generated code and decrease its size by preventing register pressure
9038 increase above the number of available hard registers and subsequent
9039 spills in register allocation.
9041 @item -fsched-spec-load
9042 @opindex fsched-spec-load
9043 Allow speculative motion of some load instructions. This only makes
9044 sense when scheduling before register allocation, i.e.@: with
9045 @option{-fschedule-insns} or at @option{-O2} or higher.
9047 @item -fsched-spec-load-dangerous
9048 @opindex fsched-spec-load-dangerous
9049 Allow speculative motion of more load instructions. This only makes
9050 sense when scheduling before register allocation, i.e.@: with
9051 @option{-fschedule-insns} or at @option{-O2} or higher.
9053 @item -fsched-stalled-insns
9054 @itemx -fsched-stalled-insns=@var{n}
9055 @opindex fsched-stalled-insns
9056 Define how many insns (if any) can be moved prematurely from the queue
9057 of stalled insns into the ready list during the second scheduling pass.
9058 @option{-fno-sched-stalled-insns} means that no insns are moved
9059 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
9060 on how many queued insns can be moved prematurely.
9061 @option{-fsched-stalled-insns} without a value is equivalent to
9062 @option{-fsched-stalled-insns=1}.
9064 @item -fsched-stalled-insns-dep
9065 @itemx -fsched-stalled-insns-dep=@var{n}
9066 @opindex fsched-stalled-insns-dep
9067 Define how many insn groups (cycles) are examined for a dependency
9068 on a stalled insn that is a candidate for premature removal from the queue
9069 of stalled insns. This has an effect only during the second scheduling pass,
9070 and only if @option{-fsched-stalled-insns} is used.
9071 @option{-fno-sched-stalled-insns-dep} is equivalent to
9072 @option{-fsched-stalled-insns-dep=0}.
9073 @option{-fsched-stalled-insns-dep} without a value is equivalent to
9074 @option{-fsched-stalled-insns-dep=1}.
9076 @item -fsched2-use-superblocks
9077 @opindex fsched2-use-superblocks
9078 When scheduling after register allocation, use superblock scheduling.
9079 This allows motion across basic block boundaries,
9080 resulting in faster schedules. This option is experimental, as not all machine
9081 descriptions used by GCC model the CPU closely enough to avoid unreliable
9082 results from the algorithm.
9084 This only makes sense when scheduling after register allocation, i.e.@: with
9085 @option{-fschedule-insns2} or at @option{-O2} or higher.
9087 @item -fsched-group-heuristic
9088 @opindex fsched-group-heuristic
9089 Enable the group heuristic in the scheduler. This heuristic favors
9090 the instruction that belongs to a schedule group. This is enabled
9091 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
9092 or @option{-fschedule-insns2} or at @option{-O2} or higher.
9094 @item -fsched-critical-path-heuristic
9095 @opindex fsched-critical-path-heuristic
9096 Enable the critical-path heuristic in the scheduler. This heuristic favors
9097 instructions on the critical path. This is enabled by default when
9098 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
9099 or @option{-fschedule-insns2} or at @option{-O2} or higher.
9101 @item -fsched-spec-insn-heuristic
9102 @opindex fsched-spec-insn-heuristic
9103 Enable the speculative instruction heuristic in the scheduler. This
9104 heuristic favors speculative instructions with greater dependency weakness.
9105 This is enabled by default when scheduling is enabled, i.e.@:
9106 with @option{-fschedule-insns} or @option{-fschedule-insns2}
9107 or at @option{-O2} or higher.
9109 @item -fsched-rank-heuristic
9110 @opindex fsched-rank-heuristic
9111 Enable the rank heuristic in the scheduler. This heuristic favors
9112 the instruction belonging to a basic block with greater size or frequency.
9113 This is enabled by default when scheduling is enabled, i.e.@:
9114 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
9115 at @option{-O2} or higher.
9117 @item -fsched-last-insn-heuristic
9118 @opindex fsched-last-insn-heuristic
9119 Enable the last-instruction heuristic in the scheduler. This heuristic
9120 favors the instruction that is less dependent on the last instruction
9121 scheduled. This is enabled by default when scheduling is enabled,
9122 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
9123 at @option{-O2} or higher.
9125 @item -fsched-dep-count-heuristic
9126 @opindex fsched-dep-count-heuristic
9127 Enable the dependent-count heuristic in the scheduler. This heuristic
9128 favors the instruction that has more instructions depending on it.
9129 This is enabled by default when scheduling is enabled, i.e.@:
9130 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
9131 at @option{-O2} or higher.
9133 @item -freschedule-modulo-scheduled-loops
9134 @opindex freschedule-modulo-scheduled-loops
9135 Modulo scheduling is performed before traditional scheduling. If a loop
9136 is modulo scheduled, later scheduling passes may change its schedule.
9137 Use this option to control that behavior.
9139 @item -fselective-scheduling
9140 @opindex fselective-scheduling
9141 Schedule instructions using selective scheduling algorithm. Selective
9142 scheduling runs instead of the first scheduler pass.
9144 @item -fselective-scheduling2
9145 @opindex fselective-scheduling2
9146 Schedule instructions using selective scheduling algorithm. Selective
9147 scheduling runs instead of the second scheduler pass.
9149 @item -fsel-sched-pipelining
9150 @opindex fsel-sched-pipelining
9151 Enable software pipelining of innermost loops during selective scheduling.
9152 This option has no effect unless one of @option{-fselective-scheduling} or
9153 @option{-fselective-scheduling2} is turned on.
9155 @item -fsel-sched-pipelining-outer-loops
9156 @opindex fsel-sched-pipelining-outer-loops
9157 When pipelining loops during selective scheduling, also pipeline outer loops.
9158 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
9160 @item -fsemantic-interposition
9161 @opindex fsemantic-interposition
9162 Some object formats, like ELF, allow interposing of symbols by the
9164 This means that for symbols exported from the DSO, the compiler cannot perform
9165 interprocedural propagation, inlining and other optimizations in anticipation
9166 that the function or variable in question may change. While this feature is
9167 useful, for example, to rewrite memory allocation functions by a debugging
9168 implementation, it is expensive in the terms of code quality.
9169 With @option{-fno-semantic-interposition} the compiler assumes that
9170 if interposition happens for functions the overwriting function will have
9171 precisely the same semantics (and side effects).
9172 Similarly if interposition happens
9173 for variables, the constructor of the variable will be the same. The flag
9174 has no effect for functions explicitly declared inline
9175 (where it is never allowed for interposition to change semantics)
9176 and for symbols explicitly declared weak.
9179 @opindex fshrink-wrap
9180 Emit function prologues only before parts of the function that need it,
9181 rather than at the top of the function. This flag is enabled by default at
9182 @option{-O} and higher.
9184 @item -fshrink-wrap-separate
9185 @opindex fshrink-wrap-separate
9186 Shrink-wrap separate parts of the prologue and epilogue separately, so that
9187 those parts are only executed when needed.
9188 This option is on by default, but has no effect unless @option{-fshrink-wrap}
9189 is also turned on and the target supports this.
9191 @item -fcaller-saves
9192 @opindex fcaller-saves
9193 Enable allocation of values to registers that are clobbered by
9194 function calls, by emitting extra instructions to save and restore the
9195 registers around such calls. Such allocation is done only when it
9196 seems to result in better code.
9198 This option is always enabled by default on certain machines, usually
9199 those which have no call-preserved registers to use instead.
9201 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9203 @item -fcombine-stack-adjustments
9204 @opindex fcombine-stack-adjustments
9205 Tracks stack adjustments (pushes and pops) and stack memory references
9206 and then tries to find ways to combine them.
9208 Enabled by default at @option{-O1} and higher.
9212 Use caller save registers for allocation if those registers are not used by
9213 any called function. In that case it is not necessary to save and restore
9214 them around calls. This is only possible if called functions are part of
9215 same compilation unit as current function and they are compiled before it.
9217 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}, however the option
9218 is disabled if generated code will be instrumented for profiling
9219 (@option{-p}, or @option{-pg}) or if callee's register usage cannot be known
9220 exactly (this happens on targets that do not expose prologues
9221 and epilogues in RTL).
9223 @item -fconserve-stack
9224 @opindex fconserve-stack
9225 Attempt to minimize stack usage. The compiler attempts to use less
9226 stack space, even if that makes the program slower. This option
9227 implies setting the @option{large-stack-frame} parameter to 100
9228 and the @option{large-stack-frame-growth} parameter to 400.
9230 @item -ftree-reassoc
9231 @opindex ftree-reassoc
9232 Perform reassociation on trees. This flag is enabled by default
9233 at @option{-O} and higher.
9235 @item -fcode-hoisting
9236 @opindex fcode-hoisting
9237 Perform code hoisting. Code hoisting tries to move the
9238 evaluation of expressions executed on all paths to the function exit
9239 as early as possible. This is especially useful as a code size
9240 optimization, but it often helps for code speed as well.
9241 This flag is enabled by default at @option{-O2} and higher.
9245 Perform partial redundancy elimination (PRE) on trees. This flag is
9246 enabled by default at @option{-O2} and @option{-O3}.
9248 @item -ftree-partial-pre
9249 @opindex ftree-partial-pre
9250 Make partial redundancy elimination (PRE) more aggressive. This flag is
9251 enabled by default at @option{-O3}.
9253 @item -ftree-forwprop
9254 @opindex ftree-forwprop
9255 Perform forward propagation on trees. This flag is enabled by default
9256 at @option{-O} and higher.
9260 Perform full redundancy elimination (FRE) on trees. The difference
9261 between FRE and PRE is that FRE only considers expressions
9262 that are computed on all paths leading to the redundant computation.
9263 This analysis is faster than PRE, though it exposes fewer redundancies.
9264 This flag is enabled by default at @option{-O} and higher.
9266 @item -ftree-phiprop
9267 @opindex ftree-phiprop
9268 Perform hoisting of loads from conditional pointers on trees. This
9269 pass is enabled by default at @option{-O} and higher.
9271 @item -fhoist-adjacent-loads
9272 @opindex fhoist-adjacent-loads
9273 Speculatively hoist loads from both branches of an if-then-else if the
9274 loads are from adjacent locations in the same structure and the target
9275 architecture has a conditional move instruction. This flag is enabled
9276 by default at @option{-O2} and higher.
9278 @item -ftree-copy-prop
9279 @opindex ftree-copy-prop
9280 Perform copy propagation on trees. This pass eliminates unnecessary
9281 copy operations. This flag is enabled by default at @option{-O} and
9284 @item -fipa-pure-const
9285 @opindex fipa-pure-const
9286 Discover which functions are pure or constant.
9287 Enabled by default at @option{-O} and higher.
9289 @item -fipa-reference
9290 @opindex fipa-reference
9291 Discover which static variables do not escape the
9293 Enabled by default at @option{-O} and higher.
9295 @item -fipa-reference-addressable
9296 @opindex fipa-reference-addressable
9297 Discover read-only, write-only and non-addressable static variables.
9298 Enabled by default at @option{-O} and higher.
9300 @item -fipa-stack-alignment
9301 @opindex fipa-stack-alignment
9302 Reduce stack alignment on call sites if possible.
9307 Perform interprocedural pointer analysis and interprocedural modification
9308 and reference analysis. This option can cause excessive memory and
9309 compile-time usage on large compilation units. It is not enabled by
9310 default at any optimization level.
9313 @opindex fipa-profile
9314 Perform interprocedural profile propagation. The functions called only from
9315 cold functions are marked as cold. Also functions executed once (such as
9316 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
9317 functions and loop less parts of functions executed once are then optimized for
9319 Enabled by default at @option{-O} and higher.
9323 Perform interprocedural constant propagation.
9324 This optimization analyzes the program to determine when values passed
9325 to functions are constants and then optimizes accordingly.
9326 This optimization can substantially increase performance
9327 if the application has constants passed to functions.
9328 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
9329 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9331 @item -fipa-cp-clone
9332 @opindex fipa-cp-clone
9333 Perform function cloning to make interprocedural constant propagation stronger.
9334 When enabled, interprocedural constant propagation performs function cloning
9335 when externally visible function can be called with constant arguments.
9336 Because this optimization can create multiple copies of functions,
9337 it may significantly increase code size
9338 (see @option{--param ipcp-unit-growth=@var{value}}).
9339 This flag is enabled by default at @option{-O3}.
9340 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9343 @opindex fipa-bit-cp
9344 When enabled, perform interprocedural bitwise constant
9345 propagation. This flag is enabled by default at @option{-O2} and
9346 by @option{-fprofile-use} and @option{-fauto-profile}.
9347 It requires that @option{-fipa-cp} is enabled.
9351 When enabled, perform interprocedural propagation of value
9352 ranges. This flag is enabled by default at @option{-O2}. It requires
9353 that @option{-fipa-cp} is enabled.
9357 Perform Identical Code Folding for functions and read-only variables.
9358 The optimization reduces code size and may disturb unwind stacks by replacing
9359 a function by equivalent one with a different name. The optimization works
9360 more effectively with link-time optimization enabled.
9362 Although the behavior is similar to the Gold Linker's ICF optimization, GCC ICF
9363 works on different levels and thus the optimizations are not same - there are
9364 equivalences that are found only by GCC and equivalences found only by Gold.
9366 This flag is enabled by default at @option{-O2} and @option{-Os}.
9368 @item -flive-patching=@var{level}
9369 @opindex flive-patching
9370 Control GCC's optimizations to provide a safe compilation for live-patching.
9372 If the compiler's optimization uses a function's body or information extracted
9373 from its body to optimize/change another function, the latter is called an
9374 impacted function of the former. If a function is patched, its impacted
9375 functions should be patched too.
9377 The impacted functions are decided by the compiler's interprocedural
9378 optimizations. For example, inlining a function into its caller, cloning
9379 a function and changing its caller to call this new clone, or extracting
9380 a function's pureness/constness information to optimize its direct or
9381 indirect callers, etc.
9383 Usually, the more IPA optimizations enabled, the larger the number of
9384 impacted functions for each function. In order to control the number of
9385 impacted functions and computed the list of impacted function easily,
9386 we provide control to partially enable IPA optimizations on two different
9389 The @var{level} argument should be one of the following:
9395 Only enable inlining and cloning optimizations, which includes inlining,
9396 cloning, interprocedural scalar replacement of aggregates and partial inlining.
9397 As a result, when patching a function, all its callers and its clones'
9398 callers need to be patched as well.
9400 @option{-flive-patching=inline-clone} disables the following optimization flags:
9401 @gccoptlist{-fwhole-program -fipa-pta -fipa-reference -fipa-ra @gol
9402 -fipa-icf -fipa-icf-functions -fipa-icf-variables @gol
9403 -fipa-bit-cp -fipa-vrp -fipa-pure-const -fipa-reference-addressable @gol
9404 -fipa-stack-alignment}
9406 @item inline-only-static
9408 Only enable inlining of static functions.
9409 As a result, when patching a static function, all its callers need to be
9412 In addition to all the flags that -flive-patching=inline-clone disables,
9413 @option{-flive-patching=inline-only-static} disables the following additional
9415 @gccoptlist{-fipa-cp-clone -fipa-sra -fpartial-inlining -fipa-cp}
9419 When -flive-patching specified without any value, the default value
9422 This flag is disabled by default.
9424 Note that -flive-patching is not supported with link-time optimizer.
9427 @item -fisolate-erroneous-paths-dereference
9428 @opindex fisolate-erroneous-paths-dereference
9429 Detect paths that trigger erroneous or undefined behavior due to
9430 dereferencing a null pointer. Isolate those paths from the main control
9431 flow and turn the statement with erroneous or undefined behavior into a trap.
9432 This flag is enabled by default at @option{-O2} and higher and depends on
9433 @option{-fdelete-null-pointer-checks} also being enabled.
9435 @item -fisolate-erroneous-paths-attribute
9436 @opindex fisolate-erroneous-paths-attribute
9437 Detect paths that trigger erroneous or undefined behavior due to a null value
9438 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
9439 attribute. Isolate those paths from the main control flow and turn the
9440 statement with erroneous or undefined behavior into a trap. This is not
9441 currently enabled, but may be enabled by @option{-O2} in the future.
9445 Perform forward store motion on trees. This flag is
9446 enabled by default at @option{-O} and higher.
9448 @item -ftree-bit-ccp
9449 @opindex ftree-bit-ccp
9450 Perform sparse conditional bit constant propagation on trees and propagate
9451 pointer alignment information.
9452 This pass only operates on local scalar variables and is enabled by default
9453 at @option{-O1} and higher, except for @option{-Og}.
9454 It requires that @option{-ftree-ccp} is enabled.
9458 Perform sparse conditional constant propagation (CCP) on trees. This
9459 pass only operates on local scalar variables and is enabled by default
9460 at @option{-O} and higher.
9462 @item -fssa-backprop
9463 @opindex fssa-backprop
9464 Propagate information about uses of a value up the definition chain
9465 in order to simplify the definitions. For example, this pass strips
9466 sign operations if the sign of a value never matters. The flag is
9467 enabled by default at @option{-O} and higher.
9470 @opindex fssa-phiopt
9471 Perform pattern matching on SSA PHI nodes to optimize conditional
9472 code. This pass is enabled by default at @option{-O1} and higher,
9473 except for @option{-Og}.
9475 @item -ftree-switch-conversion
9476 @opindex ftree-switch-conversion
9477 Perform conversion of simple initializations in a switch to
9478 initializations from a scalar array. This flag is enabled by default
9479 at @option{-O2} and higher.
9481 @item -ftree-tail-merge
9482 @opindex ftree-tail-merge
9483 Look for identical code sequences. When found, replace one with a jump to the
9484 other. This optimization is known as tail merging or cross jumping. This flag
9485 is enabled by default at @option{-O2} and higher. The compilation time
9487 be limited using @option{max-tail-merge-comparisons} parameter and
9488 @option{max-tail-merge-iterations} parameter.
9492 Perform dead code elimination (DCE) on trees. This flag is enabled by
9493 default at @option{-O} and higher.
9495 @item -ftree-builtin-call-dce
9496 @opindex ftree-builtin-call-dce
9497 Perform conditional dead code elimination (DCE) for calls to built-in functions
9498 that may set @code{errno} but are otherwise free of side effects. This flag is
9499 enabled by default at @option{-O2} and higher if @option{-Os} is not also
9502 @item -ftree-dominator-opts
9503 @opindex ftree-dominator-opts
9504 Perform a variety of simple scalar cleanups (constant/copy
9505 propagation, redundancy elimination, range propagation and expression
9506 simplification) based on a dominator tree traversal. This also
9507 performs jump threading (to reduce jumps to jumps). This flag is
9508 enabled by default at @option{-O} and higher.
9512 Perform dead store elimination (DSE) on trees. A dead store is a store into
9513 a memory location that is later overwritten by another store without
9514 any intervening loads. In this case the earlier store can be deleted. This
9515 flag is enabled by default at @option{-O} and higher.
9519 Perform loop header copying on trees. This is beneficial since it increases
9520 effectiveness of code motion optimizations. It also saves one jump. This flag
9521 is enabled by default at @option{-O} and higher. It is not enabled
9522 for @option{-Os}, since it usually increases code size.
9524 @item -ftree-loop-optimize
9525 @opindex ftree-loop-optimize
9526 Perform loop optimizations on trees. This flag is enabled by default
9527 at @option{-O} and higher.
9529 @item -ftree-loop-linear
9530 @itemx -floop-strip-mine
9532 @opindex ftree-loop-linear
9533 @opindex floop-strip-mine
9534 @opindex floop-block
9535 Perform loop nest optimizations. Same as
9536 @option{-floop-nest-optimize}. To use this code transformation, GCC has
9537 to be configured with @option{--with-isl} to enable the Graphite loop
9538 transformation infrastructure.
9540 @item -fgraphite-identity
9541 @opindex fgraphite-identity
9542 Enable the identity transformation for graphite. For every SCoP we generate
9543 the polyhedral representation and transform it back to gimple. Using
9544 @option{-fgraphite-identity} we can check the costs or benefits of the
9545 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
9546 are also performed by the code generator isl, like index splitting and
9547 dead code elimination in loops.
9549 @item -floop-nest-optimize
9550 @opindex floop-nest-optimize
9551 Enable the isl based loop nest optimizer. This is a generic loop nest
9552 optimizer based on the Pluto optimization algorithms. It calculates a loop
9553 structure optimized for data-locality and parallelism. This option
9556 @item -floop-parallelize-all
9557 @opindex floop-parallelize-all
9558 Use the Graphite data dependence analysis to identify loops that can
9559 be parallelized. Parallelize all the loops that can be analyzed to
9560 not contain loop carried dependences without checking that it is
9561 profitable to parallelize the loops.
9563 @item -ftree-coalesce-vars
9564 @opindex ftree-coalesce-vars
9565 While transforming the program out of the SSA representation, attempt to
9566 reduce copying by coalescing versions of different user-defined
9567 variables, instead of just compiler temporaries. This may severely
9568 limit the ability to debug an optimized program compiled with
9569 @option{-fno-var-tracking-assignments}. In the negated form, this flag
9570 prevents SSA coalescing of user variables. This option is enabled by
9571 default if optimization is enabled, and it does very little otherwise.
9573 @item -ftree-loop-if-convert
9574 @opindex ftree-loop-if-convert
9575 Attempt to transform conditional jumps in the innermost loops to
9576 branch-less equivalents. The intent is to remove control-flow from
9577 the innermost loops in order to improve the ability of the
9578 vectorization pass to handle these loops. This is enabled by default
9579 if vectorization is enabled.
9581 @item -ftree-loop-distribution
9582 @opindex ftree-loop-distribution
9583 Perform loop distribution. This flag can improve cache performance on
9584 big loop bodies and allow further loop optimizations, like
9585 parallelization or vectorization, to take place. For example, the loop
9601 This flag is enabled by default at @option{-O3}.
9602 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9604 @item -ftree-loop-distribute-patterns
9605 @opindex ftree-loop-distribute-patterns
9606 Perform loop distribution of patterns that can be code generated with
9607 calls to a library. This flag is enabled by default at @option{-O3}, and
9608 by @option{-fprofile-use} and @option{-fauto-profile}.
9610 This pass distributes the initialization loops and generates a call to
9611 memset zero. For example, the loop
9627 and the initialization loop is transformed into a call to memset zero.
9628 This flag is enabled by default at @option{-O3}.
9629 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9631 @item -floop-interchange
9632 @opindex floop-interchange
9633 Perform loop interchange outside of graphite. This flag can improve cache
9634 performance on loop nest and allow further loop optimizations, like
9635 vectorization, to take place. For example, the loop
9637 for (int i = 0; i < N; i++)
9638 for (int j = 0; j < N; j++)
9639 for (int k = 0; k < N; k++)
9640 c[i][j] = c[i][j] + a[i][k]*b[k][j];
9644 for (int i = 0; i < N; i++)
9645 for (int k = 0; k < N; k++)
9646 for (int j = 0; j < N; j++)
9647 c[i][j] = c[i][j] + a[i][k]*b[k][j];
9649 This flag is enabled by default at @option{-O3}.
9650 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9652 @item -floop-unroll-and-jam
9653 @opindex floop-unroll-and-jam
9654 Apply unroll and jam transformations on feasible loops. In a loop
9655 nest this unrolls the outer loop by some factor and fuses the resulting
9656 multiple inner loops. This flag is enabled by default at @option{-O3}.
9657 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9659 @item -ftree-loop-im
9660 @opindex ftree-loop-im
9661 Perform loop invariant motion on trees. This pass moves only invariants that
9662 are hard to handle at RTL level (function calls, operations that expand to
9663 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
9664 operands of conditions that are invariant out of the loop, so that we can use
9665 just trivial invariantness analysis in loop unswitching. The pass also includes
9668 @item -ftree-loop-ivcanon
9669 @opindex ftree-loop-ivcanon
9670 Create a canonical counter for number of iterations in loops for which
9671 determining number of iterations requires complicated analysis. Later
9672 optimizations then may determine the number easily. Useful especially
9673 in connection with unrolling.
9675 @item -ftree-scev-cprop
9676 @opindex ftree-scev-cprop
9677 Perform final value replacement. If a variable is modified in a loop
9678 in such a way that its value when exiting the loop can be determined using
9679 only its initial value and the number of loop iterations, replace uses of
9680 the final value by such a computation, provided it is sufficiently cheap.
9681 This reduces data dependencies and may allow further simplifications.
9682 Enabled by default at @option{-O} and higher.
9686 Perform induction variable optimizations (strength reduction, induction
9687 variable merging and induction variable elimination) on trees.
9689 @item -ftree-parallelize-loops=n
9690 @opindex ftree-parallelize-loops
9691 Parallelize loops, i.e., split their iteration space to run in n threads.
9692 This is only possible for loops whose iterations are independent
9693 and can be arbitrarily reordered. The optimization is only
9694 profitable on multiprocessor machines, for loops that are CPU-intensive,
9695 rather than constrained e.g.@: by memory bandwidth. This option
9696 implies @option{-pthread}, and thus is only supported on targets
9697 that have support for @option{-pthread}.
9701 Perform function-local points-to analysis on trees. This flag is
9702 enabled by default at @option{-O1} and higher, except for @option{-Og}.
9706 Perform scalar replacement of aggregates. This pass replaces structure
9707 references with scalars to prevent committing structures to memory too
9708 early. This flag is enabled by default at @option{-O1} and higher,
9709 except for @option{-Og}.
9711 @item -fstore-merging
9712 @opindex fstore-merging
9713 Perform merging of narrow stores to consecutive memory addresses. This pass
9714 merges contiguous stores of immediate values narrower than a word into fewer
9715 wider stores to reduce the number of instructions. This is enabled by default
9716 at @option{-O2} and higher as well as @option{-Os}.
9720 Perform temporary expression replacement during the SSA->normal phase. Single
9721 use/single def temporaries are replaced at their use location with their
9722 defining expression. This results in non-GIMPLE code, but gives the expanders
9723 much more complex trees to work on resulting in better RTL generation. This is
9724 enabled by default at @option{-O} and higher.
9728 Perform straight-line strength reduction on trees. This recognizes related
9729 expressions involving multiplications and replaces them by less expensive
9730 calculations when possible. This is enabled by default at @option{-O} and
9733 @item -ftree-vectorize
9734 @opindex ftree-vectorize
9735 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
9736 and @option{-ftree-slp-vectorize} if not explicitly specified.
9738 @item -ftree-loop-vectorize
9739 @opindex ftree-loop-vectorize
9740 Perform loop vectorization on trees. This flag is enabled by default at
9741 @option{-O3} and by @option{-ftree-vectorize}, @option{-fprofile-use},
9742 and @option{-fauto-profile}.
9744 @item -ftree-slp-vectorize
9745 @opindex ftree-slp-vectorize
9746 Perform basic block vectorization on trees. This flag is enabled by default at
9747 @option{-O3} and by @option{-ftree-vectorize}, @option{-fprofile-use},
9748 and @option{-fauto-profile}.
9750 @item -fvect-cost-model=@var{model}
9751 @opindex fvect-cost-model
9752 Alter the cost model used for vectorization. The @var{model} argument
9753 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
9754 With the @samp{unlimited} model the vectorized code-path is assumed
9755 to be profitable while with the @samp{dynamic} model a runtime check
9756 guards the vectorized code-path to enable it only for iteration
9757 counts that will likely execute faster than when executing the original
9758 scalar loop. The @samp{cheap} model disables vectorization of
9759 loops where doing so would be cost prohibitive for example due to
9760 required runtime checks for data dependence or alignment but otherwise
9761 is equal to the @samp{dynamic} model.
9762 The default cost model depends on other optimization flags and is
9763 either @samp{dynamic} or @samp{cheap}.
9765 @item -fsimd-cost-model=@var{model}
9766 @opindex fsimd-cost-model
9767 Alter the cost model used for vectorization of loops marked with the OpenMP
9768 simd directive. The @var{model} argument should be one of
9769 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
9770 have the same meaning as described in @option{-fvect-cost-model} and by
9771 default a cost model defined with @option{-fvect-cost-model} is used.
9775 Perform Value Range Propagation on trees. This is similar to the
9776 constant propagation pass, but instead of values, ranges of values are
9777 propagated. This allows the optimizers to remove unnecessary range
9778 checks like array bound checks and null pointer checks. This is
9779 enabled by default at @option{-O2} and higher. Null pointer check
9780 elimination is only done if @option{-fdelete-null-pointer-checks} is
9784 @opindex fsplit-paths
9785 Split paths leading to loop backedges. This can improve dead code
9786 elimination and common subexpression elimination. This is enabled by
9787 default at @option{-O3} and above.
9789 @item -fsplit-ivs-in-unroller
9790 @opindex fsplit-ivs-in-unroller
9791 Enables expression of values of induction variables in later iterations
9792 of the unrolled loop using the value in the first iteration. This breaks
9793 long dependency chains, thus improving efficiency of the scheduling passes.
9795 A combination of @option{-fweb} and CSE is often sufficient to obtain the
9796 same effect. However, that is not reliable in cases where the loop body
9797 is more complicated than a single basic block. It also does not work at all
9798 on some architectures due to restrictions in the CSE pass.
9800 This optimization is enabled by default.
9802 @item -fvariable-expansion-in-unroller
9803 @opindex fvariable-expansion-in-unroller
9804 With this option, the compiler creates multiple copies of some
9805 local variables when unrolling a loop, which can result in superior code.
9807 @item -fpartial-inlining
9808 @opindex fpartial-inlining
9809 Inline parts of functions. This option has any effect only
9810 when inlining itself is turned on by the @option{-finline-functions}
9811 or @option{-finline-small-functions} options.
9813 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9815 @item -fpredictive-commoning
9816 @opindex fpredictive-commoning
9817 Perform predictive commoning optimization, i.e., reusing computations
9818 (especially memory loads and stores) performed in previous
9819 iterations of loops.
9821 This option is enabled at level @option{-O3}.
9822 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9824 @item -fprefetch-loop-arrays
9825 @opindex fprefetch-loop-arrays
9826 If supported by the target machine, generate instructions to prefetch
9827 memory to improve the performance of loops that access large arrays.
9829 This option may generate better or worse code; results are highly
9830 dependent on the structure of loops within the source code.
9832 Disabled at level @option{-Os}.
9834 @item -fno-printf-return-value
9835 @opindex fno-printf-return-value
9836 @opindex fprintf-return-value
9837 Do not substitute constants for known return value of formatted output
9838 functions such as @code{sprintf}, @code{snprintf}, @code{vsprintf}, and
9839 @code{vsnprintf} (but not @code{printf} of @code{fprintf}). This
9840 transformation allows GCC to optimize or even eliminate branches based
9841 on the known return value of these functions called with arguments that
9842 are either constant, or whose values are known to be in a range that
9843 makes determining the exact return value possible. For example, when
9844 @option{-fprintf-return-value} is in effect, both the branch and the
9845 body of the @code{if} statement (but not the call to @code{snprint})
9846 can be optimized away when @code{i} is a 32-bit or smaller integer
9847 because the return value is guaranteed to be at most 8.
9851 if (snprintf (buf, "%08x", i) >= sizeof buf)
9855 The @option{-fprintf-return-value} option relies on other optimizations
9856 and yields best results with @option{-O2} and above. It works in tandem
9857 with the @option{-Wformat-overflow} and @option{-Wformat-truncation}
9858 options. The @option{-fprintf-return-value} option is enabled by default.
9861 @itemx -fno-peephole2
9862 @opindex fno-peephole
9864 @opindex fno-peephole2
9866 Disable any machine-specific peephole optimizations. The difference
9867 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
9868 are implemented in the compiler; some targets use one, some use the
9869 other, a few use both.
9871 @option{-fpeephole} is enabled by default.
9872 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9874 @item -fno-guess-branch-probability
9875 @opindex fno-guess-branch-probability
9876 @opindex fguess-branch-probability
9877 Do not guess branch probabilities using heuristics.
9879 GCC uses heuristics to guess branch probabilities if they are
9880 not provided by profiling feedback (@option{-fprofile-arcs}). These
9881 heuristics are based on the control flow graph. If some branch probabilities
9882 are specified by @code{__builtin_expect}, then the heuristics are
9883 used to guess branch probabilities for the rest of the control flow graph,
9884 taking the @code{__builtin_expect} info into account. The interactions
9885 between the heuristics and @code{__builtin_expect} can be complex, and in
9886 some cases, it may be useful to disable the heuristics so that the effects
9887 of @code{__builtin_expect} are easier to understand.
9889 It is also possible to specify expected probability of the expression
9890 with @code{__builtin_expect_with_probability} built-in function.
9892 The default is @option{-fguess-branch-probability} at levels
9893 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9895 @item -freorder-blocks
9896 @opindex freorder-blocks
9897 Reorder basic blocks in the compiled function in order to reduce number of
9898 taken branches and improve code locality.
9900 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9902 @item -freorder-blocks-algorithm=@var{algorithm}
9903 @opindex freorder-blocks-algorithm
9904 Use the specified algorithm for basic block reordering. The
9905 @var{algorithm} argument can be @samp{simple}, which does not increase
9906 code size (except sometimes due to secondary effects like alignment),
9907 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
9908 put all often executed code together, minimizing the number of branches
9909 executed by making extra copies of code.
9911 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
9912 @samp{stc} at levels @option{-O2}, @option{-O3}.
9914 @item -freorder-blocks-and-partition
9915 @opindex freorder-blocks-and-partition
9916 In addition to reordering basic blocks in the compiled function, in order
9917 to reduce number of taken branches, partitions hot and cold basic blocks
9918 into separate sections of the assembly and @file{.o} files, to improve
9919 paging and cache locality performance.
9921 This optimization is automatically turned off in the presence of
9922 exception handling or unwind tables (on targets using setjump/longjump or target specific scheme), for linkonce sections, for functions with a user-defined
9923 section attribute and on any architecture that does not support named
9924 sections. When @option{-fsplit-stack} is used this option is not
9925 enabled by default (to avoid linker errors), but may be enabled
9926 explicitly (if using a working linker).
9928 Enabled for x86 at levels @option{-O2}, @option{-O3}, @option{-Os}.
9930 @item -freorder-functions
9931 @opindex freorder-functions
9932 Reorder functions in the object file in order to
9933 improve code locality. This is implemented by using special
9934 subsections @code{.text.hot} for most frequently executed functions and
9935 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
9936 the linker so object file format must support named sections and linker must
9937 place them in a reasonable way.
9939 This option isn't effective unless you either provide profile feedback
9940 (see @option{-fprofile-arcs} for details) or manually annotate functions with
9941 @code{hot} or @code{cold} attributes (@pxref{Common Function Attributes}).
9943 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9945 @item -fstrict-aliasing
9946 @opindex fstrict-aliasing
9947 Allow the compiler to assume the strictest aliasing rules applicable to
9948 the language being compiled. For C (and C++), this activates
9949 optimizations based on the type of expressions. In particular, an
9950 object of one type is assumed never to reside at the same address as an
9951 object of a different type, unless the types are almost the same. For
9952 example, an @code{unsigned int} can alias an @code{int}, but not a
9953 @code{void*} or a @code{double}. A character type may alias any other
9956 @anchor{Type-punning}Pay special attention to code like this:
9969 The practice of reading from a different union member than the one most
9970 recently written to (called ``type-punning'') is common. Even with
9971 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
9972 is accessed through the union type. So, the code above works as
9973 expected. @xref{Structures unions enumerations and bit-fields
9974 implementation}. However, this code might not:
9985 Similarly, access by taking the address, casting the resulting pointer
9986 and dereferencing the result has undefined behavior, even if the cast
9987 uses a union type, e.g.:
9991 return ((union a_union *) &d)->i;
9995 The @option{-fstrict-aliasing} option is enabled at levels
9996 @option{-O2}, @option{-O3}, @option{-Os}.
9998 @item -falign-functions
9999 @itemx -falign-functions=@var{n}
10000 @itemx -falign-functions=@var{n}:@var{m}
10001 @itemx -falign-functions=@var{n}:@var{m}:@var{n2}
10002 @itemx -falign-functions=@var{n}:@var{m}:@var{n2}:@var{m2}
10003 @opindex falign-functions
10004 Align the start of functions to the next power-of-two greater than
10005 @var{n}, skipping up to @var{m}-1 bytes. This ensures that at least
10006 the first @var{m} bytes of the function can be fetched by the CPU
10007 without crossing an @var{n}-byte alignment boundary.
10009 If @var{m} is not specified, it defaults to @var{n}.
10011 Examples: @option{-falign-functions=32} aligns functions to the next
10012 32-byte boundary, @option{-falign-functions=24} aligns to the next
10013 32-byte boundary only if this can be done by skipping 23 bytes or less,
10014 @option{-falign-functions=32:7} aligns to the next
10015 32-byte boundary only if this can be done by skipping 6 bytes or less.
10017 The second pair of @var{n2}:@var{m2} values allows you to specify
10018 a secondary alignment: @option{-falign-functions=64:7:32:3} aligns to
10019 the next 64-byte boundary if this can be done by skipping 6 bytes or less,
10020 otherwise aligns to the next 32-byte boundary if this can be done
10021 by skipping 2 bytes or less.
10022 If @var{m2} is not specified, it defaults to @var{n2}.
10024 Some assemblers only support this flag when @var{n} is a power of two;
10025 in that case, it is rounded up.
10027 @option{-fno-align-functions} and @option{-falign-functions=1} are
10028 equivalent and mean that functions are not aligned.
10030 If @var{n} is not specified or is zero, use a machine-dependent default.
10031 The maximum allowed @var{n} option value is 65536.
10033 Enabled at levels @option{-O2}, @option{-O3}.
10035 @item -flimit-function-alignment
10036 If this option is enabled, the compiler tries to avoid unnecessarily
10037 overaligning functions. It attempts to instruct the assembler to align
10038 by the amount specified by @option{-falign-functions}, but not to
10039 skip more bytes than the size of the function.
10041 @item -falign-labels
10042 @itemx -falign-labels=@var{n}
10043 @itemx -falign-labels=@var{n}:@var{m}
10044 @itemx -falign-labels=@var{n}:@var{m}:@var{n2}
10045 @itemx -falign-labels=@var{n}:@var{m}:@var{n2}:@var{m2}
10046 @opindex falign-labels
10047 Align all branch targets to a power-of-two boundary.
10049 Parameters of this option are analogous to the @option{-falign-functions} option.
10050 @option{-fno-align-labels} and @option{-falign-labels=1} are
10051 equivalent and mean that labels are not aligned.
10053 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
10054 are greater than this value, then their values are used instead.
10056 If @var{n} is not specified or is zero, use a machine-dependent default
10057 which is very likely to be @samp{1}, meaning no alignment.
10058 The maximum allowed @var{n} option value is 65536.
10060 Enabled at levels @option{-O2}, @option{-O3}.
10062 @item -falign-loops
10063 @itemx -falign-loops=@var{n}
10064 @itemx -falign-loops=@var{n}:@var{m}
10065 @itemx -falign-loops=@var{n}:@var{m}:@var{n2}
10066 @itemx -falign-loops=@var{n}:@var{m}:@var{n2}:@var{m2}
10067 @opindex falign-loops
10068 Align loops to a power-of-two boundary. If the loops are executed
10069 many times, this makes up for any execution of the dummy padding
10072 Parameters of this option are analogous to the @option{-falign-functions} option.
10073 @option{-fno-align-loops} and @option{-falign-loops=1} are
10074 equivalent and mean that loops are not aligned.
10075 The maximum allowed @var{n} option value is 65536.
10077 If @var{n} is not specified or is zero, use a machine-dependent default.
10079 Enabled at levels @option{-O2}, @option{-O3}.
10081 @item -falign-jumps
10082 @itemx -falign-jumps=@var{n}
10083 @itemx -falign-jumps=@var{n}:@var{m}
10084 @itemx -falign-jumps=@var{n}:@var{m}:@var{n2}
10085 @itemx -falign-jumps=@var{n}:@var{m}:@var{n2}:@var{m2}
10086 @opindex falign-jumps
10087 Align branch targets to a power-of-two boundary, for branch targets
10088 where the targets can only be reached by jumping. In this case,
10089 no dummy operations need be executed.
10091 Parameters of this option are analogous to the @option{-falign-functions} option.
10092 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
10093 equivalent and mean that loops are not aligned.
10095 If @var{n} is not specified or is zero, use a machine-dependent default.
10096 The maximum allowed @var{n} option value is 65536.
10098 Enabled at levels @option{-O2}, @option{-O3}.
10100 @item -funit-at-a-time
10101 @opindex funit-at-a-time
10102 This option is left for compatibility reasons. @option{-funit-at-a-time}
10103 has no effect, while @option{-fno-unit-at-a-time} implies
10104 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
10106 Enabled by default.
10108 @item -fno-toplevel-reorder
10109 @opindex fno-toplevel-reorder
10110 @opindex ftoplevel-reorder
10111 Do not reorder top-level functions, variables, and @code{asm}
10112 statements. Output them in the same order that they appear in the
10113 input file. When this option is used, unreferenced static variables
10114 are not removed. This option is intended to support existing code
10115 that relies on a particular ordering. For new code, it is better to
10116 use attributes when possible.
10118 @option{-ftoplevel-reorder} is the default at @option{-O1} and higher, and
10119 also at @option{-O0} if @option{-fsection-anchors} is explicitly requested.
10120 Additionally @option{-fno-toplevel-reorder} implies
10121 @option{-fno-section-anchors}.
10125 Constructs webs as commonly used for register allocation purposes and assign
10126 each web individual pseudo register. This allows the register allocation pass
10127 to operate on pseudos directly, but also strengthens several other optimization
10128 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
10129 however, make debugging impossible, since variables no longer stay in a
10132 Enabled by default with @option{-funroll-loops}.
10134 @item -fwhole-program
10135 @opindex fwhole-program
10136 Assume that the current compilation unit represents the whole program being
10137 compiled. All public functions and variables with the exception of @code{main}
10138 and those merged by attribute @code{externally_visible} become static functions
10139 and in effect are optimized more aggressively by interprocedural optimizers.
10141 This option should not be used in combination with @option{-flto}.
10142 Instead relying on a linker plugin should provide safer and more precise
10145 @item -flto[=@var{n}]
10147 This option runs the standard link-time optimizer. When invoked
10148 with source code, it generates GIMPLE (one of GCC's internal
10149 representations) and writes it to special ELF sections in the object
10150 file. When the object files are linked together, all the function
10151 bodies are read from these ELF sections and instantiated as if they
10152 had been part of the same translation unit.
10154 To use the link-time optimizer, @option{-flto} and optimization
10155 options should be specified at compile time and during the final link.
10156 It is recommended that you compile all the files participating in the
10157 same link with the same options and also specify those options at
10162 gcc -c -O2 -flto foo.c
10163 gcc -c -O2 -flto bar.c
10164 gcc -o myprog -flto -O2 foo.o bar.o
10167 The first two invocations to GCC save a bytecode representation
10168 of GIMPLE into special ELF sections inside @file{foo.o} and
10169 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
10170 @file{foo.o} and @file{bar.o}, merges the two files into a single
10171 internal image, and compiles the result as usual. Since both
10172 @file{foo.o} and @file{bar.o} are merged into a single image, this
10173 causes all the interprocedural analyses and optimizations in GCC to
10174 work across the two files as if they were a single one. This means,
10175 for example, that the inliner is able to inline functions in
10176 @file{bar.o} into functions in @file{foo.o} and vice-versa.
10178 Another (simpler) way to enable link-time optimization is:
10181 gcc -o myprog -flto -O2 foo.c bar.c
10184 The above generates bytecode for @file{foo.c} and @file{bar.c},
10185 merges them together into a single GIMPLE representation and optimizes
10186 them as usual to produce @file{myprog}.
10188 The important thing to keep in mind is that to enable link-time
10189 optimizations you need to use the GCC driver to perform the link step.
10190 GCC automatically performs link-time optimization if any of the
10191 objects involved were compiled with the @option{-flto} command-line option.
10192 You can always override
10193 the automatic decision to do link-time optimization
10194 by passing @option{-fno-lto} to the link command.
10196 To make whole program optimization effective, it is necessary to make
10197 certain whole program assumptions. The compiler needs to know
10198 what functions and variables can be accessed by libraries and runtime
10199 outside of the link-time optimized unit. When supported by the linker,
10200 the linker plugin (see @option{-fuse-linker-plugin}) passes information
10201 to the compiler about used and externally visible symbols. When
10202 the linker plugin is not available, @option{-fwhole-program} should be
10203 used to allow the compiler to make these assumptions, which leads
10204 to more aggressive optimization decisions.
10206 When a file is compiled with @option{-flto} without
10207 @option{-fuse-linker-plugin}, the generated object file is larger than
10208 a regular object file because it contains GIMPLE bytecodes and the usual
10209 final code (see @option{-ffat-lto-objects}. This means that
10210 object files with LTO information can be linked as normal object
10211 files; if @option{-fno-lto} is passed to the linker, no
10212 interprocedural optimizations are applied. Note that when
10213 @option{-fno-fat-lto-objects} is enabled the compile stage is faster
10214 but you cannot perform a regular, non-LTO link on them.
10216 When producing the final binary, GCC only
10217 applies link-time optimizations to those files that contain bytecode.
10218 Therefore, you can mix and match object files and libraries with
10219 GIMPLE bytecodes and final object code. GCC automatically selects
10220 which files to optimize in LTO mode and which files to link without
10221 further processing.
10223 Generally, options specified at link time override those
10224 specified at compile time, although in some cases GCC attempts to infer
10225 link-time options from the settings used to compile the input files.
10227 If you do not specify an optimization level option @option{-O} at
10228 link time, then GCC uses the highest optimization level
10229 used when compiling the object files. Note that it is generally
10230 ineffective to specify an optimization level option only at link time and
10231 not at compile time, for two reasons. First, compiling without
10232 optimization suppresses compiler passes that gather information
10233 needed for effective optimization at link time. Second, some early
10234 optimization passes can be performed only at compile time and
10237 There are some code generation flags preserved by GCC when
10238 generating bytecodes, as they need to be used during the final link.
10239 Currently, the following options and their settings are taken from
10240 the first object file that explicitly specifies them:
10241 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
10242 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
10243 and all the @option{-m} target flags.
10245 Certain ABI-changing flags are required to match in all compilation units,
10246 and trying to override this at link time with a conflicting value
10247 is ignored. This includes options such as @option{-freg-struct-return}
10248 and @option{-fpcc-struct-return}.
10250 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
10251 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
10252 are passed through to the link stage and merged conservatively for
10253 conflicting translation units. Specifically
10254 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
10255 precedence; and for example @option{-ffp-contract=off} takes precedence
10256 over @option{-ffp-contract=fast}. You can override them at link time.
10258 If LTO encounters objects with C linkage declared with incompatible
10259 types in separate translation units to be linked together (undefined
10260 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
10261 issued. The behavior is still undefined at run time. Similar
10262 diagnostics may be raised for other languages.
10264 Another feature of LTO is that it is possible to apply interprocedural
10265 optimizations on files written in different languages:
10269 g++ -c -flto bar.cc
10270 gfortran -c -flto baz.f90
10271 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
10274 Notice that the final link is done with @command{g++} to get the C++
10275 runtime libraries and @option{-lgfortran} is added to get the Fortran
10276 runtime libraries. In general, when mixing languages in LTO mode, you
10277 should use the same link command options as when mixing languages in a
10278 regular (non-LTO) compilation.
10280 If object files containing GIMPLE bytecode are stored in a library archive, say
10281 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
10282 are using a linker with plugin support. To create static libraries suitable
10283 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
10284 and @command{ranlib};
10285 to show the symbols of object files with GIMPLE bytecode, use
10286 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
10287 and @command{nm} have been compiled with plugin support. At link time, use the
10288 flag @option{-fuse-linker-plugin} to ensure that the library participates in
10289 the LTO optimization process:
10292 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
10295 With the linker plugin enabled, the linker extracts the needed
10296 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
10297 to make them part of the aggregated GIMPLE image to be optimized.
10299 If you are not using a linker with plugin support and/or do not
10300 enable the linker plugin, then the objects inside @file{libfoo.a}
10301 are extracted and linked as usual, but they do not participate
10302 in the LTO optimization process. In order to make a static library suitable
10303 for both LTO optimization and usual linkage, compile its object files with
10304 @option{-flto} @option{-ffat-lto-objects}.
10306 Link-time optimizations do not require the presence of the whole program to
10307 operate. If the program does not require any symbols to be exported, it is
10308 possible to combine @option{-flto} and @option{-fwhole-program} to allow
10309 the interprocedural optimizers to use more aggressive assumptions which may
10310 lead to improved optimization opportunities.
10311 Use of @option{-fwhole-program} is not needed when linker plugin is
10312 active (see @option{-fuse-linker-plugin}).
10314 The current implementation of LTO makes no
10315 attempt to generate bytecode that is portable between different
10316 types of hosts. The bytecode files are versioned and there is a
10317 strict version check, so bytecode files generated in one version of
10318 GCC do not work with an older or newer version of GCC.
10320 Link-time optimization does not work well with generation of debugging
10321 information on systems other than those using a combination of ELF and
10324 If you specify the optional @var{n}, the optimization and code
10325 generation done at link time is executed in parallel using @var{n}
10326 parallel jobs by utilizing an installed @command{make} program. The
10327 environment variable @env{MAKE} may be used to override the program
10328 used. The default value for @var{n} is 1.
10330 You can also specify @option{-flto=jobserver} to use GNU make's
10331 job server mode to determine the number of parallel jobs. This
10332 is useful when the Makefile calling GCC is already executing in parallel.
10333 You must prepend a @samp{+} to the command recipe in the parent Makefile
10334 for this to work. This option likely only works if @env{MAKE} is
10337 @item -flto-partition=@var{alg}
10338 @opindex flto-partition
10339 Specify the partitioning algorithm used by the link-time optimizer.
10340 The value is either @samp{1to1} to specify a partitioning mirroring
10341 the original source files or @samp{balanced} to specify partitioning
10342 into equally sized chunks (whenever possible) or @samp{max} to create
10343 new partition for every symbol where possible. Specifying @samp{none}
10344 as an algorithm disables partitioning and streaming completely.
10345 The default value is @samp{balanced}. While @samp{1to1} can be used
10346 as an workaround for various code ordering issues, the @samp{max}
10347 partitioning is intended for internal testing only.
10348 The value @samp{one} specifies that exactly one partition should be
10349 used while the value @samp{none} bypasses partitioning and executes
10350 the link-time optimization step directly from the WPA phase.
10352 @item -flto-odr-type-merging
10353 @opindex flto-odr-type-merging
10354 Enable streaming of mangled types names of C++ types and their unification
10355 at link time. This increases size of LTO object files, but enables
10356 diagnostics about One Definition Rule violations.
10358 @item -flto-compression-level=@var{n}
10359 @opindex flto-compression-level
10360 This option specifies the level of compression used for intermediate
10361 language written to LTO object files, and is only meaningful in
10362 conjunction with LTO mode (@option{-flto}). Valid
10363 values are 0 (no compression) to 9 (maximum compression). Values
10364 outside this range are clamped to either 0 or 9. If the option is not
10365 given, a default balanced compression setting is used.
10367 @item -fuse-linker-plugin
10368 @opindex fuse-linker-plugin
10369 Enables the use of a linker plugin during link-time optimization. This
10370 option relies on plugin support in the linker, which is available in gold
10371 or in GNU ld 2.21 or newer.
10373 This option enables the extraction of object files with GIMPLE bytecode out
10374 of library archives. This improves the quality of optimization by exposing
10375 more code to the link-time optimizer. This information specifies what
10376 symbols can be accessed externally (by non-LTO object or during dynamic
10377 linking). Resulting code quality improvements on binaries (and shared
10378 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
10379 See @option{-flto} for a description of the effect of this flag and how to
10382 This option is enabled by default when LTO support in GCC is enabled
10383 and GCC was configured for use with
10384 a linker supporting plugins (GNU ld 2.21 or newer or gold).
10386 @item -ffat-lto-objects
10387 @opindex ffat-lto-objects
10388 Fat LTO objects are object files that contain both the intermediate language
10389 and the object code. This makes them usable for both LTO linking and normal
10390 linking. This option is effective only when compiling with @option{-flto}
10391 and is ignored at link time.
10393 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
10394 requires the complete toolchain to be aware of LTO. It requires a linker with
10395 linker plugin support for basic functionality. Additionally,
10396 @command{nm}, @command{ar} and @command{ranlib}
10397 need to support linker plugins to allow a full-featured build environment
10398 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
10399 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
10400 to these tools. With non fat LTO makefiles need to be modified to use them.
10402 Note that modern binutils provide plugin auto-load mechanism.
10403 Installing the linker plugin into @file{$libdir/bfd-plugins} has the same
10404 effect as usage of the command wrappers (@command{gcc-ar}, @command{gcc-nm} and
10405 @command{gcc-ranlib}).
10407 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
10410 @item -fcompare-elim
10411 @opindex fcompare-elim
10412 After register allocation and post-register allocation instruction splitting,
10413 identify arithmetic instructions that compute processor flags similar to a
10414 comparison operation based on that arithmetic. If possible, eliminate the
10415 explicit comparison operation.
10417 This pass only applies to certain targets that cannot explicitly represent
10418 the comparison operation before register allocation is complete.
10420 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
10422 @item -fcprop-registers
10423 @opindex fcprop-registers
10424 After register allocation and post-register allocation instruction splitting,
10425 perform a copy-propagation pass to try to reduce scheduling dependencies
10426 and occasionally eliminate the copy.
10428 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
10430 @item -fprofile-correction
10431 @opindex fprofile-correction
10432 Profiles collected using an instrumented binary for multi-threaded programs may
10433 be inconsistent due to missed counter updates. When this option is specified,
10434 GCC uses heuristics to correct or smooth out such inconsistencies. By
10435 default, GCC emits an error message when an inconsistent profile is detected.
10437 This option is enabled by @option{-fauto-profile}.
10439 @item -fprofile-use
10440 @itemx -fprofile-use=@var{path}
10441 @opindex fprofile-use
10442 Enable profile feedback-directed optimizations,
10443 and the following optimizations, many of which
10444 are generally profitable only with profile feedback available:
10446 @gccoptlist{-fbranch-probabilities -fprofile-values @gol
10447 -funroll-loops -fpeel-loops -ftracer -fvpt @gol
10448 -finline-functions -fipa-cp -fipa-cp-clone -fipa-bit-cp @gol
10449 -fpredictive-commoning -fsplit-loops -funswitch-loops @gol
10450 -fgcse-after-reload -ftree-loop-vectorize -ftree-slp-vectorize @gol
10451 -fvect-cost-model=dynamic -ftree-loop-distribute-patterns @gol
10452 -fprofile-reorder-functions}
10454 Before you can use this option, you must first generate profiling information.
10455 @xref{Instrumentation Options}, for information about the
10456 @option{-fprofile-generate} option.
10458 By default, GCC emits an error message if the feedback profiles do not
10459 match the source code. This error can be turned into a warning by using
10460 @option{-Wno-error=coverage-mismatch}. Note this may result in poorly
10461 optimized code. Additionally, by default, GCC also emits a warning message if
10462 the feedback profiles do not exist (see @option{-Wmissing-profile}).
10464 If @var{path} is specified, GCC looks at the @var{path} to find
10465 the profile feedback data files. See @option{-fprofile-dir}.
10467 @item -fauto-profile
10468 @itemx -fauto-profile=@var{path}
10469 @opindex fauto-profile
10470 Enable sampling-based feedback-directed optimizations,
10471 and the following optimizations,
10472 many of which are generally profitable only with profile feedback available:
10474 @gccoptlist{-fbranch-probabilities -fprofile-values @gol
10475 -funroll-loops -fpeel-loops -ftracer -fvpt @gol
10476 -finline-functions -fipa-cp -fipa-cp-clone -fipa-bit-cp @gol
10477 -fpredictive-commoning -fsplit-loops -funswitch-loops @gol
10478 -fgcse-after-reload -ftree-loop-vectorize -ftree-slp-vectorize @gol
10479 -fvect-cost-model=dynamic -ftree-loop-distribute-patterns @gol
10480 -fprofile-correction}
10482 @var{path} is the name of a file containing AutoFDO profile information.
10483 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
10485 Producing an AutoFDO profile data file requires running your program
10486 with the @command{perf} utility on a supported GNU/Linux target system.
10487 For more information, see @uref{https://perf.wiki.kernel.org/}.
10491 perf record -e br_inst_retired:near_taken -b -o perf.data \
10495 Then use the @command{create_gcov} tool to convert the raw profile data
10496 to a format that can be used by GCC.@ You must also supply the
10497 unstripped binary for your program to this tool.
10498 See @uref{https://github.com/google/autofdo}.
10502 create_gcov --binary=your_program.unstripped --profile=perf.data \
10503 --gcov=profile.afdo
10507 The following options control compiler behavior regarding floating-point
10508 arithmetic. These options trade off between speed and
10509 correctness. All must be specifically enabled.
10512 @item -ffloat-store
10513 @opindex ffloat-store
10514 Do not store floating-point variables in registers, and inhibit other
10515 options that might change whether a floating-point value is taken from a
10516 register or memory.
10518 @cindex floating-point precision
10519 This option prevents undesirable excess precision on machines such as
10520 the 68000 where the floating registers (of the 68881) keep more
10521 precision than a @code{double} is supposed to have. Similarly for the
10522 x86 architecture. For most programs, the excess precision does only
10523 good, but a few programs rely on the precise definition of IEEE floating
10524 point. Use @option{-ffloat-store} for such programs, after modifying
10525 them to store all pertinent intermediate computations into variables.
10527 @item -fexcess-precision=@var{style}
10528 @opindex fexcess-precision
10529 This option allows further control over excess precision on machines
10530 where floating-point operations occur in a format with more precision or
10531 range than the IEEE standard and interchange floating-point types. By
10532 default, @option{-fexcess-precision=fast} is in effect; this means that
10533 operations may be carried out in a wider precision than the types specified
10534 in the source if that would result in faster code, and it is unpredictable
10535 when rounding to the types specified in the source code takes place.
10536 When compiling C, if @option{-fexcess-precision=standard} is specified then
10537 excess precision follows the rules specified in ISO C99; in particular,
10538 both casts and assignments cause values to be rounded to their
10539 semantic types (whereas @option{-ffloat-store} only affects
10540 assignments). This option is enabled by default for C if a strict
10541 conformance option such as @option{-std=c99} is used.
10542 @option{-ffast-math} enables @option{-fexcess-precision=fast} by default
10543 regardless of whether a strict conformance option is used.
10546 @option{-fexcess-precision=standard} is not implemented for languages
10547 other than C. On the x86, it has no effect if @option{-mfpmath=sse}
10548 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
10549 semantics apply without excess precision, and in the latter, rounding
10553 @opindex ffast-math
10554 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
10555 @option{-ffinite-math-only}, @option{-fno-rounding-math},
10556 @option{-fno-signaling-nans}, @option{-fcx-limited-range} and
10557 @option{-fexcess-precision=fast}.
10559 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
10561 This option is not turned on by any @option{-O} option besides
10562 @option{-Ofast} since it can result in incorrect output for programs
10563 that depend on an exact implementation of IEEE or ISO rules/specifications
10564 for math functions. It may, however, yield faster code for programs
10565 that do not require the guarantees of these specifications.
10567 @item -fno-math-errno
10568 @opindex fno-math-errno
10569 @opindex fmath-errno
10570 Do not set @code{errno} after calling math functions that are executed
10571 with a single instruction, e.g., @code{sqrt}. A program that relies on
10572 IEEE exceptions for math error handling may want to use this flag
10573 for speed while maintaining IEEE arithmetic compatibility.
10575 This option is not turned on by any @option{-O} option since
10576 it can result in incorrect output for programs that depend on
10577 an exact implementation of IEEE or ISO rules/specifications for
10578 math functions. It may, however, yield faster code for programs
10579 that do not require the guarantees of these specifications.
10581 The default is @option{-fmath-errno}.
10583 On Darwin systems, the math library never sets @code{errno}. There is
10584 therefore no reason for the compiler to consider the possibility that
10585 it might, and @option{-fno-math-errno} is the default.
10587 @item -funsafe-math-optimizations
10588 @opindex funsafe-math-optimizations
10590 Allow optimizations for floating-point arithmetic that (a) assume
10591 that arguments and results are valid and (b) may violate IEEE or
10592 ANSI standards. When used at link time, it may include libraries
10593 or startup files that change the default FPU control word or other
10594 similar optimizations.
10596 This option is not turned on by any @option{-O} option since
10597 it can result in incorrect output for programs that depend on
10598 an exact implementation of IEEE or ISO rules/specifications for
10599 math functions. It may, however, yield faster code for programs
10600 that do not require the guarantees of these specifications.
10601 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
10602 @option{-fassociative-math} and @option{-freciprocal-math}.
10604 The default is @option{-fno-unsafe-math-optimizations}.
10606 @item -fassociative-math
10607 @opindex fassociative-math
10609 Allow re-association of operands in series of floating-point operations.
10610 This violates the ISO C and C++ language standard by possibly changing
10611 computation result. NOTE: re-ordering may change the sign of zero as
10612 well as ignore NaNs and inhibit or create underflow or overflow (and
10613 thus cannot be used on code that relies on rounding behavior like
10614 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
10615 and thus may not be used when ordered comparisons are required.
10616 This option requires that both @option{-fno-signed-zeros} and
10617 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
10618 much sense with @option{-frounding-math}. For Fortran the option
10619 is automatically enabled when both @option{-fno-signed-zeros} and
10620 @option{-fno-trapping-math} are in effect.
10622 The default is @option{-fno-associative-math}.
10624 @item -freciprocal-math
10625 @opindex freciprocal-math
10627 Allow the reciprocal of a value to be used instead of dividing by
10628 the value if this enables optimizations. For example @code{x / y}
10629 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
10630 is subject to common subexpression elimination. Note that this loses
10631 precision and increases the number of flops operating on the value.
10633 The default is @option{-fno-reciprocal-math}.
10635 @item -ffinite-math-only
10636 @opindex ffinite-math-only
10637 Allow optimizations for floating-point arithmetic that assume
10638 that arguments and results are not NaNs or +-Infs.
10640 This option is not turned on by any @option{-O} option since
10641 it can result in incorrect output for programs that depend on
10642 an exact implementation of IEEE or ISO rules/specifications for
10643 math functions. It may, however, yield faster code for programs
10644 that do not require the guarantees of these specifications.
10646 The default is @option{-fno-finite-math-only}.
10648 @item -fno-signed-zeros
10649 @opindex fno-signed-zeros
10650 @opindex fsigned-zeros
10651 Allow optimizations for floating-point arithmetic that ignore the
10652 signedness of zero. IEEE arithmetic specifies the behavior of
10653 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
10654 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
10655 This option implies that the sign of a zero result isn't significant.
10657 The default is @option{-fsigned-zeros}.
10659 @item -fno-trapping-math
10660 @opindex fno-trapping-math
10661 @opindex ftrapping-math
10662 Compile code assuming that floating-point operations cannot generate
10663 user-visible traps. These traps include division by zero, overflow,
10664 underflow, inexact result and invalid operation. This option requires
10665 that @option{-fno-signaling-nans} be in effect. Setting this option may
10666 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
10668 This option should never be turned on by any @option{-O} option since
10669 it can result in incorrect output for programs that depend on
10670 an exact implementation of IEEE or ISO rules/specifications for
10673 The default is @option{-ftrapping-math}.
10675 @item -frounding-math
10676 @opindex frounding-math
10677 Disable transformations and optimizations that assume default floating-point
10678 rounding behavior. This is round-to-zero for all floating point
10679 to integer conversions, and round-to-nearest for all other arithmetic
10680 truncations. This option should be specified for programs that change
10681 the FP rounding mode dynamically, or that may be executed with a
10682 non-default rounding mode. This option disables constant folding of
10683 floating-point expressions at compile time (which may be affected by
10684 rounding mode) and arithmetic transformations that are unsafe in the
10685 presence of sign-dependent rounding modes.
10687 The default is @option{-fno-rounding-math}.
10689 This option is experimental and does not currently guarantee to
10690 disable all GCC optimizations that are affected by rounding mode.
10691 Future versions of GCC may provide finer control of this setting
10692 using C99's @code{FENV_ACCESS} pragma. This command-line option
10693 will be used to specify the default state for @code{FENV_ACCESS}.
10695 @item -fsignaling-nans
10696 @opindex fsignaling-nans
10697 Compile code assuming that IEEE signaling NaNs may generate user-visible
10698 traps during floating-point operations. Setting this option disables
10699 optimizations that may change the number of exceptions visible with
10700 signaling NaNs. This option implies @option{-ftrapping-math}.
10702 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
10705 The default is @option{-fno-signaling-nans}.
10707 This option is experimental and does not currently guarantee to
10708 disable all GCC optimizations that affect signaling NaN behavior.
10710 @item -fno-fp-int-builtin-inexact
10711 @opindex fno-fp-int-builtin-inexact
10712 @opindex ffp-int-builtin-inexact
10713 Do not allow the built-in functions @code{ceil}, @code{floor},
10714 @code{round} and @code{trunc}, and their @code{float} and @code{long
10715 double} variants, to generate code that raises the ``inexact''
10716 floating-point exception for noninteger arguments. ISO C99 and C11
10717 allow these functions to raise the ``inexact'' exception, but ISO/IEC
10718 TS 18661-1:2014, the C bindings to IEEE 754-2008, does not allow these
10719 functions to do so.
10721 The default is @option{-ffp-int-builtin-inexact}, allowing the
10722 exception to be raised. This option does nothing unless
10723 @option{-ftrapping-math} is in effect.
10725 Even if @option{-fno-fp-int-builtin-inexact} is used, if the functions
10726 generate a call to a library function then the ``inexact'' exception
10727 may be raised if the library implementation does not follow TS 18661.
10729 @item -fsingle-precision-constant
10730 @opindex fsingle-precision-constant
10731 Treat floating-point constants as single precision instead of
10732 implicitly converting them to double-precision constants.
10734 @item -fcx-limited-range
10735 @opindex fcx-limited-range
10736 When enabled, this option states that a range reduction step is not
10737 needed when performing complex division. Also, there is no checking
10738 whether the result of a complex multiplication or division is @code{NaN
10739 + I*NaN}, with an attempt to rescue the situation in that case. The
10740 default is @option{-fno-cx-limited-range}, but is enabled by
10741 @option{-ffast-math}.
10743 This option controls the default setting of the ISO C99
10744 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
10747 @item -fcx-fortran-rules
10748 @opindex fcx-fortran-rules
10749 Complex multiplication and division follow Fortran rules. Range
10750 reduction is done as part of complex division, but there is no checking
10751 whether the result of a complex multiplication or division is @code{NaN
10752 + I*NaN}, with an attempt to rescue the situation in that case.
10754 The default is @option{-fno-cx-fortran-rules}.
10758 The following options control optimizations that may improve
10759 performance, but are not enabled by any @option{-O} options. This
10760 section includes experimental options that may produce broken code.
10763 @item -fbranch-probabilities
10764 @opindex fbranch-probabilities
10765 After running a program compiled with @option{-fprofile-arcs}
10766 (@pxref{Instrumentation Options}),
10767 you can compile it a second time using
10768 @option{-fbranch-probabilities}, to improve optimizations based on
10769 the number of times each branch was taken. When a program
10770 compiled with @option{-fprofile-arcs} exits, it saves arc execution
10771 counts to a file called @file{@var{sourcename}.gcda} for each source
10772 file. The information in this data file is very dependent on the
10773 structure of the generated code, so you must use the same source code
10774 and the same optimization options for both compilations.
10776 With @option{-fbranch-probabilities}, GCC puts a
10777 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
10778 These can be used to improve optimization. Currently, they are only
10779 used in one place: in @file{reorg.c}, instead of guessing which path a
10780 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
10781 exactly determine which path is taken more often.
10783 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10785 @item -fprofile-values
10786 @opindex fprofile-values
10787 If combined with @option{-fprofile-arcs}, it adds code so that some
10788 data about values of expressions in the program is gathered.
10790 With @option{-fbranch-probabilities}, it reads back the data gathered
10791 from profiling values of expressions for usage in optimizations.
10793 Enabled by @option{-fprofile-generate}, @option{-fprofile-use}, and
10794 @option{-fauto-profile}.
10796 @item -fprofile-reorder-functions
10797 @opindex fprofile-reorder-functions
10798 Function reordering based on profile instrumentation collects
10799 first time of execution of a function and orders these functions
10800 in ascending order.
10802 Enabled with @option{-fprofile-use}.
10806 If combined with @option{-fprofile-arcs}, this option instructs the compiler
10807 to add code to gather information about values of expressions.
10809 With @option{-fbranch-probabilities}, it reads back the data gathered
10810 and actually performs the optimizations based on them.
10811 Currently the optimizations include specialization of division operations
10812 using the knowledge about the value of the denominator.
10814 Enabled with @option{-fprofile-use} and @option{-fauto-profile}.
10816 @item -frename-registers
10817 @opindex frename-registers
10818 Attempt to avoid false dependencies in scheduled code by making use
10819 of registers left over after register allocation. This optimization
10820 most benefits processors with lots of registers. Depending on the
10821 debug information format adopted by the target, however, it can
10822 make debugging impossible, since variables no longer stay in
10823 a ``home register''.
10825 Enabled by default with @option{-funroll-loops}.
10827 @item -fschedule-fusion
10828 @opindex fschedule-fusion
10829 Performs a target dependent pass over the instruction stream to schedule
10830 instructions of same type together because target machine can execute them
10831 more efficiently if they are adjacent to each other in the instruction flow.
10833 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
10837 Perform tail duplication to enlarge superblock size. This transformation
10838 simplifies the control flow of the function allowing other optimizations to do
10841 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10843 @item -funroll-loops
10844 @opindex funroll-loops
10845 Unroll loops whose number of iterations can be determined at compile time or
10846 upon entry to the loop. @option{-funroll-loops} implies
10847 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
10848 It also turns on complete loop peeling (i.e.@: complete removal of loops with
10849 a small constant number of iterations). This option makes code larger, and may
10850 or may not make it run faster.
10852 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10854 @item -funroll-all-loops
10855 @opindex funroll-all-loops
10856 Unroll all loops, even if their number of iterations is uncertain when
10857 the loop is entered. This usually makes programs run more slowly.
10858 @option{-funroll-all-loops} implies the same options as
10859 @option{-funroll-loops}.
10862 @opindex fpeel-loops
10863 Peels loops for which there is enough information that they do not
10864 roll much (from profile feedback or static analysis). It also turns on
10865 complete loop peeling (i.e.@: complete removal of loops with small constant
10866 number of iterations).
10868 Enabled by @option{-O3}, @option{-fprofile-use}, and @option{-fauto-profile}.
10870 @item -fmove-loop-invariants
10871 @opindex fmove-loop-invariants
10872 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
10873 at level @option{-O1} and higher, except for @option{-Og}.
10875 @item -fsplit-loops
10876 @opindex fsplit-loops
10877 Split a loop into two if it contains a condition that's always true
10878 for one side of the iteration space and false for the other.
10880 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10882 @item -funswitch-loops
10883 @opindex funswitch-loops
10884 Move branches with loop invariant conditions out of the loop, with duplicates
10885 of the loop on both branches (modified according to result of the condition).
10887 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10889 @item -fversion-loops-for-strides
10890 @opindex fversion-loops-for-strides
10891 If a loop iterates over an array with a variable stride, create another
10892 version of the loop that assumes the stride is always one. For example:
10895 for (int i = 0; i < n; ++i)
10896 x[i * stride] = @dots{};
10903 for (int i = 0; i < n; ++i)
10906 for (int i = 0; i < n; ++i)
10907 x[i * stride] = @dots{};
10910 This is particularly useful for assumed-shape arrays in Fortran where
10911 (for example) it allows better vectorization assuming contiguous accesses.
10912 This flag is enabled by default at @option{-O3}.
10913 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10915 @item -ffunction-sections
10916 @itemx -fdata-sections
10917 @opindex ffunction-sections
10918 @opindex fdata-sections
10919 Place each function or data item into its own section in the output
10920 file if the target supports arbitrary sections. The name of the
10921 function or the name of the data item determines the section's name
10922 in the output file.
10924 Use these options on systems where the linker can perform optimizations to
10925 improve locality of reference in the instruction space. Most systems using the
10926 ELF object format have linkers with such optimizations. On AIX, the linker
10927 rearranges sections (CSECTs) based on the call graph. The performance impact
10930 Together with a linker garbage collection (linker @option{--gc-sections}
10931 option) these options may lead to smaller statically-linked executables (after
10934 On ELF/DWARF systems these options do not degenerate the quality of the debug
10935 information. There could be issues with other object files/debug info formats.
10937 Only use these options when there are significant benefits from doing so. When
10938 you specify these options, the assembler and linker create larger object and
10939 executable files and are also slower. These options affect code generation.
10940 They prevent optimizations by the compiler and assembler using relative
10941 locations inside a translation unit since the locations are unknown until
10942 link time. An example of such an optimization is relaxing calls to short call
10945 @item -fbranch-target-load-optimize
10946 @opindex fbranch-target-load-optimize
10947 Perform branch target register load optimization before prologue / epilogue
10949 The use of target registers can typically be exposed only during reload,
10950 thus hoisting loads out of loops and doing inter-block scheduling needs
10951 a separate optimization pass.
10953 @item -fbranch-target-load-optimize2
10954 @opindex fbranch-target-load-optimize2
10955 Perform branch target register load optimization after prologue / epilogue
10958 @item -fbtr-bb-exclusive
10959 @opindex fbtr-bb-exclusive
10960 When performing branch target register load optimization, don't reuse
10961 branch target registers within any basic block.
10964 @opindex fstdarg-opt
10965 Optimize the prologue of variadic argument functions with respect to usage of
10968 @item -fsection-anchors
10969 @opindex fsection-anchors
10970 Try to reduce the number of symbolic address calculations by using
10971 shared ``anchor'' symbols to address nearby objects. This transformation
10972 can help to reduce the number of GOT entries and GOT accesses on some
10975 For example, the implementation of the following function @code{foo}:
10978 static int a, b, c;
10979 int foo (void) @{ return a + b + c; @}
10983 usually calculates the addresses of all three variables, but if you
10984 compile it with @option{-fsection-anchors}, it accesses the variables
10985 from a common anchor point instead. The effect is similar to the
10986 following pseudocode (which isn't valid C):
10991 register int *xr = &x;
10992 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
10996 Not all targets support this option.
10998 @item --param @var{name}=@var{value}
11000 In some places, GCC uses various constants to control the amount of
11001 optimization that is done. For example, GCC does not inline functions
11002 that contain more than a certain number of instructions. You can
11003 control some of these constants on the command line using the
11004 @option{--param} option.
11006 The names of specific parameters, and the meaning of the values, are
11007 tied to the internals of the compiler, and are subject to change
11008 without notice in future releases.
11010 In order to get minimal, maximal and default value of a parameter,
11011 one can use @option{--help=param -Q} options.
11013 In each case, the @var{value} is an integer. The allowable choices for
11017 @item predictable-branch-outcome
11018 When branch is predicted to be taken with probability lower than this threshold
11019 (in percent), then it is considered well predictable.
11021 @item max-rtl-if-conversion-insns
11022 RTL if-conversion tries to remove conditional branches around a block and
11023 replace them with conditionally executed instructions. This parameter
11024 gives the maximum number of instructions in a block which should be
11025 considered for if-conversion. The compiler will
11026 also use other heuristics to decide whether if-conversion is likely to be
11029 @item max-rtl-if-conversion-predictable-cost
11030 @itemx max-rtl-if-conversion-unpredictable-cost
11031 RTL if-conversion will try to remove conditional branches around a block
11032 and replace them with conditionally executed instructions. These parameters
11033 give the maximum permissible cost for the sequence that would be generated
11034 by if-conversion depending on whether the branch is statically determined
11035 to be predictable or not. The units for this parameter are the same as
11036 those for the GCC internal seq_cost metric. The compiler will try to
11037 provide a reasonable default for this parameter using the BRANCH_COST
11040 @item max-crossjump-edges
11041 The maximum number of incoming edges to consider for cross-jumping.
11042 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
11043 the number of edges incoming to each block. Increasing values mean
11044 more aggressive optimization, making the compilation time increase with
11045 probably small improvement in executable size.
11047 @item min-crossjump-insns
11048 The minimum number of instructions that must be matched at the end
11049 of two blocks before cross-jumping is performed on them. This
11050 value is ignored in the case where all instructions in the block being
11051 cross-jumped from are matched.
11053 @item max-grow-copy-bb-insns
11054 The maximum code size expansion factor when copying basic blocks
11055 instead of jumping. The expansion is relative to a jump instruction.
11057 @item max-goto-duplication-insns
11058 The maximum number of instructions to duplicate to a block that jumps
11059 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
11060 passes, GCC factors computed gotos early in the compilation process,
11061 and unfactors them as late as possible. Only computed jumps at the
11062 end of a basic blocks with no more than max-goto-duplication-insns are
11065 @item max-delay-slot-insn-search
11066 The maximum number of instructions to consider when looking for an
11067 instruction to fill a delay slot. If more than this arbitrary number of
11068 instructions are searched, the time savings from filling the delay slot
11069 are minimal, so stop searching. Increasing values mean more
11070 aggressive optimization, making the compilation time increase with probably
11071 small improvement in execution time.
11073 @item max-delay-slot-live-search
11074 When trying to fill delay slots, the maximum number of instructions to
11075 consider when searching for a block with valid live register
11076 information. Increasing this arbitrarily chosen value means more
11077 aggressive optimization, increasing the compilation time. This parameter
11078 should be removed when the delay slot code is rewritten to maintain the
11079 control-flow graph.
11081 @item max-gcse-memory
11082 The approximate maximum amount of memory that can be allocated in
11083 order to perform the global common subexpression elimination
11084 optimization. If more memory than specified is required, the
11085 optimization is not done.
11087 @item max-gcse-insertion-ratio
11088 If the ratio of expression insertions to deletions is larger than this value
11089 for any expression, then RTL PRE inserts or removes the expression and thus
11090 leaves partially redundant computations in the instruction stream.
11092 @item max-pending-list-length
11093 The maximum number of pending dependencies scheduling allows
11094 before flushing the current state and starting over. Large functions
11095 with few branches or calls can create excessively large lists which
11096 needlessly consume memory and resources.
11098 @item max-modulo-backtrack-attempts
11099 The maximum number of backtrack attempts the scheduler should make
11100 when modulo scheduling a loop. Larger values can exponentially increase
11103 @item max-inline-insns-single
11104 Several parameters control the tree inliner used in GCC@.
11105 This number sets the maximum number of instructions (counted in GCC's
11106 internal representation) in a single function that the tree inliner
11107 considers for inlining. This only affects functions declared
11108 inline and methods implemented in a class declaration (C++).
11110 @item max-inline-insns-auto
11111 When you use @option{-finline-functions} (included in @option{-O3}),
11112 a lot of functions that would otherwise not be considered for inlining
11113 by the compiler are investigated. To those functions, a different
11114 (more restrictive) limit compared to functions declared inline can
11117 @item max-inline-insns-small
11118 This is bound applied to calls which are considered relevant with
11119 @option{-finline-small-functions}.
11121 @item max-inline-insns-size
11122 This is bound applied to calls which are optimized for size. Small growth
11123 may be desirable to anticipate optimization oppurtunities exposed by inlining.
11125 @item uninlined-function-insns
11126 Number of instructions accounted by inliner for function overhead such as
11127 function prologue and epilogue.
11129 @item uninlined-function-time
11130 Extra time accounted by inliner for function overhead such as time needed to
11131 execute function prologue and epilogue
11133 @item uninlined-thunk-insns
11134 @item uninlined-thunk-time
11135 Same as @option{--param uninlined-function-insns} and
11136 @option{--param uninlined-function-time} but applied to function thunks
11138 @item inline-min-speedup
11139 When estimated performance improvement of caller + callee runtime exceeds this
11140 threshold (in percent), the function can be inlined regardless of the limit on
11141 @option{--param max-inline-insns-single} and @option{--param
11142 max-inline-insns-auto}.
11144 @item large-function-insns
11145 The limit specifying really large functions. For functions larger than this
11146 limit after inlining, inlining is constrained by
11147 @option{--param large-function-growth}. This parameter is useful primarily
11148 to avoid extreme compilation time caused by non-linear algorithms used by the
11151 @item large-function-growth
11152 Specifies maximal growth of large function caused by inlining in percents.
11153 For example, parameter value 100 limits large function growth to 2.0 times
11156 @item large-unit-insns
11157 The limit specifying large translation unit. Growth caused by inlining of
11158 units larger than this limit is limited by @option{--param inline-unit-growth}.
11159 For small units this might be too tight.
11160 For example, consider a unit consisting of function A
11161 that is inline and B that just calls A three times. If B is small relative to
11162 A, the growth of unit is 300\% and yet such inlining is very sane. For very
11163 large units consisting of small inlineable functions, however, the overall unit
11164 growth limit is needed to avoid exponential explosion of code size. Thus for
11165 smaller units, the size is increased to @option{--param large-unit-insns}
11166 before applying @option{--param inline-unit-growth}.
11168 @item inline-unit-growth
11169 Specifies maximal overall growth of the compilation unit caused by inlining.
11170 For example, parameter value 20 limits unit growth to 1.2 times the original
11171 size. Cold functions (either marked cold via an attribute or by profile
11172 feedback) are not accounted into the unit size.
11174 @item ipcp-unit-growth
11175 Specifies maximal overall growth of the compilation unit caused by
11176 interprocedural constant propagation. For example, parameter value 10 limits
11177 unit growth to 1.1 times the original size.
11179 @item large-stack-frame
11180 The limit specifying large stack frames. While inlining the algorithm is trying
11181 to not grow past this limit too much.
11183 @item large-stack-frame-growth
11184 Specifies maximal growth of large stack frames caused by inlining in percents.
11185 For example, parameter value 1000 limits large stack frame growth to 11 times
11188 @item max-inline-insns-recursive
11189 @itemx max-inline-insns-recursive-auto
11190 Specifies the maximum number of instructions an out-of-line copy of a
11191 self-recursive inline
11192 function can grow into by performing recursive inlining.
11194 @option{--param max-inline-insns-recursive} applies to functions
11196 For functions not declared inline, recursive inlining
11197 happens only when @option{-finline-functions} (included in @option{-O3}) is
11198 enabled; @option{--param max-inline-insns-recursive-auto} applies instead.
11200 @item max-inline-recursive-depth
11201 @itemx max-inline-recursive-depth-auto
11202 Specifies the maximum recursion depth used for recursive inlining.
11204 @option{--param max-inline-recursive-depth} applies to functions
11205 declared inline. For functions not declared inline, recursive inlining
11206 happens only when @option{-finline-functions} (included in @option{-O3}) is
11207 enabled; @option{--param max-inline-recursive-depth-auto} applies instead.
