1 @c Copyright (C) 1988-2016 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{} 1988-2016 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), adb(1), dbx(1), sdb(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.
126 @xref{Option Index}, for an index to GCC's options.
129 * Option Summary:: Brief list of all options, without explanations.
130 * Overall Options:: Controlling the kind of output:
131 an executable, object files, assembler files,
132 or preprocessed source.
133 * Invoking G++:: Compiling C++ programs.
134 * C Dialect Options:: Controlling the variant of C language compiled.
135 * C++ Dialect Options:: Variations on C++.
136 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
138 * Diagnostic Message Formatting Options:: Controlling how diagnostics should
140 * Warning Options:: How picky should the compiler be?
141 * Debugging Options:: Producing debuggable code.
142 * Optimize Options:: How much optimization?
143 * Instrumentation Options:: Enabling profiling and extra run-time error checking.
144 * Preprocessor Options:: Controlling header files and macro definitions.
145 Also, getting dependency information for Make.
146 * Assembler Options:: Passing options to the assembler.
147 * Link Options:: Specifying libraries and so on.
148 * Directory Options:: Where to find header files and libraries.
149 Where to find the compiler executable files.
150 * Code Gen Options:: Specifying conventions for function calls, data layout
152 * Developer Options:: Printing GCC configuration info, statistics, and
154 * Submodel Options:: Target-specific options, such as compiling for a
155 specific processor variant.
156 * Spec Files:: How to pass switches to sub-processes.
157 * Environment Variables:: Env vars that affect GCC.
158 * Precompiled Headers:: Compiling a header once, and using it many times.
164 @section Option Summary
166 Here is a summary of all the options, grouped by type. Explanations are
167 in the following sections.
170 @item Overall Options
171 @xref{Overall Options,,Options Controlling the Kind of Output}.
172 @gccoptlist{-c -S -E -o @var{file} -x @var{language} @gol
173 -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help --version @gol
174 -pass-exit-codes -pipe -specs=@var{file} -wrapper @gol
175 @@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
176 -fdump-ada-spec@r{[}-slim@r{]} -fada-spec-parent=@var{unit} -fdump-go-spec=@var{file}}
178 @item C Language Options
179 @xref{C Dialect Options,,Options Controlling C Dialect}.
180 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
181 -fpermitted-flt-eval-methods=@var{standard} @gol
182 -aux-info @var{filename} -fallow-parameterless-variadic-functions @gol
183 -fno-asm -fno-builtin -fno-builtin-@var{function} -fgimple@gol
184 -fhosted -ffreestanding -fopenacc -fopenmp -fopenmp-simd @gol
185 -fms-extensions -fplan9-extensions -fsso-struct=@var{endianness}
186 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
187 -fsigned-bitfields -fsigned-char @gol
188 -funsigned-bitfields -funsigned-char @gol
189 -trigraphs -traditional -traditional-cpp}
191 @item C++ Language Options
192 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
193 @gccoptlist{-fabi-version=@var{n} -fno-access-control @gol
194 -faligned-new=@var{n} -fargs-in-order=@var{n} -fcheck-new @gol
195 -fconstexpr-depth=@var{n} -fconstexpr-loop-limit=@var{n} @gol
196 -ffriend-injection @gol
197 -fno-elide-constructors @gol
198 -fno-enforce-eh-specs @gol
199 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
200 -fno-implicit-templates @gol
201 -fno-implicit-inline-templates @gol
202 -fno-implement-inlines -fms-extensions @gol
203 -fnew-inheriting-ctors @gol
204 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
205 -fno-optional-diags -fpermissive @gol
206 -fno-pretty-templates @gol
207 -frepo -fno-rtti -fsized-deallocation @gol
208 -ftemplate-backtrace-limit=@var{n} @gol
209 -ftemplate-depth=@var{n} @gol
210 -fno-threadsafe-statics -fuse-cxa-atexit @gol
211 -fno-weak -nostdinc++ @gol
212 -fvisibility-inlines-hidden @gol
213 -fvisibility-ms-compat @gol
214 -fext-numeric-literals @gol
215 -Wabi=@var{n} -Wabi-tag -Wconversion-null -Wctor-dtor-privacy @gol
216 -Wdelete-non-virtual-dtor -Wliteral-suffix -Wmultiple-inheritance @gol
217 -Wnamespaces -Wnarrowing @gol
218 -Wnoexcept -Wnon-virtual-dtor -Wreorder -Wregister @gol
219 -Weffc++ -Wstrict-null-sentinel -Wtemplates @gol
220 -Wno-non-template-friend -Wold-style-cast @gol
221 -Woverloaded-virtual -Wno-pmf-conversions @gol
222 -Wsign-promo -Wvirtual-inheritance}
224 @item Objective-C and Objective-C++ Language Options
225 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
226 Objective-C and Objective-C++ Dialects}.
227 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
228 -fgnu-runtime -fnext-runtime @gol
229 -fno-nil-receivers @gol
230 -fobjc-abi-version=@var{n} @gol
231 -fobjc-call-cxx-cdtors @gol
232 -fobjc-direct-dispatch @gol
233 -fobjc-exceptions @gol
236 -fobjc-std=objc1 @gol
237 -fno-local-ivars @gol
238 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
239 -freplace-objc-classes @gol
242 -Wassign-intercept @gol
243 -Wno-protocol -Wselector @gol
244 -Wstrict-selector-match @gol
245 -Wundeclared-selector}
247 @item Diagnostic Message Formatting Options
248 @xref{Diagnostic Message Formatting Options,,Options to Control Diagnostic Messages Formatting}.
249 @gccoptlist{-fmessage-length=@var{n} @gol
250 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
251 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
252 -fno-diagnostics-show-option -fno-diagnostics-show-caret @gol
253 -fdiagnostics-parseable-fixits -fdiagnostics-generate-patch}
255 @item Warning Options
256 @xref{Warning Options,,Options to Request or Suppress Warnings}.