11209 @item min-inline-recursive-probability
11210 Recursive inlining is profitable only for function having deep recursion
11211 in average and can hurt for function having little recursion depth by
11212 increasing the prologue size or complexity of function body to other
11215 When profile feedback is available (see @option{-fprofile-generate}) the actual
11216 recursion depth can be guessed from the probability that function recurses
11217 via a given call expression. This parameter limits inlining only to call
11218 expressions whose probability exceeds the given threshold (in percents).
11220 @item early-inlining-insns
11221 Specify growth that the early inliner can make. In effect it increases
11222 the amount of inlining for code having a large abstraction penalty.
11224 @item max-early-inliner-iterations
11225 Limit of iterations of the early inliner. This basically bounds
11226 the number of nested indirect calls the early inliner can resolve.
11227 Deeper chains are still handled by late inlining.
11229 @item comdat-sharing-probability
11230 Probability (in percent) that C++ inline function with comdat visibility
11231 are shared across multiple compilation units.
11233 @item profile-func-internal-id
11234 A parameter to control whether to use function internal id in profile
11235 database lookup. If the value is 0, the compiler uses an id that
11236 is based on function assembler name and filename, which makes old profile
11237 data more tolerant to source changes such as function reordering etc.
11239 @item min-vect-loop-bound
11240 The minimum number of iterations under which loops are not vectorized
11241 when @option{-ftree-vectorize} is used. The number of iterations after
11242 vectorization needs to be greater than the value specified by this option
11243 to allow vectorization.
11245 @item gcse-cost-distance-ratio
11246 Scaling factor in calculation of maximum distance an expression
11247 can be moved by GCSE optimizations. This is currently supported only in the
11248 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
11249 is with simple expressions, i.e., the expressions that have cost
11250 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
11251 hoisting of simple expressions.
11253 @item gcse-unrestricted-cost
11254 Cost, roughly measured as the cost of a single typical machine
11255 instruction, at which GCSE optimizations do not constrain
11256 the distance an expression can travel. This is currently
11257 supported only in the code hoisting pass. The lesser the cost,
11258 the more aggressive code hoisting is. Specifying 0
11259 allows all expressions to travel unrestricted distances.
11261 @item max-hoist-depth
11262 The depth of search in the dominator tree for expressions to hoist.
11263 This is used to avoid quadratic behavior in hoisting algorithm.
11264 The value of 0 does not limit on the search, but may slow down compilation
11267 @item max-tail-merge-comparisons
11268 The maximum amount of similar bbs to compare a bb with. This is used to
11269 avoid quadratic behavior in tree tail merging.
11271 @item max-tail-merge-iterations
11272 The maximum amount of iterations of the pass over the function. This is used to
11273 limit compilation time in tree tail merging.
11275 @item store-merging-allow-unaligned
11276 Allow the store merging pass to introduce unaligned stores if it is legal to
11279 @item max-stores-to-merge
11280 The maximum number of stores to attempt to merge into wider stores in the store
11283 @item max-unrolled-insns
11284 The maximum number of instructions that a loop may have to be unrolled.
11285 If a loop is unrolled, this parameter also determines how many times
11286 the loop code is unrolled.
11288 @item max-average-unrolled-insns
11289 The maximum number of instructions biased by probabilities of their execution
11290 that a loop may have to be unrolled. If a loop is unrolled,
11291 this parameter also determines how many times the loop code is unrolled.
11293 @item max-unroll-times
11294 The maximum number of unrollings of a single loop.
11296 @item max-peeled-insns
11297 The maximum number of instructions that a loop may have to be peeled.
11298 If a loop is peeled, this parameter also determines how many times
11299 the loop code is peeled.
11301 @item max-peel-times
11302 The maximum number of peelings of a single loop.
11304 @item max-peel-branches
11305 The maximum number of branches on the hot path through the peeled sequence.
11307 @item max-completely-peeled-insns
11308 The maximum number of insns of a completely peeled loop.
11310 @item max-completely-peel-times
11311 The maximum number of iterations of a loop to be suitable for complete peeling.
11313 @item max-completely-peel-loop-nest-depth
11314 The maximum depth of a loop nest suitable for complete peeling.
11316 @item max-unswitch-insns
11317 The maximum number of insns of an unswitched loop.
11319 @item max-unswitch-level
11320 The maximum number of branches unswitched in a single loop.
11322 @item lim-expensive
11323 The minimum cost of an expensive expression in the loop invariant motion.
11325 @item iv-consider-all-candidates-bound
11326 Bound on number of candidates for induction variables, below which
11327 all candidates are considered for each use in induction variable
11328 optimizations. If there are more candidates than this,
11329 only the most relevant ones are considered to avoid quadratic time complexity.
11331 @item iv-max-considered-uses
11332 The induction variable optimizations give up on loops that contain more
11333 induction variable uses.
11335 @item iv-always-prune-cand-set-bound
11336 If the number of candidates in the set is smaller than this value,
11337 always try to remove unnecessary ivs from the set
11338 when adding a new one.
11340 @item avg-loop-niter
11341 Average number of iterations of a loop.
11343 @item dse-max-object-size
11344 Maximum size (in bytes) of objects tracked bytewise by dead store elimination.
11345 Larger values may result in larger compilation times.
11347 @item dse-max-alias-queries-per-store
11348 Maximum number of queries into the alias oracle per store.
11349 Larger values result in larger compilation times and may result in more
11350 removed dead stores.
11352 @item scev-max-expr-size
11353 Bound on size of expressions used in the scalar evolutions analyzer.
11354 Large expressions slow the analyzer.
11356 @item scev-max-expr-complexity
11357 Bound on the complexity of the expressions in the scalar evolutions analyzer.
11358 Complex expressions slow the analyzer.
11360 @item max-tree-if-conversion-phi-args
11361 Maximum number of arguments in a PHI supported by TREE if conversion
11362 unless the loop is marked with simd pragma.
11364 @item vect-max-version-for-alignment-checks
11365 The maximum number of run-time checks that can be performed when
11366 doing loop versioning for alignment in the vectorizer.
11368 @item vect-max-version-for-alias-checks
11369 The maximum number of run-time checks that can be performed when
11370 doing loop versioning for alias in the vectorizer.
11372 @item vect-max-peeling-for-alignment
11373 The maximum number of loop peels to enhance access alignment
11374 for vectorizer. Value -1 means no limit.
11376 @item max-iterations-to-track
11377 The maximum number of iterations of a loop the brute-force algorithm
11378 for analysis of the number of iterations of the loop tries to evaluate.
11380 @item hot-bb-count-ws-permille
11381 A basic block profile count is considered hot if it contributes to
11382 the given permillage (i.e.@: 0...1000) of the entire profiled execution.
11384 @item hot-bb-frequency-fraction
11385 Select fraction of the entry block frequency of executions of basic block in
11386 function given basic block needs to have to be considered hot.
11388 @item max-predicted-iterations
11389 The maximum number of loop iterations we predict statically. This is useful
11390 in cases where a function contains a single loop with known bound and
11391 another loop with unknown bound.
11392 The known number of iterations is predicted correctly, while
11393 the unknown number of iterations average to roughly 10. This means that the
11394 loop without bounds appears artificially cold relative to the other one.
11396 @item builtin-expect-probability
11397 Control the probability of the expression having the specified value. This
11398 parameter takes a percentage (i.e.@: 0 ... 100) as input.
11400 @item builtin-string-cmp-inline-length
11401 The maximum length of a constant string for a builtin string cmp call
11402 eligible for inlining.
11404 @item align-threshold
11406 Select fraction of the maximal frequency of executions of a basic block in
11407 a function to align the basic block.
11409 @item align-loop-iterations
11411 A loop expected to iterate at least the selected number of iterations is
11414 @item tracer-dynamic-coverage
11415 @itemx tracer-dynamic-coverage-feedback
11417 This value is used to limit superblock formation once the given percentage of
11418 executed instructions is covered. This limits unnecessary code size
11421 The @option{tracer-dynamic-coverage-feedback} parameter
11422 is used only when profile
11423 feedback is available. The real profiles (as opposed to statically estimated
11424 ones) are much less balanced allowing the threshold to be larger value.
11426 @item tracer-max-code-growth
11427 Stop tail duplication once code growth has reached given percentage. This is
11428 a rather artificial limit, as most of the duplicates are eliminated later in
11429 cross jumping, so it may be set to much higher values than is the desired code
11432 @item tracer-min-branch-ratio
11434 Stop reverse growth when the reverse probability of best edge is less than this
11435 threshold (in percent).
11437 @item tracer-min-branch-probability
11438 @itemx tracer-min-branch-probability-feedback
11440 Stop forward growth if the best edge has probability lower than this
11443 Similarly to @option{tracer-dynamic-coverage} two parameters are
11444 provided. @option{tracer-min-branch-probability-feedback} is used for
11445 compilation with profile feedback and @option{tracer-min-branch-probability}
11446 compilation without. The value for compilation with profile feedback
11447 needs to be more conservative (higher) in order to make tracer
11450 @item stack-clash-protection-guard-size
11451 Specify the size of the operating system provided stack guard as
11452 2 raised to @var{num} bytes. Higher values may reduce the
11453 number of explicit probes, but a value larger than the operating system
11454 provided guard will leave code vulnerable to stack clash style attacks.
11456 @item stack-clash-protection-probe-interval
11457 Stack clash protection involves probing stack space as it is allocated. This
11458 param controls the maximum distance between probes into the stack as 2 raised
11459 to @var{num} bytes. Higher values may reduce the number of explicit probes, but a value
11460 larger than the operating system provided guard will leave code vulnerable to
11461 stack clash style attacks.
11463 @item max-cse-path-length
11465 The maximum number of basic blocks on path that CSE considers.
11467 @item max-cse-insns
11468 The maximum number of instructions CSE processes before flushing.
11470 @item ggc-min-expand
11472 GCC uses a garbage collector to manage its own memory allocation. This
11473 parameter specifies the minimum percentage by which the garbage
11474 collector's heap should be allowed to expand between collections.
11475 Tuning this may improve compilation speed; it has no effect on code
11478 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
11479 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
11480 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
11481 GCC is not able to calculate RAM on a particular platform, the lower
11482 bound of 30% is used. Setting this parameter and
11483 @option{ggc-min-heapsize} to zero causes a full collection to occur at
11484 every opportunity. This is extremely slow, but can be useful for
11487 @item ggc-min-heapsize
11489 Minimum size of the garbage collector's heap before it begins bothering
11490 to collect garbage. The first collection occurs after the heap expands
11491 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
11492 tuning this may improve compilation speed, and has no effect on code
11495 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
11496 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
11497 with a lower bound of 4096 (four megabytes) and an upper bound of
11498 131072 (128 megabytes). If GCC is not able to calculate RAM on a
11499 particular platform, the lower bound is used. Setting this parameter
11500 very large effectively disables garbage collection. Setting this
11501 parameter and @option{ggc-min-expand} to zero causes a full collection
11502 to occur at every opportunity.
11504 @item max-reload-search-insns
11505 The maximum number of instruction reload should look backward for equivalent
11506 register. Increasing values mean more aggressive optimization, making the
11507 compilation time increase with probably slightly better performance.
11509 @item max-cselib-memory-locations
11510 The maximum number of memory locations cselib should take into account.
11511 Increasing values mean more aggressive optimization, making the compilation time
11512 increase with probably slightly better performance.
11514 @item max-sched-ready-insns
11515 The maximum number of instructions ready to be issued the scheduler should
11516 consider at any given time during the first scheduling pass. Increasing
11517 values mean more thorough searches, making the compilation time increase
11518 with probably little benefit.
11520 @item max-sched-region-blocks
11521 The maximum number of blocks in a region to be considered for
11522 interblock scheduling.
11524 @item max-pipeline-region-blocks
11525 The maximum number of blocks in a region to be considered for
11526 pipelining in the selective scheduler.
11528 @item max-sched-region-insns
11529 The maximum number of insns in a region to be considered for
11530 interblock scheduling.
11532 @item max-pipeline-region-insns
11533 The maximum number of insns in a region to be considered for
11534 pipelining in the selective scheduler.
11536 @item min-spec-prob
11537 The minimum probability (in percents) of reaching a source block
11538 for interblock speculative scheduling.
11540 @item max-sched-extend-regions-iters
11541 The maximum number of iterations through CFG to extend regions.
11542 A value of 0 disables region extensions.
11544 @item max-sched-insn-conflict-delay
11545 The maximum conflict delay for an insn to be considered for speculative motion.
11547 @item sched-spec-prob-cutoff
11548 The minimal probability of speculation success (in percents), so that
11549 speculative insns are scheduled.
11551 @item sched-state-edge-prob-cutoff
11552 The minimum probability an edge must have for the scheduler to save its
11555 @item sched-mem-true-dep-cost
11556 Minimal distance (in CPU cycles) between store and load targeting same
11559 @item selsched-max-lookahead
11560 The maximum size of the lookahead window of selective scheduling. It is a
11561 depth of search for available instructions.
11563 @item selsched-max-sched-times
11564 The maximum number of times that an instruction is scheduled during
11565 selective scheduling. This is the limit on the number of iterations
11566 through which the instruction may be pipelined.
11568 @item selsched-insns-to-rename
11569 The maximum number of best instructions in the ready list that are considered
11570 for renaming in the selective scheduler.
11573 The minimum value of stage count that swing modulo scheduler
11576 @item max-last-value-rtl
11577 The maximum size measured as number of RTLs that can be recorded in an expression
11578 in combiner for a pseudo register as last known value of that register.
11580 @item max-combine-insns
11581 The maximum number of instructions the RTL combiner tries to combine.
11583 @item integer-share-limit
11584 Small integer constants can use a shared data structure, reducing the
11585 compiler's memory usage and increasing its speed. This sets the maximum
11586 value of a shared integer constant.
11588 @item ssp-buffer-size
11589 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
11590 protection when @option{-fstack-protection} is used.
11592 @item min-size-for-stack-sharing
11593 The minimum size of variables taking part in stack slot sharing when not
11596 @item max-jump-thread-duplication-stmts
11597 Maximum number of statements allowed in a block that needs to be
11598 duplicated when threading jumps.
11600 @item max-fields-for-field-sensitive
11601 Maximum number of fields in a structure treated in
11602 a field sensitive manner during pointer analysis.
11604 @item prefetch-latency
11605 Estimate on average number of instructions that are executed before
11606 prefetch finishes. The distance prefetched ahead is proportional
11607 to this constant. Increasing this number may also lead to less
11608 streams being prefetched (see @option{simultaneous-prefetches}).
11610 @item simultaneous-prefetches
11611 Maximum number of prefetches that can run at the same time.
11613 @item l1-cache-line-size
11614 The size of cache line in L1 data cache, in bytes.
11616 @item l1-cache-size
11617 The size of L1 data cache, in kilobytes.
11619 @item l2-cache-size
11620 The size of L2 data cache, in kilobytes.
11622 @item prefetch-dynamic-strides
11623 Whether the loop array prefetch pass should issue software prefetch hints
11624 for strides that are non-constant. In some cases this may be
11625 beneficial, though the fact the stride is non-constant may make it
11626 hard to predict when there is clear benefit to issuing these hints.
11628 Set to 1 if the prefetch hints should be issued for non-constant
11629 strides. Set to 0 if prefetch hints should be issued only for strides that
11630 are known to be constant and below @option{prefetch-minimum-stride}.
11632 @item prefetch-minimum-stride
11633 Minimum constant stride, in bytes, to start using prefetch hints for. If
11634 the stride is less than this threshold, prefetch hints will not be issued.
11636 This setting is useful for processors that have hardware prefetchers, in
11637 which case there may be conflicts between the hardware prefetchers and
11638 the software prefetchers. If the hardware prefetchers have a maximum
11639 stride they can handle, it should be used here to improve the use of
11640 software prefetchers.
11642 A value of -1 means we don't have a threshold and therefore
11643 prefetch hints can be issued for any constant stride.
11645 This setting is only useful for strides that are known and constant.
11647 @item loop-interchange-max-num-stmts
11648 The maximum number of stmts in a loop to be interchanged.
11650 @item loop-interchange-stride-ratio
11651 The minimum ratio between stride of two loops for interchange to be profitable.
11653 @item min-insn-to-prefetch-ratio
11654 The minimum ratio between the number of instructions and the
11655 number of prefetches to enable prefetching in a loop.
11657 @item prefetch-min-insn-to-mem-ratio
11658 The minimum ratio between the number of instructions and the
11659 number of memory references to enable prefetching in a loop.
11661 @item use-canonical-types
11662 Whether the compiler should use the ``canonical'' type system.
11663 Should always be 1, which uses a more efficient internal
11664 mechanism for comparing types in C++ and Objective-C++. However, if
11665 bugs in the canonical type system are causing compilation failures,
11666 set this value to 0 to disable canonical types.
11668 @item switch-conversion-max-branch-ratio
11669 Switch initialization conversion refuses to create arrays that are
11670 bigger than @option{switch-conversion-max-branch-ratio} times the number of
11671 branches in the switch.
11673 @item max-partial-antic-length
11674 Maximum length of the partial antic set computed during the tree
11675 partial redundancy elimination optimization (@option{-ftree-pre}) when
11676 optimizing at @option{-O3} and above. For some sorts of source code
11677 the enhanced partial redundancy elimination optimization can run away,
11678 consuming all of the memory available on the host machine. This
11679 parameter sets a limit on the length of the sets that are computed,
11680 which prevents the runaway behavior. Setting a value of 0 for
11681 this parameter allows an unlimited set length.
11683 @item rpo-vn-max-loop-depth
11684 Maximum loop depth that is value-numbered optimistically.
11685 When the limit hits the innermost
11686 @var{rpo-vn-max-loop-depth} loops and the outermost loop in the
11687 loop nest are value-numbered optimistically and the remaining ones not.
11689 @item sccvn-max-alias-queries-per-access
11690 Maximum number of alias-oracle queries we perform when looking for
11691 redundancies for loads and stores. If this limit is hit the search
11692 is aborted and the load or store is not considered redundant. The
11693 number of queries is algorithmically limited to the number of
11694 stores on all paths from the load to the function entry.
11696 @item ira-max-loops-num
11697 IRA uses regional register allocation by default. If a function
11698 contains more loops than the number given by this parameter, only at most
11699 the given number of the most frequently-executed loops form regions
11700 for regional register allocation.
11702 @item ira-max-conflict-table-size
11703 Although IRA uses a sophisticated algorithm to compress the conflict
11704 table, the table can still require excessive amounts of memory for
11705 huge functions. If the conflict table for a function could be more
11706 than the size in MB given by this parameter, the register allocator
11707 instead uses a faster, simpler, and lower-quality
11708 algorithm that does not require building a pseudo-register conflict table.
11710 @item ira-loop-reserved-regs
11711 IRA can be used to evaluate more accurate register pressure in loops
11712 for decisions to move loop invariants (see @option{-O3}). The number
11713 of available registers reserved for some other purposes is given
11714 by this parameter. Default of the parameter
11715 is the best found from numerous experiments.
11717 @item lra-inheritance-ebb-probability-cutoff
11718 LRA tries to reuse values reloaded in registers in subsequent insns.
11719 This optimization is called inheritance. EBB is used as a region to
11720 do this optimization. The parameter defines a minimal fall-through
11721 edge probability in percentage used to add BB to inheritance EBB in
11722 LRA. The default value was chosen
11723 from numerous runs of SPEC2000 on x86-64.
11725 @item loop-invariant-max-bbs-in-loop
11726 Loop invariant motion can be very expensive, both in compilation time and
11727 in amount of needed compile-time memory, with very large loops. Loops
11728 with more basic blocks than this parameter won't have loop invariant
11729 motion optimization performed on them.
11731 @item loop-max-datarefs-for-datadeps
11732 Building data dependencies is expensive for very large loops. This
11733 parameter limits the number of data references in loops that are
11734 considered for data dependence analysis. These large loops are no
11735 handled by the optimizations using loop data dependencies.
11737 @item max-vartrack-size
11738 Sets a maximum number of hash table slots to use during variable
11739 tracking dataflow analysis of any function. If this limit is exceeded
11740 with variable tracking at assignments enabled, analysis for that
11741 function is retried without it, after removing all debug insns from
11742 the function. If the limit is exceeded even without debug insns, var
11743 tracking analysis is completely disabled for the function. Setting
11744 the parameter to zero makes it unlimited.
11746 @item max-vartrack-expr-depth
11747 Sets a maximum number of recursion levels when attempting to map
11748 variable names or debug temporaries to value expressions. This trades
11749 compilation time for more complete debug information. If this is set too
11750 low, value expressions that are available and could be represented in
11751 debug information may end up not being used; setting this higher may
11752 enable the compiler to find more complex debug expressions, but compile
11753 time and memory use may grow.
11755 @item max-debug-marker-count
11756 Sets a threshold on the number of debug markers (e.g.@: begin stmt
11757 markers) to avoid complexity explosion at inlining or expanding to RTL.
11758 If a function has more such gimple stmts than the set limit, such stmts
11759 will be dropped from the inlined copy of a function, and from its RTL
11762 @item min-nondebug-insn-uid
11763 Use uids starting at this parameter for nondebug insns. The range below
11764 the parameter is reserved exclusively for debug insns created by
11765 @option{-fvar-tracking-assignments}, but debug insns may get
11766 (non-overlapping) uids above it if the reserved range is exhausted.
11768 @item ipa-sra-ptr-growth-factor
11769 IPA-SRA replaces a pointer to an aggregate with one or more new
11770 parameters only when their cumulative size is less or equal to
11771 @option{ipa-sra-ptr-growth-factor} times the size of the original
11774 @item sra-max-scalarization-size-Ospeed
11775 @itemx sra-max-scalarization-size-Osize
11776 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
11777 replace scalar parts of aggregates with uses of independent scalar
11778 variables. These parameters control the maximum size, in storage units,
11779 of aggregate which is considered for replacement when compiling for
11781 (@option{sra-max-scalarization-size-Ospeed}) or size
11782 (@option{sra-max-scalarization-size-Osize}) respectively.
11784 @item tm-max-aggregate-size
11785 When making copies of thread-local variables in a transaction, this
11786 parameter specifies the size in bytes after which variables are
11787 saved with the logging functions as opposed to save/restore code
11788 sequence pairs. This option only applies when using
11791 @item graphite-max-nb-scop-params
11792 To avoid exponential effects in the Graphite loop transforms, the
11793 number of parameters in a Static Control Part (SCoP) is bounded.
11794 A value of zero can be used to lift
11795 the bound. A variable whose value is unknown at compilation time and
11796 defined outside a SCoP is a parameter of the SCoP.
11798 @item loop-block-tile-size
11799 Loop blocking or strip mining transforms, enabled with
11800 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
11801 loop in the loop nest by a given number of iterations. The strip
11802 length can be changed using the @option{loop-block-tile-size}
11805 @item ipa-cp-value-list-size
11806 IPA-CP attempts to track all possible values and types passed to a function's
11807 parameter in order to propagate them and perform devirtualization.
11808 @option{ipa-cp-value-list-size} is the maximum number of values and types it
11809 stores per one formal parameter of a function.
11811 @item ipa-cp-eval-threshold
11812 IPA-CP calculates its own score of cloning profitability heuristics
11813 and performs those cloning opportunities with scores that exceed
11814 @option{ipa-cp-eval-threshold}.
11816 @item ipa-cp-recursion-penalty
11817 Percentage penalty the recursive functions will receive when they
11818 are evaluated for cloning.
11820 @item ipa-cp-single-call-penalty
11821 Percentage penalty functions containing a single call to another
11822 function will receive when they are evaluated for cloning.
11824 @item ipa-max-agg-items
11825 IPA-CP is also capable to propagate a number of scalar values passed
11826 in an aggregate. @option{ipa-max-agg-items} controls the maximum
11827 number of such values per one parameter.
11829 @item ipa-cp-loop-hint-bonus
11830 When IPA-CP determines that a cloning candidate would make the number
11831 of iterations of a loop known, it adds a bonus of
11832 @option{ipa-cp-loop-hint-bonus} to the profitability score of
11835 @item ipa-cp-array-index-hint-bonus
11836 When IPA-CP determines that a cloning candidate would make the index of
11837 an array access known, it adds a bonus of
11838 @option{ipa-cp-array-index-hint-bonus} to the profitability
11839 score of the candidate.
11841 @item ipa-max-aa-steps
11842 During its analysis of function bodies, IPA-CP employs alias analysis
11843 in order to track values pointed to by function parameters. In order
11844 not spend too much time analyzing huge functions, it gives up and
11845 consider all memory clobbered after examining
11846 @option{ipa-max-aa-steps} statements modifying memory.
11848 @item lto-partitions
11849 Specify desired number of partitions produced during WHOPR compilation.
11850 The number of partitions should exceed the number of CPUs used for compilation.
11852 @item lto-min-partition
11853 Size of minimal partition for WHOPR (in estimated instructions).
11854 This prevents expenses of splitting very small programs into too many
11857 @item lto-max-partition
11858 Size of max partition for WHOPR (in estimated instructions).
11859 to provide an upper bound for individual size of partition.
11860 Meant to be used only with balanced partitioning.
11862 @item cxx-max-namespaces-for-diagnostic-help
11863 The maximum number of namespaces to consult for suggestions when C++
11864 name lookup fails for an identifier.
11866 @item sink-frequency-threshold
11867 The maximum relative execution frequency (in percents) of the target block
11868 relative to a statement's original block to allow statement sinking of a
11869 statement. Larger numbers result in more aggressive statement sinking.
11870 A small positive adjustment is applied for
11871 statements with memory operands as those are even more profitable so sink.
11873 @item max-stores-to-sink
11874 The maximum number of conditional store pairs that can be sunk. Set to 0
11875 if either vectorization (@option{-ftree-vectorize}) or if-conversion
11876 (@option{-ftree-loop-if-convert}) is disabled.
11878 @item allow-store-data-races
11879 Allow optimizers to introduce new data races on stores.
11880 Set to 1 to allow, otherwise to 0.
11882 @item case-values-threshold
11883 The smallest number of different values for which it is best to use a
11884 jump-table instead of a tree of conditional branches. If the value is
11885 0, use the default for the machine.
11887 @item tree-reassoc-width
11888 Set the maximum number of instructions executed in parallel in
11889 reassociated tree. This parameter overrides target dependent
11890 heuristics used by default if has non zero value.
11892 @item sched-pressure-algorithm
11893 Choose between the two available implementations of
11894 @option{-fsched-pressure}. Algorithm 1 is the original implementation
11895 and is the more likely to prevent instructions from being reordered.
11896 Algorithm 2 was designed to be a compromise between the relatively
11897 conservative approach taken by algorithm 1 and the rather aggressive
11898 approach taken by the default scheduler. It relies more heavily on
11899 having a regular register file and accurate register pressure classes.
11900 See @file{haifa-sched.c} in the GCC sources for more details.
11902 The default choice depends on the target.
11904 @item max-slsr-cand-scan
11905 Set the maximum number of existing candidates that are considered when
11906 seeking a basis for a new straight-line strength reduction candidate.
11909 Enable buffer overflow detection for global objects. This kind
11910 of protection is enabled by default if you are using
11911 @option{-fsanitize=address} option.
11912 To disable global objects protection use @option{--param asan-globals=0}.
11915 Enable buffer overflow detection for stack objects. This kind of
11916 protection is enabled by default when using @option{-fsanitize=address}.
11917 To disable stack protection use @option{--param asan-stack=0} option.
11919 @item asan-instrument-reads
11920 Enable buffer overflow detection for memory reads. This kind of
11921 protection is enabled by default when using @option{-fsanitize=address}.
11922 To disable memory reads protection use
11923 @option{--param asan-instrument-reads=0}.
11925 @item asan-instrument-writes
11926 Enable buffer overflow detection for memory writes. This kind of
11927 protection is enabled by default when using @option{-fsanitize=address}.
11928 To disable memory writes protection use
11929 @option{--param asan-instrument-writes=0} option.
11931 @item asan-memintrin
11932 Enable detection for built-in functions. This kind of protection
11933 is enabled by default when using @option{-fsanitize=address}.
11934 To disable built-in functions protection use
11935 @option{--param asan-memintrin=0}.
11937 @item asan-use-after-return
11938 Enable detection of use-after-return. This kind of protection
11939 is enabled by default when using the @option{-fsanitize=address} option.
11940 To disable it use @option{--param asan-use-after-return=0}.
11942 Note: By default the check is disabled at run time. To enable it,
11943 add @code{detect_stack_use_after_return=1} to the environment variable
11944 @env{ASAN_OPTIONS}.
11946 @item asan-instrumentation-with-call-threshold
11947 If number of memory accesses in function being instrumented
11948 is greater or equal to this number, use callbacks instead of inline checks.
11949 E.g. to disable inline code use
11950 @option{--param asan-instrumentation-with-call-threshold=0}.
11952 @item use-after-scope-direct-emission-threshold
11953 If the size of a local variable in bytes is smaller or equal to this
11954 number, directly poison (or unpoison) shadow memory instead of using
11955 run-time callbacks.
11957 @item max-fsm-thread-path-insns
11958 Maximum number of instructions to copy when duplicating blocks on a
11959 finite state automaton jump thread path.
11961 @item max-fsm-thread-length
11962 Maximum number of basic blocks on a finite state automaton jump thread
11965 @item max-fsm-thread-paths
11966 Maximum number of new jump thread paths to create for a finite state
11969 @item parloops-chunk-size
11970 Chunk size of omp schedule for loops parallelized by parloops.
11972 @item parloops-schedule
11973 Schedule type of omp schedule for loops parallelized by parloops (static,
11974 dynamic, guided, auto, runtime).
11976 @item parloops-min-per-thread
11977 The minimum number of iterations per thread of an innermost parallelized
11978 loop for which the parallelized variant is preferred over the single threaded
11979 one. Note that for a parallelized loop nest the
11980 minimum number of iterations of the outermost loop per thread is two.
11982 @item max-ssa-name-query-depth
11983 Maximum depth of recursion when querying properties of SSA names in things
11984 like fold routines. One level of recursion corresponds to following a
11987 @item hsa-gen-debug-stores
11988 Enable emission of special debug stores within HSA kernels which are
11989 then read and reported by libgomp plugin. Generation of these stores
11990 is disabled by default, use @option{--param hsa-gen-debug-stores=1} to
11993 @item max-speculative-devirt-maydefs
11994 The maximum number of may-defs we analyze when looking for a must-def
11995 specifying the dynamic type of an object that invokes a virtual call
11996 we may be able to devirtualize speculatively.
11998 @item max-vrp-switch-assertions
11999 The maximum number of assertions to add along the default edge of a switch
12000 statement during VRP.
12002 @item unroll-jam-min-percent
12003 The minimum percentage of memory references that must be optimized
12004 away for the unroll-and-jam transformation to be considered profitable.
12006 @item unroll-jam-max-unroll
12007 The maximum number of times the outer loop should be unrolled by
12008 the unroll-and-jam transformation.
12010 @item max-rtl-if-conversion-unpredictable-cost
12011 Maximum permissible cost for the sequence that would be generated
12012 by the RTL if-conversion pass for a branch that is considered unpredictable.
12014 @item max-variable-expansions-in-unroller
12015 If @option{-fvariable-expansion-in-unroller} is used, the maximum number
12016 of times that an individual variable will be expanded during loop unrolling.
12018 @item tracer-min-branch-probability-feedback
12019 Stop forward growth if the probability of best edge is less than
12020 this threshold (in percent). Used when profile feedback is available.
12022 @item partial-inlining-entry-probability
12023 Maximum probability of the entry BB of split region
12024 (in percent relative to entry BB of the function)
12025 to make partial inlining happen.
12027 @item max-tracked-strlens
12028 Maximum number of strings for which strlen optimization pass will
12029 track string lengths.
12031 @item gcse-after-reload-partial-fraction
12032 The threshold ratio for performing partial redundancy
12033 elimination after reload.
12035 @item gcse-after-reload-critical-fraction
12036 The threshold ratio of critical edges execution count that
12037 permit performing redundancy elimination after reload.
12039 @item max-loop-header-insns
12040 The maximum number of insns in loop header duplicated
12041 by the copy loop headers pass.
12043 @item vect-epilogues-nomask
12044 Enable loop epilogue vectorization using smaller vector size.
12046 @item slp-max-insns-in-bb
12047 Maximum number of instructions in basic block to be
12048 considered for SLP vectorization.
12050 @item avoid-fma-max-bits
12051 Maximum number of bits for which we avoid creating FMAs.
12053 @item sms-loop-average-count-threshold
12054 A threshold on the average loop count considered by the swing modulo scheduler.
12056 @item sms-dfa-history
12057 The number of cycles the swing modulo scheduler considers when checking
12058 conflicts using DFA.
12060 @item hot-bb-count-fraction
12061 Select fraction of the maximal count of repetitions of basic block
12062 in program given basic block needs
12063 to have to be considered hot (used in non-LTO mode)
12065 @item max-inline-insns-recursive-auto
12066 The maximum number of instructions non-inline function
12067 can grow to via recursive inlining.
12069 @item graphite-allow-codegen-errors
12070 Whether codegen errors should be ICEs when @option{-fchecking}.
12072 @item sms-max-ii-factor
12073 A factor for tuning the upper bound that swing modulo scheduler
12074 uses for scheduling a loop.
12076 @item lra-max-considered-reload-pseudos
12077 The max number of reload pseudos which are considered during
12078 spilling a non-reload pseudo.
12080 @item max-pow-sqrt-depth
12081 Maximum depth of sqrt chains to use when synthesizing exponentiation
12082 by a real constant.
12084 @item max-dse-active-local-stores
12085 Maximum number of active local stores in RTL dead store elimination.
12087 @item asan-instrument-allocas
12088 Enable asan allocas/VLAs protection.
12090 @item max-iterations-computation-cost
12091 Bound on the cost of an expression to compute the number of iterations.
12093 @item max-isl-operations
12094 Maximum number of isl operations, 0 means unlimited.
12096 @item graphite-max-arrays-per-scop
12097 Maximum number of arrays per scop.
12099 @item max-vartrack-reverse-op-size
12100 Max. size of loc list for which reverse ops should be added.
12102 @item unlikely-bb-count-fraction
12103 The minimum fraction of profile runs a given basic block execution count
12104 must be not to be considered unlikely.
12106 @item tracer-dynamic-coverage-feedback
12107 The percentage of function, weighted by execution frequency,
12108 that must be covered by trace formation.
12109 Used when profile feedback is available.
12111 @item max-inline-recursive-depth-auto
12112 The maximum depth of recursive inlining for non-inline functions.
12114 @item fsm-scale-path-stmts
12115 Scale factor to apply to the number of statements in a threading path
12116 when comparing to the number of (scaled) blocks.
12118 @item fsm-maximum-phi-arguments
12119 Maximum number of arguments a PHI may have before the FSM threader
12120 will not try to thread through its block.
12122 @item uninit-control-dep-attempts
12123 Maximum number of nested calls to search for control dependencies
12124 during uninitialized variable analysis.
12126 @item indir-call-topn-profile
12127 Track top N target addresses in indirect-call profile.
12129 @item max-once-peeled-insns
12130 The maximum number of insns of a peeled loop that rolls only once.
12132 @item sra-max-scalarization-size-Osize
12133 Maximum size, in storage units, of an aggregate
12134 which should be considered for scalarization when compiling for size.
12136 @item fsm-scale-path-blocks
12137 Scale factor to apply to the number of blocks in a threading path
12138 when comparing to the number of (scaled) statements.
12140 @item sched-autopref-queue-depth
12141 Hardware autoprefetcher scheduler model control flag.
12142 Number of lookahead cycles the model looks into; at '
12143 ' only enable instruction sorting heuristic.
12145 @item loop-versioning-max-inner-insns
12146 The maximum number of instructions that an inner loop can have
12147 before the loop versioning pass considers it too big to copy.
12149 @item loop-versioning-max-outer-insns
12150 The maximum number of instructions that an outer loop can have
12151 before the loop versioning pass considers it too big to copy,
12152 discounting any instructions in inner loops that directly benefit
12158 @node Instrumentation Options
12159 @section Program Instrumentation Options
12160 @cindex instrumentation options
12161 @cindex program instrumentation options
12162 @cindex run-time error checking options
12163 @cindex profiling options
12164 @cindex options, program instrumentation
12165 @cindex options, run-time error checking
12166 @cindex options, profiling
12168 GCC supports a number of command-line options that control adding
12169 run-time instrumentation to the code it normally generates.
12170 For example, one purpose of instrumentation is collect profiling
12171 statistics for use in finding program hot spots, code coverage
12172 analysis, or profile-guided optimizations.
12173 Another class of program instrumentation is adding run-time checking
12174 to detect programming errors like invalid pointer
12175 dereferences or out-of-bounds array accesses, as well as deliberately
12176 hostile attacks such as stack smashing or C++ vtable hijacking.
12177 There is also a general hook which can be used to implement other
12178 forms of tracing or function-level instrumentation for debug or
12179 program analysis purposes.
12182 @cindex @command{prof}
12183 @cindex @command{gprof}
12188 Generate extra code to write profile information suitable for the
12189 analysis program @command{prof} (for @option{-p}) or @command{gprof}
12190 (for @option{-pg}). You must use this option when compiling
12191 the source files you want data about, and you must also use it when
12194 You can use the function attribute @code{no_instrument_function} to
12195 suppress profiling of individual functions when compiling with these options.
12196 @xref{Common Function Attributes}.
12198 @item -fprofile-arcs
12199 @opindex fprofile-arcs
12200 Add code so that program flow @dfn{arcs} are instrumented. During
12201 execution the program records how many times each branch and call is
12202 executed and how many times it is taken or returns. On targets that support
12203 constructors with priority support, profiling properly handles constructors,
12204 destructors and C++ constructors (and destructors) of classes which are used
12205 as a type of a global variable.
12208 program exits it saves this data to a file called
12209 @file{@var{auxname}.gcda} for each source file. The data may be used for
12210 profile-directed optimizations (@option{-fbranch-probabilities}), or for
12211 test coverage analysis (@option{-ftest-coverage}). Each object file's
12212 @var{auxname} is generated from the name of the output file, if
12213 explicitly specified and it is not the final executable, otherwise it is
12214 the basename of the source file. In both cases any suffix is removed
12215 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
12216 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
12217 @xref{Cross-profiling}.
12219 @cindex @command{gcov}
12223 This option is used to compile and link code instrumented for coverage
12224 analysis. The option is a synonym for @option{-fprofile-arcs}
12225 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
12226 linking). See the documentation for those options for more details.
12231 Compile the source files with @option{-fprofile-arcs} plus optimization
12232 and code generation options. For test coverage analysis, use the
12233 additional @option{-ftest-coverage} option. You do not need to profile
12234 every source file in a program.
12237 Compile the source files additionally with @option{-fprofile-abs-path}
12238 to create absolute path names in the @file{.gcno} files. This allows
12239 @command{gcov} to find the correct sources in projects where compilations
12240 occur with different working directories.
12243 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
12244 (the latter implies the former).
12247 Run the program on a representative workload to generate the arc profile
12248 information. This may be repeated any number of times. You can run
12249 concurrent instances of your program, and provided that the file system
12250 supports locking, the data files will be correctly updated. Unless
12251 a strict ISO C dialect option is in effect, @code{fork} calls are
12252 detected and correctly handled without double counting.
12255 For profile-directed optimizations, compile the source files again with
12256 the same optimization and code generation options plus
12257 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
12258 Control Optimization}).
12261 For test coverage analysis, use @command{gcov} to produce human readable
12262 information from the @file{.gcno} and @file{.gcda} files. Refer to the
12263 @command{gcov} documentation for further information.
12267 With @option{-fprofile-arcs}, for each function of your program GCC
12268 creates a program flow graph, then finds a spanning tree for the graph.
12269 Only arcs that are not on the spanning tree have to be instrumented: the
12270 compiler adds code to count the number of times that these arcs are
12271 executed. When an arc is the only exit or only entrance to a block, the
12272 instrumentation code can be added to the block; otherwise, a new basic
12273 block must be created to hold the instrumentation code.
12276 @item -ftest-coverage
12277 @opindex ftest-coverage
12278 Produce a notes file that the @command{gcov} code-coverage utility
12279 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
12280 show program coverage. Each source file's note file is called
12281 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
12282 above for a description of @var{auxname} and instructions on how to
12283 generate test coverage data. Coverage data matches the source files
12284 more closely if you do not optimize.
12286 @item -fprofile-abs-path
12287 @opindex fprofile-abs-path
12288 Automatically convert relative source file names to absolute path names
12289 in the @file{.gcno} files. This allows @command{gcov} to find the correct
12290 sources in projects where compilations occur with different working
12293 @item -fprofile-dir=@var{path}
12294 @opindex fprofile-dir
12296 Set the directory to search for the profile data files in to @var{path}.
12297 This option affects only the profile data generated by
12298 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
12299 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
12300 and its related options. Both absolute and relative paths can be used.
12301 By default, GCC uses the current directory as @var{path}, thus the
12302 profile data file appears in the same directory as the object file.
12303 In order to prevent the file name clashing, if the object file name is
12304 not an absolute path, we mangle the absolute path of the
12305 @file{@var{sourcename}.gcda} file and use it as the file name of a
12308 When an executable is run in a massive parallel environment, it is recommended
12309 to save profile to different folders. That can be done with variables
12310 in @var{path} that are exported during run-time:
12318 value of environment variable @var{VAR}
12322 @item -fprofile-generate
12323 @itemx -fprofile-generate=@var{path}
12324 @opindex fprofile-generate
12326 Enable options usually used for instrumenting application to produce
12327 profile useful for later recompilation with profile feedback based
12328 optimization. You must use @option{-fprofile-generate} both when
12329 compiling and when linking your program.
12331 The following options are enabled:
12332 @option{-fprofile-arcs}, @option{-fprofile-values},
12333 @option{-finline-functions}, and @option{-fipa-bit-cp}.
12335 If @var{path} is specified, GCC looks at the @var{path} to find
12336 the profile feedback data files. See @option{-fprofile-dir}.
12338 To optimize the program based on the collected profile information, use
12339 @option{-fprofile-use}. @xref{Optimize Options}, for more information.
12341 @item -fprofile-update=@var{method}
12342 @opindex fprofile-update
12344 Alter the update method for an application instrumented for profile
12345 feedback based optimization. The @var{method} argument should be one of
12346 @samp{single}, @samp{atomic} or @samp{prefer-atomic}.
12347 The first one is useful for single-threaded applications,
12348 while the second one prevents profile corruption by emitting thread-safe code.
12350 @strong{Warning:} When an application does not properly join all threads
12351 (or creates an detached thread), a profile file can be still corrupted.
12353 Using @samp{prefer-atomic} would be transformed either to @samp{atomic},
12354 when supported by a target, or to @samp{single} otherwise. The GCC driver
12355 automatically selects @samp{prefer-atomic} when @option{-pthread}
12356 is present in the command line.
12358 @item -fprofile-filter-files=@var{regex}
12359 @opindex fprofile-filter-files
12361 Instrument only functions from files where names match
12362 any regular expression (separated by a semi-colon).
12364 For example, @option{-fprofile-filter-files=main.c;module.*.c} will instrument
12365 only @file{main.c} and all C files starting with 'module'.
12367 @item -fprofile-exclude-files=@var{regex}
12368 @opindex fprofile-exclude-files
12370 Instrument only functions from files where names do not match
12371 all the regular expressions (separated by a semi-colon).
12373 For example, @option{-fprofile-exclude-files=/usr/*} will prevent instrumentation
12374 of all files that are located in @file{/usr/} folder.
12376 @item -fsanitize=address
12377 @opindex fsanitize=address
12378 Enable AddressSanitizer, a fast memory error detector.
12379 Memory access instructions are instrumented to detect
12380 out-of-bounds and use-after-free bugs.
12381 The option enables @option{-fsanitize-address-use-after-scope}.
12382 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
12383 more details. The run-time behavior can be influenced using the
12384 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
12385 the available options are shown at startup of the instrumented program. See
12386 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
12387 for a list of supported options.
12388 The option cannot be combined with @option{-fsanitize=thread}.
12390 @item -fsanitize=kernel-address
12391 @opindex fsanitize=kernel-address
12392 Enable AddressSanitizer for Linux kernel.
12393 See @uref{https://github.com/google/kasan/wiki} for more details.
12395 @item -fsanitize=pointer-compare
12396 @opindex fsanitize=pointer-compare
12397 Instrument comparison operation (<, <=, >, >=) with pointer operands.
12398 The option must be combined with either @option{-fsanitize=kernel-address} or
12399 @option{-fsanitize=address}
12400 The option cannot be combined with @option{-fsanitize=thread}.
12401 Note: By default the check is disabled at run time. To enable it,
12402 add @code{detect_invalid_pointer_pairs=2} to the environment variable
12403 @env{ASAN_OPTIONS}. Using @code{detect_invalid_pointer_pairs=1} detects
12404 invalid operation only when both pointers are non-null.
12406 @item -fsanitize=pointer-subtract
12407 @opindex fsanitize=pointer-subtract
12408 Instrument subtraction with pointer operands.
12409 The option must be combined with either @option{-fsanitize=kernel-address} or
12410 @option{-fsanitize=address}
12411 The option cannot be combined with @option{-fsanitize=thread}.
12412 Note: By default the check is disabled at run time. To enable it,
12413 add @code{detect_invalid_pointer_pairs=2} to the environment variable
12414 @env{ASAN_OPTIONS}. Using @code{detect_invalid_pointer_pairs=1} detects
12415 invalid operation only when both pointers are non-null.
12417 @item -fsanitize=thread
12418 @opindex fsanitize=thread
12419 Enable ThreadSanitizer, a fast data race detector.
12420 Memory access instructions are instrumented to detect
12421 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
12422 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
12423 environment variable; see
12424 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
12426 The option cannot be combined with @option{-fsanitize=address},
12427 @option{-fsanitize=leak}.
12429 Note that sanitized atomic builtins cannot throw exceptions when
12430 operating on invalid memory addresses with non-call exceptions
12431 (@option{-fnon-call-exceptions}).
12433 @item -fsanitize=leak
12434 @opindex fsanitize=leak
12435 Enable LeakSanitizer, a memory leak detector.
12436 This option only matters for linking of executables and
12437 the executable is linked against a library that overrides @code{malloc}
12438 and other allocator functions. See
12439 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
12440 details. The run-time behavior can be influenced using the
12441 @env{LSAN_OPTIONS} environment variable.
12442 The option cannot be combined with @option{-fsanitize=thread}.
12444 @item -fsanitize=undefined
12445 @opindex fsanitize=undefined
12446 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
12447 Various computations are instrumented to detect undefined behavior
12448 at runtime. Current suboptions are:
12452 @item -fsanitize=shift
12453 @opindex fsanitize=shift
12454 This option enables checking that the result of a shift operation is
12455 not undefined. Note that what exactly is considered undefined differs
12456 slightly between C and C++, as well as between ISO C90 and C99, etc.
12457 This option has two suboptions, @option{-fsanitize=shift-base} and
12458 @option{-fsanitize=shift-exponent}.
12460 @item -fsanitize=shift-exponent
12461 @opindex fsanitize=shift-exponent
12462 This option enables checking that the second argument of a shift operation
12463 is not negative and is smaller than the precision of the promoted first
12466 @item -fsanitize=shift-base
12467 @opindex fsanitize=shift-base
12468 If the second argument of a shift operation is within range, check that the
12469 result of a shift operation is not undefined. Note that what exactly is
12470 considered undefined differs slightly between C and C++, as well as between
12471 ISO C90 and C99, etc.
12473 @item -fsanitize=integer-divide-by-zero
12474 @opindex fsanitize=integer-divide-by-zero
12475 Detect integer division by zero as well as @code{INT_MIN / -1} division.
12477 @item -fsanitize=unreachable
12478 @opindex fsanitize=unreachable
12479 With this option, the compiler turns the @code{__builtin_unreachable}
12480 call into a diagnostics message call instead. When reaching the
12481 @code{__builtin_unreachable} call, the behavior is undefined.
12483 @item -fsanitize=vla-bound
12484 @opindex fsanitize=vla-bound
12485 This option instructs the compiler to check that the size of a variable
12486 length array is positive.
12488 @item -fsanitize=null
12489 @opindex fsanitize=null
12490 This option enables pointer checking. Particularly, the application
12491 built with this option turned on will issue an error message when it
12492 tries to dereference a NULL pointer, or if a reference (possibly an
12493 rvalue reference) is bound to a NULL pointer, or if a method is invoked
12494 on an object pointed by a NULL pointer.
12496 @item -fsanitize=return
12497 @opindex fsanitize=return
12498 This option enables return statement checking. Programs
12499 built with this option turned on will issue an error message
12500 when the end of a non-void function is reached without actually
12501 returning a value. This option works in C++ only.
12503 @item -fsanitize=signed-integer-overflow
12504 @opindex fsanitize=signed-integer-overflow
12505 This option enables signed integer overflow checking. We check that
12506 the result of @code{+}, @code{*}, and both unary and binary @code{-}
12507 does not overflow in the signed arithmetics. Note, integer promotion
12508 rules must be taken into account. That is, the following is not an
12511 signed char a = SCHAR_MAX;
12515 @item -fsanitize=bounds
12516 @opindex fsanitize=bounds
12517 This option enables instrumentation of array bounds. Various out of bounds
12518 accesses are detected. Flexible array members, flexible array member-like
12519 arrays, and initializers of variables with static storage are not instrumented.
12521 @item -fsanitize=bounds-strict
12522 @opindex fsanitize=bounds-strict
12523 This option enables strict instrumentation of array bounds. Most out of bounds
12524 accesses are detected, including flexible array members and flexible array
12525 member-like arrays. Initializers of variables with static storage are not
12528 @item -fsanitize=alignment
12529 @opindex fsanitize=alignment
12531 This option enables checking of alignment of pointers when they are
12532 dereferenced, or when a reference is bound to insufficiently aligned target,
12533 or when a method or constructor is invoked on insufficiently aligned object.
12535 @item -fsanitize=object-size
12536 @opindex fsanitize=object-size
12537 This option enables instrumentation of memory references using the
12538 @code{__builtin_object_size} function. Various out of bounds pointer
12539 accesses are detected.
12541 @item -fsanitize=float-divide-by-zero
12542 @opindex fsanitize=float-divide-by-zero
12543 Detect floating-point division by zero. Unlike other similar options,
12544 @option{-fsanitize=float-divide-by-zero} is not enabled by
12545 @option{-fsanitize=undefined}, since floating-point division by zero can
12546 be a legitimate way of obtaining infinities and NaNs.
12548 @item -fsanitize=float-cast-overflow
12549 @opindex fsanitize=float-cast-overflow
12550 This option enables floating-point type to integer conversion checking.
12551 We check that the result of the conversion does not overflow.
12552 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
12553 not enabled by @option{-fsanitize=undefined}.
12554 This option does not work well with @code{FE_INVALID} exceptions enabled.
12556 @item -fsanitize=nonnull-attribute
12557 @opindex fsanitize=nonnull-attribute
12559 This option enables instrumentation of calls, checking whether null values
12560 are not passed to arguments marked as requiring a non-null value by the
12561 @code{nonnull} function attribute.
12563 @item -fsanitize=returns-nonnull-attribute
12564 @opindex fsanitize=returns-nonnull-attribute
12566 This option enables instrumentation of return statements in functions
12567 marked with @code{returns_nonnull} function attribute, to detect returning
12568 of null values from such functions.
12570 @item -fsanitize=bool
12571 @opindex fsanitize=bool
12573 This option enables instrumentation of loads from bool. If a value other
12574 than 0/1 is loaded, a run-time error is issued.
12576 @item -fsanitize=enum
12577 @opindex fsanitize=enum
12579 This option enables instrumentation of loads from an enum type. If
12580 a value outside the range of values for the enum type is loaded,
12581 a run-time error is issued.
12583 @item -fsanitize=vptr
12584 @opindex fsanitize=vptr
12586 This option enables instrumentation of C++ member function calls, member
12587 accesses and some conversions between pointers to base and derived classes,
12588 to verify the referenced object has the correct dynamic type.
12590 @item -fsanitize=pointer-overflow
12591 @opindex fsanitize=pointer-overflow
12593 This option enables instrumentation of pointer arithmetics. If the pointer
12594 arithmetics overflows, a run-time error is issued.
12596 @item -fsanitize=builtin
12597 @opindex fsanitize=builtin
12599 This option enables instrumentation of arguments to selected builtin
12600 functions. If an invalid value is passed to such arguments, a run-time
12601 error is issued. E.g.@ passing 0 as the argument to @code{__builtin_ctz}
12602 or @code{__builtin_clz} invokes undefined behavior and is diagnosed
12607 While @option{-ftrapv} causes traps for signed overflows to be emitted,
12608 @option{-fsanitize=undefined} gives a diagnostic message.
12609 This currently works only for the C family of languages.
12611 @item -fno-sanitize=all
12612 @opindex fno-sanitize=all
12614 This option disables all previously enabled sanitizers.
12615 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
12618 @item -fasan-shadow-offset=@var{number}
12619 @opindex fasan-shadow-offset
12620 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
12621 It is useful for experimenting with different shadow memory layouts in
12622 Kernel AddressSanitizer.
12624 @item -fsanitize-sections=@var{s1},@var{s2},...
12625 @opindex fsanitize-sections
12626 Sanitize global variables in selected user-defined sections. @var{si} may
12629 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
12630 @opindex fsanitize-recover
12631 @opindex fno-sanitize-recover
12632 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
12633 mentioned in comma-separated list of @var{opts}. Enabling this option
12634 for a sanitizer component causes it to attempt to continue
12635 running the program as if no error happened. This means multiple
12636 runtime errors can be reported in a single program run, and the exit
12637 code of the program may indicate success even when errors
12638 have been reported. The @option{-fno-sanitize-recover=} option
12639 can be used to alter
12640 this behavior: only the first detected error is reported
12641 and program then exits with a non-zero exit code.
12643 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
12644 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
12645 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
12646 @option{-fsanitize=bounds-strict},
12647 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
12648 For these sanitizers error recovery is turned on by default,
12649 except @option{-fsanitize=address}, for which this feature is experimental.
12650 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
12651 accepted, the former enables recovery for all sanitizers that support it,
12652 the latter disables recovery for all sanitizers that support it.
12654 Even if a recovery mode is turned on the compiler side, it needs to be also
12655 enabled on the runtime library side, otherwise the failures are still fatal.
12656 The runtime library defaults to @code{halt_on_error=0} for
12657 ThreadSanitizer and UndefinedBehaviorSanitizer, while default value for
12658 AddressSanitizer is @code{halt_on_error=1}. This can be overridden through
12659 setting the @code{halt_on_error} flag in the corresponding environment variable.
12661 Syntax without an explicit @var{opts} parameter is deprecated. It is
12662 equivalent to specifying an @var{opts} list of:
12665 undefined,float-cast-overflow,float-divide-by-zero,bounds-strict
12668 @item -fsanitize-address-use-after-scope
12669 @opindex fsanitize-address-use-after-scope
12670 Enable sanitization of local variables to detect use-after-scope bugs.
12671 The option sets @option{-fstack-reuse} to @samp{none}.
12673 @item -fsanitize-undefined-trap-on-error
12674 @opindex fsanitize-undefined-trap-on-error
12675 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
12676 report undefined behavior using @code{__builtin_trap} rather than
12677 a @code{libubsan} library routine. The advantage of this is that the
12678 @code{libubsan} library is not needed and is not linked in, so this
12679 is usable even in freestanding environments.
12681 @item -fsanitize-coverage=trace-pc
12682 @opindex fsanitize-coverage=trace-pc
12683 Enable coverage-guided fuzzing code instrumentation.
12684 Inserts a call to @code{__sanitizer_cov_trace_pc} into every basic block.
12686 @item -fsanitize-coverage=trace-cmp
12687 @opindex fsanitize-coverage=trace-cmp
12688 Enable dataflow guided fuzzing code instrumentation.
12689 Inserts a call to @code{__sanitizer_cov_trace_cmp1},
12690 @code{__sanitizer_cov_trace_cmp2}, @code{__sanitizer_cov_trace_cmp4} or
12691 @code{__sanitizer_cov_trace_cmp8} for integral comparison with both operands
12692 variable or @code{__sanitizer_cov_trace_const_cmp1},
12693 @code{__sanitizer_cov_trace_const_cmp2},
12694 @code{__sanitizer_cov_trace_const_cmp4} or
12695 @code{__sanitizer_cov_trace_const_cmp8} for integral comparison with one
12696 operand constant, @code{__sanitizer_cov_trace_cmpf} or
12697 @code{__sanitizer_cov_trace_cmpd} for float or double comparisons and
12698 @code{__sanitizer_cov_trace_switch} for switch statements.
12700 @item -fcf-protection=@r{[}full@r{|}branch@r{|}return@r{|}none@r{]}
12701 @opindex fcf-protection
12702 Enable code instrumentation of control-flow transfers to increase
12703 program security by checking that target addresses of control-flow
12704 transfer instructions (such as indirect function call, function return,
12705 indirect jump) are valid. This prevents diverting the flow of control
12706 to an unexpected target. This is intended to protect against such
12707 threats as Return-oriented Programming (ROP), and similarly
12708 call/jmp-oriented programming (COP/JOP).
12710 The value @code{branch} tells the compiler to implement checking of
12711 validity of control-flow transfer at the point of indirect branch
12712 instructions, i.e.@: call/jmp instructions. The value @code{return}
12713 implements checking of validity at the point of returning from a
12714 function. The value @code{full} is an alias for specifying both
12715 @code{branch} and @code{return}. The value @code{none} turns off
12718 The macro @code{__CET__} is defined when @option{-fcf-protection} is
12719 used. The first bit of @code{__CET__} is set to 1 for the value
12720 @code{branch} and the second bit of @code{__CET__} is set to 1 for
12723 You can also use the @code{nocf_check} attribute to identify
12724 which functions and calls should be skipped from instrumentation
12725 (@pxref{Function Attributes}).
12727 Currently the x86 GNU/Linux target provides an implementation based
12728 on Intel Control-flow Enforcement Technology (CET).
12730 @item -fstack-protector
12731 @opindex fstack-protector
12732 Emit extra code to check for buffer overflows, such as stack smashing
12733 attacks. This is done by adding a guard variable to functions with
12734 vulnerable objects. This includes functions that call @code{alloca}, and
12735 functions with buffers larger than 8 bytes. The guards are initialized
12736 when a function is entered and then checked when the function exits.
12737 If a guard check fails, an error message is printed and the program exits.
12739 @item -fstack-protector-all
12740 @opindex fstack-protector-all
12741 Like @option{-fstack-protector} except that all functions are protected.
12743 @item -fstack-protector-strong
12744 @opindex fstack-protector-strong
12745 Like @option{-fstack-protector} but includes additional functions to
12746 be protected --- those that have local array definitions, or have
12747 references to local frame addresses.
12749 @item -fstack-protector-explicit
12750 @opindex fstack-protector-explicit
12751 Like @option{-fstack-protector} but only protects those functions which
12752 have the @code{stack_protect} attribute.
12754 @item -fstack-check
12755 @opindex fstack-check
12756 Generate code to verify that you do not go beyond the boundary of the
12757 stack. You should specify this flag if you are running in an
12758 environment with multiple threads, but you only rarely need to specify it in
12759 a single-threaded environment since stack overflow is automatically
12760 detected on nearly all systems if there is only one stack.
12762 Note that this switch does not actually cause checking to be done; the
12763 operating system or the language runtime must do that. The switch causes
12764 generation of code to ensure that they see the stack being extended.
12766 You can additionally specify a string parameter: @samp{no} means no
12767 checking, @samp{generic} means force the use of old-style checking,
12768 @samp{specific} means use the best checking method and is equivalent
12769 to bare @option{-fstack-check}.
12771 Old-style checking is a generic mechanism that requires no specific
12772 target support in the compiler but comes with the following drawbacks:
12776 Modified allocation strategy for large objects: they are always
12777 allocated dynamically if their size exceeds a fixed threshold. Note this
12778 may change the semantics of some code.
12781 Fixed limit on the size of the static frame of functions: when it is
12782 topped by a particular function, stack checking is not reliable and
12783 a warning is issued by the compiler.
12786 Inefficiency: because of both the modified allocation strategy and the
12787 generic implementation, code performance is hampered.
12790 Note that old-style stack checking is also the fallback method for
12791 @samp{specific} if no target support has been added in the compiler.
12793 @samp{-fstack-check=} is designed for Ada's needs to detect infinite recursion
12794 and stack overflows. @samp{specific} is an excellent choice when compiling
12795 Ada code. It is not generally sufficient to protect against stack-clash
12796 attacks. To protect against those you want @samp{-fstack-clash-protection}.
12798 @item -fstack-clash-protection
12799 @opindex fstack-clash-protection
12800 Generate code to prevent stack clash style attacks. When this option is
12801 enabled, the compiler will only allocate one page of stack space at a time
12802 and each page is accessed immediately after allocation. Thus, it prevents
12803 allocations from jumping over any stack guard page provided by the
12806 Most targets do not fully support stack clash protection. However, on
12807 those targets @option{-fstack-clash-protection} will protect dynamic stack
12808 allocations. @option{-fstack-clash-protection} may also provide limited
12809 protection for static stack allocations if the target supports
12810 @option{-fstack-check=specific}.
12812 @item -fstack-limit-register=@var{reg}
12813 @itemx -fstack-limit-symbol=@var{sym}
12814 @itemx -fno-stack-limit
12815 @opindex fstack-limit-register
12816 @opindex fstack-limit-symbol
12817 @opindex fno-stack-limit
12818 Generate code to ensure that the stack does not grow beyond a certain value,
12819 either the value of a register or the address of a symbol. If a larger
12820 stack is required, a signal is raised at run time. For most targets,
12821 the signal is raised before the stack overruns the boundary, so
12822 it is possible to catch the signal without taking special precautions.
12824 For instance, if the stack starts at absolute address @samp{0x80000000}
12825 and grows downwards, you can use the flags
12826 @option{-fstack-limit-symbol=__stack_limit} and
12827 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
12828 of 128KB@. Note that this may only work with the GNU linker.
12830 You can locally override stack limit checking by using the
12831 @code{no_stack_limit} function attribute (@pxref{Function Attributes}).
12833 @item -fsplit-stack
12834 @opindex fsplit-stack
12835 Generate code to automatically split the stack before it overflows.
12836 The resulting program has a discontiguous stack which can only
12837 overflow if the program is unable to allocate any more memory. This
12838 is most useful when running threaded programs, as it is no longer
12839 necessary to calculate a good stack size to use for each thread. This
12840 is currently only implemented for the x86 targets running
12843 When code compiled with @option{-fsplit-stack} calls code compiled
12844 without @option{-fsplit-stack}, there may not be much stack space
12845 available for the latter code to run. If compiling all code,
12846 including library code, with @option{-fsplit-stack} is not an option,
12847 then the linker can fix up these calls so that the code compiled
12848 without @option{-fsplit-stack} always has a large stack. Support for
12849 this is implemented in the gold linker in GNU binutils release 2.21
12852 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
12853 @opindex fvtable-verify
12854 This option is only available when compiling C++ code.
12855 It turns on (or off, if using @option{-fvtable-verify=none}) the security
12856 feature that verifies at run time, for every virtual call, that
12857 the vtable pointer through which the call is made is valid for the type of
12858 the object, and has not been corrupted or overwritten. If an invalid vtable
12859 pointer is detected at run time, an error is reported and execution of the
12860 program is immediately halted.
12862 This option causes run-time data structures to be built at program startup,
12863 which are used for verifying the vtable pointers.
12864 The options @samp{std} and @samp{preinit}
12865 control the timing of when these data structures are built. In both cases the
12866 data structures are built before execution reaches @code{main}. Using
12867 @option{-fvtable-verify=std} causes the data structures to be built after
12868 shared libraries have been loaded and initialized.
12869 @option{-fvtable-verify=preinit} causes them to be built before shared
12870 libraries have been loaded and initialized.
12872 If this option appears multiple times in the command line with different
12873 values specified, @samp{none} takes highest priority over both @samp{std} and
12874 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
12877 @opindex fvtv-debug
12878 When used in conjunction with @option{-fvtable-verify=std} or
12879 @option{-fvtable-verify=preinit}, causes debug versions of the
12880 runtime functions for the vtable verification feature to be called.
12881 This flag also causes the compiler to log information about which
12882 vtable pointers it finds for each class.
12883 This information is written to a file named @file{vtv_set_ptr_data.log}
12884 in the directory named by the environment variable @env{VTV_LOGS_DIR}
12885 if that is defined or the current working directory otherwise.
12887 Note: This feature @emph{appends} data to the log file. If you want a fresh log
12888 file, be sure to delete any existing one.
12891 @opindex fvtv-counts
12892 This is a debugging flag. When used in conjunction with
12893 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
12894 causes the compiler to keep track of the total number of virtual calls
12895 it encounters and the number of verifications it inserts. It also
12896 counts the number of calls to certain run-time library functions
12897 that it inserts and logs this information for each compilation unit.
12898 The compiler writes this information to a file named
12899 @file{vtv_count_data.log} in the directory named by the environment
12900 variable @env{VTV_LOGS_DIR} if that is defined or the current working
12901 directory otherwise. It also counts the size of the vtable pointer sets
12902 for each class, and writes this information to @file{vtv_class_set_sizes.log}
12903 in the same directory.
12905 Note: This feature @emph{appends} data to the log files. To get fresh log
12906 files, be sure to delete any existing ones.
12908 @item -finstrument-functions
12909 @opindex finstrument-functions
12910 Generate instrumentation calls for entry and exit to functions. Just
12911 after function entry and just before function exit, the following
12912 profiling functions are called with the address of the current
12913 function and its call site. (On some platforms,
12914 @code{__builtin_return_address} does not work beyond the current
12915 function, so the call site information may not be available to the
12916 profiling functions otherwise.)
12919 void __cyg_profile_func_enter (void *this_fn,
12921 void __cyg_profile_func_exit (void *this_fn,
12925 The first argument is the address of the start of the current function,
12926 which may be looked up exactly in the symbol table.
12928 This instrumentation is also done for functions expanded inline in other
12929 functions. The profiling calls indicate where, conceptually, the
12930 inline function is entered and exited. This means that addressable
12931 versions of such functions must be available. If all your uses of a
12932 function are expanded inline, this may mean an additional expansion of
12933 code size. If you use @code{extern inline} in your C code, an
12934 addressable version of such functions must be provided. (This is
12935 normally the case anyway, but if you get lucky and the optimizer always
12936 expands the functions inline, you might have gotten away without
12937 providing static copies.)
12939 A function may be given the attribute @code{no_instrument_function}, in
12940 which case this instrumentation is not done. This can be used, for
12941 example, for the profiling functions listed above, high-priority
12942 interrupt routines, and any functions from which the profiling functions
12943 cannot safely be called (perhaps signal handlers, if the profiling
12944 routines generate output or allocate memory).
12945 @xref{Common Function Attributes}.
12947 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
12948 @opindex finstrument-functions-exclude-file-list
12950 Set the list of functions that are excluded from instrumentation (see
12951 the description of @option{-finstrument-functions}). If the file that
12952 contains a function definition matches with one of @var{file}, then
12953 that function is not instrumented. The match is done on substrings:
12954 if the @var{file} parameter is a substring of the file name, it is
12955 considered to be a match.
12960 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
12964 excludes any inline function defined in files whose pathnames
12965 contain @file{/bits/stl} or @file{include/sys}.
12967 If, for some reason, you want to include letter @samp{,} in one of
12968 @var{sym}, write @samp{\,}. For example,
12969 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
12970 (note the single quote surrounding the option).
12972 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
12973 @opindex finstrument-functions-exclude-function-list
12975 This is similar to @option{-finstrument-functions-exclude-file-list},
12976 but this option sets the list of function names to be excluded from
12977 instrumentation. The function name to be matched is its user-visible
12978 name, such as @code{vector<int> blah(const vector<int> &)}, not the
12979 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
12980 match is done on substrings: if the @var{sym} parameter is a substring
12981 of the function name, it is considered to be a match. For C99 and C++
12982 extended identifiers, the function name must be given in UTF-8, not
12983 using universal character names.
12985 @item -fpatchable-function-entry=@var{N}[,@var{M}]
12986 @opindex fpatchable-function-entry
12987 Generate @var{N} NOPs right at the beginning
12988 of each function, with the function entry point before the @var{M}th NOP.
12989 If @var{M} is omitted, it defaults to @code{0} so the
12990 function entry points to the address just at the first NOP.
12991 The NOP instructions reserve extra space which can be used to patch in
12992 any desired instrumentation at run time, provided that the code segment
12993 is writable. The amount of space is controllable indirectly via
12994 the number of NOPs; the NOP instruction used corresponds to the instruction
12995 emitted by the internal GCC back-end interface @code{gen_nop}. This behavior
12996 is target-specific and may also depend on the architecture variant and/or
12997 other compilation options.
12999 For run-time identification, the starting addresses of these areas,
13000 which correspond to their respective function entries minus @var{M},
13001 are additionally collected in the @code{__patchable_function_entries}
13002 section of the resulting binary.
13004 Note that the value of @code{__attribute__ ((patchable_function_entry
13005 (N,M)))} takes precedence over command-line option
13006 @option{-fpatchable-function-entry=N,M}. This can be used to increase
13007 the area size or to remove it completely on a single function.
13008 If @code{N=0}, no pad location is recorded.
13010 The NOP instructions are inserted at---and maybe before, depending on
13011 @var{M}---the function entry address, even before the prologue.
13016 @node Preprocessor Options
13017 @section Options Controlling the Preprocessor
13018 @cindex preprocessor options
13019 @cindex options, preprocessor
13021 These options control the C preprocessor, which is run on each C source
13022 file before actual compilation.
13024 If you use the @option{-E} option, nothing is done except preprocessing.
13025 Some of these options make sense only together with @option{-E} because
13026 they cause the preprocessor output to be unsuitable for actual
13029 In addition to the options listed here, there are a number of options
13030 to control search paths for include files documented in
13031 @ref{Directory Options}.
13032 Options to control preprocessor diagnostics are listed in
13033 @ref{Warning Options}.
13036 @include cppopts.texi
13038 @item -Wp,@var{option}
13040 You can use @option{-Wp,@var{option}} to bypass the compiler driver
13041 and pass @var{option} directly through to the preprocessor. If
13042 @var{option} contains commas, it is split into multiple options at the
13043 commas. However, many options are modified, translated or interpreted
13044 by the compiler driver before being passed to the preprocessor, and
13045 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
13046 interface is undocumented and subject to change, so whenever possible
13047 you should avoid using @option{-Wp} and let the driver handle the
13050 @item -Xpreprocessor @var{option}
13051 @opindex Xpreprocessor
13052 Pass @var{option} as an option to the preprocessor. You can use this to
13053 supply system-specific preprocessor options that GCC does not
13056 If you want to pass an option that takes an argument, you must use
13057 @option{-Xpreprocessor} twice, once for the option and once for the argument.
13059 @item -no-integrated-cpp
13060 @opindex no-integrated-cpp
13061 Perform preprocessing as a separate pass before compilation.
13062 By default, GCC performs preprocessing as an integrated part of
13063 input tokenization and parsing.
13064 If this option is provided, the appropriate language front end
13065 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
13066 and Objective-C, respectively) is instead invoked twice,
13067 once for preprocessing only and once for actual compilation
13068 of the preprocessed input.
13069 This option may be useful in conjunction with the @option{-B} or
13070 @option{-wrapper} options to specify an alternate preprocessor or
13071 perform additional processing of the program source between
13072 normal preprocessing and compilation.
13076 @node Assembler Options
13077 @section Passing Options to the Assembler
13079 @c prevent bad page break with this line
13080 You can pass options to the assembler.
13083 @item -Wa,@var{option}
13085 Pass @var{option} as an option to the assembler. If @var{option}
13086 contains commas, it is split into multiple options at the commas.
13088 @item -Xassembler @var{option}
13089 @opindex Xassembler
13090 Pass @var{option} as an option to the assembler. You can use this to
13091 supply system-specific assembler options that GCC does not
13094 If you want to pass an option that takes an argument, you must use
13095 @option{-Xassembler} twice, once for the option and once for the argument.
13100 @section Options for Linking
13101 @cindex link options
13102 @cindex options, linking
13104 These options come into play when the compiler links object files into
13105 an executable output file. They are meaningless if the compiler is
13106 not doing a link step.
13110 @item @var{object-file-name}
13111 A file name that does not end in a special recognized suffix is
13112 considered to name an object file or library. (Object files are
13113 distinguished from libraries by the linker according to the file
13114 contents.) If linking is done, these object files are used as input
13123 If any of these options is used, then the linker is not run, and
13124 object file names should not be used as arguments. @xref{Overall
13127 @item -flinker-output=@var{type}
13128 @opindex flinker-output
13129 This option controls the code generation of the link time optimizer. By
13130 default the linker output is determined by the linker plugin automatically. For
13131 debugging the compiler and in the case of incremental linking to non-lto object
13132 file is desired, it may be useful to control the type manually.
13134 If @var{type} is @samp{exec} the code generation is configured to produce static
13135 binary. In this case @option{-fpic} and @option{-fpie} are both disabled.
13137 If @var{type} is @samp{dyn} the code generation is configured to produce shared
13138 library. In this case @option{-fpic} or @option{-fPIC} is preserved, but not
13139 enabled automatically. This makes it possible to build shared libraries without
13140 position independent code on architectures this is possible, i.e.@: on x86.
13142 If @var{type} is @samp{pie} the code generation is configured to produce
13143 @option{-fpie} executable. This result in similar optimizations as @samp{exec}
13144 except that @option{-fpie} is not disabled if specified at compilation time.
13146 If @var{type} is @samp{rel} the compiler assumes that incremental linking is
13147 done. The sections containing intermediate code for link-time optimization are
13148 merged, pre-optimized, and output to the resulting object file. In addition, if
13149 @option{-ffat-lto-objects} is specified the binary code is produced for future
13150 non-lto linking. The object file produced by incremental linking will be smaller
13151 than a static library produced from the same object files. At link-time the
13152 result of incremental linking will also load faster to compiler than a static
13153 library assuming that majority of objects in the library are used.
13155 Finally @samp{nolto-rel} configure compiler to for incremental linking where
13156 code generation is forced, final binary is produced and the intermediate code
13157 for later link-time optimization is stripped. When multiple object files are
13158 linked together the resulting code will be optimized better than with link time
13159 optimizations disabled (for example, the cross-module inlining will happen),
13160 most of benefits of whole program optimizations are however lost.
13162 During the incremental link (by @option{-r}) the linker plugin will default to
13163 @option{rel}. With current interfaces to GNU Binutils it is however not
13164 possible to link incrementally LTO objects and non-LTO objects into a single
13165 mixed object file. In the case any of object files in incremental link cannot
13166 be used for link-time optimization the linker plugin will output warning and
13167 use @samp{nolto-rel}. To maintain the whole program optimization it is
13168 recommended to link such objects into static library instead. Alternatively it
13169 is possible to use H.J. Lu's binutils with support for mixed objects.
13172 @opindex fuse-ld=bfd
13173 Use the @command{bfd} linker instead of the default linker.
13175 @item -fuse-ld=gold
13176 @opindex fuse-ld=gold
13177 Use the @command{gold} linker instead of the default linker.
13180 @opindex fuse-ld=lld
13181 Use the LLVM @command{lld} linker instead of the default linker.
13184 @item -l@var{library}
13185 @itemx -l @var{library}
13187 Search the library named @var{library} when linking. (The second
13188 alternative with the library as a separate argument is only for
13189 POSIX compliance and is not recommended.)
13191 The @option{-l} option is passed directly to the linker by GCC. Refer
13192 to your linker documentation for exact details. The general
13193 description below applies to the GNU linker.
13195 The linker searches a standard list of directories for the library.
13196 The directories searched include several standard system directories
13197 plus any that you specify with @option{-L}.
13199 Static libraries are archives of object files, and have file names
13200 like @file{lib@var{library}.a}. Some targets also support shared
13201 libraries, which typically have names like @file{lib@var{library}.so}.
13202 If both static and shared libraries are found, the linker gives
13203 preference to linking with the shared library unless the
13204 @option{-static} option is used.
13206 It makes a difference where in the command you write this option; the
13207 linker searches and processes libraries and object files in the order they
13208 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
13209 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
13210 to functions in @samp{z}, those functions may not be loaded.
13214 You need this special case of the @option{-l} option in order to
13215 link an Objective-C or Objective-C++ program.
13217 @item -nostartfiles
13218 @opindex nostartfiles
13219 Do not use the standard system startup files when linking.
13220 The standard system libraries are used normally, unless @option{-nostdlib},
13221 @option{-nolibc}, or @option{-nodefaultlibs} is used.
13223 @item -nodefaultlibs
13224 @opindex nodefaultlibs
13225 Do not use the standard system libraries when linking.
13226 Only the libraries you specify are passed to the linker, and options
13227 specifying linkage of the system libraries, such as @option{-static-libgcc}
13228 or @option{-shared-libgcc}, are ignored.
13229 The standard startup files are used normally, unless @option{-nostartfiles}
13232 The compiler may generate calls to @code{memcmp},
13233 @code{memset}, @code{memcpy} and @code{memmove}.
13234 These entries are usually resolved by entries in
13235 libc. These entry points should be supplied through some other
13236 mechanism when this option is specified.
13240 Do not use the C library or system libraries tightly coupled with it when
13241 linking. Still link with the startup files, @file{libgcc} or toolchain
13242 provided language support libraries such as @file{libgnat}, @file{libgfortran}
13243 or @file{libstdc++} unless options preventing their inclusion are used as
13244 well. This typically removes @option{-lc} from the link command line, as well
13245 as system libraries that normally go with it and become meaningless when
13246 absence of a C library is assumed, for example @option{-lpthread} or
13247 @option{-lm} in some configurations. This is intended for bare-board
13248 targets when there is indeed no C library available.
13252 Do not use the standard system startup files or libraries when linking.
13253 No startup files and only the libraries you specify are passed to
13254 the linker, and options specifying linkage of the system libraries, such as
13255 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
13257 The compiler may generate calls to @code{memcmp}, @code{memset},
13258 @code{memcpy} and @code{memmove}.
13259 These entries are usually resolved by entries in
13260 libc. These entry points should be supplied through some other
13261 mechanism when this option is specified.
13263 @cindex @option{-lgcc}, use with @option{-nostdlib}
13264 @cindex @option{-nostdlib} and unresolved references
13265 @cindex unresolved references and @option{-nostdlib}
13266 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
13267 @cindex @option{-nodefaultlibs} and unresolved references
13268 @cindex unresolved references and @option{-nodefaultlibs}
13269 One of the standard libraries bypassed by @option{-nostdlib} and
13270 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
13271 which GCC uses to overcome shortcomings of particular machines, or special
13272 needs for some languages.
13273 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
13274 Collection (GCC) Internals},
13275 for more discussion of @file{libgcc.a}.)
13276 In most cases, you need @file{libgcc.a} even when you want to avoid
13277 other standard libraries. In other words, when you specify @option{-nostdlib}
13278 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
13279 This ensures that you have no unresolved references to internal GCC
13280 library subroutines.
13281 (An example of such an internal subroutine is @code{__main}, used to ensure C++
13282 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
13283 GNU Compiler Collection (GCC) Internals}.)
13285 @item -e @var{entry}
13286 @itemx --entry=@var{entry}
13290 Specify that the program entry point is @var{entry}. The argument is
13291 interpreted by the linker; the GNU linker accepts either a symbol name
13296 Produce a dynamically linked position independent executable on targets
13297 that support it. For predictable results, you must also specify the same
13298 set of options used for compilation (@option{-fpie}, @option{-fPIE},
13299 or model suboptions) when you specify this linker option.
13303 Don't produce a dynamically linked position independent executable.
13306 @opindex static-pie
13307 Produce a static position independent executable on targets that support
13308 it. A static position independent executable is similar to a static
13309 executable, but can be loaded at any address without a dynamic linker.
13310 For predictable results, you must also specify the same set of options
13311 used for compilation (@option{-fpie}, @option{-fPIE}, or model
13312 suboptions) when you specify this linker option.
13316 Link with the POSIX threads library. This option is supported on
13317 GNU/Linux targets, most other Unix derivatives, and also on
13318 x86 Cygwin and MinGW targets. On some targets this option also sets
13319 flags for the preprocessor, so it should be used consistently for both
13320 compilation and linking.
13324 Produce a relocatable object as output. This is also known as partial
13329 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
13330 that support it. This instructs the linker to add all symbols, not
13331 only used ones, to the dynamic symbol table. This option is needed
13332 for some uses of @code{dlopen} or to allow obtaining backtraces
13333 from within a program.
13337 Remove all symbol table and relocation information from the executable.
13341 On systems that support dynamic linking, this overrides @option{-pie}
13342 and prevents linking with the shared libraries. On other systems, this
13343 option has no effect.
13347 Produce a shared object which can then be linked with other objects to
13348 form an executable. Not all systems support this option. For predictable
13349 results, you must also specify the same set of options used for compilation
13350 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
13351 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
13352 needs to build supplementary stub code for constructors to work. On
13353 multi-libbed systems, @samp{gcc -shared} must select the correct support
13354 libraries to link against. Failing to supply the correct flags may lead
13355 to subtle defects. Supplying them in cases where they are not necessary
13358 @item -shared-libgcc
13359 @itemx -static-libgcc
13360 @opindex shared-libgcc
13361 @opindex static-libgcc
13362 On systems that provide @file{libgcc} as a shared library, these options
13363 force the use of either the shared or static version, respectively.
13364 If no shared version of @file{libgcc} was built when the compiler was
13365 configured, these options have no effect.
13367 There are several situations in which an application should use the
13368 shared @file{libgcc} instead of the static version. The most common
13369 of these is when the application wishes to throw and catch exceptions
13370 across different shared libraries. In that case, each of the libraries
13371 as well as the application itself should use the shared @file{libgcc}.
13373 Therefore, the G++ driver automatically adds @option{-shared-libgcc}
13374 whenever you build a shared library or a main executable, because C++
13375 programs typically use exceptions, so this is the right thing to do.
13377 If, instead, you use the GCC driver to create shared libraries, you may
13378 find that they are not always linked with the shared @file{libgcc}.
13379 If GCC finds, at its configuration time, that you have a non-GNU linker
13380 or a GNU linker that does not support option @option{--eh-frame-hdr},
13381 it links the shared version of @file{libgcc} into shared libraries
13382 by default. Otherwise, it takes advantage of the linker and optimizes
13383 away the linking with the shared version of @file{libgcc}, linking with
13384 the static version of libgcc by default. This allows exceptions to
13385 propagate through such shared libraries, without incurring relocation
13386 costs at library load time.
13388 However, if a library or main executable is supposed to throw or catch
13389 exceptions, you must link it using the G++ driver, or using the option
13390 @option{-shared-libgcc}, such that it is linked with the shared
13393 @item -static-libasan
13394 @opindex static-libasan
13395 When the @option{-fsanitize=address} option is used to link a program,
13396 the GCC driver automatically links against @option{libasan}. If
13397 @file{libasan} is available as a shared library, and the @option{-static}
13398 option is not used, then this links against the shared version of
13399 @file{libasan}. The @option{-static-libasan} option directs the GCC
13400 driver to link @file{libasan} statically, without necessarily linking
13401 other libraries statically.
13403 @item -static-libtsan
13404 @opindex static-libtsan
13405 When the @option{-fsanitize=thread} option is used to link a program,
13406 the GCC driver automatically links against @option{libtsan}. If
13407 @file{libtsan} is available as a shared library, and the @option{-static}
13408 option is not used, then this links against the shared version of
13409 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
13410 driver to link @file{libtsan} statically, without necessarily linking
13411 other libraries statically.
13413 @item -static-liblsan
13414 @opindex static-liblsan
13415 When the @option{-fsanitize=leak} option is used to link a program,
13416 the GCC driver automatically links against @option{liblsan}. If
13417 @file{liblsan} is available as a shared library, and the @option{-static}
13418 option is not used, then this links against the shared version of
13419 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
13420 driver to link @file{liblsan} statically, without necessarily linking
13421 other libraries statically.
13423 @item -static-libubsan
13424 @opindex static-libubsan
13425 When the @option{-fsanitize=undefined} option is used to link a program,
13426 the GCC driver automatically links against @option{libubsan}. If
13427 @file{libubsan} is available as a shared library, and the @option{-static}
13428 option is not used, then this links against the shared version of
13429 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
13430 driver to link @file{libubsan} statically, without necessarily linking
13431 other libraries statically.
13433 @item -static-libstdc++
13434 @opindex static-libstdc++
13435 When the @command{g++} program is used to link a C++ program, it
13436 normally automatically links against @option{libstdc++}. If
13437 @file{libstdc++} is available as a shared library, and the
13438 @option{-static} option is not used, then this links against the
13439 shared version of @file{libstdc++}. That is normally fine. However, it
13440 is sometimes useful to freeze the version of @file{libstdc++} used by
13441 the program without going all the way to a fully static link. The
13442 @option{-static-libstdc++} option directs the @command{g++} driver to
13443 link @file{libstdc++} statically, without necessarily linking other
13444 libraries statically.
13448 Bind references to global symbols when building a shared object. Warn
13449 about any unresolved references (unless overridden by the link editor
13450 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
13453 @item -T @var{script}
13455 @cindex linker script
13456 Use @var{script} as the linker script. This option is supported by most
13457 systems using the GNU linker. On some targets, such as bare-board
13458 targets without an operating system, the @option{-T} option may be required
13459 when linking to avoid references to undefined symbols.
13461 @item -Xlinker @var{option}
13463 Pass @var{option} as an option to the linker. You can use this to
13464 supply system-specific linker options that GCC does not recognize.
13466 If you want to pass an option that takes a separate argument, you must use
13467 @option{-Xlinker} twice, once for the option and once for the argument.
13468 For example, to pass @option{-assert definitions}, you must write
13469 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
13470 @option{-Xlinker "-assert definitions"}, because this passes the entire
13471 string as a single argument, which is not what the linker expects.
13473 When using the GNU linker, it is usually more convenient to pass
13474 arguments to linker options using the @option{@var{option}=@var{value}}
13475 syntax than as separate arguments. For example, you can specify
13476 @option{-Xlinker -Map=output.map} rather than
13477 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
13478 this syntax for command-line options.
13480 @item -Wl,@var{option}
13482 Pass @var{option} as an option to the linker. If @var{option} contains
13483 commas, it is split into multiple options at the commas. You can use this
13484 syntax to pass an argument to the option.
13485 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
13486 linker. When using the GNU linker, you can also get the same effect with
13487 @option{-Wl,-Map=output.map}.
13489 @item -u @var{symbol}
13491 Pretend the symbol @var{symbol} is undefined, to force linking of
13492 library modules to define it. You can use @option{-u} multiple times with
13493 different symbols to force loading of additional library modules.
13495 @item -z @var{keyword}
13497 @option{-z} is passed directly on to the linker along with the keyword
13498 @var{keyword}. See the section in the documentation of your linker for
13499 permitted values and their meanings.
13502 @node Directory Options
13503 @section Options for Directory Search
13504 @cindex directory options
13505 @cindex options, directory search
13506 @cindex search path
13508 These options specify directories to search for header files, for
13509 libraries and for parts of the compiler:
13512 @include cppdiropts.texi
13514 @item -iplugindir=@var{dir}
13515 @opindex iplugindir=
13516 Set the directory to search for plugins that are passed
13517 by @option{-fplugin=@var{name}} instead of
13518 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
13519 to be used by the user, but only passed by the driver.
13523 Add directory @var{dir} to the list of directories to be searched
13526 @item -B@var{prefix}
13528 This option specifies where to find the executables, libraries,
13529 include files, and data files of the compiler itself.
13531 The compiler driver program runs one or more of the subprograms
13532 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
13533 @var{prefix} as a prefix for each program it tries to run, both with and
13534 without @samp{@var{machine}/@var{version}/} for the corresponding target
13535 machine and compiler version.
13537 For each subprogram to be run, the compiler driver first tries the
13538 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
13539 is not specified, the driver tries two standard prefixes,
13540 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
13541 those results in a file name that is found, the unmodified program
13542 name is searched for using the directories specified in your
13543 @env{PATH} environment variable.
13545 The compiler checks to see if the path provided by @option{-B}
13546 refers to a directory, and if necessary it adds a directory
13547 separator character at the end of the path.
13549 @option{-B} prefixes that effectively specify directory names also apply
13550 to libraries in the linker, because the compiler translates these
13551 options into @option{-L} options for the linker. They also apply to
13552 include files in the preprocessor, because the compiler translates these
13553 options into @option{-isystem} options for the preprocessor. In this case,
13554 the compiler appends @samp{include} to the prefix.
13556 The runtime support file @file{libgcc.a} can also be searched for using
13557 the @option{-B} prefix, if needed. If it is not found there, the two
13558 standard prefixes above are tried, and that is all. The file is left
13559 out of the link if it is not found by those means.
13561 Another way to specify a prefix much like the @option{-B} prefix is to use
13562 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
13565 As a special kludge, if the path provided by @option{-B} is
13566 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
13567 9, then it is replaced by @file{[dir/]include}. This is to help
13568 with boot-strapping the compiler.
13570 @item -no-canonical-prefixes
13571 @opindex no-canonical-prefixes
13572 Do not expand any symbolic links, resolve references to @samp{/../}
13573 or @samp{/./}, or make the path absolute when generating a relative
13576 @item --sysroot=@var{dir}
13578 Use @var{dir} as the logical root directory for headers and libraries.
13579 For example, if the compiler normally searches for headers in
13580 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
13581 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
13583 If you use both this option and the @option{-isysroot} option, then
13584 the @option{--sysroot} option applies to libraries, but the
13585 @option{-isysroot} option applies to header files.
13587 The GNU linker (beginning with version 2.16) has the necessary support
13588 for this option. If your linker does not support this option, the
13589 header file aspect of @option{--sysroot} still works, but the
13590 library aspect does not.
13592 @item --no-sysroot-suffix
13593 @opindex no-sysroot-suffix
13594 For some targets, a suffix is added to the root directory specified
13595 with @option{--sysroot}, depending on the other options used, so that
13596 headers may for example be found in
13597 @file{@var{dir}/@var{suffix}/usr/include} instead of
13598 @file{@var{dir}/usr/include}. This option disables the addition of
13603 @node Code Gen Options
13604 @section Options for Code Generation Conventions
13605 @cindex code generation conventions
13606 @cindex options, code generation
13607 @cindex run-time options
13609 These machine-independent options control the interface conventions
13610 used in code generation.
13612 Most of them have both positive and negative forms; the negative form
13613 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
13614 one of the forms is listed---the one that is not the default. You
13615 can figure out the other form by either removing @samp{no-} or adding
13619 @item -fstack-reuse=@var{reuse-level}
13620 @opindex fstack_reuse
13621 This option controls stack space reuse for user declared local/auto variables
13622 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
13623 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
13624 local variables and temporaries, @samp{named_vars} enables the reuse only for
13625 user defined local variables with names, and @samp{none} disables stack reuse
13626 completely. The default value is @samp{all}. The option is needed when the
13627 program extends the lifetime of a scoped local variable or a compiler generated
13628 temporary beyond the end point defined by the language. When a lifetime of
13629 a variable ends, and if the variable lives in memory, the optimizing compiler
13630 has the freedom to reuse its stack space with other temporaries or scoped
13631 local variables whose live range does not overlap with it. Legacy code extending
13632 local lifetime is likely to break with the stack reuse optimization.
13651 if (*p == 10) // out of scope use of local1
13662 A(int k) : i(k), j(k) @{ @}
13669 void foo(const A& ar)
13676 foo(A(10)); // temp object's lifetime ends when foo returns
13682 ap->i+= 10; // ap references out of scope temp whose space
13683 // is reused with a. What is the value of ap->i?
13688 The lifetime of a compiler generated temporary is well defined by the C++
13689 standard. When a lifetime of a temporary ends, and if the temporary lives
13690 in memory, the optimizing compiler has the freedom to reuse its stack
13691 space with other temporaries or scoped local variables whose live range
13692 does not overlap with it. However some of the legacy code relies on
13693 the behavior of older compilers in which temporaries' stack space is
13694 not reused, the aggressive stack reuse can lead to runtime errors. This
13695 option is used to control the temporary stack reuse optimization.
13699 This option generates traps for signed overflow on addition, subtraction,
13700 multiplication operations.
13701 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
13702 @option{-ftrapv} @option{-fwrapv} on the command-line results in
13703 @option{-fwrapv} being effective. Note that only active options override, so
13704 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
13705 results in @option{-ftrapv} being effective.
13709 This option instructs the compiler to assume that signed arithmetic
13710 overflow of addition, subtraction and multiplication wraps around
13711 using twos-complement representation. This flag enables some optimizations
13712 and disables others.
13713 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
13714 @option{-ftrapv} @option{-fwrapv} on the command-line results in
13715 @option{-fwrapv} being effective. Note that only active options override, so
13716 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
13717 results in @option{-ftrapv} being effective.
13719 @item -fwrapv-pointer
13720 @opindex fwrapv-pointer
13721 This option instructs the compiler to assume that pointer arithmetic
13722 overflow on addition and subtraction wraps around using twos-complement
13723 representation. This flag disables some optimizations which assume
13724 pointer overflow is invalid.
13726 @item -fstrict-overflow
13727 @opindex fstrict-overflow
13728 This option implies @option{-fno-wrapv} @option{-fno-wrapv-pointer} and when
13729 negated implies @option{-fwrapv} @option{-fwrapv-pointer}.
13732 @opindex fexceptions
13733 Enable exception handling. Generates extra code needed to propagate
13734 exceptions. For some targets, this implies GCC generates frame
13735 unwind information for all functions, which can produce significant data
13736 size overhead, although it does not affect execution. If you do not
13737 specify this option, GCC enables it by default for languages like
13738 C++ that normally require exception handling, and disables it for
13739 languages like C that do not normally require it. However, you may need
13740 to enable this option when compiling C code that needs to interoperate
13741 properly with exception handlers written in C++. You may also wish to
13742 disable this option if you are compiling older C++ programs that don't
13743 use exception handling.
13745 @item -fnon-call-exceptions
13746 @opindex fnon-call-exceptions
13747 Generate code that allows trapping instructions to throw exceptions.
13748 Note that this requires platform-specific runtime support that does
13749 not exist everywhere. Moreover, it only allows @emph{trapping}
13750 instructions to throw exceptions, i.e.@: memory references or floating-point
13751 instructions. It does not allow exceptions to be thrown from
13752 arbitrary signal handlers such as @code{SIGALRM}.
13754 @item -fdelete-dead-exceptions
13755 @opindex fdelete-dead-exceptions
13756 Consider that instructions that may throw exceptions but don't otherwise
13757 contribute to the execution of the program can be optimized away.
13758 This option is enabled by default for the Ada front end, as permitted by
13759 the Ada language specification.
13760 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
13762 @item -funwind-tables
13763 @opindex funwind-tables
13764 Similar to @option{-fexceptions}, except that it just generates any needed
13765 static data, but does not affect the generated code in any other way.
13766 You normally do not need to enable this option; instead, a language processor
13767 that needs this handling enables it on your behalf.
13769 @item -fasynchronous-unwind-tables
13770 @opindex fasynchronous-unwind-tables
13771 Generate unwind table in DWARF format, if supported by target machine. The
13772 table is exact at each instruction boundary, so it can be used for stack
13773 unwinding from asynchronous events (such as debugger or garbage collector).
13775 @item -fno-gnu-unique
13776 @opindex fno-gnu-unique
13777 @opindex fgnu-unique
13778 On systems with recent GNU assembler and C library, the C++ compiler
13779 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
13780 of template static data members and static local variables in inline
13781 functions are unique even in the presence of @code{RTLD_LOCAL}; this
13782 is necessary to avoid problems with a library used by two different
13783 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
13784 therefore disagreeing with the other one about the binding of the
13785 symbol. But this causes @code{dlclose} to be ignored for affected
13786 DSOs; if your program relies on reinitialization of a DSO via
13787 @code{dlclose} and @code{dlopen}, you can use
13788 @option{-fno-gnu-unique}.
13790 @item -fpcc-struct-return
13791 @opindex fpcc-struct-return
13792 Return ``short'' @code{struct} and @code{union} values in memory like
13793 longer ones, rather than in registers. This convention is less
13794 efficient, but it has the advantage of allowing intercallability between
13795 GCC-compiled files and files compiled with other compilers, particularly
13796 the Portable C Compiler (pcc).
13798 The precise convention for returning structures in memory depends
13799 on the target configuration macros.
13801 Short structures and unions are those whose size and alignment match
13802 that of some integer type.
13804 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
13805 switch is not binary compatible with code compiled with the
13806 @option{-freg-struct-return} switch.
13807 Use it to conform to a non-default application binary interface.
13809 @item -freg-struct-return
13810 @opindex freg-struct-return
13811 Return @code{struct} and @code{union} values in registers when possible.
13812 This is more efficient for small structures than
13813 @option{-fpcc-struct-return}.
13815 If you specify neither @option{-fpcc-struct-return} nor
13816 @option{-freg-struct-return}, GCC defaults to whichever convention is
13817 standard for the target. If there is no standard convention, GCC
13818 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
13819 the principal compiler. In those cases, we can choose the standard, and
13820 we chose the more efficient register return alternative.
13822 @strong{Warning:} code compiled with the @option{-freg-struct-return}
13823 switch is not binary compatible with code compiled with the
13824 @option{-fpcc-struct-return} switch.
13825 Use it to conform to a non-default application binary interface.
13827 @item -fshort-enums
13828 @opindex fshort-enums
13829 Allocate to an @code{enum} type only as many bytes as it needs for the
13830 declared range of possible values. Specifically, the @code{enum} type
13831 is equivalent to the smallest integer type that has enough room.
13833 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
13834 code that is not binary compatible with code generated without that switch.
13835 Use it to conform to a non-default application binary interface.
13837 @item -fshort-wchar
13838 @opindex fshort-wchar
13839 Override the underlying type for @code{wchar_t} to be @code{short
13840 unsigned int} instead of the default for the target. This option is
13841 useful for building programs to run under WINE@.
13843 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
13844 code that is not binary compatible with code generated without that switch.
13845 Use it to conform to a non-default application binary interface.
13848 @opindex fno-common
13850 @cindex tentative definitions
13851 In C code, this option controls the placement of global variables
13852 defined without an initializer, known as @dfn{tentative definitions}
13853 in the C standard. Tentative definitions are distinct from declarations
13854 of a variable with the @code{extern} keyword, which do not allocate storage.
13856 Unix C compilers have traditionally allocated storage for
13857 uninitialized global variables in a common block. This allows the
13858 linker to resolve all tentative definitions of the same variable
13859 in different compilation units to the same object, or to a non-tentative
13861 This is the behavior specified by @option{-fcommon}, and is the default for
13862 GCC on most targets.
13863 On the other hand, this behavior is not required by ISO
13864 C, and on some targets may carry a speed or code size penalty on
13865 variable references.
13867 The @option{-fno-common} option specifies that the compiler should instead
13868 place uninitialized global variables in the BSS section of the object file.
13869 This inhibits the merging of tentative definitions by the linker so
13870 you get a multiple-definition error if the same
13871 variable is defined in more than one compilation unit.
13872 Compiling with @option{-fno-common} is useful on targets for which
13873 it provides better performance, or if you wish to verify that the
13874 program will work on other systems that always treat uninitialized
13875 variable definitions this way.
13880 Ignore the @code{#ident} directive.
13882 @item -finhibit-size-directive
13883 @opindex finhibit-size-directive
13884 Don't output a @code{.size} assembler directive, or anything else that
13885 would cause trouble if the function is split in the middle, and the
13886 two halves are placed at locations far apart in memory. This option is
13887 used when compiling @file{crtstuff.c}; you should not need to use it
13890 @item -fverbose-asm
13891 @opindex fverbose-asm
13892 Put extra commentary information in the generated assembly code to
13893 make it more readable. This option is generally only of use to those
13894 who actually need to read the generated assembly code (perhaps while
13895 debugging the compiler itself).
13897 @option{-fno-verbose-asm}, the default, causes the
13898 extra information to be omitted and is useful when comparing two assembler
13901 The added comments include:
13906 information on the compiler version and command-line options,
13909 the source code lines associated with the assembly instructions,
13910 in the form FILENAME:LINENUMBER:CONTENT OF LINE,
13913 hints on which high-level expressions correspond to
13914 the various assembly instruction operands.
13918 For example, given this C source file:
13926 for (i = 0; i < n; i++)
13933 compiling to (x86_64) assembly via @option{-S} and emitting the result
13934 direct to stdout via @option{-o} @option{-}
13937 gcc -S test.c -fverbose-asm -Os -o -
13940 gives output similar to this:
13944 # GNU C11 (GCC) version 7.0.0 20160809 (experimental) (x86_64-pc-linux-gnu)
13951 .type test, @@function
13955 # test.c:4: int total = 0;
13956 xorl %eax, %eax # <retval>
13957 # test.c:6: for (i = 0; i < n; i++)
13958 xorl %edx, %edx # i
13960 # test.c:6: for (i = 0; i < n; i++)
13961 cmpl %edi, %edx # n, i
13963 # test.c:7: total += i * i;
13964 movl %edx, %ecx # i, tmp92
13965 imull %edx, %ecx # i, tmp92
13966 # test.c:6: for (i = 0; i < n; i++)
13968 # test.c:7: total += i * i;
13969 addl %ecx, %eax # tmp92, <retval>
13977 .ident "GCC: (GNU) 7.0.0 20160809 (experimental)"
13978 .section .note.GNU-stack,"",@@progbits
13981 The comments are intended for humans rather than machines and hence the
13982 precise format of the comments is subject to change.
13984 @item -frecord-gcc-switches
13985 @opindex frecord-gcc-switches
13986 This switch causes the command line used to invoke the
13987 compiler to be recorded into the object file that is being created.
13988 This switch is only implemented on some targets and the exact format
13989 of the recording is target and binary file format dependent, but it
13990 usually takes the form of a section containing ASCII text. This
13991 switch is related to the @option{-fverbose-asm} switch, but that
13992 switch only records information in the assembler output file as
13993 comments, so it never reaches the object file.
13994 See also @option{-grecord-gcc-switches} for another
13995 way of storing compiler options into the object file.
13999 @cindex global offset table
14001 Generate position-independent code (PIC) suitable for use in a shared
14002 library, if supported for the target machine. Such code accesses all
14003 constant addresses through a global offset table (GOT)@. The dynamic
14004 loader resolves the GOT entries when the program starts (the dynamic
14005 loader is not part of GCC; it is part of the operating system). If
14006 the GOT size for the linked executable exceeds a machine-specific
14007 maximum size, you get an error message from the linker indicating that
14008 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
14009 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
14010 on the m68k and RS/6000. The x86 has no such limit.)
14012 Position-independent code requires special support, and therefore works
14013 only on certain machines. For the x86, GCC supports PIC for System V
14014 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
14015 position-independent.
14017 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
14022 If supported for the target machine, emit position-independent code,
14023 suitable for dynamic linking and avoiding any limit on the size of the
14024 global offset table. This option makes a difference on AArch64, m68k,
14025 PowerPC and SPARC@.
14027 Position-independent code requires special support, and therefore works
14028 only on certain machines.
14030 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
14037 These options are similar to @option{-fpic} and @option{-fPIC}, but the
14038 generated position-independent code can be only linked into executables.
14039 Usually these options are used to compile code that will be linked using
14040 the @option{-pie} GCC option.
14042 @option{-fpie} and @option{-fPIE} both define the macros
14043 @code{__pie__} and @code{__PIE__}. The macros have the value 1
14044 for @option{-fpie} and 2 for @option{-fPIE}.
14049 Do not use the PLT for external function calls in position-independent code.
14050 Instead, load the callee address at call sites from the GOT and branch to it.
14051 This leads to more efficient code by eliminating PLT stubs and exposing
14052 GOT loads to optimizations. On architectures such as 32-bit x86 where
14053 PLT stubs expect the GOT pointer in a specific register, this gives more
14054 register allocation freedom to the compiler.
14055 Lazy binding requires use of the PLT;
14056 with @option{-fno-plt} all external symbols are resolved at load time.
14058 Alternatively, the function attribute @code{noplt} can be used to avoid calls
14059 through the PLT for specific external functions.
14061 In position-dependent code, a few targets also convert calls to
14062 functions that are marked to not use the PLT to use the GOT instead.
14064 @item -fno-jump-tables
14065 @opindex fno-jump-tables
14066 @opindex fjump-tables
14067 Do not use jump tables for switch statements even where it would be
14068 more efficient than other code generation strategies. This option is
14069 of use in conjunction with @option{-fpic} or @option{-fPIC} for
14070 building code that forms part of a dynamic linker and cannot
14071 reference the address of a jump table. On some targets, jump tables
14072 do not require a GOT and this option is not needed.
14074 @item -ffixed-@var{reg}
14076 Treat the register named @var{reg} as a fixed register; generated code
14077 should never refer to it (except perhaps as a stack pointer, frame
14078 pointer or in some other fixed role).
14080 @var{reg} must be the name of a register. The register names accepted
14081 are machine-specific and are defined in the @code{REGISTER_NAMES}
14082 macro in the machine description macro file.
14084 This flag does not have a negative form, because it specifies a
14087 @item -fcall-used-@var{reg}
14088 @opindex fcall-used
14089 Treat the register named @var{reg} as an allocable register that is
14090 clobbered by function calls. It may be allocated for temporaries or
14091 variables that do not live across a call. Functions compiled this way
14092 do not save and restore the register @var{reg}.
14094 It is an error to use this flag with the frame pointer or stack pointer.
14095 Use of this flag for other registers that have fixed pervasive roles in
14096 the machine's execution model produces disastrous results.
14098 This flag does not have a negative form, because it specifies a
14101 @item -fcall-saved-@var{reg}
14102 @opindex fcall-saved
14103 Treat the register named @var{reg} as an allocable register saved by
14104 functions. It may be allocated even for temporaries or variables that
14105 live across a call. Functions compiled this way save and restore
14106 the register @var{reg} if they use it.
14108 It is an error to use this flag with the frame pointer or stack pointer.
14109 Use of this flag for other registers that have fixed pervasive roles in
14110 the machine's execution model produces disastrous results.
14112 A different sort of disaster results from the use of this flag for
14113 a register in which function values may be returned.
14115 This flag does not have a negative form, because it specifies a
14118 @item -fpack-struct[=@var{n}]
14119 @opindex fpack-struct
14120 Without a value specified, pack all structure members together without
14121 holes. When a value is specified (which must be a small power of two), pack
14122 structure members according to this value, representing the maximum
14123 alignment (that is, objects with default alignment requirements larger than
14124 this are output potentially unaligned at the next fitting location.
14126 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
14127 code that is not binary compatible with code generated without that switch.
14128 Additionally, it makes the code suboptimal.
14129 Use it to conform to a non-default application binary interface.
14131 @item -fleading-underscore
14132 @opindex fleading-underscore
14133 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
14134 change the way C symbols are represented in the object file. One use
14135 is to help link with legacy assembly code.
14137 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
14138 generate code that is not binary compatible with code generated without that
14139 switch. Use it to conform to a non-default application binary interface.
14140 Not all targets provide complete support for this switch.
14142 @item -ftls-model=@var{model}
14143 @opindex ftls-model
14144 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
14145 The @var{model} argument should be one of @samp{global-dynamic},
14146 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
14147 Note that the choice is subject to optimization: the compiler may use
14148 a more efficient model for symbols not visible outside of the translation
14149 unit, or if @option{-fpic} is not given on the command line.
14151 The default without @option{-fpic} is @samp{initial-exec}; with
14152 @option{-fpic} the default is @samp{global-dynamic}.
14154 @item -ftrampolines
14155 @opindex ftrampolines
14156 For targets that normally need trampolines for nested functions, always
14157 generate them instead of using descriptors. Otherwise, for targets that
14158 do not need them, like for example HP-PA or IA-64, do nothing.
14160 A trampoline is a small piece of code that is created at run time on the
14161 stack when the address of a nested function is taken, and is used to call
14162 the nested function indirectly. Therefore, it requires the stack to be
14163 made executable in order for the program to work properly.
14165 @option{-fno-trampolines} is enabled by default on a language by language
14166 basis to let the compiler avoid generating them, if it computes that this
14167 is safe, and replace them with descriptors. Descriptors are made up of data
14168 only, but the generated code must be prepared to deal with them. As of this
14169 writing, @option{-fno-trampolines} is enabled by default only for Ada.
14171 Moreover, code compiled with @option{-ftrampolines} and code compiled with
14172 @option{-fno-trampolines} are not binary compatible if nested functions are
14173 present. This option must therefore be used on a program-wide basis and be
14174 manipulated with extreme care.
14176 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
14177 @opindex fvisibility
14178 Set the default ELF image symbol visibility to the specified option---all
14179 symbols are marked with this unless overridden within the code.
14180 Using this feature can very substantially improve linking and
14181 load times of shared object libraries, produce more optimized
14182 code, provide near-perfect API export and prevent symbol clashes.
14183 It is @strong{strongly} recommended that you use this in any shared objects
14186 Despite the nomenclature, @samp{default} always means public; i.e.,
14187 available to be linked against from outside the shared object.
14188 @samp{protected} and @samp{internal} are pretty useless in real-world
14189 usage so the only other commonly used option is @samp{hidden}.
14190 The default if @option{-fvisibility} isn't specified is
14191 @samp{default}, i.e., make every symbol public.
14193 A good explanation of the benefits offered by ensuring ELF
14194 symbols have the correct visibility is given by ``How To Write
14195 Shared Libraries'' by Ulrich Drepper (which can be found at
14196 @w{@uref{https://www.akkadia.org/drepper/}})---however a superior
14197 solution made possible by this option to marking things hidden when
14198 the default is public is to make the default hidden and mark things
14199 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
14200 and @code{__attribute__ ((visibility("default")))} instead of
14201 @code{__declspec(dllexport)} you get almost identical semantics with
14202 identical syntax. This is a great boon to those working with
14203 cross-platform projects.
14205 For those adding visibility support to existing code, you may find
14206 @code{#pragma GCC visibility} of use. This works by you enclosing
14207 the declarations you wish to set visibility for with (for example)
14208 @code{#pragma GCC visibility push(hidden)} and
14209 @code{#pragma GCC visibility pop}.
14210 Bear in mind that symbol visibility should be viewed @strong{as
14211 part of the API interface contract} and thus all new code should
14212 always specify visibility when it is not the default; i.e., declarations
14213 only for use within the local DSO should @strong{always} be marked explicitly
14214 as hidden as so to avoid PLT indirection overheads---making this
14215 abundantly clear also aids readability and self-documentation of the code.
14216 Note that due to ISO C++ specification requirements, @code{operator new} and
14217 @code{operator delete} must always be of default visibility.
14219 Be aware that headers from outside your project, in particular system
14220 headers and headers from any other library you use, may not be
14221 expecting to be compiled with visibility other than the default. You
14222 may need to explicitly say @code{#pragma GCC visibility push(default)}
14223 before including any such headers.
14225 @code{extern} declarations are not affected by @option{-fvisibility}, so
14226 a lot of code can be recompiled with @option{-fvisibility=hidden} with
14227 no modifications. However, this means that calls to @code{extern}
14228 functions with no explicit visibility use the PLT, so it is more
14229 effective to use @code{__attribute ((visibility))} and/or
14230 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
14231 declarations should be treated as hidden.
14233 Note that @option{-fvisibility} does affect C++ vague linkage
14234 entities. This means that, for instance, an exception class that is
14235 be thrown between DSOs must be explicitly marked with default
14236 visibility so that the @samp{type_info} nodes are unified between
14239 An overview of these techniques, their benefits and how to use them
14240 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
14242 @item -fstrict-volatile-bitfields
14243 @opindex fstrict-volatile-bitfields
14244 This option should be used if accesses to volatile bit-fields (or other
14245 structure fields, although the compiler usually honors those types
14246 anyway) should use a single access of the width of the
14247 field's type, aligned to a natural alignment if possible. For
14248 example, targets with memory-mapped peripheral registers might require
14249 all such accesses to be 16 bits wide; with this flag you can
14250 declare all peripheral bit-fields as @code{unsigned short} (assuming short
14251 is 16 bits on these targets) to force GCC to use 16-bit accesses
14252 instead of, perhaps, a more efficient 32-bit access.
14254 If this option is disabled, the compiler uses the most efficient
14255 instruction. In the previous example, that might be a 32-bit load
14256 instruction, even though that accesses bytes that do not contain
14257 any portion of the bit-field, or memory-mapped registers unrelated to
14258 the one being updated.
14260 In some cases, such as when the @code{packed} attribute is applied to a
14261 structure field, it may not be possible to access the field with a single
14262 read or write that is correctly aligned for the target machine. In this
14263 case GCC falls back to generating multiple accesses rather than code that
14264 will fault or truncate the result at run time.
14266 Note: Due to restrictions of the C/C++11 memory model, write accesses are
14267 not allowed to touch non bit-field members. It is therefore recommended
14268 to define all bits of the field's type as bit-field members.
14270 The default value of this option is determined by the application binary
14271 interface for the target processor.
14273 @item -fsync-libcalls
14274 @opindex fsync-libcalls
14275 This option controls whether any out-of-line instance of the @code{__sync}
14276 family of functions may be used to implement the C++11 @code{__atomic}
14277 family of functions.
14279 The default value of this option is enabled, thus the only useful form
14280 of the option is @option{-fno-sync-libcalls}. This option is used in
14281 the implementation of the @file{libatomic} runtime library.
14285 @node Developer Options
14286 @section GCC Developer Options
14287 @cindex developer options
14288 @cindex debugging GCC
14289 @cindex debug dump options
14290 @cindex dump options
14291 @cindex compilation statistics
14293 This section describes command-line options that are primarily of
14294 interest to GCC developers, including options to support compiler
14295 testing and investigation of compiler bugs and compile-time
14296 performance problems. This includes options that produce debug dumps
14297 at various points in the compilation; that print statistics such as
14298 memory use and execution time; and that print information about GCC's
14299 configuration, such as where it searches for libraries. You should
14300 rarely need to use any of these options for ordinary compilation and
14303 Many developer options that cause GCC to dump output to a file take an
14304 optional @samp{=@var{filename}} suffix. You can specify @samp{stdout}
14305 or @samp{-} to dump to standard output, and @samp{stderr} for standard
14308 If @samp{=@var{filename}} is omitted, a default dump file name is
14309 constructed by concatenating the base dump file name, a pass number,
14310 phase letter, and pass name. The base dump file name is the name of
14311 output file produced by the compiler if explicitly specified and not
14312 an executable; otherwise it is the source file name.
14313 The pass number is determined by the order passes are registered with
14314 the compiler's pass manager.
14315 This is generally the same as the order of execution, but passes
14316 registered by plugins, target-specific passes, or passes that are
14317 otherwise registered late are numbered higher than the pass named
14318 @samp{final}, even if they are executed earlier. The phase letter is
14319 one of @samp{i} (inter-procedural analysis), @samp{l}
14320 (language-specific), @samp{r} (RTL), or @samp{t} (tree).
14321 The files are created in the directory of the output file.
14325 @item -d@var{letters}
14326 @itemx -fdump-rtl-@var{pass}
14327 @itemx -fdump-rtl-@var{pass}=@var{filename}
14329 @opindex fdump-rtl-@var{pass}
14330 Says to make debugging dumps during compilation at times specified by
14331 @var{letters}. This is used for debugging the RTL-based passes of the
14334 Some @option{-d@var{letters}} switches have different meaning when
14335 @option{-E} is used for preprocessing. @xref{Preprocessor Options},
14336 for information about preprocessor-specific dump options.
14338 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
14339 @option{-d} option @var{letters}. Here are the possible
14340 letters for use in @var{pass} and @var{letters}, and their meanings:
14344 @item -fdump-rtl-alignments
14345 @opindex fdump-rtl-alignments
14346 Dump after branch alignments have been computed.
14348 @item -fdump-rtl-asmcons
14349 @opindex fdump-rtl-asmcons
14350 Dump after fixing rtl statements that have unsatisfied in/out constraints.
14352 @item -fdump-rtl-auto_inc_dec
14353 @opindex fdump-rtl-auto_inc_dec
14354 Dump after auto-inc-dec discovery. This pass is only run on
14355 architectures that have auto inc or auto dec instructions.
14357 @item -fdump-rtl-barriers
14358 @opindex fdump-rtl-barriers
14359 Dump after cleaning up the barrier instructions.
14361 @item -fdump-rtl-bbpart
14362 @opindex fdump-rtl-bbpart
14363 Dump after partitioning hot and cold basic blocks.
14365 @item -fdump-rtl-bbro
14366 @opindex fdump-rtl-bbro
14367 Dump after block reordering.
14369 @item -fdump-rtl-btl1
14370 @itemx -fdump-rtl-btl2
14371 @opindex fdump-rtl-btl2
14372 @opindex fdump-rtl-btl2
14373 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
14374 after the two branch
14375 target load optimization passes.
14377 @item -fdump-rtl-bypass
14378 @opindex fdump-rtl-bypass
14379 Dump after jump bypassing and control flow optimizations.
14381 @item -fdump-rtl-combine
14382 @opindex fdump-rtl-combine
14383 Dump after the RTL instruction combination pass.
14385 @item -fdump-rtl-compgotos
14386 @opindex fdump-rtl-compgotos
14387 Dump after duplicating the computed gotos.
14389 @item -fdump-rtl-ce1
14390 @itemx -fdump-rtl-ce2
14391 @itemx -fdump-rtl-ce3
14392 @opindex fdump-rtl-ce1
14393 @opindex fdump-rtl-ce2
14394 @opindex fdump-rtl-ce3
14395 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
14396 @option{-fdump-rtl-ce3} enable dumping after the three
14397 if conversion passes.
14399 @item -fdump-rtl-cprop_hardreg
14400 @opindex fdump-rtl-cprop_hardreg
14401 Dump after hard register copy propagation.
14403 @item -fdump-rtl-csa
14404 @opindex fdump-rtl-csa
14405 Dump after combining stack adjustments.
14407 @item -fdump-rtl-cse1
14408 @itemx -fdump-rtl-cse2
14409 @opindex fdump-rtl-cse1
14410 @opindex fdump-rtl-cse2
14411 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
14412 the two common subexpression elimination passes.
14414 @item -fdump-rtl-dce
14415 @opindex fdump-rtl-dce
14416 Dump after the standalone dead code elimination passes.
14418 @item -fdump-rtl-dbr
14419 @opindex fdump-rtl-dbr
14420 Dump after delayed branch scheduling.
14422 @item -fdump-rtl-dce1
14423 @itemx -fdump-rtl-dce2
14424 @opindex fdump-rtl-dce1
14425 @opindex fdump-rtl-dce2
14426 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
14427 the two dead store elimination passes.
14429 @item -fdump-rtl-eh
14430 @opindex fdump-rtl-eh
14431 Dump after finalization of EH handling code.
14433 @item -fdump-rtl-eh_ranges
14434 @opindex fdump-rtl-eh_ranges
14435 Dump after conversion of EH handling range regions.
14437 @item -fdump-rtl-expand
14438 @opindex fdump-rtl-expand
14439 Dump after RTL generation.
14441 @item -fdump-rtl-fwprop1
14442 @itemx -fdump-rtl-fwprop2
14443 @opindex fdump-rtl-fwprop1
14444 @opindex fdump-rtl-fwprop2
14445 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
14446 dumping after the two forward propagation passes.
14448 @item -fdump-rtl-gcse1
14449 @itemx -fdump-rtl-gcse2
14450 @opindex fdump-rtl-gcse1
14451 @opindex fdump-rtl-gcse2
14452 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
14453 after global common subexpression elimination.
14455 @item -fdump-rtl-init-regs
14456 @opindex fdump-rtl-init-regs
14457 Dump after the initialization of the registers.
14459 @item -fdump-rtl-initvals
14460 @opindex fdump-rtl-initvals
14461 Dump after the computation of the initial value sets.
14463 @item -fdump-rtl-into_cfglayout
14464 @opindex fdump-rtl-into_cfglayout
14465 Dump after converting to cfglayout mode.
14467 @item -fdump-rtl-ira
14468 @opindex fdump-rtl-ira
14469 Dump after iterated register allocation.
14471 @item -fdump-rtl-jump
14472 @opindex fdump-rtl-jump
14473 Dump after the second jump optimization.
14475 @item -fdump-rtl-loop2
14476 @opindex fdump-rtl-loop2
14477 @option{-fdump-rtl-loop2} enables dumping after the rtl
14478 loop optimization passes.
14480 @item -fdump-rtl-mach
14481 @opindex fdump-rtl-mach
14482 Dump after performing the machine dependent reorganization pass, if that
14485 @item -fdump-rtl-mode_sw
14486 @opindex fdump-rtl-mode_sw
14487 Dump after removing redundant mode switches.
14489 @item -fdump-rtl-rnreg
14490 @opindex fdump-rtl-rnreg
14491 Dump after register renumbering.
14493 @item -fdump-rtl-outof_cfglayout
14494 @opindex fdump-rtl-outof_cfglayout
14495 Dump after converting from cfglayout mode.
14497 @item -fdump-rtl-peephole2
14498 @opindex fdump-rtl-peephole2
14499 Dump after the peephole pass.
14501 @item -fdump-rtl-postreload
14502 @opindex fdump-rtl-postreload
14503 Dump after post-reload optimizations.
14505 @item -fdump-rtl-pro_and_epilogue
14506 @opindex fdump-rtl-pro_and_epilogue
14507 Dump after generating the function prologues and epilogues.
14509 @item -fdump-rtl-sched1
14510 @itemx -fdump-rtl-sched2
14511 @opindex fdump-rtl-sched1
14512 @opindex fdump-rtl-sched2
14513 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
14514 after the basic block scheduling passes.
14516 @item -fdump-rtl-ree
14517 @opindex fdump-rtl-ree
14518 Dump after sign/zero extension elimination.
14520 @item -fdump-rtl-seqabstr
14521 @opindex fdump-rtl-seqabstr
14522 Dump after common sequence discovery.
14524 @item -fdump-rtl-shorten
14525 @opindex fdump-rtl-shorten
14526 Dump after shortening branches.
14528 @item -fdump-rtl-sibling
14529 @opindex fdump-rtl-sibling
14530 Dump after sibling call optimizations.
14532 @item -fdump-rtl-split1
14533 @itemx -fdump-rtl-split2
14534 @itemx -fdump-rtl-split3
14535 @itemx -fdump-rtl-split4
14536 @itemx -fdump-rtl-split5
14537 @opindex fdump-rtl-split1
14538 @opindex fdump-rtl-split2
14539 @opindex fdump-rtl-split3
14540 @opindex fdump-rtl-split4
14541 @opindex fdump-rtl-split5
14542 These options enable dumping after five rounds of
14543 instruction splitting.
14545 @item -fdump-rtl-sms
14546 @opindex fdump-rtl-sms
14547 Dump after modulo scheduling. This pass is only run on some
14550 @item -fdump-rtl-stack
14551 @opindex fdump-rtl-stack
14552 Dump after conversion from GCC's ``flat register file'' registers to the
14553 x87's stack-like registers. This pass is only run on x86 variants.
14555 @item -fdump-rtl-subreg1
14556 @itemx -fdump-rtl-subreg2
14557 @opindex fdump-rtl-subreg1
14558 @opindex fdump-rtl-subreg2
14559 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
14560 the two subreg expansion passes.
14562 @item -fdump-rtl-unshare
14563 @opindex fdump-rtl-unshare
14564 Dump after all rtl has been unshared.
14566 @item -fdump-rtl-vartrack
14567 @opindex fdump-rtl-vartrack
14568 Dump after variable tracking.
14570 @item -fdump-rtl-vregs
14571 @opindex fdump-rtl-vregs
14572 Dump after converting virtual registers to hard registers.
14574 @item -fdump-rtl-web
14575 @opindex fdump-rtl-web
14576 Dump after live range splitting.
14578 @item -fdump-rtl-regclass
14579 @itemx -fdump-rtl-subregs_of_mode_init
14580 @itemx -fdump-rtl-subregs_of_mode_finish
14581 @itemx -fdump-rtl-dfinit
14582 @itemx -fdump-rtl-dfinish
14583 @opindex fdump-rtl-regclass
14584 @opindex fdump-rtl-subregs_of_mode_init
14585 @opindex fdump-rtl-subregs_of_mode_finish
14586 @opindex fdump-rtl-dfinit
14587 @opindex fdump-rtl-dfinish
14588 These dumps are defined but always produce empty files.
14591 @itemx -fdump-rtl-all
14593 @opindex fdump-rtl-all
14594 Produce all the dumps listed above.
14598 Annotate the assembler output with miscellaneous debugging information.
14602 Dump all macro definitions, at the end of preprocessing, in addition to
14607 Produce a core dump whenever an error occurs.
14611 Annotate the assembler output with a comment indicating which
14612 pattern and alternative is used. The length and cost of each instruction are
14617 Dump the RTL in the assembler output as a comment before each instruction.
14618 Also turns on @option{-dp} annotation.
14622 Just generate RTL for a function instead of compiling it. Usually used
14623 with @option{-fdump-rtl-expand}.
14627 @opindex fdump-debug
14628 Dump debugging information generated during the debug
14631 @item -fdump-earlydebug
14632 @opindex fdump-earlydebug
14633 Dump debugging information generated during the early debug
14636 @item -fdump-noaddr
14637 @opindex fdump-noaddr
14638 When doing debugging dumps, suppress address output. This makes it more
14639 feasible to use diff on debugging dumps for compiler invocations with
14640 different compiler binaries and/or different
14641 text / bss / data / heap / stack / dso start locations.
14644 @opindex freport-bug
14645 Collect and dump debug information into a temporary file if an
14646 internal compiler error (ICE) occurs.
14648 @item -fdump-unnumbered
14649 @opindex fdump-unnumbered
14650 When doing debugging dumps, suppress instruction numbers and address output.
14651 This makes it more feasible to use diff on debugging dumps for compiler
14652 invocations with different options, in particular with and without
14655 @item -fdump-unnumbered-links
14656 @opindex fdump-unnumbered-links
14657 When doing debugging dumps (see @option{-d} option above), suppress
14658 instruction numbers for the links to the previous and next instructions
14661 @item -fdump-ipa-@var{switch}
14662 @itemx -fdump-ipa-@var{switch}-@var{options}
14664 Control the dumping at various stages of inter-procedural analysis
14665 language tree to a file. The file name is generated by appending a
14666 switch specific suffix to the source file name, and the file is created
14667 in the same directory as the output file. The following dumps are
14672 Enables all inter-procedural analysis dumps.
14675 Dumps information about call-graph optimization, unused function removal,
14676 and inlining decisions.
14679 Dump after function inlining.
14683 Additionally, the options @option{-optimized}, @option{-missed},
14684 @option{-note}, and @option{-all} can be provided, with the same meaning
14685 as for @option{-fopt-info}, defaulting to @option{-optimized}.
14687 For example, @option{-fdump-ipa-inline-optimized-missed} will emit
14688 information on callsites that were inlined, along with callsites
14689 that were not inlined.
14691 By default, the dump will contain messages about successful
14692 optimizations (equivalent to @option{-optimized}) together with
14693 low-level details about the analysis.
14695 @item -fdump-lang-all
14696 @itemx -fdump-lang-@var{switch}
14697 @itemx -fdump-lang-@var{switch}-@var{options}
14698 @itemx -fdump-lang-@var{switch}-@var{options}=@var{filename}
14699 @opindex fdump-lang-all
14700 @opindex fdump-lang
14701 Control the dumping of language-specific information. The @var{options}
14702 and @var{filename} portions behave as described in the
14703 @option{-fdump-tree} option. The following @var{switch} values are
14709 Enable all language-specific dumps.
14712 Dump class hierarchy information. Virtual table information is emitted
14713 unless '@option{slim}' is specified. This option is applicable to C++ only.
14716 Dump the raw internal tree data. This option is applicable to C++ only.
14720 @item -fdump-passes
14721 @opindex fdump-passes
14722 Print on @file{stderr} the list of optimization passes that are turned
14723 on and off by the current command-line options.
14725 @item -fdump-statistics-@var{option}
14726 @opindex fdump-statistics
14727 Enable and control dumping of pass statistics in a separate file. The
14728 file name is generated by appending a suffix ending in
14729 @samp{.statistics} to the source file name, and the file is created in
14730 the same directory as the output file. If the @samp{-@var{option}}
14731 form is used, @samp{-stats} causes counters to be summed over the
14732 whole compilation unit while @samp{-details} dumps every event as
14733 the passes generate them. The default with no option is to sum
14734 counters for each function compiled.
14736 @item -fdump-tree-all
14737 @itemx -fdump-tree-@var{switch}
14738 @itemx -fdump-tree-@var{switch}-@var{options}
14739 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
14740 @opindex fdump-tree-all
14741 @opindex fdump-tree
14742 Control the dumping at various stages of processing the intermediate
14743 language tree to a file. If the @samp{-@var{options}}
14744 form is used, @var{options} is a list of @samp{-} separated options
14745 which control the details of the dump. Not all options are applicable
14746 to all dumps; those that are not meaningful are ignored. The
14747 following options are available
14751 Print the address of each node. Usually this is not meaningful as it
14752 changes according to the environment and source file. Its primary use
14753 is for tying up a dump file with a debug environment.
14755 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
14756 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
14757 use working backward from mangled names in the assembly file.
14759 When dumping front-end intermediate representations, inhibit dumping
14760 of members of a scope or body of a function merely because that scope
14761 has been reached. Only dump such items when they are directly reachable
14762 by some other path.
14764 When dumping pretty-printed trees, this option inhibits dumping the
14765 bodies of control structures.
14767 When dumping RTL, print the RTL in slim (condensed) form instead of
14768 the default LISP-like representation.
14770 Print a raw representation of the tree. By default, trees are
14771 pretty-printed into a C-like representation.
14773 Enable more detailed dumps (not honored by every dump option). Also
14774 include information from the optimization passes.
14776 Enable dumping various statistics about the pass (not honored by every dump
14779 Enable showing basic block boundaries (disabled in raw dumps).
14781 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
14782 dump a representation of the control flow graph suitable for viewing with
14783 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
14784 the file is pretty-printed as a subgraph, so that GraphViz can render them
14785 all in a single plot.
14787 This option currently only works for RTL dumps, and the RTL is always
14788 dumped in slim form.
14790 Enable showing virtual operands for every statement.
14792 Enable showing line numbers for statements.
14794 Enable showing the unique ID (@code{DECL_UID}) for each variable.
14796 Enable showing the tree dump for each statement.
14798 Enable showing the EH region number holding each statement.
14800 Enable showing scalar evolution analysis details.
14802 Enable showing optimization information (only available in certain
14805 Enable showing missed optimization information (only available in certain
14808 Enable other detailed optimization information (only available in
14811 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
14812 and @option{lineno}.
14814 Turn on all optimization options, i.e., @option{optimized},
14815 @option{missed}, and @option{note}.
14818 To determine what tree dumps are available or find the dump for a pass
14819 of interest follow the steps below.
14823 Invoke GCC with @option{-fdump-passes} and in the @file{stderr} output
14824 look for a code that corresponds to the pass you are interested in.
14825 For example, the codes @code{tree-evrp}, @code{tree-vrp1}, and
14826 @code{tree-vrp2} correspond to the three Value Range Propagation passes.
14827 The number at the end distinguishes distinct invocations of the same pass.
14829 To enable the creation of the dump file, append the pass code to
14830 the @option{-fdump-} option prefix and invoke GCC with it. For example,
14831 to enable the dump from the Early Value Range Propagation pass, invoke
14832 GCC with the @option{-fdump-tree-evrp} option. Optionally, you may
14833 specify the name of the dump file. If you don't specify one, GCC
14834 creates as described below.
14836 Find the pass dump in a file whose name is composed of three components
14837 separated by a period: the name of the source file GCC was invoked to
14838 compile, a numeric suffix indicating the pass number followed by the
14839 letter @samp{t} for tree passes (and the letter @samp{r} for RTL passes),
14840 and finally the pass code. For example, the Early VRP pass dump might
14841 be in a file named @file{myfile.c.038t.evrp} in the current working
14842 directory. Note that the numeric codes are not stable and may change
14843 from one version of GCC to another.
14847 @itemx -fopt-info-@var{options}
14848 @itemx -fopt-info-@var{options}=@var{filename}
14850 Controls optimization dumps from various optimization passes. If the
14851 @samp{-@var{options}} form is used, @var{options} is a list of
14852 @samp{-} separated option keywords to select the dump details and
14855 The @var{options} can be divided into three groups:
14858 options describing what kinds of messages should be emitted,
14860 options describing the verbosity of the dump, and
14862 options describing which optimizations should be included.
14864 The options from each group can be freely mixed as they are
14865 non-overlapping. However, in case of any conflicts,
14866 the later options override the earlier options on the command
14869 The following options control which kinds of messages should be emitted:
14873 Print information when an optimization is successfully applied. It is
14874 up to a pass to decide which information is relevant. For example, the
14875 vectorizer passes print the source location of loops which are
14876 successfully vectorized.
14878 Print information about missed optimizations. Individual passes
14879 control which information to include in the output.
14881 Print verbose information about optimizations, such as certain
14882 transformations, more detailed messages about decisions etc.
14884 Print detailed optimization information. This includes
14885 @samp{optimized}, @samp{missed}, and @samp{note}.
14888 The following option controls the dump verbosity:
14892 By default, only ``high-level'' messages are emitted. This option enables
14893 additional, more detailed, messages, which are likely to only be of interest
14897 One or more of the following option keywords can be used to describe a
14898 group of optimizations:
14902 Enable dumps from all interprocedural optimizations.
14904 Enable dumps from all loop optimizations.
14906 Enable dumps from all inlining optimizations.
14908 Enable dumps from all OMP (Offloading and Multi Processing) optimizations.
14910 Enable dumps from all vectorization optimizations.
14912 Enable dumps from all optimizations. This is a superset of
14913 the optimization groups listed above.
14916 If @var{options} is
14917 omitted, it defaults to @samp{optimized-optall}, which means to dump messages
14918 about successful optimizations from all the passes, omitting messages
14919 that are treated as ``internals''.
14921 If the @var{filename} is provided, then the dumps from all the
14922 applicable optimizations are concatenated into the @var{filename}.
14923 Otherwise the dump is output onto @file{stderr}. Though multiple
14924 @option{-fopt-info} options are accepted, only one of them can include
14925 a @var{filename}. If other filenames are provided then all but the
14926 first such option are ignored.
14928 Note that the output @var{filename} is overwritten
14929 in case of multiple translation units. If a combined output from
14930 multiple translation units is desired, @file{stderr} should be used
14933 In the following example, the optimization info is output to
14942 gcc -O3 -fopt-info-missed=missed.all
14946 outputs missed optimization report from all the passes into
14947 @file{missed.all}, and this one:
14950 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
14954 prints information about missed optimization opportunities from
14955 vectorization passes on @file{stderr}.
14956 Note that @option{-fopt-info-vec-missed} is equivalent to
14957 @option{-fopt-info-missed-vec}. The order of the optimization group
14958 names and message types listed after @option{-fopt-info} does not matter.
14960 As another example,
14962 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
14966 outputs information about missed optimizations as well as
14967 optimized locations from all the inlining passes into
14973 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
14977 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
14978 in conflict since only one output file is allowed. In this case, only
14979 the first option takes effect and the subsequent options are
14980 ignored. Thus only @file{vec.miss} is produced which contains
14981 dumps from the vectorizer about missed opportunities.
14983 @item -fsave-optimization-record
14984 @opindex fsave-optimization-record
14985 Write a SRCFILE.opt-record.json.gz file detailing what optimizations
14986 were performed, for those optimizations that support @option{-fopt-info}.
14988 This option is experimental and the format of the data within the
14989 compressed JSON file is subject to change.
14991 It is roughly equivalent to a machine-readable version of
14992 @option{-fopt-info-all}, as a collection of messages with source file,
14993 line number and column number, with the following additional data for
14999 the execution count of the code being optimized, along with metadata about
15000 whether this was from actual profile data, or just an estimate, allowing
15001 consumers to prioritize messages by code hotness,
15004 the function name of the code being optimized, where applicable,
15007 the ``inlining chain'' for the code being optimized, so that when
15008 a function is inlined into several different places (which might
15009 themselves be inlined), the reader can distinguish between the copies,
15012 objects identifying those parts of the message that refer to expressions,
15013 statements or symbol-table nodes, which of these categories they are, and,
15014 when available, their source code location,
15017 the GCC pass that emitted the message, and
15020 the location in GCC's own code from which the message was emitted
15024 Additionally, some messages are logically nested within other
15025 messages, reflecting implementation details of the optimization
15028 @item -fsched-verbose=@var{n}
15029 @opindex fsched-verbose
15030 On targets that use instruction scheduling, this option controls the
15031 amount of debugging output the scheduler prints to the dump files.
15033 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
15034 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
15035 For @var{n} greater than one, it also output basic block probabilities,
15036 detailed ready list information and unit/insn info. For @var{n} greater
15037 than two, it includes RTL at abort point, control-flow and regions info.
15038 And for @var{n} over four, @option{-fsched-verbose} also includes
15043 @item -fenable-@var{kind}-@var{pass}
15044 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
15048 This is a set of options that are used to explicitly disable/enable
15049 optimization passes. These options are intended for use for debugging GCC.
15050 Compiler users should use regular options for enabling/disabling
15055 @item -fdisable-ipa-@var{pass}
15056 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
15057 statically invoked in the compiler multiple times, the pass name should be
15058 appended with a sequential number starting from 1.
15060 @item -fdisable-rtl-@var{pass}
15061 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
15062 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
15063 statically invoked in the compiler multiple times, the pass name should be
15064 appended with a sequential number starting from 1. @var{range-list} is a
15065 comma-separated list of function ranges or assembler names. Each range is a number
15066 pair separated by a colon. The range is inclusive in both ends. If the range
15067 is trivial, the number pair can be simplified as a single number. If the
15068 function's call graph node's @var{uid} falls within one of the specified ranges,
15069 the @var{pass} is disabled for that function. The @var{uid} is shown in the
15070 function header of a dump file, and the pass names can be dumped by using
15071 option @option{-fdump-passes}.
15073 @item -fdisable-tree-@var{pass}
15074 @itemx -fdisable-tree-@var{pass}=@var{range-list}
15075 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
15078 @item -fenable-ipa-@var{pass}
15079 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
15080 statically invoked in the compiler multiple times, the pass name should be
15081 appended with a sequential number starting from 1.
15083 @item -fenable-rtl-@var{pass}
15084 @itemx -fenable-rtl-@var{pass}=@var{range-list}
15085 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
15086 description and examples.
15088 @item -fenable-tree-@var{pass}
15089 @itemx -fenable-tree-@var{pass}=@var{range-list}
15090 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
15091 of option arguments.
15095 Here are some examples showing uses of these options.
15099 # disable ccp1 for all functions
15100 -fdisable-tree-ccp1
15101 # disable complete unroll for function whose cgraph node uid is 1
15102 -fenable-tree-cunroll=1
15103 # disable gcse2 for functions at the following ranges [1,1],
15104 # [300,400], and [400,1000]
15105 # disable gcse2 for functions foo and foo2
15106 -fdisable-rtl-gcse2=foo,foo2
15107 # disable early inlining
15108 -fdisable-tree-einline
15109 # disable ipa inlining
15110 -fdisable-ipa-inline
15111 # enable tree full unroll
15112 -fenable-tree-unroll
15117 @itemx -fchecking=@var{n}
15119 @opindex fno-checking
15120 Enable internal consistency checking. The default depends on
15121 the compiler configuration. @option{-fchecking=2} enables further
15122 internal consistency checking that might affect code generation.
15124 @item -frandom-seed=@var{string}
15125 @opindex frandom-seed
15126 This option provides a seed that GCC uses in place of
15127 random numbers in generating certain symbol names
15128 that have to be different in every compiled file. It is also used to
15129 place unique stamps in coverage data files and the object files that
15130 produce them. You can use the @option{-frandom-seed} option to produce
15131 reproducibly identical object files.
15133 The @var{string} can either be a number (decimal, octal or hex) or an
15134 arbitrary string (in which case it's converted to a number by
15137 The @var{string} should be different for every file you compile.
15140 @itemx -save-temps=cwd
15141 @opindex save-temps
15142 Store the usual ``temporary'' intermediate files permanently; place them
15143 in the current directory and name them based on the source file. Thus,
15144 compiling @file{foo.c} with @option{-c -save-temps} produces files
15145 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
15146 preprocessed @file{foo.i} output file even though the compiler now
15147 normally uses an integrated preprocessor.
15149 When used in combination with the @option{-x} command-line option,
15150 @option{-save-temps} is sensible enough to avoid over writing an
15151 input source file with the same extension as an intermediate file.
15152 The corresponding intermediate file may be obtained by renaming the
15153 source file before using @option{-save-temps}.
15155 If you invoke GCC in parallel, compiling several different source
15156 files that share a common base name in different subdirectories or the
15157 same source file compiled for multiple output destinations, it is
15158 likely that the different parallel compilers will interfere with each
15159 other, and overwrite the temporary files. For instance:
15162 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
15163 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
15166 may result in @file{foo.i} and @file{foo.o} being written to
15167 simultaneously by both compilers.
15169 @item -save-temps=obj
15170 @opindex save-temps=obj
15171 Store the usual ``temporary'' intermediate files permanently. If the
15172 @option{-o} option is used, the temporary files are based on the
15173 object file. If the @option{-o} option is not used, the
15174 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
15179 gcc -save-temps=obj -c foo.c
15180 gcc -save-temps=obj -c bar.c -o dir/xbar.o
15181 gcc -save-temps=obj foobar.c -o dir2/yfoobar
15185 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
15186 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
15187 @file{dir2/yfoobar.o}.
15189 @item -time@r{[}=@var{file}@r{]}
15191 Report the CPU time taken by each subprocess in the compilation
15192 sequence. For C source files, this is the compiler proper and assembler
15193 (plus the linker if linking is done).
15195 Without the specification of an output file, the output looks like this:
15202 The first number on each line is the ``user time'', that is time spent
15203 executing the program itself. The second number is ``system time'',
15204 time spent executing operating system routines on behalf of the program.
15205 Both numbers are in seconds.
15207 With the specification of an output file, the output is appended to the
15208 named file, and it looks like this:
15211 0.12 0.01 cc1 @var{options}
15212 0.00 0.01 as @var{options}
15215 The ``user time'' and the ``system time'' are moved before the program
15216 name, and the options passed to the program are displayed, so that one
15217 can later tell what file was being compiled, and with which options.
15219 @item -fdump-final-insns@r{[}=@var{file}@r{]}
15220 @opindex fdump-final-insns
15221 Dump the final internal representation (RTL) to @var{file}. If the
15222 optional argument is omitted (or if @var{file} is @code{.}), the name
15223 of the dump file is determined by appending @code{.gkd} to the
15224 compilation output file name.
15226 @item -fcompare-debug@r{[}=@var{opts}@r{]}
15227 @opindex fcompare-debug
15228 @opindex fno-compare-debug
15229 If no error occurs during compilation, run the compiler a second time,
15230 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
15231 passed to the second compilation. Dump the final internal
15232 representation in both compilations, and print an error if they differ.
15234 If the equal sign is omitted, the default @option{-gtoggle} is used.
15236 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
15237 and nonzero, implicitly enables @option{-fcompare-debug}. If
15238 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
15239 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
15242 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
15243 is equivalent to @option{-fno-compare-debug}, which disables the dumping
15244 of the final representation and the second compilation, preventing even
15245 @env{GCC_COMPARE_DEBUG} from taking effect.
15247 To verify full coverage during @option{-fcompare-debug} testing, set
15248 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
15249 which GCC rejects as an invalid option in any actual compilation
15250 (rather than preprocessing, assembly or linking). To get just a
15251 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
15252 not overridden} will do.
15254 @item -fcompare-debug-second
15255 @opindex fcompare-debug-second
15256 This option is implicitly passed to the compiler for the second
15257 compilation requested by @option{-fcompare-debug}, along with options to
15258 silence warnings, and omitting other options that would cause the compiler
15259 to produce output to files or to standard output as a side effect. Dump
15260 files and preserved temporary files are renamed so as to contain the
15261 @code{.gk} additional extension during the second compilation, to avoid
15262 overwriting those generated by the first.
15264 When this option is passed to the compiler driver, it causes the
15265 @emph{first} compilation to be skipped, which makes it useful for little
15266 other than debugging the compiler proper.
15270 Turn off generation of debug info, if leaving out this option
15271 generates it, or turn it on at level 2 otherwise. The position of this
15272 argument in the command line does not matter; it takes effect after all
15273 other options are processed, and it does so only once, no matter how
15274 many times it is given. This is mainly intended to be used with
15275 @option{-fcompare-debug}.
15277 @item -fvar-tracking-assignments-toggle
15278 @opindex fvar-tracking-assignments-toggle
15279 @opindex fno-var-tracking-assignments-toggle
15280 Toggle @option{-fvar-tracking-assignments}, in the same way that
15281 @option{-gtoggle} toggles @option{-g}.
15285 Makes the compiler print out each function name as it is compiled, and
15286 print some statistics about each pass when it finishes.
15288 @item -ftime-report
15289 @opindex ftime-report
15290 Makes the compiler print some statistics about the time consumed by each
15291 pass when it finishes.
15293 @item -ftime-report-details
15294 @opindex ftime-report-details
15295 Record the time consumed by infrastructure parts separately for each pass.
15297 @item -fira-verbose=@var{n}
15298 @opindex fira-verbose
15299 Control the verbosity of the dump file for the integrated register allocator.
15300 The default value is 5. If the value @var{n} is greater or equal to 10,
15301 the dump output is sent to stderr using the same format as @var{n} minus 10.
15304 @opindex flto-report
15305 Prints a report with internal details on the workings of the link-time
15306 optimizer. The contents of this report vary from version to version.
15307 It is meant to be useful to GCC developers when processing object
15308 files in LTO mode (via @option{-flto}).
15310 Disabled by default.
15312 @item -flto-report-wpa
15313 @opindex flto-report-wpa
15314 Like @option{-flto-report}, but only print for the WPA phase of Link
15318 @opindex fmem-report
15319 Makes the compiler print some statistics about permanent memory
15320 allocation when it finishes.
15322 @item -fmem-report-wpa
15323 @opindex fmem-report-wpa
15324 Makes the compiler print some statistics about permanent memory
15325 allocation for the WPA phase only.
15327 @item -fpre-ipa-mem-report
15328 @opindex fpre-ipa-mem-report
15329 @item -fpost-ipa-mem-report
15330 @opindex fpost-ipa-mem-report
15331 Makes the compiler print some statistics about permanent memory
15332 allocation before or after interprocedural optimization.
15334 @item -fprofile-report
15335 @opindex fprofile-report
15336 Makes the compiler print some statistics about consistency of the
15337 (estimated) profile and effect of individual passes.
15339 @item -fstack-usage
15340 @opindex fstack-usage
15341 Makes the compiler output stack usage information for the program, on a
15342 per-function basis. The filename for the dump is made by appending
15343 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
15344 the output file, if explicitly specified and it is not an executable,
15345 otherwise it is the basename of the source file. An entry is made up
15350 The name of the function.
15354 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
15357 The qualifier @code{static} means that the function manipulates the stack
15358 statically: a fixed number of bytes are allocated for the frame on function
15359 entry and released on function exit; no stack adjustments are otherwise made
15360 in the function. The second field is this fixed number of bytes.
15362 The qualifier @code{dynamic} means that the function manipulates the stack
15363 dynamically: in addition to the static allocation described above, stack
15364 adjustments are made in the body of the function, for example to push/pop
15365 arguments around function calls. If the qualifier @code{bounded} is also
15366 present, the amount of these adjustments is bounded at compile time and
15367 the second field is an upper bound of the total amount of stack used by
15368 the function. If it is not present, the amount of these adjustments is
15369 not bounded at compile time and the second field only represents the
15374 Emit statistics about front-end processing at the end of the compilation.
15375 This option is supported only by the C++ front end, and
15376 the information is generally only useful to the G++ development team.
15378 @item -fdbg-cnt-list
15379 @opindex fdbg-cnt-list
15380 Print the name and the counter upper bound for all debug counters.
15383 @item -fdbg-cnt=@var{counter-value-list}
15385 Set the internal debug counter lower and upper bound. @var{counter-value-list}
15386 is a comma-separated list of @var{name}:@var{lower_bound}:@var{upper_bound}
15387 tuples which sets the lower and the upper bound of each debug
15388 counter @var{name}. The @var{lower_bound} is optional and is zero
15389 initialized if not set.
15390 All debug counters have the initial upper bound of @code{UINT_MAX};
15391 thus @code{dbg_cnt} returns true always unless the upper bound
15392 is set by this option.
15393 For example, with @option{-fdbg-cnt=dce:2:4,tail_call:10},
15394 @code{dbg_cnt(dce)} returns true only for third and fourth invocation.
15395 For @code{dbg_cnt(tail_call)} true is returned for first 10 invocations.
15397 @item -print-file-name=@var{library}
15398 @opindex print-file-name
15399 Print the full absolute name of the library file @var{library} that
15400 would be used when linking---and don't do anything else. With this
15401 option, GCC does not compile or link anything; it just prints the
15404 @item -print-multi-directory
15405 @opindex print-multi-directory
15406 Print the directory name corresponding to the multilib selected by any
15407 other switches present in the command line. This directory is supposed
15408 to exist in @env{GCC_EXEC_PREFIX}.
15410 @item -print-multi-lib
15411 @opindex print-multi-lib
15412 Print the mapping from multilib directory names to compiler switches
15413 that enable them. The directory name is separated from the switches by
15414 @samp{;}, and each switch starts with an @samp{@@} instead of the
15415 @samp{-}, without spaces between multiple switches. This is supposed to
15416 ease shell processing.
15418 @item -print-multi-os-directory
15419 @opindex print-multi-os-directory
15420 Print the path to OS libraries for the selected
15421 multilib, relative to some @file{lib} subdirectory. If OS libraries are
15422 present in the @file{lib} subdirectory and no multilibs are used, this is
15423 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
15424 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
15425 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
15426 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
15428 @item -print-multiarch
15429 @opindex print-multiarch
15430 Print the path to OS libraries for the selected multiarch,
15431 relative to some @file{lib} subdirectory.
15433 @item -print-prog-name=@var{program}
15434 @opindex print-prog-name
15435 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
15437 @item -print-libgcc-file-name
15438 @opindex print-libgcc-file-name
15439 Same as @option{-print-file-name=libgcc.a}.
15441 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
15442 but you do want to link with @file{libgcc.a}. You can do:
15445 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
15448 @item -print-search-dirs
15449 @opindex print-search-dirs
15450 Print the name of the configured installation directory and a list of
15451 program and library directories @command{gcc} searches---and don't do anything else.
15453 This is useful when @command{gcc} prints the error message
15454 @samp{installation problem, cannot exec cpp0: No such file or directory}.
15455 To resolve this you either need to put @file{cpp0} and the other compiler
15456 components where @command{gcc} expects to find them, or you can set the environment
15457 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
15458 Don't forget the trailing @samp{/}.
15459 @xref{Environment Variables}.
15461 @item -print-sysroot
15462 @opindex print-sysroot
15463 Print the target sysroot directory that is used during
15464 compilation. This is the target sysroot specified either at configure
15465 time or using the @option{--sysroot} option, possibly with an extra
15466 suffix that depends on compilation options. If no target sysroot is
15467 specified, the option prints nothing.
15469 @item -print-sysroot-headers-suffix
15470 @opindex print-sysroot-headers-suffix
15471 Print the suffix added to the target sysroot when searching for
15472 headers, or give an error if the compiler is not configured with such
15473 a suffix---and don't do anything else.
15476 @opindex dumpmachine
15477 Print the compiler's target machine (for example,
15478 @samp{i686-pc-linux-gnu})---and don't do anything else.
15481 @opindex dumpversion
15482 Print the compiler version (for example, @code{3.0}, @code{6.3.0} or @code{7})---and don't do
15483 anything else. This is the compiler version used in filesystem paths and
15484 specs. Depending on how the compiler has been configured it can be just
15485 a single number (major version), two numbers separated by a dot (major and
15486 minor version) or three numbers separated by dots (major, minor and patchlevel
15489 @item -dumpfullversion
15490 @opindex dumpfullversion
15491 Print the full compiler version---and don't do anything else. The output is
15492 always three numbers separated by dots, major, minor and patchlevel version.
15496 Print the compiler's built-in specs---and don't do anything else. (This
15497 is used when GCC itself is being built.) @xref{Spec Files}.
15500 @node Submodel Options
15501 @section Machine-Dependent Options
15502 @cindex submodel options
15503 @cindex specifying hardware config
15504 @cindex hardware models and configurations, specifying
15505 @cindex target-dependent options
15506 @cindex machine-dependent options
15508 Each target machine supported by GCC can have its own options---for
15509 example, to allow you to compile for a particular processor variant or
15510 ABI, or to control optimizations specific to that machine. By
15511 convention, the names of machine-specific options start with
15514 Some configurations of the compiler also support additional target-specific
15515 options, usually for compatibility with other compilers on the same
15518 @c This list is ordered alphanumerically by subsection name.
15519 @c It should be the same order and spelling as these options are listed
15520 @c in Machine Dependent Options
15523 * AArch64 Options::
15524 * Adapteva Epiphany Options::
15525 * AMD GCN Options::
15529 * Blackfin Options::
15535 * DEC Alpha Options::
15539 * GNU/Linux Options::
15549 * MicroBlaze Options::
15552 * MN10300 Options::
15556 * Nios II Options::
15557 * Nvidia PTX Options::
15558 * OpenRISC Options::
15560 * picoChip Options::
15561 * PowerPC Options::
15564 * RS/6000 and PowerPC Options::
15566 * S/390 and zSeries Options::
15569 * Solaris 2 Options::
15572 * System V Options::
15573 * TILE-Gx Options::
15574 * TILEPro Options::
15579 * VxWorks Options::
15581 * x86 Windows Options::
15582 * Xstormy16 Options::
15584 * zSeries Options::
15587 @node AArch64 Options
15588 @subsection AArch64 Options
15589 @cindex AArch64 Options
15591 These options are defined for AArch64 implementations:
15595 @item -mabi=@var{name}
15597 Generate code for the specified data model. Permissible values
15598 are @samp{ilp32} for SysV-like data model where int, long int and pointers
15599 are 32 bits, and @samp{lp64} for SysV-like data model where int is 32 bits,
15600 but long int and pointers are 64 bits.
15602 The default depends on the specific target configuration. Note that
15603 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
15604 entire program with the same ABI, and link with a compatible set of libraries.
15607 @opindex mbig-endian
15608 Generate big-endian code. This is the default when GCC is configured for an
15609 @samp{aarch64_be-*-*} target.
15611 @item -mgeneral-regs-only
15612 @opindex mgeneral-regs-only
15613 Generate code which uses only the general-purpose registers. This will prevent
15614 the compiler from using floating-point and Advanced SIMD registers but will not
15615 impose any restrictions on the assembler.
15617 @item -mlittle-endian
15618 @opindex mlittle-endian
15619 Generate little-endian code. This is the default when GCC is configured for an
15620 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
15622 @item -mcmodel=tiny
15623 @opindex mcmodel=tiny
15624 Generate code for the tiny code model. The program and its statically defined
15625 symbols must be within 1MB of each other. Programs can be statically or
15626 dynamically linked.
15628 @item -mcmodel=small
15629 @opindex mcmodel=small
15630 Generate code for the small code model. The program and its statically defined
15631 symbols must be within 4GB of each other. Programs can be statically or
15632 dynamically linked. This is the default code model.
15634 @item -mcmodel=large
15635 @opindex mcmodel=large
15636 Generate code for the large code model. This makes no assumptions about
15637 addresses and sizes of sections. Programs can be statically linked only.
15639 @item -mstrict-align
15640 @itemx -mno-strict-align
15641 @opindex mstrict-align
15642 @opindex mno-strict-align
15643 Avoid or allow generating memory accesses that may not be aligned on a natural
15644 object boundary as described in the architecture specification.
15646 @item -momit-leaf-frame-pointer
15647 @itemx -mno-omit-leaf-frame-pointer
15648 @opindex momit-leaf-frame-pointer
15649 @opindex mno-omit-leaf-frame-pointer
15650 Omit or keep the frame pointer in leaf functions. The former behavior is the
15653 @item -mstack-protector-guard=@var{guard}
15654 @itemx -mstack-protector-guard-reg=@var{reg}
15655 @itemx -mstack-protector-guard-offset=@var{offset}
15656 @opindex mstack-protector-guard
15657 @opindex mstack-protector-guard-reg
15658 @opindex mstack-protector-guard-offset
15659 Generate stack protection code using canary at @var{guard}. Supported
15660 locations are @samp{global} for a global canary or @samp{sysreg} for a
15661 canary in an appropriate system register.
15663 With the latter choice the options
15664 @option{-mstack-protector-guard-reg=@var{reg}} and
15665 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
15666 which system register to use as base register for reading the canary,
15667 and from what offset from that base register. There is no default
15668 register or offset as this is entirely for use within the Linux
15671 @item -mstack-protector-guard=@var{guard}
15672 @itemx -mstack-protector-guard-reg=@var{reg}
15673 @itemx -mstack-protector-guard-offset=@var{offset}
15674 @opindex mstack-protector-guard
15675 @opindex mstack-protector-guard-reg
15676 @opindex mstack-protector-guard-offset
15677 Generate stack protection code using canary at @var{guard}. Supported
15678 locations are @samp{global} for a global canary or @samp{sysreg} for a
15679 canary in an appropriate system register.
15681 With the latter choice the options
15682 @option{-mstack-protector-guard-reg=@var{reg}} and
15683 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
15684 which system register to use as base register for reading the canary,
15685 and from what offset from that base register. There is no default
15686 register or offset as this is entirely for use within the Linux
15689 @item -mtls-dialect=desc
15690 @opindex mtls-dialect=desc
15691 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
15692 of TLS variables. This is the default.
15694 @item -mtls-dialect=traditional
15695 @opindex mtls-dialect=traditional
15696 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
15699 @item -mtls-size=@var{size}
15701 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
15702 This option requires binutils 2.26 or newer.
15704 @item -mfix-cortex-a53-835769
15705 @itemx -mno-fix-cortex-a53-835769
15706 @opindex mfix-cortex-a53-835769
15707 @opindex mno-fix-cortex-a53-835769
15708 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
15709 This involves inserting a NOP instruction between memory instructions and
15710 64-bit integer multiply-accumulate instructions.
15712 @item -mfix-cortex-a53-843419
15713 @itemx -mno-fix-cortex-a53-843419
15714 @opindex mfix-cortex-a53-843419
15715 @opindex mno-fix-cortex-a53-843419
15716 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
15717 This erratum workaround is made at link time and this will only pass the
15718 corresponding flag to the linker.
15720 @item -mlow-precision-recip-sqrt
15721 @itemx -mno-low-precision-recip-sqrt
15722 @opindex mlow-precision-recip-sqrt
15723 @opindex mno-low-precision-recip-sqrt
15724 Enable or disable the reciprocal square root approximation.
15725 This option only has an effect if @option{-ffast-math} or
15726 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
15727 precision of reciprocal square root results to about 16 bits for
15728 single precision and to 32 bits for double precision.
15730 @item -mlow-precision-sqrt
15731 @itemx -mno-low-precision-sqrt
15732 @opindex mlow-precision-sqrt
15733 @opindex mno-low-precision-sqrt
15734 Enable or disable the square root approximation.
15735 This option only has an effect if @option{-ffast-math} or
15736 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
15737 precision of square root results to about 16 bits for
15738 single precision and to 32 bits for double precision.
15739 If enabled, it implies @option{-mlow-precision-recip-sqrt}.
15741 @item -mlow-precision-div
15742 @itemx -mno-low-precision-div
15743 @opindex mlow-precision-div
15744 @opindex mno-low-precision-div
15745 Enable or disable the division approximation.
15746 This option only has an effect if @option{-ffast-math} or
15747 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
15748 precision of division results to about 16 bits for
15749 single precision and to 32 bits for double precision.
15751 @item -mtrack-speculation
15752 @itemx -mno-track-speculation
15753 Enable or disable generation of additional code to track speculative
15754 execution through conditional branches. The tracking state can then
15755 be used by the compiler when expanding calls to
15756 @code{__builtin_speculation_safe_copy} to permit a more efficient code
15757 sequence to be generated.
15759 @item -march=@var{name}
15761 Specify the name of the target architecture and, optionally, one or
15762 more feature modifiers. This option has the form
15763 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
15765 The permissible values for @var{arch} are @samp{armv8-a},
15766 @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a}, @samp{armv8.4-a},
15767 @samp{armv8.5-a} or @var{native}.
15769 The value @samp{armv8.5-a} implies @samp{armv8.4-a} and enables compiler
15770 support for the ARMv8.5-A architecture extensions.
15772 The value @samp{armv8.4-a} implies @samp{armv8.3-a} and enables compiler
15773 support for the ARMv8.4-A architecture extensions.
15775 The value @samp{armv8.3-a} implies @samp{armv8.2-a} and enables compiler
15776 support for the ARMv8.3-A architecture extensions.
15778 The value @samp{armv8.2-a} implies @samp{armv8.1-a} and enables compiler
15779 support for the ARMv8.2-A architecture extensions.
15781 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
15782 support for the ARMv8.1-A architecture extension. In particular, it
15783 enables the @samp{+crc}, @samp{+lse}, and @samp{+rdma} features.
15785 The value @samp{native} is available on native AArch64 GNU/Linux and
15786 causes the compiler to pick the architecture of the host system. This
15787 option has no effect if the compiler is unable to recognize the
15788 architecture of the host system,
15790 The permissible values for @var{feature} are listed in the sub-section
15791 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
15792 Feature Modifiers}. Where conflicting feature modifiers are
15793 specified, the right-most feature is used.
15795 GCC uses @var{name} to determine what kind of instructions it can emit
15796 when generating assembly code. If @option{-march} is specified
15797 without either of @option{-mtune} or @option{-mcpu} also being
15798 specified, the code is tuned to perform well across a range of target
15799 processors implementing the target architecture.
15801 @item -mtune=@var{name}
15803 Specify the name of the target processor for which GCC should tune the
15804 performance of the code. Permissible values for this option are:
15805 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
15806 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
15807 @samp{cortex-a76}, @samp{ares}, @samp{exynos-m1}, @samp{emag}, @samp{falkor},
15808 @samp{neoverse-e1},@samp{neoverse-n1},@samp{qdf24xx}, @samp{saphira},
15809 @samp{phecda}, @samp{xgene1}, @samp{vulcan}, @samp{octeontx},
15810 @samp{octeontx81}, @samp{octeontx83}, @samp{thunderx}, @samp{thunderxt88},
15811 @samp{thunderxt88p1}, @samp{thunderxt81}, @samp{tsv110},
15812 @samp{thunderxt83}, @samp{thunderx2t99},
15813 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15814 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53},
15815 @samp{cortex-a75.cortex-a55}, @samp{cortex-a76.cortex-a55}
15818 The values @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15819 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53},
15820 @samp{cortex-a75.cortex-a55}, @samp{cortex-a76.cortex-a55} specify that GCC
15821 should tune for a big.LITTLE system.
15823 Additionally on native AArch64 GNU/Linux systems the value
15824 @samp{native} tunes performance to the host system. This option has no effect
15825 if the compiler is unable to recognize the processor of the host system.
15827 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
15828 are specified, the code is tuned to perform well across a range
15829 of target processors.
15831 This option cannot be suffixed by feature modifiers.
15833 @item -mcpu=@var{name}
15835 Specify the name of the target processor, optionally suffixed by one
15836 or more feature modifiers. This option has the form
15837 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
15838 the permissible values for @var{cpu} are the same as those available
15839 for @option{-mtune}. The permissible values for @var{feature} are
15840 documented in the sub-section on
15841 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
15842 Feature Modifiers}. Where conflicting feature modifiers are
15843 specified, the right-most feature is used.
15845 GCC uses @var{name} to determine what kind of instructions it can emit when
15846 generating assembly code (as if by @option{-march}) and to determine
15847 the target processor for which to tune for performance (as if
15848 by @option{-mtune}). Where this option is used in conjunction
15849 with @option{-march} or @option{-mtune}, those options take precedence
15850 over the appropriate part of this option.
15852 @item -moverride=@var{string}
15854 Override tuning decisions made by the back-end in response to a
15855 @option{-mtune=} switch. The syntax, semantics, and accepted values
15856 for @var{string} in this option are not guaranteed to be consistent
15859 This option is only intended to be useful when developing GCC.
15861 @item -mverbose-cost-dump
15862 @opindex mverbose-cost-dump
15863 Enable verbose cost model dumping in the debug dump files. This option is
15864 provided for use in debugging the compiler.
15866 @item -mpc-relative-literal-loads
15867 @itemx -mno-pc-relative-literal-loads
15868 @opindex mpc-relative-literal-loads
15869 @opindex mno-pc-relative-literal-loads
15870 Enable or disable PC-relative literal loads. With this option literal pools are
15871 accessed using a single instruction and emitted after each function. This
15872 limits the maximum size of functions to 1MB. This is enabled by default for
15873 @option{-mcmodel=tiny}.
15875 @item -msign-return-address=@var{scope}
15876 @opindex msign-return-address
15877 Select the function scope on which return address signing will be applied.
15878 Permissible values are @samp{none}, which disables return address signing,
15879 @samp{non-leaf}, which enables pointer signing for functions which are not leaf
15880 functions, and @samp{all}, which enables pointer signing for all functions. The
15881 default value is @samp{none}. This option has been deprecated by
15882 -mbranch-protection.