257 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
258 -pedantic-errors @gol
259 -w -Wextra -Wall -Waddress -Waggregate-return @gol
260 -Walloc-zero -Walloc-size-larger-than=@var{n}
261 -Walloca -Walloca-larger-than=@var{n} @gol
262 -Wno-aggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
263 -Wno-attributes -Wbool-compare -Wbool-operation @gol
264 -Wno-builtin-declaration-mismatch @gol
265 -Wno-builtin-macro-redefined -Wc90-c99-compat -Wc99-c11-compat @gol
266 -Wc++-compat -Wc++11-compat -Wc++14-compat -Wcast-align -Wcast-qual @gol
267 -Wchar-subscripts -Wclobbered -Wcomment -Wconditionally-supported @gol
268 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wdangling-else -Wdate-time @gol
269 -Wdelete-incomplete @gol
270 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
271 -Wdisabled-optimization @gol
272 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
273 -Wno-div-by-zero -Wdouble-promotion -Wduplicated-cond @gol
274 -Wempty-body -Wenum-compare -Wno-endif-labels @gol
275 -Werror -Werror=* -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
276 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-length=@var{n} @gol
277 -Wformat-nonliteral @gol
278 -Wformat-security -Wformat-signedness -Wformat-y2k -Wframe-address @gol
279 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
280 -Wignored-qualifiers -Wignored-attributes -Wincompatible-pointer-types @gol
281 -Wimplicit -Wimplicit-fallthrough -Wimplicit-fallthrough=@var{n} @gol
282 -Wimplicit-function-declaration -Wimplicit-int @gol
283 -Winit-self -Winline -Wno-int-conversion -Wint-in-bool-context @gol
284 -Wno-int-to-pointer-cast -Winvalid-memory-model -Wno-invalid-offsetof @gol
285 -Winvalid-pch -Wlarger-than=@var{len} @gol
286 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
287 -Wmain -Wmaybe-uninitialized -Wmemset-elt-size -Wmemset-transposed-args @gol
288 -Wmisleading-indentation -Wmissing-braces @gol
289 -Wmissing-field-initializers -Wmissing-include-dirs @gol
290 -Wno-multichar -Wnonnull -Wnonnull-compare @gol
291 -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
292 -Wnull-dereference -Wodr -Wno-overflow -Wopenmp-simd @gol
293 -Woverride-init-side-effects -Woverlength-strings @gol
294 -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
295 -Wparentheses -Wno-pedantic-ms-format @gol
296 -Wplacement-new -Wplacement-new=@var{n} @gol
297 -Wpointer-arith -Wno-pointer-to-int-cast @gol
298 -Wno-pragmas -Wredundant-decls -Wrestrict -Wno-return-local-addr @gol
299 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
300 -Wshadow=global, -Wshadow=local, -Wshadow=compatible-local @gol
301 -Wshift-overflow -Wshift-overflow=@var{n} @gol
302 -Wshift-count-negative -Wshift-count-overflow -Wshift-negative-value @gol
303 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
304 -Wno-scalar-storage-order @gol
305 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
306 -Wstack-protector -Wstack-usage=@var{len} -Wstrict-aliasing @gol
307 -Wstrict-aliasing=n -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
308 -Wstringop-overflow=@var{n} @gol
309 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]} @gol
310 -Wsuggest-final-types @gol -Wsuggest-final-methods -Wsuggest-override @gol
311 -Wmissing-format-attribute -Wsubobject-linkage @gol
312 -Wswitch -Wswitch-bool -Wswitch-default -Wswitch-enum @gol
313 -Wswitch-unreachable -Wsync-nand @gol
314 -Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs @gol
315 -Wtype-limits -Wundef @gol
316 -Wuninitialized -Wunknown-pragmas -Wunsafe-loop-optimizations @gol
317 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
318 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
319 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
320 -Wunused-const-variable -Wunused-const-variable=@var{n} @gol
321 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
322 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
323 -Wvla -Wvla-larger-than=@var{n} -Wvolatile-register-var -Wwrite-strings @gol
324 -Wzero-as-null-pointer-constant -Whsa}
326 @item C and Objective-C-only Warning Options
327 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
328 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
329 -Wold-style-declaration -Wold-style-definition @gol
330 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
331 -Wdeclaration-after-statement -Wpointer-sign}
333 @item Debugging Options
334 @xref{Debugging Options,,Options for Debugging Your Program}.
335 @gccoptlist{-g -g@var{level} -gcoff -gdwarf -gdwarf-@var{version} @gol
336 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
337 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
338 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
339 -fdebug-prefix-map=@var{old}=@var{new} -fdebug-types-section @gol
340 -feliminate-dwarf2-dups -fno-eliminate-unused-debug-types @gol
341 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
342 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
343 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
344 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
345 -fvar-tracking -fvar-tracking-assignments}
347 @item Optimization Options
348 @xref{Optimize Options,,Options that Control Optimization}.
349 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
350 -falign-jumps[=@var{n}] @gol
351 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
352 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
353 -fauto-inc-dec -fbranch-probabilities @gol
354 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
355 -fbtr-bb-exclusive -fcaller-saves @gol
356 -fcombine-stack-adjustments -fconserve-stack @gol
357 -fcompare-elim -fcprop-registers -fcrossjumping @gol
358 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
359 -fcx-limited-range @gol
360 -fdata-sections -fdce -fdelayed-branch @gol
361 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
362 -fdevirtualize-at-ltrans -fdse @gol
363 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
364 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
365 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
366 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
367 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
368 -fif-conversion2 -findirect-inlining @gol
369 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
370 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-cp-alignment -fipa-bit-cp @gol
371 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference -fipa-icf @gol
372 -fira-algorithm=@var{algorithm} @gol
373 -fira-region=@var{region} -fira-hoist-pressure @gol
374 -fira-loop-pressure -fno-ira-share-save-slots @gol
375 -fno-ira-share-spill-slots @gol
376 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
377 -fivopts -fkeep-inline-functions -fkeep-static-functions @gol
378 -fkeep-static-consts -flimit-function-alignment -flive-range-shrinkage @gol
379 -floop-block -floop-interchange -floop-strip-mine @gol
380 -floop-unroll-and-jam -floop-nest-optimize @gol
381 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
382 -flto-partition=@var{alg} -fmerge-all-constants @gol
383 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
384 -fmove-loop-invariants -fno-branch-count-reg @gol
385 -fno-defer-pop -fno-fp-int-builtin-inexact -fno-function-cse @gol
386 -fno-guess-branch-probability -fno-inline -fno-math-errno -fno-peephole @gol
387 -fno-peephole2 -fno-printf-return-value -fno-sched-interblock @gol
388 -fno-sched-spec -fno-signed-zeros @gol
389 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
390 -fomit-frame-pointer -foptimize-sibling-calls @gol
391 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
392 -fprefetch-loop-arrays @gol
393 -fprofile-correction @gol
394 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
395 -fprofile-reorder-functions @gol
396 -freciprocal-math -free -frename-registers -freorder-blocks @gol
397 -freorder-blocks-algorithm=@var{algorithm} @gol
398 -freorder-blocks-and-partition -freorder-functions @gol
399 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
400 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
401 -fsched-spec-load -fsched-spec-load-dangerous @gol
402 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
403 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
404 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
405 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
406 -fschedule-fusion @gol
407 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
408 -fselective-scheduling -fselective-scheduling2 @gol
409 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
410 -fsemantic-interposition -fshrink-wrap -fshrink-wrap-separate @gol
411 -fsignaling-nans @gol
412 -fsingle-precision-constant -fsplit-ivs-in-unroller -fsplit-loops@gol
414 -fsplit-wide-types -fssa-backprop -fssa-phiopt @gol
415 -fstdarg-opt -fstore-merging -fstrict-aliasing @gol
416 -fstrict-overflow -fthread-jumps -ftracer -ftree-bit-ccp @gol
417 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
418 -ftree-coalesce-vars -ftree-copy-prop -ftree-dce -ftree-dominator-opts @gol
419 -ftree-dse -ftree-forwprop -ftree-fre -fcode-hoisting -ftree-loop-if-convert @gol
421 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
422 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
423 -ftree-loop-vectorize @gol
424 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
425 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
426 -ftree-switch-conversion -ftree-tail-merge @gol
427 -ftree-ter -ftree-vectorize -ftree-vrp -funconstrained-commons @gol
428 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
429 -funsafe-math-optimizations -funswitch-loops @gol
430 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
431 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
432 --param @var{name}=@var{value}
433 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
435 @item Program Instrumentation Options
436 @xref{Instrumentation Options,,Program Instrumentation Options}.