15884 @item -mbranch-protection=@var{none}|@var{standard}|@var{pac-ret}[+@var{leaf}]|@var{bti}
15885 @opindex mbranch-protection
15886 Select the branch protection features to use.
15887 @samp{none} is the default and turns off all types of branch protection.
15888 @samp{standard} turns on all types of branch protection features. If a feature
15889 has additional tuning options, then @samp{standard} sets it to its standard
15891 @samp{pac-ret[+@var{leaf}]} turns on return address signing to its standard
15892 level: signing functions that save the return address to memory (non-leaf
15893 functions will practically always do this) using the a-key. The optional
15894 argument @samp{leaf} can be used to extend the signing to include leaf
15896 @samp{bti} turns on branch target identification mechanism.
15898 @item -msve-vector-bits=@var{bits}
15899 @opindex msve-vector-bits
15900 Specify the number of bits in an SVE vector register. This option only has
15901 an effect when SVE is enabled.
15903 GCC supports two forms of SVE code generation: ``vector-length
15904 agnostic'' output that works with any size of vector register and
15905 ``vector-length specific'' output that allows GCC to make assumptions
15906 about the vector length when it is useful for optimization reasons.
15907 The possible values of @samp{bits} are: @samp{scalable}, @samp{128},
15908 @samp{256}, @samp{512}, @samp{1024} and @samp{2048}.
15909 Specifying @samp{scalable} selects vector-length agnostic
15910 output. At present @samp{-msve-vector-bits=128} also generates vector-length
15911 agnostic output. All other values generate vector-length specific code.
15912 The behavior of these values may change in future releases and no value except
15913 @samp{scalable} should be relied on for producing code that is portable across
15914 different hardware SVE vector lengths.
15916 The default is @samp{-msve-vector-bits=scalable}, which produces
15917 vector-length agnostic code.
15920 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
15921 @anchor{aarch64-feature-modifiers}
15922 @cindex @option{-march} feature modifiers
15923 @cindex @option{-mcpu} feature modifiers
15924 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
15925 the following and their inverses @option{no@var{feature}}:
15929 Enable CRC extension. This is on by default for
15930 @option{-march=armv8.1-a}.
15932 Enable Crypto extension. This also enables Advanced SIMD and floating-point
15935 Enable floating-point instructions. This is on by default for all possible
15936 values for options @option{-march} and @option{-mcpu}.
15938 Enable Advanced SIMD instructions. This also enables floating-point
15939 instructions. This is on by default for all possible values for options
15940 @option{-march} and @option{-mcpu}.
15942 Enable Scalable Vector Extension instructions. This also enables Advanced
15943 SIMD and floating-point instructions.
15945 Enable Large System Extension instructions. This is on by default for
15946 @option{-march=armv8.1-a}.
15948 Enable Round Double Multiply Accumulate instructions. This is on by default
15949 for @option{-march=armv8.1-a}.
15951 Enable FP16 extension. This also enables floating-point instructions.
15953 Enable FP16 fmla extension. This also enables FP16 extensions and
15954 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.
15957 Enable the RcPc extension. This does not change code generation from GCC,
15958 but is passed on to the assembler, enabling inline asm statements to use
15959 instructions from the RcPc extension.
15961 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
15963 Enable the Armv8-a aes and pmull crypto extension. This also enables Advanced
15966 Enable the Armv8-a sha2 crypto extension. This also enables Advanced SIMD instructions.
15968 Enable the sha512 and sha3 crypto extension. This also enables Advanced SIMD
15969 instructions. Use of this option with architectures prior to Armv8.2-A is not supported.
15971 Enable the sm3 and sm4 crypto extension. This also enables Advanced SIMD instructions.
15972 Use of this option with architectures prior to Armv8.2-A is not supported.
15974 Enable the Statistical Profiling extension. This option is only to enable the
15975 extension at the assembler level and does not affect code generation.
15977 Enable the Armv8.5-a Random Number instructions. This option is only to
15978 enable the extension at the assembler level and does not affect code
15981 Enable the Armv8.5-a Memory Tagging Extensions. This option is only to
15982 enable the extension at the assembler level and does not affect code
15985 Enable the Armv8-a Speculation Barrier instruction. This option is only to
15986 enable the extension at the assembler level and does not affect code
15987 generation. This option is enabled by default for @option{-march=armv8.5-a}.
15989 Enable the Armv8-a Speculative Store Bypass Safe instruction. This option
15990 is only to enable the extension at the assembler level and does not affect code
15991 generation. This option is enabled by default for @option{-march=armv8.5-a}.
15993 Enable the Armv8-a Execution and Data Prediction Restriction instructions.
15994 This option is only to enable the extension at the assembler level and does
15995 not affect code generation. This option is enabled by default for
15996 @option{-march=armv8.5-a}.
16000 Feature @option{crypto} implies @option{aes}, @option{sha2}, and @option{simd},
16001 which implies @option{fp}.
16002 Conversely, @option{nofp} implies @option{nosimd}, which implies
16003 @option{nocrypto}, @option{noaes} and @option{nosha2}.
16005 @node Adapteva Epiphany Options
16006 @subsection Adapteva Epiphany Options
16008 These @samp{-m} options are defined for Adapteva Epiphany:
16011 @item -mhalf-reg-file
16012 @opindex mhalf-reg-file
16013 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
16014 That allows code to run on hardware variants that lack these registers.
16016 @item -mprefer-short-insn-regs
16017 @opindex mprefer-short-insn-regs
16018 Preferentially allocate registers that allow short instruction generation.
16019 This can result in increased instruction count, so this may either reduce or
16020 increase overall code size.
16022 @item -mbranch-cost=@var{num}
16023 @opindex mbranch-cost
16024 Set the cost of branches to roughly @var{num} ``simple'' instructions.
16025 This cost is only a heuristic and is not guaranteed to produce
16026 consistent results across releases.
16030 Enable the generation of conditional moves.
16032 @item -mnops=@var{num}
16034 Emit @var{num} NOPs before every other generated instruction.
16036 @item -mno-soft-cmpsf
16037 @opindex mno-soft-cmpsf
16038 @opindex msoft-cmpsf
16039 For single-precision floating-point comparisons, emit an @code{fsub} instruction
16040 and test the flags. This is faster than a software comparison, but can
16041 get incorrect results in the presence of NaNs, or when two different small
16042 numbers are compared such that their difference is calculated as zero.
16043 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
16044 software comparisons.
16046 @item -mstack-offset=@var{num}
16047 @opindex mstack-offset
16048 Set the offset between the top of the stack and the stack pointer.
16049 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
16050 can be used by leaf functions without stack allocation.
16051 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
16052 Note also that this option changes the ABI; compiling a program with a
16053 different stack offset than the libraries have been compiled with
16054 generally does not work.
16055 This option can be useful if you want to evaluate if a different stack
16056 offset would give you better code, but to actually use a different stack
16057 offset to build working programs, it is recommended to configure the
16058 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
16060 @item -mno-round-nearest
16061 @opindex mno-round-nearest
16062 @opindex mround-nearest
16063 Make the scheduler assume that the rounding mode has been set to
16064 truncating. The default is @option{-mround-nearest}.
16067 @opindex mlong-calls
16068 If not otherwise specified by an attribute, assume all calls might be beyond
16069 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
16070 function address into a register before performing a (otherwise direct) call.
16071 This is the default.
16073 @item -mshort-calls
16074 @opindex short-calls
16075 If not otherwise specified by an attribute, assume all direct calls are
16076 in the range of the @code{b} / @code{bl} instructions, so use these instructions
16077 for direct calls. The default is @option{-mlong-calls}.
16081 Assume addresses can be loaded as 16-bit unsigned values. This does not
16082 apply to function addresses for which @option{-mlong-calls} semantics
16085 @item -mfp-mode=@var{mode}
16087 Set the prevailing mode of the floating-point unit.
16088 This determines the floating-point mode that is provided and expected
16089 at function call and return time. Making this mode match the mode you
16090 predominantly need at function start can make your programs smaller and
16091 faster by avoiding unnecessary mode switches.
16093 @var{mode} can be set to one the following values:
16097 Any mode at function entry is valid, and retained or restored when
16098 the function returns, and when it calls other functions.
16099 This mode is useful for compiling libraries or other compilation units
16100 you might want to incorporate into different programs with different
16101 prevailing FPU modes, and the convenience of being able to use a single
16102 object file outweighs the size and speed overhead for any extra
16103 mode switching that might be needed, compared with what would be needed
16104 with a more specific choice of prevailing FPU mode.
16107 This is the mode used for floating-point calculations with
16108 truncating (i.e.@: round towards zero) rounding mode. That includes
16109 conversion from floating point to integer.
16111 @item round-nearest
16112 This is the mode used for floating-point calculations with
16113 round-to-nearest-or-even rounding mode.
16116 This is the mode used to perform integer calculations in the FPU, e.g.@:
16117 integer multiply, or integer multiply-and-accumulate.
16120 The default is @option{-mfp-mode=caller}
16122 @item -mno-split-lohi
16123 @itemx -mno-postinc
16124 @itemx -mno-postmodify
16125 @opindex mno-split-lohi
16126 @opindex msplit-lohi
16127 @opindex mno-postinc
16129 @opindex mno-postmodify
16130 @opindex mpostmodify
16131 Code generation tweaks that disable, respectively, splitting of 32-bit
16132 loads, generation of post-increment addresses, and generation of
16133 post-modify addresses. The defaults are @option{msplit-lohi},
16134 @option{-mpost-inc}, and @option{-mpost-modify}.
16136 @item -mnovect-double
16137 @opindex mno-vect-double
16138 @opindex mvect-double
16139 Change the preferred SIMD mode to SImode. The default is
16140 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
16142 @item -max-vect-align=@var{num}
16143 @opindex max-vect-align
16144 The maximum alignment for SIMD vector mode types.
16145 @var{num} may be 4 or 8. The default is 8.
16146 Note that this is an ABI change, even though many library function
16147 interfaces are unaffected if they don't use SIMD vector modes
16148 in places that affect size and/or alignment of relevant types.
16150 @item -msplit-vecmove-early
16151 @opindex msplit-vecmove-early
16152 Split vector moves into single word moves before reload. In theory this
16153 can give better register allocation, but so far the reverse seems to be
16154 generally the case.
16156 @item -m1reg-@var{reg}
16158 Specify a register to hold the constant @minus{}1, which makes loading small negative
16159 constants and certain bitmasks faster.
16160 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
16161 which specify use of that register as a fixed register,
16162 and @samp{none}, which means that no register is used for this
16163 purpose. The default is @option{-m1reg-none}.
16167 @node AMD GCN Options
16168 @subsection AMD GCN Options
16169 @cindex AMD GCN Options
16171 These options are defined specifically for the AMD GCN port.
16175 @item -march=@var{gpu}
16177 @itemx -mtune=@var{gpu}
16179 Set architecture type or tuning for @var{gpu}. Supported values for @var{gpu}
16185 Compile for GCN3 Fiji devices (gfx803).
16188 Compile for GCN5 Vega 10 devices (gfx900).
16192 @item -mstack-size=@var{bytes}
16193 @opindex mstack-size
16194 Specify how many @var{bytes} of stack space will be requested for each GPU
16195 thread (wave-front). Beware that there may be many threads and limited memory
16196 available. The size of the stack allocation may also have an impact on
16197 run-time performance. The default is 32KB when using OpenACC or OpenMP, and
16203 @subsection ARC Options
16204 @cindex ARC options
16206 The following options control the architecture variant for which code
16209 @c architecture variants
16212 @item -mbarrel-shifter
16213 @opindex mbarrel-shifter
16214 Generate instructions supported by barrel shifter. This is the default
16215 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
16218 @opindex mjli-alawys
16219 Force to call a function using jli_s instruction. This option is
16220 valid only for ARCv2 architecture.
16222 @item -mcpu=@var{cpu}
16224 Set architecture type, register usage, and instruction scheduling
16225 parameters for @var{cpu}. There are also shortcut alias options
16226 available for backward compatibility and convenience. Supported
16227 values for @var{cpu} are
16233 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
16237 Compile for ARC601. Alias: @option{-mARC601}.
16242 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
16243 This is the default when configured with @option{--with-cpu=arc700}@.
16246 Compile for ARC EM.
16249 Compile for ARC HS.
16252 Compile for ARC EM CPU with no hardware extensions.
16255 Compile for ARC EM4 CPU.
16258 Compile for ARC EM4 DMIPS CPU.
16261 Compile for ARC EM4 DMIPS CPU with the single-precision floating-point
16265 Compile for ARC EM4 DMIPS CPU with single-precision floating-point and
16266 double assist instructions.
16269 Compile for ARC HS CPU with no hardware extensions except the atomic
16273 Compile for ARC HS34 CPU.
16276 Compile for ARC HS38 CPU.
16279 Compile for ARC HS38 CPU with all hardware extensions on.
16282 Compile for ARC 600 CPU with @code{norm} instructions enabled.
16284 @item arc600_mul32x16
16285 Compile for ARC 600 CPU with @code{norm} and 32x16-bit multiply
16286 instructions enabled.
16289 Compile for ARC 600 CPU with @code{norm} and @code{mul64}-family
16290 instructions enabled.
16293 Compile for ARC 601 CPU with @code{norm} instructions enabled.
16295 @item arc601_mul32x16
16296 Compile for ARC 601 CPU with @code{norm} and 32x16-bit multiply
16297 instructions enabled.
16300 Compile for ARC 601 CPU with @code{norm} and @code{mul64}-family
16301 instructions enabled.
16304 Compile for ARC 700 on NPS400 chip.
16307 Compile for ARC EM minimalist configuration featuring reduced register
16314 @itemx -mdpfp-compact
16315 @opindex mdpfp-compact
16316 Generate double-precision FPX instructions, tuned for the compact
16320 @opindex mdpfp-fast
16321 Generate double-precision FPX instructions, tuned for the fast
16324 @item -mno-dpfp-lrsr
16325 @opindex mno-dpfp-lrsr
16326 Disable @code{lr} and @code{sr} instructions from using FPX extension
16331 Generate extended arithmetic instructions. Currently only
16332 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
16333 supported. This is always enabled for @option{-mcpu=ARC700}.
16338 Do not generate @code{mpy}-family instructions for ARC700. This option is
16343 Generate 32x16-bit multiply and multiply-accumulate instructions.
16347 Generate @code{mul64} and @code{mulu64} instructions.
16348 Only valid for @option{-mcpu=ARC600}.
16352 Generate @code{norm} instructions. This is the default if @option{-mcpu=ARC700}
16357 @itemx -mspfp-compact
16358 @opindex mspfp-compact
16359 Generate single-precision FPX instructions, tuned for the compact
16363 @opindex mspfp-fast
16364 Generate single-precision FPX instructions, tuned for the fast
16369 Enable generation of ARC SIMD instructions via target-specific
16370 builtins. Only valid for @option{-mcpu=ARC700}.
16373 @opindex msoft-float
16374 This option ignored; it is provided for compatibility purposes only.
16375 Software floating-point code is emitted by default, and this default
16376 can overridden by FPX options; @option{-mspfp}, @option{-mspfp-compact}, or
16377 @option{-mspfp-fast} for single precision, and @option{-mdpfp},
16378 @option{-mdpfp-compact}, or @option{-mdpfp-fast} for double precision.
16382 Generate @code{swap} instructions.
16386 This enables use of the locked load/store conditional extension to implement
16387 atomic memory built-in functions. Not available for ARC 6xx or ARC
16392 Enable @code{div} and @code{rem} instructions for ARCv2 cores.
16394 @item -mcode-density
16395 @opindex mcode-density
16396 Enable code density instructions for ARC EM.
16397 This option is on by default for ARC HS.
16401 Enable double load/store operations for ARC HS cores.
16403 @item -mtp-regno=@var{regno}
16405 Specify thread pointer register number.
16407 @item -mmpy-option=@var{multo}
16408 @opindex mmpy-option
16409 Compile ARCv2 code with a multiplier design option. You can specify
16410 the option using either a string or numeric value for @var{multo}.
16411 @samp{wlh1} is the default value. The recognized values are:
16416 No multiplier available.
16420 16x16 multiplier, fully pipelined.
16421 The following instructions are enabled: @code{mpyw} and @code{mpyuw}.
16425 32x32 multiplier, fully
16426 pipelined (1 stage). The following instructions are additionally
16427 enabled: @code{mpy}, @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16431 32x32 multiplier, fully pipelined
16432 (2 stages). The following instructions are additionally enabled: @code{mpy},
16433 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16437 Two 16x16 multipliers, blocking,
16438 sequential. The following instructions are additionally enabled: @code{mpy},
16439 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16443 One 16x16 multiplier, blocking,
16444 sequential. The following instructions are additionally enabled: @code{mpy},
16445 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16449 One 32x4 multiplier, blocking,
16450 sequential. The following instructions are additionally enabled: @code{mpy},
16451 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16455 ARC HS SIMD support.
16459 ARC HS SIMD support.
16463 ARC HS SIMD support.
16467 This option is only available for ARCv2 cores@.
16469 @item -mfpu=@var{fpu}
16471 Enables support for specific floating-point hardware extensions for ARCv2
16472 cores. Supported values for @var{fpu} are:
16477 Enables support for single-precision floating-point hardware
16481 Enables support for double-precision floating-point hardware
16482 extensions. The single-precision floating-point extension is also
16483 enabled. Not available for ARC EM@.
16486 Enables support for double-precision floating-point hardware
16487 extensions using double-precision assist instructions. The single-precision
16488 floating-point extension is also enabled. This option is
16489 only available for ARC EM@.
16492 Enables support for double-precision floating-point hardware
16493 extensions using double-precision assist instructions.
16494 The single-precision floating-point, square-root, and divide
16495 extensions are also enabled. This option is
16496 only available for ARC EM@.
16499 Enables support for double-precision floating-point hardware
16500 extensions using double-precision assist instructions.
16501 The single-precision floating-point and fused multiply and add
16502 hardware extensions are also enabled. This option is
16503 only available for ARC EM@.
16506 Enables support for double-precision floating-point hardware
16507 extensions using double-precision assist instructions.
16508 All single-precision floating-point hardware extensions are also
16509 enabled. This option is only available for ARC EM@.
16512 Enables support for single-precision floating-point, square-root and divide
16513 hardware extensions@.
16516 Enables support for double-precision floating-point, square-root and divide
16517 hardware extensions. This option
16518 includes option @samp{fpus_div}. Not available for ARC EM@.
16521 Enables support for single-precision floating-point and
16522 fused multiply and add hardware extensions@.
16525 Enables support for double-precision floating-point and
16526 fused multiply and add hardware extensions. This option
16527 includes option @samp{fpus_fma}. Not available for ARC EM@.
16530 Enables support for all single-precision floating-point hardware
16534 Enables support for all single- and double-precision floating-point
16535 hardware extensions. Not available for ARC EM@.
16539 @item -mirq-ctrl-saved=@var{register-range}, @var{blink}, @var{lp_count}
16540 @opindex mirq-ctrl-saved
16541 Specifies general-purposes registers that the processor automatically
16542 saves/restores on interrupt entry and exit. @var{register-range} is
16543 specified as two registers separated by a dash. The register range
16544 always starts with @code{r0}, the upper limit is @code{fp} register.
16545 @var{blink} and @var{lp_count} are optional. This option is only
16546 valid for ARC EM and ARC HS cores.
16548 @item -mrgf-banked-regs=@var{number}
16549 @opindex mrgf-banked-regs
16550 Specifies the number of registers replicated in second register bank
16551 on entry to fast interrupt. Fast interrupts are interrupts with the
16552 highest priority level P0. These interrupts save only PC and STATUS32
16553 registers to avoid memory transactions during interrupt entry and exit
16554 sequences. Use this option when you are using fast interrupts in an
16555 ARC V2 family processor. Permitted values are 4, 8, 16, and 32.
16557 @item -mlpc-width=@var{width}
16558 @opindex mlpc-width
16559 Specify the width of the @code{lp_count} register. Valid values for
16560 @var{width} are 8, 16, 20, 24, 28 and 32 bits. The default width is
16561 fixed to 32 bits. If the width is less than 32, the compiler does not
16562 attempt to transform loops in your program to use the zero-delay loop
16563 mechanism unless it is known that the @code{lp_count} register can
16564 hold the required loop-counter value. Depending on the width
16565 specified, the compiler and run-time library might continue to use the
16566 loop mechanism for various needs. This option defines macro
16567 @code{__ARC_LPC_WIDTH__} with the value of @var{width}.
16571 This option instructs the compiler to generate code for a 16-entry
16572 register file. This option defines the @code{__ARC_RF16__}
16573 preprocessor macro.
16575 @item -mbranch-index
16576 @opindex mbranch-index
16577 Enable use of @code{bi} or @code{bih} instructions to implement jump
16582 The following options are passed through to the assembler, and also
16583 define preprocessor macro symbols.
16585 @c Flags used by the assembler, but for which we define preprocessor
16586 @c macro symbols as well.
16589 @opindex mdsp-packa
16590 Passed down to the assembler to enable the DSP Pack A extensions.
16591 Also sets the preprocessor symbol @code{__Xdsp_packa}. This option is
16596 Passed down to the assembler to enable the dual Viterbi butterfly
16597 extension. Also sets the preprocessor symbol @code{__Xdvbf}. This
16598 option is deprecated.
16600 @c ARC700 4.10 extension instruction
16603 Passed down to the assembler to enable the locked load/store
16604 conditional extension. Also sets the preprocessor symbol
16609 Passed down to the assembler. Also sets the preprocessor symbol
16610 @code{__Xxmac_d16}. This option is deprecated.
16614 Passed down to the assembler. Also sets the preprocessor symbol
16615 @code{__Xxmac_24}. This option is deprecated.
16617 @c ARC700 4.10 extension instruction
16620 Passed down to the assembler to enable the 64-bit time-stamp counter
16621 extension instruction. Also sets the preprocessor symbol
16622 @code{__Xrtsc}. This option is deprecated.
16624 @c ARC700 4.10 extension instruction
16627 Passed down to the assembler to enable the swap byte ordering
16628 extension instruction. Also sets the preprocessor symbol
16632 @opindex mtelephony
16633 Passed down to the assembler to enable dual- and single-operand
16634 instructions for telephony. Also sets the preprocessor symbol
16635 @code{__Xtelephony}. This option is deprecated.
16639 Passed down to the assembler to enable the XY memory extension. Also
16640 sets the preprocessor symbol @code{__Xxy}.
16644 The following options control how the assembly code is annotated:
16646 @c Assembly annotation options
16650 Annotate assembler instructions with estimated addresses.
16652 @item -mannotate-align
16653 @opindex mannotate-align
16654 Explain what alignment considerations lead to the decision to make an
16655 instruction short or long.
16659 The following options are passed through to the linker:
16661 @c options passed through to the linker
16665 Passed through to the linker, to specify use of the @code{arclinux} emulation.
16666 This option is enabled by default in tool chains built for
16667 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
16668 when profiling is not requested.
16670 @item -marclinux_prof
16671 @opindex marclinux_prof
16672 Passed through to the linker, to specify use of the
16673 @code{arclinux_prof} emulation. This option is enabled by default in
16674 tool chains built for @w{@code{arc-linux-uclibc}} and
16675 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
16679 The following options control the semantics of generated code:
16681 @c semantically relevant code generation options
16684 @opindex mlong-calls
16685 Generate calls as register indirect calls, thus providing access
16686 to the full 32-bit address range.
16688 @item -mmedium-calls
16689 @opindex mmedium-calls
16690 Don't use less than 25-bit addressing range for calls, which is the
16691 offset available for an unconditional branch-and-link
16692 instruction. Conditional execution of function calls is suppressed, to
16693 allow use of the 25-bit range, rather than the 21-bit range with
16694 conditional branch-and-link. This is the default for tool chains built
16695 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
16699 Put definitions of externally-visible data in a small data section if
16700 that data is no bigger than @var{num} bytes. The default value of
16701 @var{num} is 4 for any ARC configuration, or 8 when we have double
16702 load/store operations.
16707 Do not generate sdata references. This is the default for tool chains
16708 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
16711 @item -mvolatile-cache
16712 @opindex mvolatile-cache
16713 Use ordinarily cached memory accesses for volatile references. This is the
16716 @item -mno-volatile-cache
16717 @opindex mno-volatile-cache
16718 @opindex mvolatile-cache
16719 Enable cache bypass for volatile references.
16723 The following options fine tune code generation:
16724 @c code generation tuning options
16727 @opindex malign-call
16728 Do alignment optimizations for call instructions.
16730 @item -mauto-modify-reg
16731 @opindex mauto-modify-reg
16732 Enable the use of pre/post modify with register displacement.
16734 @item -mbbit-peephole
16735 @opindex mbbit-peephole
16736 Enable bbit peephole2.
16740 This option disables a target-specific pass in @file{arc_reorg} to
16741 generate compare-and-branch (@code{br@var{cc}}) instructions.
16742 It has no effect on
16743 generation of these instructions driven by the combiner pass.
16745 @item -mcase-vector-pcrel
16746 @opindex mcase-vector-pcrel
16747 Use PC-relative switch case tables to enable case table shortening.
16748 This is the default for @option{-Os}.
16750 @item -mcompact-casesi
16751 @opindex mcompact-casesi
16752 Enable compact @code{casesi} pattern. This is the default for @option{-Os},
16753 and only available for ARCv1 cores. This option is deprecated.
16755 @item -mno-cond-exec
16756 @opindex mno-cond-exec
16757 Disable the ARCompact-specific pass to generate conditional
16758 execution instructions.
16760 Due to delay slot scheduling and interactions between operand numbers,
16761 literal sizes, instruction lengths, and the support for conditional execution,
16762 the target-independent pass to generate conditional execution is often lacking,
16763 so the ARC port has kept a special pass around that tries to find more
16764 conditional execution generation opportunities after register allocation,
16765 branch shortening, and delay slot scheduling have been done. This pass
16766 generally, but not always, improves performance and code size, at the cost of
16767 extra compilation time, which is why there is an option to switch it off.
16768 If you have a problem with call instructions exceeding their allowable
16769 offset range because they are conditionalized, you should consider using
16770 @option{-mmedium-calls} instead.
16772 @item -mearly-cbranchsi
16773 @opindex mearly-cbranchsi
16774 Enable pre-reload use of the @code{cbranchsi} pattern.
16776 @item -mexpand-adddi
16777 @opindex mexpand-adddi
16778 Expand @code{adddi3} and @code{subdi3} at RTL generation time into
16779 @code{add.f}, @code{adc} etc. This option is deprecated.
16781 @item -mindexed-loads
16782 @opindex mindexed-loads
16783 Enable the use of indexed loads. This can be problematic because some
16784 optimizers then assume that indexed stores exist, which is not
16789 Enable Local Register Allocation. This is still experimental for ARC,
16790 so by default the compiler uses standard reload
16791 (i.e.@: @option{-mno-lra}).
16793 @item -mlra-priority-none
16794 @opindex mlra-priority-none
16795 Don't indicate any priority for target registers.
16797 @item -mlra-priority-compact
16798 @opindex mlra-priority-compact
16799 Indicate target register priority for r0..r3 / r12..r15.
16801 @item -mlra-priority-noncompact
16802 @opindex mlra-priority-noncompact
16803 Reduce target register priority for r0..r3 / r12..r15.
16806 @opindex mmillicode
16807 When optimizing for size (using @option{-Os}), prologues and epilogues
16808 that have to save or restore a large number of registers are often
16809 shortened by using call to a special function in libgcc; this is
16810 referred to as a @emph{millicode} call. As these calls can pose
16811 performance issues, and/or cause linking issues when linking in a
16812 nonstandard way, this option is provided to turn on or off millicode
16815 @item -mcode-density-frame
16816 @opindex mcode-density-frame
16817 This option enable the compiler to emit @code{enter} and @code{leave}
16818 instructions. These instructions are only valid for CPUs with
16819 code-density feature.
16822 @opindex mmixed-code
16823 Tweak register allocation to help 16-bit instruction generation.
16824 This generally has the effect of decreasing the average instruction size
16825 while increasing the instruction count.
16829 Enable @samp{q} instruction alternatives.
16830 This is the default for @option{-Os}.
16834 Enable @samp{Rcq} constraint handling.
16835 Most short code generation depends on this.
16836 This is the default.
16840 Enable @samp{Rcw} constraint handling.
16841 Most ccfsm condexec mostly depends on this.
16842 This is the default.
16844 @item -msize-level=@var{level}
16845 @opindex msize-level
16846 Fine-tune size optimization with regards to instruction lengths and alignment.
16847 The recognized values for @var{level} are:
16850 No size optimization. This level is deprecated and treated like @samp{1}.
16853 Short instructions are used opportunistically.
16856 In addition, alignment of loops and of code after barriers are dropped.
16859 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
16863 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
16864 the behavior when this is not set is equivalent to level @samp{1}.
16866 @item -mtune=@var{cpu}
16868 Set instruction scheduling parameters for @var{cpu}, overriding any implied
16869 by @option{-mcpu=}.
16871 Supported values for @var{cpu} are
16875 Tune for ARC600 CPU.
16878 Tune for ARC601 CPU.
16881 Tune for ARC700 CPU with standard multiplier block.
16884 Tune for ARC700 CPU with XMAC block.
16887 Tune for ARC725D CPU.
16890 Tune for ARC750D CPU.
16894 @item -mmultcost=@var{num}
16896 Cost to assume for a multiply instruction, with @samp{4} being equal to a
16897 normal instruction.
16899 @item -munalign-prob-threshold=@var{probability}
16900 @opindex munalign-prob-threshold
16901 Set probability threshold for unaligning branches.
16902 When tuning for @samp{ARC700} and optimizing for speed, branches without
16903 filled delay slot are preferably emitted unaligned and long, unless
16904 profiling indicates that the probability for the branch to be taken
16905 is below @var{probability}. @xref{Cross-profiling}.
16906 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
16910 The following options are maintained for backward compatibility, but
16911 are now deprecated and will be removed in a future release:
16913 @c Deprecated options
16921 @opindex mbig-endian
16924 Compile code for big-endian targets. Use of these options is now
16925 deprecated. Big-endian code is supported by configuring GCC to build
16926 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets,
16927 for which big endian is the default.
16929 @item -mlittle-endian
16930 @opindex mlittle-endian
16933 Compile code for little-endian targets. Use of these options is now
16934 deprecated. Little-endian code is supported by configuring GCC to build
16935 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets,
16936 for which little endian is the default.
16938 @item -mbarrel_shifter
16939 @opindex mbarrel_shifter
16940 Replaced by @option{-mbarrel-shifter}.
16942 @item -mdpfp_compact
16943 @opindex mdpfp_compact
16944 Replaced by @option{-mdpfp-compact}.
16947 @opindex mdpfp_fast
16948 Replaced by @option{-mdpfp-fast}.
16951 @opindex mdsp_packa
16952 Replaced by @option{-mdsp-packa}.
16956 Replaced by @option{-mea}.
16960 Replaced by @option{-mmac-24}.
16964 Replaced by @option{-mmac-d16}.
16966 @item -mspfp_compact
16967 @opindex mspfp_compact
16968 Replaced by @option{-mspfp-compact}.
16971 @opindex mspfp_fast
16972 Replaced by @option{-mspfp-fast}.
16974 @item -mtune=@var{cpu}
16976 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
16977 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
16978 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively.
16980 @item -multcost=@var{num}
16982 Replaced by @option{-mmultcost}.
16987 @subsection ARM Options
16988 @cindex ARM options
16990 These @samp{-m} options are defined for the ARM port:
16993 @item -mabi=@var{name}
16995 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
16996 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
16999 @opindex mapcs-frame
17000 Generate a stack frame that is compliant with the ARM Procedure Call
17001 Standard for all functions, even if this is not strictly necessary for
17002 correct execution of the code. Specifying @option{-fomit-frame-pointer}
17003 with this option causes the stack frames not to be generated for
17004 leaf functions. The default is @option{-mno-apcs-frame}.
17005 This option is deprecated.
17009 This is a synonym for @option{-mapcs-frame} and is deprecated.
17012 @c not currently implemented
17013 @item -mapcs-stack-check
17014 @opindex mapcs-stack-check
17015 Generate code to check the amount of stack space available upon entry to
17016 every function (that actually uses some stack space). If there is
17017 insufficient space available then either the function
17018 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
17019 called, depending upon the amount of stack space required. The runtime
17020 system is required to provide these functions. The default is
17021 @option{-mno-apcs-stack-check}, since this produces smaller code.
17023 @c not currently implemented
17024 @item -mapcs-reentrant
17025 @opindex mapcs-reentrant
17026 Generate reentrant, position-independent code. The default is
17027 @option{-mno-apcs-reentrant}.
17030 @item -mthumb-interwork
17031 @opindex mthumb-interwork
17032 Generate code that supports calling between the ARM and Thumb
17033 instruction sets. Without this option, on pre-v5 architectures, the
17034 two instruction sets cannot be reliably used inside one program. The
17035 default is @option{-mno-thumb-interwork}, since slightly larger code
17036 is generated when @option{-mthumb-interwork} is specified. In AAPCS
17037 configurations this option is meaningless.
17039 @item -mno-sched-prolog
17040 @opindex mno-sched-prolog
17041 @opindex msched-prolog
17042 Prevent the reordering of instructions in the function prologue, or the
17043 merging of those instruction with the instructions in the function's
17044 body. This means that all functions start with a recognizable set
17045 of instructions (or in fact one of a choice from a small set of
17046 different function prologues), and this information can be used to
17047 locate the start of functions inside an executable piece of code. The
17048 default is @option{-msched-prolog}.
17050 @item -mfloat-abi=@var{name}
17051 @opindex mfloat-abi
17052 Specifies which floating-point ABI to use. Permissible values
17053 are: @samp{soft}, @samp{softfp} and @samp{hard}.
17055 Specifying @samp{soft} causes GCC to generate output containing
17056 library calls for floating-point operations.
17057 @samp{softfp} allows the generation of code using hardware floating-point
17058 instructions, but still uses the soft-float calling conventions.
17059 @samp{hard} allows generation of floating-point instructions
17060 and uses FPU-specific calling conventions.
17062 The default depends on the specific target configuration. Note that
17063 the hard-float and soft-float ABIs are not link-compatible; you must
17064 compile your entire program with the same ABI, and link with a
17065 compatible set of libraries.
17067 @item -mlittle-endian
17068 @opindex mlittle-endian
17069 Generate code for a processor running in little-endian mode. This is
17070 the default for all standard configurations.
17073 @opindex mbig-endian
17074 Generate code for a processor running in big-endian mode; the default is
17075 to compile code for a little-endian processor.
17080 When linking a big-endian image select between BE8 and BE32 formats.
17081 The option has no effect for little-endian images and is ignored. The
17082 default is dependent on the selected target architecture. For ARMv6
17083 and later architectures the default is BE8, for older architectures
17084 the default is BE32. BE32 format has been deprecated by ARM.
17086 @item -march=@var{name}@r{[}+extension@dots{}@r{]}
17088 This specifies the name of the target ARM architecture. GCC uses this
17089 name to determine what kind of instructions it can emit when generating
17090 assembly code. This option can be used in conjunction with or instead
17091 of the @option{-mcpu=} option.
17093 Permissible names are:
17095 @samp{armv5t}, @samp{armv5te},
17096 @samp{armv6}, @samp{armv6j}, @samp{armv6k}, @samp{armv6kz}, @samp{armv6t2},
17097 @samp{armv6z}, @samp{armv6zk},
17098 @samp{armv7}, @samp{armv7-a}, @samp{armv7ve},
17099 @samp{armv8-a}, @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a},
17104 @samp{armv6-m}, @samp{armv6s-m},
17105 @samp{armv7-m}, @samp{armv7e-m},
17106 @samp{armv8-m.base}, @samp{armv8-m.main},
17107 @samp{iwmmxt} and @samp{iwmmxt2}.
17109 Additionally, the following architectures, which lack support for the
17110 Thumb execution state, are recognized but support is deprecated: @samp{armv4}.
17112 Many of the architectures support extensions. These can be added by
17113 appending @samp{+@var{extension}} to the architecture name. Extension
17114 options are processed in order and capabilities accumulate. An extension
17115 will also enable any necessary base extensions
17116 upon which it depends. For example, the @samp{+crypto} extension
17117 will always enable the @samp{+simd} extension. The exception to the
17118 additive construction is for extensions that are prefixed with
17119 @samp{+no@dots{}}: these extensions disable the specified option and
17120 any other extensions that may depend on the presence of that
17123 For example, @samp{-march=armv7-a+simd+nofp+vfpv4} is equivalent to
17124 writing @samp{-march=armv7-a+vfpv4} since the @samp{+simd} option is
17125 entirely disabled by the @samp{+nofp} option that follows it.
17127 Most extension names are generically named, but have an effect that is
17128 dependent upon the architecture to which it is applied. For example,
17129 the @samp{+simd} option can be applied to both @samp{armv7-a} and
17130 @samp{armv8-a} architectures, but will enable the original ARMv7-A
17131 Advanced SIMD (Neon) extensions for @samp{armv7-a} and the ARMv8-A
17132 variant for @samp{armv8-a}.
17134 The table below lists the supported extensions for each architecture.
17135 Architectures not mentioned do not support any extensions.
17148 The VFPv2 floating-point instructions. The extension @samp{+vfpv2} can be
17149 used as an alias for this extension.
17152 Disable the floating-point instructions.
17156 The common subset of the ARMv7-A, ARMv7-R and ARMv7-M architectures.
17159 The VFPv3 floating-point instructions, with 16 double-precision
17160 registers. The extension @samp{+vfpv3-d16} can be used as an alias
17161 for this extension. Note that floating-point is not supported by the
17162 base ARMv7-M architecture, but is compatible with both the ARMv7-A and
17163 ARMv7-R architectures.
17166 Disable the floating-point instructions.
17172 The multiprocessing extension.
17175 The security extension.
17178 The VFPv3 floating-point instructions, with 16 double-precision
17179 registers. The extension @samp{+vfpv3-d16} can be used as an alias
17180 for this extension.
17183 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
17184 The extensions @samp{+neon} and @samp{+neon-vfpv3} can be used as aliases
17185 for this extension.
17188 The VFPv3 floating-point instructions, with 32 double-precision
17191 @item +vfpv3-d16-fp16
17192 The VFPv3 floating-point instructions, with 16 double-precision
17193 registers and the half-precision floating-point conversion operations.
17196 The VFPv3 floating-point instructions, with 32 double-precision
17197 registers and the half-precision floating-point conversion operations.
17200 The VFPv4 floating-point instructions, with 16 double-precision
17204 The VFPv4 floating-point instructions, with 32 double-precision
17208 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
17209 the half-precision floating-point conversion operations.
17212 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions.
17215 Disable the Advanced SIMD instructions (does not disable floating point).
17218 Disable the floating-point and Advanced SIMD instructions.
17222 The extended version of the ARMv7-A architecture with support for
17226 The VFPv4 floating-point instructions, with 16 double-precision registers.
17227 The extension @samp{+vfpv4-d16} can be used as an alias for this extension.
17230 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions. The
17231 extension @samp{+neon-vfpv4} can be used as an alias for this extension.
17234 The VFPv3 floating-point instructions, with 16 double-precision
17238 The VFPv3 floating-point instructions, with 32 double-precision
17241 @item +vfpv3-d16-fp16
17242 The VFPv3 floating-point instructions, with 16 double-precision
17243 registers and the half-precision floating-point conversion operations.
17246 The VFPv3 floating-point instructions, with 32 double-precision
17247 registers and the half-precision floating-point conversion operations.
17250 The VFPv4 floating-point instructions, with 16 double-precision
17254 The VFPv4 floating-point instructions, with 32 double-precision
17258 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
17259 The extension @samp{+neon-vfpv3} can be used as an alias for this extension.
17262 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
17263 the half-precision floating-point conversion operations.
17266 Disable the Advanced SIMD instructions (does not disable floating point).
17269 Disable the floating-point and Advanced SIMD instructions.
17275 The Cyclic Redundancy Check (CRC) instructions.
17277 The ARMv8-A Advanced SIMD and floating-point instructions.
17279 The cryptographic instructions.
17281 Disable the cryptographic instructions.
17283 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17285 Speculation Barrier Instruction.
17287 Execution and Data Prediction Restriction Instructions.
17293 The ARMv8.1-A Advanced SIMD and floating-point instructions.
17296 The cryptographic instructions. This also enables the Advanced SIMD and
17297 floating-point instructions.
17300 Disable the cryptographic instructions.
17303 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17306 Speculation Barrier Instruction.
17309 Execution and Data Prediction Restriction Instructions.
17316 The half-precision floating-point data processing instructions.
17317 This also enables the Advanced SIMD and floating-point instructions.
17320 The half-precision floating-point fmla extension. This also enables
17321 the half-precision floating-point extension and Advanced SIMD and
17322 floating-point instructions.
17325 The ARMv8.1-A Advanced SIMD and floating-point instructions.
17328 The cryptographic instructions. This also enables the Advanced SIMD and
17329 floating-point instructions.
17332 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
17335 Disable the cryptographic extension.
17338 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17341 Speculation Barrier Instruction.
17344 Execution and Data Prediction Restriction Instructions.
17350 The half-precision floating-point data processing instructions.
17351 This also enables the Advanced SIMD and floating-point instructions as well
17352 as the Dot Product extension and the half-precision floating-point fmla
17356 The ARMv8.3-A Advanced SIMD and floating-point instructions as well as the
17357 Dot Product extension.
17360 The cryptographic instructions. This also enables the Advanced SIMD and
17361 floating-point instructions as well as the Dot Product extension.
17364 Disable the cryptographic extension.
17367 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17370 Speculation Barrier Instruction.
17373 Execution and Data Prediction Restriction Instructions.
17379 The half-precision floating-point data processing instructions.
17380 This also enables the Advanced SIMD and floating-point instructions as well
17381 as the Dot Product extension and the half-precision floating-point fmla
17385 The ARMv8.3-A Advanced SIMD and floating-point instructions as well as the
17386 Dot Product extension.
17389 The cryptographic instructions. This also enables the Advanced SIMD and
17390 floating-point instructions as well as the Dot Product extension.
17393 Disable the cryptographic extension.
17396 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17402 The single-precision VFPv3 floating-point instructions. The extension
17403 @samp{+vfpv3xd} can be used as an alias for this extension.
17406 The VFPv3 floating-point instructions with 16 double-precision registers.
17407 The extension +vfpv3-d16 can be used as an alias for this extension.
17409 @item +vfpv3xd-d16-fp16
17410 The single-precision VFPv3 floating-point instructions with 16 double-precision
17411 registers and the half-precision floating-point conversion operations.
17413 @item +vfpv3-d16-fp16
17414 The VFPv3 floating-point instructions with 16 double-precision
17415 registers and the half-precision floating-point conversion operations.
17418 Disable the floating-point extension.
17421 The ARM-state integer division instructions.
17424 Disable the ARM-state integer division extension.
17430 The single-precision VFPv4 floating-point instructions.
17433 The single-precision FPv5 floating-point instructions.
17436 The single- and double-precision FPv5 floating-point instructions.
17439 Disable the floating-point extensions.
17445 The DSP instructions.
17448 Disable the DSP extension.
17451 The single-precision floating-point instructions.
17454 The single- and double-precision floating-point instructions.
17457 Disable the floating-point extension.
17463 The Cyclic Redundancy Check (CRC) instructions.
17465 The single-precision FPv5 floating-point instructions.
17467 The ARMv8-A Advanced SIMD and floating-point instructions.
17469 The cryptographic instructions.
17471 Disable the cryptographic instructions.
17473 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17478 @option{-march=native} causes the compiler to auto-detect the architecture
17479 of the build computer. At present, this feature is only supported on
17480 GNU/Linux, and not all architectures are recognized. If the auto-detect
17481 is unsuccessful the option has no effect.
17483 @item -mtune=@var{name}
17485 This option specifies the name of the target ARM processor for
17486 which GCC should tune the performance of the code.
17487 For some ARM implementations better performance can be obtained by using
17489 Permissible names are: @samp{arm7tdmi}, @samp{arm7tdmi-s}, @samp{arm710t},
17490 @samp{arm720t}, @samp{arm740t}, @samp{strongarm}, @samp{strongarm110},
17491 @samp{strongarm1100}, 0@samp{strongarm1110}, @samp{arm8}, @samp{arm810},
17492 @samp{arm9}, @samp{arm9e}, @samp{arm920}, @samp{arm920t}, @samp{arm922t},
17493 @samp{arm946e-s}, @samp{arm966e-s}, @samp{arm968e-s}, @samp{arm926ej-s},
17494 @samp{arm940t}, @samp{arm9tdmi}, @samp{arm10tdmi}, @samp{arm1020t},
17495 @samp{arm1026ej-s}, @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
17496 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
17497 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
17498 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
17499 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
17500 @samp{cortex-a32}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
17501 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
17502 @samp{cortex-a76}, @samp{ares}, @samp{cortex-r4}, @samp{cortex-r4f},
17503 @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8}, @samp{cortex-r52},
17504 @samp{cortex-m0}, @samp{cortex-m0plus}, @samp{cortex-m1}, @samp{cortex-m3},
17505 @samp{cortex-m4}, @samp{cortex-m7}, @samp{cortex-m23}, @samp{cortex-m33},
17506 @samp{cortex-m1.small-multiply}, @samp{cortex-m0.small-multiply},
17507 @samp{cortex-m0plus.small-multiply}, @samp{exynos-m1}, @samp{marvell-pj4},
17508 @samp{neoverse-n1}, @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2},
17509 @samp{ep9312}, @samp{fa526}, @samp{fa626}, @samp{fa606te}, @samp{fa626te},
17510 @samp{fmp626}, @samp{fa726te}, @samp{xgene1}.
17512 Additionally, this option can specify that GCC should tune the performance
17513 of the code for a big.LITTLE system. Permissible names are:
17514 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
17515 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
17516 @samp{cortex-a72.cortex-a35}, @samp{cortex-a73.cortex-a53},
17517 @samp{cortex-a75.cortex-a55}, @samp{cortex-a76.cortex-a55}.
17519 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
17520 performance for a blend of processors within architecture @var{arch}.
17521 The aim is to generate code that run well on the current most popular
17522 processors, balancing between optimizations that benefit some CPUs in the
17523 range, and avoiding performance pitfalls of other CPUs. The effects of
17524 this option may change in future GCC versions as CPU models come and go.
17526 @option{-mtune} permits the same extension options as @option{-mcpu}, but
17527 the extension options do not affect the tuning of the generated code.
17529 @option{-mtune=native} causes the compiler to auto-detect the CPU
17530 of the build computer. At present, this feature is only supported on
17531 GNU/Linux, and not all architectures are recognized. If the auto-detect is
17532 unsuccessful the option has no effect.
17534 @item -mcpu=@var{name}@r{[}+extension@dots{}@r{]}
17536 This specifies the name of the target ARM processor. GCC uses this name
17537 to derive the name of the target ARM architecture (as if specified
17538 by @option{-march}) and the ARM processor type for which to tune for
17539 performance (as if specified by @option{-mtune}). Where this option
17540 is used in conjunction with @option{-march} or @option{-mtune},
17541 those options take precedence over the appropriate part of this option.
17543 Many of the supported CPUs implement optional architectural
17544 extensions. Where this is so the architectural extensions are
17545 normally enabled by default. If implementations that lack the
17546 extension exist, then the extension syntax can be used to disable
17547 those extensions that have been omitted. For floating-point and
17548 Advanced SIMD (Neon) instructions, the settings of the options
17549 @option{-mfloat-abi} and @option{-mfpu} must also be considered:
17550 floating-point and Advanced SIMD instructions will only be used if
17551 @option{-mfloat-abi} is not set to @samp{soft}; and any setting of
17552 @option{-mfpu} other than @samp{auto} will override the available
17553 floating-point and SIMD extension instructions.
17555 For example, @samp{cortex-a9} can be found in three major
17556 configurations: integer only, with just a floating-point unit or with
17557 floating-point and Advanced SIMD. The default is to enable all the
17558 instructions, but the extensions @samp{+nosimd} and @samp{+nofp} can
17559 be used to disable just the SIMD or both the SIMD and floating-point
17560 instructions respectively.
17562 Permissible names for this option are the same as those for
17565 The following extension options are common to the listed CPUs:
17569 Disable the DSP instructions on @samp{cortex-m33}.
17572 Disables the floating-point instructions on @samp{arm9e},
17573 @samp{arm946e-s}, @samp{arm966e-s}, @samp{arm968e-s}, @samp{arm10e},
17574 @samp{arm1020e}, @samp{arm1022e}, @samp{arm926ej-s},
17575 @samp{arm1026ej-s}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8},
17576 @samp{cortex-m4}, @samp{cortex-m7} and @samp{cortex-m33}.
17577 Disables the floating-point and SIMD instructions on
17578 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7},
17579 @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a12},
17580 @samp{cortex-a15}, @samp{cortex-a17}, @samp{cortex-a15.cortex-a7},
17581 @samp{cortex-a17.cortex-a7}, @samp{cortex-a32}, @samp{cortex-a35},
17582 @samp{cortex-a53} and @samp{cortex-a55}.
17585 Disables the double-precision component of the floating-point instructions
17586 on @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8}, @samp{cortex-r52} and
17590 Disables the SIMD (but not floating-point) instructions on
17591 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}
17592 and @samp{cortex-a9}.
17595 Enables the cryptographic instructions on @samp{cortex-a32},
17596 @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55}, @samp{cortex-a57},
17597 @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75}, @samp{exynos-m1},
17598 @samp{xgene1}, @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
17599 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53} and
17600 @samp{cortex-a75.cortex-a55}.
17603 Additionally the @samp{generic-armv7-a} pseudo target defaults to
17604 VFPv3 with 16 double-precision registers. It supports the following
17605 extension options: @samp{mp}, @samp{sec}, @samp{vfpv3-d16},
17606 @samp{vfpv3}, @samp{vfpv3-d16-fp16}, @samp{vfpv3-fp16},
17607 @samp{vfpv4-d16}, @samp{vfpv4}, @samp{neon}, @samp{neon-vfpv3},
17608 @samp{neon-fp16}, @samp{neon-vfpv4}. The meanings are the same as for
17609 the extensions to @option{-march=armv7-a}.
17611 @option{-mcpu=generic-@var{arch}} is also permissible, and is
17612 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
17613 See @option{-mtune} for more information.
17615 @option{-mcpu=native} causes the compiler to auto-detect the CPU
17616 of the build computer. At present, this feature is only supported on
17617 GNU/Linux, and not all architectures are recognized. If the auto-detect
17618 is unsuccessful the option has no effect.
17620 @item -mfpu=@var{name}
17622 This specifies what floating-point hardware (or hardware emulation) is
17623 available on the target. Permissible names are: @samp{auto}, @samp{vfpv2},
17625 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
17626 @samp{vfpv3xd-fp16}, @samp{neon-vfpv3}, @samp{neon-fp16}, @samp{vfpv4},
17627 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
17628 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
17629 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
17630 Note that @samp{neon} is an alias for @samp{neon-vfpv3} and @samp{vfp}
17631 is an alias for @samp{vfpv2}.
17633 The setting @samp{auto} is the default and is special. It causes the
17634 compiler to select the floating-point and Advanced SIMD instructions
17635 based on the settings of @option{-mcpu} and @option{-march}.
17637 If the selected floating-point hardware includes the NEON extension
17638 (e.g.@: @option{-mfpu=neon}), note that floating-point
17639 operations are not generated by GCC's auto-vectorization pass unless
17640 @option{-funsafe-math-optimizations} is also specified. This is
17641 because NEON hardware does not fully implement the IEEE 754 standard for
17642 floating-point arithmetic (in particular denormal values are treated as
17643 zero), so the use of NEON instructions may lead to a loss of precision.
17645 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}).
17647 @item -mfp16-format=@var{name}
17648 @opindex mfp16-format
17649 Specify the format of the @code{__fp16} half-precision floating-point type.
17650 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
17651 the default is @samp{none}, in which case the @code{__fp16} type is not
17652 defined. @xref{Half-Precision}, for more information.
17654 @item -mstructure-size-boundary=@var{n}
17655 @opindex mstructure-size-boundary
17656 The sizes of all structures and unions are rounded up to a multiple
17657 of the number of bits set by this option. Permissible values are 8, 32
17658 and 64. The default value varies for different toolchains. For the COFF
17659 targeted toolchain the default value is 8. A value of 64 is only allowed
17660 if the underlying ABI supports it.
17662 Specifying a larger number can produce faster, more efficient code, but
17663 can also increase the size of the program. Different values are potentially
17664 incompatible. Code compiled with one value cannot necessarily expect to
17665 work with code or libraries compiled with another value, if they exchange
17666 information using structures or unions.
17668 This option is deprecated.
17670 @item -mabort-on-noreturn
17671 @opindex mabort-on-noreturn
17672 Generate a call to the function @code{abort} at the end of a
17673 @code{noreturn} function. It is executed if the function tries to
17677 @itemx -mno-long-calls
17678 @opindex mlong-calls
17679 @opindex mno-long-calls
17680 Tells the compiler to perform function calls by first loading the
17681 address of the function into a register and then performing a subroutine
17682 call on this register. This switch is needed if the target function
17683 lies outside of the 64-megabyte addressing range of the offset-based
17684 version of subroutine call instruction.
17686 Even if this switch is enabled, not all function calls are turned
17687 into long calls. The heuristic is that static functions, functions
17688 that have the @code{short_call} attribute, functions that are inside
17689 the scope of a @code{#pragma no_long_calls} directive, and functions whose
17690 definitions have already been compiled within the current compilation
17691 unit are not turned into long calls. The exceptions to this rule are
17692 that weak function definitions, functions with the @code{long_call}
17693 attribute or the @code{section} attribute, and functions that are within
17694 the scope of a @code{#pragma long_calls} directive are always
17695 turned into long calls.
17697 This feature is not enabled by default. Specifying
17698 @option{-mno-long-calls} restores the default behavior, as does
17699 placing the function calls within the scope of a @code{#pragma
17700 long_calls_off} directive. Note these switches have no effect on how
17701 the compiler generates code to handle function calls via function
17704 @item -msingle-pic-base
17705 @opindex msingle-pic-base
17706 Treat the register used for PIC addressing as read-only, rather than
17707 loading it in the prologue for each function. The runtime system is
17708 responsible for initializing this register with an appropriate value
17709 before execution begins.
17711 @item -mpic-register=@var{reg}
17712 @opindex mpic-register
17713 Specify the register to be used for PIC addressing.
17714 For standard PIC base case, the default is any suitable register
17715 determined by compiler. For single PIC base case, the default is
17716 @samp{R9} if target is EABI based or stack-checking is enabled,
17717 otherwise the default is @samp{R10}.
17719 @item -mpic-data-is-text-relative
17720 @opindex mpic-data-is-text-relative
17721 Assume that the displacement between the text and data segments is fixed
17722 at static link time. This permits using PC-relative addressing
17723 operations to access data known to be in the data segment. For
17724 non-VxWorks RTP targets, this option is enabled by default. When
17725 disabled on such targets, it will enable @option{-msingle-pic-base} by
17728 @item -mpoke-function-name
17729 @opindex mpoke-function-name
17730 Write the name of each function into the text section, directly
17731 preceding the function prologue. The generated code is similar to this:
17735 .ascii "arm_poke_function_name", 0
17738 .word 0xff000000 + (t1 - t0)
17739 arm_poke_function_name
17741 stmfd sp!, @{fp, ip, lr, pc@}
17745 When performing a stack backtrace, code can inspect the value of
17746 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
17747 location @code{pc - 12} and the top 8 bits are set, then we know that
17748 there is a function name embedded immediately preceding this location
17749 and has length @code{((pc[-3]) & 0xff000000)}.
17756 Select between generating code that executes in ARM and Thumb
17757 states. The default for most configurations is to generate code
17758 that executes in ARM state, but the default can be changed by
17759 configuring GCC with the @option{--with-mode=}@var{state}
17762 You can also override the ARM and Thumb mode for each function
17763 by using the @code{target("thumb")} and @code{target("arm")} function attributes
17764 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
17767 @opindex mflip-thumb
17768 Switch ARM/Thumb modes on alternating functions.
17769 This option is provided for regression testing of mixed Thumb/ARM code
17770 generation, and is not intended for ordinary use in compiling code.
17773 @opindex mtpcs-frame
17774 Generate a stack frame that is compliant with the Thumb Procedure Call
17775 Standard for all non-leaf functions. (A leaf function is one that does
17776 not call any other functions.) The default is @option{-mno-tpcs-frame}.
17778 @item -mtpcs-leaf-frame
17779 @opindex mtpcs-leaf-frame
17780 Generate a stack frame that is compliant with the Thumb Procedure Call
17781 Standard for all leaf functions. (A leaf function is one that does
17782 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
17784 @item -mcallee-super-interworking
17785 @opindex mcallee-super-interworking
17786 Gives all externally visible functions in the file being compiled an ARM
17787 instruction set header which switches to Thumb mode before executing the
17788 rest of the function. This allows these functions to be called from
17789 non-interworking code. This option is not valid in AAPCS configurations
17790 because interworking is enabled by default.
17792 @item -mcaller-super-interworking
17793 @opindex mcaller-super-interworking
17794 Allows calls via function pointers (including virtual functions) to
17795 execute correctly regardless of whether the target code has been
17796 compiled for interworking or not. There is a small overhead in the cost
17797 of executing a function pointer if this option is enabled. This option
17798 is not valid in AAPCS configurations because interworking is enabled
17801 @item -mtp=@var{name}
17803 Specify the access model for the thread local storage pointer. The valid
17804 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
17805 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
17806 (supported in the arm6k architecture), and @samp{auto}, which uses the
17807 best available method for the selected processor. The default setting is
17810 @item -mtls-dialect=@var{dialect}
17811 @opindex mtls-dialect
17812 Specify the dialect to use for accessing thread local storage. Two
17813 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
17814 @samp{gnu} dialect selects the original GNU scheme for supporting
17815 local and global dynamic TLS models. The @samp{gnu2} dialect
17816 selects the GNU descriptor scheme, which provides better performance
17817 for shared libraries. The GNU descriptor scheme is compatible with
17818 the original scheme, but does require new assembler, linker and
17819 library support. Initial and local exec TLS models are unaffected by
17820 this option and always use the original scheme.
17822 @item -mword-relocations
17823 @opindex mword-relocations
17824 Only generate absolute relocations on word-sized values (i.e.@: R_ARM_ABS32).
17825 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
17826 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
17827 is specified. This option conflicts with @option{-mslow-flash-data}.
17829 @item -mfix-cortex-m3-ldrd
17830 @opindex mfix-cortex-m3-ldrd
17831 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
17832 with overlapping destination and base registers are used. This option avoids
17833 generating these instructions. This option is enabled by default when
17834 @option{-mcpu=cortex-m3} is specified.
17836 @item -munaligned-access
17837 @itemx -mno-unaligned-access
17838 @opindex munaligned-access
17839 @opindex mno-unaligned-access
17840 Enables (or disables) reading and writing of 16- and 32- bit values
17841 from addresses that are not 16- or 32- bit aligned. By default
17842 unaligned access is disabled for all pre-ARMv6, all ARMv6-M and for
17843 ARMv8-M Baseline architectures, and enabled for all other
17844 architectures. If unaligned access is not enabled then words in packed
17845 data structures are accessed a byte at a time.
17847 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
17848 generated object file to either true or false, depending upon the
17849 setting of this option. If unaligned access is enabled then the
17850 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
17853 @item -mneon-for-64bits
17854 @opindex mneon-for-64bits
17855 Enables using Neon to handle scalar 64-bits operations. This is
17856 disabled by default since the cost of moving data from core registers
17859 @item -mslow-flash-data
17860 @opindex mslow-flash-data
17861 Assume loading data from flash is slower than fetching instruction.
17862 Therefore literal load is minimized for better performance.
17863 This option is only supported when compiling for ARMv7 M-profile and
17864 off by default. It conflicts with @option{-mword-relocations}.
17866 @item -masm-syntax-unified
17867 @opindex masm-syntax-unified
17868 Assume inline assembler is using unified asm syntax. The default is
17869 currently off which implies divided syntax. This option has no impact
17870 on Thumb2. However, this may change in future releases of GCC.
17871 Divided syntax should be considered deprecated.
17873 @item -mrestrict-it
17874 @opindex mrestrict-it
17875 Restricts generation of IT blocks to conform to the rules of ARMv8-A.
17876 IT blocks can only contain a single 16-bit instruction from a select
17877 set of instructions. This option is on by default for ARMv8-A Thumb mode.
17879 @item -mprint-tune-info
17880 @opindex mprint-tune-info
17881 Print CPU tuning information as comment in assembler file. This is
17882 an option used only for regression testing of the compiler and not
17883 intended for ordinary use in compiling code. This option is disabled
17886 @item -mverbose-cost-dump
17887 @opindex mverbose-cost-dump
17888 Enable verbose cost model dumping in the debug dump files. This option is
17889 provided for use in debugging the compiler.
17892 @opindex mpure-code
17893 Do not allow constant data to be placed in code sections.
17894 Additionally, when compiling for ELF object format give all text sections the
17895 ELF processor-specific section attribute @code{SHF_ARM_PURECODE}. This option
17896 is only available when generating non-pic code for M-profile targets with the
17901 Generate secure code as per the "ARMv8-M Security Extensions: Requirements on
17902 Development Tools Engineering Specification", which can be found on
17903 @url{http://infocenter.arm.com/help/topic/com.arm.doc.ecm0359818/ECM0359818_armv8m_security_extensions_reqs_on_dev_tools_1_0.pdf}.
17907 @subsection AVR Options
17908 @cindex AVR Options
17910 These options are defined for AVR implementations:
17913 @item -mmcu=@var{mcu}
17915 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
17917 The default for this option is@tie{}@samp{avr2}.
17919 GCC supports the following AVR devices and ISAs:
17921 @include avr-mmcu.texi
17926 Assume that all data in static storage can be accessed by LDS / STS
17927 instructions. This option has only an effect on reduced Tiny devices like
17928 ATtiny40. See also the @code{absdata}
17929 @ref{AVR Variable Attributes,variable attribute}.
17931 @item -maccumulate-args
17932 @opindex maccumulate-args
17933 Accumulate outgoing function arguments and acquire/release the needed
17934 stack space for outgoing function arguments once in function
17935 prologue/epilogue. Without this option, outgoing arguments are pushed
17936 before calling a function and popped afterwards.
17938 Popping the arguments after the function call can be expensive on
17939 AVR so that accumulating the stack space might lead to smaller
17940 executables because arguments need not be removed from the
17941 stack after such a function call.
17943 This option can lead to reduced code size for functions that perform
17944 several calls to functions that get their arguments on the stack like
17945 calls to printf-like functions.
17947 @item -mbranch-cost=@var{cost}
17948 @opindex mbranch-cost
17949 Set the branch costs for conditional branch instructions to
17950 @var{cost}. Reasonable values for @var{cost} are small, non-negative
17951 integers. The default branch cost is 0.
17953 @item -mcall-prologues
17954 @opindex mcall-prologues
17955 Functions prologues/epilogues are expanded as calls to appropriate
17956 subroutines. Code size is smaller.
17958 @item -mgas-isr-prologues
17959 @opindex mgas-isr-prologues
17960 Interrupt service routines (ISRs) may use the @code{__gcc_isr} pseudo
17961 instruction supported by GNU Binutils.
17962 If this option is on, the feature can still be disabled for individual
17963 ISRs by means of the @ref{AVR Function Attributes,,@code{no_gccisr}}
17964 function attribute. This feature is activated per default
17965 if optimization is on (but not with @option{-Og}, @pxref{Optimize Options}),
17966 and if GNU Binutils support @w{@uref{https://sourceware.org/PR21683,PR21683}}.
17970 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
17971 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
17972 and @code{long long} is 4 bytes. Please note that this option does not
17973 conform to the C standards, but it results in smaller code
17976 @item -mmain-is-OS_task
17977 @opindex mmain-is-OS_task
17978 Do not save registers in @code{main}. The effect is the same like
17979 attaching attribute @ref{AVR Function Attributes,,@code{OS_task}}
17980 to @code{main}. It is activated per default if optimization is on.
17982 @item -mn-flash=@var{num}
17984 Assume that the flash memory has a size of
17985 @var{num} times 64@tie{}KiB.
17987 @item -mno-interrupts
17988 @opindex mno-interrupts
17989 Generated code is not compatible with hardware interrupts.
17990 Code size is smaller.
17994 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
17995 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
17996 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
17997 the assembler's command line and the @option{--relax} option to the
17998 linker's command line.
18000 Jump relaxing is performed by the linker because jump offsets are not
18001 known before code is located. Therefore, the assembler code generated by the
18002 compiler is the same, but the instructions in the executable may
18003 differ from instructions in the assembler code.
18005 Relaxing must be turned on if linker stubs are needed, see the
18006 section on @code{EIND} and linker stubs below.
18010 Assume that the device supports the Read-Modify-Write
18011 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
18013 @item -mshort-calls
18014 @opindex mshort-calls
18016 Assume that @code{RJMP} and @code{RCALL} can target the whole
18019 This option is used internally for multilib selection. It is
18020 not an optimization option, and you don't need to set it by hand.
18024 Treat the stack pointer register as an 8-bit register,
18025 i.e.@: assume the high byte of the stack pointer is zero.
18026 In general, you don't need to set this option by hand.
18028 This option is used internally by the compiler to select and
18029 build multilibs for architectures @code{avr2} and @code{avr25}.
18030 These architectures mix devices with and without @code{SPH}.
18031 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
18032 the compiler driver adds or removes this option from the compiler
18033 proper's command line, because the compiler then knows if the device
18034 or architecture has an 8-bit stack pointer and thus no @code{SPH}
18039 Use address register @code{X} in a way proposed by the hardware. This means
18040 that @code{X} is only used in indirect, post-increment or
18041 pre-decrement addressing.
18043 Without this option, the @code{X} register may be used in the same way
18044 as @code{Y} or @code{Z} which then is emulated by additional
18046 For example, loading a value with @code{X+const} addressing with a
18047 small non-negative @code{const < 64} to a register @var{Rn} is
18051 adiw r26, const ; X += const
18052 ld @var{Rn}, X ; @var{Rn} = *X
18053 sbiw r26, const ; X -= const
18057 @opindex mtiny-stack
18058 Only change the lower 8@tie{}bits of the stack pointer.
18060 @item -mfract-convert-truncate
18061 @opindex mfract-convert-truncate
18062 Allow to use truncation instead of rounding towards zero for fractional fixed-point types.
18065 @opindex nodevicelib
18066 Don't link against AVR-LibC's device specific library @code{lib<mcu>.a}.
18068 @item -Waddr-space-convert
18069 @opindex Waddr-space-convert
18070 @opindex Wno-addr-space-convert
18071 Warn about conversions between address spaces in the case where the
18072 resulting address space is not contained in the incoming address space.
18074 @item -Wmisspelled-isr
18075 @opindex Wmisspelled-isr
18076 @opindex Wno-misspelled-isr
18077 Warn if the ISR is misspelled, i.e.@: without __vector prefix.
18078 Enabled by default.
18081 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
18082 @cindex @code{EIND}
18083 Pointers in the implementation are 16@tie{}bits wide.
18084 The address of a function or label is represented as word address so
18085 that indirect jumps and calls can target any code address in the
18086 range of 64@tie{}Ki words.
18088 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
18089 bytes of program memory space, there is a special function register called
18090 @code{EIND} that serves as most significant part of the target address
18091 when @code{EICALL} or @code{EIJMP} instructions are used.
18093 Indirect jumps and calls on these devices are handled as follows by
18094 the compiler and are subject to some limitations:
18099 The compiler never sets @code{EIND}.
18102 The compiler uses @code{EIND} implicitly in @code{EICALL}/@code{EIJMP}
18103 instructions or might read @code{EIND} directly in order to emulate an
18104 indirect call/jump by means of a @code{RET} instruction.
18107 The compiler assumes that @code{EIND} never changes during the startup
18108 code or during the application. In particular, @code{EIND} is not
18109 saved/restored in function or interrupt service routine
18113 For indirect calls to functions and computed goto, the linker
18114 generates @emph{stubs}. Stubs are jump pads sometimes also called
18115 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
18116 The stub contains a direct jump to the desired address.
18119 Linker relaxation must be turned on so that the linker generates
18120 the stubs correctly in all situations. See the compiler option
18121 @option{-mrelax} and the linker option @option{--relax}.
18122 There are corner cases where the linker is supposed to generate stubs
18123 but aborts without relaxation and without a helpful error message.
18126 The default linker script is arranged for code with @code{EIND = 0}.
18127 If code is supposed to work for a setup with @code{EIND != 0}, a custom
18128 linker script has to be used in order to place the sections whose
18129 name start with @code{.trampolines} into the segment where @code{EIND}
18133 The startup code from libgcc never sets @code{EIND}.
18134 Notice that startup code is a blend of code from libgcc and AVR-LibC.
18135 For the impact of AVR-LibC on @code{EIND}, see the
18136 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
18139 It is legitimate for user-specific startup code to set up @code{EIND}
18140 early, for example by means of initialization code located in
18141 section @code{.init3}. Such code runs prior to general startup code
18142 that initializes RAM and calls constructors, but after the bit
18143 of startup code from AVR-LibC that sets @code{EIND} to the segment
18144 where the vector table is located.
18146 #include <avr/io.h>
18149 __attribute__((section(".init3"),naked,used,no_instrument_function))
18150 init3_set_eind (void)
18152 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
18153 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
18158 The @code{__trampolines_start} symbol is defined in the linker script.
18161 Stubs are generated automatically by the linker if
18162 the following two conditions are met:
18165 @item The address of a label is taken by means of the @code{gs} modifier
18166 (short for @emph{generate stubs}) like so:
18168 LDI r24, lo8(gs(@var{func}))
18169 LDI r25, hi8(gs(@var{func}))
18171 @item The final location of that label is in a code segment
18172 @emph{outside} the segment where the stubs are located.
18176 The compiler emits such @code{gs} modifiers for code labels in the
18177 following situations:
18179 @item Taking address of a function or code label.
18180 @item Computed goto.
18181 @item If prologue-save function is used, see @option{-mcall-prologues}
18182 command-line option.
18183 @item Switch/case dispatch tables. If you do not want such dispatch
18184 tables you can specify the @option{-fno-jump-tables} command-line option.
18185 @item C and C++ constructors/destructors called during startup/shutdown.
18186 @item If the tools hit a @code{gs()} modifier explained above.
18190 Jumping to non-symbolic addresses like so is @emph{not} supported:
18195 /* Call function at word address 0x2 */
18196 return ((int(*)(void)) 0x2)();
18200 Instead, a stub has to be set up, i.e.@: the function has to be called
18201 through a symbol (@code{func_4} in the example):
18206 extern int func_4 (void);
18208 /* Call function at byte address 0x4 */
18213 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
18214 Alternatively, @code{func_4} can be defined in the linker script.
18217 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
18218 @cindex @code{RAMPD}
18219 @cindex @code{RAMPX}
18220 @cindex @code{RAMPY}
18221 @cindex @code{RAMPZ}
18222 Some AVR devices support memories larger than the 64@tie{}KiB range
18223 that can be accessed with 16-bit pointers. To access memory locations
18224 outside this 64@tie{}KiB range, the content of a @code{RAMP}
18225 register is used as high part of the address:
18226 The @code{X}, @code{Y}, @code{Z} address register is concatenated
18227 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
18228 register, respectively, to get a wide address. Similarly,
18229 @code{RAMPD} is used together with direct addressing.
18233 The startup code initializes the @code{RAMP} special function
18234 registers with zero.
18237 If a @ref{AVR Named Address Spaces,named address space} other than
18238 generic or @code{__flash} is used, then @code{RAMPZ} is set
18239 as needed before the operation.
18242 If the device supports RAM larger than 64@tie{}KiB and the compiler
18243 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
18244 is reset to zero after the operation.
18247 If the device comes with a specific @code{RAMP} register, the ISR
18248 prologue/epilogue saves/restores that SFR and initializes it with
18249 zero in case the ISR code might (implicitly) use it.
18252 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
18253 If you use inline assembler to read from locations outside the
18254 16-bit address range and change one of the @code{RAMP} registers,
18255 you must reset it to zero after the access.
18259 @subsubsection AVR Built-in Macros
18261 GCC defines several built-in macros so that the user code can test
18262 for the presence or absence of features. Almost any of the following
18263 built-in macros are deduced from device capabilities and thus
18264 triggered by the @option{-mmcu=} command-line option.
18266 For even more AVR-specific built-in macros see
18267 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
18272 Build-in macro that resolves to a decimal number that identifies the
18273 architecture and depends on the @option{-mmcu=@var{mcu}} option.
18274 Possible values are:
18276 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
18277 @code{4}, @code{5}, @code{51}, @code{6}
18279 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
18280 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
18285 @code{102}, @code{103}, @code{104},
18286 @code{105}, @code{106}, @code{107}
18288 for @var{mcu}=@code{avrtiny},
18289 @code{avrxmega2}, @code{avrxmega3}, @code{avrxmega4},
18290 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
18291 If @var{mcu} specifies a device, this built-in macro is set
18292 accordingly. For example, with @option{-mmcu=atmega8} the macro is
18293 defined to @code{4}.
18295 @item __AVR_@var{Device}__
18296 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
18297 the device's name. For example, @option{-mmcu=atmega8} defines the
18298 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
18299 @code{__AVR_ATtiny261A__}, etc.
18301 The built-in macros' names follow
18302 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
18303 the device name as from the AVR user manual. The difference between
18304 @var{Device} in the built-in macro and @var{device} in
18305 @option{-mmcu=@var{device}} is that the latter is always lowercase.
18307 If @var{device} is not a device but only a core architecture like
18308 @samp{avr51}, this macro is not defined.
18310 @item __AVR_DEVICE_NAME__
18311 Setting @option{-mmcu=@var{device}} defines this built-in macro to
18312 the device's name. For example, with @option{-mmcu=atmega8} the macro
18313 is defined to @code{atmega8}.
18315 If @var{device} is not a device but only a core architecture like
18316 @samp{avr51}, this macro is not defined.
18318 @item __AVR_XMEGA__
18319 The device / architecture belongs to the XMEGA family of devices.
18321 @item __AVR_HAVE_ELPM__
18322 The device has the @code{ELPM} instruction.
18324 @item __AVR_HAVE_ELPMX__
18325 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
18326 R@var{n},Z+} instructions.
18328 @item __AVR_HAVE_MOVW__
18329 The device has the @code{MOVW} instruction to perform 16-bit
18330 register-register moves.
18332 @item __AVR_HAVE_LPMX__
18333 The device has the @code{LPM R@var{n},Z} and
18334 @code{LPM R@var{n},Z+} instructions.
18336 @item __AVR_HAVE_MUL__
18337 The device has a hardware multiplier.
18339 @item __AVR_HAVE_JMP_CALL__
18340 The device has the @code{JMP} and @code{CALL} instructions.
18341 This is the case for devices with more than 8@tie{}KiB of program
18344 @item __AVR_HAVE_EIJMP_EICALL__
18345 @itemx __AVR_3_BYTE_PC__
18346 The device has the @code{EIJMP} and @code{EICALL} instructions.
18347 This is the case for devices with more than 128@tie{}KiB of program memory.
18348 This also means that the program counter
18349 (PC) is 3@tie{}bytes wide.
18351 @item __AVR_2_BYTE_PC__
18352 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
18353 with up to 128@tie{}KiB of program memory.
18355 @item __AVR_HAVE_8BIT_SP__
18356 @itemx __AVR_HAVE_16BIT_SP__
18357 The stack pointer (SP) register is treated as 8-bit respectively
18358 16-bit register by the compiler.
18359 The definition of these macros is affected by @option{-mtiny-stack}.
18361 @item __AVR_HAVE_SPH__
18363 The device has the SPH (high part of stack pointer) special function
18364 register or has an 8-bit stack pointer, respectively.
18365 The definition of these macros is affected by @option{-mmcu=} and
18366 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
18369 @item __AVR_HAVE_RAMPD__
18370 @itemx __AVR_HAVE_RAMPX__
18371 @itemx __AVR_HAVE_RAMPY__
18372 @itemx __AVR_HAVE_RAMPZ__
18373 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
18374 @code{RAMPZ} special function register, respectively.
18376 @item __NO_INTERRUPTS__
18377 This macro reflects the @option{-mno-interrupts} command-line option.
18379 @item __AVR_ERRATA_SKIP__
18380 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
18381 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
18382 instructions because of a hardware erratum. Skip instructions are
18383 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
18384 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
18387 @item __AVR_ISA_RMW__
18388 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
18390 @item __AVR_SFR_OFFSET__=@var{offset}
18391 Instructions that can address I/O special function registers directly
18392 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
18393 address as if addressed by an instruction to access RAM like @code{LD}
18394 or @code{STS}. This offset depends on the device architecture and has
18395 to be subtracted from the RAM address in order to get the
18396 respective I/O@tie{}address.
18398 @item __AVR_SHORT_CALLS__
18399 The @option{-mshort-calls} command line option is set.
18401 @item __AVR_PM_BASE_ADDRESS__=@var{addr}
18402 Some devices support reading from flash memory by means of @code{LD*}
18403 instructions. The flash memory is seen in the data address space
18404 at an offset of @code{__AVR_PM_BASE_ADDRESS__}. If this macro
18405 is not defined, this feature is not available. If defined,
18406 the address space is linear and there is no need to put
18407 @code{.rodata} into RAM. This is handled by the default linker
18408 description file, and is currently available for
18409 @code{avrtiny} and @code{avrxmega3}. Even more convenient,
18410 there is no need to use address spaces like @code{__flash} or
18411 features like attribute @code{progmem} and @code{pgm_read_*}.
18413 @item __WITH_AVRLIBC__
18414 The compiler is configured to be used together with AVR-Libc.
18415 See the @option{--with-avrlibc} configure option.
18419 @node Blackfin Options
18420 @subsection Blackfin Options
18421 @cindex Blackfin Options
18424 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
18426 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
18427 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
18428 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
18429 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
18430 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
18431 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
18432 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
18433 @samp{bf561}, @samp{bf592}.
18435 The optional @var{sirevision} specifies the silicon revision of the target
18436 Blackfin processor. Any workarounds available for the targeted silicon revision
18437 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
18438 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
18439 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
18440 hexadecimal digits representing the major and minor numbers in the silicon
18441 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
18442 is not defined. If @var{sirevision} is @samp{any}, the
18443 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
18444 If this optional @var{sirevision} is not used, GCC assumes the latest known
18445 silicon revision of the targeted Blackfin processor.
18447 GCC defines a preprocessor macro for the specified @var{cpu}.
18448 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
18449 provided by libgloss to be linked in if @option{-msim} is not given.
18451 Without this option, @samp{bf532} is used as the processor by default.
18453 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
18454 only the preprocessor macro is defined.
18458 Specifies that the program will be run on the simulator. This causes
18459 the simulator BSP provided by libgloss to be linked in. This option
18460 has effect only for @samp{bfin-elf} toolchain.
18461 Certain other options, such as @option{-mid-shared-library} and
18462 @option{-mfdpic}, imply @option{-msim}.
18464 @item -momit-leaf-frame-pointer
18465 @opindex momit-leaf-frame-pointer
18466 Don't keep the frame pointer in a register for leaf functions. This
18467 avoids the instructions to save, set up and restore frame pointers and
18468 makes an extra register available in leaf functions.
18470 @item -mspecld-anomaly
18471 @opindex mspecld-anomaly
18472 When enabled, the compiler ensures that the generated code does not
18473 contain speculative loads after jump instructions. If this option is used,
18474 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
18476 @item -mno-specld-anomaly
18477 @opindex mno-specld-anomaly
18478 @opindex mspecld-anomaly
18479 Don't generate extra code to prevent speculative loads from occurring.
18481 @item -mcsync-anomaly
18482 @opindex mcsync-anomaly
18483 When enabled, the compiler ensures that the generated code does not
18484 contain CSYNC or SSYNC instructions too soon after conditional branches.
18485 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
18487 @item -mno-csync-anomaly
18488 @opindex mno-csync-anomaly
18489 @opindex mcsync-anomaly
18490 Don't generate extra code to prevent CSYNC or SSYNC instructions from
18491 occurring too soon after a conditional branch.
18495 When enabled, the compiler is free to take advantage of the knowledge that
18496 the entire program fits into the low 64k of memory.
18499 @opindex mno-low64k
18500 Assume that the program is arbitrarily large. This is the default.
18502 @item -mstack-check-l1
18503 @opindex mstack-check-l1
18504 Do stack checking using information placed into L1 scratchpad memory by the
18507 @item -mid-shared-library
18508 @opindex mid-shared-library
18509 Generate code that supports shared libraries via the library ID method.
18510 This allows for execute in place and shared libraries in an environment
18511 without virtual memory management. This option implies @option{-fPIC}.
18512 With a @samp{bfin-elf} target, this option implies @option{-msim}.
18514 @item -mno-id-shared-library
18515 @opindex mno-id-shared-library
18516 @opindex mid-shared-library
18517 Generate code that doesn't assume ID-based shared libraries are being used.
18518 This is the default.
18520 @item -mleaf-id-shared-library
18521 @opindex mleaf-id-shared-library
18522 Generate code that supports shared libraries via the library ID method,
18523 but assumes that this library or executable won't link against any other
18524 ID shared libraries. That allows the compiler to use faster code for jumps
18527 @item -mno-leaf-id-shared-library
18528 @opindex mno-leaf-id-shared-library
18529 @opindex mleaf-id-shared-library
18530 Do not assume that the code being compiled won't link against any ID shared
18531 libraries. Slower code is generated for jump and call insns.
18533 @item -mshared-library-id=n
18534 @opindex mshared-library-id
18535 Specifies the identification number of the ID-based shared library being
18536 compiled. Specifying a value of 0 generates more compact code; specifying
18537 other values forces the allocation of that number to the current
18538 library but is no more space- or time-efficient than omitting this option.
18542 Generate code that allows the data segment to be located in a different
18543 area of memory from the text segment. This allows for execute in place in
18544 an environment without virtual memory management by eliminating relocations
18545 against the text section.
18547 @item -mno-sep-data
18548 @opindex mno-sep-data
18550 Generate code that assumes that the data segment follows the text segment.
18551 This is the default.
18554 @itemx -mno-long-calls
18555 @opindex mlong-calls
18556 @opindex mno-long-calls
18557 Tells the compiler to perform function calls by first loading the
18558 address of the function into a register and then performing a subroutine
18559 call on this register. This switch is needed if the target function
18560 lies outside of the 24-bit addressing range of the offset-based
18561 version of subroutine call instruction.
18563 This feature is not enabled by default. Specifying
18564 @option{-mno-long-calls} restores the default behavior. Note these
18565 switches have no effect on how the compiler generates code to handle
18566 function calls via function pointers.
18570 Link with the fast floating-point library. This library relaxes some of
18571 the IEEE floating-point standard's rules for checking inputs against
18572 Not-a-Number (NAN), in the interest of performance.
18575 @opindex minline-plt
18576 Enable inlining of PLT entries in function calls to functions that are
18577 not known to bind locally. It has no effect without @option{-mfdpic}.
18580 @opindex mmulticore
18581 Build a standalone application for multicore Blackfin processors.
18582 This option causes proper start files and link scripts supporting
18583 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
18584 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
18586 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
18587 selects the one-application-per-core programming model. Without
18588 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
18589 programming model is used. In this model, the main function of Core B
18590 should be named as @code{coreb_main}.
18592 If this option is not used, the single-core application programming
18597 Build a standalone application for Core A of BF561 when using
18598 the one-application-per-core programming model. Proper start files
18599 and link scripts are used to support Core A, and the macro
18600 @code{__BFIN_COREA} is defined.
18601 This option can only be used in conjunction with @option{-mmulticore}.
18605 Build a standalone application for Core B of BF561 when using
18606 the one-application-per-core programming model. Proper start files
18607 and link scripts are used to support Core B, and the macro
18608 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
18609 should be used instead of @code{main}.
18610 This option can only be used in conjunction with @option{-mmulticore}.
18614 Build a standalone application for SDRAM. Proper start files and
18615 link scripts are used to put the application into SDRAM, and the macro
18616 @code{__BFIN_SDRAM} is defined.
18617 The loader should initialize SDRAM before loading the application.
18621 Assume that ICPLBs are enabled at run time. This has an effect on certain
18622 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
18623 are enabled; for standalone applications the default is off.
18627 @subsection C6X Options
18628 @cindex C6X Options
18631 @item -march=@var{name}
18633 This specifies the name of the target architecture. GCC uses this
18634 name to determine what kind of instructions it can emit when generating
18635 assembly code. Permissible names are: @samp{c62x},
18636 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
18639 @opindex mbig-endian
18640 Generate code for a big-endian target.
18642 @item -mlittle-endian
18643 @opindex mlittle-endian
18644 Generate code for a little-endian target. This is the default.
18648 Choose startup files and linker script suitable for the simulator.
18650 @item -msdata=default
18651 @opindex msdata=default
18652 Put small global and static data in the @code{.neardata} section,
18653 which is pointed to by register @code{B14}. Put small uninitialized
18654 global and static data in the @code{.bss} section, which is adjacent
18655 to the @code{.neardata} section. Put small read-only data into the
18656 @code{.rodata} section. The corresponding sections used for large
18657 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
18660 @opindex msdata=all
18661 Put all data, not just small objects, into the sections reserved for
18662 small data, and use addressing relative to the @code{B14} register to
18666 @opindex msdata=none
18667 Make no use of the sections reserved for small data, and use absolute
18668 addresses to access all data. Put all initialized global and static
18669 data in the @code{.fardata} section, and all uninitialized data in the
18670 @code{.far} section. Put all constant data into the @code{.const}
18675 @subsection CRIS Options
18676 @cindex CRIS Options
18678 These options are defined specifically for the CRIS ports.
18681 @item -march=@var{architecture-type}
18682 @itemx -mcpu=@var{architecture-type}
18685 Generate code for the specified architecture. The choices for
18686 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
18687 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
18688 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
18691 @item -mtune=@var{architecture-type}
18693 Tune to @var{architecture-type} everything applicable about the generated
18694 code, except for the ABI and the set of available instructions. The
18695 choices for @var{architecture-type} are the same as for
18696 @option{-march=@var{architecture-type}}.
18698 @item -mmax-stack-frame=@var{n}
18699 @opindex mmax-stack-frame
18700 Warn when the stack frame of a function exceeds @var{n} bytes.
18706 The options @option{-metrax4} and @option{-metrax100} are synonyms for
18707 @option{-march=v3} and @option{-march=v8} respectively.
18709 @item -mmul-bug-workaround
18710 @itemx -mno-mul-bug-workaround
18711 @opindex mmul-bug-workaround
18712 @opindex mno-mul-bug-workaround
18713 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
18714 models where it applies. This option is active by default.
18718 Enable CRIS-specific verbose debug-related information in the assembly
18719 code. This option also has the effect of turning off the @samp{#NO_APP}
18720 formatted-code indicator to the assembler at the beginning of the
18725 Do not use condition-code results from previous instruction; always emit
18726 compare and test instructions before use of condition codes.
18728 @item -mno-side-effects
18729 @opindex mno-side-effects
18730 @opindex mside-effects
18731 Do not emit instructions with side effects in addressing modes other than
18734 @item -mstack-align
18735 @itemx -mno-stack-align
18736 @itemx -mdata-align
18737 @itemx -mno-data-align
18738 @itemx -mconst-align
18739 @itemx -mno-const-align
18740 @opindex mstack-align
18741 @opindex mno-stack-align
18742 @opindex mdata-align
18743 @opindex mno-data-align
18744 @opindex mconst-align
18745 @opindex mno-const-align
18746 These options (@samp{no-} options) arrange (eliminate arrangements) for the
18747 stack frame, individual data and constants to be aligned for the maximum
18748 single data access size for the chosen CPU model. The default is to
18749 arrange for 32-bit alignment. ABI details such as structure layout are
18750 not affected by these options.
18758 Similar to the stack- data- and const-align options above, these options
18759 arrange for stack frame, writable data and constants to all be 32-bit,
18760 16-bit or 8-bit aligned. The default is 32-bit alignment.
18762 @item -mno-prologue-epilogue
18763 @itemx -mprologue-epilogue
18764 @opindex mno-prologue-epilogue
18765 @opindex mprologue-epilogue
18766 With @option{-mno-prologue-epilogue}, the normal function prologue and
18767 epilogue which set up the stack frame are omitted and no return
18768 instructions or return sequences are generated in the code. Use this
18769 option only together with visual inspection of the compiled code: no
18770 warnings or errors are generated when call-saved registers must be saved,
18771 or storage for local variables needs to be allocated.
18775 @opindex mno-gotplt
18777 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
18778 instruction sequences that load addresses for functions from the PLT part
18779 of the GOT rather than (traditional on other architectures) calls to the
18780 PLT@. The default is @option{-mgotplt}.
18784 Legacy no-op option only recognized with the cris-axis-elf and
18785 cris-axis-linux-gnu targets.
18789 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
18793 This option, recognized for the cris-axis-elf, arranges
18794 to link with input-output functions from a simulator library. Code,
18795 initialized data and zero-initialized data are allocated consecutively.
18799 Like @option{-sim}, but pass linker options to locate initialized data at
18800 0x40000000 and zero-initialized data at 0x80000000.
18804 @subsection CR16 Options
18805 @cindex CR16 Options
18807 These options are defined specifically for the CR16 ports.
18813 Enable the use of multiply-accumulate instructions. Disabled by default.
18817 @opindex mcr16cplus
18819 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
18824 Links the library libsim.a which is in compatible with simulator. Applicable
18825 to ELF compiler only.
18829 Choose integer type as 32-bit wide.
18833 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
18835 @item -mdata-model=@var{model}
18836 @opindex mdata-model
18837 Choose a data model. The choices for @var{model} are @samp{near},
18838 @samp{far} or @samp{medium}. @samp{medium} is default.
18839 However, @samp{far} is not valid with @option{-mcr16c}, as the
18840 CR16C architecture does not support the far data model.
18843 @node C-SKY Options
18844 @subsection C-SKY Options
18845 @cindex C-SKY Options
18847 GCC supports these options when compiling for C-SKY V2 processors.
18851 @item -march=@var{arch}
18853 Specify the C-SKY target architecture. Valid values for @var{arch} are:
18854 @samp{ck801}, @samp{ck802}, @samp{ck803}, @samp{ck807}, and @samp{ck810}.
18855 The default is @samp{ck810}.
18857 @item -mcpu=@var{cpu}
18859 Specify the C-SKY target processor. Valid values for @var{cpu} are:
18860 @samp{ck801}, @samp{ck801t},
18861 @samp{ck802}, @samp{ck802t}, @samp{ck802j},
18862 @samp{ck803}, @samp{ck803h}, @samp{ck803t}, @samp{ck803ht},
18863 @samp{ck803f}, @samp{ck803fh}, @samp{ck803e}, @samp{ck803eh},
18864 @samp{ck803et}, @samp{ck803eht}, @samp{ck803ef}, @samp{ck803efh},
18865 @samp{ck803ft}, @samp{ck803eft}, @samp{ck803efht}, @samp{ck803r1},
18866 @samp{ck803hr1}, @samp{ck803tr1}, @samp{ck803htr1}, @samp{ck803fr1},
18867 @samp{ck803fhr1}, @samp{ck803er1}, @samp{ck803ehr1}, @samp{ck803etr1},
18868 @samp{ck803ehtr1}, @samp{ck803efr1}, @samp{ck803efhr1}, @samp{ck803ftr1},
18869 @samp{ck803eftr1}, @samp{ck803efhtr1},
18870 @samp{ck803s}, @samp{ck803st}, @samp{ck803se}, @samp{ck803sf},
18871 @samp{ck803sef}, @samp{ck803seft},
18872 @samp{ck807e}, @samp{ck807ef}, @samp{ck807}, @samp{ck807f},
18873 @samp{ck810e}, @samp{ck810et}, @samp{ck810ef}, @samp{ck810eft},
18874 @samp{ck810}, @samp{ck810v}, @samp{ck810f}, @samp{ck810t}, @samp{ck810fv},
18875 @samp{ck810tv}, @samp{ck810ft}, and @samp{ck810ftv}.
18878 @opindex mbig-endian
18881 @itemx -mlittle-endian
18882 @opindex mlittle-endian
18886 Select big- or little-endian code. The default is little-endian.
18889 @opindex mhard-float
18890 @itemx -msoft-float
18891 @opindex msoft-float
18893 Select hardware or software floating-point implementations.
18894 The default is soft float.
18896 @item -mdouble-float
18897 @itemx -mno-double-float
18898 @opindex mdouble-float
18899 When @option{-mhard-float} is in effect, enable generation of
18900 double-precision float instructions. This is the default except
18901 when compiling for CK803.
18906 When @option{-mhard-float} is in effect, enable generation of
18907 @code{frecipd}, @code{fsqrtd}, and @code{fdivd} instructions.
18908 This is the default except when compiling for CK803.
18910 @item -mfpu=@var{fpu}
18912 Select the floating-point processor. This option can only be used with
18913 @option{-mhard-float}.
18914 Values for @var{fpu} are
18915 @samp{fpv2_sf} (equivalent to @samp{-mno-double-float -mno-fdivdu}),
18916 @samp{fpv2} (@samp{-mdouble-float -mno-divdu}), and
18917 @samp{fpv2_divd} (@samp{-mdouble-float -mdivdu}).
18922 Enable the extended @code{lrw} instruction. This option defaults to on
18923 for CK801 and off otherwise.
18928 Enable interrupt stack instructions; the default is off.
18930 The @option{-mistack} option is required to handle the
18931 @code{interrupt} and @code{isr} function attributes
18932 (@pxref{C-SKY Function Attributes}).
18936 Enable multiprocessor instructions; the default is off.
18940 Enable coprocessor instructions; the default is off.
18944 Enable coprocessor instructions; the default is off.
18948 Enable C-SKY security instructions; the default is off.
18952 Enable C-SKY trust instructions; the default is off.
18960 Enable C-SKY DSP, Enhanced DSP, or Vector DSP instructions, respectively.
18961 All of these options default to off.
18966 Generate divide instructions. Default is off.
18971 Generate code for Smart Mode, using only registers numbered 0-7 to allow
18972 use of 16-bit instructions. This option is ignored for CK801 where this
18973 is the required behavior, and it defaults to on for CK802.
18974 For other targets, the default is off.
18976 @item -mhigh-registers
18977 @itemx -mno-high-registers
18978 @opindex mhigh-registers
18979 Generate code using the high registers numbered 16-31. This option
18980 is not supported on CK801, CK802, or CK803, and is enabled by default
18981 for other processors.
18986 Generate code using global anchor symbol addresses.
18989 @itemx -mno-pushpop
18991 Generate code using @code{push} and @code{pop} instructions. This option
18994 @item -mmultiple-stld
18996 @itemx -mno-multiple-stld
18998 @opindex mmultiple-stld
18999 Generate code using @code{stm} and @code{ldm} instructions. This option
19000 isn't supported on CK801 but is enabled by default on other processors.
19003 @itemx -mno-constpool
19004 @opindex mconstpool
19005 Create constant pools in the compiler instead of deferring it to the
19006 assembler. This option is the default and required for correct code
19007 generation on CK801 and CK802, and is optional on other processors.
19010 @item -mno-stack-size
19011 @opindex mstack-size
19012 Emit @code{.stack_size} directives for each function in the assembly
19013 output. This option defaults to off.
19018 Generate code for the C-SKY compiler runtime instead of libgcc. This
19019 option defaults to off.
19021 @item -mbranch-cost=@var{n}
19022 @opindex mbranch-cost=
19023 Set the branch costs to roughly @code{n} instructions. The default is 1.
19025 @item -msched-prolog
19026 @itemx -mno-sched-prolog
19027 @opindex msched-prolog
19028 Permit scheduling of function prologue and epilogue sequences. Using
19029 this option can result in code that is not compliant with the C-SKY V2 ABI
19030 prologue requirements and that cannot be debugged or backtraced.
19031 It is disabled by default.
19035 @node Darwin Options
19036 @subsection Darwin Options
19037 @cindex Darwin options
19039 These options are defined for all architectures running the Darwin operating
19042 FSF GCC on Darwin does not create ``fat'' object files; it creates
19043 an object file for the single architecture that GCC was built to
19044 target. Apple's GCC on Darwin does create ``fat'' files if multiple
19045 @option{-arch} options are used; it does so by running the compiler or
19046 linker multiple times and joining the results together with
19049 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
19050 @samp{i686}) is determined by the flags that specify the ISA
19051 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
19052 @option{-force_cpusubtype_ALL} option can be used to override this.
19054 The Darwin tools vary in their behavior when presented with an ISA
19055 mismatch. The assembler, @file{as}, only permits instructions to
19056 be used that are valid for the subtype of the file it is generating,
19057 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
19058 The linker for shared libraries, @file{/usr/bin/libtool}, fails
19059 and prints an error if asked to create a shared library with a less
19060 restrictive subtype than its input files (for instance, trying to put
19061 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
19062 for executables, @command{ld}, quietly gives the executable the most
19063 restrictive subtype of any of its input files.
19068 Add the framework directory @var{dir} to the head of the list of
19069 directories to be searched for header files. These directories are
19070 interleaved with those specified by @option{-I} options and are
19071 scanned in a left-to-right order.
19073 A framework directory is a directory with frameworks in it. A
19074 framework is a directory with a @file{Headers} and/or
19075 @file{PrivateHeaders} directory contained directly in it that ends
19076 in @file{.framework}. The name of a framework is the name of this
19077 directory excluding the @file{.framework}. Headers associated with
19078 the framework are found in one of those two directories, with
19079 @file{Headers} being searched first. A subframework is a framework
19080 directory that is in a framework's @file{Frameworks} directory.
19081 Includes of subframework headers can only appear in a header of a
19082 framework that contains the subframework, or in a sibling subframework
19083 header. Two subframeworks are siblings if they occur in the same
19084 framework. A subframework should not have the same name as a
19085 framework; a warning is issued if this is violated. Currently a
19086 subframework cannot have subframeworks; in the future, the mechanism
19087 may be extended to support this. The standard frameworks can be found
19088 in @file{/System/Library/Frameworks} and
19089 @file{/Library/Frameworks}. An example include looks like
19090 @code{#include <Framework/header.h>}, where @file{Framework} denotes
19091 the name of the framework and @file{header.h} is found in the
19092 @file{PrivateHeaders} or @file{Headers} directory.
19094 @item -iframework@var{dir}
19095 @opindex iframework
19096 Like @option{-F} except the directory is a treated as a system
19097 directory. The main difference between this @option{-iframework} and
19098 @option{-F} is that with @option{-iframework} the compiler does not
19099 warn about constructs contained within header files found via
19100 @var{dir}. This option is valid only for the C family of languages.
19104 Emit debugging information for symbols that are used. For stabs
19105 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
19106 This is by default ON@.
19110 Emit debugging information for all symbols and types.
19112 @item -mmacosx-version-min=@var{version}
19113 The earliest version of MacOS X that this executable will run on
19114 is @var{version}. Typical values of @var{version} include @code{10.1},
19115 @code{10.2}, and @code{10.3.9}.
19117 If the compiler was built to use the system's headers by default,
19118 then the default for this option is the system version on which the
19119 compiler is running, otherwise the default is to make choices that
19120 are compatible with as many systems and code bases as possible.
19124 Enable kernel development mode. The @option{-mkernel} option sets
19125 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
19126 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
19127 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
19128 applicable. This mode also sets @option{-mno-altivec},
19129 @option{-msoft-float}, @option{-fno-builtin} and
19130 @option{-mlong-branch} for PowerPC targets.
19132 @item -mone-byte-bool
19133 @opindex mone-byte-bool
19134 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
19135 By default @code{sizeof(bool)} is @code{4} when compiling for
19136 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
19137 option has no effect on x86.
19139 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
19140 to generate code that is not binary compatible with code generated
19141 without that switch. Using this switch may require recompiling all
19142 other modules in a program, including system libraries. Use this
19143 switch to conform to a non-default data model.
19145 @item -mfix-and-continue
19146 @itemx -ffix-and-continue
19147 @itemx -findirect-data
19148 @opindex mfix-and-continue
19149 @opindex ffix-and-continue
19150 @opindex findirect-data
19151 Generate code suitable for fast turnaround development, such as to
19152 allow GDB to dynamically load @file{.o} files into already-running
19153 programs. @option{-findirect-data} and @option{-ffix-and-continue}
19154 are provided for backwards compatibility.
19158 Loads all members of static archive libraries.
19159 See man ld(1) for more information.
19161 @item -arch_errors_fatal
19162 @opindex arch_errors_fatal
19163 Cause the errors having to do with files that have the wrong architecture
19166 @item -bind_at_load
19167 @opindex bind_at_load
19168 Causes the output file to be marked such that the dynamic linker will
19169 bind all undefined references when the file is loaded or launched.
19173 Produce a Mach-o bundle format file.
19174 See man ld(1) for more information.
19176 @item -bundle_loader @var{executable}
19177 @opindex bundle_loader
19178 This option specifies the @var{executable} that will load the build
19179 output file being linked. See man ld(1) for more information.
19182 @opindex dynamiclib
19183 When passed this option, GCC produces a dynamic library instead of
19184 an executable when linking, using the Darwin @file{libtool} command.
19186 @item -force_cpusubtype_ALL
19187 @opindex force_cpusubtype_ALL
19188 This causes GCC's output file to have the @samp{ALL} subtype, instead of
19189 one controlled by the @option{-mcpu} or @option{-march} option.
19191 @item -allowable_client @var{client_name}
19192 @itemx -client_name
19193 @itemx -compatibility_version
19194 @itemx -current_version
19196 @itemx -dependency-file
19198 @itemx -dylinker_install_name
19200 @itemx -exported_symbols_list
19203 @itemx -flat_namespace
19204 @itemx -force_flat_namespace
19205 @itemx -headerpad_max_install_names
19208 @itemx -install_name
19209 @itemx -keep_private_externs
19210 @itemx -multi_module
19211 @itemx -multiply_defined
19212 @itemx -multiply_defined_unused
19215 @itemx -no_dead_strip_inits_and_terms
19216 @itemx -nofixprebinding
19217 @itemx -nomultidefs
19219 @itemx -noseglinkedit
19220 @itemx -pagezero_size
19222 @itemx -prebind_all_twolevel_modules
19223 @itemx -private_bundle
19225 @itemx -read_only_relocs
19227 @itemx -sectobjectsymbols
19231 @itemx -sectobjectsymbols
19234 @itemx -segs_read_only_addr
19236 @itemx -segs_read_write_addr
19237 @itemx -seg_addr_table
19238 @itemx -seg_addr_table_filename
19239 @itemx -seglinkedit
19241 @itemx -segs_read_only_addr
19242 @itemx -segs_read_write_addr
19243 @itemx -single_module
19245 @itemx -sub_library
19247 @itemx -sub_umbrella
19248 @itemx -twolevel_namespace
19251 @itemx -unexported_symbols_list
19252 @itemx -weak_reference_mismatches
19253 @itemx -whatsloaded
19254 @opindex allowable_client
19255 @opindex client_name
19256 @opindex compatibility_version
19257 @opindex current_version
19258 @opindex dead_strip
19259 @opindex dependency-file
19260 @opindex dylib_file
19261 @opindex dylinker_install_name
19263 @opindex exported_symbols_list
19265 @opindex flat_namespace
19266 @opindex force_flat_namespace
19267 @opindex headerpad_max_install_names
19268 @opindex image_base
19270 @opindex install_name
19271 @opindex keep_private_externs
19272 @opindex multi_module
19273 @opindex multiply_defined
19274 @opindex multiply_defined_unused
19275 @opindex noall_load
19276 @opindex no_dead_strip_inits_and_terms
19277 @opindex nofixprebinding
19278 @opindex nomultidefs
19280 @opindex noseglinkedit
19281 @opindex pagezero_size
19283 @opindex prebind_all_twolevel_modules
19284 @opindex private_bundle
19285 @opindex read_only_relocs
19287 @opindex sectobjectsymbols
19290 @opindex sectcreate
19291 @opindex sectobjectsymbols
19294 @opindex segs_read_only_addr
19295 @opindex segs_read_write_addr
19296 @opindex seg_addr_table
19297 @opindex seg_addr_table_filename
19298 @opindex seglinkedit
19300 @opindex segs_read_only_addr
19301 @opindex segs_read_write_addr
19302 @opindex single_module
19304 @opindex sub_library
19305 @opindex sub_umbrella
19306 @opindex twolevel_namespace
19309 @opindex unexported_symbols_list
19310 @opindex weak_reference_mismatches
19311 @opindex whatsloaded
19312 These options are passed to the Darwin linker. The Darwin linker man page
19313 describes them in detail.
19316 @node DEC Alpha Options
19317 @subsection DEC Alpha Options
19319 These @samp{-m} options are defined for the DEC Alpha implementations:
19322 @item -mno-soft-float
19323 @itemx -msoft-float
19324 @opindex mno-soft-float
19325 @opindex msoft-float
19326 Use (do not use) the hardware floating-point instructions for
19327 floating-point operations. When @option{-msoft-float} is specified,
19328 functions in @file{libgcc.a} are used to perform floating-point
19329 operations. Unless they are replaced by routines that emulate the
19330 floating-point operations, or compiled in such a way as to call such
19331 emulations routines, these routines issue floating-point
19332 operations. If you are compiling for an Alpha without floating-point
19333 operations, you must ensure that the library is built so as not to call
19336 Note that Alpha implementations without floating-point operations are
19337 required to have floating-point registers.
19340 @itemx -mno-fp-regs
19342 @opindex mno-fp-regs
19343 Generate code that uses (does not use) the floating-point register set.
19344 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
19345 register set is not used, floating-point operands are passed in integer
19346 registers as if they were integers and floating-point results are passed
19347 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
19348 so any function with a floating-point argument or return value called by code
19349 compiled with @option{-mno-fp-regs} must also be compiled with that
19352 A typical use of this option is building a kernel that does not use,
19353 and hence need not save and restore, any floating-point registers.
19357 The Alpha architecture implements floating-point hardware optimized for
19358 maximum performance. It is mostly compliant with the IEEE floating-point
19359 standard. However, for full compliance, software assistance is
19360 required. This option generates code fully IEEE-compliant code
19361 @emph{except} that the @var{inexact-flag} is not maintained (see below).
19362 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
19363 defined during compilation. The resulting code is less efficient but is
19364 able to correctly support denormalized numbers and exceptional IEEE
19365 values such as not-a-number and plus/minus infinity. Other Alpha
19366 compilers call this option @option{-ieee_with_no_inexact}.
19368 @item -mieee-with-inexact
19369 @opindex mieee-with-inexact
19370 This is like @option{-mieee} except the generated code also maintains
19371 the IEEE @var{inexact-flag}. Turning on this option causes the
19372 generated code to implement fully-compliant IEEE math. In addition to
19373 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
19374 macro. On some Alpha implementations the resulting code may execute
19375 significantly slower than the code generated by default. Since there is
19376 very little code that depends on the @var{inexact-flag}, you should
19377 normally not specify this option. Other Alpha compilers call this
19378 option @option{-ieee_with_inexact}.
19380 @item -mfp-trap-mode=@var{trap-mode}
19381 @opindex mfp-trap-mode
19382 This option controls what floating-point related traps are enabled.
19383 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
19384 The trap mode can be set to one of four values:
19388 This is the default (normal) setting. The only traps that are enabled
19389 are the ones that cannot be disabled in software (e.g., division by zero
19393 In addition to the traps enabled by @samp{n}, underflow traps are enabled
19397 Like @samp{u}, but the instructions are marked to be safe for software
19398 completion (see Alpha architecture manual for details).
19401 Like @samp{su}, but inexact traps are enabled as well.
19404 @item -mfp-rounding-mode=@var{rounding-mode}
19405 @opindex mfp-rounding-mode
19406 Selects the IEEE rounding mode. Other Alpha compilers call this option
19407 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
19412 Normal IEEE rounding mode. Floating-point numbers are rounded towards
19413 the nearest machine number or towards the even machine number in case
19417 Round towards minus infinity.
19420 Chopped rounding mode. Floating-point numbers are rounded towards zero.
19423 Dynamic rounding mode. A field in the floating-point control register
19424 (@var{fpcr}, see Alpha architecture reference manual) controls the
19425 rounding mode in effect. The C library initializes this register for
19426 rounding towards plus infinity. Thus, unless your program modifies the
19427 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
19430 @item -mtrap-precision=@var{trap-precision}
19431 @opindex mtrap-precision
19432 In the Alpha architecture, floating-point traps are imprecise. This
19433 means without software assistance it is impossible to recover from a
19434 floating trap and program execution normally needs to be terminated.
19435 GCC can generate code that can assist operating system trap handlers
19436 in determining the exact location that caused a floating-point trap.
19437 Depending on the requirements of an application, different levels of
19438 precisions can be selected:
19442 Program precision. This option is the default and means a trap handler
19443 can only identify which program caused a floating-point exception.
19446 Function precision. The trap handler can determine the function that
19447 caused a floating-point exception.
19450 Instruction precision. The trap handler can determine the exact
19451 instruction that caused a floating-point exception.
19454 Other Alpha compilers provide the equivalent options called
19455 @option{-scope_safe} and @option{-resumption_safe}.
19457 @item -mieee-conformant
19458 @opindex mieee-conformant
19459 This option marks the generated code as IEEE conformant. You must not
19460 use this option unless you also specify @option{-mtrap-precision=i} and either
19461 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
19462 is to emit the line @samp{.eflag 48} in the function prologue of the
19463 generated assembly file.
19465 @item -mbuild-constants
19466 @opindex mbuild-constants
19467 Normally GCC examines a 32- or 64-bit integer constant to
19468 see if it can construct it from smaller constants in two or three
19469 instructions. If it cannot, it outputs the constant as a literal and
19470 generates code to load it from the data segment at run time.
19472 Use this option to require GCC to construct @emph{all} integer constants
19473 using code, even if it takes more instructions (the maximum is six).
19475 You typically use this option to build a shared library dynamic
19476 loader. Itself a shared library, it must relocate itself in memory
19477 before it can find the variables and constants in its own data segment.
19495 Indicate whether GCC should generate code to use the optional BWX,
19496 CIX, FIX and MAX instruction sets. The default is to use the instruction
19497 sets supported by the CPU type specified via @option{-mcpu=} option or that
19498 of the CPU on which GCC was built if none is specified.
19501 @itemx -mfloat-ieee
19502 @opindex mfloat-vax
19503 @opindex mfloat-ieee
19504 Generate code that uses (does not use) VAX F and G floating-point
19505 arithmetic instead of IEEE single and double precision.
19507 @item -mexplicit-relocs
19508 @itemx -mno-explicit-relocs
19509 @opindex mexplicit-relocs
19510 @opindex mno-explicit-relocs
19511 Older Alpha assemblers provided no way to generate symbol relocations
19512 except via assembler macros. Use of these macros does not allow
19513 optimal instruction scheduling. GNU binutils as of version 2.12
19514 supports a new syntax that allows the compiler to explicitly mark
19515 which relocations should apply to which instructions. This option
19516 is mostly useful for debugging, as GCC detects the capabilities of
19517 the assembler when it is built and sets the default accordingly.
19520 @itemx -mlarge-data
19521 @opindex msmall-data
19522 @opindex mlarge-data
19523 When @option{-mexplicit-relocs} is in effect, static data is
19524 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
19525 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
19526 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
19527 16-bit relocations off of the @code{$gp} register. This limits the
19528 size of the small data area to 64KB, but allows the variables to be
19529 directly accessed via a single instruction.
19531 The default is @option{-mlarge-data}. With this option the data area
19532 is limited to just below 2GB@. Programs that require more than 2GB of
19533 data must use @code{malloc} or @code{mmap} to allocate the data in the
19534 heap instead of in the program's data segment.
19536 When generating code for shared libraries, @option{-fpic} implies
19537 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
19540 @itemx -mlarge-text
19541 @opindex msmall-text
19542 @opindex mlarge-text
19543 When @option{-msmall-text} is used, the compiler assumes that the
19544 code of the entire program (or shared library) fits in 4MB, and is
19545 thus reachable with a branch instruction. When @option{-msmall-data}
19546 is used, the compiler can assume that all local symbols share the
19547 same @code{$gp} value, and thus reduce the number of instructions
19548 required for a function call from 4 to 1.
19550 The default is @option{-mlarge-text}.
19552 @item -mcpu=@var{cpu_type}
19554 Set the instruction set and instruction scheduling parameters for
19555 machine type @var{cpu_type}. You can specify either the @samp{EV}
19556 style name or the corresponding chip number. GCC supports scheduling
19557 parameters for the EV4, EV5 and EV6 family of processors and
19558 chooses the default values for the instruction set from the processor
19559 you specify. If you do not specify a processor type, GCC defaults
19560 to the processor on which the compiler was built.
19562 Supported values for @var{cpu_type} are
19568 Schedules as an EV4 and has no instruction set extensions.
19572 Schedules as an EV5 and has no instruction set extensions.
19576 Schedules as an EV5 and supports the BWX extension.
19581 Schedules as an EV5 and supports the BWX and MAX extensions.
19585 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
19589 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
19592 Native toolchains also support the value @samp{native},
19593 which selects the best architecture option for the host processor.
19594 @option{-mcpu=native} has no effect if GCC does not recognize
19597 @item -mtune=@var{cpu_type}
19599 Set only the instruction scheduling parameters for machine type
19600 @var{cpu_type}. The instruction set is not changed.
19602 Native toolchains also support the value @samp{native},
19603 which selects the best architecture option for the host processor.
19604 @option{-mtune=native} has no effect if GCC does not recognize
19607 @item -mmemory-latency=@var{time}
19608 @opindex mmemory-latency
19609 Sets the latency the scheduler should assume for typical memory
19610 references as seen by the application. This number is highly
19611 dependent on the memory access patterns used by the application
19612 and the size of the external cache on the machine.
19614 Valid options for @var{time} are
19618 A decimal number representing clock cycles.
19624 The compiler contains estimates of the number of clock cycles for
19625 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
19626 (also called Dcache, Scache, and Bcache), as well as to main memory.
19627 Note that L3 is only valid for EV5.
19633 @subsection FR30 Options
19634 @cindex FR30 Options
19636 These options are defined specifically for the FR30 port.
19640 @item -msmall-model
19641 @opindex msmall-model
19642 Use the small address space model. This can produce smaller code, but
19643 it does assume that all symbolic values and addresses fit into a
19648 Assume that runtime support has been provided and so there is no need
19649 to include the simulator library (@file{libsim.a}) on the linker
19655 @subsection FT32 Options
19656 @cindex FT32 Options
19658 These options are defined specifically for the FT32 port.
19664 Specifies that the program will be run on the simulator. This causes
19665 an alternate runtime startup and library to be linked.
19666 You must not use this option when generating programs that will run on
19667 real hardware; you must provide your own runtime library for whatever
19668 I/O functions are needed.
19672 Enable Local Register Allocation. This is still experimental for FT32,
19673 so by default the compiler uses standard reload.
19677 Do not use div and mod instructions.
19681 Enable use of the extended instructions of the FT32B processor.
19685 Compress all code using the Ft32B code compression scheme.
19689 Do not generate code that reads program memory.
19694 @subsection FRV Options
19695 @cindex FRV Options
19701 Only use the first 32 general-purpose registers.
19706 Use all 64 general-purpose registers.
19711 Use only the first 32 floating-point registers.
19716 Use all 64 floating-point registers.
19719 @opindex mhard-float
19721 Use hardware instructions for floating-point operations.
19724 @opindex msoft-float
19726 Use library routines for floating-point operations.
19731 Dynamically allocate condition code registers.
19736 Do not try to dynamically allocate condition code registers, only
19737 use @code{icc0} and @code{fcc0}.
19742 Change ABI to use double word insns.
19748 Do not use double word instructions.
19753 Use floating-point double instructions.
19756 @opindex mno-double
19758 Do not use floating-point double instructions.
19763 Use media instructions.
19768 Do not use media instructions.
19773 Use multiply and add/subtract instructions.
19776 @opindex mno-muladd
19778 Do not use multiply and add/subtract instructions.
19783 Select the FDPIC ABI, which uses function descriptors to represent
19784 pointers to functions. Without any PIC/PIE-related options, it
19785 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
19786 assumes GOT entries and small data are within a 12-bit range from the
19787 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
19788 are computed with 32 bits.
19789 With a @samp{bfin-elf} target, this option implies @option{-msim}.
19792 @opindex minline-plt
19794 Enable inlining of PLT entries in function calls to functions that are
19795 not known to bind locally. It has no effect without @option{-mfdpic}.
19796 It's enabled by default if optimizing for speed and compiling for
19797 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
19798 optimization option such as @option{-O3} or above is present in the
19804 Assume a large TLS segment when generating thread-local code.
19809 Do not assume a large TLS segment when generating thread-local code.
19814 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
19815 that is known to be in read-only sections. It's enabled by default,
19816 except for @option{-fpic} or @option{-fpie}: even though it may help
19817 make the global offset table smaller, it trades 1 instruction for 4.
19818 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
19819 one of which may be shared by multiple symbols, and it avoids the need
19820 for a GOT entry for the referenced symbol, so it's more likely to be a
19821 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
19823 @item -multilib-library-pic
19824 @opindex multilib-library-pic
19826 Link with the (library, not FD) pic libraries. It's implied by
19827 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
19828 @option{-fpic} without @option{-mfdpic}. You should never have to use
19832 @opindex mlinked-fp
19834 Follow the EABI requirement of always creating a frame pointer whenever
19835 a stack frame is allocated. This option is enabled by default and can
19836 be disabled with @option{-mno-linked-fp}.
19839 @opindex mlong-calls
19841 Use indirect addressing to call functions outside the current
19842 compilation unit. This allows the functions to be placed anywhere
19843 within the 32-bit address space.
19845 @item -malign-labels
19846 @opindex malign-labels
19848 Try to align labels to an 8-byte boundary by inserting NOPs into the
19849 previous packet. This option only has an effect when VLIW packing
19850 is enabled. It doesn't create new packets; it merely adds NOPs to
19853 @item -mlibrary-pic
19854 @opindex mlibrary-pic
19856 Generate position-independent EABI code.
19861 Use only the first four media accumulator registers.
19866 Use all eight media accumulator registers.
19871 Pack VLIW instructions.
19876 Do not pack VLIW instructions.
19879 @opindex mno-eflags
19881 Do not mark ABI switches in e_flags.
19884 @opindex mcond-move
19886 Enable the use of conditional-move instructions (default).
19888 This switch is mainly for debugging the compiler and will likely be removed
19889 in a future version.
19891 @item -mno-cond-move
19892 @opindex mno-cond-move
19894 Disable the use of conditional-move instructions.
19896 This switch is mainly for debugging the compiler and will likely be removed
19897 in a future version.
19902 Enable the use of conditional set instructions (default).
19904 This switch is mainly for debugging the compiler and will likely be removed
19905 in a future version.
19910 Disable the use of conditional set instructions.
19912 This switch is mainly for debugging the compiler and will likely be removed
19913 in a future version.
19916 @opindex mcond-exec
19918 Enable the use of conditional execution (default).
19920 This switch is mainly for debugging the compiler and will likely be removed
19921 in a future version.
19923 @item -mno-cond-exec
19924 @opindex mno-cond-exec
19926 Disable the use of conditional execution.
19928 This switch is mainly for debugging the compiler and will likely be removed
19929 in a future version.
19931 @item -mvliw-branch
19932 @opindex mvliw-branch
19934 Run a pass to pack branches into VLIW instructions (default).
19936 This switch is mainly for debugging the compiler and will likely be removed
19937 in a future version.
19939 @item -mno-vliw-branch
19940 @opindex mno-vliw-branch
19942 Do not run a pass to pack branches into VLIW instructions.
19944 This switch is mainly for debugging the compiler and will likely be removed
19945 in a future version.
19947 @item -mmulti-cond-exec
19948 @opindex mmulti-cond-exec
19950 Enable optimization of @code{&&} and @code{||} in conditional execution
19953 This switch is mainly for debugging the compiler and will likely be removed
19954 in a future version.
19956 @item -mno-multi-cond-exec
19957 @opindex mno-multi-cond-exec
19959 Disable optimization of @code{&&} and @code{||} in conditional execution.
19961 This switch is mainly for debugging the compiler and will likely be removed
19962 in a future version.
19964 @item -mnested-cond-exec
19965 @opindex mnested-cond-exec
19967 Enable nested conditional execution optimizations (default).
19969 This switch is mainly for debugging the compiler and will likely be removed
19970 in a future version.
19972 @item -mno-nested-cond-exec
19973 @opindex mno-nested-cond-exec
19975 Disable nested conditional execution optimizations.
19977 This switch is mainly for debugging the compiler and will likely be removed
19978 in a future version.
19980 @item -moptimize-membar
19981 @opindex moptimize-membar
19983 This switch removes redundant @code{membar} instructions from the
19984 compiler-generated code. It is enabled by default.
19986 @item -mno-optimize-membar
19987 @opindex mno-optimize-membar
19988 @opindex moptimize-membar
19990 This switch disables the automatic removal of redundant @code{membar}
19991 instructions from the generated code.
19993 @item -mtomcat-stats
19994 @opindex mtomcat-stats
19996 Cause gas to print out tomcat statistics.
19998 @item -mcpu=@var{cpu}
20001 Select the processor type for which to generate code. Possible values are
20002 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
20003 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
20007 @node GNU/Linux Options
20008 @subsection GNU/Linux Options
20010 These @samp{-m} options are defined for GNU/Linux targets:
20015 Use the GNU C library. This is the default except
20016 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
20017 @samp{*-*-linux-*android*} targets.
20021 Use uClibc C library. This is the default on
20022 @samp{*-*-linux-*uclibc*} targets.
20026 Use the musl C library. This is the default on
20027 @samp{*-*-linux-*musl*} targets.
20031 Use Bionic C library. This is the default on
20032 @samp{*-*-linux-*android*} targets.
20036 Compile code compatible with Android platform. This is the default on
20037 @samp{*-*-linux-*android*} targets.
20039 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
20040 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
20041 this option makes the GCC driver pass Android-specific options to the linker.
20042 Finally, this option causes the preprocessor macro @code{__ANDROID__}
20045 @item -tno-android-cc
20046 @opindex tno-android-cc
20047 Disable compilation effects of @option{-mandroid}, i.e., do not enable
20048 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
20049 @option{-fno-rtti} by default.
20051 @item -tno-android-ld
20052 @opindex tno-android-ld
20053 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
20054 linking options to the linker.
20058 @node H8/300 Options
20059 @subsection H8/300 Options
20061 These @samp{-m} options are defined for the H8/300 implementations:
20066 Shorten some address references at link time, when possible; uses the
20067 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
20068 ld, Using ld}, for a fuller description.
20072 Generate code for the H8/300H@.
20076 Generate code for the H8S@.
20080 Generate code for the H8S and H8/300H in the normal mode. This switch
20081 must be used either with @option{-mh} or @option{-ms}.
20085 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
20089 Extended registers are stored on stack before execution of function
20090 with monitor attribute. Default option is @option{-mexr}.
20091 This option is valid only for H8S targets.
20096 Extended registers are not stored on stack before execution of function
20097 with monitor attribute. Default option is @option{-mno-exr}.
20098 This option is valid only for H8S targets.
20102 Make @code{int} data 32 bits by default.
20105 @opindex malign-300
20106 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
20107 The default for the H8/300H and H8S is to align longs and floats on
20109 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
20110 This option has no effect on the H8/300.
20114 @subsection HPPA Options
20115 @cindex HPPA Options
20117 These @samp{-m} options are defined for the HPPA family of computers:
20120 @item -march=@var{architecture-type}
20122 Generate code for the specified architecture. The choices for
20123 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
20124 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
20125 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
20126 architecture option for your machine. Code compiled for lower numbered
20127 architectures runs on higher numbered architectures, but not the
20130 @item -mpa-risc-1-0
20131 @itemx -mpa-risc-1-1
20132 @itemx -mpa-risc-2-0
20133 @opindex mpa-risc-1-0
20134 @opindex mpa-risc-1-1
20135 @opindex mpa-risc-2-0
20136 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
20138 @item -mcaller-copies
20139 @opindex mcaller-copies
20140 The caller copies function arguments passed by hidden reference. This
20141 option should be used with care as it is not compatible with the default
20142 32-bit runtime. However, only aggregates larger than eight bytes are
20143 passed by hidden reference and the option provides better compatibility
20146 @item -mjump-in-delay
20147 @opindex mjump-in-delay
20148 This option is ignored and provided for compatibility purposes only.
20150 @item -mdisable-fpregs
20151 @opindex mdisable-fpregs
20152 Prevent floating-point registers from being used in any manner. This is
20153 necessary for compiling kernels that perform lazy context switching of
20154 floating-point registers. If you use this option and attempt to perform
20155 floating-point operations, the compiler aborts.
20157 @item -mdisable-indexing
20158 @opindex mdisable-indexing
20159 Prevent the compiler from using indexing address modes. This avoids some
20160 rather obscure problems when compiling MIG generated code under MACH@.
20162 @item -mno-space-regs
20163 @opindex mno-space-regs
20164 @opindex mspace-regs
20165 Generate code that assumes the target has no space registers. This allows
20166 GCC to generate faster indirect calls and use unscaled index address modes.
20168 Such code is suitable for level 0 PA systems and kernels.
20170 @item -mfast-indirect-calls
20171 @opindex mfast-indirect-calls
20172 Generate code that assumes calls never cross space boundaries. This
20173 allows GCC to emit code that performs faster indirect calls.
20175 This option does not work in the presence of shared libraries or nested
20178 @item -mfixed-range=@var{register-range}
20179 @opindex mfixed-range
20180 Generate code treating the given register range as fixed registers.
20181 A fixed register is one that the register allocator cannot use. This is
20182 useful when compiling kernel code. A register range is specified as
20183 two registers separated by a dash. Multiple register ranges can be
20184 specified separated by a comma.
20186 @item -mlong-load-store
20187 @opindex mlong-load-store
20188 Generate 3-instruction load and store sequences as sometimes required by
20189 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
20192 @item -mportable-runtime
20193 @opindex mportable-runtime
20194 Use the portable calling conventions proposed by HP for ELF systems.
20198 Enable the use of assembler directives only GAS understands.
20200 @item -mschedule=@var{cpu-type}
20202 Schedule code according to the constraints for the machine type
20203 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
20204 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
20205 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
20206 proper scheduling option for your machine. The default scheduling is
20210 @opindex mlinker-opt
20211 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
20212 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
20213 linkers in which they give bogus error messages when linking some programs.
20216 @opindex msoft-float
20217 Generate output containing library calls for floating point.
20218 @strong{Warning:} the requisite libraries are not available for all HPPA
20219 targets. Normally the facilities of the machine's usual C compiler are
20220 used, but this cannot be done directly in cross-compilation. You must make
20221 your own arrangements to provide suitable library functions for
20224 @option{-msoft-float} changes the calling convention in the output file;
20225 therefore, it is only useful if you compile @emph{all} of a program with
20226 this option. In particular, you need to compile @file{libgcc.a}, the
20227 library that comes with GCC, with @option{-msoft-float} in order for
20232 Generate the predefine, @code{_SIO}, for server IO@. The default is
20233 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
20234 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
20235 options are available under HP-UX and HI-UX@.
20239 Use options specific to GNU @command{ld}.
20240 This passes @option{-shared} to @command{ld} when
20241 building a shared library. It is the default when GCC is configured,
20242 explicitly or implicitly, with the GNU linker. This option does not
20243 affect which @command{ld} is called; it only changes what parameters
20244 are passed to that @command{ld}.
20245 The @command{ld} that is called is determined by the
20246 @option{--with-ld} configure option, GCC's program search path, and
20247 finally by the user's @env{PATH}. The linker used by GCC can be printed
20248 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
20249 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
20253 Use options specific to HP @command{ld}.
20254 This passes @option{-b} to @command{ld} when building
20255 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
20256 links. It is the default when GCC is configured, explicitly or
20257 implicitly, with the HP linker. This option does not affect
20258 which @command{ld} is called; it only changes what parameters are passed to that
20260 The @command{ld} that is called is determined by the @option{--with-ld}
20261 configure option, GCC's program search path, and finally by the user's
20262 @env{PATH}. The linker used by GCC can be printed using @samp{which
20263 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
20264 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
20267 @opindex mno-long-calls
20268 @opindex mlong-calls
20269 Generate code that uses long call sequences. This ensures that a call
20270 is always able to reach linker generated stubs. The default is to generate
20271 long calls only when the distance from the call site to the beginning
20272 of the function or translation unit, as the case may be, exceeds a
20273 predefined limit set by the branch type being used. The limits for
20274 normal calls are 7,600,000 and 240,000 bytes, respectively for the
20275 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
20278 Distances are measured from the beginning of functions when using the
20279 @option{-ffunction-sections} option, or when using the @option{-mgas}
20280 and @option{-mno-portable-runtime} options together under HP-UX with
20283 It is normally not desirable to use this option as it degrades
20284 performance. However, it may be useful in large applications,
20285 particularly when partial linking is used to build the application.
20287 The types of long calls used depends on the capabilities of the
20288 assembler and linker, and the type of code being generated. The
20289 impact on systems that support long absolute calls, and long pic
20290 symbol-difference or pc-relative calls should be relatively small.
20291 However, an indirect call is used on 32-bit ELF systems in pic code
20292 and it is quite long.
20294 @item -munix=@var{unix-std}
20296 Generate compiler predefines and select a startfile for the specified
20297 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
20298 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
20299 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
20300 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
20301 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
20304 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
20305 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
20306 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
20307 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
20308 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
20309 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
20311 It is @emph{important} to note that this option changes the interfaces
20312 for various library routines. It also affects the operational behavior
20313 of the C library. Thus, @emph{extreme} care is needed in using this
20316 Library code that is intended to operate with more than one UNIX
20317 standard must test, set and restore the variable @code{__xpg4_extended_mask}
20318 as appropriate. Most GNU software doesn't provide this capability.
20322 Suppress the generation of link options to search libdld.sl when the
20323 @option{-static} option is specified on HP-UX 10 and later.
20327 The HP-UX implementation of setlocale in libc has a dependency on
20328 libdld.sl. There isn't an archive version of libdld.sl. Thus,
20329 when the @option{-static} option is specified, special link options
20330 are needed to resolve this dependency.
20332 On HP-UX 10 and later, the GCC driver adds the necessary options to
20333 link with libdld.sl when the @option{-static} option is specified.
20334 This causes the resulting binary to be dynamic. On the 64-bit port,
20335 the linkers generate dynamic binaries by default in any case. The
20336 @option{-nolibdld} option can be used to prevent the GCC driver from
20337 adding these link options.
20341 Add support for multithreading with the @dfn{dce thread} library
20342 under HP-UX@. This option sets flags for both the preprocessor and
20346 @node IA-64 Options
20347 @subsection IA-64 Options
20348 @cindex IA-64 Options
20350 These are the @samp{-m} options defined for the Intel IA-64 architecture.
20354 @opindex mbig-endian
20355 Generate code for a big-endian target. This is the default for HP-UX@.
20357 @item -mlittle-endian
20358 @opindex mlittle-endian
20359 Generate code for a little-endian target. This is the default for AIX5
20365 @opindex mno-gnu-as
20366 Generate (or don't) code for the GNU assembler. This is the default.
20367 @c Also, this is the default if the configure option @option{--with-gnu-as}
20373 @opindex mno-gnu-ld
20374 Generate (or don't) code for the GNU linker. This is the default.
20375 @c Also, this is the default if the configure option @option{--with-gnu-ld}
20380 Generate code that does not use a global pointer register. The result
20381 is not position independent code, and violates the IA-64 ABI@.
20383 @item -mvolatile-asm-stop
20384 @itemx -mno-volatile-asm-stop
20385 @opindex mvolatile-asm-stop
20386 @opindex mno-volatile-asm-stop
20387 Generate (or don't) a stop bit immediately before and after volatile asm
20390 @item -mregister-names
20391 @itemx -mno-register-names
20392 @opindex mregister-names
20393 @opindex mno-register-names
20394 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
20395 the stacked registers. This may make assembler output more readable.
20401 Disable (or enable) optimizations that use the small data section. This may
20402 be useful for working around optimizer bugs.
20404 @item -mconstant-gp
20405 @opindex mconstant-gp
20406 Generate code that uses a single constant global pointer value. This is
20407 useful when compiling kernel code.
20411 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
20412 This is useful when compiling firmware code.
20414 @item -minline-float-divide-min-latency
20415 @opindex minline-float-divide-min-latency
20416 Generate code for inline divides of floating-point values
20417 using the minimum latency algorithm.
20419 @item -minline-float-divide-max-throughput
20420 @opindex minline-float-divide-max-throughput
20421 Generate code for inline divides of floating-point values
20422 using the maximum throughput algorithm.
20424 @item -mno-inline-float-divide
20425 @opindex mno-inline-float-divide
20426 Do not generate inline code for divides of floating-point values.
20428 @item -minline-int-divide-min-latency
20429 @opindex minline-int-divide-min-latency
20430 Generate code for inline divides of integer values
20431 using the minimum latency algorithm.
20433 @item -minline-int-divide-max-throughput
20434 @opindex minline-int-divide-max-throughput
20435 Generate code for inline divides of integer values
20436 using the maximum throughput algorithm.
20438 @item -mno-inline-int-divide
20439 @opindex mno-inline-int-divide
20440 @opindex minline-int-divide
20441 Do not generate inline code for divides of integer values.
20443 @item -minline-sqrt-min-latency
20444 @opindex minline-sqrt-min-latency
20445 Generate code for inline square roots
20446 using the minimum latency algorithm.
20448 @item -minline-sqrt-max-throughput
20449 @opindex minline-sqrt-max-throughput
20450 Generate code for inline square roots
20451 using the maximum throughput algorithm.
20453 @item -mno-inline-sqrt
20454 @opindex mno-inline-sqrt
20455 Do not generate inline code for @code{sqrt}.
20458 @itemx -mno-fused-madd
20459 @opindex mfused-madd
20460 @opindex mno-fused-madd
20461 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
20462 instructions. The default is to use these instructions.
20464 @item -mno-dwarf2-asm
20465 @itemx -mdwarf2-asm
20466 @opindex mno-dwarf2-asm
20467 @opindex mdwarf2-asm
20468 Don't (or do) generate assembler code for the DWARF line number debugging
20469 info. This may be useful when not using the GNU assembler.
20471 @item -mearly-stop-bits
20472 @itemx -mno-early-stop-bits
20473 @opindex mearly-stop-bits
20474 @opindex mno-early-stop-bits
20475 Allow stop bits to be placed earlier than immediately preceding the
20476 instruction that triggered the stop bit. This can improve instruction
20477 scheduling, but does not always do so.
20479 @item -mfixed-range=@var{register-range}
20480 @opindex mfixed-range
20481 Generate code treating the given register range as fixed registers.
20482 A fixed register is one that the register allocator cannot use. This is
20483 useful when compiling kernel code. A register range is specified as
20484 two registers separated by a dash. Multiple register ranges can be
20485 specified separated by a comma.
20487 @item -mtls-size=@var{tls-size}
20489 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
20492 @item -mtune=@var{cpu-type}
20494 Tune the instruction scheduling for a particular CPU, Valid values are
20495 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
20496 and @samp{mckinley}.
20502 Generate code for a 32-bit or 64-bit environment.
20503 The 32-bit environment sets int, long and pointer to 32 bits.
20504 The 64-bit environment sets int to 32 bits and long and pointer
20505 to 64 bits. These are HP-UX specific flags.
20507 @item -mno-sched-br-data-spec
20508 @itemx -msched-br-data-spec
20509 @opindex mno-sched-br-data-spec
20510 @opindex msched-br-data-spec
20511 (Dis/En)able data speculative scheduling before reload.
20512 This results in generation of @code{ld.a} instructions and
20513 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
20514 The default setting is disabled.
20516 @item -msched-ar-data-spec
20517 @itemx -mno-sched-ar-data-spec
20518 @opindex msched-ar-data-spec
20519 @opindex mno-sched-ar-data-spec
20520 (En/Dis)able data speculative scheduling after reload.
20521 This results in generation of @code{ld.a} instructions and
20522 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
20523 The default setting is enabled.
20525 @item -mno-sched-control-spec
20526 @itemx -msched-control-spec
20527 @opindex mno-sched-control-spec
20528 @opindex msched-control-spec
20529 (Dis/En)able control speculative scheduling. This feature is
20530 available only during region scheduling (i.e.@: before reload).
20531 This results in generation of the @code{ld.s} instructions and
20532 the corresponding check instructions @code{chk.s}.
20533 The default setting is disabled.
20535 @item -msched-br-in-data-spec
20536 @itemx -mno-sched-br-in-data-spec
20537 @opindex msched-br-in-data-spec
20538 @opindex mno-sched-br-in-data-spec
20539 (En/Dis)able speculative scheduling of the instructions that
20540 are dependent on the data speculative loads before reload.
20541 This is effective only with @option{-msched-br-data-spec} enabled.
20542 The default setting is enabled.
20544 @item -msched-ar-in-data-spec
20545 @itemx -mno-sched-ar-in-data-spec
20546 @opindex msched-ar-in-data-spec
20547 @opindex mno-sched-ar-in-data-spec
20548 (En/Dis)able speculative scheduling of the instructions that
20549 are dependent on the data speculative loads after reload.
20550 This is effective only with @option{-msched-ar-data-spec} enabled.
20551 The default setting is enabled.
20553 @item -msched-in-control-spec
20554 @itemx -mno-sched-in-control-spec
20555 @opindex msched-in-control-spec
20556 @opindex mno-sched-in-control-spec
20557 (En/Dis)able speculative scheduling of the instructions that
20558 are dependent on the control speculative loads.
20559 This is effective only with @option{-msched-control-spec} enabled.
20560 The default setting is enabled.
20562 @item -mno-sched-prefer-non-data-spec-insns
20563 @itemx -msched-prefer-non-data-spec-insns
20564 @opindex mno-sched-prefer-non-data-spec-insns
20565 @opindex msched-prefer-non-data-spec-insns
20566 If enabled, data-speculative instructions are chosen for schedule
20567 only if there are no other choices at the moment. This makes
20568 the use of the data speculation much more conservative.
20569 The default setting is disabled.
20571 @item -mno-sched-prefer-non-control-spec-insns
20572 @itemx -msched-prefer-non-control-spec-insns
20573 @opindex mno-sched-prefer-non-control-spec-insns
20574 @opindex msched-prefer-non-control-spec-insns
20575 If enabled, control-speculative instructions are chosen for schedule
20576 only if there are no other choices at the moment. This makes
20577 the use of the control speculation much more conservative.
20578 The default setting is disabled.
20580 @item -mno-sched-count-spec-in-critical-path
20581 @itemx -msched-count-spec-in-critical-path
20582 @opindex mno-sched-count-spec-in-critical-path
20583 @opindex msched-count-spec-in-critical-path
20584 If enabled, speculative dependencies are considered during
20585 computation of the instructions priorities. This makes the use of the
20586 speculation a bit more conservative.
20587 The default setting is disabled.
20589 @item -msched-spec-ldc
20590 @opindex msched-spec-ldc
20591 Use a simple data speculation check. This option is on by default.
20593 @item -msched-control-spec-ldc
20594 @opindex msched-spec-ldc
20595 Use a simple check for control speculation. This option is on by default.
20597 @item -msched-stop-bits-after-every-cycle
20598 @opindex msched-stop-bits-after-every-cycle
20599 Place a stop bit after every cycle when scheduling. This option is on
20602 @item -msched-fp-mem-deps-zero-cost
20603 @opindex msched-fp-mem-deps-zero-cost
20604 Assume that floating-point stores and loads are not likely to cause a conflict
20605 when placed into the same instruction group. This option is disabled by
20608 @item -msel-sched-dont-check-control-spec
20609 @opindex msel-sched-dont-check-control-spec
20610 Generate checks for control speculation in selective scheduling.
20611 This flag is disabled by default.
20613 @item -msched-max-memory-insns=@var{max-insns}
20614 @opindex msched-max-memory-insns
20615 Limit on the number of memory insns per instruction group, giving lower
20616 priority to subsequent memory insns attempting to schedule in the same
20617 instruction group. Frequently useful to prevent cache bank conflicts.
20618 The default value is 1.
20620 @item -msched-max-memory-insns-hard-limit
20621 @opindex msched-max-memory-insns-hard-limit
20622 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
20623 disallowing more than that number in an instruction group.
20624 Otherwise, the limit is ``soft'', meaning that non-memory operations
20625 are preferred when the limit is reached, but memory operations may still
20631 @subsection LM32 Options
20632 @cindex LM32 options
20634 These @option{-m} options are defined for the LatticeMico32 architecture:
20637 @item -mbarrel-shift-enabled
20638 @opindex mbarrel-shift-enabled
20639 Enable barrel-shift instructions.
20641 @item -mdivide-enabled
20642 @opindex mdivide-enabled
20643 Enable divide and modulus instructions.
20645 @item -mmultiply-enabled
20646 @opindex multiply-enabled
20647 Enable multiply instructions.
20649 @item -msign-extend-enabled
20650 @opindex msign-extend-enabled
20651 Enable sign extend instructions.
20653 @item -muser-enabled
20654 @opindex muser-enabled
20655 Enable user-defined instructions.
20660 @subsection M32C Options
20661 @cindex M32C options
20664 @item -mcpu=@var{name}
20666 Select the CPU for which code is generated. @var{name} may be one of
20667 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
20668 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
20669 the M32C/80 series.
20673 Specifies that the program will be run on the simulator. This causes
20674 an alternate runtime library to be linked in which supports, for
20675 example, file I/O@. You must not use this option when generating
20676 programs that will run on real hardware; you must provide your own
20677 runtime library for whatever I/O functions are needed.
20679 @item -memregs=@var{number}
20681 Specifies the number of memory-based pseudo-registers GCC uses
20682 during code generation. These pseudo-registers are used like real
20683 registers, so there is a tradeoff between GCC's ability to fit the
20684 code into available registers, and the performance penalty of using
20685 memory instead of registers. Note that all modules in a program must
20686 be compiled with the same value for this option. Because of that, you
20687 must not use this option with GCC's default runtime libraries.
20691 @node M32R/D Options
20692 @subsection M32R/D Options
20693 @cindex M32R/D options
20695 These @option{-m} options are defined for Renesas M32R/D architectures:
20700 Generate code for the M32R/2@.
20704 Generate code for the M32R/X@.
20708 Generate code for the M32R@. This is the default.
20710 @item -mmodel=small
20711 @opindex mmodel=small
20712 Assume all objects live in the lower 16MB of memory (so that their addresses
20713 can be loaded with the @code{ld24} instruction), and assume all subroutines
20714 are reachable with the @code{bl} instruction.
20715 This is the default.
20717 The addressability of a particular object can be set with the
20718 @code{model} attribute.
20720 @item -mmodel=medium
20721 @opindex mmodel=medium
20722 Assume objects may be anywhere in the 32-bit address space (the compiler
20723 generates @code{seth/add3} instructions to load their addresses), and
20724 assume all subroutines are reachable with the @code{bl} instruction.
20726 @item -mmodel=large
20727 @opindex mmodel=large
20728 Assume objects may be anywhere in the 32-bit address space (the compiler
20729 generates @code{seth/add3} instructions to load their addresses), and
20730 assume subroutines may not be reachable with the @code{bl} instruction
20731 (the compiler generates the much slower @code{seth/add3/jl}
20732 instruction sequence).
20735 @opindex msdata=none
20736 Disable use of the small data area. Variables are put into
20737 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
20738 @code{section} attribute has been specified).
20739 This is the default.
20741 The small data area consists of sections @code{.sdata} and @code{.sbss}.
20742 Objects may be explicitly put in the small data area with the
20743 @code{section} attribute using one of these sections.
20745 @item -msdata=sdata
20746 @opindex msdata=sdata
20747 Put small global and static data in the small data area, but do not
20748 generate special code to reference them.
20751 @opindex msdata=use
20752 Put small global and static data in the small data area, and generate
20753 special instructions to reference them.
20757 @cindex smaller data references
20758 Put global and static objects less than or equal to @var{num} bytes
20759 into the small data or BSS sections instead of the normal data or BSS
20760 sections. The default value of @var{num} is 8.
20761 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
20762 for this option to have any effect.
20764 All modules should be compiled with the same @option{-G @var{num}} value.
20765 Compiling with different values of @var{num} may or may not work; if it
20766 doesn't the linker gives an error message---incorrect code is not
20771 Makes the M32R-specific code in the compiler display some statistics
20772 that might help in debugging programs.
20774 @item -malign-loops
20775 @opindex malign-loops
20776 Align all loops to a 32-byte boundary.
20778 @item -mno-align-loops
20779 @opindex mno-align-loops
20780 Do not enforce a 32-byte alignment for loops. This is the default.
20782 @item -missue-rate=@var{number}
20783 @opindex missue-rate=@var{number}
20784 Issue @var{number} instructions per cycle. @var{number} can only be 1
20787 @item -mbranch-cost=@var{number}
20788 @opindex mbranch-cost=@var{number}
20789 @var{number} can only be 1 or 2. If it is 1 then branches are
20790 preferred over conditional code, if it is 2, then the opposite applies.
20792 @item -mflush-trap=@var{number}
20793 @opindex mflush-trap=@var{number}
20794 Specifies the trap number to use to flush the cache. The default is
20795 12. Valid numbers are between 0 and 15 inclusive.
20797 @item -mno-flush-trap
20798 @opindex mno-flush-trap
20799 Specifies that the cache cannot be flushed by using a trap.
20801 @item -mflush-func=@var{name}
20802 @opindex mflush-func=@var{name}
20803 Specifies the name of the operating system function to call to flush
20804 the cache. The default is @samp{_flush_cache}, but a function call
20805 is only used if a trap is not available.
20807 @item -mno-flush-func
20808 @opindex mno-flush-func
20809 Indicates that there is no OS function for flushing the cache.
20813 @node M680x0 Options
20814 @subsection M680x0 Options
20815 @cindex M680x0 options
20817 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
20818 The default settings depend on which architecture was selected when
20819 the compiler was configured; the defaults for the most common choices
20823 @item -march=@var{arch}
20825 Generate code for a specific M680x0 or ColdFire instruction set
20826 architecture. Permissible values of @var{arch} for M680x0
20827 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
20828 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
20829 architectures are selected according to Freescale's ISA classification
20830 and the permissible values are: @samp{isaa}, @samp{isaaplus},
20831 @samp{isab} and @samp{isac}.
20833 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
20834 code for a ColdFire target. The @var{arch} in this macro is one of the
20835 @option{-march} arguments given above.
20837 When used together, @option{-march} and @option{-mtune} select code
20838 that runs on a family of similar processors but that is optimized
20839 for a particular microarchitecture.
20841 @item -mcpu=@var{cpu}
20843 Generate code for a specific M680x0 or ColdFire processor.
20844 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
20845 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
20846 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
20847 below, which also classifies the CPUs into families:
20849 @multitable @columnfractions 0.20 0.80
20850 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
20851 @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}
20852 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
20853 @item @samp{5206e} @tab @samp{5206e}
20854 @item @samp{5208} @tab @samp{5207} @samp{5208}
20855 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
20856 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
20857 @item @samp{5216} @tab @samp{5214} @samp{5216}
20858 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
20859 @item @samp{5225} @tab @samp{5224} @samp{5225}
20860 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
20861 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
20862 @item @samp{5249} @tab @samp{5249}
20863 @item @samp{5250} @tab @samp{5250}
20864 @item @samp{5271} @tab @samp{5270} @samp{5271}
20865 @item @samp{5272} @tab @samp{5272}
20866 @item @samp{5275} @tab @samp{5274} @samp{5275}
20867 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
20868 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
20869 @item @samp{5307} @tab @samp{5307}
20870 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
20871 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
20872 @item @samp{5407} @tab @samp{5407}
20873 @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}
20876 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
20877 @var{arch} is compatible with @var{cpu}. Other combinations of
20878 @option{-mcpu} and @option{-march} are rejected.
20880 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
20881 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
20882 where the value of @var{family} is given by the table above.
20884 @item -mtune=@var{tune}
20886 Tune the code for a particular microarchitecture within the
20887 constraints set by @option{-march} and @option{-mcpu}.
20888 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
20889 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
20890 and @samp{cpu32}. The ColdFire microarchitectures
20891 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
20893 You can also use @option{-mtune=68020-40} for code that needs
20894 to run relatively well on 68020, 68030 and 68040 targets.
20895 @option{-mtune=68020-60} is similar but includes 68060 targets
20896 as well. These two options select the same tuning decisions as
20897 @option{-m68020-40} and @option{-m68020-60} respectively.
20899 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
20900 when tuning for 680x0 architecture @var{arch}. It also defines
20901 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
20902 option is used. If GCC is tuning for a range of architectures,
20903 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
20904 it defines the macros for every architecture in the range.
20906 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
20907 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
20908 of the arguments given above.
20914 Generate output for a 68000. This is the default
20915 when the compiler is configured for 68000-based systems.
20916 It is equivalent to @option{-march=68000}.
20918 Use this option for microcontrollers with a 68000 or EC000 core,
20919 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
20923 Generate output for a 68010. This is the default
20924 when the compiler is configured for 68010-based systems.
20925 It is equivalent to @option{-march=68010}.
20931 Generate output for a 68020. This is the default
20932 when the compiler is configured for 68020-based systems.
20933 It is equivalent to @option{-march=68020}.
20937 Generate output for a 68030. This is the default when the compiler is
20938 configured for 68030-based systems. It is equivalent to
20939 @option{-march=68030}.
20943 Generate output for a 68040. This is the default when the compiler is
20944 configured for 68040-based systems. It is equivalent to
20945 @option{-march=68040}.
20947 This option inhibits the use of 68881/68882 instructions that have to be
20948 emulated by software on the 68040. Use this option if your 68040 does not
20949 have code to emulate those instructions.
20953 Generate output for a 68060. This is the default when the compiler is
20954 configured for 68060-based systems. It is equivalent to
20955 @option{-march=68060}.
20957 This option inhibits the use of 68020 and 68881/68882 instructions that
20958 have to be emulated by software on the 68060. Use this option if your 68060
20959 does not have code to emulate those instructions.
20963 Generate output for a CPU32. This is the default
20964 when the compiler is configured for CPU32-based systems.
20965 It is equivalent to @option{-march=cpu32}.
20967 Use this option for microcontrollers with a
20968 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
20969 68336, 68340, 68341, 68349 and 68360.
20973 Generate output for a 520X ColdFire CPU@. This is the default
20974 when the compiler is configured for 520X-based systems.
20975 It is equivalent to @option{-mcpu=5206}, and is now deprecated
20976 in favor of that option.
20978 Use this option for microcontroller with a 5200 core, including
20979 the MCF5202, MCF5203, MCF5204 and MCF5206.
20983 Generate output for a 5206e ColdFire CPU@. The option is now
20984 deprecated in favor of the equivalent @option{-mcpu=5206e}.
20988 Generate output for a member of the ColdFire 528X family.
20989 The option is now deprecated in favor of the equivalent
20990 @option{-mcpu=528x}.
20994 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
20995 in favor of the equivalent @option{-mcpu=5307}.
20999 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
21000 in favor of the equivalent @option{-mcpu=5407}.
21004 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
21005 This includes use of hardware floating-point instructions.
21006 The option is equivalent to @option{-mcpu=547x}, and is now
21007 deprecated in favor of that option.
21011 Generate output for a 68040, without using any of the new instructions.
21012 This results in code that can run relatively efficiently on either a
21013 68020/68881 or a 68030 or a 68040. The generated code does use the
21014 68881 instructions that are emulated on the 68040.
21016 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
21020 Generate output for a 68060, without using any of the new instructions.
21021 This results in code that can run relatively efficiently on either a
21022 68020/68881 or a 68030 or a 68040. The generated code does use the
21023 68881 instructions that are emulated on the 68060.
21025 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
21029 @opindex mhard-float
21031 Generate floating-point instructions. This is the default for 68020
21032 and above, and for ColdFire devices that have an FPU@. It defines the
21033 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
21034 on ColdFire targets.
21037 @opindex msoft-float
21038 Do not generate floating-point instructions; use library calls instead.
21039 This is the default for 68000, 68010, and 68832 targets. It is also
21040 the default for ColdFire devices that have no FPU.
21046 Generate (do not generate) ColdFire hardware divide and remainder
21047 instructions. If @option{-march} is used without @option{-mcpu},
21048 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
21049 architectures. Otherwise, the default is taken from the target CPU
21050 (either the default CPU, or the one specified by @option{-mcpu}). For
21051 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
21052 @option{-mcpu=5206e}.
21054 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
21058 Consider type @code{int} to be 16 bits wide, like @code{short int}.
21059 Additionally, parameters passed on the stack are also aligned to a
21060 16-bit boundary even on targets whose API mandates promotion to 32-bit.
21064 Do not consider type @code{int} to be 16 bits wide. This is the default.
21067 @itemx -mno-bitfield
21068 @opindex mnobitfield
21069 @opindex mno-bitfield
21070 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
21071 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
21075 Do use the bit-field instructions. The @option{-m68020} option implies
21076 @option{-mbitfield}. This is the default if you use a configuration
21077 designed for a 68020.
21081 Use a different function-calling convention, in which functions
21082 that take a fixed number of arguments return with the @code{rtd}
21083 instruction, which pops their arguments while returning. This
21084 saves one instruction in the caller since there is no need to pop
21085 the arguments there.
21087 This calling convention is incompatible with the one normally
21088 used on Unix, so you cannot use it if you need to call libraries
21089 compiled with the Unix compiler.
21091 Also, you must provide function prototypes for all functions that
21092 take variable numbers of arguments (including @code{printf});
21093 otherwise incorrect code is generated for calls to those
21096 In addition, seriously incorrect code results if you call a
21097 function with too many arguments. (Normally, extra arguments are
21098 harmlessly ignored.)
21100 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
21101 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
21103 The default is @option{-mno-rtd}.
21106 @itemx -mno-align-int
21107 @opindex malign-int
21108 @opindex mno-align-int
21109 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
21110 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
21111 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
21112 Aligning variables on 32-bit boundaries produces code that runs somewhat
21113 faster on processors with 32-bit busses at the expense of more memory.
21115 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
21116 aligns structures containing the above types differently than
21117 most published application binary interface specifications for the m68k.
21121 Use the pc-relative addressing mode of the 68000 directly, instead of
21122 using a global offset table. At present, this option implies @option{-fpic},
21123 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
21124 not presently supported with @option{-mpcrel}, though this could be supported for
21125 68020 and higher processors.
21127 @item -mno-strict-align
21128 @itemx -mstrict-align
21129 @opindex mno-strict-align
21130 @opindex mstrict-align
21131 Do not (do) assume that unaligned memory references are handled by
21135 Generate code that allows the data segment to be located in a different
21136 area of memory from the text segment. This allows for execute-in-place in
21137 an environment without virtual memory management. This option implies
21140 @item -mno-sep-data
21141 Generate code that assumes that the data segment follows the text segment.
21142 This is the default.
21144 @item -mid-shared-library
21145 Generate code that supports shared libraries via the library ID method.
21146 This allows for execute-in-place and shared libraries in an environment
21147 without virtual memory management. This option implies @option{-fPIC}.
21149 @item -mno-id-shared-library
21150 Generate code that doesn't assume ID-based shared libraries are being used.
21151 This is the default.
21153 @item -mshared-library-id=n
21154 Specifies the identification number of the ID-based shared library being
21155 compiled. Specifying a value of 0 generates more compact code; specifying
21156 other values forces the allocation of that number to the current
21157 library, but is no more space- or time-efficient than omitting this option.
21163 When generating position-independent code for ColdFire, generate code
21164 that works if the GOT has more than 8192 entries. This code is
21165 larger and slower than code generated without this option. On M680x0
21166 processors, this option is not needed; @option{-fPIC} suffices.
21168 GCC normally uses a single instruction to load values from the GOT@.
21169 While this is relatively efficient, it only works if the GOT
21170 is smaller than about 64k. Anything larger causes the linker
21171 to report an error such as:
21173 @cindex relocation truncated to fit (ColdFire)
21175 relocation truncated to fit: R_68K_GOT16O foobar
21178 If this happens, you should recompile your code with @option{-mxgot}.
21179 It should then work with very large GOTs. However, code generated with
21180 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
21181 the value of a global symbol.
21183 Note that some linkers, including newer versions of the GNU linker,
21184 can create multiple GOTs and sort GOT entries. If you have such a linker,
21185 you should only need to use @option{-mxgot} when compiling a single
21186 object file that accesses more than 8192 GOT entries. Very few do.
21188 These options have no effect unless GCC is generating
21189 position-independent code.
21191 @item -mlong-jump-table-offsets
21192 @opindex mlong-jump-table-offsets
21193 Use 32-bit offsets in @code{switch} tables. The default is to use
21198 @node MCore Options
21199 @subsection MCore Options
21200 @cindex MCore options
21202 These are the @samp{-m} options defined for the Motorola M*Core
21208 @itemx -mno-hardlit
21210 @opindex mno-hardlit
21211 Inline constants into the code stream if it can be done in two
21212 instructions or less.
21218 Use the divide instruction. (Enabled by default).
21220 @item -mrelax-immediate
21221 @itemx -mno-relax-immediate
21222 @opindex mrelax-immediate
21223 @opindex mno-relax-immediate
21224 Allow arbitrary-sized immediates in bit operations.
21226 @item -mwide-bitfields
21227 @itemx -mno-wide-bitfields
21228 @opindex mwide-bitfields
21229 @opindex mno-wide-bitfields
21230 Always treat bit-fields as @code{int}-sized.
21232 @item -m4byte-functions
21233 @itemx -mno-4byte-functions
21234 @opindex m4byte-functions
21235 @opindex mno-4byte-functions
21236 Force all functions to be aligned to a 4-byte boundary.
21238 @item -mcallgraph-data
21239 @itemx -mno-callgraph-data
21240 @opindex mcallgraph-data
21241 @opindex mno-callgraph-data
21242 Emit callgraph information.
21245 @itemx -mno-slow-bytes
21246 @opindex mslow-bytes
21247 @opindex mno-slow-bytes
21248 Prefer word access when reading byte quantities.
21250 @item -mlittle-endian
21251 @itemx -mbig-endian
21252 @opindex mlittle-endian
21253 @opindex mbig-endian
21254 Generate code for a little-endian target.
21260 Generate code for the 210 processor.
21264 Assume that runtime support has been provided and so omit the
21265 simulator library (@file{libsim.a)} from the linker command line.
21267 @item -mstack-increment=@var{size}
21268 @opindex mstack-increment
21269 Set the maximum amount for a single stack increment operation. Large
21270 values can increase the speed of programs that contain functions
21271 that need a large amount of stack space, but they can also trigger a
21272 segmentation fault if the stack is extended too much. The default
21278 @subsection MeP Options
21279 @cindex MeP options
21285 Enables the @code{abs} instruction, which is the absolute difference
21286 between two registers.
21290 Enables all the optional instructions---average, multiply, divide, bit
21291 operations, leading zero, absolute difference, min/max, clip, and
21297 Enables the @code{ave} instruction, which computes the average of two
21300 @item -mbased=@var{n}
21302 Variables of size @var{n} bytes or smaller are placed in the
21303 @code{.based} section by default. Based variables use the @code{$tp}
21304 register as a base register, and there is a 128-byte limit to the
21305 @code{.based} section.
21309 Enables the bit operation instructions---bit test (@code{btstm}), set
21310 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
21311 test-and-set (@code{tas}).
21313 @item -mc=@var{name}
21315 Selects which section constant data is placed in. @var{name} may
21316 be @samp{tiny}, @samp{near}, or @samp{far}.
21320 Enables the @code{clip} instruction. Note that @option{-mclip} is not
21321 useful unless you also provide @option{-mminmax}.
21323 @item -mconfig=@var{name}
21325 Selects one of the built-in core configurations. Each MeP chip has
21326 one or more modules in it; each module has a core CPU and a variety of
21327 coprocessors, optional instructions, and peripherals. The
21328 @code{MeP-Integrator} tool, not part of GCC, provides these
21329 configurations through this option; using this option is the same as
21330 using all the corresponding command-line options. The default
21331 configuration is @samp{default}.
21335 Enables the coprocessor instructions. By default, this is a 32-bit
21336 coprocessor. Note that the coprocessor is normally enabled via the
21337 @option{-mconfig=} option.
21341 Enables the 32-bit coprocessor's instructions.
21345 Enables the 64-bit coprocessor's instructions.
21349 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
21353 Causes constant variables to be placed in the @code{.near} section.
21357 Enables the @code{div} and @code{divu} instructions.
21361 Generate big-endian code.
21365 Generate little-endian code.
21367 @item -mio-volatile
21368 @opindex mio-volatile
21369 Tells the compiler that any variable marked with the @code{io}
21370 attribute is to be considered volatile.
21374 Causes variables to be assigned to the @code{.far} section by default.
21378 Enables the @code{leadz} (leading zero) instruction.
21382 Causes variables to be assigned to the @code{.near} section by default.
21386 Enables the @code{min} and @code{max} instructions.
21390 Enables the multiplication and multiply-accumulate instructions.
21394 Disables all the optional instructions enabled by @option{-mall-opts}.
21398 Enables the @code{repeat} and @code{erepeat} instructions, used for
21399 low-overhead looping.
21403 Causes all variables to default to the @code{.tiny} section. Note
21404 that there is a 65536-byte limit to this section. Accesses to these
21405 variables use the @code{%gp} base register.
21409 Enables the saturation instructions. Note that the compiler does not
21410 currently generate these itself, but this option is included for
21411 compatibility with other tools, like @code{as}.
21415 Link the SDRAM-based runtime instead of the default ROM-based runtime.
21419 Link the simulator run-time libraries.
21423 Link the simulator runtime libraries, excluding built-in support
21424 for reset and exception vectors and tables.
21428 Causes all functions to default to the @code{.far} section. Without
21429 this option, functions default to the @code{.near} section.
21431 @item -mtiny=@var{n}
21433 Variables that are @var{n} bytes or smaller are allocated to the
21434 @code{.tiny} section. These variables use the @code{$gp} base
21435 register. The default for this option is 4, but note that there's a
21436 65536-byte limit to the @code{.tiny} section.
21440 @node MicroBlaze Options
21441 @subsection MicroBlaze Options
21442 @cindex MicroBlaze Options
21447 @opindex msoft-float
21448 Use software emulation for floating point (default).
21451 @opindex mhard-float
21452 Use hardware floating-point instructions.
21456 Do not optimize block moves, use @code{memcpy}.
21458 @item -mno-clearbss
21459 @opindex mno-clearbss
21460 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
21462 @item -mcpu=@var{cpu-type}
21464 Use features of, and schedule code for, the given CPU.
21465 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
21466 where @var{X} is a major version, @var{YY} is the minor version, and
21467 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
21468 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v6.00.a}.
21470 @item -mxl-soft-mul
21471 @opindex mxl-soft-mul
21472 Use software multiply emulation (default).
21474 @item -mxl-soft-div
21475 @opindex mxl-soft-div
21476 Use software emulation for divides (default).
21478 @item -mxl-barrel-shift
21479 @opindex mxl-barrel-shift
21480 Use the hardware barrel shifter.
21482 @item -mxl-pattern-compare
21483 @opindex mxl-pattern-compare
21484 Use pattern compare instructions.
21486 @item -msmall-divides
21487 @opindex msmall-divides
21488 Use table lookup optimization for small signed integer divisions.
21490 @item -mxl-stack-check
21491 @opindex mxl-stack-check
21492 This option is deprecated. Use @option{-fstack-check} instead.
21495 @opindex mxl-gp-opt
21496 Use GP-relative @code{.sdata}/@code{.sbss} sections.
21498 @item -mxl-multiply-high
21499 @opindex mxl-multiply-high
21500 Use multiply high instructions for high part of 32x32 multiply.
21502 @item -mxl-float-convert
21503 @opindex mxl-float-convert
21504 Use hardware floating-point conversion instructions.
21506 @item -mxl-float-sqrt
21507 @opindex mxl-float-sqrt
21508 Use hardware floating-point square root instruction.
21511 @opindex mbig-endian
21512 Generate code for a big-endian target.
21514 @item -mlittle-endian
21515 @opindex mlittle-endian
21516 Generate code for a little-endian target.
21519 @opindex mxl-reorder
21520 Use reorder instructions (swap and byte reversed load/store).
21522 @item -mxl-mode-@var{app-model}
21523 Select application model @var{app-model}. Valid models are
21526 normal executable (default), uses startup code @file{crt0.o}.
21528 @item -mpic-data-is-text-relative
21529 @opindex mpic-data-is-text-relative
21530 Assume that the displacement between the text and data segments is fixed
21531 at static link time. This allows data to be referenced by offset from start of
21532 text address instead of GOT since PC-relative addressing is not supported.
21535 for use with Xilinx Microprocessor Debugger (XMD) based
21536 software intrusive debug agent called xmdstub. This uses startup file
21537 @file{crt1.o} and sets the start address of the program to 0x800.
21540 for applications that are loaded using a bootloader.
21541 This model uses startup file @file{crt2.o} which does not contain a processor
21542 reset vector handler. This is suitable for transferring control on a
21543 processor reset to the bootloader rather than the application.
21546 for applications that do not require any of the
21547 MicroBlaze vectors. This option may be useful for applications running
21548 within a monitoring application. This model uses @file{crt3.o} as a startup file.