437 @gccoptlist{-p -pg -fprofile-arcs --coverage -ftest-coverage @gol
438 -fprofile-dir=@var{path} -fprofile-generate -fprofile-generate=@var{path} @gol
439 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
440 -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
441 -fsanitize-undefined-trap-on-error -fbounds-check @gol
442 -fcheck-pointer-bounds -fchkp-check-incomplete-type @gol
443 -fchkp-first-field-has-own-bounds -fchkp-narrow-bounds @gol
444 -fchkp-narrow-to-innermost-array -fchkp-optimize @gol
445 -fchkp-use-fast-string-functions -fchkp-use-nochk-string-functions @gol
446 -fchkp-use-static-bounds -fchkp-use-static-const-bounds @gol
447 -fchkp-treat-zero-dynamic-size-as-infinite -fchkp-check-read @gol
448 -fchkp-check-read -fchkp-check-write -fchkp-store-bounds @gol
449 -fchkp-instrument-calls -fchkp-instrument-marked-only @gol
450 -fchkp-use-wrappers @gol
451 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
452 -fstack-protector-explicit -fstack-check @gol
453 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
454 -fno-stack-limit -fsplit-stack @gol
455 -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} @gol
456 -fvtv-counts -fvtv-debug @gol
457 -finstrument-functions @gol
458 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
459 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}}
461 @item Preprocessor Options
462 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
463 @gccoptlist{-A@var{question}=@var{answer} @gol
464 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
465 -C -dD -dI -dM -dN @gol
466 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
467 -idirafter @var{dir} @gol
468 -include @var{file} -imacros @var{file} @gol
469 -iprefix @var{file} -iwithprefix @var{dir} @gol
470 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
471 -imultilib @var{dir} -isysroot @var{dir} @gol
472 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
473 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
474 -remap -trigraphs -undef -U@var{macro} @gol
475 -Wp,@var{option} -Xpreprocessor @var{option} -no-integrated-cpp}
477 @item Assembler Option
478 @xref{Assembler Options,,Passing Options to the Assembler}.
479 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
482 @xref{Link Options,,Options for Linking}.
483 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
484 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
485 -s -static -static-libgcc -static-libstdc++ @gol
486 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
487 -static-libmpx -static-libmpxwrappers @gol
488 -shared -shared-libgcc -symbolic @gol
489 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
490 -u @var{symbol} -z @var{keyword}}
492 @item Directory Options
493 @xref{Directory Options,,Options for Directory Search}.
494 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
495 -iquote@var{dir} -L@var{dir} -no-canonical-prefixes -I- @gol
496 --sysroot=@var{dir} --no-sysroot-suffix}
498 @item Code Generation Options
499 @xref{Code Gen Options,,Options for Code Generation Conventions}.
500 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
501 -ffixed-@var{reg} -fexceptions @gol
502 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
503 -fasynchronous-unwind-tables @gol
505 -finhibit-size-directive -fno-common -fno-ident @gol
506 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol
507 -fno-jump-tables @gol
508 -frecord-gcc-switches @gol
509 -freg-struct-return -fshort-enums -fshort-wchar @gol
510 -fverbose-asm -fpack-struct[=@var{n}] @gol
511 -fleading-underscore -ftls-model=@var{model} @gol
512 -fstack-reuse=@var{reuse_level} @gol
513 -ftrampolines -ftrapv -fwrapv @gol
514 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
515 -fstrict-volatile-bitfields -fsync-libcalls}
517 @item Developer Options
518 @xref{Developer Options,,GCC Developer Options}.
519 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
520 -fchecking -fchecking=@var{n} -fdbg-cnt-list @gol
521 -fdbg-cnt=@var{counter-value-list} @gol
522 -fdisable-ipa-@var{pass_name} @gol
523 -fdisable-rtl-@var{pass_name} @gol
524 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
525 -fdisable-tree-@var{pass_name} @gol
526 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
527 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
528 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
529 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
530 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
532 -fdump-rtl-@var{pass} -fdump-rtl-@var{pass}=@var{filename} @gol
533 -fdump-statistics @gol
535 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
536 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
537 -fdump-tree-cfg -fdump-tree-alias @gol
539 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
540 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
541 -fdump-tree-gimple@r{[}-raw@r{]} @gol
542 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
543 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
544 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
545 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
546 -fdump-tree-backprop@r{[}-@var{n}@r{]} @gol
547 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
548 -fdump-tree-nrv -fdump-tree-vect @gol
549 -fdump-tree-sink @gol
550 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
551 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
552 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
553 -fdump-tree-vtable-verify @gol
554 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
555 -fdump-tree-split-paths@r{[}-@var{n}@r{]} @gol
556 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
557 -fdump-final-insns=@var{file} @gol
558 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
559 -fenable-@var{kind}-@var{pass} @gol
560 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
561 -fira-verbose=@var{n} @gol
562 -flto-report -flto-report-wpa -fmem-report-wpa @gol
563 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report @gol
564 -fopt-info -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
565 -fprofile-report @gol
566 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
567 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
568 -fstats -fstack-usage -ftime-report -ftime-report-details @gol
569 -fvar-tracking-assignments-toggle -gtoggle @gol
570 -print-file-name=@var{library} -print-libgcc-file-name @gol
571 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
572 -print-prog-name=@var{program} -print-search-dirs -Q @gol
573 -print-sysroot -print-sysroot-headers-suffix @gol
574 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
576 @item Machine-Dependent Options
577 @xref{Submodel Options,,Machine-Dependent Options}.
578 @c This list is ordered alphanumerically by subsection name.
579 @c Try and put the significant identifier (CPU or system) first,
580 @c so users have a clue at guessing where the ones they want will be.
582 @emph{AArch64 Options}
583 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
584 -mgeneral-regs-only @gol
585 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
587 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
588 -mtls-dialect=desc -mtls-dialect=traditional @gol
589 -mtls-size=@var{size} @gol
590 -mfix-cortex-a53-835769 -mno-fix-cortex-a53-835769 @gol
591 -mfix-cortex-a53-843419 -mno-fix-cortex-a53-843419 @gol
592 -mlow-precision-recip-sqrt -mno-low-precision-recip-sqrt@gol
593 -mlow-precision-sqrt -mno-low-precision-sqrt@gol
594 -mlow-precision-div -mno-low-precision-div @gol
595 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
597 @emph{Adapteva Epiphany Options}
598 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
599 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
600 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
601 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
602 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
603 -msplit-vecmove-early -m1reg-@var{reg}}
606 @gccoptlist{-mbarrel-shifter @gol
607 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
608 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
609 -mea -mno-mpy -mmul32x16 -mmul64 -matomic @gol
610 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
611 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
612 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
613 -mlong-calls -mmedium-calls -msdata @gol
614 -mucb-mcount -mvolatile-cache -mtp-regno=@var{regno} @gol
615 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
616 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
617 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
618 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
619 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
620 -mtune=@var{cpu} -mmultcost=@var{num} @gol
621 -munalign-prob-threshold=@var{probability} -mmpy-option=@var{multo} @gol
622 -mdiv-rem -mcode-density -mll64 -mfpu=@var{fpu}}
625 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
626 -mabi=@var{name} @gol
627 -mapcs-stack-check -mno-apcs-stack-check @gol
628 -mapcs-reentrant -mno-apcs-reentrant @gol
629 -msched-prolog -mno-sched-prolog @gol
630 -mlittle-endian -mbig-endian @gol
631 -mfloat-abi=@var{name} @gol
632 -mfp16-format=@var{name}
633 -mthumb-interwork -mno-thumb-interwork @gol
634 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
635 -mtune=@var{name} -mprint-tune-info @gol
636 -mstructure-size-boundary=@var{n} @gol
637 -mabort-on-noreturn @gol
638 -mlong-calls -mno-long-calls @gol
639 -msingle-pic-base -mno-single-pic-base @gol
640 -mpic-register=@var{reg} @gol
641 -mnop-fun-dllimport @gol
642 -mpoke-function-name @gol
644 -mtpcs-frame -mtpcs-leaf-frame @gol
645 -mcaller-super-interworking -mcallee-super-interworking @gol
646 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
647 -mword-relocations @gol
648 -mfix-cortex-m3-ldrd @gol
649 -munaligned-access @gol
650 -mneon-for-64bits @gol
651 -mslow-flash-data @gol
652 -masm-syntax-unified @gol
658 @gccoptlist{-mmcu=@var{mcu} -mabsdata -maccumulate-args @gol
659 -mbranch-cost=@var{cost} @gol
660 -mcall-prologues -mint8 -mn_flash=@var{size} -mno-interrupts @gol
661 -mrelax -mrmw -mstrict-X -mtiny-stack -mfract-convert-truncate -nodevicelib @gol