21551 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
21552 @option{-mxl-mode-@var{app-model}}.
21557 @subsection MIPS Options
21558 @cindex MIPS options
21564 Generate big-endian code.
21568 Generate little-endian code. This is the default for @samp{mips*el-*-*}
21571 @item -march=@var{arch}
21573 Generate code that runs on @var{arch}, which can be the name of a
21574 generic MIPS ISA, or the name of a particular processor.
21576 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
21577 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
21578 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
21579 @samp{mips64r5} and @samp{mips64r6}.
21580 The processor names are:
21581 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
21582 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
21583 @samp{5kc}, @samp{5kf},
21585 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
21586 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
21587 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
21588 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
21589 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
21590 @samp{i6400}, @samp{i6500},
21592 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a}, @samp{gs464},
21593 @samp{gs464e}, @samp{gs264e},
21595 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
21596 @samp{m5100}, @samp{m5101},
21597 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
21599 @samp{p5600}, @samp{p6600},
21600 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
21601 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r5900},
21602 @samp{r6000}, @samp{r8000},
21603 @samp{rm7000}, @samp{rm9000},
21604 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
21607 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
21608 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
21609 @samp{xlr} and @samp{xlp}.
21610 The special value @samp{from-abi} selects the
21611 most compatible architecture for the selected ABI (that is,
21612 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
21614 The native Linux/GNU toolchain also supports the value @samp{native},
21615 which selects the best architecture option for the host processor.
21616 @option{-march=native} has no effect if GCC does not recognize
21619 In processor names, a final @samp{000} can be abbreviated as @samp{k}
21620 (for example, @option{-march=r2k}). Prefixes are optional, and
21621 @samp{vr} may be written @samp{r}.
21623 Names of the form @samp{@var{n}f2_1} refer to processors with
21624 FPUs clocked at half the rate of the core, names of the form
21625 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
21626 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
21627 processors with FPUs clocked a ratio of 3:2 with respect to the core.
21628 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
21629 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
21630 accepted as synonyms for @samp{@var{n}f1_1}.
21632 GCC defines two macros based on the value of this option. The first
21633 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
21634 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
21635 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
21636 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
21637 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
21639 Note that the @code{_MIPS_ARCH} macro uses the processor names given
21640 above. In other words, it has the full prefix and does not
21641 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
21642 the macro names the resolved architecture (either @code{"mips1"} or
21643 @code{"mips3"}). It names the default architecture when no
21644 @option{-march} option is given.
21646 @item -mtune=@var{arch}
21648 Optimize for @var{arch}. Among other things, this option controls
21649 the way instructions are scheduled, and the perceived cost of arithmetic
21650 operations. The list of @var{arch} values is the same as for
21653 When this option is not used, GCC optimizes for the processor
21654 specified by @option{-march}. By using @option{-march} and
21655 @option{-mtune} together, it is possible to generate code that
21656 runs on a family of processors, but optimize the code for one
21657 particular member of that family.
21659 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
21660 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
21661 @option{-march} ones described above.
21665 Equivalent to @option{-march=mips1}.
21669 Equivalent to @option{-march=mips2}.
21673 Equivalent to @option{-march=mips3}.
21677 Equivalent to @option{-march=mips4}.
21681 Equivalent to @option{-march=mips32}.
21685 Equivalent to @option{-march=mips32r3}.
21689 Equivalent to @option{-march=mips32r5}.
21693 Equivalent to @option{-march=mips32r6}.
21697 Equivalent to @option{-march=mips64}.
21701 Equivalent to @option{-march=mips64r2}.
21705 Equivalent to @option{-march=mips64r3}.
21709 Equivalent to @option{-march=mips64r5}.
21713 Equivalent to @option{-march=mips64r6}.
21718 @opindex mno-mips16
21719 Generate (do not generate) MIPS16 code. If GCC is targeting a
21720 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
21722 MIPS16 code generation can also be controlled on a per-function basis
21723 by means of @code{mips16} and @code{nomips16} attributes.
21724 @xref{Function Attributes}, for more information.
21726 @item -mflip-mips16
21727 @opindex mflip-mips16
21728 Generate MIPS16 code on alternating functions. This option is provided
21729 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
21730 not intended for ordinary use in compiling user code.
21732 @item -minterlink-compressed
21733 @itemx -mno-interlink-compressed
21734 @opindex minterlink-compressed
21735 @opindex mno-interlink-compressed
21736 Require (do not require) that code using the standard (uncompressed) MIPS ISA
21737 be link-compatible with MIPS16 and microMIPS code, and vice versa.
21739 For example, code using the standard ISA encoding cannot jump directly
21740 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
21741 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
21742 knows that the target of the jump is not compressed.
21744 @item -minterlink-mips16
21745 @itemx -mno-interlink-mips16
21746 @opindex minterlink-mips16
21747 @opindex mno-interlink-mips16
21748 Aliases of @option{-minterlink-compressed} and
21749 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
21750 and are retained for backwards compatibility.
21762 Generate code for the given ABI@.
21764 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
21765 generates 64-bit code when you select a 64-bit architecture, but you
21766 can use @option{-mgp32} to get 32-bit code instead.
21768 For information about the O64 ABI, see
21769 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
21771 GCC supports a variant of the o32 ABI in which floating-point registers
21772 are 64 rather than 32 bits wide. You can select this combination with
21773 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
21774 and @code{mfhc1} instructions and is therefore only supported for
21775 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
21777 The register assignments for arguments and return values remain the
21778 same, but each scalar value is passed in a single 64-bit register
21779 rather than a pair of 32-bit registers. For example, scalar
21780 floating-point values are returned in @samp{$f0} only, not a
21781 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
21782 remains the same in that the even-numbered double-precision registers
21785 Two additional variants of the o32 ABI are supported to enable
21786 a transition from 32-bit to 64-bit registers. These are FPXX
21787 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
21788 The FPXX extension mandates that all code must execute correctly
21789 when run using 32-bit or 64-bit registers. The code can be interlinked
21790 with either FP32 or FP64, but not both.
21791 The FP64A extension is similar to the FP64 extension but forbids the
21792 use of odd-numbered single-precision registers. This can be used
21793 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
21794 processors and allows both FP32 and FP64A code to interlink and
21795 run in the same process without changing FPU modes.
21798 @itemx -mno-abicalls
21800 @opindex mno-abicalls
21801 Generate (do not generate) code that is suitable for SVR4-style
21802 dynamic objects. @option{-mabicalls} is the default for SVR4-based
21807 Generate (do not generate) code that is fully position-independent,
21808 and that can therefore be linked into shared libraries. This option
21809 only affects @option{-mabicalls}.
21811 All @option{-mabicalls} code has traditionally been position-independent,
21812 regardless of options like @option{-fPIC} and @option{-fpic}. However,
21813 as an extension, the GNU toolchain allows executables to use absolute
21814 accesses for locally-binding symbols. It can also use shorter GP
21815 initialization sequences and generate direct calls to locally-defined
21816 functions. This mode is selected by @option{-mno-shared}.
21818 @option{-mno-shared} depends on binutils 2.16 or higher and generates
21819 objects that can only be linked by the GNU linker. However, the option
21820 does not affect the ABI of the final executable; it only affects the ABI
21821 of relocatable objects. Using @option{-mno-shared} generally makes
21822 executables both smaller and quicker.
21824 @option{-mshared} is the default.
21830 Assume (do not assume) that the static and dynamic linkers
21831 support PLTs and copy relocations. This option only affects
21832 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
21833 has no effect without @option{-msym32}.
21835 You can make @option{-mplt} the default by configuring
21836 GCC with @option{--with-mips-plt}. The default is
21837 @option{-mno-plt} otherwise.
21843 Lift (do not lift) the usual restrictions on the size of the global
21846 GCC normally uses a single instruction to load values from the GOT@.
21847 While this is relatively efficient, it only works if the GOT
21848 is smaller than about 64k. Anything larger causes the linker
21849 to report an error such as:
21851 @cindex relocation truncated to fit (MIPS)
21853 relocation truncated to fit: R_MIPS_GOT16 foobar
21856 If this happens, you should recompile your code with @option{-mxgot}.
21857 This works with very large GOTs, although the code is also
21858 less efficient, since it takes three instructions to fetch the
21859 value of a global symbol.
21861 Note that some linkers can create multiple GOTs. If you have such a
21862 linker, you should only need to use @option{-mxgot} when a single object
21863 file accesses more than 64k's worth of GOT entries. Very few do.
21865 These options have no effect unless GCC is generating position
21870 Assume that general-purpose registers are 32 bits wide.
21874 Assume that general-purpose registers are 64 bits wide.
21878 Assume that floating-point registers are 32 bits wide.
21882 Assume that floating-point registers are 64 bits wide.
21886 Do not assume the width of floating-point registers.
21889 @opindex mhard-float
21890 Use floating-point coprocessor instructions.
21893 @opindex msoft-float
21894 Do not use floating-point coprocessor instructions. Implement
21895 floating-point calculations using library calls instead.
21899 Equivalent to @option{-msoft-float}, but additionally asserts that the
21900 program being compiled does not perform any floating-point operations.
21901 This option is presently supported only by some bare-metal MIPS
21902 configurations, where it may select a special set of libraries
21903 that lack all floating-point support (including, for example, the
21904 floating-point @code{printf} formats).
21905 If code compiled with @option{-mno-float} accidentally contains
21906 floating-point operations, it is likely to suffer a link-time
21907 or run-time failure.
21909 @item -msingle-float
21910 @opindex msingle-float
21911 Assume that the floating-point coprocessor only supports single-precision
21914 @item -mdouble-float
21915 @opindex mdouble-float
21916 Assume that the floating-point coprocessor supports double-precision
21917 operations. This is the default.
21920 @itemx -mno-odd-spreg
21921 @opindex modd-spreg
21922 @opindex mno-odd-spreg
21923 Enable the use of odd-numbered single-precision floating-point registers
21924 for the o32 ABI. This is the default for processors that are known to
21925 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
21929 @itemx -mabs=legacy
21931 @opindex mabs=legacy
21932 These options control the treatment of the special not-a-number (NaN)
21933 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
21934 @code{neg.@i{fmt}} machine instructions.
21936 By default or when @option{-mabs=legacy} is used the legacy
21937 treatment is selected. In this case these instructions are considered
21938 arithmetic and avoided where correct operation is required and the
21939 input operand might be a NaN. A longer sequence of instructions that
21940 manipulate the sign bit of floating-point datum manually is used
21941 instead unless the @option{-ffinite-math-only} option has also been
21944 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
21945 this case these instructions are considered non-arithmetic and therefore
21946 operating correctly in all cases, including in particular where the
21947 input operand is a NaN. These instructions are therefore always used
21948 for the respective operations.
21951 @itemx -mnan=legacy
21953 @opindex mnan=legacy
21954 These options control the encoding of the special not-a-number (NaN)
21955 IEEE 754 floating-point data.
21957 The @option{-mnan=legacy} option selects the legacy encoding. In this
21958 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
21959 significand field being 0, whereas signaling NaNs (sNaNs) are denoted
21960 by the first bit of their trailing significand field being 1.
21962 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
21963 this case qNaNs are denoted by the first bit of their trailing
21964 significand field being 1, whereas sNaNs are denoted by the first bit of
21965 their trailing significand field being 0.
21967 The default is @option{-mnan=legacy} unless GCC has been configured with
21968 @option{--with-nan=2008}.
21974 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
21975 implement atomic memory built-in functions. When neither option is
21976 specified, GCC uses the instructions if the target architecture
21979 @option{-mllsc} is useful if the runtime environment can emulate the
21980 instructions and @option{-mno-llsc} can be useful when compiling for
21981 nonstandard ISAs. You can make either option the default by
21982 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
21983 respectively. @option{--with-llsc} is the default for some
21984 configurations; see the installation documentation for details.
21990 Use (do not use) revision 1 of the MIPS DSP ASE@.
21991 @xref{MIPS DSP Built-in Functions}. This option defines the
21992 preprocessor macro @code{__mips_dsp}. It also defines
21993 @code{__mips_dsp_rev} to 1.
21999 Use (do not use) revision 2 of the MIPS DSP ASE@.
22000 @xref{MIPS DSP Built-in Functions}. This option defines the
22001 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
22002 It also defines @code{__mips_dsp_rev} to 2.
22005 @itemx -mno-smartmips
22006 @opindex msmartmips
22007 @opindex mno-smartmips
22008 Use (do not use) the MIPS SmartMIPS ASE.
22010 @item -mpaired-single
22011 @itemx -mno-paired-single
22012 @opindex mpaired-single
22013 @opindex mno-paired-single
22014 Use (do not use) paired-single floating-point instructions.
22015 @xref{MIPS Paired-Single Support}. This option requires
22016 hardware floating-point support to be enabled.
22022 Use (do not use) MIPS Digital Media Extension instructions.
22023 This option can only be used when generating 64-bit code and requires
22024 hardware floating-point support to be enabled.
22029 @opindex mno-mips3d
22030 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
22031 The option @option{-mips3d} implies @option{-mpaired-single}.
22034 @itemx -mno-micromips
22035 @opindex mmicromips
22036 @opindex mno-mmicromips
22037 Generate (do not generate) microMIPS code.
22039 MicroMIPS code generation can also be controlled on a per-function basis
22040 by means of @code{micromips} and @code{nomicromips} attributes.
22041 @xref{Function Attributes}, for more information.
22047 Use (do not use) MT Multithreading instructions.
22053 Use (do not use) the MIPS MCU ASE instructions.
22059 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
22065 Use (do not use) the MIPS Virtualization (VZ) instructions.
22071 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
22077 Use (do not use) the MIPS Cyclic Redundancy Check (CRC) instructions.
22083 Use (do not use) the MIPS Global INValidate (GINV) instructions.
22085 @item -mloongson-mmi
22086 @itemx -mno-loongson-mmi
22087 @opindex mloongson-mmi
22088 @opindex mno-loongson-mmi
22089 Use (do not use) the MIPS Loongson MultiMedia extensions Instructions (MMI).
22091 @item -mloongson-ext
22092 @itemx -mno-loongson-ext
22093 @opindex mloongson-ext
22094 @opindex mno-loongson-ext
22095 Use (do not use) the MIPS Loongson EXTensions (EXT) instructions.
22097 @item -mloongson-ext2
22098 @itemx -mno-loongson-ext2
22099 @opindex mloongson-ext2
22100 @opindex mno-loongson-ext2
22101 Use (do not use) the MIPS Loongson EXTensions r2 (EXT2) instructions.
22105 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
22106 an explanation of the default and the way that the pointer size is
22111 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
22113 The default size of @code{int}s, @code{long}s and pointers depends on
22114 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
22115 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
22116 32-bit @code{long}s. Pointers are the same size as @code{long}s,
22117 or the same size as integer registers, whichever is smaller.
22123 Assume (do not assume) that all symbols have 32-bit values, regardless
22124 of the selected ABI@. This option is useful in combination with
22125 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
22126 to generate shorter and faster references to symbolic addresses.
22130 Put definitions of externally-visible data in a small data section
22131 if that data is no bigger than @var{num} bytes. GCC can then generate
22132 more efficient accesses to the data; see @option{-mgpopt} for details.
22134 The default @option{-G} option depends on the configuration.
22136 @item -mlocal-sdata
22137 @itemx -mno-local-sdata
22138 @opindex mlocal-sdata
22139 @opindex mno-local-sdata
22140 Extend (do not extend) the @option{-G} behavior to local data too,
22141 such as to static variables in C@. @option{-mlocal-sdata} is the
22142 default for all configurations.
22144 If the linker complains that an application is using too much small data,
22145 you might want to try rebuilding the less performance-critical parts with
22146 @option{-mno-local-sdata}. You might also want to build large
22147 libraries with @option{-mno-local-sdata}, so that the libraries leave
22148 more room for the main program.
22150 @item -mextern-sdata
22151 @itemx -mno-extern-sdata
22152 @opindex mextern-sdata
22153 @opindex mno-extern-sdata
22154 Assume (do not assume) that externally-defined data is in
22155 a small data section if the size of that data is within the @option{-G} limit.
22156 @option{-mextern-sdata} is the default for all configurations.
22158 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
22159 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
22160 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
22161 is placed in a small data section. If @var{Var} is defined by another
22162 module, you must either compile that module with a high-enough
22163 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
22164 definition. If @var{Var} is common, you must link the application
22165 with a high-enough @option{-G} setting.
22167 The easiest way of satisfying these restrictions is to compile
22168 and link every module with the same @option{-G} option. However,
22169 you may wish to build a library that supports several different
22170 small data limits. You can do this by compiling the library with
22171 the highest supported @option{-G} setting and additionally using
22172 @option{-mno-extern-sdata} to stop the library from making assumptions
22173 about externally-defined data.
22179 Use (do not use) GP-relative accesses for symbols that are known to be
22180 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
22181 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
22184 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
22185 might not hold the value of @code{_gp}. For example, if the code is
22186 part of a library that might be used in a boot monitor, programs that
22187 call boot monitor routines pass an unknown value in @code{$gp}.
22188 (In such situations, the boot monitor itself is usually compiled
22189 with @option{-G0}.)
22191 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
22192 @option{-mno-extern-sdata}.
22194 @item -membedded-data
22195 @itemx -mno-embedded-data
22196 @opindex membedded-data
22197 @opindex mno-embedded-data
22198 Allocate variables to the read-only data section first if possible, then
22199 next in the small data section if possible, otherwise in data. This gives
22200 slightly slower code than the default, but reduces the amount of RAM required
22201 when executing, and thus may be preferred for some embedded systems.
22203 @item -muninit-const-in-rodata
22204 @itemx -mno-uninit-const-in-rodata
22205 @opindex muninit-const-in-rodata
22206 @opindex mno-uninit-const-in-rodata
22207 Put uninitialized @code{const} variables in the read-only data section.
22208 This option is only meaningful in conjunction with @option{-membedded-data}.
22210 @item -mcode-readable=@var{setting}
22211 @opindex mcode-readable
22212 Specify whether GCC may generate code that reads from executable sections.
22213 There are three possible settings:
22216 @item -mcode-readable=yes
22217 Instructions may freely access executable sections. This is the
22220 @item -mcode-readable=pcrel
22221 MIPS16 PC-relative load instructions can access executable sections,
22222 but other instructions must not do so. This option is useful on 4KSc
22223 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
22224 It is also useful on processors that can be configured to have a dual
22225 instruction/data SRAM interface and that, like the M4K, automatically
22226 redirect PC-relative loads to the instruction RAM.
22228 @item -mcode-readable=no
22229 Instructions must not access executable sections. This option can be
22230 useful on targets that are configured to have a dual instruction/data
22231 SRAM interface but that (unlike the M4K) do not automatically redirect
22232 PC-relative loads to the instruction RAM.
22235 @item -msplit-addresses
22236 @itemx -mno-split-addresses
22237 @opindex msplit-addresses
22238 @opindex mno-split-addresses
22239 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
22240 relocation operators. This option has been superseded by
22241 @option{-mexplicit-relocs} but is retained for backwards compatibility.
22243 @item -mexplicit-relocs
22244 @itemx -mno-explicit-relocs
22245 @opindex mexplicit-relocs
22246 @opindex mno-explicit-relocs
22247 Use (do not use) assembler relocation operators when dealing with symbolic
22248 addresses. The alternative, selected by @option{-mno-explicit-relocs},
22249 is to use assembler macros instead.
22251 @option{-mexplicit-relocs} is the default if GCC was configured
22252 to use an assembler that supports relocation operators.
22254 @item -mcheck-zero-division
22255 @itemx -mno-check-zero-division
22256 @opindex mcheck-zero-division
22257 @opindex mno-check-zero-division
22258 Trap (do not trap) on integer division by zero.
22260 The default is @option{-mcheck-zero-division}.
22262 @item -mdivide-traps
22263 @itemx -mdivide-breaks
22264 @opindex mdivide-traps
22265 @opindex mdivide-breaks
22266 MIPS systems check for division by zero by generating either a
22267 conditional trap or a break instruction. Using traps results in
22268 smaller code, but is only supported on MIPS II and later. Also, some
22269 versions of the Linux kernel have a bug that prevents trap from
22270 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
22271 allow conditional traps on architectures that support them and
22272 @option{-mdivide-breaks} to force the use of breaks.
22274 The default is usually @option{-mdivide-traps}, but this can be
22275 overridden at configure time using @option{--with-divide=breaks}.
22276 Divide-by-zero checks can be completely disabled using
22277 @option{-mno-check-zero-division}.
22279 @item -mload-store-pairs
22280 @itemx -mno-load-store-pairs
22281 @opindex mload-store-pairs
22282 @opindex mno-load-store-pairs
22283 Enable (disable) an optimization that pairs consecutive load or store
22284 instructions to enable load/store bonding. This option is enabled by
22285 default but only takes effect when the selected architecture is known
22286 to support bonding.
22291 @opindex mno-memcpy
22292 Force (do not force) the use of @code{memcpy} for non-trivial block
22293 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
22294 most constant-sized copies.
22297 @itemx -mno-long-calls
22298 @opindex mlong-calls
22299 @opindex mno-long-calls
22300 Disable (do not disable) use of the @code{jal} instruction. Calling
22301 functions using @code{jal} is more efficient but requires the caller
22302 and callee to be in the same 256 megabyte segment.
22304 This option has no effect on abicalls code. The default is
22305 @option{-mno-long-calls}.
22311 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
22312 instructions, as provided by the R4650 ISA@.
22318 Enable (disable) use of the @code{madd} and @code{msub} integer
22319 instructions. The default is @option{-mimadd} on architectures
22320 that support @code{madd} and @code{msub} except for the 74k
22321 architecture where it was found to generate slower code.
22324 @itemx -mno-fused-madd
22325 @opindex mfused-madd
22326 @opindex mno-fused-madd
22327 Enable (disable) use of the floating-point multiply-accumulate
22328 instructions, when they are available. The default is
22329 @option{-mfused-madd}.
22331 On the R8000 CPU when multiply-accumulate instructions are used,
22332 the intermediate product is calculated to infinite precision
22333 and is not subject to the FCSR Flush to Zero bit. This may be
22334 undesirable in some circumstances. On other processors the result
22335 is numerically identical to the equivalent computation using
22336 separate multiply, add, subtract and negate instructions.
22340 Tell the MIPS assembler to not run its preprocessor over user
22341 assembler files (with a @samp{.s} suffix) when assembling them.
22344 @itemx -mno-fix-24k
22346 @opindex mno-fix-24k
22347 Work around the 24K E48 (lost data on stores during refill) errata.
22348 The workarounds are implemented by the assembler rather than by GCC@.
22351 @itemx -mno-fix-r4000
22352 @opindex mfix-r4000
22353 @opindex mno-fix-r4000
22354 Work around certain R4000 CPU errata:
22357 A double-word or a variable shift may give an incorrect result if executed
22358 immediately after starting an integer division.
22360 A double-word or a variable shift may give an incorrect result if executed
22361 while an integer multiplication is in progress.
22363 An integer division may give an incorrect result if started in a delay slot
22364 of a taken branch or a jump.
22368 @itemx -mno-fix-r4400
22369 @opindex mfix-r4400
22370 @opindex mno-fix-r4400
22371 Work around certain R4400 CPU errata:
22374 A double-word or a variable shift may give an incorrect result if executed
22375 immediately after starting an integer division.
22379 @itemx -mno-fix-r10000
22380 @opindex mfix-r10000
22381 @opindex mno-fix-r10000
22382 Work around certain R10000 errata:
22385 @code{ll}/@code{sc} sequences may not behave atomically on revisions
22386 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
22389 This option can only be used if the target architecture supports
22390 branch-likely instructions. @option{-mfix-r10000} is the default when
22391 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
22395 @itemx -mno-fix-r5900
22396 @opindex mfix-r5900
22397 Do not attempt to schedule the preceding instruction into the delay slot
22398 of a branch instruction placed at the end of a short loop of six
22399 instructions or fewer and always schedule a @code{nop} instruction there
22400 instead. The short loop bug under certain conditions causes loops to
22401 execute only once or twice, due to a hardware bug in the R5900 chip. The
22402 workaround is implemented by the assembler rather than by GCC@.
22405 @itemx -mno-fix-rm7000
22406 @opindex mfix-rm7000
22407 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
22408 workarounds are implemented by the assembler rather than by GCC@.
22411 @itemx -mno-fix-vr4120
22412 @opindex mfix-vr4120
22413 Work around certain VR4120 errata:
22416 @code{dmultu} does not always produce the correct result.
22418 @code{div} and @code{ddiv} do not always produce the correct result if one
22419 of the operands is negative.
22421 The workarounds for the division errata rely on special functions in
22422 @file{libgcc.a}. At present, these functions are only provided by
22423 the @code{mips64vr*-elf} configurations.
22425 Other VR4120 errata require a NOP to be inserted between certain pairs of
22426 instructions. These errata are handled by the assembler, not by GCC itself.
22429 @opindex mfix-vr4130
22430 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
22431 workarounds are implemented by the assembler rather than by GCC,
22432 although GCC avoids using @code{mflo} and @code{mfhi} if the
22433 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
22434 instructions are available instead.
22437 @itemx -mno-fix-sb1
22439 Work around certain SB-1 CPU core errata.
22440 (This flag currently works around the SB-1 revision 2
22441 ``F1'' and ``F2'' floating-point errata.)
22443 @item -mr10k-cache-barrier=@var{setting}
22444 @opindex mr10k-cache-barrier
22445 Specify whether GCC should insert cache barriers to avoid the
22446 side effects of speculation on R10K processors.
22448 In common with many processors, the R10K tries to predict the outcome
22449 of a conditional branch and speculatively executes instructions from
22450 the ``taken'' branch. It later aborts these instructions if the
22451 predicted outcome is wrong. However, on the R10K, even aborted
22452 instructions can have side effects.
22454 This problem only affects kernel stores and, depending on the system,
22455 kernel loads. As an example, a speculatively-executed store may load
22456 the target memory into cache and mark the cache line as dirty, even if
22457 the store itself is later aborted. If a DMA operation writes to the
22458 same area of memory before the ``dirty'' line is flushed, the cached
22459 data overwrites the DMA-ed data. See the R10K processor manual
22460 for a full description, including other potential problems.
22462 One workaround is to insert cache barrier instructions before every memory
22463 access that might be speculatively executed and that might have side
22464 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
22465 controls GCC's implementation of this workaround. It assumes that
22466 aborted accesses to any byte in the following regions does not have
22471 the memory occupied by the current function's stack frame;
22474 the memory occupied by an incoming stack argument;
22477 the memory occupied by an object with a link-time-constant address.
22480 It is the kernel's responsibility to ensure that speculative
22481 accesses to these regions are indeed safe.
22483 If the input program contains a function declaration such as:
22489 then the implementation of @code{foo} must allow @code{j foo} and
22490 @code{jal foo} to be executed speculatively. GCC honors this
22491 restriction for functions it compiles itself. It expects non-GCC
22492 functions (such as hand-written assembly code) to do the same.
22494 The option has three forms:
22497 @item -mr10k-cache-barrier=load-store
22498 Insert a cache barrier before a load or store that might be
22499 speculatively executed and that might have side effects even
22502 @item -mr10k-cache-barrier=store
22503 Insert a cache barrier before a store that might be speculatively
22504 executed and that might have side effects even if aborted.
22506 @item -mr10k-cache-barrier=none
22507 Disable the insertion of cache barriers. This is the default setting.
22510 @item -mflush-func=@var{func}
22511 @itemx -mno-flush-func
22512 @opindex mflush-func
22513 Specifies the function to call to flush the I and D caches, or to not
22514 call any such function. If called, the function must take the same
22515 arguments as the common @code{_flush_func}, that is, the address of the
22516 memory range for which the cache is being flushed, the size of the
22517 memory range, and the number 3 (to flush both caches). The default
22518 depends on the target GCC was configured for, but commonly is either
22519 @code{_flush_func} or @code{__cpu_flush}.
22521 @item mbranch-cost=@var{num}
22522 @opindex mbranch-cost
22523 Set the cost of branches to roughly @var{num} ``simple'' instructions.
22524 This cost is only a heuristic and is not guaranteed to produce
22525 consistent results across releases. A zero cost redundantly selects
22526 the default, which is based on the @option{-mtune} setting.
22528 @item -mbranch-likely
22529 @itemx -mno-branch-likely
22530 @opindex mbranch-likely
22531 @opindex mno-branch-likely
22532 Enable or disable use of Branch Likely instructions, regardless of the
22533 default for the selected architecture. By default, Branch Likely
22534 instructions may be generated if they are supported by the selected
22535 architecture. An exception is for the MIPS32 and MIPS64 architectures
22536 and processors that implement those architectures; for those, Branch
22537 Likely instructions are not be generated by default because the MIPS32
22538 and MIPS64 architectures specifically deprecate their use.
22540 @item -mcompact-branches=never
22541 @itemx -mcompact-branches=optimal
22542 @itemx -mcompact-branches=always
22543 @opindex mcompact-branches=never
22544 @opindex mcompact-branches=optimal
22545 @opindex mcompact-branches=always
22546 These options control which form of branches will be generated. The
22547 default is @option{-mcompact-branches=optimal}.
22549 The @option{-mcompact-branches=never} option ensures that compact branch
22550 instructions will never be generated.
22552 The @option{-mcompact-branches=always} option ensures that a compact
22553 branch instruction will be generated if available. If a compact branch
22554 instruction is not available, a delay slot form of the branch will be
22557 This option is supported from MIPS Release 6 onwards.
22559 The @option{-mcompact-branches=optimal} option will cause a delay slot
22560 branch to be used if one is available in the current ISA and the delay
22561 slot is successfully filled. If the delay slot is not filled, a compact
22562 branch will be chosen if one is available.
22564 @item -mfp-exceptions
22565 @itemx -mno-fp-exceptions
22566 @opindex mfp-exceptions
22567 Specifies whether FP exceptions are enabled. This affects how
22568 FP instructions are scheduled for some processors.
22569 The default is that FP exceptions are
22572 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
22573 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
22576 @item -mvr4130-align
22577 @itemx -mno-vr4130-align
22578 @opindex mvr4130-align
22579 The VR4130 pipeline is two-way superscalar, but can only issue two
22580 instructions together if the first one is 8-byte aligned. When this
22581 option is enabled, GCC aligns pairs of instructions that it
22582 thinks should execute in parallel.
22584 This option only has an effect when optimizing for the VR4130.
22585 It normally makes code faster, but at the expense of making it bigger.
22586 It is enabled by default at optimization level @option{-O3}.
22591 Enable (disable) generation of @code{synci} instructions on
22592 architectures that support it. The @code{synci} instructions (if
22593 enabled) are generated when @code{__builtin___clear_cache} is
22596 This option defaults to @option{-mno-synci}, but the default can be
22597 overridden by configuring GCC with @option{--with-synci}.
22599 When compiling code for single processor systems, it is generally safe
22600 to use @code{synci}. However, on many multi-core (SMP) systems, it
22601 does not invalidate the instruction caches on all cores and may lead
22602 to undefined behavior.
22604 @item -mrelax-pic-calls
22605 @itemx -mno-relax-pic-calls
22606 @opindex mrelax-pic-calls
22607 Try to turn PIC calls that are normally dispatched via register
22608 @code{$25} into direct calls. This is only possible if the linker can
22609 resolve the destination at link time and if the destination is within
22610 range for a direct call.
22612 @option{-mrelax-pic-calls} is the default if GCC was configured to use
22613 an assembler and a linker that support the @code{.reloc} assembly
22614 directive and @option{-mexplicit-relocs} is in effect. With
22615 @option{-mno-explicit-relocs}, this optimization can be performed by the
22616 assembler and the linker alone without help from the compiler.
22618 @item -mmcount-ra-address
22619 @itemx -mno-mcount-ra-address
22620 @opindex mmcount-ra-address
22621 @opindex mno-mcount-ra-address
22622 Emit (do not emit) code that allows @code{_mcount} to modify the
22623 calling function's return address. When enabled, this option extends
22624 the usual @code{_mcount} interface with a new @var{ra-address}
22625 parameter, which has type @code{intptr_t *} and is passed in register
22626 @code{$12}. @code{_mcount} can then modify the return address by
22627 doing both of the following:
22630 Returning the new address in register @code{$31}.
22632 Storing the new address in @code{*@var{ra-address}},
22633 if @var{ra-address} is nonnull.
22636 The default is @option{-mno-mcount-ra-address}.
22638 @item -mframe-header-opt
22639 @itemx -mno-frame-header-opt
22640 @opindex mframe-header-opt
22641 Enable (disable) frame header optimization in the o32 ABI. When using the
22642 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
22643 function to write out register arguments. When enabled, this optimization
22644 will suppress the allocation of the frame header if it can be determined that
22647 This optimization is off by default at all optimization levels.
22650 @itemx -mno-lxc1-sxc1
22651 @opindex mlxc1-sxc1
22652 When applicable, enable (disable) the generation of @code{lwxc1},
22653 @code{swxc1}, @code{ldxc1}, @code{sdxc1} instructions. Enabled by default.
22658 When applicable, enable (disable) the generation of 4-operand @code{madd.s},
22659 @code{madd.d} and related instructions. Enabled by default.
22664 @subsection MMIX Options
22665 @cindex MMIX Options
22667 These options are defined for the MMIX:
22671 @itemx -mno-libfuncs
22673 @opindex mno-libfuncs
22674 Specify that intrinsic library functions are being compiled, passing all
22675 values in registers, no matter the size.
22678 @itemx -mno-epsilon
22680 @opindex mno-epsilon
22681 Generate floating-point comparison instructions that compare with respect
22682 to the @code{rE} epsilon register.
22684 @item -mabi=mmixware
22686 @opindex mabi=mmixware
22688 Generate code that passes function parameters and return values that (in
22689 the called function) are seen as registers @code{$0} and up, as opposed to
22690 the GNU ABI which uses global registers @code{$231} and up.
22692 @item -mzero-extend
22693 @itemx -mno-zero-extend
22694 @opindex mzero-extend
22695 @opindex mno-zero-extend
22696 When reading data from memory in sizes shorter than 64 bits, use (do not
22697 use) zero-extending load instructions by default, rather than
22698 sign-extending ones.
22701 @itemx -mno-knuthdiv
22703 @opindex mno-knuthdiv
22704 Make the result of a division yielding a remainder have the same sign as
22705 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
22706 remainder follows the sign of the dividend. Both methods are
22707 arithmetically valid, the latter being almost exclusively used.
22709 @item -mtoplevel-symbols
22710 @itemx -mno-toplevel-symbols
22711 @opindex mtoplevel-symbols
22712 @opindex mno-toplevel-symbols
22713 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
22714 code can be used with the @code{PREFIX} assembly directive.
22718 Generate an executable in the ELF format, rather than the default
22719 @samp{mmo} format used by the @command{mmix} simulator.
22721 @item -mbranch-predict
22722 @itemx -mno-branch-predict
22723 @opindex mbranch-predict
22724 @opindex mno-branch-predict
22725 Use (do not use) the probable-branch instructions, when static branch
22726 prediction indicates a probable branch.
22728 @item -mbase-addresses
22729 @itemx -mno-base-addresses
22730 @opindex mbase-addresses
22731 @opindex mno-base-addresses
22732 Generate (do not generate) code that uses @emph{base addresses}. Using a
22733 base address automatically generates a request (handled by the assembler
22734 and the linker) for a constant to be set up in a global register. The
22735 register is used for one or more base address requests within the range 0
22736 to 255 from the value held in the register. The generally leads to short
22737 and fast code, but the number of different data items that can be
22738 addressed is limited. This means that a program that uses lots of static
22739 data may require @option{-mno-base-addresses}.
22741 @item -msingle-exit
22742 @itemx -mno-single-exit
22743 @opindex msingle-exit
22744 @opindex mno-single-exit
22745 Force (do not force) generated code to have a single exit point in each
22749 @node MN10300 Options
22750 @subsection MN10300 Options
22751 @cindex MN10300 options
22753 These @option{-m} options are defined for Matsushita MN10300 architectures:
22758 Generate code to avoid bugs in the multiply instructions for the MN10300
22759 processors. This is the default.
22761 @item -mno-mult-bug
22762 @opindex mno-mult-bug
22763 Do not generate code to avoid bugs in the multiply instructions for the
22764 MN10300 processors.
22768 Generate code using features specific to the AM33 processor.
22772 Do not generate code using features specific to the AM33 processor. This
22777 Generate code using features specific to the AM33/2.0 processor.
22781 Generate code using features specific to the AM34 processor.
22783 @item -mtune=@var{cpu-type}
22785 Use the timing characteristics of the indicated CPU type when
22786 scheduling instructions. This does not change the targeted processor
22787 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
22788 @samp{am33-2} or @samp{am34}.
22790 @item -mreturn-pointer-on-d0
22791 @opindex mreturn-pointer-on-d0
22792 When generating a function that returns a pointer, return the pointer
22793 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
22794 only in @code{a0}, and attempts to call such functions without a prototype
22795 result in errors. Note that this option is on by default; use
22796 @option{-mno-return-pointer-on-d0} to disable it.
22800 Do not link in the C run-time initialization object file.
22804 Indicate to the linker that it should perform a relaxation optimization pass
22805 to shorten branches, calls and absolute memory addresses. This option only
22806 has an effect when used on the command line for the final link step.
22808 This option makes symbolic debugging impossible.
22812 Allow the compiler to generate @emph{Long Instruction Word}
22813 instructions if the target is the @samp{AM33} or later. This is the
22814 default. This option defines the preprocessor macro @code{__LIW__}.
22818 Do not allow the compiler to generate @emph{Long Instruction Word}
22819 instructions. This option defines the preprocessor macro
22824 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
22825 instructions if the target is the @samp{AM33} or later. This is the
22826 default. This option defines the preprocessor macro @code{__SETLB__}.
22830 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
22831 instructions. This option defines the preprocessor macro
22832 @code{__NO_SETLB__}.
22836 @node Moxie Options
22837 @subsection Moxie Options
22838 @cindex Moxie Options
22844 Generate big-endian code. This is the default for @samp{moxie-*-*}
22849 Generate little-endian code.
22853 Generate mul.x and umul.x instructions. This is the default for
22854 @samp{moxiebox-*-*} configurations.
22858 Do not link in the C run-time initialization object file.
22862 @node MSP430 Options
22863 @subsection MSP430 Options
22864 @cindex MSP430 Options
22866 These options are defined for the MSP430:
22872 Force assembly output to always use hex constants. Normally such
22873 constants are signed decimals, but this option is available for
22874 testsuite and/or aesthetic purposes.
22878 Select the MCU to target. This is used to create a C preprocessor
22879 symbol based upon the MCU name, converted to upper case and pre- and
22880 post-fixed with @samp{__}. This in turn is used by the
22881 @file{msp430.h} header file to select an MCU-specific supplementary
22884 The option also sets the ISA to use. If the MCU name is one that is
22885 known to only support the 430 ISA then that is selected, otherwise the
22886 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
22887 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
22888 name selects the 430X ISA.
22890 In addition an MCU-specific linker script is added to the linker
22891 command line. The script's name is the name of the MCU with
22892 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
22893 command line defines the C preprocessor symbol @code{__XXX__} and
22894 cause the linker to search for a script called @file{xxx.ld}.
22896 This option is also passed on to the assembler.
22899 @itemx -mno-warn-mcu
22901 @opindex mno-warn-mcu
22902 This option enables or disables warnings about conflicts between the
22903 MCU name specified by the @option{-mmcu} option and the ISA set by the
22904 @option{-mcpu} option and/or the hardware multiply support set by the
22905 @option{-mhwmult} option. It also toggles warnings about unrecognized
22906 MCU names. This option is on by default.
22910 Specifies the ISA to use. Accepted values are @samp{msp430},
22911 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
22912 @option{-mmcu=} option should be used to select the ISA.
22916 Link to the simulator runtime libraries and linker script. Overrides
22917 any scripts that would be selected by the @option{-mmcu=} option.
22921 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
22925 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
22929 This option is passed to the assembler and linker, and allows the
22930 linker to perform certain optimizations that cannot be done until
22935 Describes the type of hardware multiply supported by the target.
22936 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
22937 for the original 16-bit-only multiply supported by early MCUs.
22938 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
22939 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
22940 A value of @samp{auto} can also be given. This tells GCC to deduce
22941 the hardware multiply support based upon the MCU name provided by the
22942 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
22943 the MCU name is not recognized then no hardware multiply support is
22944 assumed. @code{auto} is the default setting.
22946 Hardware multiplies are normally performed by calling a library
22947 routine. This saves space in the generated code. When compiling at
22948 @option{-O3} or higher however the hardware multiplier is invoked
22949 inline. This makes for bigger, but faster code.
22951 The hardware multiply routines disable interrupts whilst running and
22952 restore the previous interrupt state when they finish. This makes
22953 them safe to use inside interrupt handlers as well as in normal code.
22957 Enable the use of a minimum runtime environment - no static
22958 initializers or constructors. This is intended for memory-constrained
22959 devices. The compiler includes special symbols in some objects
22960 that tell the linker and runtime which code fragments are required.
22962 @item -mcode-region=
22963 @itemx -mdata-region=
22964 @opindex mcode-region
22965 @opindex mdata-region
22966 These options tell the compiler where to place functions and data that
22967 do not have one of the @code{lower}, @code{upper}, @code{either} or
22968 @code{section} attributes. Possible values are @code{lower},
22969 @code{upper}, @code{either} or @code{any}. The first three behave
22970 like the corresponding attribute. The fourth possible value -
22971 @code{any} - is the default. It leaves placement entirely up to the
22972 linker script and how it assigns the standard sections
22973 (@code{.text}, @code{.data}, etc) to the memory regions.
22975 @item -msilicon-errata=
22976 @opindex msilicon-errata
22977 This option passes on a request to assembler to enable the fixes for
22978 the named silicon errata.
22980 @item -msilicon-errata-warn=
22981 @opindex msilicon-errata-warn
22982 This option passes on a request to the assembler to enable warning
22983 messages when a silicon errata might need to be applied.
22987 @node NDS32 Options
22988 @subsection NDS32 Options
22989 @cindex NDS32 Options
22991 These options are defined for NDS32 implementations:
22996 @opindex mbig-endian
22997 Generate code in big-endian mode.
22999 @item -mlittle-endian
23000 @opindex mlittle-endian
23001 Generate code in little-endian mode.
23003 @item -mreduced-regs
23004 @opindex mreduced-regs
23005 Use reduced-set registers for register allocation.
23008 @opindex mfull-regs
23009 Use full-set registers for register allocation.
23013 Generate conditional move instructions.
23017 Do not generate conditional move instructions.
23021 Generate performance extension instructions.
23023 @item -mno-ext-perf
23024 @opindex mno-ext-perf
23025 Do not generate performance extension instructions.
23028 @opindex mext-perf2
23029 Generate performance extension 2 instructions.
23031 @item -mno-ext-perf2
23032 @opindex mno-ext-perf2
23033 Do not generate performance extension 2 instructions.
23036 @opindex mext-string
23037 Generate string extension instructions.
23039 @item -mno-ext-string
23040 @opindex mno-ext-string
23041 Do not generate string extension instructions.
23045 Generate v3 push25/pop25 instructions.
23048 @opindex mno-v3push
23049 Do not generate v3 push25/pop25 instructions.
23053 Generate 16-bit instructions.
23056 @opindex mno-16-bit
23057 Do not generate 16-bit instructions.
23059 @item -misr-vector-size=@var{num}
23060 @opindex misr-vector-size
23061 Specify the size of each interrupt vector, which must be 4 or 16.
23063 @item -mcache-block-size=@var{num}
23064 @opindex mcache-block-size
23065 Specify the size of each cache block,
23066 which must be a power of 2 between 4 and 512.
23068 @item -march=@var{arch}
23070 Specify the name of the target architecture.
23072 @item -mcmodel=@var{code-model}
23074 Set the code model to one of
23077 All the data and read-only data segments must be within 512KB addressing space.
23078 The text segment must be within 16MB addressing space.
23079 @item @samp{medium}
23080 The data segment must be within 512KB while the read-only data segment can be
23081 within 4GB addressing space. The text segment should be still within 16MB
23084 All the text and data segments can be within 4GB addressing space.
23088 @opindex mctor-dtor
23089 Enable constructor/destructor feature.
23093 Guide linker to relax instructions.
23097 @node Nios II Options
23098 @subsection Nios II Options
23099 @cindex Nios II options
23100 @cindex Altera Nios II options
23102 These are the options defined for the Altera Nios II processor.
23108 @cindex smaller data references
23109 Put global and static objects less than or equal to @var{num} bytes
23110 into the small data or BSS sections instead of the normal data or BSS
23111 sections. The default value of @var{num} is 8.
23113 @item -mgpopt=@var{option}
23118 Generate (do not generate) GP-relative accesses. The following
23119 @var{option} names are recognized:
23124 Do not generate GP-relative accesses.
23127 Generate GP-relative accesses for small data objects that are not
23128 external, weak, or uninitialized common symbols.
23129 Also use GP-relative addressing for objects that
23130 have been explicitly placed in a small data section via a @code{section}
23134 As for @samp{local}, but also generate GP-relative accesses for
23135 small data objects that are external, weak, or common. If you use this option,
23136 you must ensure that all parts of your program (including libraries) are
23137 compiled with the same @option{-G} setting.
23140 Generate GP-relative accesses for all data objects in the program. If you
23141 use this option, the entire data and BSS segments
23142 of your program must fit in 64K of memory and you must use an appropriate
23143 linker script to allocate them within the addressable range of the
23147 Generate GP-relative addresses for function pointers as well as data
23148 pointers. If you use this option, the entire text, data, and BSS segments
23149 of your program must fit in 64K of memory and you must use an appropriate
23150 linker script to allocate them within the addressable range of the
23155 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
23156 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
23158 The default is @option{-mgpopt} except when @option{-fpic} or
23159 @option{-fPIC} is specified to generate position-independent code.
23160 Note that the Nios II ABI does not permit GP-relative accesses from
23163 You may need to specify @option{-mno-gpopt} explicitly when building
23164 programs that include large amounts of small data, including large
23165 GOT data sections. In this case, the 16-bit offset for GP-relative
23166 addressing may not be large enough to allow access to the entire
23167 small data section.
23169 @item -mgprel-sec=@var{regexp}
23170 @opindex mgprel-sec
23171 This option specifies additional section names that can be accessed via
23172 GP-relative addressing. It is most useful in conjunction with
23173 @code{section} attributes on variable declarations
23174 (@pxref{Common Variable Attributes}) and a custom linker script.
23175 The @var{regexp} is a POSIX Extended Regular Expression.
23177 This option does not affect the behavior of the @option{-G} option, and
23178 the specified sections are in addition to the standard @code{.sdata}
23179 and @code{.sbss} small-data sections that are recognized by @option{-mgpopt}.
23181 @item -mr0rel-sec=@var{regexp}
23182 @opindex mr0rel-sec
23183 This option specifies names of sections that can be accessed via a
23184 16-bit offset from @code{r0}; that is, in the low 32K or high 32K
23185 of the 32-bit address space. It is most useful in conjunction with
23186 @code{section} attributes on variable declarations
23187 (@pxref{Common Variable Attributes}) and a custom linker script.
23188 The @var{regexp} is a POSIX Extended Regular Expression.
23190 In contrast to the use of GP-relative addressing for small data,
23191 zero-based addressing is never generated by default and there are no
23192 conventional section names used in standard linker scripts for sections
23193 in the low or high areas of memory.
23199 Generate little-endian (default) or big-endian (experimental) code,
23202 @item -march=@var{arch}
23204 This specifies the name of the target Nios II architecture. GCC uses this
23205 name to determine what kind of instructions it can emit when generating
23206 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
23208 The preprocessor macro @code{__nios2_arch__} is available to programs,
23209 with value 1 or 2, indicating the targeted ISA level.
23211 @item -mbypass-cache
23212 @itemx -mno-bypass-cache
23213 @opindex mno-bypass-cache
23214 @opindex mbypass-cache
23215 Force all load and store instructions to always bypass cache by
23216 using I/O variants of the instructions. The default is not to
23219 @item -mno-cache-volatile
23220 @itemx -mcache-volatile
23221 @opindex mcache-volatile
23222 @opindex mno-cache-volatile
23223 Volatile memory access bypass the cache using the I/O variants of
23224 the load and store instructions. The default is not to bypass the cache.
23226 @item -mno-fast-sw-div
23227 @itemx -mfast-sw-div
23228 @opindex mno-fast-sw-div
23229 @opindex mfast-sw-div
23230 Do not use table-based fast divide for small numbers. The default
23231 is to use the fast divide at @option{-O3} and above.
23235 @itemx -mno-hw-mulx
23239 @opindex mno-hw-mul
23241 @opindex mno-hw-mulx
23243 @opindex mno-hw-div
23245 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
23246 instructions by the compiler. The default is to emit @code{mul}
23247 and not emit @code{div} and @code{mulx}.
23253 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
23254 CDX (code density) instructions. Enabling these instructions also
23255 requires @option{-march=r2}. Since these instructions are optional
23256 extensions to the R2 architecture, the default is not to emit them.
23258 @item -mcustom-@var{insn}=@var{N}
23259 @itemx -mno-custom-@var{insn}
23260 @opindex mcustom-@var{insn}
23261 @opindex mno-custom-@var{insn}
23262 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
23263 custom instruction with encoding @var{N} when generating code that uses
23264 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
23265 instruction 253 for single-precision floating-point add operations instead
23266 of the default behavior of using a library call.
23268 The following values of @var{insn} are supported. Except as otherwise
23269 noted, floating-point operations are expected to be implemented with
23270 normal IEEE 754 semantics and correspond directly to the C operators or the
23271 equivalent GCC built-in functions (@pxref{Other Builtins}).
23273 Single-precision floating point:
23276 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
23277 Binary arithmetic operations.
23283 Unary absolute value.
23285 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
23286 Comparison operations.
23288 @item @samp{fmins}, @samp{fmaxs}
23289 Floating-point minimum and maximum. These instructions are only
23290 generated if @option{-ffinite-math-only} is specified.
23292 @item @samp{fsqrts}
23293 Unary square root operation.
23295 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
23296 Floating-point trigonometric and exponential functions. These instructions
23297 are only generated if @option{-funsafe-math-optimizations} is also specified.
23301 Double-precision floating point:
23304 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
23305 Binary arithmetic operations.
23311 Unary absolute value.
23313 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
23314 Comparison operations.
23316 @item @samp{fmind}, @samp{fmaxd}
23317 Double-precision minimum and maximum. These instructions are only
23318 generated if @option{-ffinite-math-only} is specified.
23320 @item @samp{fsqrtd}
23321 Unary square root operation.
23323 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
23324 Double-precision trigonometric and exponential functions. These instructions
23325 are only generated if @option{-funsafe-math-optimizations} is also specified.
23331 @item @samp{fextsd}
23332 Conversion from single precision to double precision.
23334 @item @samp{ftruncds}
23335 Conversion from double precision to single precision.
23337 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
23338 Conversion from floating point to signed or unsigned integer types, with
23339 truncation towards zero.
23342 Conversion from single-precision floating point to signed integer,
23343 rounding to the nearest integer and ties away from zero.
23344 This corresponds to the @code{__builtin_lroundf} function when
23345 @option{-fno-math-errno} is used.
23347 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
23348 Conversion from signed or unsigned integer types to floating-point types.
23352 In addition, all of the following transfer instructions for internal
23353 registers X and Y must be provided to use any of the double-precision
23354 floating-point instructions. Custom instructions taking two
23355 double-precision source operands expect the first operand in the
23356 64-bit register X. The other operand (or only operand of a unary
23357 operation) is given to the custom arithmetic instruction with the
23358 least significant half in source register @var{src1} and the most
23359 significant half in @var{src2}. A custom instruction that returns a
23360 double-precision result returns the most significant 32 bits in the
23361 destination register and the other half in 32-bit register Y.
23362 GCC automatically generates the necessary code sequences to write
23363 register X and/or read register Y when double-precision floating-point
23364 instructions are used.
23369 Write @var{src1} into the least significant half of X and @var{src2} into
23370 the most significant half of X.
23373 Write @var{src1} into Y.
23375 @item @samp{frdxhi}, @samp{frdxlo}
23376 Read the most or least (respectively) significant half of X and store it in
23380 Read the value of Y and store it into @var{dest}.
23383 Note that you can gain more local control over generation of Nios II custom
23384 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
23385 and @code{target("no-custom-@var{insn}")} function attributes
23386 (@pxref{Function Attributes})
23387 or pragmas (@pxref{Function Specific Option Pragmas}).
23389 @item -mcustom-fpu-cfg=@var{name}
23390 @opindex mcustom-fpu-cfg
23392 This option enables a predefined, named set of custom instruction encodings
23393 (see @option{-mcustom-@var{insn}} above).
23394 Currently, the following sets are defined:
23396 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
23397 @gccoptlist{-mcustom-fmuls=252 @gol
23398 -mcustom-fadds=253 @gol
23399 -mcustom-fsubs=254 @gol
23400 -fsingle-precision-constant}
23402 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
23403 @gccoptlist{-mcustom-fmuls=252 @gol
23404 -mcustom-fadds=253 @gol
23405 -mcustom-fsubs=254 @gol
23406 -mcustom-fdivs=255 @gol
23407 -fsingle-precision-constant}
23409 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
23410 @gccoptlist{-mcustom-floatus=243 @gol
23411 -mcustom-fixsi=244 @gol
23412 -mcustom-floatis=245 @gol
23413 -mcustom-fcmpgts=246 @gol
23414 -mcustom-fcmples=249 @gol
23415 -mcustom-fcmpeqs=250 @gol
23416 -mcustom-fcmpnes=251 @gol
23417 -mcustom-fmuls=252 @gol
23418 -mcustom-fadds=253 @gol
23419 -mcustom-fsubs=254 @gol
23420 -mcustom-fdivs=255 @gol
23421 -fsingle-precision-constant}
23423 Custom instruction assignments given by individual
23424 @option{-mcustom-@var{insn}=} options override those given by
23425 @option{-mcustom-fpu-cfg=}, regardless of the
23426 order of the options on the command line.
23428 Note that you can gain more local control over selection of a FPU
23429 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
23430 function attribute (@pxref{Function Attributes})
23431 or pragma (@pxref{Function Specific Option Pragmas}).
23435 These additional @samp{-m} options are available for the Altera Nios II
23436 ELF (bare-metal) target:
23442 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
23443 startup and termination code, and is typically used in conjunction with
23444 @option{-msys-crt0=} to specify the location of the alternate startup code
23445 provided by the HAL BSP.
23449 Link with a limited version of the C library, @option{-lsmallc}, rather than
23452 @item -msys-crt0=@var{startfile}
23454 @var{startfile} is the file name of the startfile (crt0) to use
23455 when linking. This option is only useful in conjunction with @option{-mhal}.
23457 @item -msys-lib=@var{systemlib}
23459 @var{systemlib} is the library name of the library that provides
23460 low-level system calls required by the C library,
23461 e.g.@: @code{read} and @code{write}.
23462 This option is typically used to link with a library provided by a HAL BSP.
23466 @node Nvidia PTX Options
23467 @subsection Nvidia PTX Options
23468 @cindex Nvidia PTX options
23469 @cindex nvptx options
23471 These options are defined for Nvidia PTX:
23479 Generate code for 32-bit or 64-bit ABI.
23481 @item -misa=@var{ISA-string}
23483 Generate code for given the specified PTX ISA (e.g.@: @samp{sm_35}). ISA
23484 strings must be lower-case. Valid ISA strings include @samp{sm_30} and
23485 @samp{sm_35}. The default ISA is sm_30.
23488 @opindex mmainkernel
23489 Link in code for a __main kernel. This is for stand-alone instead of
23490 offloading execution.
23494 Apply partitioned execution optimizations. This is the default when any
23495 level of optimization is selected.
23498 @opindex msoft-stack
23499 Generate code that does not use @code{.local} memory
23500 directly for stack storage. Instead, a per-warp stack pointer is
23501 maintained explicitly. This enables variable-length stack allocation (with
23502 variable-length arrays or @code{alloca}), and when global memory is used for
23503 underlying storage, makes it possible to access automatic variables from other
23504 threads, or with atomic instructions. This code generation variant is used
23505 for OpenMP offloading, but the option is exposed on its own for the purpose
23506 of testing the compiler; to generate code suitable for linking into programs
23507 using OpenMP offloading, use option @option{-mgomp}.
23509 @item -muniform-simt
23510 @opindex muniform-simt
23511 Switch to code generation variant that allows to execute all threads in each
23512 warp, while maintaining memory state and side effects as if only one thread
23513 in each warp was active outside of OpenMP SIMD regions. All atomic operations
23514 and calls to runtime (malloc, free, vprintf) are conditionally executed (iff
23515 current lane index equals the master lane index), and the register being
23516 assigned is copied via a shuffle instruction from the master lane. Outside of
23517 SIMD regions lane 0 is the master; inside, each thread sees itself as the
23518 master. Shared memory array @code{int __nvptx_uni[]} stores all-zeros or
23519 all-ones bitmasks for each warp, indicating current mode (0 outside of SIMD
23520 regions). Each thread can bitwise-and the bitmask at position @code{tid.y}
23521 with current lane index to compute the master lane index.
23525 Generate code for use in OpenMP offloading: enables @option{-msoft-stack} and
23526 @option{-muniform-simt} options, and selects corresponding multilib variant.
23530 @node OpenRISC Options
23531 @subsection OpenRISC Options
23532 @cindex OpenRISC Options
23534 These options are defined for OpenRISC:
23538 @item -mboard=@var{name}
23540 Configure a board specific runtime. This will be passed to the linker for
23541 newlib board library linking. The default is @code{or1ksim}.
23545 For compatibility, it's always newlib for elf now.
23549 Generate code for hardware which supports divide instructions. This is the
23554 Generate code for hardware which supports multiply instructions. This is the
23559 Generate code for hardware which supports the conditional move (@code{l.cmov})
23564 Generate code for hardware which supports rotate right instructions.
23568 Generate code for hardware which supports sign-extension instructions.
23572 Generate code for hardware which supports set flag immediate (@code{l.sf*i})
23577 Generate code for hardware which supports shift immediate related instructions
23578 (i.e. @code{l.srai}, @code{l.srli}, @code{l.slli}, @code{1.rori}). Note, to
23579 enable generation of the @code{l.rori} instruction the @option{-mror} flag must
23584 Generate code for hardware which requires divide instruction emulation.
23588 Generate code for hardware which requires multiply instruction emulation.
23592 @node PDP-11 Options
23593 @subsection PDP-11 Options
23594 @cindex PDP-11 Options
23596 These options are defined for the PDP-11:
23601 Use hardware FPP floating point. This is the default. (FIS floating
23602 point on the PDP-11/40 is not supported.) Implies -m45.
23605 @opindex msoft-float
23606 Do not use hardware floating point.
23610 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
23614 Return floating-point results in memory. This is the default.
23618 Generate code for a PDP-11/40. Implies -msoft-float -mno-split.
23622 Generate code for a PDP-11/45. This is the default.
23626 Generate code for a PDP-11/10. Implies -msoft-float -mno-split.
23632 Use 16-bit @code{int}. This is the default.
23638 Use 32-bit @code{int}.
23642 Target has split instruction and data space. Implies -m45.
23646 Use Unix assembler syntax.
23650 Use DEC assembler syntax.
23654 Use GNU assembler syntax. This is the default.
23658 Use the new LRA register allocator. By default, the old ``reload''
23662 @node picoChip Options
23663 @subsection picoChip Options
23664 @cindex picoChip options
23666 These @samp{-m} options are defined for picoChip implementations:
23670 @item -mae=@var{ae_type}
23672 Set the instruction set, register set, and instruction scheduling
23673 parameters for array element type @var{ae_type}. Supported values
23674 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
23676 @option{-mae=ANY} selects a completely generic AE type. Code
23677 generated with this option runs on any of the other AE types. The
23678 code is not as efficient as it would be if compiled for a specific
23679 AE type, and some types of operation (e.g., multiplication) do not
23680 work properly on all types of AE.
23682 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
23683 for compiled code, and is the default.
23685 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
23686 option may suffer from poor performance of byte (char) manipulation,
23687 since the DSP AE does not provide hardware support for byte load/stores.
23689 @item -msymbol-as-address
23690 Enable the compiler to directly use a symbol name as an address in a
23691 load/store instruction, without first loading it into a
23692 register. Typically, the use of this option generates larger
23693 programs, which run faster than when the option isn't used. However, the
23694 results vary from program to program, so it is left as a user option,
23695 rather than being permanently enabled.
23697 @item -mno-inefficient-warnings
23698 Disables warnings about the generation of inefficient code. These
23699 warnings can be generated, for example, when compiling code that
23700 performs byte-level memory operations on the MAC AE type. The MAC AE has
23701 no hardware support for byte-level memory operations, so all byte
23702 load/stores must be synthesized from word load/store operations. This is
23703 inefficient and a warning is generated to indicate
23704 that you should rewrite the code to avoid byte operations, or to target
23705 an AE type that has the necessary hardware support. This option disables
23710 @node PowerPC Options
23711 @subsection PowerPC Options
23712 @cindex PowerPC options
23714 These are listed under @xref{RS/6000 and PowerPC Options}.
23716 @node RISC-V Options
23717 @subsection RISC-V Options
23718 @cindex RISC-V Options
23720 These command-line options are defined for RISC-V targets:
23723 @item -mbranch-cost=@var{n}
23724 @opindex mbranch-cost
23725 Set the cost of branches to roughly @var{n} instructions.
23730 When generating PIC code, do or don't allow the use of PLTs. Ignored for
23731 non-PIC. The default is @option{-mplt}.
23733 @item -mabi=@var{ABI-string}
23735 Specify integer and floating-point calling convention. @var{ABI-string}
23736 contains two parts: the size of integer types and the registers used for
23737 floating-point types. For example @samp{-march=rv64ifd -mabi=lp64d} means that
23738 @samp{long} and pointers are 64-bit (implicitly defining @samp{int} to be
23739 32-bit), and that floating-point values up to 64 bits wide are passed in F
23740 registers. Contrast this with @samp{-march=rv64ifd -mabi=lp64f}, which still
23741 allows the compiler to generate code that uses the F and D extensions but only
23742 allows floating-point values up to 32 bits long to be passed in registers; or
23743 @samp{-march=rv64ifd -mabi=lp64}, in which no floating-point arguments will be
23744 passed in registers.
23746 The default for this argument is system dependent, users who want a specific
23747 calling convention should specify one explicitly. The valid calling
23748 conventions are: @samp{ilp32}, @samp{ilp32f}, @samp{ilp32d}, @samp{lp64},
23749 @samp{lp64f}, and @samp{lp64d}. Some calling conventions are impossible to
23750 implement on some ISAs: for example, @samp{-march=rv32if -mabi=ilp32d} is
23751 invalid because the ABI requires 64-bit values be passed in F registers, but F
23752 registers are only 32 bits wide. There is also the @samp{ilp32e} ABI that can
23753 only be used with the @samp{rv32e} architecture. This ABI is not well
23754 specified at present, and is subject to change.
23759 Do or don't use hardware floating-point divide and square root instructions.
23760 This requires the F or D extensions for floating-point registers. The default
23761 is to use them if the specified architecture has these instructions.
23766 Do or don't use hardware instructions for integer division. This requires the
23767 M extension. The default is to use them if the specified architecture has
23768 these instructions.
23770 @item -march=@var{ISA-string}
23772 Generate code for given RISC-V ISA (e.g.@: @samp{rv64im}). ISA strings must be
23773 lower-case. Examples include @samp{rv64i}, @samp{rv32g}, @samp{rv32e}, and
23776 @item -mtune=@var{processor-string}
23778 Optimize the output for the given processor, specified by microarchitecture
23779 name. Permissible values for this option are: @samp{rocket},
23780 @samp{sifive-3-series}, @samp{sifive-5-series}, @samp{sifive-7-series},
23783 When @option{-mtune=} is not specified, the default is @samp{rocket}.
23785 The @samp{size} choice is not intended for use by end-users. This is used
23786 when @option{-Os} is specified. It overrides the instruction cost info
23787 provided by @option{-mtune=}, but does not override the pipeline info. This
23788 helps reduce code size while still giving good performance.
23790 @item -mpreferred-stack-boundary=@var{num}
23791 @opindex mpreferred-stack-boundary
23792 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
23793 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
23794 the default is 4 (16 bytes or 128-bits).
23796 @strong{Warning:} If you use this switch, then you must build all modules with
23797 the same value, including any libraries. This includes the system libraries
23798 and startup modules.
23800 @item -msmall-data-limit=@var{n}
23801 @opindex msmall-data-limit
23802 Put global and static data smaller than @var{n} bytes into a special section
23805 @item -msave-restore
23806 @itemx -mno-save-restore
23807 @opindex msave-restore
23808 Do or don't use smaller but slower prologue and epilogue code that uses
23809 library function calls. The default is to use fast inline prologues and
23812 @item -mstrict-align
23813 @itemx -mno-strict-align
23814 @opindex mstrict-align
23815 Do not or do generate unaligned memory accesses. The default is set depending
23816 on whether the processor we are optimizing for supports fast unaligned access
23819 @item -mcmodel=medlow
23820 @opindex mcmodel=medlow
23821 Generate code for the medium-low code model. The program and its statically
23822 defined symbols must lie within a single 2 GiB address range and must lie
23823 between absolute addresses @minus{}2 GiB and +2 GiB. Programs can be
23824 statically or dynamically linked. This is the default code model.
23826 @item -mcmodel=medany
23827 @opindex mcmodel=medany
23828 Generate code for the medium-any code model. The program and its statically
23829 defined symbols must be within any single 2 GiB address range. Programs can be
23830 statically or dynamically linked.
23832 @item -mexplicit-relocs
23833 @itemx -mno-exlicit-relocs
23834 Use or do not use assembler relocation operators when dealing with symbolic
23835 addresses. The alternative is to use assembler macros instead, which may
23836 limit optimization.
23840 Take advantage of linker relaxations to reduce the number of instructions
23841 required to materialize symbol addresses. The default is to take advantage of
23842 linker relaxations.
23844 @item -memit-attribute
23845 @itemx -mno-emit-attribute
23846 Emit (do not emit) RISC-V attribute to record extra information into ELF
23847 objects. This feature requires at least binutils 2.32.
23851 @subsection RL78 Options
23852 @cindex RL78 Options
23858 Links in additional target libraries to support operation within a
23867 Specifies the type of hardware multiplication and division support to
23868 be used. The simplest is @code{none}, which uses software for both
23869 multiplication and division. This is the default. The @code{g13}
23870 value is for the hardware multiply/divide peripheral found on the
23871 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
23872 the multiplication and division instructions supported by the RL78/G14
23873 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
23874 the value @code{mg10} is an alias for @code{none}.
23876 In addition a C preprocessor macro is defined, based upon the setting
23877 of this option. Possible values are: @code{__RL78_MUL_NONE__},
23878 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
23885 Specifies the RL78 core to target. The default is the G14 core, also
23886 known as an S3 core or just RL78. The G13 or S2 core does not have
23887 multiply or divide instructions, instead it uses a hardware peripheral
23888 for these operations. The G10 or S1 core does not have register
23889 banks, so it uses a different calling convention.
23891 If this option is set it also selects the type of hardware multiply
23892 support to use, unless this is overridden by an explicit
23893 @option{-mmul=none} option on the command line. Thus specifying
23894 @option{-mcpu=g13} enables the use of the G13 hardware multiply
23895 peripheral and specifying @option{-mcpu=g10} disables the use of
23896 hardware multiplications altogether.
23898 Note, although the RL78/G14 core is the default target, specifying
23899 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
23900 change the behavior of the toolchain since it also enables G14
23901 hardware multiply support. If these options are not specified on the
23902 command line then software multiplication routines will be used even
23903 though the code targets the RL78 core. This is for backwards
23904 compatibility with older toolchains which did not have hardware
23905 multiply and divide support.
23907 In addition a C preprocessor macro is defined, based upon the setting
23908 of this option. Possible values are: @code{__RL78_G10__},
23909 @code{__RL78_G13__} or @code{__RL78_G14__}.
23919 These are aliases for the corresponding @option{-mcpu=} option. They
23920 are provided for backwards compatibility.
23924 Allow the compiler to use all of the available registers. By default
23925 registers @code{r24..r31} are reserved for use in interrupt handlers.
23926 With this option enabled these registers can be used in ordinary
23929 @item -m64bit-doubles
23930 @itemx -m32bit-doubles
23931 @opindex m64bit-doubles
23932 @opindex m32bit-doubles
23933 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
23934 or 32 bits (@option{-m32bit-doubles}) in size. The default is
23935 @option{-m32bit-doubles}.
23937 @item -msave-mduc-in-interrupts
23938 @itemx -mno-save-mduc-in-interrupts
23939 @opindex msave-mduc-in-interrupts
23940 @opindex mno-save-mduc-in-interrupts
23941 Specifies that interrupt handler functions should preserve the
23942 MDUC registers. This is only necessary if normal code might use
23943 the MDUC registers, for example because it performs multiplication
23944 and division operations. The default is to ignore the MDUC registers
23945 as this makes the interrupt handlers faster. The target option -mg13
23946 needs to be passed for this to work as this feature is only available
23947 on the G13 target (S2 core). The MDUC registers will only be saved
23948 if the interrupt handler performs a multiplication or division
23949 operation or it calls another function.
23953 @node RS/6000 and PowerPC Options
23954 @subsection IBM RS/6000 and PowerPC Options
23955 @cindex RS/6000 and PowerPC Options
23956 @cindex IBM RS/6000 and PowerPC Options
23958 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
23960 @item -mpowerpc-gpopt
23961 @itemx -mno-powerpc-gpopt
23962 @itemx -mpowerpc-gfxopt
23963 @itemx -mno-powerpc-gfxopt
23966 @itemx -mno-powerpc64
23970 @itemx -mno-popcntb
23972 @itemx -mno-popcntd
23981 @itemx -mno-hard-dfp
23982 @opindex mpowerpc-gpopt
23983 @opindex mno-powerpc-gpopt
23984 @opindex mpowerpc-gfxopt
23985 @opindex mno-powerpc-gfxopt
23986 @opindex mpowerpc64
23987 @opindex mno-powerpc64
23991 @opindex mno-popcntb
23993 @opindex mno-popcntd
23999 @opindex mno-mfpgpr
24001 @opindex mno-hard-dfp
24002 You use these options to specify which instructions are available on the
24003 processor you are using. The default value of these options is
24004 determined when configuring GCC@. Specifying the
24005 @option{-mcpu=@var{cpu_type}} overrides the specification of these
24006 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
24007 rather than the options listed above.
24009 Specifying @option{-mpowerpc-gpopt} allows
24010 GCC to use the optional PowerPC architecture instructions in the
24011 General Purpose group, including floating-point square root. Specifying
24012 @option{-mpowerpc-gfxopt} allows GCC to
24013 use the optional PowerPC architecture instructions in the Graphics
24014 group, including floating-point select.
24016 The @option{-mmfcrf} option allows GCC to generate the move from
24017 condition register field instruction implemented on the POWER4
24018 processor and other processors that support the PowerPC V2.01
24020 The @option{-mpopcntb} option allows GCC to generate the popcount and
24021 double-precision FP reciprocal estimate instruction implemented on the
24022 POWER5 processor and other processors that support the PowerPC V2.02
24024 The @option{-mpopcntd} option allows GCC to generate the popcount
24025 instruction implemented on the POWER7 processor and other processors
24026 that support the PowerPC V2.06 architecture.
24027 The @option{-mfprnd} option allows GCC to generate the FP round to
24028 integer instructions implemented on the POWER5+ processor and other
24029 processors that support the PowerPC V2.03 architecture.
24030 The @option{-mcmpb} option allows GCC to generate the compare bytes
24031 instruction implemented on the POWER6 processor and other processors
24032 that support the PowerPC V2.05 architecture.
24033 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
24034 general-purpose register instructions implemented on the POWER6X
24035 processor and other processors that support the extended PowerPC V2.05
24037 The @option{-mhard-dfp} option allows GCC to generate the decimal
24038 floating-point instructions implemented on some POWER processors.
24040 The @option{-mpowerpc64} option allows GCC to generate the additional
24041 64-bit instructions that are found in the full PowerPC64 architecture
24042 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
24043 @option{-mno-powerpc64}.
24045 @item -mcpu=@var{cpu_type}
24047 Set architecture type, register usage, and
24048 instruction scheduling parameters for machine type @var{cpu_type}.
24049 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
24050 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
24051 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
24052 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
24053 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
24054 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
24055 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
24056 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
24057 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
24058 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
24059 @samp{power9}, @samp{powerpc}, @samp{powerpc64}, @samp{powerpc64le},
24060 @samp{rs64}, and @samp{native}.
24062 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
24063 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
24064 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
24065 architecture machine types, with an appropriate, generic processor
24066 model assumed for scheduling purposes.
24068 Specifying @samp{native} as cpu type detects and selects the
24069 architecture option that corresponds to the host processor of the
24070 system performing the compilation.
24071 @option{-mcpu=native} has no effect if GCC does not recognize the
24074 The other options specify a specific processor. Code generated under
24075 those options runs best on that processor, and may not run at all on
24078 The @option{-mcpu} options automatically enable or disable the
24081 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
24082 -mpopcntb -mpopcntd -mpowerpc64 @gol
24083 -mpowerpc-gpopt -mpowerpc-gfxopt @gol
24084 -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
24085 -mcrypto -mhtm -mpower8-fusion -mpower8-vector @gol
24086 -mquad-memory -mquad-memory-atomic -mfloat128 -mfloat128-hardware}
24088 The particular options set for any particular CPU varies between
24089 compiler versions, depending on what setting seems to produce optimal
24090 code for that CPU; it doesn't necessarily reflect the actual hardware's
24091 capabilities. If you wish to set an individual option to a particular
24092 value, you may specify it after the @option{-mcpu} option, like
24093 @option{-mcpu=970 -mno-altivec}.
24095 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
24096 not enabled or disabled by the @option{-mcpu} option at present because
24097 AIX does not have full support for these options. You may still
24098 enable or disable them individually if you're sure it'll work in your
24101 @item -mtune=@var{cpu_type}
24103 Set the instruction scheduling parameters for machine type
24104 @var{cpu_type}, but do not set the architecture type or register usage,
24105 as @option{-mcpu=@var{cpu_type}} does. The same
24106 values for @var{cpu_type} are used for @option{-mtune} as for
24107 @option{-mcpu}. If both are specified, the code generated uses the
24108 architecture and registers set by @option{-mcpu}, but the
24109 scheduling parameters set by @option{-mtune}.
24111 @item -mcmodel=small
24112 @opindex mcmodel=small
24113 Generate PowerPC64 code for the small model: The TOC is limited to
24116 @item -mcmodel=medium
24117 @opindex mcmodel=medium
24118 Generate PowerPC64 code for the medium model: The TOC and other static
24119 data may be up to a total of 4G in size. This is the default for 64-bit
24122 @item -mcmodel=large
24123 @opindex mcmodel=large
24124 Generate PowerPC64 code for the large model: The TOC may be up to 4G
24125 in size. Other data and code is only limited by the 64-bit address
24129 @itemx -mno-altivec
24131 @opindex mno-altivec
24132 Generate code that uses (does not use) AltiVec instructions, and also
24133 enable the use of built-in functions that allow more direct access to
24134 the AltiVec instruction set. You may also need to set
24135 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
24138 When @option{-maltivec} is used, the element order for AltiVec intrinsics
24139 such as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
24140 match array element order corresponding to the endianness of the
24141 target. That is, element zero identifies the leftmost element in a
24142 vector register when targeting a big-endian platform, and identifies
24143 the rightmost element in a vector register when targeting a
24144 little-endian platform.
24149 @opindex mno-vrsave
24150 Generate VRSAVE instructions when generating AltiVec code.
24153 @opindex msecure-plt
24154 Generate code that allows @command{ld} and @command{ld.so}
24155 to build executables and shared
24156 libraries with non-executable @code{.plt} and @code{.got} sections.
24158 32-bit SYSV ABI option.
24162 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
24164 requires @code{.plt} and @code{.got}
24165 sections that are both writable and executable.
24166 This is a PowerPC 32-bit SYSV ABI option.
24172 This switch enables or disables the generation of ISEL instructions.
24178 Generate code that uses (does not use) vector/scalar (VSX)
24179 instructions, and also enable the use of built-in functions that allow
24180 more direct access to the VSX instruction set.
24185 @opindex mno-crypto
24186 Enable the use (disable) of the built-in functions that allow direct
24187 access to the cryptographic instructions that were added in version
24188 2.07 of the PowerPC ISA.
24194 Enable (disable) the use of the built-in functions that allow direct
24195 access to the Hardware Transactional Memory (HTM) instructions that
24196 were added in version 2.07 of the PowerPC ISA.
24198 @item -mpower8-fusion
24199 @itemx -mno-power8-fusion
24200 @opindex mpower8-fusion
24201 @opindex mno-power8-fusion
24202 Generate code that keeps (does not keeps) some integer operations
24203 adjacent so that the instructions can be fused together on power8 and
24206 @item -mpower8-vector
24207 @itemx -mno-power8-vector
24208 @opindex mpower8-vector
24209 @opindex mno-power8-vector
24210 Generate code that uses (does not use) the vector and scalar
24211 instructions that were added in version 2.07 of the PowerPC ISA. Also
24212 enable the use of built-in functions that allow more direct access to
24213 the vector instructions.
24215 @item -mquad-memory
24216 @itemx -mno-quad-memory
24217 @opindex mquad-memory
24218 @opindex mno-quad-memory
24219 Generate code that uses (does not use) the non-atomic quad word memory
24220 instructions. The @option{-mquad-memory} option requires use of
24223 @item -mquad-memory-atomic
24224 @itemx -mno-quad-memory-atomic
24225 @opindex mquad-memory-atomic
24226 @opindex mno-quad-memory-atomic
24227 Generate code that uses (does not use) the atomic quad word memory
24228 instructions. The @option{-mquad-memory-atomic} option requires use of
24232 @itemx -mno-float128
24234 @opindex mno-float128
24235 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
24236 and use either software emulation for IEEE 128-bit floating point or
24237 hardware instructions.
24239 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7},
24240 @option{-mcpu=power8}), or @option{-mcpu=power9} must be enabled to
24241 use the IEEE 128-bit floating point support. The IEEE 128-bit
24242 floating point support only works on PowerPC Linux systems.
24244 The default for @option{-mfloat128} is enabled on PowerPC Linux
24245 systems using the VSX instruction set, and disabled on other systems.
24247 If you use the ISA 3.0 instruction set (@option{-mpower9-vector} or
24248 @option{-mcpu=power9}) on a 64-bit system, the IEEE 128-bit floating
24249 point support will also enable the generation of ISA 3.0 IEEE 128-bit
24250 floating point instructions. Otherwise, if you do not specify to
24251 generate ISA 3.0 instructions or you are targeting a 32-bit big endian
24252 system, IEEE 128-bit floating point will be done with software
24255 @item -mfloat128-hardware
24256 @itemx -mno-float128-hardware
24257 @opindex mfloat128-hardware
24258 @opindex mno-float128-hardware
24259 Enable/disable using ISA 3.0 hardware instructions to support the
24260 @var{__float128} data type.
24262 The default for @option{-mfloat128-hardware} is enabled on PowerPC
24263 Linux systems using the ISA 3.0 instruction set, and disabled on other
24270 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
24271 targets (including GNU/Linux). The 32-bit environment sets int, long
24272 and pointer to 32 bits and generates code that runs on any PowerPC
24273 variant. The 64-bit environment sets int to 32 bits and long and
24274 pointer to 64 bits, and generates code for PowerPC64, as for
24275 @option{-mpowerpc64}.
24278 @itemx -mno-fp-in-toc
24279 @itemx -mno-sum-in-toc
24280 @itemx -mminimal-toc
24282 @opindex mno-fp-in-toc
24283 @opindex mno-sum-in-toc
24284 @opindex mminimal-toc
24285 Modify generation of the TOC (Table Of Contents), which is created for
24286 every executable file. The @option{-mfull-toc} option is selected by
24287 default. In that case, GCC allocates at least one TOC entry for
24288 each unique non-automatic variable reference in your program. GCC
24289 also places floating-point constants in the TOC@. However, only
24290 16,384 entries are available in the TOC@.
24292 If you receive a linker error message that saying you have overflowed
24293 the available TOC space, you can reduce the amount of TOC space used
24294 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
24295 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
24296 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
24297 generate code to calculate the sum of an address and a constant at
24298 run time instead of putting that sum into the TOC@. You may specify one
24299 or both of these options. Each causes GCC to produce very slightly
24300 slower and larger code at the expense of conserving TOC space.
24302 If you still run out of space in the TOC even when you specify both of
24303 these options, specify @option{-mminimal-toc} instead. This option causes
24304 GCC to make only one TOC entry for every file. When you specify this
24305 option, GCC produces code that is slower and larger but which
24306 uses extremely little TOC space. You may wish to use this option
24307 only on files that contain less frequently-executed code.
24313 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
24314 @code{long} type, and the infrastructure needed to support them.
24315 Specifying @option{-maix64} implies @option{-mpowerpc64},
24316 while @option{-maix32} disables the 64-bit ABI and
24317 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
24320 @itemx -mno-xl-compat
24321 @opindex mxl-compat
24322 @opindex mno-xl-compat
24323 Produce code that conforms more closely to IBM XL compiler semantics
24324 when using AIX-compatible ABI@. Pass floating-point arguments to
24325 prototyped functions beyond the register save area (RSA) on the stack
24326 in addition to argument FPRs. Do not assume that most significant
24327 double in 128-bit long double value is properly rounded when comparing
24328 values and converting to double. Use XL symbol names for long double
24331 The AIX calling convention was extended but not initially documented to
24332 handle an obscure K&R C case of calling a function that takes the
24333 address of its arguments with fewer arguments than declared. IBM XL
24334 compilers access floating-point arguments that do not fit in the
24335 RSA from the stack when a subroutine is compiled without
24336 optimization. Because always storing floating-point arguments on the
24337 stack is inefficient and rarely needed, this option is not enabled by
24338 default and only is necessary when calling subroutines compiled by IBM
24339 XL compilers without optimization.
24343 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
24344 application written to use message passing with special startup code to
24345 enable the application to run. The system must have PE installed in the
24346 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
24347 must be overridden with the @option{-specs=} option to specify the
24348 appropriate directory location. The Parallel Environment does not
24349 support threads, so the @option{-mpe} option and the @option{-pthread}
24350 option are incompatible.
24352 @item -malign-natural
24353 @itemx -malign-power
24354 @opindex malign-natural
24355 @opindex malign-power
24356 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
24357 @option{-malign-natural} overrides the ABI-defined alignment of larger
24358 types, such as floating-point doubles, on their natural size-based boundary.
24359 The option @option{-malign-power} instructs GCC to follow the ABI-specified
24360 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
24362 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
24366 @itemx -mhard-float
24367 @opindex msoft-float
24368 @opindex mhard-float
24369 Generate code that does not use (uses) the floating-point register set.
24370 Software floating-point emulation is provided if you use the
24371 @option{-msoft-float} option, and pass the option to GCC when linking.
24374 @itemx -mno-multiple
24376 @opindex mno-multiple
24377 Generate code that uses (does not use) the load multiple word
24378 instructions and the store multiple word instructions. These
24379 instructions are generated by default on POWER systems, and not
24380 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
24381 PowerPC systems, since those instructions do not work when the
24382 processor is in little-endian mode. The exceptions are PPC740 and
24383 PPC750 which permit these instructions in little-endian mode.
24388 @opindex mno-update
24389 Generate code that uses (does not use) the load or store instructions
24390 that update the base register to the address of the calculated memory
24391 location. These instructions are generated by default. If you use
24392 @option{-mno-update}, there is a small window between the time that the
24393 stack pointer is updated and the address of the previous frame is
24394 stored, which means code that walks the stack frame across interrupts or
24395 signals may get corrupted data.
24397 @item -mavoid-indexed-addresses
24398 @itemx -mno-avoid-indexed-addresses
24399 @opindex mavoid-indexed-addresses
24400 @opindex mno-avoid-indexed-addresses
24401 Generate code that tries to avoid (not avoid) the use of indexed load
24402 or store instructions. These instructions can incur a performance
24403 penalty on Power6 processors in certain situations, such as when
24404 stepping through large arrays that cross a 16M boundary. This option
24405 is enabled by default when targeting Power6 and disabled otherwise.
24408 @itemx -mno-fused-madd
24409 @opindex mfused-madd
24410 @opindex mno-fused-madd
24411 Generate code that uses (does not use) the floating-point multiply and
24412 accumulate instructions. These instructions are generated by default
24413 if hardware floating point is used. The machine-dependent
24414 @option{-mfused-madd} option is now mapped to the machine-independent
24415 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
24416 mapped to @option{-ffp-contract=off}.
24422 Generate code that uses (does not use) the half-word multiply and
24423 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
24424 These instructions are generated by default when targeting those
24431 Generate code that uses (does not use) the string-search @samp{dlmzb}
24432 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
24433 generated by default when targeting those processors.
24435 @item -mno-bit-align
24437 @opindex mno-bit-align
24438 @opindex mbit-align
24439 On System V.4 and embedded PowerPC systems do not (do) force structures
24440 and unions that contain bit-fields to be aligned to the base type of the
24443 For example, by default a structure containing nothing but 8
24444 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
24445 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
24446 the structure is aligned to a 1-byte boundary and is 1 byte in
24449 @item -mno-strict-align
24450 @itemx -mstrict-align
24451 @opindex mno-strict-align
24452 @opindex mstrict-align
24453 On System V.4 and embedded PowerPC systems do not (do) assume that
24454 unaligned memory references are handled by the system.
24456 @item -mrelocatable
24457 @itemx -mno-relocatable
24458 @opindex mrelocatable
24459 @opindex mno-relocatable
24460 Generate code that allows (does not allow) a static executable to be
24461 relocated to a different address at run time. A simple embedded
24462 PowerPC system loader should relocate the entire contents of
24463 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
24464 a table of 32-bit addresses generated by this option. For this to
24465 work, all objects linked together must be compiled with
24466 @option{-mrelocatable} or @option{-mrelocatable-lib}.
24467 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
24469 @item -mrelocatable-lib
24470 @itemx -mno-relocatable-lib
24471 @opindex mrelocatable-lib
24472 @opindex mno-relocatable-lib
24473 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
24474 @code{.fixup} section to allow static executables to be relocated at
24475 run time, but @option{-mrelocatable-lib} does not use the smaller stack
24476 alignment of @option{-mrelocatable}. Objects compiled with
24477 @option{-mrelocatable-lib} may be linked with objects compiled with
24478 any combination of the @option{-mrelocatable} options.
24484 On System V.4 and embedded PowerPC systems do not (do) assume that
24485 register 2 contains a pointer to a global area pointing to the addresses
24486 used in the program.
24489 @itemx -mlittle-endian
24491 @opindex mlittle-endian
24492 On System V.4 and embedded PowerPC systems compile code for the
24493 processor in little-endian mode. The @option{-mlittle-endian} option is
24494 the same as @option{-mlittle}.
24497 @itemx -mbig-endian
24499 @opindex mbig-endian
24500 On System V.4 and embedded PowerPC systems compile code for the
24501 processor in big-endian mode. The @option{-mbig-endian} option is
24502 the same as @option{-mbig}.
24504 @item -mdynamic-no-pic
24505 @opindex mdynamic-no-pic
24506 On Darwin and Mac OS X systems, compile code so that it is not
24507 relocatable, but that its external references are relocatable. The
24508 resulting code is suitable for applications, but not shared
24511 @item -msingle-pic-base
24512 @opindex msingle-pic-base
24513 Treat the register used for PIC addressing as read-only, rather than
24514 loading it in the prologue for each function. The runtime system is
24515 responsible for initializing this register with an appropriate value
24516 before execution begins.
24518 @item -mprioritize-restricted-insns=@var{priority}
24519 @opindex mprioritize-restricted-insns
24520 This option controls the priority that is assigned to
24521 dispatch-slot restricted instructions during the second scheduling
24522 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
24523 or @samp{2} to assign no, highest, or second-highest (respectively)
24524 priority to dispatch-slot restricted
24527 @item -msched-costly-dep=@var{dependence_type}
24528 @opindex msched-costly-dep
24529 This option controls which dependences are considered costly
24530 by the target during instruction scheduling. The argument
24531 @var{dependence_type} takes one of the following values:
24535 No dependence is costly.
24538 All dependences are costly.
24540 @item @samp{true_store_to_load}
24541 A true dependence from store to load is costly.
24543 @item @samp{store_to_load}
24544 Any dependence from store to load is costly.
24547 Any dependence for which the latency is greater than or equal to
24548 @var{number} is costly.
24551 @item -minsert-sched-nops=@var{scheme}
24552 @opindex minsert-sched-nops
24553 This option controls which NOP insertion scheme is used during
24554 the second scheduling pass. The argument @var{scheme} takes one of the
24562 Pad with NOPs any dispatch group that has vacant issue slots,
24563 according to the scheduler's grouping.
24565 @item @samp{regroup_exact}
24566 Insert NOPs to force costly dependent insns into
24567 separate groups. Insert exactly as many NOPs as needed to force an insn
24568 to a new group, according to the estimated processor grouping.
24571 Insert NOPs to force costly dependent insns into
24572 separate groups. Insert @var{number} NOPs to force an insn to a new group.
24576 @opindex mcall-sysv
24577 On System V.4 and embedded PowerPC systems compile code using calling
24578 conventions that adhere to the March 1995 draft of the System V
24579 Application Binary Interface, PowerPC processor supplement. This is the
24580 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
24582 @item -mcall-sysv-eabi
24584 @opindex mcall-sysv-eabi
24585 @opindex mcall-eabi
24586 Specify both @option{-mcall-sysv} and @option{-meabi} options.
24588 @item -mcall-sysv-noeabi
24589 @opindex mcall-sysv-noeabi
24590 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
24592 @item -mcall-aixdesc
24594 On System V.4 and embedded PowerPC systems compile code for the AIX
24598 @opindex mcall-linux
24599 On System V.4 and embedded PowerPC systems compile code for the
24600 Linux-based GNU system.
24602 @item -mcall-freebsd
24603 @opindex mcall-freebsd
24604 On System V.4 and embedded PowerPC systems compile code for the
24605 FreeBSD operating system.
24607 @item -mcall-netbsd
24608 @opindex mcall-netbsd
24609 On System V.4 and embedded PowerPC systems compile code for the
24610 NetBSD operating system.
24612 @item -mcall-openbsd
24613 @opindex mcall-netbsd
24614 On System V.4 and embedded PowerPC systems compile code for the
24615 OpenBSD operating system.
24617 @item -mtraceback=@var{traceback_type}
24618 @opindex mtraceback
24619 Select the type of traceback table. Valid values for @var{traceback_type}
24620 are @samp{full}, @samp{part}, and @samp{no}.
24622 @item -maix-struct-return
24623 @opindex maix-struct-return
24624 Return all structures in memory (as specified by the AIX ABI)@.
24626 @item -msvr4-struct-return
24627 @opindex msvr4-struct-return
24628 Return structures smaller than 8 bytes in registers (as specified by the
24631 @item -mabi=@var{abi-type}
24633 Extend the current ABI with a particular extension, or remove such extension.
24634 Valid values are @samp{altivec}, @samp{no-altivec},
24635 @samp{ibmlongdouble}, @samp{ieeelongdouble},
24636 @samp{elfv1}, @samp{elfv2}@.
24638 @item -mabi=ibmlongdouble
24639 @opindex mabi=ibmlongdouble
24640 Change the current ABI to use IBM extended-precision long double.
24641 This is not likely to work if your system defaults to using IEEE
24642 extended-precision long double. If you change the long double type
24643 from IEEE extended-precision, the compiler will issue a warning unless
24644 you use the @option{-Wno-psabi} option. Requires @option{-mlong-double-128}
24647 @item -mabi=ieeelongdouble
24648 @opindex mabi=ieeelongdouble
24649 Change the current ABI to use IEEE extended-precision long double.
24650 This is not likely to work if your system defaults to using IBM
24651 extended-precision long double. If you change the long double type
24652 from IBM extended-precision, the compiler will issue a warning unless
24653 you use the @option{-Wno-psabi} option. Requires @option{-mlong-double-128}
24657 @opindex mabi=elfv1
24658 Change the current ABI to use the ELFv1 ABI.
24659 This is the default ABI for big-endian PowerPC 64-bit Linux.
24660 Overriding the default ABI requires special system support and is
24661 likely to fail in spectacular ways.
24664 @opindex mabi=elfv2
24665 Change the current ABI to use the ELFv2 ABI.
24666 This is the default ABI for little-endian PowerPC 64-bit Linux.
24667 Overriding the default ABI requires special system support and is
24668 likely to fail in spectacular ways.
24670 @item -mgnu-attribute
24671 @itemx -mno-gnu-attribute
24672 @opindex mgnu-attribute
24673 @opindex mno-gnu-attribute
24674 Emit .gnu_attribute assembly directives to set tag/value pairs in a
24675 .gnu.attributes section that specify ABI variations in function
24676 parameters or return values.
24679 @itemx -mno-prototype
24680 @opindex mprototype
24681 @opindex mno-prototype
24682 On System V.4 and embedded PowerPC systems assume that all calls to
24683 variable argument functions are properly prototyped. Otherwise, the
24684 compiler must insert an instruction before every non-prototyped call to
24685 set or clear bit 6 of the condition code register (@code{CR}) to
24686 indicate whether floating-point values are passed in the floating-point
24687 registers in case the function takes variable arguments. With
24688 @option{-mprototype}, only calls to prototyped variable argument functions
24689 set or clear the bit.
24693 On embedded PowerPC systems, assume that the startup module is called
24694 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
24695 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
24700 On embedded PowerPC systems, assume that the startup module is called
24701 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
24706 On embedded PowerPC systems, assume that the startup module is called
24707 @file{crt0.o} and the standard C libraries are @file{libads.a} and
24710 @item -myellowknife
24711 @opindex myellowknife
24712 On embedded PowerPC systems, assume that the startup module is called
24713 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
24718 On System V.4 and embedded PowerPC systems, specify that you are
24719 compiling for a VxWorks system.
24723 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
24724 header to indicate that @samp{eabi} extended relocations are used.
24730 On System V.4 and embedded PowerPC systems do (do not) adhere to the
24731 Embedded Applications Binary Interface (EABI), which is a set of
24732 modifications to the System V.4 specifications. Selecting @option{-meabi}
24733 means that the stack is aligned to an 8-byte boundary, a function
24734 @code{__eabi} is called from @code{main} to set up the EABI
24735 environment, and the @option{-msdata} option can use both @code{r2} and
24736 @code{r13} to point to two separate small data areas. Selecting
24737 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
24738 no EABI initialization function is called from @code{main}, and the
24739 @option{-msdata} option only uses @code{r13} to point to a single
24740 small data area. The @option{-meabi} option is on by default if you
24741 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
24744 @opindex msdata=eabi
24745 On System V.4 and embedded PowerPC systems, put small initialized
24746 @code{const} global and static data in the @code{.sdata2} section, which
24747 is pointed to by register @code{r2}. Put small initialized
24748 non-@code{const} global and static data in the @code{.sdata} section,
24749 which is pointed to by register @code{r13}. Put small uninitialized
24750 global and static data in the @code{.sbss} section, which is adjacent to
24751 the @code{.sdata} section. The @option{-msdata=eabi} option is
24752 incompatible with the @option{-mrelocatable} option. The
24753 @option{-msdata=eabi} option also sets the @option{-memb} option.
24756 @opindex msdata=sysv
24757 On System V.4 and embedded PowerPC systems, put small global and static
24758 data in the @code{.sdata} section, which is pointed to by register
24759 @code{r13}. Put small uninitialized global and static data in the
24760 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
24761 The @option{-msdata=sysv} option is incompatible with the
24762 @option{-mrelocatable} option.
24764 @item -msdata=default
24766 @opindex msdata=default
24768 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
24769 compile code the same as @option{-msdata=eabi}, otherwise compile code the
24770 same as @option{-msdata=sysv}.
24773 @opindex msdata=data
24774 On System V.4 and embedded PowerPC systems, put small global
24775 data in the @code{.sdata} section. Put small uninitialized global
24776 data in the @code{.sbss} section. Do not use register @code{r13}
24777 to address small data however. This is the default behavior unless
24778 other @option{-msdata} options are used.
24782 @opindex msdata=none
24784 On embedded PowerPC systems, put all initialized global and static data
24785 in the @code{.data} section, and all uninitialized data in the
24786 @code{.bss} section.
24788 @item -mreadonly-in-sdata
24789 @opindex mreadonly-in-sdata
24790 @opindex mno-readonly-in-sdata
24791 Put read-only objects in the @code{.sdata} section as well. This is the
24794 @item -mblock-move-inline-limit=@var{num}
24795 @opindex mblock-move-inline-limit
24796 Inline all block moves (such as calls to @code{memcpy} or structure
24797 copies) less than or equal to @var{num} bytes. The minimum value for
24798 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
24799 targets. The default value is target-specific.
24801 @item -mblock-compare-inline-limit=@var{num}
24802 @opindex mblock-compare-inline-limit
24803 Generate non-looping inline code for all block compares (such as calls
24804 to @code{memcmp} or structure compares) less than or equal to @var{num}
24805 bytes. If @var{num} is 0, all inline expansion (non-loop and loop) of
24806 block compare is disabled. The default value is target-specific.
24808 @item -mblock-compare-inline-loop-limit=@var{num}
24809 @opindex mblock-compare-inline-loop-limit
24810 Generate an inline expansion using loop code for all block compares that
24811 are less than or equal to @var{num} bytes, but greater than the limit
24812 for non-loop inline block compare expansion. If the block length is not
24813 constant, at most @var{num} bytes will be compared before @code{memcmp}
24814 is called to compare the remainder of the block. The default value is
24817 @item -mstring-compare-inline-limit=@var{num}
24818 @opindex mstring-compare-inline-limit
24819 Compare at most @var{num} string bytes with inline code.
24820 If the difference or end of string is not found at the
24821 end of the inline compare a call to @code{strcmp} or @code{strncmp} will
24822 take care of the rest of the comparison. The default is 64 bytes.
24826 @cindex smaller data references (PowerPC)
24827 @cindex .sdata/.sdata2 references (PowerPC)
24828 On embedded PowerPC systems, put global and static items less than or
24829 equal to @var{num} bytes into the small data or BSS sections instead of
24830 the normal data or BSS section. By default, @var{num} is 8. The
24831 @option{-G @var{num}} switch is also passed to the linker.
24832 All modules should be compiled with the same @option{-G @var{num}} value.
24835 @itemx -mno-regnames
24837 @opindex mno-regnames
24838 On System V.4 and embedded PowerPC systems do (do not) emit register
24839 names in the assembly language output using symbolic forms.
24842 @itemx -mno-longcall
24844 @opindex mno-longcall
24845 By default assume that all calls are far away so that a longer and more
24846 expensive calling sequence is required. This is required for calls
24847 farther than 32 megabytes (33,554,432 bytes) from the current location.
24848 A short call is generated if the compiler knows
24849 the call cannot be that far away. This setting can be overridden by
24850 the @code{shortcall} function attribute, or by @code{#pragma
24853 Some linkers are capable of detecting out-of-range calls and generating
24854 glue code on the fly. On these systems, long calls are unnecessary and
24855 generate slower code. As of this writing, the AIX linker can do this,
24856 as can the GNU linker for PowerPC/64. It is planned to add this feature
24857 to the GNU linker for 32-bit PowerPC systems as well.
24859 On PowerPC64 ELFv2 and 32-bit PowerPC systems with newer GNU linkers,
24860 GCC can generate long calls using an inline PLT call sequence (see
24861 @option{-mpltseq}). PowerPC with @option{-mbss-plt} and PowerPC64
24862 ELFv1 (big-endian) do not support inline PLT calls.
24864 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
24865 callee, L42}, plus a @dfn{branch island} (glue code). The two target
24866 addresses represent the callee and the branch island. The
24867 Darwin/PPC linker prefers the first address and generates a @code{bl
24868 callee} if the PPC @code{bl} instruction reaches the callee directly;
24869 otherwise, the linker generates @code{bl L42} to call the branch
24870 island. The branch island is appended to the body of the
24871 calling function; it computes the full 32-bit address of the callee
24874 On Mach-O (Darwin) systems, this option directs the compiler emit to
24875 the glue for every direct call, and the Darwin linker decides whether
24876 to use or discard it.
24878 In the future, GCC may ignore all longcall specifications
24879 when the linker is known to generate glue.
24884 @opindex mno-pltseq
24885 Implement (do not implement) -fno-plt and long calls using an inline
24886 PLT call sequence that supports lazy linking and long calls to
24887 functions in dlopen'd shared libraries. Inline PLT calls are only
24888 supported on PowerPC64 ELFv2 and 32-bit PowerPC systems with newer GNU
24889 linkers, and are enabled by default if the support is detected when
24890 configuring GCC, and, in the case of 32-bit PowerPC, if GCC is
24891 configured with @option{--enable-secureplt}. @option{-mpltseq} code
24892 and @option{-mbss-plt} 32-bit PowerPC relocatable objects may not be
24895 @item -mtls-markers
24896 @itemx -mno-tls-markers
24897 @opindex mtls-markers
24898 @opindex mno-tls-markers
24899 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
24900 specifying the function argument. The relocation allows the linker to
24901 reliably associate function call with argument setup instructions for
24902 TLS optimization, which in turn allows GCC to better schedule the
24908 This option enables use of the reciprocal estimate and
24909 reciprocal square root estimate instructions with additional
24910 Newton-Raphson steps to increase precision instead of doing a divide or
24911 square root and divide for floating-point arguments. You should use
24912 the @option{-ffast-math} option when using @option{-mrecip} (or at
24913 least @option{-funsafe-math-optimizations},
24914 @option{-ffinite-math-only}, @option{-freciprocal-math} and
24915 @option{-fno-trapping-math}). Note that while the throughput of the
24916 sequence is generally higher than the throughput of the non-reciprocal
24917 instruction, the precision of the sequence can be decreased by up to 2
24918 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
24921 @item -mrecip=@var{opt}
24922 @opindex mrecip=opt
24923 This option controls which reciprocal estimate instructions
24924 may be used. @var{opt} is a comma-separated list of options, which may
24925 be preceded by a @code{!} to invert the option:
24930 Enable all estimate instructions.
24933 Enable the default instructions, equivalent to @option{-mrecip}.
24936 Disable all estimate instructions, equivalent to @option{-mno-recip}.
24939 Enable the reciprocal approximation instructions for both
24940 single and double precision.
24943 Enable the single-precision reciprocal approximation instructions.
24946 Enable the double-precision reciprocal approximation instructions.
24949 Enable the reciprocal square root approximation instructions for both
24950 single and double precision.
24953 Enable the single-precision reciprocal square root approximation instructions.
24956 Enable the double-precision reciprocal square root approximation instructions.
24960 So, for example, @option{-mrecip=all,!rsqrtd} enables
24961 all of the reciprocal estimate instructions, except for the
24962 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
24963 which handle the double-precision reciprocal square root calculations.
24965 @item -mrecip-precision
24966 @itemx -mno-recip-precision
24967 @opindex mrecip-precision
24968 Assume (do not assume) that the reciprocal estimate instructions
24969 provide higher-precision estimates than is mandated by the PowerPC
24970 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
24971 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
24972 The double-precision square root estimate instructions are not generated by
24973 default on low-precision machines, since they do not provide an
24974 estimate that converges after three steps.
24976 @item -mveclibabi=@var{type}
24977 @opindex mveclibabi
24978 Specifies the ABI type to use for vectorizing intrinsics using an
24979 external library. The only type supported at present is @samp{mass},
24980 which specifies to use IBM's Mathematical Acceleration Subsystem
24981 (MASS) libraries for vectorizing intrinsics using external libraries.
24982 GCC currently emits calls to @code{acosd2}, @code{acosf4},
24983 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
24984 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
24985 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
24986 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
24987 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
24988 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
24989 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
24990 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
24991 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
24992 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
24993 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
24994 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
24995 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
24996 for power7. Both @option{-ftree-vectorize} and
24997 @option{-funsafe-math-optimizations} must also be enabled. The MASS
24998 libraries must be specified at link time.
25003 Generate (do not generate) the @code{friz} instruction when the
25004 @option{-funsafe-math-optimizations} option is used to optimize
25005 rounding of floating-point values to 64-bit integer and back to floating
25006 point. The @code{friz} instruction does not return the same value if
25007 the floating-point number is too large to fit in an integer.
25009 @item -mpointers-to-nested-functions
25010 @itemx -mno-pointers-to-nested-functions
25011 @opindex mpointers-to-nested-functions
25012 Generate (do not generate) code to load up the static chain register
25013 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
25014 systems where a function pointer points to a 3-word descriptor giving
25015 the function address, TOC value to be loaded in register @code{r2}, and
25016 static chain value to be loaded in register @code{r11}. The
25017 @option{-mpointers-to-nested-functions} is on by default. You cannot
25018 call through pointers to nested functions or pointers
25019 to functions compiled in other languages that use the static chain if
25020 you use @option{-mno-pointers-to-nested-functions}.
25022 @item -msave-toc-indirect
25023 @itemx -mno-save-toc-indirect
25024 @opindex msave-toc-indirect
25025 Generate (do not generate) code to save the TOC value in the reserved
25026 stack location in the function prologue if the function calls through
25027 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
25028 saved in the prologue, it is saved just before the call through the
25029 pointer. The @option{-mno-save-toc-indirect} option is the default.
25031 @item -mcompat-align-parm
25032 @itemx -mno-compat-align-parm
25033 @opindex mcompat-align-parm
25034 Generate (do not generate) code to pass structure parameters with a
25035 maximum alignment of 64 bits, for compatibility with older versions
25038 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
25039 structure parameter on a 128-bit boundary when that structure contained
25040 a member requiring 128-bit alignment. This is corrected in more
25041 recent versions of GCC. This option may be used to generate code
25042 that is compatible with functions compiled with older versions of
25045 The @option{-mno-compat-align-parm} option is the default.
25047 @item -mstack-protector-guard=@var{guard}
25048 @itemx -mstack-protector-guard-reg=@var{reg}
25049 @itemx -mstack-protector-guard-offset=@var{offset}
25050 @itemx -mstack-protector-guard-symbol=@var{symbol}
25051 @opindex mstack-protector-guard
25052 @opindex mstack-protector-guard-reg
25053 @opindex mstack-protector-guard-offset
25054 @opindex mstack-protector-guard-symbol
25055 Generate stack protection code using canary at @var{guard}. Supported
25056 locations are @samp{global} for global canary or @samp{tls} for per-thread
25057 canary in the TLS block (the default with GNU libc version 2.4 or later).
25059 With the latter choice the options
25060 @option{-mstack-protector-guard-reg=@var{reg}} and
25061 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
25062 which register to use as base register for reading the canary, and from what
25063 offset from that base register. The default for those is as specified in the
25064 relevant ABI. @option{-mstack-protector-guard-symbol=@var{symbol}} overrides
25065 the offset with a symbol reference to a canary in the TLS block.
25069 @subsection RX Options
25072 These command-line options are defined for RX targets:
25075 @item -m64bit-doubles
25076 @itemx -m32bit-doubles
25077 @opindex m64bit-doubles
25078 @opindex m32bit-doubles
25079 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
25080 or 32 bits (@option{-m32bit-doubles}) in size. The default is
25081 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
25082 works on 32-bit values, which is why the default is
25083 @option{-m32bit-doubles}.
25089 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
25090 floating-point hardware. The default is enabled for the RX600
25091 series and disabled for the RX200 series.
25093 Floating-point instructions are only generated for 32-bit floating-point
25094 values, however, so the FPU hardware is not used for doubles if the
25095 @option{-m64bit-doubles} option is used.
25097 @emph{Note} If the @option{-fpu} option is enabled then
25098 @option{-funsafe-math-optimizations} is also enabled automatically.
25099 This is because the RX FPU instructions are themselves unsafe.
25101 @item -mcpu=@var{name}
25103 Selects the type of RX CPU to be targeted. Currently three types are
25104 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
25105 the specific @samp{RX610} CPU. The default is @samp{RX600}.
25107 The only difference between @samp{RX600} and @samp{RX610} is that the
25108 @samp{RX610} does not support the @code{MVTIPL} instruction.
25110 The @samp{RX200} series does not have a hardware floating-point unit
25111 and so @option{-nofpu} is enabled by default when this type is
25114 @item -mbig-endian-data
25115 @itemx -mlittle-endian-data
25116 @opindex mbig-endian-data
25117 @opindex mlittle-endian-data
25118 Store data (but not code) in the big-endian format. The default is
25119 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
25122 @item -msmall-data-limit=@var{N}
25123 @opindex msmall-data-limit
25124 Specifies the maximum size in bytes of global and static variables
25125 which can be placed into the small data area. Using the small data
25126 area can lead to smaller and faster code, but the size of area is
25127 limited and it is up to the programmer to ensure that the area does
25128 not overflow. Also when the small data area is used one of the RX's
25129 registers (usually @code{r13}) is reserved for use pointing to this
25130 area, so it is no longer available for use by the compiler. This
25131 could result in slower and/or larger code if variables are pushed onto
25132 the stack instead of being held in this register.
25134 Note, common variables (variables that have not been initialized) and
25135 constants are not placed into the small data area as they are assigned
25136 to other sections in the output executable.
25138 The default value is zero, which disables this feature. Note, this
25139 feature is not enabled by default with higher optimization levels
25140 (@option{-O2} etc) because of the potentially detrimental effects of
25141 reserving a register. It is up to the programmer to experiment and
25142 discover whether this feature is of benefit to their program. See the
25143 description of the @option{-mpid} option for a description of how the
25144 actual register to hold the small data area pointer is chosen.
25150 Use the simulator runtime. The default is to use the libgloss
25151 board-specific runtime.
25153 @item -mas100-syntax
25154 @itemx -mno-as100-syntax
25155 @opindex mas100-syntax
25156 @opindex mno-as100-syntax
25157 When generating assembler output use a syntax that is compatible with
25158 Renesas's AS100 assembler. This syntax can also be handled by the GAS
25159 assembler, but it has some restrictions so it is not generated by default.
25161 @item -mmax-constant-size=@var{N}
25162 @opindex mmax-constant-size
25163 Specifies the maximum size, in bytes, of a constant that can be used as
25164 an operand in a RX instruction. Although the RX instruction set does
25165 allow constants of up to 4 bytes in length to be used in instructions,
25166 a longer value equates to a longer instruction. Thus in some
25167 circumstances it can be beneficial to restrict the size of constants
25168 that are used in instructions. Constants that are too big are instead
25169 placed into a constant pool and referenced via register indirection.
25171 The value @var{N} can be between 0 and 4. A value of 0 (the default)
25172 or 4 means that constants of any size are allowed.
25176 Enable linker relaxation. Linker relaxation is a process whereby the
25177 linker attempts to reduce the size of a program by finding shorter
25178 versions of various instructions. Disabled by default.
25180 @item -mint-register=@var{N}
25181 @opindex mint-register
25182 Specify the number of registers to reserve for fast interrupt handler
25183 functions. The value @var{N} can be between 0 and 4. A value of 1
25184 means that register @code{r13} is reserved for the exclusive use
25185 of fast interrupt handlers. A value of 2 reserves @code{r13} and
25186 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
25187 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
25188 A value of 0, the default, does not reserve any registers.
25190 @item -msave-acc-in-interrupts
25191 @opindex msave-acc-in-interrupts
25192 Specifies that interrupt handler functions should preserve the
25193 accumulator register. This is only necessary if normal code might use
25194 the accumulator register, for example because it performs 64-bit
25195 multiplications. The default is to ignore the accumulator as this
25196 makes the interrupt handlers faster.
25202 Enables the generation of position independent data. When enabled any
25203 access to constant data is done via an offset from a base address
25204 held in a register. This allows the location of constant data to be
25205 determined at run time without requiring the executable to be
25206 relocated, which is a benefit to embedded applications with tight
25207 memory constraints. Data that can be modified is not affected by this
25210 Note, using this feature reserves a register, usually @code{r13}, for
25211 the constant data base address. This can result in slower and/or
25212 larger code, especially in complicated functions.
25214 The actual register chosen to hold the constant data base address
25215 depends upon whether the @option{-msmall-data-limit} and/or the
25216 @option{-mint-register} command-line options are enabled. Starting
25217 with register @code{r13} and proceeding downwards, registers are
25218 allocated first to satisfy the requirements of @option{-mint-register},
25219 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
25220 is possible for the small data area register to be @code{r8} if both
25221 @option{-mint-register=4} and @option{-mpid} are specified on the
25224 By default this feature is not enabled. The default can be restored
25225 via the @option{-mno-pid} command-line option.
25227 @item -mno-warn-multiple-fast-interrupts
25228 @itemx -mwarn-multiple-fast-interrupts
25229 @opindex mno-warn-multiple-fast-interrupts
25230 @opindex mwarn-multiple-fast-interrupts
25231 Prevents GCC from issuing a warning message if it finds more than one
25232 fast interrupt handler when it is compiling a file. The default is to
25233 issue a warning for each extra fast interrupt handler found, as the RX
25234 only supports one such interrupt.
25236 @item -mallow-string-insns
25237 @itemx -mno-allow-string-insns
25238 @opindex mallow-string-insns
25239 @opindex mno-allow-string-insns
25240 Enables or disables the use of the string manipulation instructions
25241 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
25242 @code{SWHILE} and also the @code{RMPA} instruction. These
25243 instructions may prefetch data, which is not safe to do if accessing
25244 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
25245 for more information).
25247 The default is to allow these instructions, but it is not possible for
25248 GCC to reliably detect all circumstances where a string instruction
25249 might be used to access an I/O register, so their use cannot be
25250 disabled automatically. Instead it is reliant upon the programmer to
25251 use the @option{-mno-allow-string-insns} option if their program
25252 accesses I/O space.
25254 When the instructions are enabled GCC defines the C preprocessor
25255 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
25256 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
25262 Use only (or not only) @code{JSR} instructions to access functions.
25263 This option can be used when code size exceeds the range of @code{BSR}
25264 instructions. Note that @option{-mno-jsr} does not mean to not use
25265 @code{JSR} but instead means that any type of branch may be used.
25268 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
25269 has special significance to the RX port when used with the
25270 @code{interrupt} function attribute. This attribute indicates a
25271 function intended to process fast interrupts. GCC ensures
25272 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
25273 and/or @code{r13} and only provided that the normal use of the
25274 corresponding registers have been restricted via the
25275 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
25278 @node S/390 and zSeries Options
25279 @subsection S/390 and zSeries Options
25280 @cindex S/390 and zSeries Options
25282 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
25286 @itemx -msoft-float
25287 @opindex mhard-float
25288 @opindex msoft-float
25289 Use (do not use) the hardware floating-point instructions and registers
25290 for floating-point operations. When @option{-msoft-float} is specified,
25291 functions in @file{libgcc.a} are used to perform floating-point
25292 operations. When @option{-mhard-float} is specified, the compiler
25293 generates IEEE floating-point instructions. This is the default.
25296 @itemx -mno-hard-dfp
25298 @opindex mno-hard-dfp
25299 Use (do not use) the hardware decimal-floating-point instructions for
25300 decimal-floating-point operations. When @option{-mno-hard-dfp} is
25301 specified, functions in @file{libgcc.a} are used to perform
25302 decimal-floating-point operations. When @option{-mhard-dfp} is
25303 specified, the compiler generates decimal-floating-point hardware
25304 instructions. This is the default for @option{-march=z9-ec} or higher.
25306 @item -mlong-double-64
25307 @itemx -mlong-double-128
25308 @opindex mlong-double-64
25309 @opindex mlong-double-128
25310 These switches control the size of @code{long double} type. A size
25311 of 64 bits makes the @code{long double} type equivalent to the @code{double}
25312 type. This is the default.
25315 @itemx -mno-backchain
25316 @opindex mbackchain
25317 @opindex mno-backchain
25318 Store (do not store) the address of the caller's frame as backchain pointer
25319 into the callee's stack frame.
25320 A backchain may be needed to allow debugging using tools that do not understand
25321 DWARF call frame information.
25322 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
25323 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
25324 the backchain is placed into the topmost word of the 96/160 byte register
25327 In general, code compiled with @option{-mbackchain} is call-compatible with
25328 code compiled with @option{-mmo-backchain}; however, use of the backchain
25329 for debugging purposes usually requires that the whole binary is built with
25330 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
25331 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
25332 to build a linux kernel use @option{-msoft-float}.
25334 The default is to not maintain the backchain.
25336 @item -mpacked-stack
25337 @itemx -mno-packed-stack
25338 @opindex mpacked-stack
25339 @opindex mno-packed-stack
25340 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
25341 specified, the compiler uses the all fields of the 96/160 byte register save
25342 area only for their default purpose; unused fields still take up stack space.
25343 When @option{-mpacked-stack} is specified, register save slots are densely
25344 packed at the top of the register save area; unused space is reused for other
25345 purposes, allowing for more efficient use of the available stack space.
25346 However, when @option{-mbackchain} is also in effect, the topmost word of
25347 the save area is always used to store the backchain, and the return address
25348 register is always saved two words below the backchain.
25350 As long as the stack frame backchain is not used, code generated with
25351 @option{-mpacked-stack} is call-compatible with code generated with
25352 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
25353 S/390 or zSeries generated code that uses the stack frame backchain at run
25354 time, not just for debugging purposes. Such code is not call-compatible
25355 with code compiled with @option{-mpacked-stack}. Also, note that the
25356 combination of @option{-mbackchain},
25357 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
25358 to build a linux kernel use @option{-msoft-float}.
25360 The default is to not use the packed stack layout.
25363 @itemx -mno-small-exec
25364 @opindex msmall-exec
25365 @opindex mno-small-exec
25366 Generate (or do not generate) code using the @code{bras} instruction
25367 to do subroutine calls.
25368 This only works reliably if the total executable size does not
25369 exceed 64k. The default is to use the @code{basr} instruction instead,
25370 which does not have this limitation.
25376 When @option{-m31} is specified, generate code compliant to the
25377 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
25378 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
25379 particular to generate 64-bit instructions. For the @samp{s390}
25380 targets, the default is @option{-m31}, while the @samp{s390x}
25381 targets default to @option{-m64}.
25387 When @option{-mzarch} is specified, generate code using the
25388 instructions available on z/Architecture.
25389 When @option{-mesa} is specified, generate code using the
25390 instructions available on ESA/390. Note that @option{-mesa} is
25391 not possible with @option{-m64}.
25392 When generating code compliant to the GNU/Linux for S/390 ABI,
25393 the default is @option{-mesa}. When generating code compliant
25394 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
25400 The @option{-mhtm} option enables a set of builtins making use of
25401 instructions available with the transactional execution facility
25402 introduced with the IBM zEnterprise EC12 machine generation
25403 @ref{S/390 System z Built-in Functions}.
25404 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
25410 When @option{-mvx} is specified, generate code using the instructions
25411 available with the vector extension facility introduced with the IBM
25412 z13 machine generation.
25413 This option changes the ABI for some vector type values with regard to
25414 alignment and calling conventions. In case vector type values are
25415 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
25416 command will be added to mark the resulting binary with the ABI used.
25417 @option{-mvx} is enabled by default when using @option{-march=z13}.
25420 @itemx -mno-zvector
25422 @opindex mno-zvector
25423 The @option{-mzvector} option enables vector language extensions and
25424 builtins using instructions available with the vector extension
25425 facility introduced with the IBM z13 machine generation.
25426 This option adds support for @samp{vector} to be used as a keyword to
25427 define vector type variables and arguments. @samp{vector} is only
25428 available when GNU extensions are enabled. It will not be expanded
25429 when requesting strict standard compliance e.g.@: with @option{-std=c99}.
25430 In addition to the GCC low-level builtins @option{-mzvector} enables
25431 a set of builtins added for compatibility with AltiVec-style
25432 implementations like Power and Cell. In order to make use of these
25433 builtins the header file @file{vecintrin.h} needs to be included.
25434 @option{-mzvector} is disabled by default.
25440 Generate (or do not generate) code using the @code{mvcle} instruction
25441 to perform block moves. When @option{-mno-mvcle} is specified,
25442 use a @code{mvc} loop instead. This is the default unless optimizing for
25449 Print (or do not print) additional debug information when compiling.
25450 The default is to not print debug information.
25452 @item -march=@var{cpu-type}
25454 Generate code that runs on @var{cpu-type}, which is the name of a
25455 system representing a certain processor type. Possible values for
25456 @var{cpu-type} are @samp{z900}/@samp{arch5}, @samp{z990}/@samp{arch6},
25457 @samp{z9-109}, @samp{z9-ec}/@samp{arch7}, @samp{z10}/@samp{arch8},
25458 @samp{z196}/@samp{arch9}, @samp{zEC12}, @samp{z13}/@samp{arch11},
25459 @samp{z14}/@samp{arch12}, and @samp{native}.
25461 The default is @option{-march=z900}.
25463 Specifying @samp{native} as cpu type can be used to select the best
25464 architecture option for the host processor.
25465 @option{-march=native} has no effect if GCC does not recognize the
25468 @item -mtune=@var{cpu-type}
25470 Tune to @var{cpu-type} everything applicable about the generated code,
25471 except for the ABI and the set of available instructions.
25472 The list of @var{cpu-type} values is the same as for @option{-march}.
25473 The default is the value used for @option{-march}.
25476 @itemx -mno-tpf-trace
25477 @opindex mtpf-trace
25478 @opindex mno-tpf-trace
25479 Generate code that adds (does not add) in TPF OS specific branches to trace
25480 routines in the operating system. This option is off by default, even
25481 when compiling for the TPF OS@.
25484 @itemx -mno-fused-madd
25485 @opindex mfused-madd
25486 @opindex mno-fused-madd
25487 Generate code that uses (does not use) the floating-point multiply and
25488 accumulate instructions. These instructions are generated by default if
25489 hardware floating point is used.
25491 @item -mwarn-framesize=@var{framesize}
25492 @opindex mwarn-framesize
25493 Emit a warning if the current function exceeds the given frame size. Because
25494 this is a compile-time check it doesn't need to be a real problem when the program
25495 runs. It is intended to identify functions that most probably cause
25496 a stack overflow. It is useful to be used in an environment with limited stack
25497 size e.g.@: the linux kernel.
25499 @item -mwarn-dynamicstack
25500 @opindex mwarn-dynamicstack
25501 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
25502 arrays. This is generally a bad idea with a limited stack size.
25504 @item -mstack-guard=@var{stack-guard}
25505 @itemx -mstack-size=@var{stack-size}
25506 @opindex mstack-guard
25507 @opindex mstack-size
25508 If these options are provided the S/390 back end emits additional instructions in
25509 the function prologue that trigger a trap if the stack size is @var{stack-guard}
25510 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
25511 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
25512 the frame size of the compiled function is chosen.
25513 These options are intended to be used to help debugging stack overflow problems.
25514 The additionally emitted code causes only little overhead and hence can also be
25515 used in production-like systems without greater performance degradation. The given
25516 values have to be exact powers of 2 and @var{stack-size} has to be greater than
25517 @var{stack-guard} without exceeding 64k.
25518 In order to be efficient the extra code makes the assumption that the stack starts
25519 at an address aligned to the value given by @var{stack-size}.
25520 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
25522 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
25524 If the hotpatch option is enabled, a ``hot-patching'' function
25525 prologue is generated for all functions in the compilation unit.
25526 The funtion label is prepended with the given number of two-byte
25527 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
25528 the label, 2 * @var{post-halfwords} bytes are appended, using the
25529 largest NOP like instructions the architecture allows (maximum
25532 If both arguments are zero, hotpatching is disabled.
25534 This option can be overridden for individual functions with the
25535 @code{hotpatch} attribute.
25538 @node Score Options
25539 @subsection Score Options
25540 @cindex Score Options
25542 These options are defined for Score implementations:
25547 Compile code for big-endian mode. This is the default.
25551 Compile code for little-endian mode.
25555 Disable generation of @code{bcnz} instructions.
25559 Enable generation of unaligned load and store instructions.
25563 Enable the use of multiply-accumulate instructions. Disabled by default.
25567 Specify the SCORE5 as the target architecture.
25571 Specify the SCORE5U of the target architecture.
25575 Specify the SCORE7 as the target architecture. This is the default.
25579 Specify the SCORE7D as the target architecture.
25583 @subsection SH Options
25585 These @samp{-m} options are defined for the SH implementations:
25590 Generate code for the SH1.
25594 Generate code for the SH2.
25597 Generate code for the SH2e.
25601 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
25602 that the floating-point unit is not used.
25604 @item -m2a-single-only
25605 @opindex m2a-single-only
25606 Generate code for the SH2a-FPU, in such a way that no double-precision
25607 floating-point operations are used.
25610 @opindex m2a-single
25611 Generate code for the SH2a-FPU assuming the floating-point unit is in
25612 single-precision mode by default.
25616 Generate code for the SH2a-FPU assuming the floating-point unit is in
25617 double-precision mode by default.
25621 Generate code for the SH3.
25625 Generate code for the SH3e.
25629 Generate code for the SH4 without a floating-point unit.
25631 @item -m4-single-only
25632 @opindex m4-single-only
25633 Generate code for the SH4 with a floating-point unit that only
25634 supports single-precision arithmetic.
25638 Generate code for the SH4 assuming the floating-point unit is in
25639 single-precision mode by default.
25643 Generate code for the SH4.
25647 Generate code for SH4-100.
25649 @item -m4-100-nofpu
25650 @opindex m4-100-nofpu
25651 Generate code for SH4-100 in such a way that the
25652 floating-point unit is not used.
25654 @item -m4-100-single
25655 @opindex m4-100-single
25656 Generate code for SH4-100 assuming the floating-point unit is in
25657 single-precision mode by default.
25659 @item -m4-100-single-only
25660 @opindex m4-100-single-only
25661 Generate code for SH4-100 in such a way that no double-precision
25662 floating-point operations are used.
25666 Generate code for SH4-200.
25668 @item -m4-200-nofpu
25669 @opindex m4-200-nofpu
25670 Generate code for SH4-200 without in such a way that the
25671 floating-point unit is not used.
25673 @item -m4-200-single
25674 @opindex m4-200-single
25675 Generate code for SH4-200 assuming the floating-point unit is in
25676 single-precision mode by default.
25678 @item -m4-200-single-only
25679 @opindex m4-200-single-only
25680 Generate code for SH4-200 in such a way that no double-precision
25681 floating-point operations are used.
25685 Generate code for SH4-300.
25687 @item -m4-300-nofpu
25688 @opindex m4-300-nofpu
25689 Generate code for SH4-300 without in such a way that the
25690 floating-point unit is not used.
25692 @item -m4-300-single
25693 @opindex m4-300-single
25694 Generate code for SH4-300 in such a way that no double-precision
25695 floating-point operations are used.
25697 @item -m4-300-single-only
25698 @opindex m4-300-single-only
25699 Generate code for SH4-300 in such a way that no double-precision
25700 floating-point operations are used.
25704 Generate code for SH4-340 (no MMU, no FPU).
25708 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
25713 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
25714 floating-point unit is not used.
25716 @item -m4a-single-only
25717 @opindex m4a-single-only
25718 Generate code for the SH4a, in such a way that no double-precision
25719 floating-point operations are used.
25722 @opindex m4a-single
25723 Generate code for the SH4a assuming the floating-point unit is in
25724 single-precision mode by default.
25728 Generate code for the SH4a.
25732 Same as @option{-m4a-nofpu}, except that it implicitly passes
25733 @option{-dsp} to the assembler. GCC doesn't generate any DSP
25734 instructions at the moment.
25738 Compile code for the processor in big-endian mode.
25742 Compile code for the processor in little-endian mode.
25746 Align doubles at 64-bit boundaries. Note that this changes the calling
25747 conventions, and thus some functions from the standard C library do
25748 not work unless you recompile it first with @option{-mdalign}.
25752 Shorten some address references at link time, when possible; uses the
25753 linker option @option{-relax}.
25757 Use 32-bit offsets in @code{switch} tables. The default is to use
25762 Enable the use of bit manipulation instructions on SH2A.
25766 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
25767 alignment constraints.
25771 Comply with the calling conventions defined by Renesas.
25774 @opindex mno-renesas
25775 Comply with the calling conventions defined for GCC before the Renesas
25776 conventions were available. This option is the default for all
25777 targets of the SH toolchain.
25780 @opindex mnomacsave
25781 Mark the @code{MAC} register as call-clobbered, even if
25782 @option{-mrenesas} is given.
25788 Control the IEEE compliance of floating-point comparisons, which affects the
25789 handling of cases where the result of a comparison is unordered. By default
25790 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
25791 enabled @option{-mno-ieee} is implicitly set, which results in faster
25792 floating-point greater-equal and less-equal comparisons. The implicit settings
25793 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
25795 @item -minline-ic_invalidate
25796 @opindex minline-ic_invalidate
25797 Inline code to invalidate instruction cache entries after setting up
25798 nested function trampolines.
25799 This option has no effect if @option{-musermode} is in effect and the selected
25800 code generation option (e.g.@: @option{-m4}) does not allow the use of the @code{icbi}
25802 If the selected code generation option does not allow the use of the @code{icbi}
25803 instruction, and @option{-musermode} is not in effect, the inlined code
25804 manipulates the instruction cache address array directly with an associative
25805 write. This not only requires privileged mode at run time, but it also
25806 fails if the cache line had been mapped via the TLB and has become unmapped.
25810 Dump instruction size and location in the assembly code.
25813 @opindex mpadstruct
25814 This option is deprecated. It pads structures to multiple of 4 bytes,
25815 which is incompatible with the SH ABI@.
25817 @item -matomic-model=@var{model}
25818 @opindex matomic-model=@var{model}
25819 Sets the model of atomic operations and additional parameters as a comma
25820 separated list. For details on the atomic built-in functions see
25821 @ref{__atomic Builtins}. The following models and parameters are supported:
25826 Disable compiler generated atomic sequences and emit library calls for atomic
25827 operations. This is the default if the target is not @code{sh*-*-linux*}.
25830 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
25831 built-in functions. The generated atomic sequences require additional support
25832 from the interrupt/exception handling code of the system and are only suitable
25833 for SH3* and SH4* single-core systems. This option is enabled by default when
25834 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
25835 this option also partially utilizes the hardware atomic instructions
25836 @code{movli.l} and @code{movco.l} to create more efficient code, unless
25837 @samp{strict} is specified.
25840 Generate software atomic sequences that use a variable in the thread control
25841 block. This is a variation of the gUSA sequences which can also be used on
25842 SH1* and SH2* targets. The generated atomic sequences require additional
25843 support from the interrupt/exception handling code of the system and are only
25844 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
25845 parameter has to be specified as well.
25848 Generate software atomic sequences that temporarily disable interrupts by
25849 setting @code{SR.IMASK = 1111}. This model works only when the program runs
25850 in privileged mode and is only suitable for single-core systems. Additional
25851 support from the interrupt/exception handling code of the system is not
25852 required. This model is enabled by default when the target is
25853 @code{sh*-*-linux*} and SH1* or SH2*.
25856 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
25857 instructions only. This is only available on SH4A and is suitable for
25858 multi-core systems. Since the hardware instructions support only 32 bit atomic
25859 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
25860 Code compiled with this option is also compatible with other software
25861 atomic model interrupt/exception handling systems if executed on an SH4A
25862 system. Additional support from the interrupt/exception handling code of the
25863 system is not required for this model.
25866 This parameter specifies the offset in bytes of the variable in the thread
25867 control block structure that should be used by the generated atomic sequences
25868 when the @samp{soft-tcb} model has been selected. For other models this
25869 parameter is ignored. The specified value must be an integer multiple of four
25870 and in the range 0-1020.
25873 This parameter prevents mixed usage of multiple atomic models, even if they
25874 are compatible, and makes the compiler generate atomic sequences of the
25875 specified model only.
25881 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
25882 Notice that depending on the particular hardware and software configuration
25883 this can degrade overall performance due to the operand cache line flushes
25884 that are implied by the @code{tas.b} instruction. On multi-core SH4A
25885 processors the @code{tas.b} instruction must be used with caution since it
25886 can result in data corruption for certain cache configurations.
25889 @opindex mprefergot
25890 When generating position-independent code, emit function calls using
25891 the Global Offset Table instead of the Procedure Linkage Table.
25894 @itemx -mno-usermode
25896 @opindex mno-usermode
25897 Don't allow (allow) the compiler generating privileged mode code. Specifying
25898 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
25899 inlined code would not work in user mode. @option{-musermode} is the default
25900 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
25901 @option{-musermode} has no effect, since there is no user mode.
25903 @item -multcost=@var{number}
25904 @opindex multcost=@var{number}
25905 Set the cost to assume for a multiply insn.
25907 @item -mdiv=@var{strategy}
25908 @opindex mdiv=@var{strategy}
25909 Set the division strategy to be used for integer division operations.
25910 @var{strategy} can be one of:
25915 Calls a library function that uses the single-step division instruction
25916 @code{div1} to perform the operation. Division by zero calculates an
25917 unspecified result and does not trap. This is the default except for SH4,
25918 SH2A and SHcompact.
25921 Calls a library function that performs the operation in double precision
25922 floating point. Division by zero causes a floating-point exception. This is
25923 the default for SHcompact with FPU. Specifying this for targets that do not
25924 have a double precision FPU defaults to @code{call-div1}.
25927 Calls a library function that uses a lookup table for small divisors and
25928 the @code{div1} instruction with case distinction for larger divisors. Division
25929 by zero calculates an unspecified result and does not trap. This is the default
25930 for SH4. Specifying this for targets that do not have dynamic shift
25931 instructions defaults to @code{call-div1}.
25935 When a division strategy has not been specified the default strategy is
25936 selected based on the current target. For SH2A the default strategy is to
25937 use the @code{divs} and @code{divu} instructions instead of library function
25940 @item -maccumulate-outgoing-args
25941 @opindex maccumulate-outgoing-args
25942 Reserve space once for outgoing arguments in the function prologue rather
25943 than around each call. Generally beneficial for performance and size. Also
25944 needed for unwinding to avoid changing the stack frame around conditional code.
25946 @item -mdivsi3_libfunc=@var{name}
25947 @opindex mdivsi3_libfunc=@var{name}
25948 Set the name of the library function used for 32-bit signed division to
25950 This only affects the name used in the @samp{call} division strategies, and
25951 the compiler still expects the same sets of input/output/clobbered registers as
25952 if this option were not present.
25954 @item -mfixed-range=@var{register-range}
25955 @opindex mfixed-range
25956 Generate code treating the given register range as fixed registers.
25957 A fixed register is one that the register allocator cannot use. This is
25958 useful when compiling kernel code. A register range is specified as
25959 two registers separated by a dash. Multiple register ranges can be
25960 specified separated by a comma.
25962 @item -mbranch-cost=@var{num}
25963 @opindex mbranch-cost=@var{num}
25964 Assume @var{num} to be the cost for a branch instruction. Higher numbers
25965 make the compiler try to generate more branch-free code if possible.
25966 If not specified the value is selected depending on the processor type that
25967 is being compiled for.
25970 @itemx -mno-zdcbranch
25971 @opindex mzdcbranch
25972 @opindex mno-zdcbranch
25973 Assume (do not assume) that zero displacement conditional branch instructions
25974 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
25975 compiler prefers zero displacement branch code sequences. This is
25976 enabled by default when generating code for SH4 and SH4A. It can be explicitly
25977 disabled by specifying @option{-mno-zdcbranch}.
25979 @item -mcbranch-force-delay-slot
25980 @opindex mcbranch-force-delay-slot
25981 Force the usage of delay slots for conditional branches, which stuffs the delay
25982 slot with a @code{nop} if a suitable instruction cannot be found. By default
25983 this option is disabled. It can be enabled to work around hardware bugs as
25984 found in the original SH7055.
25987 @itemx -mno-fused-madd
25988 @opindex mfused-madd
25989 @opindex mno-fused-madd
25990 Generate code that uses (does not use) the floating-point multiply and
25991 accumulate instructions. These instructions are generated by default
25992 if hardware floating point is used. The machine-dependent
25993 @option{-mfused-madd} option is now mapped to the machine-independent
25994 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
25995 mapped to @option{-ffp-contract=off}.
26001 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
26002 and cosine approximations. The option @option{-mfsca} must be used in
26003 combination with @option{-funsafe-math-optimizations}. It is enabled by default
26004 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
26005 approximations even if @option{-funsafe-math-optimizations} is in effect.
26011 Allow or disallow the compiler to emit the @code{fsrra} instruction for
26012 reciprocal square root approximations. The option @option{-mfsrra} must be used
26013 in combination with @option{-funsafe-math-optimizations} and
26014 @option{-ffinite-math-only}. It is enabled by default when generating code for
26015 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
26016 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
26019 @item -mpretend-cmove
26020 @opindex mpretend-cmove
26021 Prefer zero-displacement conditional branches for conditional move instruction
26022 patterns. This can result in faster code on the SH4 processor.
26026 Generate code using the FDPIC ABI.
26030 @node Solaris 2 Options
26031 @subsection Solaris 2 Options
26032 @cindex Solaris 2 options
26034 These @samp{-m} options are supported on Solaris 2:
26037 @item -mclear-hwcap
26038 @opindex mclear-hwcap
26039 @option{-mclear-hwcap} tells the compiler to remove the hardware
26040 capabilities generated by the Solaris assembler. This is only necessary
26041 when object files use ISA extensions not supported by the current
26042 machine, but check at runtime whether or not to use them.
26044 @item -mimpure-text
26045 @opindex mimpure-text
26046 @option{-mimpure-text}, used in addition to @option{-shared}, tells
26047 the compiler to not pass @option{-z text} to the linker when linking a
26048 shared object. Using this option, you can link position-dependent
26049 code into a shared object.
26051 @option{-mimpure-text} suppresses the ``relocations remain against
26052 allocatable but non-writable sections'' linker error message.
26053 However, the necessary relocations trigger copy-on-write, and the
26054 shared object is not actually shared across processes. Instead of
26055 using @option{-mimpure-text}, you should compile all source code with
26056 @option{-fpic} or @option{-fPIC}.
26060 These switches are supported in addition to the above on Solaris 2:
26065 This is a synonym for @option{-pthread}.
26068 @node SPARC Options
26069 @subsection SPARC Options
26070 @cindex SPARC options
26072 These @samp{-m} options are supported on the SPARC:
26075 @item -mno-app-regs
26077 @opindex mno-app-regs
26079 Specify @option{-mapp-regs} to generate output using the global registers
26080 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
26081 global register 1, each global register 2 through 4 is then treated as an
26082 allocable register that is clobbered by function calls. This is the default.
26084 To be fully SVR4 ABI-compliant at the cost of some performance loss,
26085 specify @option{-mno-app-regs}. You should compile libraries and system
26086 software with this option.
26092 With @option{-mflat}, the compiler does not generate save/restore instructions
26093 and uses a ``flat'' or single register window model. This model is compatible
26094 with the regular register window model. The local registers and the input
26095 registers (0--5) are still treated as ``call-saved'' registers and are
26096 saved on the stack as needed.
26098 With @option{-mno-flat} (the default), the compiler generates save/restore
26099 instructions (except for leaf functions). This is the normal operating mode.
26102 @itemx -mhard-float
26104 @opindex mhard-float
26105 Generate output containing floating-point instructions. This is the
26109 @itemx -msoft-float
26111 @opindex msoft-float
26112 Generate output containing library calls for floating point.
26113 @strong{Warning:} the requisite libraries are not available for all SPARC
26114 targets. Normally the facilities of the machine's usual C compiler are
26115 used, but this cannot be done directly in cross-compilation. You must make
26116 your own arrangements to provide suitable library functions for
26117 cross-compilation. The embedded targets @samp{sparc-*-aout} and
26118 @samp{sparclite-*-*} do provide software floating-point support.
26120 @option{-msoft-float} changes the calling convention in the output file;
26121 therefore, it is only useful if you compile @emph{all} of a program with
26122 this option. In particular, you need to compile @file{libgcc.a}, the
26123 library that comes with GCC, with @option{-msoft-float} in order for
26126 @item -mhard-quad-float
26127 @opindex mhard-quad-float
26128 Generate output containing quad-word (long double) floating-point
26131 @item -msoft-quad-float
26132 @opindex msoft-quad-float
26133 Generate output containing library calls for quad-word (long double)
26134 floating-point instructions. The functions called are those specified
26135 in the SPARC ABI@. This is the default.
26137 As of this writing, there are no SPARC implementations that have hardware
26138 support for the quad-word floating-point instructions. They all invoke
26139 a trap handler for one of these instructions, and then the trap handler
26140 emulates the effect of the instruction. Because of the trap handler overhead,
26141 this is much slower than calling the ABI library routines. Thus the
26142 @option{-msoft-quad-float} option is the default.
26144 @item -mno-unaligned-doubles
26145 @itemx -munaligned-doubles
26146 @opindex mno-unaligned-doubles
26147 @opindex munaligned-doubles
26148 Assume that doubles have 8-byte alignment. This is the default.
26150 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
26151 alignment only if they are contained in another type, or if they have an
26152 absolute address. Otherwise, it assumes they have 4-byte alignment.
26153 Specifying this option avoids some rare compatibility problems with code
26154 generated by other compilers. It is not the default because it results
26155 in a performance loss, especially for floating-point code.
26158 @itemx -mno-user-mode
26159 @opindex muser-mode
26160 @opindex mno-user-mode
26161 Do not generate code that can only run in supervisor mode. This is relevant
26162 only for the @code{casa} instruction emitted for the LEON3 processor. This
26165 @item -mfaster-structs
26166 @itemx -mno-faster-structs
26167 @opindex mfaster-structs
26168 @opindex mno-faster-structs
26169 With @option{-mfaster-structs}, the compiler assumes that structures
26170 should have 8-byte alignment. This enables the use of pairs of
26171 @code{ldd} and @code{std} instructions for copies in structure
26172 assignment, in place of twice as many @code{ld} and @code{st} pairs.
26173 However, the use of this changed alignment directly violates the SPARC
26174 ABI@. Thus, it's intended only for use on targets where the developer
26175 acknowledges that their resulting code is not directly in line with
26176 the rules of the ABI@.
26178 @item -mstd-struct-return
26179 @itemx -mno-std-struct-return
26180 @opindex mstd-struct-return
26181 @opindex mno-std-struct-return
26182 With @option{-mstd-struct-return}, the compiler generates checking code
26183 in functions returning structures or unions to detect size mismatches
26184 between the two sides of function calls, as per the 32-bit ABI@.
26186 The default is @option{-mno-std-struct-return}. This option has no effect
26193 Enable Local Register Allocation. This is the default for SPARC since GCC 7
26194 so @option{-mno-lra} needs to be passed to get old Reload.
26196 @item -mcpu=@var{cpu_type}
26198 Set the instruction set, register set, and instruction scheduling parameters
26199 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
26200 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
26201 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
26202 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
26203 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
26204 @samp{niagara3}, @samp{niagara4}, @samp{niagara7} and @samp{m8}.
26206 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
26207 which selects the best architecture option for the host processor.
26208 @option{-mcpu=native} has no effect if GCC does not recognize
26211 Default instruction scheduling parameters are used for values that select
26212 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
26213 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
26215 Here is a list of each supported architecture and their supported
26223 supersparc, hypersparc, leon, leon3
26226 f930, f934, sparclite86x
26232 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4,
26236 By default (unless configured otherwise), GCC generates code for the V7
26237 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
26238 additionally optimizes it for the Cypress CY7C602 chip, as used in the
26239 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
26240 SPARCStation 1, 2, IPX etc.
26242 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
26243 architecture. The only difference from V7 code is that the compiler emits
26244 the integer multiply and integer divide instructions which exist in SPARC-V8
26245 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
26246 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
26249 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
26250 the SPARC architecture. This adds the integer multiply, integer divide step
26251 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
26252 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
26253 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
26254 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
26255 MB86934 chip, which is the more recent SPARClite with FPU@.
26257 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
26258 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
26259 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
26260 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
26261 optimizes it for the TEMIC SPARClet chip.
26263 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
26264 architecture. This adds 64-bit integer and floating-point move instructions,
26265 3 additional floating-point condition code registers and conditional move
26266 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
26267 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
26268 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
26269 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
26270 @option{-mcpu=niagara}, the compiler additionally optimizes it for
26271 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
26272 additionally optimizes it for Sun UltraSPARC T2 chips. With
26273 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
26274 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
26275 additionally optimizes it for Sun UltraSPARC T4 chips. With
26276 @option{-mcpu=niagara7}, the compiler additionally optimizes it for
26277 Oracle SPARC M7 chips. With @option{-mcpu=m8}, the compiler
26278 additionally optimizes it for Oracle M8 chips.
26280 @item -mtune=@var{cpu_type}
26282 Set the instruction scheduling parameters for machine type
26283 @var{cpu_type}, but do not set the instruction set or register set that the
26284 option @option{-mcpu=@var{cpu_type}} does.
26286 The same values for @option{-mcpu=@var{cpu_type}} can be used for
26287 @option{-mtune=@var{cpu_type}}, but the only useful values are those
26288 that select a particular CPU implementation. Those are
26289 @samp{cypress}, @samp{supersparc}, @samp{hypersparc}, @samp{leon},
26290 @samp{leon3}, @samp{leon3v7}, @samp{f930}, @samp{f934},
26291 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
26292 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
26293 @samp{niagara4}, @samp{niagara7} and @samp{m8}. With native Solaris
26294 and GNU/Linux toolchains, @samp{native} can also be used.
26299 @opindex mno-v8plus
26300 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
26301 difference from the V8 ABI is that the global and out registers are
26302 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
26303 mode for all SPARC-V9 processors.
26309 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
26310 Visual Instruction Set extensions. The default is @option{-mno-vis}.
26316 With @option{-mvis2}, GCC generates code that takes advantage of
26317 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
26318 default is @option{-mvis2} when targeting a cpu that supports such
26319 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
26320 also sets @option{-mvis}.
26326 With @option{-mvis3}, GCC generates code that takes advantage of
26327 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
26328 default is @option{-mvis3} when targeting a cpu that supports such
26329 instructions, such as niagara-3 and later. Setting @option{-mvis3}
26330 also sets @option{-mvis2} and @option{-mvis}.
26336 With @option{-mvis4}, GCC generates code that takes advantage of
26337 version 4.0 of the UltraSPARC Visual Instruction Set extensions. The
26338 default is @option{-mvis4} when targeting a cpu that supports such
26339 instructions, such as niagara-7 and later. Setting @option{-mvis4}
26340 also sets @option{-mvis3}, @option{-mvis2} and @option{-mvis}.
26346 With @option{-mvis4b}, GCC generates code that takes advantage of
26347 version 4.0 of the UltraSPARC Visual Instruction Set extensions, plus
26348 the additional VIS instructions introduced in the Oracle SPARC
26349 Architecture 2017. The default is @option{-mvis4b} when targeting a
26350 cpu that supports such instructions, such as m8 and later. Setting
26351 @option{-mvis4b} also sets @option{-mvis4}, @option{-mvis3},
26352 @option{-mvis2} and @option{-mvis}.
26357 @opindex mno-cbcond
26358 With @option{-mcbcond}, GCC generates code that takes advantage of the UltraSPARC
26359 Compare-and-Branch-on-Condition instructions. The default is @option{-mcbcond}
26360 when targeting a CPU that supports such instructions, such as Niagara-4 and
26367 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
26368 Fused Multiply-Add Floating-point instructions. The default is @option{-mfmaf}
26369 when targeting a CPU that supports such instructions, such as Niagara-3 and
26375 @opindex mno-fsmuld
26376 With @option{-mfsmuld}, GCC generates code that takes advantage of the
26377 Floating-point Multiply Single to Double (FsMULd) instruction. The default is
26378 @option{-mfsmuld} when targeting a CPU supporting the architecture versions V8
26379 or V9 with FPU except @option{-mcpu=leon}.
26385 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
26386 Population Count instruction. The default is @option{-mpopc}
26387 when targeting a CPU that supports such an instruction, such as Niagara-2 and
26394 With @option{-msubxc}, GCC generates code that takes advantage of the UltraSPARC
26395 Subtract-Extended-with-Carry instruction. The default is @option{-msubxc}
26396 when targeting a CPU that supports such an instruction, such as Niagara-7 and
26400 @opindex mfix-at697f
26401 Enable the documented workaround for the single erratum of the Atmel AT697F
26402 processor (which corresponds to erratum #13 of the AT697E processor).
26405 @opindex mfix-ut699
26406 Enable the documented workarounds for the floating-point errata and the data
26407 cache nullify errata of the UT699 processor.
26410 @opindex mfix-ut700
26411 Enable the documented workaround for the back-to-back store errata of
26412 the UT699E/UT700 processor.
26414 @item -mfix-gr712rc
26415 @opindex mfix-gr712rc
26416 Enable the documented workaround for the back-to-back store errata of
26417 the GR712RC processor.
26420 These @samp{-m} options are supported in addition to the above
26421 on SPARC-V9 processors in 64-bit environments:
26428 Generate code for a 32-bit or 64-bit environment.
26429 The 32-bit environment sets int, long and pointer to 32 bits.
26430 The 64-bit environment sets int to 32 bits and long and pointer
26433 @item -mcmodel=@var{which}
26435 Set the code model to one of
26439 The Medium/Low code model: 64-bit addresses, programs
26440 must be linked in the low 32 bits of memory. Programs can be statically
26441 or dynamically linked.
26444 The Medium/Middle code model: 64-bit addresses, programs
26445 must be linked in the low 44 bits of memory, the text and data segments must
26446 be less than 2GB in size and the data segment must be located within 2GB of
26450 The Medium/Anywhere code model: 64-bit addresses, programs
26451 may be linked anywhere in memory, the text and data segments must be less
26452 than 2GB in size and the data segment must be located within 2GB of the
26456 The Medium/Anywhere code model for embedded systems:
26457 64-bit addresses, the text and data segments must be less than 2GB in
26458 size, both starting anywhere in memory (determined at link time). The
26459 global register %g4 points to the base of the data segment. Programs
26460 are statically linked and PIC is not supported.
26463 @item -mmemory-model=@var{mem-model}
26464 @opindex mmemory-model
26465 Set the memory model in force on the processor to one of
26469 The default memory model for the processor and operating system.
26472 Relaxed Memory Order
26475 Partial Store Order
26481 Sequential Consistency
26484 These memory models are formally defined in Appendix D of the SPARC-V9
26485 architecture manual, as set in the processor's @code{PSTATE.MM} field.
26488 @itemx -mno-stack-bias
26489 @opindex mstack-bias
26490 @opindex mno-stack-bias
26491 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
26492 frame pointer if present, are offset by @minus{}2047 which must be added back
26493 when making stack frame references. This is the default in 64-bit mode.
26494 Otherwise, assume no such offset is present.
26498 @subsection SPU Options
26499 @cindex SPU options
26501 These @samp{-m} options are supported on the SPU:
26505 @itemx -merror-reloc
26506 @opindex mwarn-reloc
26507 @opindex merror-reloc
26509 The loader for SPU does not handle dynamic relocations. By default, GCC
26510 gives an error when it generates code that requires a dynamic
26511 relocation. @option{-mno-error-reloc} disables the error,
26512 @option{-mwarn-reloc} generates a warning instead.
26515 @itemx -munsafe-dma
26517 @opindex munsafe-dma
26519 Instructions that initiate or test completion of DMA must not be
26520 reordered with respect to loads and stores of the memory that is being
26522 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
26523 memory accesses, but that can lead to inefficient code in places where the
26524 memory is known to not change. Rather than mark the memory as volatile,
26525 you can use @option{-msafe-dma} to tell the compiler to treat
26526 the DMA instructions as potentially affecting all memory.
26528 @item -mbranch-hints
26529 @opindex mbranch-hints
26531 By default, GCC generates a branch hint instruction to avoid
26532 pipeline stalls for always-taken or probably-taken branches. A hint
26533 is not generated closer than 8 instructions away from its branch.
26534 There is little reason to disable them, except for debugging purposes,
26535 or to make an object a little bit smaller.
26539 @opindex msmall-mem
26540 @opindex mlarge-mem
26542 By default, GCC generates code assuming that addresses are never larger
26543 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
26544 a full 32-bit address.
26549 By default, GCC links against startup code that assumes the SPU-style
26550 main function interface (which has an unconventional parameter list).
26551 With @option{-mstdmain}, GCC links your program against startup
26552 code that assumes a C99-style interface to @code{main}, including a
26553 local copy of @code{argv} strings.
26555 @item -mfixed-range=@var{register-range}
26556 @opindex mfixed-range
26557 Generate code treating the given register range as fixed registers.
26558 A fixed register is one that the register allocator cannot use. This is
26559 useful when compiling kernel code. A register range is specified as
26560 two registers separated by a dash. Multiple register ranges can be
26561 specified separated by a comma.
26567 Compile code assuming that pointers to the PPU address space accessed
26568 via the @code{__ea} named address space qualifier are either 32 or 64
26569 bits wide. The default is 32 bits. As this is an ABI-changing option,
26570 all object code in an executable must be compiled with the same setting.
26572 @item -maddress-space-conversion
26573 @itemx -mno-address-space-conversion
26574 @opindex maddress-space-conversion
26575 @opindex mno-address-space-conversion
26576 Allow/disallow treating the @code{__ea} address space as superset
26577 of the generic address space. This enables explicit type casts
26578 between @code{__ea} and generic pointer as well as implicit
26579 conversions of generic pointers to @code{__ea} pointers. The
26580 default is to allow address space pointer conversions.
26582 @item -mcache-size=@var{cache-size}
26583 @opindex mcache-size
26584 This option controls the version of libgcc that the compiler links to an
26585 executable and selects a software-managed cache for accessing variables
26586 in the @code{__ea} address space with a particular cache size. Possible
26587 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
26588 and @samp{128}. The default cache size is 64KB.
26590 @item -matomic-updates
26591 @itemx -mno-atomic-updates
26592 @opindex matomic-updates
26593 @opindex mno-atomic-updates
26594 This option controls the version of libgcc that the compiler links to an
26595 executable and selects whether atomic updates to the software-managed
26596 cache of PPU-side variables are used. If you use atomic updates, changes
26597 to a PPU variable from SPU code using the @code{__ea} named address space
26598 qualifier do not interfere with changes to other PPU variables residing
26599 in the same cache line from PPU code. If you do not use atomic updates,
26600 such interference may occur; however, writing back cache lines is
26601 more efficient. The default behavior is to use atomic updates.
26604 @itemx -mdual-nops=@var{n}
26605 @opindex mdual-nops
26606 By default, GCC inserts NOPs to increase dual issue when it expects
26607 it to increase performance. @var{n} can be a value from 0 to 10. A
26608 smaller @var{n} inserts fewer NOPs. 10 is the default, 0 is the
26609 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
26611 @item -mhint-max-nops=@var{n}
26612 @opindex mhint-max-nops
26613 Maximum number of NOPs to insert for a branch hint. A branch hint must
26614 be at least 8 instructions away from the branch it is affecting. GCC
26615 inserts up to @var{n} NOPs to enforce this, otherwise it does not
26616 generate the branch hint.
26618 @item -mhint-max-distance=@var{n}
26619 @opindex mhint-max-distance
26620 The encoding of the branch hint instruction limits the hint to be within
26621 256 instructions of the branch it is affecting. By default, GCC makes
26622 sure it is within 125.
26625 @opindex msafe-hints
26626 Work around a hardware bug that causes the SPU to stall indefinitely.
26627 By default, GCC inserts the @code{hbrp} instruction to make sure
26628 this stall won't happen.
26632 @node System V Options
26633 @subsection Options for System V
26635 These additional options are available on System V Release 4 for
26636 compatibility with other compilers on those systems:
26641 Create a shared object.
26642 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
26646 Identify the versions of each tool used by the compiler, in a
26647 @code{.ident} assembler directive in the output.
26651 Refrain from adding @code{.ident} directives to the output file (this is
26654 @item -YP,@var{dirs}
26656 Search the directories @var{dirs}, and no others, for libraries
26657 specified with @option{-l}.
26659 @item -Ym,@var{dir}
26661 Look in the directory @var{dir} to find the M4 preprocessor.
26662 The assembler uses this option.
26663 @c This is supposed to go with a -Yd for predefined M4 macro files, but
26664 @c the generic assembler that comes with Solaris takes just -Ym.
26667 @node TILE-Gx Options
26668 @subsection TILE-Gx Options
26669 @cindex TILE-Gx options
26671 These @samp{-m} options are supported on the TILE-Gx:
26674 @item -mcmodel=small
26675 @opindex mcmodel=small
26676 Generate code for the small model. The distance for direct calls is
26677 limited to 500M in either direction. PC-relative addresses are 32
26678 bits. Absolute addresses support the full address range.
26680 @item -mcmodel=large
26681 @opindex mcmodel=large
26682 Generate code for the large model. There is no limitation on call
26683 distance, pc-relative addresses, or absolute addresses.
26685 @item -mcpu=@var{name}
26687 Selects the type of CPU to be targeted. Currently the only supported
26688 type is @samp{tilegx}.
26694 Generate code for a 32-bit or 64-bit environment. The 32-bit
26695 environment sets int, long, and pointer to 32 bits. The 64-bit
26696 environment sets int to 32 bits and long and pointer to 64 bits.
26699 @itemx -mlittle-endian
26700 @opindex mbig-endian
26701 @opindex mlittle-endian
26702 Generate code in big/little endian mode, respectively.
26705 @node TILEPro Options
26706 @subsection TILEPro Options
26707 @cindex TILEPro options
26709 These @samp{-m} options are supported on the TILEPro:
26712 @item -mcpu=@var{name}
26714 Selects the type of CPU to be targeted. Currently the only supported
26715 type is @samp{tilepro}.
26719 Generate code for a 32-bit environment, which sets int, long, and
26720 pointer to 32 bits. This is the only supported behavior so the flag
26721 is essentially ignored.
26725 @subsection V850 Options
26726 @cindex V850 Options
26728 These @samp{-m} options are defined for V850 implementations:
26732 @itemx -mno-long-calls
26733 @opindex mlong-calls
26734 @opindex mno-long-calls
26735 Treat all calls as being far away (near). If calls are assumed to be
26736 far away, the compiler always loads the function's address into a
26737 register, and calls indirect through the pointer.
26743 Do not optimize (do optimize) basic blocks that use the same index
26744 pointer 4 or more times to copy pointer into the @code{ep} register, and
26745 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
26746 option is on by default if you optimize.
26748 @item -mno-prolog-function
26749 @itemx -mprolog-function
26750 @opindex mno-prolog-function
26751 @opindex mprolog-function
26752 Do not use (do use) external functions to save and restore registers
26753 at the prologue and epilogue of a function. The external functions
26754 are slower, but use less code space if more than one function saves
26755 the same number of registers. The @option{-mprolog-function} option
26756 is on by default if you optimize.
26760 Try to make the code as small as possible. At present, this just turns
26761 on the @option{-mep} and @option{-mprolog-function} options.
26763 @item -mtda=@var{n}
26765 Put static or global variables whose size is @var{n} bytes or less into
26766 the tiny data area that register @code{ep} points to. The tiny data
26767 area can hold up to 256 bytes in total (128 bytes for byte references).
26769 @item -msda=@var{n}
26771 Put static or global variables whose size is @var{n} bytes or less into
26772 the small data area that register @code{gp} points to. The small data
26773 area can hold up to 64 kilobytes.
26775 @item -mzda=@var{n}
26777 Put static or global variables whose size is @var{n} bytes or less into
26778 the first 32 kilobytes of memory.
26782 Specify that the target processor is the V850.
26786 Specify that the target processor is the V850E3V5. The preprocessor
26787 constant @code{__v850e3v5__} is defined if this option is used.
26791 Specify that the target processor is the V850E3V5. This is an alias for
26792 the @option{-mv850e3v5} option.
26796 Specify that the target processor is the V850E2V3. The preprocessor
26797 constant @code{__v850e2v3__} is defined if this option is used.
26801 Specify that the target processor is the V850E2. The preprocessor
26802 constant @code{__v850e2__} is defined if this option is used.
26806 Specify that the target processor is the V850E1. The preprocessor
26807 constants @code{__v850e1__} and @code{__v850e__} are defined if
26808 this option is used.
26812 Specify that the target processor is the V850ES. This is an alias for
26813 the @option{-mv850e1} option.
26817 Specify that the target processor is the V850E@. The preprocessor
26818 constant @code{__v850e__} is defined if this option is used.
26820 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
26821 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
26822 are defined then a default target processor is chosen and the
26823 relevant @samp{__v850*__} preprocessor constant is defined.
26825 The preprocessor constants @code{__v850} and @code{__v851__} are always
26826 defined, regardless of which processor variant is the target.
26828 @item -mdisable-callt
26829 @itemx -mno-disable-callt
26830 @opindex mdisable-callt
26831 @opindex mno-disable-callt
26832 This option suppresses generation of the @code{CALLT} instruction for the
26833 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
26836 This option is enabled by default when the RH850 ABI is
26837 in use (see @option{-mrh850-abi}), and disabled by default when the
26838 GCC ABI is in use. If @code{CALLT} instructions are being generated
26839 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
26845 Pass on (or do not pass on) the @option{-mrelax} command-line option
26849 @itemx -mno-long-jumps
26850 @opindex mlong-jumps
26851 @opindex mno-long-jumps
26852 Disable (or re-enable) the generation of PC-relative jump instructions.
26855 @itemx -mhard-float
26856 @opindex msoft-float
26857 @opindex mhard-float
26858 Disable (or re-enable) the generation of hardware floating point
26859 instructions. This option is only significant when the target
26860 architecture is @samp{V850E2V3} or higher. If hardware floating point
26861 instructions are being generated then the C preprocessor symbol
26862 @code{__FPU_OK__} is defined, otherwise the symbol
26863 @code{__NO_FPU__} is defined.
26867 Enables the use of the e3v5 LOOP instruction. The use of this
26868 instruction is not enabled by default when the e3v5 architecture is
26869 selected because its use is still experimental.
26873 @opindex mrh850-abi
26875 Enables support for the RH850 version of the V850 ABI. This is the
26876 default. With this version of the ABI the following rules apply:
26880 Integer sized structures and unions are returned via a memory pointer
26881 rather than a register.
26884 Large structures and unions (more than 8 bytes in size) are passed by
26888 Functions are aligned to 16-bit boundaries.
26891 The @option{-m8byte-align} command-line option is supported.
26894 The @option{-mdisable-callt} command-line option is enabled by
26895 default. The @option{-mno-disable-callt} command-line option is not
26899 When this version of the ABI is enabled the C preprocessor symbol
26900 @code{__V850_RH850_ABI__} is defined.
26904 Enables support for the old GCC version of the V850 ABI. With this
26905 version of the ABI the following rules apply:
26909 Integer sized structures and unions are returned in register @code{r10}.
26912 Large structures and unions (more than 8 bytes in size) are passed by
26916 Functions are aligned to 32-bit boundaries, unless optimizing for
26920 The @option{-m8byte-align} command-line option is not supported.
26923 The @option{-mdisable-callt} command-line option is supported but not
26924 enabled by default.
26927 When this version of the ABI is enabled the C preprocessor symbol
26928 @code{__V850_GCC_ABI__} is defined.
26930 @item -m8byte-align
26931 @itemx -mno-8byte-align
26932 @opindex m8byte-align
26933 @opindex mno-8byte-align
26934 Enables support for @code{double} and @code{long long} types to be
26935 aligned on 8-byte boundaries. The default is to restrict the
26936 alignment of all objects to at most 4-bytes. When
26937 @option{-m8byte-align} is in effect the C preprocessor symbol
26938 @code{__V850_8BYTE_ALIGN__} is defined.
26941 @opindex mbig-switch
26942 Generate code suitable for big switch tables. Use this option only if
26943 the assembler/linker complain about out of range branches within a switch
26948 This option causes r2 and r5 to be used in the code generated by
26949 the compiler. This setting is the default.
26951 @item -mno-app-regs
26952 @opindex mno-app-regs
26953 This option causes r2 and r5 to be treated as fixed registers.
26958 @subsection VAX Options
26959 @cindex VAX options
26961 These @samp{-m} options are defined for the VAX:
26966 Do not output certain jump instructions (@code{aobleq} and so on)
26967 that the Unix assembler for the VAX cannot handle across long
26972 Do output those jump instructions, on the assumption that the
26973 GNU assembler is being used.
26977 Output code for G-format floating-point numbers instead of D-format.
26980 @node Visium Options
26981 @subsection Visium Options
26982 @cindex Visium options
26988 A program which performs file I/O and is destined to run on an MCM target
26989 should be linked with this option. It causes the libraries libc.a and
26990 libdebug.a to be linked. The program should be run on the target under
26991 the control of the GDB remote debugging stub.
26995 A program which performs file I/O and is destined to run on the simulator
26996 should be linked with option. This causes libraries libc.a and libsim.a to
27000 @itemx -mhard-float
27002 @opindex mhard-float
27003 Generate code containing floating-point instructions. This is the
27007 @itemx -msoft-float
27009 @opindex msoft-float
27010 Generate code containing library calls for floating-point.
27012 @option{-msoft-float} changes the calling convention in the output file;
27013 therefore, it is only useful if you compile @emph{all} of a program with
27014 this option. In particular, you need to compile @file{libgcc.a}, the
27015 library that comes with GCC, with @option{-msoft-float} in order for
27018 @item -mcpu=@var{cpu_type}
27020 Set the instruction set, register set, and instruction scheduling parameters
27021 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
27022 @samp{mcm}, @samp{gr5} and @samp{gr6}.