662 -Waddr-space-convert -Wmisspelled-isr}
664 @emph{Blackfin Options}
665 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
666 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
667 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
668 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
669 -mno-id-shared-library -mshared-library-id=@var{n} @gol
670 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
671 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
672 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
676 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
677 -msim -msdata=@var{sdata-type}}
680 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
681 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
682 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
683 -mstack-align -mdata-align -mconst-align @gol
684 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
685 -melf -maout -melinux -mlinux -sim -sim2 @gol
686 -mmul-bug-workaround -mno-mul-bug-workaround}
689 @gccoptlist{-mmac @gol
690 -mcr16cplus -mcr16c @gol
691 -msim -mint32 -mbit-ops
692 -mdata-model=@var{model}}
694 @emph{Darwin Options}
695 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
696 -arch_only -bind_at_load -bundle -bundle_loader @gol
697 -client_name -compatibility_version -current_version @gol
699 -dependency-file -dylib_file -dylinker_install_name @gol
700 -dynamic -dynamiclib -exported_symbols_list @gol
701 -filelist -flat_namespace -force_cpusubtype_ALL @gol
702 -force_flat_namespace -headerpad_max_install_names @gol
704 -image_base -init -install_name -keep_private_externs @gol
705 -multi_module -multiply_defined -multiply_defined_unused @gol
706 -noall_load -no_dead_strip_inits_and_terms @gol
707 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
708 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
709 -private_bundle -read_only_relocs -sectalign @gol
710 -sectobjectsymbols -whyload -seg1addr @gol
711 -sectcreate -sectobjectsymbols -sectorder @gol
712 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
713 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
714 -segprot -segs_read_only_addr -segs_read_write_addr @gol
715 -single_module -static -sub_library -sub_umbrella @gol
716 -twolevel_namespace -umbrella -undefined @gol
717 -unexported_symbols_list -weak_reference_mismatches @gol
718 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
719 -mkernel -mone-byte-bool}
721 @emph{DEC Alpha Options}
722 @gccoptlist{-mno-fp-regs -msoft-float @gol
723 -mieee -mieee-with-inexact -mieee-conformant @gol
724 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
725 -mtrap-precision=@var{mode} -mbuild-constants @gol
726 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
727 -mbwx -mmax -mfix -mcix @gol
728 -mfloat-vax -mfloat-ieee @gol
729 -mexplicit-relocs -msmall-data -mlarge-data @gol
730 -msmall-text -mlarge-text @gol
731 -mmemory-latency=@var{time}}
734 @gccoptlist{-msmall-model -mno-lsim}
737 @gccoptlist{-msim -mlra -mnodiv}
740 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
741 -mhard-float -msoft-float @gol
742 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
743 -mdouble -mno-double @gol
744 -mmedia -mno-media -mmuladd -mno-muladd @gol
745 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
746 -mlinked-fp -mlong-calls -malign-labels @gol
747 -mlibrary-pic -macc-4 -macc-8 @gol
748 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
749 -moptimize-membar -mno-optimize-membar @gol
750 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
751 -mvliw-branch -mno-vliw-branch @gol
752 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
753 -mno-nested-cond-exec -mtomcat-stats @gol
757 @emph{GNU/Linux Options}
758 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
759 -tno-android-cc -tno-android-ld}
761 @emph{H8/300 Options}
762 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
765 @gccoptlist{-march=@var{architecture-type} @gol
766 -mcaller-copies -mdisable-fpregs -mdisable-indexing @gol
767 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
768 -mfixed-range=@var{register-range} @gol
769 -mjump-in-delay -mlinker-opt -mlong-calls @gol
770 -mlong-load-store -mno-disable-fpregs @gol
771 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
772 -mno-jump-in-delay -mno-long-load-store @gol
773 -mno-portable-runtime -mno-soft-float @gol
774 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
775 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
776 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
777 -munix=@var{unix-std} -nolibdld -static -threads}
780 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
781 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
782 -mconstant-gp -mauto-pic -mfused-madd @gol
783 -minline-float-divide-min-latency @gol
784 -minline-float-divide-max-throughput @gol
785 -mno-inline-float-divide @gol
786 -minline-int-divide-min-latency @gol
787 -minline-int-divide-max-throughput @gol
788 -mno-inline-int-divide @gol
789 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
790 -mno-inline-sqrt @gol
791 -mdwarf2-asm -mearly-stop-bits @gol
792 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
793 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
794 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
795 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
796 -msched-spec-ldc -msched-spec-control-ldc @gol
797 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
798 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
799 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
800 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
803 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
804 -msign-extend-enabled -muser-enabled}
806 @emph{M32R/D Options}
807 @gccoptlist{-m32r2 -m32rx -m32r @gol
809 -malign-loops -mno-align-loops @gol
810 -missue-rate=@var{number} @gol
811 -mbranch-cost=@var{number} @gol
812 -mmodel=@var{code-size-model-type} @gol
813 -msdata=@var{sdata-type} @gol
814 -mno-flush-func -mflush-func=@var{name} @gol
815 -mno-flush-trap -mflush-trap=@var{number} @gol
819 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
821 @emph{M680x0 Options}
822 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
823 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
824 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
825 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
826 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
827 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
828 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
829 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
833 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
834 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
835 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
836 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
837 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
840 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
841 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
842 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
843 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
846 @emph{MicroBlaze Options}
847 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
848 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
849 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
850 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
851 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
854 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
855 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
856 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
857 -mips16 -mno-mips16 -mflip-mips16 @gol
858 -minterlink-compressed -mno-interlink-compressed @gol
859 -minterlink-mips16 -mno-interlink-mips16 @gol
860 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
861 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
862 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
863 -mno-float -msingle-float -mdouble-float @gol
864 -modd-spreg -mno-odd-spreg @gol
865 -mabs=@var{mode} -mnan=@var{encoding} @gol
866 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
869 -mvirt -mno-virt @gol
871 -mmicromips -mno-micromips @gol
873 -mfpu=@var{fpu-type} @gol
874 -msmartmips -mno-smartmips @gol
875 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
876 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
877 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
878 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
879 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
880 -membedded-data -mno-embedded-data @gol
881 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
882 -mcode-readable=@var{setting} @gol
883 -msplit-addresses -mno-split-addresses @gol
884 -mexplicit-relocs -mno-explicit-relocs @gol
885 -mcheck-zero-division -mno-check-zero-division @gol
886 -mdivide-traps -mdivide-breaks @gol
887 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
888 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
889 -mfix-24k -mno-fix-24k @gol
890 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
891 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
892 -mfix-vr4120 -mno-fix-vr4120 @gol
893 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
894 -mflush-func=@var{func} -mno-flush-func @gol
895 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
896 -mcompact-branches=@var{policy} @gol
897 -mfp-exceptions -mno-fp-exceptions @gol
898 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
899 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address @gol
900 -mframe-header-opt -mno-frame-header-opt}
903 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
904 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
905 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
906 -mno-base-addresses -msingle-exit -mno-single-exit}
908 @emph{MN10300 Options}
909 @gccoptlist{-mmult-bug -mno-mult-bug @gol
910 -mno-am33 -mam33 -mam33-2 -mam34 @gol
911 -mtune=@var{cpu-type} @gol
912 -mreturn-pointer-on-d0 @gol
913 -mno-crt0 -mrelax -mliw -msetlb}
916 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
918 @emph{MSP430 Options}
919 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
921 -mcode-region= -mdata-region= @gol
922 -msilicon-errata= -msilicon-errata-warn= @gol
926 @gccoptlist{-mbig-endian -mlittle-endian @gol
927 -mreduced-regs -mfull-regs @gol
928 -mcmov -mno-cmov @gol
929 -mperf-ext -mno-perf-ext @gol
930 -mv3push -mno-v3push @gol
931 -m16bit -mno-16bit @gol
932 -misr-vector-size=@var{num} @gol
933 -mcache-block-size=@var{num} @gol
934 -march=@var{arch} @gol
935 -mcmodel=@var{code-model} @gol
938 @emph{Nios II Options}
939 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
941 -mno-bypass-cache -mbypass-cache @gol
942 -mno-cache-volatile -mcache-volatile @gol
943 -mno-fast-sw-div -mfast-sw-div @gol
944 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
945 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
946 -mcustom-fpu-cfg=@var{name} @gol
947 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
948 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
950 @emph{Nvidia PTX Options}
951 @gccoptlist{-m32 -m64 -mmainkernel -moptimize}
953 @emph{PDP-11 Options}
954 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
955 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
956 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
957 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
958 -mbranch-expensive -mbranch-cheap @gol
959 -munix-asm -mdec-asm}
961 @emph{picoChip Options}
962 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
963 -msymbol-as-address -mno-inefficient-warnings}
965 @emph{PowerPC Options}
966 See RS/6000 and PowerPC Options.
969 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
970 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
971 -m64bit-doubles -m32bit-doubles -msave-mduc-in-interrupts}
973 @emph{RS/6000 and PowerPC Options}
974 @gccoptlist{-mcpu=@var{cpu-type} @gol
975 -mtune=@var{cpu-type} @gol
976 -mcmodel=@var{code-model} @gol
978 -maltivec -mno-altivec @gol
979 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
980 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
981 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
982 -mfprnd -mno-fprnd @gol
983 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
984 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
985 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
986 -malign-power -malign-natural @gol
987 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
988 -msingle-float -mdouble-float -msimple-fpu @gol
989 -mstring -mno-string -mupdate -mno-update @gol
990 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
991 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
992 -mstrict-align -mno-strict-align -mrelocatable @gol
993 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
994 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
995 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
996 -mprioritize-restricted-insns=@var{priority} @gol
997 -msched-costly-dep=@var{dependence_type} @gol
998 -minsert-sched-nops=@var{scheme} @gol
999 -mcall-sysv -mcall-netbsd @gol
1000 -maix-struct-return -msvr4-struct-return @gol
1001 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
1002 -mblock-move-inline-limit=@var{num} @gol
1003 -misel -mno-isel @gol
1004 -misel=yes -misel=no @gol
1006 -mspe=yes -mspe=no @gol
1008 -mgen-cell-microcode -mwarn-cell-microcode @gol
1009 -mvrsave -mno-vrsave @gol
1010 -mmulhw -mno-mulhw @gol
1011 -mdlmzb -mno-dlmzb @gol
1012 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
1013 -mprototype -mno-prototype @gol
1014 -msim -mmvme -mads -myellowknife -memb -msdata @gol
1015 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
1016 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
1017 -mno-recip-precision @gol
1018 -mveclibabi=@var{type} -mfriz -mno-friz @gol
1019 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
1020 -msave-toc-indirect -mno-save-toc-indirect @gol
1021 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
1022 -mcrypto -mno-crypto -mhtm -mno-htm -mdirect-move -mno-direct-move @gol
1023 -mquad-memory -mno-quad-memory @gol
1024 -mquad-memory-atomic -mno-quad-memory-atomic @gol
1025 -mcompat-align-parm -mno-compat-align-parm @gol
1026 -mupper-regs-df -mno-upper-regs-df -mupper-regs-sf -mno-upper-regs-sf @gol
1027 -mupper-regs-di -mno-upper-regs-di @gol
1028 -mupper-regs -mno-upper-regs @gol
1029 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol
1030 -mgnu-attribute -mno-gnu-attribute @gol
1034 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
1036 -mbig-endian-data -mlittle-endian-data @gol
1039 -mas100-syntax -mno-as100-syntax@gol
1041 -mmax-constant-size=@gol
1044 -mallow-string-insns -mno-allow-string-insns@gol
1046 -mno-warn-multiple-fast-interrupts@gol
1047 -msave-acc-in-interrupts}
1049 @emph{S/390 and zSeries Options}
1050 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1051 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
1052 -mlong-double-64 -mlong-double-128 @gol
1053 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
1054 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
1055 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
1056 -mhtm -mvx -mzvector @gol
1057 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
1058 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
1059 -mhotpatch=@var{halfwords},@var{halfwords}}
1061 @emph{Score Options}
1062 @gccoptlist{-meb -mel @gol
1066 -mscore5 -mscore5u -mscore7 -mscore7d}
1069 @gccoptlist{-m1 -m2 -m2e @gol
1070 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
1072 -m4-nofpu -m4-single-only -m4-single -m4 @gol
1073 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
1074 -mb -ml -mdalign -mrelax @gol
1075 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
1076 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
1077 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1078 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1079 -maccumulate-outgoing-args @gol
1080 -matomic-model=@var{atomic-model} @gol
1081 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1082 -mcbranch-force-delay-slot @gol
1083 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1084 -mpretend-cmove -mtas}
1086 @emph{Solaris 2 Options}
1087 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1090 @emph{SPARC Options}
1091 @gccoptlist{-mcpu=@var{cpu-type} @gol
1092 -mtune=@var{cpu-type} @gol
1093 -mcmodel=@var{code-model} @gol
1094 -mmemory-model=@var{mem-model} @gol
1095 -m32 -m64 -mapp-regs -mno-app-regs @gol
1096 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1097 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1098 -mhard-quad-float -msoft-quad-float @gol
1099 -mstack-bias -mno-stack-bias @gol
1100 -mstd-struct-return -mno-std-struct-return @gol
1101 -munaligned-doubles -mno-unaligned-doubles @gol
1102 -muser-mode -mno-user-mode @gol
1103 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1104 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1105 -mcbcond -mno-cbcond -mfmaf -mno-fmaf @gol
1106 -mpopc -mno-popc -msubxc -mno-subxc@gol
1107 -mfix-at697f -mfix-ut699}
1110 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1111 -msafe-dma -munsafe-dma @gol
1113 -msmall-mem -mlarge-mem -mstdmain @gol
1114 -mfixed-range=@var{register-range} @gol
1116 -maddress-space-conversion -mno-address-space-conversion @gol
1117 -mcache-size=@var{cache-size} @gol
1118 -matomic-updates -mno-atomic-updates}
1120 @emph{System V Options}
1121 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1123 @emph{TILE-Gx Options}
1124 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1125 -mcmodel=@var{code-model}}
1127 @emph{TILEPro Options}
1128 @gccoptlist{-mcpu=@var{cpu} -m32}
1131 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1132 -mprolog-function -mno-prolog-function -mspace @gol
1133 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1134 -mapp-regs -mno-app-regs @gol
1135 -mdisable-callt -mno-disable-callt @gol
1136 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1137 -mv850e -mv850 -mv850e3v5 @gol
1148 @gccoptlist{-mg -mgnu -munix}
1150 @emph{Visium Options}
1151 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1152 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1155 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1156 -mpointer-size=@var{size}}
1158 @emph{VxWorks Options}
1159 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1160 -Xbind-lazy -Xbind-now}
1163 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1164 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1165 -mfpmath=@var{unit} @gol
1166 -masm=@var{dialect} -mno-fancy-math-387 @gol
1167 -mno-fp-ret-in-387 -m80387 -mhard-float -msoft-float @gol
1168 -mno-wide-multiply -mrtd -malign-double @gol
1169 -mpreferred-stack-boundary=@var{num} @gol
1170 -mincoming-stack-boundary=@var{num} @gol
1171 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1172 -mrecip -mrecip=@var{opt} @gol
1173 -mvzeroupper -mprefer-avx128 @gol
1174 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1175 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1176 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1177 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
1178 -mprefetchwt1 -mclflushopt -mxsavec -mxsaves @gol
1179 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
1180 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx -mmwaitx -mclzero
1181 -mpku -mthreads @gol
1182 -mms-bitfields -mno-align-stringops -minline-all-stringops @gol
1183 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1184 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1185 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1186 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1187 -mregparm=@var{num} -msseregparm @gol
1188 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1189 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1190 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1191 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1192 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1193 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1194 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1195 -malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol
1196 -mmitigate-rop -mgeneral-regs-only}
1198 @emph{x86 Windows Options}
1199 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1200 -mnop-fun-dllimport -mthread @gol