27024 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
27026 By default (unless configured otherwise), GCC generates code for the GR5
27027 variant of the Visium architecture.
27029 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
27030 architecture. The only difference from GR5 code is that the compiler will
27031 generate block move instructions.
27033 @item -mtune=@var{cpu_type}
27035 Set the instruction scheduling parameters for machine type @var{cpu_type},
27036 but do not set the instruction set or register set that the option
27037 @option{-mcpu=@var{cpu_type}} would.
27041 Generate code for the supervisor mode, where there are no restrictions on
27042 the access to general registers. This is the default.
27045 @opindex muser-mode
27046 Generate code for the user mode, where the access to some general registers
27047 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
27048 mode; on the GR6, only registers r29 to r31 are affected.
27052 @subsection VMS Options
27054 These @samp{-m} options are defined for the VMS implementations:
27057 @item -mvms-return-codes
27058 @opindex mvms-return-codes
27059 Return VMS condition codes from @code{main}. The default is to return POSIX-style
27060 condition (e.g.@: error) codes.
27062 @item -mdebug-main=@var{prefix}
27063 @opindex mdebug-main=@var{prefix}
27064 Flag the first routine whose name starts with @var{prefix} as the main
27065 routine for the debugger.
27069 Default to 64-bit memory allocation routines.
27071 @item -mpointer-size=@var{size}
27072 @opindex mpointer-size=@var{size}
27073 Set the default size of pointers. Possible options for @var{size} are
27074 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
27075 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
27076 The later option disables @code{pragma pointer_size}.
27079 @node VxWorks Options
27080 @subsection VxWorks Options
27081 @cindex VxWorks Options
27083 The options in this section are defined for all VxWorks targets.
27084 Options specific to the target hardware are listed with the other
27085 options for that target.
27090 GCC can generate code for both VxWorks kernels and real time processes
27091 (RTPs). This option switches from the former to the latter. It also
27092 defines the preprocessor macro @code{__RTP__}.
27095 @opindex non-static
27096 Link an RTP executable against shared libraries rather than static
27097 libraries. The options @option{-static} and @option{-shared} can
27098 also be used for RTPs (@pxref{Link Options}); @option{-static}
27105 These options are passed down to the linker. They are defined for
27106 compatibility with Diab.
27109 @opindex Xbind-lazy
27110 Enable lazy binding of function calls. This option is equivalent to
27111 @option{-Wl,-z,now} and is defined for compatibility with Diab.
27115 Disable lazy binding of function calls. This option is the default and
27116 is defined for compatibility with Diab.
27120 @subsection x86 Options
27121 @cindex x86 Options
27123 These @samp{-m} options are defined for the x86 family of computers.
27127 @item -march=@var{cpu-type}
27129 Generate instructions for the machine type @var{cpu-type}. In contrast to
27130 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
27131 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
27132 to generate code that may not run at all on processors other than the one
27133 indicated. Specifying @option{-march=@var{cpu-type}} implies
27134 @option{-mtune=@var{cpu-type}}.
27136 The choices for @var{cpu-type} are:
27140 This selects the CPU to generate code for at compilation time by determining
27141 the processor type of the compiling machine. Using @option{-march=native}
27142 enables all instruction subsets supported by the local machine (hence
27143 the result might not run on different machines). Using @option{-mtune=native}
27144 produces code optimized for the local machine under the constraints
27145 of the selected instruction set.
27148 A generic CPU with 64-bit extensions.
27151 Original Intel i386 CPU@.
27154 Intel i486 CPU@. (No scheduling is implemented for this chip.)
27158 Intel Pentium CPU with no MMX support.
27161 Intel Lakemont MCU, based on Intel Pentium CPU.
27164 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
27167 Intel Pentium Pro CPU@.
27170 When used with @option{-march}, the Pentium Pro
27171 instruction set is used, so the code runs on all i686 family chips.
27172 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
27175 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
27180 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
27184 Intel Pentium M; low-power version of Intel Pentium III CPU
27185 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
27189 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
27192 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
27196 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
27197 SSE2 and SSE3 instruction set support.
27200 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
27201 instruction set support.
27204 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27205 SSE4.1, SSE4.2 and POPCNT instruction set support.
27208 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27209 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
27212 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27213 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
27216 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27217 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
27218 instruction set support.
27221 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27222 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27223 BMI, BMI2 and F16C instruction set support.
27226 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27227 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27228 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
27231 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27232 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27233 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
27234 XSAVES instruction set support.
27237 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
27238 instruction set support.
27241 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27242 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
27245 Intel Goldmont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27246 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL, RDRND, XSAVE, XSAVEOPT and FSGSBASE
27247 instruction set support.
27249 @item goldmont-plus
27250 Intel Goldmont Plus CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27251 SSSE3, SSE4.1, SSE4.2, POPCNT, AES, PCLMUL, RDRND, XSAVE, XSAVEOPT, FSGSBASE,
27252 PTWRITE, RDPID, SGX and UMIP instruction set support.
27255 Intel Tremont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27256 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL, RDRND, XSAVE, XSAVEOPT, FSGSBASE, PTWRITE,
27257 RDPID, SGX, UMIP, GFNI-SSE, CLWB and ENCLV instruction set support.
27260 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27261 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27262 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
27263 AVX512CD instruction set support.
27266 Intel Knights Mill CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27267 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27268 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER, AVX512CD,
27269 AVX5124VNNIW, AVX5124FMAPS and AVX512VPOPCNTDQ instruction set support.
27271 @item skylake-avx512
27272 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27273 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27274 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
27275 CLWB, AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
27278 Intel Cannonlake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
27279 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
27280 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
27281 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
27282 AVX512IFMA, SHA and UMIP instruction set support.
27284 @item icelake-client
27285 Intel Icelake Client CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
27286 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
27287 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
27288 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
27289 AVX512IFMA, SHA, CLWB, UMIP, RDPID, GFNI, AVX512VBMI2, AVX512VPOPCNTDQ,
27290 AVX512BITALG, AVX512VNNI, VPCLMULQDQ, VAES instruction set support.
27292 @item icelake-server
27293 Intel Icelake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
27294 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
27295 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
27296 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
27297 AVX512IFMA, SHA, CLWB, UMIP, RDPID, GFNI, AVX512VBMI2, AVX512VPOPCNTDQ,
27298 AVX512BITALG, AVX512VNNI, VPCLMULQDQ, VAES, PCONFIG and WBNOINVD instruction
27302 Intel Cascadelake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27303 SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI,
27304 BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F, CLWB,
27305 AVX512VL, AVX512BW, AVX512DQ, AVX512CD and AVX512VNNI instruction set support.
27308 AMD K6 CPU with MMX instruction set support.
27312 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
27315 @itemx athlon-tbird
27316 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
27322 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
27323 instruction set support.
27329 Processors based on the AMD K8 core with x86-64 instruction set support,
27330 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
27331 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
27332 instruction set extensions.)
27335 @itemx opteron-sse3
27336 @itemx athlon64-sse3
27337 Improved versions of AMD K8 cores with SSE3 instruction set support.
27341 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
27342 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
27343 instruction set extensions.)
27346 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
27347 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
27348 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
27350 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
27351 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
27352 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
27355 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
27356 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
27357 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
27358 64-bit instruction set extensions.
27360 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
27361 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
27362 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
27363 SSE4.2, ABM and 64-bit instruction set extensions.
27366 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
27367 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
27368 SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
27369 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
27370 instruction set extensions.
27372 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
27373 supersets BMI, BMI2, ,CLWB, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED,
27374 MWAITX, SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A,
27375 SSSE3, SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
27376 instruction set extensions.)
27380 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
27381 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
27382 instruction set extensions.)
27385 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
27386 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
27387 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
27390 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
27394 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
27395 instruction set support.
27398 VIA C3 CPU with MMX and 3DNow!@: instruction set support.
27399 (No scheduling is implemented for this chip.)
27402 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
27403 (No scheduling is implemented for this chip.)
27406 VIA C7 (Esther) CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
27407 (No scheduling is implemented for this chip.)
27410 VIA Eden Samuel 2 CPU with MMX and 3DNow!@: instruction set support.
27411 (No scheduling is implemented for this chip.)
27414 VIA Eden Nehemiah CPU with MMX and SSE instruction set support.
27415 (No scheduling is implemented for this chip.)
27418 VIA Eden Esther CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
27419 (No scheduling is implemented for this chip.)
27422 VIA Eden X2 CPU with x86-64, MMX, SSE, SSE2 and SSE3 instruction set support.
27423 (No scheduling is implemented for this chip.)
27426 VIA Eden X4 CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2,
27427 AVX and AVX2 instruction set support.
27428 (No scheduling is implemented for this chip.)
27431 Generic VIA Nano CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
27432 instruction set support.
27433 (No scheduling is implemented for this chip.)
27436 VIA Nano 1xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
27437 instruction set support.
27438 (No scheduling is implemented for this chip.)
27441 VIA Nano 2xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
27442 instruction set support.
27443 (No scheduling is implemented for this chip.)
27446 VIA Nano 3xxx CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
27447 instruction set support.
27448 (No scheduling is implemented for this chip.)
27451 VIA Nano Dual Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
27452 instruction set support.
27453 (No scheduling is implemented for this chip.)
27456 VIA Nano Quad Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
27457 instruction set support.
27458 (No scheduling is implemented for this chip.)
27461 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
27464 @item -mtune=@var{cpu-type}
27466 Tune to @var{cpu-type} everything applicable about the generated code, except
27467 for the ABI and the set of available instructions.
27468 While picking a specific @var{cpu-type} schedules things appropriately
27469 for that particular chip, the compiler does not generate any code that
27470 cannot run on the default machine type unless you use a
27471 @option{-march=@var{cpu-type}} option.
27472 For example, if GCC is configured for i686-pc-linux-gnu
27473 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
27474 but still runs on i686 machines.
27476 The choices for @var{cpu-type} are the same as for @option{-march}.
27477 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
27481 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
27482 If you know the CPU on which your code will run, then you should use
27483 the corresponding @option{-mtune} or @option{-march} option instead of
27484 @option{-mtune=generic}. But, if you do not know exactly what CPU users
27485 of your application will have, then you should use this option.
27487 As new processors are deployed in the marketplace, the behavior of this
27488 option will change. Therefore, if you upgrade to a newer version of
27489 GCC, code generation controlled by this option will change to reflect
27491 that are most common at the time that version of GCC is released.
27493 There is no @option{-march=generic} option because @option{-march}
27494 indicates the instruction set the compiler can use, and there is no
27495 generic instruction set applicable to all processors. In contrast,
27496 @option{-mtune} indicates the processor (or, in this case, collection of
27497 processors) for which the code is optimized.
27500 Produce code optimized for the most current Intel processors, which are
27501 Haswell and Silvermont for this version of GCC. If you know the CPU
27502 on which your code will run, then you should use the corresponding
27503 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
27504 But, if you want your application performs better on both Haswell and
27505 Silvermont, then you should use this option.
27507 As new Intel processors are deployed in the marketplace, the behavior of
27508 this option will change. Therefore, if you upgrade to a newer version of
27509 GCC, code generation controlled by this option will change to reflect
27510 the most current Intel processors at the time that version of GCC is
27513 There is no @option{-march=intel} option because @option{-march} indicates
27514 the instruction set the compiler can use, and there is no common
27515 instruction set applicable to all processors. In contrast,
27516 @option{-mtune} indicates the processor (or, in this case, collection of
27517 processors) for which the code is optimized.
27520 @item -mcpu=@var{cpu-type}
27522 A deprecated synonym for @option{-mtune}.
27524 @item -mfpmath=@var{unit}
27526 Generate floating-point arithmetic for selected unit @var{unit}. The choices
27527 for @var{unit} are:
27531 Use the standard 387 floating-point coprocessor present on the majority of chips and
27532 emulated otherwise. Code compiled with this option runs almost everywhere.
27533 The temporary results are computed in 80-bit precision instead of the precision
27534 specified by the type, resulting in slightly different results compared to most
27535 of other chips. See @option{-ffloat-store} for more detailed description.
27537 This is the default choice for non-Darwin x86-32 targets.
27540 Use scalar floating-point instructions present in the SSE instruction set.
27541 This instruction set is supported by Pentium III and newer chips,
27542 and in the AMD line
27543 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
27544 instruction set supports only single-precision arithmetic, thus the double and
27545 extended-precision arithmetic are still done using 387. A later version, present
27546 only in Pentium 4 and AMD x86-64 chips, supports double-precision
27549 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
27550 or @option{-msse2} switches to enable SSE extensions and make this option
27551 effective. For the x86-64 compiler, these extensions are enabled by default.
27553 The resulting code should be considerably faster in the majority of cases and avoid
27554 the numerical instability problems of 387 code, but may break some existing
27555 code that expects temporaries to be 80 bits.
27557 This is the default choice for the x86-64 compiler, Darwin x86-32 targets,
27558 and the default choice for x86-32 targets with the SSE2 instruction set
27559 when @option{-ffast-math} is enabled.
27564 Attempt to utilize both instruction sets at once. This effectively doubles the
27565 amount of available registers, and on chips with separate execution units for
27566 387 and SSE the execution resources too. Use this option with care, as it is
27567 still experimental, because the GCC register allocator does not model separate
27568 functional units well, resulting in unstable performance.
27571 @item -masm=@var{dialect}
27572 @opindex masm=@var{dialect}
27573 Output assembly instructions using selected @var{dialect}. Also affects
27574 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
27575 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
27576 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
27577 not support @samp{intel}.
27580 @itemx -mno-ieee-fp
27582 @opindex mno-ieee-fp
27583 Control whether or not the compiler uses IEEE floating-point
27584 comparisons. These correctly handle the case where the result of a
27585 comparison is unordered.
27588 @itemx -mhard-float
27590 @opindex mhard-float
27591 Generate output containing 80387 instructions for floating point.
27594 @itemx -msoft-float
27596 @opindex msoft-float
27597 Generate output containing library calls for floating point.
27599 @strong{Warning:} the requisite libraries are not part of GCC@.
27600 Normally the facilities of the machine's usual C compiler are used, but
27601 this cannot be done directly in cross-compilation. You must make your
27602 own arrangements to provide suitable library functions for
27605 On machines where a function returns floating-point results in the 80387
27606 register stack, some floating-point opcodes may be emitted even if
27607 @option{-msoft-float} is used.
27609 @item -mno-fp-ret-in-387
27610 @opindex mno-fp-ret-in-387
27611 @opindex mfp-ret-in-387
27612 Do not use the FPU registers for return values of functions.
27614 The usual calling convention has functions return values of types
27615 @code{float} and @code{double} in an FPU register, even if there
27616 is no FPU@. The idea is that the operating system should emulate
27619 The option @option{-mno-fp-ret-in-387} causes such values to be returned
27620 in ordinary CPU registers instead.
27622 @item -mno-fancy-math-387
27623 @opindex mno-fancy-math-387
27624 @opindex mfancy-math-387
27625 Some 387 emulators do not support the @code{sin}, @code{cos} and
27626 @code{sqrt} instructions for the 387. Specify this option to avoid
27627 generating those instructions.
27628 This option is overridden when @option{-march}
27629 indicates that the target CPU always has an FPU and so the
27630 instruction does not need emulation. These
27631 instructions are not generated unless you also use the
27632 @option{-funsafe-math-optimizations} switch.
27634 @item -malign-double
27635 @itemx -mno-align-double
27636 @opindex malign-double
27637 @opindex mno-align-double
27638 Control whether GCC aligns @code{double}, @code{long double}, and
27639 @code{long long} variables on a two-word boundary or a one-word
27640 boundary. Aligning @code{double} variables on a two-word boundary
27641 produces code that runs somewhat faster on a Pentium at the
27642 expense of more memory.
27644 On x86-64, @option{-malign-double} is enabled by default.
27646 @strong{Warning:} if you use the @option{-malign-double} switch,
27647 structures containing the above types are aligned differently than
27648 the published application binary interface specifications for the x86-32
27649 and are not binary compatible with structures in code compiled
27650 without that switch.
27652 @item -m96bit-long-double
27653 @itemx -m128bit-long-double
27654 @opindex m96bit-long-double
27655 @opindex m128bit-long-double
27656 These switches control the size of @code{long double} type. The x86-32
27657 application binary interface specifies the size to be 96 bits,
27658 so @option{-m96bit-long-double} is the default in 32-bit mode.
27660 Modern architectures (Pentium and newer) prefer @code{long double}
27661 to be aligned to an 8- or 16-byte boundary. In arrays or structures
27662 conforming to the ABI, this is not possible. So specifying
27663 @option{-m128bit-long-double} aligns @code{long double}
27664 to a 16-byte boundary by padding the @code{long double} with an additional
27667 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
27668 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
27670 Notice that neither of these options enable any extra precision over the x87
27671 standard of 80 bits for a @code{long double}.
27673 @strong{Warning:} if you override the default value for your target ABI, this
27674 changes the size of
27675 structures and arrays containing @code{long double} variables,
27676 as well as modifying the function calling convention for functions taking
27677 @code{long double}. Hence they are not binary-compatible
27678 with code compiled without that switch.
27680 @item -mlong-double-64
27681 @itemx -mlong-double-80
27682 @itemx -mlong-double-128
27683 @opindex mlong-double-64
27684 @opindex mlong-double-80
27685 @opindex mlong-double-128
27686 These switches control the size of @code{long double} type. A size
27687 of 64 bits makes the @code{long double} type equivalent to the @code{double}
27688 type. This is the default for 32-bit Bionic C library. A size
27689 of 128 bits makes the @code{long double} type equivalent to the
27690 @code{__float128} type. This is the default for 64-bit Bionic C library.
27692 @strong{Warning:} if you override the default value for your target ABI, this
27693 changes the size of
27694 structures and arrays containing @code{long double} variables,
27695 as well as modifying the function calling convention for functions taking
27696 @code{long double}. Hence they are not binary-compatible
27697 with code compiled without that switch.
27699 @item -malign-data=@var{type}
27700 @opindex malign-data
27701 Control how GCC aligns variables. Supported values for @var{type} are
27702 @samp{compat} uses increased alignment value compatible uses GCC 4.8
27703 and earlier, @samp{abi} uses alignment value as specified by the
27704 psABI, and @samp{cacheline} uses increased alignment value to match
27705 the cache line size. @samp{compat} is the default.
27707 @item -mlarge-data-threshold=@var{threshold}
27708 @opindex mlarge-data-threshold
27709 When @option{-mcmodel=medium} is specified, data objects larger than
27710 @var{threshold} are placed in the large data section. This value must be the
27711 same across all objects linked into the binary, and defaults to 65535.
27715 Use a different function-calling convention, in which functions that
27716 take a fixed number of arguments return with the @code{ret @var{num}}
27717 instruction, which pops their arguments while returning. This saves one
27718 instruction in the caller since there is no need to pop the arguments
27721 You can specify that an individual function is called with this calling
27722 sequence with the function attribute @code{stdcall}. You can also
27723 override the @option{-mrtd} option by using the function attribute
27724 @code{cdecl}. @xref{Function Attributes}.
27726 @strong{Warning:} this calling convention is incompatible with the one
27727 normally used on Unix, so you cannot use it if you need to call
27728 libraries compiled with the Unix compiler.
27730 Also, you must provide function prototypes for all functions that
27731 take variable numbers of arguments (including @code{printf});
27732 otherwise incorrect code is generated for calls to those
27735 In addition, seriously incorrect code results if you call a
27736 function with too many arguments. (Normally, extra arguments are
27737 harmlessly ignored.)
27739 @item -mregparm=@var{num}
27741 Control how many registers are used to pass integer arguments. By
27742 default, no registers are used to pass arguments, and at most 3
27743 registers can be used. You can control this behavior for a specific
27744 function by using the function attribute @code{regparm}.
27745 @xref{Function Attributes}.
27747 @strong{Warning:} if you use this switch, and
27748 @var{num} is nonzero, then you must build all modules with the same
27749 value, including any libraries. This includes the system libraries and
27753 @opindex msseregparm
27754 Use SSE register passing conventions for float and double arguments
27755 and return values. You can control this behavior for a specific
27756 function by using the function attribute @code{sseregparm}.
27757 @xref{Function Attributes}.
27759 @strong{Warning:} if you use this switch then you must build all
27760 modules with the same value, including any libraries. This includes
27761 the system libraries and startup modules.
27763 @item -mvect8-ret-in-mem
27764 @opindex mvect8-ret-in-mem
27765 Return 8-byte vectors in memory instead of MMX registers. This is the
27766 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
27767 Studio compilers until version 12. Later compiler versions (starting
27768 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
27769 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
27770 you need to remain compatible with existing code produced by those
27771 previous compiler versions or older versions of GCC@.
27780 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
27781 is specified, the significands of results of floating-point operations are
27782 rounded to 24 bits (single precision); @option{-mpc64} rounds the
27783 significands of results of floating-point operations to 53 bits (double
27784 precision) and @option{-mpc80} rounds the significands of results of
27785 floating-point operations to 64 bits (extended double precision), which is
27786 the default. When this option is used, floating-point operations in higher
27787 precisions are not available to the programmer without setting the FPU
27788 control word explicitly.
27790 Setting the rounding of floating-point operations to less than the default
27791 80 bits can speed some programs by 2% or more. Note that some mathematical
27792 libraries assume that extended-precision (80-bit) floating-point operations
27793 are enabled by default; routines in such libraries could suffer significant
27794 loss of accuracy, typically through so-called ``catastrophic cancellation'',
27795 when this option is used to set the precision to less than extended precision.
27797 @item -mstackrealign
27798 @opindex mstackrealign
27799 Realign the stack at entry. On the x86, the @option{-mstackrealign}
27800 option generates an alternate prologue and epilogue that realigns the
27801 run-time stack if necessary. This supports mixing legacy codes that keep
27802 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
27803 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
27804 applicable to individual functions.
27806 @item -mpreferred-stack-boundary=@var{num}
27807 @opindex mpreferred-stack-boundary
27808 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
27809 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
27810 the default is 4 (16 bytes or 128 bits).
27812 @strong{Warning:} When generating code for the x86-64 architecture with
27813 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
27814 used to keep the stack boundary aligned to 8 byte boundary. Since
27815 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
27816 intended to be used in controlled environment where stack space is
27817 important limitation. This option leads to wrong code when functions
27818 compiled with 16 byte stack alignment (such as functions from a standard
27819 library) are called with misaligned stack. In this case, SSE
27820 instructions may lead to misaligned memory access traps. In addition,
27821 variable arguments are handled incorrectly for 16 byte aligned
27822 objects (including x87 long double and __int128), leading to wrong
27823 results. You must build all modules with
27824 @option{-mpreferred-stack-boundary=3}, including any libraries. This
27825 includes the system libraries and startup modules.
27827 @item -mincoming-stack-boundary=@var{num}
27828 @opindex mincoming-stack-boundary
27829 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
27830 boundary. If @option{-mincoming-stack-boundary} is not specified,
27831 the one specified by @option{-mpreferred-stack-boundary} is used.
27833 On Pentium and Pentium Pro, @code{double} and @code{long double} values
27834 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
27835 suffer significant run time performance penalties. On Pentium III, the
27836 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
27837 properly if it is not 16-byte aligned.
27839 To ensure proper alignment of this values on the stack, the stack boundary
27840 must be as aligned as that required by any value stored on the stack.
27841 Further, every function must be generated such that it keeps the stack
27842 aligned. Thus calling a function compiled with a higher preferred
27843 stack boundary from a function compiled with a lower preferred stack
27844 boundary most likely misaligns the stack. It is recommended that
27845 libraries that use callbacks always use the default setting.
27847 This extra alignment does consume extra stack space, and generally
27848 increases code size. Code that is sensitive to stack space usage, such
27849 as embedded systems and operating system kernels, may want to reduce the
27850 preferred alignment to @option{-mpreferred-stack-boundary=2}.
27907 @itemx -mavx512ifma
27908 @opindex mavx512ifma
27910 @itemx -mavx512vbmi
27911 @opindex mavx512vbmi
27922 @itemx -mclflushopt
27923 @opindex mclflushopt
27958 @itemx -mprefetchwt1
27959 @opindex mprefetchwt1
28030 @itemx -mavx512vbmi2
28031 @opindex mavx512vbmi2
28042 @itemx -mvpclmulqdq
28043 @opindex mvpclmulqdq
28045 @itemx -mavx512bitalg
28046 @opindex mavx512bitalg
28052 @opindex mmovdir64b
28054 @itemx -mavx512vpopcntdq
28055 @opindex mavx512vpopcntdq
28057 @itemx -mavx5124fmaps
28058 @opindex mavx5124fmaps
28060 @itemx -mavx512vnni
28061 @opindex mavx512vnni
28063 @itemx -mavx5124vnniw
28064 @opindex mavx5124vnniw
28068 These switches enable the use of instructions in the MMX, SSE,
28069 SSE2, SSE3, SSSE3, SSE4, SSE4A, SSE4.1, SSE4.2, AVX, AVX2, AVX512F, AVX512PF,
28070 AVX512ER, AVX512CD, AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA, AVX512VBMI, SHA,
28071 AES, PCLMUL, CLFLUSHOPT, CLWB, FSGSBASE, PTWRITE, RDRND, F16C, FMA, PCONFIG,
28072 WBNOINVD, FMA4, PREFETCHW, RDPID, PREFETCHWT1, RDSEED, SGX, XOP, LWP,
28073 3DNow!@:, enhanced 3DNow!@:, POPCNT, ABM, ADX, BMI, BMI2, LZCNT, FXSR, XSAVE,
28074 XSAVEOPT, XSAVEC, XSAVES, RTM, HLE, TBM, MWAITX, CLZERO, PKU, AVX512VBMI2,
28075 GFNI, VAES, WAITPKG, VPCLMULQDQ, AVX512BITALG, MOVDIRI, MOVDIR64B,
28076 AVX512VPOPCNTDQ, AVX5124FMAPS, AVX512VNNI, AVX5124VNNIW, or CLDEMOTE
28077 extended instruction sets. Each has a corresponding @option{-mno-} option to
28078 disable use of these instructions.
28080 These extensions are also available as built-in functions: see
28081 @ref{x86 Built-in Functions}, for details of the functions enabled and
28082 disabled by these switches.
28084 To generate SSE/SSE2 instructions automatically from floating-point
28085 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
28087 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
28088 generates new AVX instructions or AVX equivalence for all SSEx instructions
28091 These options enable GCC to use these extended instructions in
28092 generated code, even without @option{-mfpmath=sse}. Applications that
28093 perform run-time CPU detection must compile separate files for each
28094 supported architecture, using the appropriate flags. In particular,
28095 the file containing the CPU detection code should be compiled without
28098 @item -mdump-tune-features
28099 @opindex mdump-tune-features
28100 This option instructs GCC to dump the names of the x86 performance
28101 tuning features and default settings. The names can be used in
28102 @option{-mtune-ctrl=@var{feature-list}}.
28104 @item -mtune-ctrl=@var{feature-list}
28105 @opindex mtune-ctrl=@var{feature-list}
28106 This option is used to do fine grain control of x86 code generation features.
28107 @var{feature-list} is a comma separated list of @var{feature} names. See also
28108 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
28109 on if it is not preceded with @samp{^}, otherwise, it is turned off.
28110 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
28111 developers. Using it may lead to code paths not covered by testing and can
28112 potentially result in compiler ICEs or runtime errors.
28115 @opindex mno-default
28116 This option instructs GCC to turn off all tunable features. See also
28117 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
28121 This option instructs GCC to emit a @code{cld} instruction in the prologue
28122 of functions that use string instructions. String instructions depend on
28123 the DF flag to select between autoincrement or autodecrement mode. While the
28124 ABI specifies the DF flag to be cleared on function entry, some operating
28125 systems violate this specification by not clearing the DF flag in their
28126 exception dispatchers. The exception handler can be invoked with the DF flag
28127 set, which leads to wrong direction mode when string instructions are used.
28128 This option can be enabled by default on 32-bit x86 targets by configuring
28129 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
28130 instructions can be suppressed with the @option{-mno-cld} compiler option
28134 @opindex mvzeroupper
28135 This option instructs GCC to emit a @code{vzeroupper} instruction
28136 before a transfer of control flow out of the function to minimize
28137 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
28140 @item -mprefer-avx128
28141 @opindex mprefer-avx128
28142 This option instructs GCC to use 128-bit AVX instructions instead of
28143 256-bit AVX instructions in the auto-vectorizer.
28145 @item -mprefer-vector-width=@var{opt}
28146 @opindex mprefer-vector-width
28147 This option instructs GCC to use @var{opt}-bit vector width in instructions
28148 instead of default on the selected platform.
28152 No extra limitations applied to GCC other than defined by the selected platform.
28155 Prefer 128-bit vector width for instructions.
28158 Prefer 256-bit vector width for instructions.
28161 Prefer 512-bit vector width for instructions.
28166 This option enables GCC to generate @code{CMPXCHG16B} instructions in 64-bit
28167 code to implement compare-and-exchange operations on 16-byte aligned 128-bit
28168 objects. This is useful for atomic updates of data structures exceeding one
28169 machine word in size. The compiler uses this instruction to implement
28170 @ref{__sync Builtins}. However, for @ref{__atomic Builtins} operating on
28171 128-bit integers, a library call is always used.
28175 This option enables generation of @code{SAHF} instructions in 64-bit code.
28176 Early Intel Pentium 4 CPUs with Intel 64 support,
28177 prior to the introduction of Pentium 4 G1 step in December 2005,
28178 lacked the @code{LAHF} and @code{SAHF} instructions
28179 which are supported by AMD64.
28180 These are load and store instructions, respectively, for certain status flags.
28181 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
28182 @code{drem}, and @code{remainder} built-in functions;
28183 see @ref{Other Builtins} for details.
28187 This option enables use of the @code{movbe} instruction to implement
28188 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
28192 The @option{-mshstk} option enables shadow stack built-in functions
28193 from x86 Control-flow Enforcement Technology (CET).
28197 This option enables built-in functions @code{__builtin_ia32_crc32qi},
28198 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
28199 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
28203 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
28204 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
28205 with an additional Newton-Raphson step
28206 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
28207 (and their vectorized
28208 variants) for single-precision floating-point arguments. These instructions
28209 are generated only when @option{-funsafe-math-optimizations} is enabled
28210 together with @option{-ffinite-math-only} and @option{-fno-trapping-math}.
28211 Note that while the throughput of the sequence is higher than the throughput
28212 of the non-reciprocal instruction, the precision of the sequence can be
28213 decreased by up to 2 ulp (i.e.@: the inverse of 1.0 equals 0.99999994).
28215 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
28216 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
28217 combination), and doesn't need @option{-mrecip}.
28219 Also note that GCC emits the above sequence with additional Newton-Raphson step
28220 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
28221 already with @option{-ffast-math} (or the above option combination), and
28222 doesn't need @option{-mrecip}.
28224 @item -mrecip=@var{opt}
28225 @opindex mrecip=opt
28226 This option controls which reciprocal estimate instructions
28227 may be used. @var{opt} is a comma-separated list of options, which may
28228 be preceded by a @samp{!} to invert the option:
28232 Enable all estimate instructions.
28235 Enable the default instructions, equivalent to @option{-mrecip}.
28238 Disable all estimate instructions, equivalent to @option{-mno-recip}.
28241 Enable the approximation for scalar division.
28244 Enable the approximation for vectorized division.
28247 Enable the approximation for scalar square root.
28250 Enable the approximation for vectorized square root.
28253 So, for example, @option{-mrecip=all,!sqrt} enables
28254 all of the reciprocal approximations, except for square root.
28256 @item -mveclibabi=@var{type}
28257 @opindex mveclibabi
28258 Specifies the ABI type to use for vectorizing intrinsics using an
28259 external library. Supported values for @var{type} are @samp{svml}
28260 for the Intel short
28261 vector math library and @samp{acml} for the AMD math core library.
28262 To use this option, both @option{-ftree-vectorize} and
28263 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
28264 ABI-compatible library must be specified at link time.
28266 GCC currently emits calls to @code{vmldExp2},
28267 @code{vmldLn2}, @code{vmldLog102}, @code{vmldPow2},
28268 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
28269 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
28270 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
28271 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4},
28272 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
28273 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
28274 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
28275 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
28276 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
28277 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
28278 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
28279 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
28280 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
28281 when @option{-mveclibabi=acml} is used.
28283 @item -mabi=@var{name}
28285 Generate code for the specified calling convention. Permissible values
28286 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
28287 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
28288 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
28289 You can control this behavior for specific functions by
28290 using the function attributes @code{ms_abi} and @code{sysv_abi}.
28291 @xref{Function Attributes}.
28293 @item -mforce-indirect-call
28294 @opindex mforce-indirect-call
28295 Force all calls to functions to be indirect. This is useful
28296 when using Intel Processor Trace where it generates more precise timing
28297 information for function calls.
28299 @item -mmanual-endbr
28300 @opindex mmanual-endbr
28301 Insert ENDBR instruction at function entry only via the @code{cf_check}
28302 function attribute. This is useful when used with the option
28303 @option{-fcf-protection=branch} to control ENDBR insertion at the
28306 @item -mcall-ms2sysv-xlogues
28307 @opindex mcall-ms2sysv-xlogues
28308 @opindex mno-call-ms2sysv-xlogues
28309 Due to differences in 64-bit ABIs, any Microsoft ABI function that calls a
28310 System V ABI function must consider RSI, RDI and XMM6-15 as clobbered. By
28311 default, the code for saving and restoring these registers is emitted inline,
28312 resulting in fairly lengthy prologues and epilogues. Using
28313 @option{-mcall-ms2sysv-xlogues} emits prologues and epilogues that
28314 use stubs in the static portion of libgcc to perform these saves and restores,
28315 thus reducing function size at the cost of a few extra instructions.
28317 @item -mtls-dialect=@var{type}
28318 @opindex mtls-dialect
28319 Generate code to access thread-local storage using the @samp{gnu} or
28320 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
28321 @samp{gnu2} is more efficient, but it may add compile- and run-time
28322 requirements that cannot be satisfied on all systems.
28325 @itemx -mno-push-args
28326 @opindex mpush-args
28327 @opindex mno-push-args
28328 Use PUSH operations to store outgoing parameters. This method is shorter
28329 and usually equally fast as method using SUB/MOV operations and is enabled
28330 by default. In some cases disabling it may improve performance because of
28331 improved scheduling and reduced dependencies.
28333 @item -maccumulate-outgoing-args
28334 @opindex maccumulate-outgoing-args
28335 If enabled, the maximum amount of space required for outgoing arguments is
28336 computed in the function prologue. This is faster on most modern CPUs
28337 because of reduced dependencies, improved scheduling and reduced stack usage
28338 when the preferred stack boundary is not equal to 2. The drawback is a notable
28339 increase in code size. This switch implies @option{-mno-push-args}.
28343 Support thread-safe exception handling on MinGW. Programs that rely
28344 on thread-safe exception handling must compile and link all code with the
28345 @option{-mthreads} option. When compiling, @option{-mthreads} defines
28346 @option{-D_MT}; when linking, it links in a special thread helper library
28347 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
28349 @item -mms-bitfields
28350 @itemx -mno-ms-bitfields
28351 @opindex mms-bitfields
28352 @opindex mno-ms-bitfields
28354 Enable/disable bit-field layout compatible with the native Microsoft
28357 If @code{packed} is used on a structure, or if bit-fields are used,
28358 it may be that the Microsoft ABI lays out the structure differently
28359 than the way GCC normally does. Particularly when moving packed
28360 data between functions compiled with GCC and the native Microsoft compiler
28361 (either via function call or as data in a file), it may be necessary to access
28364 This option is enabled by default for Microsoft Windows
28365 targets. This behavior can also be controlled locally by use of variable
28366 or type attributes. For more information, see @ref{x86 Variable Attributes}
28367 and @ref{x86 Type Attributes}.
28369 The Microsoft structure layout algorithm is fairly simple with the exception
28370 of the bit-field packing.
28371 The padding and alignment of members of structures and whether a bit-field
28372 can straddle a storage-unit boundary are determine by these rules:
28375 @item Structure members are stored sequentially in the order in which they are
28376 declared: the first member has the lowest memory address and the last member
28379 @item Every data object has an alignment requirement. The alignment requirement
28380 for all data except structures, unions, and arrays is either the size of the
28381 object or the current packing size (specified with either the
28382 @code{aligned} attribute or the @code{pack} pragma),
28383 whichever is less. For structures, unions, and arrays,
28384 the alignment requirement is the largest alignment requirement of its members.
28385 Every object is allocated an offset so that:
28388 offset % alignment_requirement == 0
28391 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
28392 unit if the integral types are the same size and if the next bit-field fits
28393 into the current allocation unit without crossing the boundary imposed by the
28394 common alignment requirements of the bit-fields.
28397 MSVC interprets zero-length bit-fields in the following ways:
28400 @item If a zero-length bit-field is inserted between two bit-fields that
28401 are normally coalesced, the bit-fields are not coalesced.
28408 unsigned long bf_1 : 12;
28410 unsigned long bf_2 : 12;
28415 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
28416 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
28418 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
28419 alignment of the zero-length bit-field is greater than the member that follows it,
28420 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
28441 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
28442 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
28443 bit-field does not affect the alignment of @code{bar} or, as a result, the size
28446 Taking this into account, it is important to note the following:
28449 @item If a zero-length bit-field follows a normal bit-field, the type of the
28450 zero-length bit-field may affect the alignment of the structure as whole. For
28451 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
28452 normal bit-field, and is of type short.
28454 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
28455 still affect the alignment of the structure:
28466 Here, @code{t4} takes up 4 bytes.
28469 @item Zero-length bit-fields following non-bit-field members are ignored:
28481 Here, @code{t5} takes up 2 bytes.
28485 @item -mno-align-stringops
28486 @opindex mno-align-stringops
28487 @opindex malign-stringops
28488 Do not align the destination of inlined string operations. This switch reduces
28489 code size and improves performance in case the destination is already aligned,
28490 but GCC doesn't know about it.
28492 @item -minline-all-stringops
28493 @opindex minline-all-stringops
28494 By default GCC inlines string operations only when the destination is
28495 known to be aligned to least a 4-byte boundary.
28496 This enables more inlining and increases code
28497 size, but may improve performance of code that depends on fast
28498 @code{memcpy}, @code{strlen},
28499 and @code{memset} for short lengths.
28501 @item -minline-stringops-dynamically
28502 @opindex minline-stringops-dynamically
28503 For string operations of unknown size, use run-time checks with
28504 inline code for small blocks and a library call for large blocks.
28506 @item -mstringop-strategy=@var{alg}
28507 @opindex mstringop-strategy=@var{alg}
28508 Override the internal decision heuristic for the particular algorithm to use
28509 for inlining string operations. The allowed values for @var{alg} are:
28515 Expand using i386 @code{rep} prefix of the specified size.
28519 @itemx unrolled_loop
28520 Expand into an inline loop.
28523 Always use a library call.
28526 @item -mmemcpy-strategy=@var{strategy}
28527 @opindex mmemcpy-strategy=@var{strategy}
28528 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
28529 should be inlined and what inline algorithm to use when the expected size
28530 of the copy operation is known. @var{strategy}
28531 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
28532 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
28533 the max byte size with which inline algorithm @var{alg} is allowed. For the last
28534 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
28535 in the list must be specified in increasing order. The minimal byte size for
28536 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
28539 @item -mmemset-strategy=@var{strategy}
28540 @opindex mmemset-strategy=@var{strategy}
28541 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
28542 @code{__builtin_memset} expansion.
28544 @item -momit-leaf-frame-pointer
28545 @opindex momit-leaf-frame-pointer
28546 Don't keep the frame pointer in a register for leaf functions. This
28547 avoids the instructions to save, set up, and restore frame pointers and
28548 makes an extra register available in leaf functions. The option
28549 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
28550 which might make debugging harder.
28552 @item -mtls-direct-seg-refs
28553 @itemx -mno-tls-direct-seg-refs
28554 @opindex mtls-direct-seg-refs
28555 Controls whether TLS variables may be accessed with offsets from the
28556 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
28557 or whether the thread base pointer must be added. Whether or not this
28558 is valid depends on the operating system, and whether it maps the
28559 segment to cover the entire TLS area.
28561 For systems that use the GNU C Library, the default is on.
28564 @itemx -mno-sse2avx
28566 Specify that the assembler should encode SSE instructions with VEX
28567 prefix. The option @option{-mavx} turns this on by default.
28572 If profiling is active (@option{-pg}), put the profiling
28573 counter call before the prologue.
28574 Note: On x86 architectures the attribute @code{ms_hook_prologue}
28575 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
28577 @item -mrecord-mcount
28578 @itemx -mno-record-mcount
28579 @opindex mrecord-mcount
28580 If profiling is active (@option{-pg}), generate a __mcount_loc section
28581 that contains pointers to each profiling call. This is useful for
28582 automatically patching and out calls.
28585 @itemx -mno-nop-mcount
28586 @opindex mnop-mcount
28587 If profiling is active (@option{-pg}), generate the calls to
28588 the profiling functions as NOPs. This is useful when they
28589 should be patched in later dynamically. This is likely only
28590 useful together with @option{-mrecord-mcount}.
28592 @item -minstrument-return=@var{type}
28593 @opindex minstrument-return
28594 Instrument function exit in -pg -mfentry instrumented functions with
28595 call to specified function. This only instruments true returns ending
28596 with ret, but not sibling calls ending with jump. Valid types
28597 are @var{none} to not instrument, @var{call} to generate a call to __return__,
28598 or @var{nop5} to generate a 5 byte nop.
28600 @item -mrecord-return
28601 @itemx -mno-record-return
28602 @opindex mrecord-return
28603 Generate a __return_loc section pointing to all return instrumentation code.
28605 @item -mfentry-name=@var{name}
28606 @opindex mfentry-name
28607 Set name of __fentry__ symbol called at function entry for -pg -mfentry functions.
28609 @item -mfentry-section=@var{name}
28610 @opindex mfentry-section
28611 Set name of section to record -mrecord-mcount calls (default __mcount_loc).
28613 @item -mskip-rax-setup
28614 @itemx -mno-skip-rax-setup
28615 @opindex mskip-rax-setup
28616 When generating code for the x86-64 architecture with SSE extensions
28617 disabled, @option{-mskip-rax-setup} can be used to skip setting up RAX
28618 register when there are no variable arguments passed in vector registers.
28620 @strong{Warning:} Since RAX register is used to avoid unnecessarily
28621 saving vector registers on stack when passing variable arguments, the
28622 impacts of this option are callees may waste some stack space,
28623 misbehave or jump to a random location. GCC 4.4 or newer don't have
28624 those issues, regardless the RAX register value.
28627 @itemx -mno-8bit-idiv
28628 @opindex m8bit-idiv
28629 On some processors, like Intel Atom, 8-bit unsigned integer divide is
28630 much faster than 32-bit/64-bit integer divide. This option generates a
28631 run-time check. If both dividend and divisor are within range of 0
28632 to 255, 8-bit unsigned integer divide is used instead of
28633 32-bit/64-bit integer divide.
28635 @item -mavx256-split-unaligned-load
28636 @itemx -mavx256-split-unaligned-store
28637 @opindex mavx256-split-unaligned-load
28638 @opindex mavx256-split-unaligned-store
28639 Split 32-byte AVX unaligned load and store.
28641 @item -mstack-protector-guard=@var{guard}
28642 @itemx -mstack-protector-guard-reg=@var{reg}
28643 @itemx -mstack-protector-guard-offset=@var{offset}
28644 @opindex mstack-protector-guard
28645 @opindex mstack-protector-guard-reg
28646 @opindex mstack-protector-guard-offset
28647 Generate stack protection code using canary at @var{guard}. Supported
28648 locations are @samp{global} for global canary or @samp{tls} for per-thread
28649 canary in the TLS block (the default). This option has effect only when
28650 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
28652 With the latter choice the options
28653 @option{-mstack-protector-guard-reg=@var{reg}} and
28654 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
28655 which segment register (@code{%fs} or @code{%gs}) to use as base register
28656 for reading the canary, and from what offset from that base register.
28657 The default for those is as specified in the relevant ABI.
28659 @item -mgeneral-regs-only
28660 @opindex mgeneral-regs-only
28661 Generate code that uses only the general-purpose registers. This
28662 prevents the compiler from using floating-point, vector, mask and bound
28665 @item -mindirect-branch=@var{choice}
28666 @opindex mindirect-branch
28667 Convert indirect call and jump with @var{choice}. The default is
28668 @samp{keep}, which keeps indirect call and jump unmodified.
28669 @samp{thunk} converts indirect call and jump to call and return thunk.
28670 @samp{thunk-inline} converts indirect call and jump to inlined call
28671 and return thunk. @samp{thunk-extern} converts indirect call and jump
28672 to external call and return thunk provided in a separate object file.
28673 You can control this behavior for a specific function by using the
28674 function attribute @code{indirect_branch}. @xref{Function Attributes}.
28676 Note that @option{-mcmodel=large} is incompatible with
28677 @option{-mindirect-branch=thunk} and
28678 @option{-mindirect-branch=thunk-extern} since the thunk function may
28679 not be reachable in the large code model.
28681 Note that @option{-mindirect-branch=thunk-extern} is incompatible with
28682 @option{-fcf-protection=branch} since the external thunk cannot be modified
28683 to disable control-flow check.
28685 @item -mfunction-return=@var{choice}
28686 @opindex mfunction-return
28687 Convert function return with @var{choice}. The default is @samp{keep},
28688 which keeps function return unmodified. @samp{thunk} converts function
28689 return to call and return thunk. @samp{thunk-inline} converts function
28690 return to inlined call and return thunk. @samp{thunk-extern} converts
28691 function return to external call and return thunk provided in a separate
28692 object file. You can control this behavior for a specific function by
28693 using the function attribute @code{function_return}.
28694 @xref{Function Attributes}.
28696 Note that @option{-mcmodel=large} is incompatible with
28697 @option{-mfunction-return=thunk} and
28698 @option{-mfunction-return=thunk-extern} since the thunk function may
28699 not be reachable in the large code model.
28702 @item -mindirect-branch-register
28703 @opindex mindirect-branch-register
28704 Force indirect call and jump via register.
28708 These @samp{-m} switches are supported in addition to the above
28709 on x86-64 processors in 64-bit environments.
28722 Generate code for a 16-bit, 32-bit or 64-bit environment.
28723 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
28725 generates code that runs on any i386 system.
28727 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
28728 types to 64 bits, and generates code for the x86-64 architecture.
28729 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
28730 and @option{-mdynamic-no-pic} options.
28732 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
28734 generates code for the x86-64 architecture.
28736 The @option{-m16} option is the same as @option{-m32}, except for that
28737 it outputs the @code{.code16gcc} assembly directive at the beginning of
28738 the assembly output so that the binary can run in 16-bit mode.
28740 The @option{-miamcu} option generates code which conforms to Intel MCU
28741 psABI. It requires the @option{-m32} option to be turned on.
28743 @item -mno-red-zone
28744 @opindex mno-red-zone
28746 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
28747 by the x86-64 ABI; it is a 128-byte area beyond the location of the
28748 stack pointer that is not modified by signal or interrupt handlers
28749 and therefore can be used for temporary data without adjusting the stack
28750 pointer. The flag @option{-mno-red-zone} disables this red zone.
28752 @item -mcmodel=small
28753 @opindex mcmodel=small
28754 Generate code for the small code model: the program and its symbols must
28755 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
28756 Programs can be statically or dynamically linked. This is the default
28759 @item -mcmodel=kernel
28760 @opindex mcmodel=kernel
28761 Generate code for the kernel code model. The kernel runs in the
28762 negative 2 GB of the address space.
28763 This model has to be used for Linux kernel code.
28765 @item -mcmodel=medium
28766 @opindex mcmodel=medium
28767 Generate code for the medium model: the program is linked in the lower 2
28768 GB of the address space. Small symbols are also placed there. Symbols
28769 with sizes larger than @option{-mlarge-data-threshold} are put into
28770 large data or BSS sections and can be located above 2GB. Programs can
28771 be statically or dynamically linked.
28773 @item -mcmodel=large
28774 @opindex mcmodel=large
28775 Generate code for the large model. This model makes no assumptions
28776 about addresses and sizes of sections.
28778 @item -maddress-mode=long
28779 @opindex maddress-mode=long
28780 Generate code for long address mode. This is only supported for 64-bit
28781 and x32 environments. It is the default address mode for 64-bit
28784 @item -maddress-mode=short
28785 @opindex maddress-mode=short
28786 Generate code for short address mode. This is only supported for 32-bit
28787 and x32 environments. It is the default address mode for 32-bit and
28791 @node x86 Windows Options
28792 @subsection x86 Windows Options
28793 @cindex x86 Windows Options
28794 @cindex Windows Options for x86
28796 These additional options are available for Microsoft Windows targets:
28802 specifies that a console application is to be generated, by
28803 instructing the linker to set the PE header subsystem type
28804 required for console applications.
28805 This option is available for Cygwin and MinGW targets and is
28806 enabled by default on those targets.
28810 This option is available for Cygwin and MinGW targets. It
28811 specifies that a DLL---a dynamic link library---is to be
28812 generated, enabling the selection of the required runtime
28813 startup object and entry point.
28815 @item -mnop-fun-dllimport
28816 @opindex mnop-fun-dllimport
28817 This option is available for Cygwin and MinGW targets. It
28818 specifies that the @code{dllimport} attribute should be ignored.
28822 This option is available for MinGW targets. It specifies
28823 that MinGW-specific thread support is to be used.
28827 This option is available for MinGW-w64 targets. It causes
28828 the @code{UNICODE} preprocessor macro to be predefined, and
28829 chooses Unicode-capable runtime startup code.
28833 This option is available for Cygwin and MinGW targets. It
28834 specifies that the typical Microsoft Windows predefined macros are to
28835 be set in the pre-processor, but does not influence the choice
28836 of runtime library/startup code.
28840 This option is available for Cygwin and MinGW targets. It
28841 specifies that a GUI application is to be generated by
28842 instructing the linker to set the PE header subsystem type
28845 @item -fno-set-stack-executable
28846 @opindex fno-set-stack-executable
28847 @opindex fset-stack-executable
28848 This option is available for MinGW targets. It specifies that
28849 the executable flag for the stack used by nested functions isn't
28850 set. This is necessary for binaries running in kernel mode of
28851 Microsoft Windows, as there the User32 API, which is used to set executable
28852 privileges, isn't available.
28854 @item -fwritable-relocated-rdata
28855 @opindex fno-writable-relocated-rdata
28856 @opindex fwritable-relocated-rdata
28857 This option is available for MinGW and Cygwin targets. It specifies
28858 that relocated-data in read-only section is put into the @code{.data}
28859 section. This is a necessary for older runtimes not supporting
28860 modification of @code{.rdata} sections for pseudo-relocation.
28862 @item -mpe-aligned-commons
28863 @opindex mpe-aligned-commons
28864 This option is available for Cygwin and MinGW targets. It
28865 specifies that the GNU extension to the PE file format that
28866 permits the correct alignment of COMMON variables should be
28867 used when generating code. It is enabled by default if
28868 GCC detects that the target assembler found during configuration
28869 supports the feature.
28872 See also under @ref{x86 Options} for standard options.
28874 @node Xstormy16 Options
28875 @subsection Xstormy16 Options
28876 @cindex Xstormy16 Options
28878 These options are defined for Xstormy16:
28883 Choose startup files and linker script suitable for the simulator.
28886 @node Xtensa Options
28887 @subsection Xtensa Options
28888 @cindex Xtensa Options
28890 These options are supported for Xtensa targets:
28894 @itemx -mno-const16
28896 @opindex mno-const16
28897 Enable or disable use of @code{CONST16} instructions for loading
28898 constant values. The @code{CONST16} instruction is currently not a
28899 standard option from Tensilica. When enabled, @code{CONST16}
28900 instructions are always used in place of the standard @code{L32R}
28901 instructions. The use of @code{CONST16} is enabled by default only if
28902 the @code{L32R} instruction is not available.
28905 @itemx -mno-fused-madd
28906 @opindex mfused-madd
28907 @opindex mno-fused-madd
28908 Enable or disable use of fused multiply/add and multiply/subtract
28909 instructions in the floating-point option. This has no effect if the
28910 floating-point option is not also enabled. Disabling fused multiply/add
28911 and multiply/subtract instructions forces the compiler to use separate
28912 instructions for the multiply and add/subtract operations. This may be
28913 desirable in some cases where strict IEEE 754-compliant results are
28914 required: the fused multiply add/subtract instructions do not round the
28915 intermediate result, thereby producing results with @emph{more} bits of
28916 precision than specified by the IEEE standard. Disabling fused multiply
28917 add/subtract instructions also ensures that the program output is not
28918 sensitive to the compiler's ability to combine multiply and add/subtract
28921 @item -mserialize-volatile
28922 @itemx -mno-serialize-volatile
28923 @opindex mserialize-volatile
28924 @opindex mno-serialize-volatile
28925 When this option is enabled, GCC inserts @code{MEMW} instructions before
28926 @code{volatile} memory references to guarantee sequential consistency.
28927 The default is @option{-mserialize-volatile}. Use
28928 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
28930 @item -mforce-no-pic
28931 @opindex mforce-no-pic
28932 For targets, like GNU/Linux, where all user-mode Xtensa code must be
28933 position-independent code (PIC), this option disables PIC for compiling
28936 @item -mtext-section-literals
28937 @itemx -mno-text-section-literals
28938 @opindex mtext-section-literals
28939 @opindex mno-text-section-literals
28940 These options control the treatment of literal pools. The default is
28941 @option{-mno-text-section-literals}, which places literals in a separate
28942 section in the output file. This allows the literal pool to be placed
28943 in a data RAM/ROM, and it also allows the linker to combine literal
28944 pools from separate object files to remove redundant literals and
28945 improve code size. With @option{-mtext-section-literals}, the literals
28946 are interspersed in the text section in order to keep them as close as
28947 possible to their references. This may be necessary for large assembly
28948 files. Literals for each function are placed right before that function.
28950 @item -mauto-litpools
28951 @itemx -mno-auto-litpools
28952 @opindex mauto-litpools
28953 @opindex mno-auto-litpools
28954 These options control the treatment of literal pools. The default is
28955 @option{-mno-auto-litpools}, which places literals in a separate
28956 section in the output file unless @option{-mtext-section-literals} is
28957 used. With @option{-mauto-litpools} the literals are interspersed in
28958 the text section by the assembler. Compiler does not produce explicit
28959 @code{.literal} directives and loads literals into registers with
28960 @code{MOVI} instructions instead of @code{L32R} to let the assembler
28961 do relaxation and place literals as necessary. This option allows
28962 assembler to create several literal pools per function and assemble
28963 very big functions, which may not be possible with
28964 @option{-mtext-section-literals}.
28966 @item -mtarget-align
28967 @itemx -mno-target-align
28968 @opindex mtarget-align
28969 @opindex mno-target-align
28970 When this option is enabled, GCC instructs the assembler to
28971 automatically align instructions to reduce branch penalties at the
28972 expense of some code density. The assembler attempts to widen density
28973 instructions to align branch targets and the instructions following call
28974 instructions. If there are not enough preceding safe density
28975 instructions to align a target, no widening is performed. The
28976 default is @option{-mtarget-align}. These options do not affect the
28977 treatment of auto-aligned instructions like @code{LOOP}, which the
28978 assembler always aligns, either by widening density instructions or
28979 by inserting NOP instructions.
28982 @itemx -mno-longcalls
28983 @opindex mlongcalls
28984 @opindex mno-longcalls
28985 When this option is enabled, GCC instructs the assembler to translate
28986 direct calls to indirect calls unless it can determine that the target
28987 of a direct call is in the range allowed by the call instruction. This
28988 translation typically occurs for calls to functions in other source
28989 files. Specifically, the assembler translates a direct @code{CALL}
28990 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
28991 The default is @option{-mno-longcalls}. This option should be used in
28992 programs where the call target can potentially be out of range. This
28993 option is implemented in the assembler, not the compiler, so the
28994 assembly code generated by GCC still shows direct call
28995 instructions---look at the disassembled object code to see the actual
28996 instructions. Note that the assembler uses an indirect call for
28997 every cross-file call, not just those that really are out of range.
29000 @node zSeries Options
29001 @subsection zSeries Options
29002 @cindex zSeries options
29004 These are listed under @xref{S/390 and zSeries Options}.
29010 @section Specifying Subprocesses and the Switches to Pass to Them
29013 @command{gcc} is a driver program. It performs its job by invoking a
29014 sequence of other programs to do the work of compiling, assembling and
29015 linking. GCC interprets its command-line parameters and uses these to
29016 deduce which programs it should invoke, and which command-line options
29017 it ought to place on their command lines. This behavior is controlled
29018 by @dfn{spec strings}. In most cases there is one spec string for each
29019 program that GCC can invoke, but a few programs have multiple spec
29020 strings to control their behavior. The spec strings built into GCC can
29021 be overridden by using the @option{-specs=} command-line switch to specify
29024 @dfn{Spec files} are plain-text files that are used to construct spec
29025 strings. They consist of a sequence of directives separated by blank
29026 lines. The type of directive is determined by the first non-whitespace
29027 character on the line, which can be one of the following:
29030 @item %@var{command}
29031 Issues a @var{command} to the spec file processor. The commands that can
29035 @item %include <@var{file}>
29036 @cindex @code{%include}
29037 Search for @var{file} and insert its text at the current point in the
29040 @item %include_noerr <@var{file}>
29041 @cindex @code{%include_noerr}
29042 Just like @samp{%include}, but do not generate an error message if the include
29043 file cannot be found.
29045 @item %rename @var{old_name} @var{new_name}
29046 @cindex @code{%rename}
29047 Rename the spec string @var{old_name} to @var{new_name}.
29051 @item *[@var{spec_name}]:
29052 This tells the compiler to create, override or delete the named spec
29053 string. All lines after this directive up to the next directive or
29054 blank line are considered to be the text for the spec string. If this
29055 results in an empty string then the spec is deleted. (Or, if the
29056 spec did not exist, then nothing happens.) Otherwise, if the spec
29057 does not currently exist a new spec is created. If the spec does
29058 exist then its contents are overridden by the text of this
29059 directive, unless the first character of that text is the @samp{+}
29060 character, in which case the text is appended to the spec.
29062 @item [@var{suffix}]:
29063 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
29064 and up to the next directive or blank line are considered to make up the
29065 spec string for the indicated suffix. When the compiler encounters an
29066 input file with the named suffix, it processes the spec string in
29067 order to work out how to compile that file. For example:
29071 z-compile -input %i
29074 This says that any input file whose name ends in @samp{.ZZ} should be
29075 passed to the program @samp{z-compile}, which should be invoked with the
29076 command-line switch @option{-input} and with the result of performing the
29077 @samp{%i} substitution. (See below.)
29079 As an alternative to providing a spec string, the text following a
29080 suffix directive can be one of the following:
29083 @item @@@var{language}
29084 This says that the suffix is an alias for a known @var{language}. This is
29085 similar to using the @option{-x} command-line switch to GCC to specify a
29086 language explicitly. For example:
29093 Says that .ZZ files are, in fact, C++ source files.
29096 This causes an error messages saying:
29099 @var{name} compiler not installed on this system.
29103 GCC already has an extensive list of suffixes built into it.
29104 This directive adds an entry to the end of the list of suffixes, but
29105 since the list is searched from the end backwards, it is effectively
29106 possible to override earlier entries using this technique.
29110 GCC has the following spec strings built into it. Spec files can
29111 override these strings or create their own. Note that individual
29112 targets can also add their own spec strings to this list.
29115 asm Options to pass to the assembler
29116 asm_final Options to pass to the assembler post-processor
29117 cpp Options to pass to the C preprocessor
29118 cc1 Options to pass to the C compiler
29119 cc1plus Options to pass to the C++ compiler
29120 endfile Object files to include at the end of the link
29121 link Options to pass to the linker
29122 lib Libraries to include on the command line to the linker
29123 libgcc Decides which GCC support library to pass to the linker
29124 linker Sets the name of the linker
29125 predefines Defines to be passed to the C preprocessor
29126 signed_char Defines to pass to CPP to say whether @code{char} is signed
29128 startfile Object files to include at the start of the link
29131 Here is a small example of a spec file:
29134 %rename lib old_lib
29137 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
29140 This example renames the spec called @samp{lib} to @samp{old_lib} and
29141 then overrides the previous definition of @samp{lib} with a new one.
29142 The new definition adds in some extra command-line options before
29143 including the text of the old definition.
29145 @dfn{Spec strings} are a list of command-line options to be passed to their
29146 corresponding program. In addition, the spec strings can contain
29147 @samp{%}-prefixed sequences to substitute variable text or to
29148 conditionally insert text into the command line. Using these constructs
29149 it is possible to generate quite complex command lines.
29151 Here is a table of all defined @samp{%}-sequences for spec
29152 strings. Note that spaces are not generated automatically around the
29153 results of expanding these sequences. Therefore you can concatenate them
29154 together or combine them with constant text in a single argument.
29158 Substitute one @samp{%} into the program name or argument.
29161 Substitute the name of the input file being processed.
29164 Substitute the basename of the input file being processed.
29165 This is the substring up to (and not including) the last period
29166 and not including the directory.
29169 This is the same as @samp{%b}, but include the file suffix (text after
29173 Marks the argument containing or following the @samp{%d} as a
29174 temporary file name, so that that file is deleted if GCC exits
29175 successfully. Unlike @samp{%g}, this contributes no text to the
29178 @item %g@var{suffix}
29179 Substitute a file name that has suffix @var{suffix} and is chosen
29180 once per compilation, and mark the argument in the same way as
29181 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
29182 name is now chosen in a way that is hard to predict even when previously
29183 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
29184 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
29185 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
29186 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
29187 was simply substituted with a file name chosen once per compilation,
29188 without regard to any appended suffix (which was therefore treated
29189 just like ordinary text), making such attacks more likely to succeed.
29191 @item %u@var{suffix}
29192 Like @samp{%g}, but generates a new temporary file name
29193 each time it appears instead of once per compilation.
29195 @item %U@var{suffix}
29196 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
29197 new one if there is no such last file name. In the absence of any
29198 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
29199 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
29200 involves the generation of two distinct file names, one
29201 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
29202 simply substituted with a file name chosen for the previous @samp{%u},
29203 without regard to any appended suffix.
29205 @item %j@var{suffix}
29206 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
29207 writable, and if @option{-save-temps} is not used;
29208 otherwise, substitute the name
29209 of a temporary file, just like @samp{%u}. This temporary file is not
29210 meant for communication between processes, but rather as a junk
29211 disposal mechanism.
29213 @item %|@var{suffix}
29214 @itemx %m@var{suffix}
29215 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
29216 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
29217 all. These are the two most common ways to instruct a program that it
29218 should read from standard input or write to standard output. If you
29219 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
29220 construct: see for example @file{f/lang-specs.h}.
29222 @item %.@var{SUFFIX}
29223 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
29224 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
29225 terminated by the next space or %.
29228 Marks the argument containing or following the @samp{%w} as the
29229 designated output file of this compilation. This puts the argument
29230 into the sequence of arguments that @samp{%o} substitutes.
29233 Substitutes the names of all the output files, with spaces
29234 automatically placed around them. You should write spaces
29235 around the @samp{%o} as well or the results are undefined.
29236 @samp{%o} is for use in the specs for running the linker.
29237 Input files whose names have no recognized suffix are not compiled
29238 at all, but they are included among the output files, so they are
29242 Substitutes the suffix for object files. Note that this is
29243 handled specially when it immediately follows @samp{%g, %u, or %U},
29244 because of the need for those to form complete file names. The
29245 handling is such that @samp{%O} is treated exactly as if it had already
29246 been substituted, except that @samp{%g, %u, and %U} do not currently
29247 support additional @var{suffix} characters following @samp{%O} as they do
29248 following, for example, @samp{.o}.
29251 Substitutes the standard macro predefinitions for the
29252 current target machine. Use this when running @command{cpp}.
29255 Like @samp{%p}, but puts @samp{__} before and after the name of each
29256 predefined macro, except for macros that start with @samp{__} or with
29257 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
29261 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
29262 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
29263 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
29264 and @option{-imultilib} as necessary.
29267 Current argument is the name of a library or startup file of some sort.
29268 Search for that file in a standard list of directories and substitute
29269 the full name found. The current working directory is included in the
29270 list of directories scanned.
29273 Current argument is the name of a linker script. Search for that file
29274 in the current list of directories to scan for libraries. If the file
29275 is located insert a @option{--script} option into the command line
29276 followed by the full path name found. If the file is not found then
29277 generate an error message. Note: the current working directory is not
29281 Print @var{str} as an error message. @var{str} is terminated by a newline.
29282 Use this when inconsistent options are detected.
29284 @item %(@var{name})
29285 Substitute the contents of spec string @var{name} at this point.
29287 @item %x@{@var{option}@}
29288 Accumulate an option for @samp{%X}.
29291 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
29295 Output the accumulated assembler options specified by @option{-Wa}.
29298 Output the accumulated preprocessor options specified by @option{-Wp}.
29301 Process the @code{asm} spec. This is used to compute the
29302 switches to be passed to the assembler.
29305 Process the @code{asm_final} spec. This is a spec string for
29306 passing switches to an assembler post-processor, if such a program is
29310 Process the @code{link} spec. This is the spec for computing the
29311 command line passed to the linker. Typically it makes use of the
29312 @samp{%L %G %S %D and %E} sequences.
29315 Dump out a @option{-L} option for each directory that GCC believes might
29316 contain startup files. If the target supports multilibs then the
29317 current multilib directory is prepended to each of these paths.
29320 Process the @code{lib} spec. This is a spec string for deciding which
29321 libraries are included on the command line to the linker.
29324 Process the @code{libgcc} spec. This is a spec string for deciding
29325 which GCC support library is included on the command line to the linker.
29328 Process the @code{startfile} spec. This is a spec for deciding which
29329 object files are the first ones passed to the linker. Typically
29330 this might be a file named @file{crt0.o}.
29333 Process the @code{endfile} spec. This is a spec string that specifies
29334 the last object files that are passed to the linker.
29337 Process the @code{cpp} spec. This is used to construct the arguments
29338 to be passed to the C preprocessor.
29341 Process the @code{cc1} spec. This is used to construct the options to be
29342 passed to the actual C compiler (@command{cc1}).
29345 Process the @code{cc1plus} spec. This is used to construct the options to be
29346 passed to the actual C++ compiler (@command{cc1plus}).
29349 Substitute the variable part of a matched option. See below.
29350 Note that each comma in the substituted string is replaced by
29354 Remove all occurrences of @code{-S} from the command line. Note---this
29355 command is position dependent. @samp{%} commands in the spec string
29356 before this one see @code{-S}, @samp{%} commands in the spec string
29357 after this one do not.
29359 @item %:@var{function}(@var{args})
29360 Call the named function @var{function}, passing it @var{args}.
29361 @var{args} is first processed as a nested spec string, then split
29362 into an argument vector in the usual fashion. The function returns
29363 a string which is processed as if it had appeared literally as part
29364 of the current spec.
29366 The following built-in spec functions are provided:
29369 @item @code{getenv}
29370 The @code{getenv} spec function takes two arguments: an environment
29371 variable name and a string. If the environment variable is not
29372 defined, a fatal error is issued. Otherwise, the return value is the
29373 value of the environment variable concatenated with the string. For
29374 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
29377 %:getenv(TOPDIR /include)
29380 expands to @file{/path/to/top/include}.
29382 @item @code{if-exists}
29383 The @code{if-exists} spec function takes one argument, an absolute
29384 pathname to a file. If the file exists, @code{if-exists} returns the
29385 pathname. Here is a small example of its usage:
29389 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
29392 @item @code{if-exists-else}
29393 The @code{if-exists-else} spec function is similar to the @code{if-exists}
29394 spec function, except that it takes two arguments. The first argument is
29395 an absolute pathname to a file. If the file exists, @code{if-exists-else}
29396 returns the pathname. If it does not exist, it returns the second argument.
29397 This way, @code{if-exists-else} can be used to select one file or another,
29398 based on the existence of the first. Here is a small example of its usage:
29402 crt0%O%s %:if-exists(crti%O%s) \
29403 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
29406 @item @code{replace-outfile}
29407 The @code{replace-outfile} spec function takes two arguments. It looks for the
29408 first argument in the outfiles array and replaces it with the second argument. Here
29409 is a small example of its usage:
29412 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
29415 @item @code{remove-outfile}
29416 The @code{remove-outfile} spec function takes one argument. It looks for the
29417 first argument in the outfiles array and removes it. Here is a small example
29421 %:remove-outfile(-lm)
29424 @item @code{pass-through-libs}
29425 The @code{pass-through-libs} spec function takes any number of arguments. It
29426 finds any @option{-l} options and any non-options ending in @file{.a} (which it
29427 assumes are the names of linker input library archive files) and returns a
29428 result containing all the found arguments each prepended by
29429 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
29430 intended to be passed to the LTO linker plugin.
29433 %:pass-through-libs(%G %L %G)
29436 @item @code{print-asm-header}
29437 The @code{print-asm-header} function takes no arguments and simply
29438 prints a banner like:
29444 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
29447 It is used to separate compiler options from assembler options
29448 in the @option{--target-help} output.
29452 Substitutes the @code{-S} switch, if that switch is given to GCC@.
29453 If that switch is not specified, this substitutes nothing. Note that
29454 the leading dash is omitted when specifying this option, and it is
29455 automatically inserted if the substitution is performed. Thus the spec
29456 string @samp{%@{foo@}} matches the command-line option @option{-foo}
29457 and outputs the command-line option @option{-foo}.
29460 Like %@{@code{S}@} but mark last argument supplied within as a file to be
29461 deleted on failure.
29464 Substitutes all the switches specified to GCC whose names start
29465 with @code{-S}, but which also take an argument. This is used for
29466 switches like @option{-o}, @option{-D}, @option{-I}, etc.
29467 GCC considers @option{-o foo} as being
29468 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
29469 text, including the space. Thus two arguments are generated.
29472 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
29473 (the order of @code{S} and @code{T} in the spec is not significant).
29474 There can be any number of ampersand-separated variables; for each the
29475 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
29478 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
29481 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
29484 Substitutes @code{X} if one or more switches whose names start with
29485 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
29486 once, no matter how many such switches appeared. However, if @code{%*}
29487 appears somewhere in @code{X}, then @code{X} is substituted once
29488 for each matching switch, with the @code{%*} replaced by the part of
29489 that switch matching the @code{*}.
29491 If @code{%*} appears as the last part of a spec sequence then a space
29492 is added after the end of the last substitution. If there is more
29493 text in the sequence, however, then a space is not generated. This
29494 allows the @code{%*} substitution to be used as part of a larger
29495 string. For example, a spec string like this:
29498 %@{mcu=*:--script=%*/memory.ld@}
29502 when matching an option like @option{-mcu=newchip} produces:
29505 --script=newchip/memory.ld
29509 Substitutes @code{X}, if processing a file with suffix @code{S}.
29512 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
29515 Substitutes @code{X}, if processing a file for language @code{S}.
29518 Substitutes @code{X}, if not processing a file for language @code{S}.
29521 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
29522 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
29523 @code{*} sequences as well, although they have a stronger binding than
29524 the @samp{|}. If @code{%*} appears in @code{X}, all of the
29525 alternatives must be starred, and only the first matching alternative
29528 For example, a spec string like this:
29531 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
29535 outputs the following command-line options from the following input
29536 command-line options:
29541 -d fred.c -foo -baz -boggle
29542 -d jim.d -bar -baz -boggle
29545 @item %@{S:X; T:Y; :D@}
29547 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
29548 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
29549 be as many clauses as you need. This may be combined with @code{.},
29550 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
29555 The switch matching text @code{S} in a @samp{%@{S@}}, @samp{%@{S:X@}}
29556 or similar construct can use a backslash to ignore the special meaning
29557 of the character following it, thus allowing literal matching of a
29558 character that is otherwise specially treated. For example,
29559 @samp{%@{std=iso9899\:1999:X@}} substitutes @code{X} if the
29560 @option{-std=iso9899:1999} option is given.
29562 The conditional text @code{X} in a @samp{%@{S:X@}} or similar
29563 construct may contain other nested @samp{%} constructs or spaces, or
29564 even newlines. They are processed as usual, as described above.
29565 Trailing white space in @code{X} is ignored. White space may also
29566 appear anywhere on the left side of the colon in these constructs,
29567 except between @code{.} or @code{*} and the corresponding word.
29569 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
29570 handled specifically in these constructs. If another value of
29571 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
29572 @option{-W} switch is found later in the command line, the earlier
29573 switch value is ignored, except with @{@code{S}*@} where @code{S} is
29574 just one letter, which passes all matching options.
29576 The character @samp{|} at the beginning of the predicate text is used to
29577 indicate that a command should be piped to the following command, but
29578 only if @option{-pipe} is specified.
29580 It is built into GCC which switches take arguments and which do not.
29581 (You might think it would be useful to generalize this to allow each
29582 compiler's spec to say which switches take arguments. But this cannot
29583 be done in a consistent fashion. GCC cannot even decide which input
29584 files have been specified without knowing which switches take arguments,
29585 and it must know which input files to compile in order to tell which
29588 GCC also knows implicitly that arguments starting in @option{-l} are to be
29589 treated as compiler output files, and passed to the linker in their
29590 proper position among the other output files.
29592 @node Environment Variables
29593 @section Environment Variables Affecting GCC
29594 @cindex environment variables
29596 @c man begin ENVIRONMENT
29597 This section describes several environment variables that affect how GCC
29598 operates. Some of them work by specifying directories or prefixes to use
29599 when searching for various kinds of files. Some are used to specify other
29600 aspects of the compilation environment.
29602 Note that you can also specify places to search using options such as
29603 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
29604 take precedence over places specified using environment variables, which
29605 in turn take precedence over those specified by the configuration of GCC@.
29606 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
29607 GNU Compiler Collection (GCC) Internals}.
29612 @c @itemx LC_COLLATE
29614 @c @itemx LC_MONETARY
29615 @c @itemx LC_NUMERIC
29620 @c @findex LC_COLLATE
29621 @findex LC_MESSAGES
29622 @c @findex LC_MONETARY
29623 @c @findex LC_NUMERIC
29627 These environment variables control the way that GCC uses
29628 localization information which allows GCC to work with different
29629 national conventions. GCC inspects the locale categories
29630 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
29631 so. These locale categories can be set to any value supported by your
29632 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
29633 Kingdom encoded in UTF-8.
29635 The @env{LC_CTYPE} environment variable specifies character
29636 classification. GCC uses it to determine the character boundaries in
29637 a string; this is needed for some multibyte encodings that contain quote
29638 and escape characters that are otherwise interpreted as a string
29641 The @env{LC_MESSAGES} environment variable specifies the language to
29642 use in diagnostic messages.
29644 If the @env{LC_ALL} environment variable is set, it overrides the value
29645 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
29646 and @env{LC_MESSAGES} default to the value of the @env{LANG}
29647 environment variable. If none of these variables are set, GCC
29648 defaults to traditional C English behavior.
29652 If @env{TMPDIR} is set, it specifies the directory to use for temporary
29653 files. GCC uses temporary files to hold the output of one stage of
29654 compilation which is to be used as input to the next stage: for example,
29655 the output of the preprocessor, which is the input to the compiler
29658 @item GCC_COMPARE_DEBUG
29659 @findex GCC_COMPARE_DEBUG
29660 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
29661 @option{-fcompare-debug} to the compiler driver. See the documentation
29662 of this option for more details.
29664 @item GCC_EXEC_PREFIX
29665 @findex GCC_EXEC_PREFIX
29666 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
29667 names of the subprograms executed by the compiler. No slash is added
29668 when this prefix is combined with the name of a subprogram, but you can
29669 specify a prefix that ends with a slash if you wish.
29671 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
29672 an appropriate prefix to use based on the pathname it is invoked with.
29674 If GCC cannot find the subprogram using the specified prefix, it
29675 tries looking in the usual places for the subprogram.
29677 The default value of @env{GCC_EXEC_PREFIX} is
29678 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
29679 the installed compiler. In many cases @var{prefix} is the value
29680 of @code{prefix} when you ran the @file{configure} script.
29682 Other prefixes specified with @option{-B} take precedence over this prefix.
29684 This prefix is also used for finding files such as @file{crt0.o} that are
29687 In addition, the prefix is used in an unusual way in finding the
29688 directories to search for header files. For each of the standard
29689 directories whose name normally begins with @samp{/usr/local/lib/gcc}
29690 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
29691 replacing that beginning with the specified prefix to produce an
29692 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
29693 @file{foo/bar} just before it searches the standard directory
29694 @file{/usr/local/lib/bar}.
29695 If a standard directory begins with the configured
29696 @var{prefix} then the value of @var{prefix} is replaced by
29697 @env{GCC_EXEC_PREFIX} when looking for header files.
29699 @item COMPILER_PATH
29700 @findex COMPILER_PATH
29701 The value of @env{COMPILER_PATH} is a colon-separated list of
29702 directories, much like @env{PATH}. GCC tries the directories thus
29703 specified when searching for subprograms, if it cannot find the
29704 subprograms using @env{GCC_EXEC_PREFIX}.
29707 @findex LIBRARY_PATH
29708 The value of @env{LIBRARY_PATH} is a colon-separated list of
29709 directories, much like @env{PATH}. When configured as a native compiler,
29710 GCC tries the directories thus specified when searching for special
29711 linker files, if it cannot find them using @env{GCC_EXEC_PREFIX}. Linking
29712 using GCC also uses these directories when searching for ordinary
29713 libraries for the @option{-l} option (but directories specified with
29714 @option{-L} come first).
29718 @cindex locale definition
29719 This variable is used to pass locale information to the compiler. One way in
29720 which this information is used is to determine the character set to be used
29721 when character literals, string literals and comments are parsed in C and C++.
29722 When the compiler is configured to allow multibyte characters,
29723 the following values for @env{LANG} are recognized:
29727 Recognize JIS characters.
29729 Recognize SJIS characters.
29731 Recognize EUCJP characters.
29734 If @env{LANG} is not defined, or if it has some other value, then the
29735 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
29736 recognize and translate multibyte characters.
29740 Some additional environment variables affect the behavior of the
29743 @include cppenv.texi
29747 @node Precompiled Headers
29748 @section Using Precompiled Headers
29749 @cindex precompiled headers
29750 @cindex speed of compilation
29752 Often large projects have many header files that are included in every
29753 source file. The time the compiler takes to process these header files
29754 over and over again can account for nearly all of the time required to
29755 build the project. To make builds faster, GCC allows you to
29756 @dfn{precompile} a header file.
29758 To create a precompiled header file, simply compile it as you would any
29759 other file, if necessary using the @option{-x} option to make the driver
29760 treat it as a C or C++ header file. You may want to use a
29761 tool like @command{make} to keep the precompiled header up-to-date when
29762 the headers it contains change.
29764 A precompiled header file is searched for when @code{#include} is
29765 seen in the compilation. As it searches for the included file
29766 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
29767 compiler looks for a precompiled header in each directory just before it
29768 looks for the include file in that directory. The name searched for is
29769 the name specified in the @code{#include} with @samp{.gch} appended. If
29770 the precompiled header file cannot be used, it is ignored.
29772 For instance, if you have @code{#include "all.h"}, and you have
29773 @file{all.h.gch} in the same directory as @file{all.h}, then the
29774 precompiled header file is used if possible, and the original
29775 header is used otherwise.
29777 Alternatively, you might decide to put the precompiled header file in a
29778 directory and use @option{-I} to ensure that directory is searched
29779 before (or instead of) the directory containing the original header.
29780 Then, if you want to check that the precompiled header file is always
29781 used, you can put a file of the same name as the original header in this
29782 directory containing an @code{#error} command.
29784 This also works with @option{-include}. So yet another way to use
29785 precompiled headers, good for projects not designed with precompiled
29786 header files in mind, is to simply take most of the header files used by
29787 a project, include them from another header file, precompile that header
29788 file, and @option{-include} the precompiled header. If the header files
29789 have guards against multiple inclusion, they are skipped because
29790 they've already been included (in the precompiled header).
29792 If you need to precompile the same header file for different
29793 languages, targets, or compiler options, you can instead make a
29794 @emph{directory} named like @file{all.h.gch}, and put each precompiled
29795 header in the directory, perhaps using @option{-o}. It doesn't matter
29796 what you call the files in the directory; every precompiled header in
29797 the directory is considered. The first precompiled header
29798 encountered in the directory that is valid for this compilation is
29799 used; they're searched in no particular order.
29801 There are many other possibilities, limited only by your imagination,
29802 good sense, and the constraints of your build system.
29804 A precompiled header file can be used only when these conditions apply:
29808 Only one precompiled header can be used in a particular compilation.
29811 A precompiled header cannot be used once the first C token is seen. You
29812 can have preprocessor directives before a precompiled header; you cannot
29813 include a precompiled header from inside another header.
29816 The precompiled header file must be produced for the same language as
29817 the current compilation. You cannot use a C precompiled header for a C++
29821 The precompiled header file must have been produced by the same compiler
29822 binary as the current compilation is using.
29825 Any macros defined before the precompiled header is included must
29826 either be defined in the same way as when the precompiled header was
29827 generated, or must not affect the precompiled header, which usually
29828 means that they don't appear in the precompiled header at all.
29830 The @option{-D} option is one way to define a macro before a
29831 precompiled header is included; using a @code{#define} can also do it.
29832 There are also some options that define macros implicitly, like
29833 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
29836 @item If debugging information is output when using the precompiled
29837 header, using @option{-g} or similar, the same kind of debugging information
29838 must have been output when building the precompiled header. However,
29839 a precompiled header built using @option{-g} can be used in a compilation
29840 when no debugging information is being output.
29842 @item The same @option{-m} options must generally be used when building
29843 and using the precompiled header. @xref{Submodel Options},
29844 for any cases where this rule is relaxed.
29846 @item Each of the following options must be the same when building and using
29847 the precompiled header:
29849 @gccoptlist{-fexceptions}
29852 Some other command-line options starting with @option{-f},
29853 @option{-p}, or @option{-O} must be defined in the same way as when
29854 the precompiled header was generated. At present, it's not clear
29855 which options are safe to change and which are not; the safest choice
29856 is to use exactly the same options when generating and using the
29857 precompiled header. The following are known to be safe:
29859 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
29860 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
29861 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
29866 For all of these except the last, the compiler automatically
29867 ignores the precompiled header if the conditions aren't met. If you
29868 find an option combination that doesn't work and doesn't cause the
29869 precompiled header to be ignored, please consider filing a bug report,
29872 If you do use differing options when generating and using the
29873 precompiled header, the actual behavior is a mixture of the
29874 behavior for the options. For instance, if you use @option{-g} to
29875 generate the precompiled header but not when using it, you may or may
29876 not get debugging information for routines in the precompiled header.