1201 -municode -mwin32 -mwindows -fno-set-stack-executable}
1203 @emph{Xstormy16 Options}
1206 @emph{Xtensa Options}
1207 @gccoptlist{-mconst16 -mno-const16 @gol
1208 -mfused-madd -mno-fused-madd @gol
1210 -mserialize-volatile -mno-serialize-volatile @gol
1211 -mtext-section-literals -mno-text-section-literals @gol
1212 -mauto-litpools -mno-auto-litpools @gol
1213 -mtarget-align -mno-target-align @gol
1214 -mlongcalls -mno-longcalls}
1216 @emph{zSeries Options}
1217 See S/390 and zSeries Options.
1221 @node Overall Options
1222 @section Options Controlling the Kind of Output
1224 Compilation can involve up to four stages: preprocessing, compilation
1225 proper, assembly and linking, always in that order. GCC is capable of
1226 preprocessing and compiling several files either into several
1227 assembler input files, or into one assembler input file; then each
1228 assembler input file produces an object file, and linking combines all
1229 the object files (those newly compiled, and those specified as input)
1230 into an executable file.
1232 @cindex file name suffix
1233 For any given input file, the file name suffix determines what kind of
1234 compilation is done:
1238 C source code that must be preprocessed.
1241 C source code that should not be preprocessed.
1244 C++ source code that should not be preprocessed.
1247 Objective-C source code. Note that you must link with the @file{libobjc}
1248 library to make an Objective-C program work.
1251 Objective-C source code that should not be preprocessed.
1255 Objective-C++ source code. Note that you must link with the @file{libobjc}
1256 library to make an Objective-C++ program work. Note that @samp{.M} refers
1257 to a literal capital M@.
1259 @item @var{file}.mii
1260 Objective-C++ source code that should not be preprocessed.
1263 C, C++, Objective-C or Objective-C++ header file to be turned into a
1264 precompiled header (default), or C, C++ header file to be turned into an
1265 Ada spec (via the @option{-fdump-ada-spec} switch).
1268 @itemx @var{file}.cp
1269 @itemx @var{file}.cxx
1270 @itemx @var{file}.cpp
1271 @itemx @var{file}.CPP
1272 @itemx @var{file}.c++
1274 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1275 the last two letters must both be literally @samp{x}. Likewise,
1276 @samp{.C} refers to a literal capital C@.
1280 Objective-C++ source code that must be preprocessed.
1282 @item @var{file}.mii
1283 Objective-C++ source code that should not be preprocessed.
1287 @itemx @var{file}.hp
1288 @itemx @var{file}.hxx
1289 @itemx @var{file}.hpp
1290 @itemx @var{file}.HPP
1291 @itemx @var{file}.h++
1292 @itemx @var{file}.tcc
1293 C++ header file to be turned into a precompiled header or Ada spec.
1296 @itemx @var{file}.for
1297 @itemx @var{file}.ftn
1298 Fixed form Fortran source code that should not be preprocessed.
1301 @itemx @var{file}.FOR
1302 @itemx @var{file}.fpp
1303 @itemx @var{file}.FPP
1304 @itemx @var{file}.FTN
1305 Fixed form Fortran source code that must be preprocessed (with the traditional
1308 @item @var{file}.f90
1309 @itemx @var{file}.f95
1310 @itemx @var{file}.f03
1311 @itemx @var{file}.f08
1312 Free form Fortran source code that should not be preprocessed.
1314 @item @var{file}.F90
1315 @itemx @var{file}.F95
1316 @itemx @var{file}.F03
1317 @itemx @var{file}.F08
1318 Free form Fortran source code that must be preprocessed (with the
1319 traditional preprocessor).
1324 @item @var{file}.ads
1325 Ada source code file that contains a library unit declaration (a
1326 declaration of a package, subprogram, or generic, or a generic
1327 instantiation), or a library unit renaming declaration (a package,
1328 generic, or subprogram renaming declaration). Such files are also
1331 @item @var{file}.adb
1332 Ada source code file containing a library unit body (a subprogram or
1333 package body). Such files are also called @dfn{bodies}.
1335 @c GCC also knows about some suffixes for languages not yet included:
1346 @itemx @var{file}.sx
1347 Assembler code that must be preprocessed.
1350 An object file to be fed straight into linking.
1351 Any file name with no recognized suffix is treated this way.
1355 You can specify the input language explicitly with the @option{-x} option:
1358 @item -x @var{language}
1359 Specify explicitly the @var{language} for the following input files
1360 (rather than letting the compiler choose a default based on the file
1361 name suffix). This option applies to all following input files until
1362 the next @option{-x} option. Possible values for @var{language} are:
1364 c c-header cpp-output
1365 c++ c++-header c++-cpp-output
1366 objective-c objective-c-header objective-c-cpp-output
1367 objective-c++ objective-c++-header objective-c++-cpp-output
1368 assembler assembler-with-cpp
1370 f77 f77-cpp-input f95 f95-cpp-input
1375 Turn off any specification of a language, so that subsequent files are
1376 handled according to their file name suffixes (as they are if @option{-x}
1377 has not been used at all).
1380 If you only want some of the stages of compilation, you can use
1381 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1382 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1383 @command{gcc} is to stop. Note that some combinations (for example,
1384 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1389 Compile or assemble the source files, but do not link. The linking
1390 stage simply is not done. The ultimate output is in the form of an
1391 object file for each source file.
1393 By default, the object file name for a source file is made by replacing
1394 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1396 Unrecognized input files, not requiring compilation or assembly, are
1401 Stop after the stage of compilation proper; do not assemble. The output
1402 is in the form of an assembler code file for each non-assembler input
1405 By default, the assembler file name for a source file is made by
1406 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1408 Input files that don't require compilation are ignored.
1412 Stop after the preprocessing stage; do not run the compiler proper. The
1413 output is in the form of preprocessed source code, which is sent to the
1416 Input files that don't require preprocessing are ignored.
1418 @cindex output file option
1421 Place output in file @var{file}. This applies to whatever
1422 sort of output is being produced, whether it be an executable file,
1423 an object file, an assembler file or preprocessed C code.
1425 If @option{-o} is not specified, the default is to put an executable
1426 file in @file{a.out}, the object file for
1427 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1428 assembler file in @file{@var{source}.s}, a precompiled header file in
1429 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1434 Print (on standard error output) the commands executed to run the stages
1435 of compilation. Also print the version number of the compiler driver
1436 program and of the preprocessor and the compiler proper.
1440 Like @option{-v} except the commands are not executed and arguments
1441 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1442 This is useful for shell scripts to capture the driver-generated command lines.
1446 Print (on the standard output) a description of the command-line options
1447 understood by @command{gcc}. If the @option{-v} option is also specified
1448 then @option{--help} is also passed on to the various processes
1449 invoked by @command{gcc}, so that they can display the command-line options
1450 they accept. If the @option{-Wextra} option has also been specified
1451 (prior to the @option{--help} option), then command-line options that
1452 have no documentation associated with them are also displayed.
1455 @opindex target-help
1456 Print (on the standard output) a description of target-specific command-line
1457 options for each tool. For some targets extra target-specific
1458 information may also be printed.
1460 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1461 Print (on the standard output) a description of the command-line
1462 options understood by the compiler that fit into all specified classes
1463 and qualifiers. These are the supported classes:
1466 @item @samp{optimizers}
1467 Display all of the optimization options supported by the
1470 @item @samp{warnings}
1471 Display all of the options controlling warning messages
1472 produced by the compiler.
1475 Display target-specific options. Unlike the
1476 @option{--target-help} option however, target-specific options of the
1477 linker and assembler are not displayed. This is because those
1478 tools do not currently support the extended @option{--help=} syntax.
1481 Display the values recognized by the @option{--param}
1484 @item @var{language}
1485 Display the options supported for @var{language}, where
1486 @var{language} is the name of one of the languages supported in this
1490 Display the options that are common to all languages.
1493 These are the supported qualifiers:
1496 @item @samp{undocumented}
1497 Display only those options that are undocumented.
1500 Display options taking an argument that appears after an equal
1501 sign in the same continuous piece of text, such as:
1502 @samp{--help=target}.
1504 @item @samp{separate}
1505 Display options taking an argument that appears as a separate word
1506 following the original option, such as: @samp{-o output-file}.
1509 Thus for example to display all the undocumented target-specific
1510 switches supported by the compiler, use:
1513 --help=target,undocumented
1516 The sense of a qualifier can be inverted by prefixing it with the
1517 @samp{^} character, so for example to display all binary warning
1518 options (i.e., ones that are either on or off and that do not take an
1519 argument) that have a description, use:
1522 --help=warnings,^joined,^undocumented
1525 The argument to @option{--help=} should not consist solely of inverted
1528 Combining several classes is possible, although this usually
1529 restricts the output so much that there is nothing to display. One
1530 case where it does work, however, is when one of the classes is
1531 @var{target}. For example, to display all the target-specific
1532 optimization options, use:
1535 --help=target,optimizers
1538 The @option{--help=} option can be repeated on the command line. Each
1539 successive use displays its requested class of options, skipping
1540 those that have already been displayed.
1542 If the @option{-Q} option appears on the command line before the
1543 @option{--help=} option, then the descriptive text displayed by
1544 @option{--help=} is changed. Instead of describing the displayed
1545 options, an indication is given as to whether the option is enabled,
1546 disabled or set to a specific value (assuming that the compiler
1547 knows this at the point where the @option{--help=} option is used).
1549 Here is a truncated example from the ARM port of @command{gcc}:
1552 % gcc -Q -mabi=2 --help=target -c
1553 The following options are target specific:
1555 -mabort-on-noreturn [disabled]
1559 The output is sensitive to the effects of previous command-line
1560 options, so for example it is possible to find out which optimizations
1561 are enabled at @option{-O2} by using:
1564 -Q -O2 --help=optimizers
1567 Alternatively you can discover which binary optimizations are enabled
1568 by @option{-O3} by using:
1571 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1572 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1573 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1578 Display the version number and copyrights of the invoked GCC@.
1580 @item -pass-exit-codes
1581 @opindex pass-exit-codes
1582 Normally the @command{gcc} program exits with the code of 1 if any
1583 phase of the compiler returns a non-success return code. If you specify
1584 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1585 the numerically highest error produced by any phase returning an error
1586 indication. The C, C++, and Fortran front ends return 4 if an internal
1587 compiler error is encountered.
1591 Use pipes rather than temporary files for communication between the
1592 various stages of compilation. This fails to work on some systems where
1593 the assembler is unable to read from a pipe; but the GNU assembler has
1596 @item -specs=@var{file}
1598 Process @var{file} after the compiler reads in the standard @file{specs}
1599 file, in order to override the defaults which the @command{gcc} driver
1600 program uses when determining what switches to pass to @command{cc1},
1601 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
1602 @option{-specs=@var{file}} can be specified on the command line, and they
1603 are processed in order, from left to right. @xref{Spec Files}, for
1604 information about the format of the @var{file}.
1608 Invoke all subcommands under a wrapper program. The name of the
1609 wrapper program and its parameters are passed as a comma separated
1613 gcc -c t.c -wrapper gdb,--args
1617 This invokes all subprograms of @command{gcc} under
1618 @samp{gdb --args}, thus the invocation of @command{cc1} is
1619 @samp{gdb --args cc1 @dots{}}.
1621 @item -fplugin=@var{name}.so
1623 Load the plugin code in file @var{name}.so, assumed to be a
1624 shared object to be dlopen'd by the compiler. The base name of
1625 the shared object file is used to identify the plugin for the
1626 purposes of argument parsing (See
1627 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1628 Each plugin should define the callback functions specified in the
1631 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1632 @opindex fplugin-arg
1633 Define an argument called @var{key} with a value of @var{value}
1634 for the plugin called @var{name}.
1636 @item -fdump-ada-spec@r{[}-slim@r{]}
1637 @opindex fdump-ada-spec
1638 For C and C++ source and include files, generate corresponding Ada specs.
1639 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1640 GNAT User's Guide}, which provides detailed documentation on this feature.
1642 @item -fada-spec-parent=@var{unit}
1643 @opindex fada-spec-parent
1644 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1645 Ada specs as child units of parent @var{unit}.
1647 @item -fdump-go-spec=@var{file}
1648 @opindex fdump-go-spec
1649 For input files in any language, generate corresponding Go
1650 declarations in @var{file}. This generates Go @code{const},
1651 @code{type}, @code{var}, and @code{func} declarations which may be a
1652 useful way to start writing a Go interface to code written in some
1655 @include @value{srcdir}/../libiberty/at-file.texi
1659 @section Compiling C++ Programs
1661 @cindex suffixes for C++ source
1662 @cindex C++ source file suffixes
1663 C++ source files conventionally use one of the suffixes @samp{.C},
1664 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1665 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1666 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1667 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1668 files with these names and compiles them as C++ programs even if you
1669 call the compiler the same way as for compiling C programs (usually
1670 with the name @command{gcc}).
1674 However, the use of @command{gcc} does not add the C++ library.
1675 @command{g++} is a program that calls GCC and automatically specifies linking
1676 against the C++ library. It treats @samp{.c},
1677 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1678 files unless @option{-x} is used. This program is also useful when
1679 precompiling a C header file with a @samp{.h} extension for use in C++
1680 compilations. On many systems, @command{g++} is also installed with
1681 the name @command{c++}.
1683 @cindex invoking @command{g++}
1684 When you compile C++ programs, you may specify many of the same
1685 command-line options that you use for compiling programs in any
1686 language; or command-line options meaningful for C and related
1687 languages; or options that are meaningful only for C++ programs.
1688 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1689 explanations of options for languages related to C@.
1690 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1691 explanations of options that are meaningful only for C++ programs.
1693 @node C Dialect Options
1694 @section Options Controlling C Dialect
1695 @cindex dialect options
1696 @cindex language dialect options
1697 @cindex options, dialect
1699 The following options control the dialect of C (or languages derived
1700 from C, such as C++, Objective-C and Objective-C++) that the compiler
1704 @cindex ANSI support
1708 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1709 equivalent to @option{-std=c++98}.
1711 This turns off certain features of GCC that are incompatible with ISO
1712 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1713 such as the @code{asm} and @code{typeof} keywords, and
1714 predefined macros such as @code{unix} and @code{vax} that identify the
1715 type of system you are using. It also enables the undesirable and
1716 rarely used ISO trigraph feature. For the C compiler,
1717 it disables recognition of C++ style @samp{//} comments as well as
1718 the @code{inline} keyword.
1720 The alternate keywords @code{__asm__}, @code{__extension__},
1721 @code{__inline__} and @code{__typeof__} continue to work despite
1722 @option{-ansi}. You would not want to use them in an ISO C program, of
1723 course, but it is useful to put them in header files that might be included
1724 in compilations done with @option{-ansi}. Alternate predefined macros
1725 such as @code{__unix__} and @code{__vax__} are also available, with or
1726 without @option{-ansi}.
1728 The @option{-ansi} option does not cause non-ISO programs to be
1729 rejected gratuitously. For that, @option{-Wpedantic} is required in
1730 addition to @option{-ansi}. @xref{Warning Options}.
1732 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1733 option is used. Some header files may notice this macro and refrain
1734 from declaring certain functions or defining certain macros that the
1735 ISO standard doesn't call for; this is to avoid interfering with any
1736 programs that might use these names for other things.
1738 Functions that are normally built in but do not have semantics
1739 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1740 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1741 built-in functions provided by GCC}, for details of the functions
1746 Determine the language standard. @xref{Standards,,Language Standards
1747 Supported by GCC}, for details of these standard versions. This option
1748 is currently only supported when compiling C or C++.
1750 The compiler can accept several base standards, such as @samp{c90} or
1751 @samp{c++98}, and GNU dialects of those standards, such as
1752 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1753 compiler accepts all programs following that standard plus those
1754 using GNU extensions that do not contradict it. For example,
1755 @option{-std=c90} turns off certain features of GCC that are
1756 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1757 keywords, but not other GNU extensions that do not have a meaning in
1758 ISO C90, such as omitting the middle term of a @code{?:}
1759 expression. On the other hand, when a GNU dialect of a standard is
1760 specified, all features supported by the compiler are enabled, even when
1761 those features change the meaning of the base standard. As a result, some
1762 strict-conforming programs may be rejected. The particular standard
1763 is used by @option{-Wpedantic} to identify which features are GNU
1764 extensions given that version of the standard. For example
1765 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1766 comments, while @option{-std=gnu99 -Wpedantic} does not.
1768 A value for this option must be provided; possible values are
1774 Support all ISO C90 programs (certain GNU extensions that conflict
1775 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1777 @item iso9899:199409
1778 ISO C90 as modified in amendment 1.
1784 ISO C99. This standard is substantially completely supported, modulo
1785 bugs and floating-point issues
1786 (mainly but not entirely relating to optional C99 features from
1787 Annexes F and G). See
1788 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1789 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1794 ISO C11, the 2011 revision of the ISO C standard. This standard is
1795 substantially completely supported, modulo bugs, floating-point issues
1796 (mainly but not entirely relating to optional C11 features from
1797 Annexes F and G) and the optional Annexes K (Bounds-checking
1798 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1802 GNU dialect of ISO C90 (including some C99 features).
1806 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1810 GNU dialect of ISO C11. This is the default for C code.
1811 The name @samp{gnu1x} is deprecated.
1815 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1816 additional defect reports. Same as @option{-ansi} for C++ code.
1820 GNU dialect of @option{-std=c++98}.
1824 The 2011 ISO C++ standard plus amendments.
1825 The name @samp{c++0x} is deprecated.
1829 GNU dialect of @option{-std=c++11}.
1830 The name @samp{gnu++0x} is deprecated.
1834 The 2014 ISO C++ standard plus amendments.
1835 The name @samp{c++1y} is deprecated.
1839 GNU dialect of @option{-std=c++14}.
1840 This is the default for C++ code.
1841 The name @samp{gnu++1y} is deprecated.
1844 The next revision of the ISO C++ standard, tentatively planned for
1845 2017. Support is highly experimental, and will almost certainly
1846 change in incompatible ways in future releases.
1849 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1850 and will almost certainly change in incompatible ways in future
1854 @item -fgnu89-inline
1855 @opindex fgnu89-inline
1856 The option @option{-fgnu89-inline} tells GCC to use the traditional
1857 GNU semantics for @code{inline} functions when in C99 mode.
1858 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1859 Using this option is roughly equivalent to adding the
1860 @code{gnu_inline} function attribute to all inline functions
1861 (@pxref{Function Attributes}).
1863 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1864 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1865 specifies the default behavior).
1866 This option is not supported in @option{-std=c90} or
1867 @option{-std=gnu90} mode.
1869 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1870 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1871 in effect for @code{inline} functions. @xref{Common Predefined
1872 Macros,,,cpp,The C Preprocessor}.
1874 @item -fpermitted-flt-eval-methods=@var{style}
1875 @opindex fpermitted-flt-eval-methods
1876 @opindex fpermitted-flt-eval-methods=c11
1877 @opindex fpermitted-flt-eval-methods=ts-18661-3
1878 ISO/IEC TS 18661-3 defines new permissible values for
1879 @code{FLT_EVAL_METHOD} that indicate that operations and constants with
1880 a semantic type that is an interchange or extended format should be
1881 evaluated to the precision and range of that type. These new values are
1882 a superset of those permitted under C99/C11, which does not specify the
1883 meaning of other positive values of @code{FLT_EVAL_METHOD}. As such, code
1884 conforming to C11 may not have been written expecting the possibility of
1887 @option{-fpermitted-flt-eval-methods} specifies whether the compiler
1888 should allow only the values of @code{FLT_EVAL_METHOD} specified in C99/C11,
1889 or the extended set of values specified in ISO/IEC TS 18661-3.
1891 @var{style} is either @code{c11} or @code{ts-18661-3} as appropriate.
1893 The default when in a standards compliant mode (@option{-std=c11} or similar)
1894 is @option{-fpermitted-flt-eval-methods=c11}. The default when in a GNU
1895 dialect (@option{-std=gnu11} or similar) is
1896 @option{-fpermitted-flt-eval-methods=ts-18661-3}.
1898 @item -aux-info @var{filename}
1900 Output to the given filename prototyped declarations for all functions
1901 declared and/or defined in a translation unit, including those in header
1902 files. This option is silently ignored in any language other than C@.
1904 Besides declarations, the file indicates, in comments, the origin of
1905 each declaration (source file and line), whether the declaration was
1906 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1907 @samp{O} for old, respectively, in the first character after the line
1908 number and the colon), and whether it came from a declaration or a
1909 definition (@samp{C} or @samp{F}, respectively, in the following
1910 character). In the case of function definitions, a K&R-style list of
1911 arguments followed by their declarations is also provided, inside
1912 comments, after the declaration.
1914 @item -fallow-parameterless-variadic-functions
1915 @opindex fallow-parameterless-variadic-functions
1916 Accept variadic functions without named parameters.
1918 Although it is possible to define such a function, this is not very
1919 useful as it is not possible to read the arguments. This is only
1920 supported for C as this construct is allowed by C++.
1924 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1925 keyword, so that code can use these words as identifiers. You can use
1926 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1927 instead. @option{-ansi} implies @option{-fno-asm}.
1929 In C++, this switch only affects the @code{typeof} keyword, since
1930 @code{asm} and @code{inline} are standard keywords. You may want to
1931 use the @option{-fno-gnu-keywords} flag instead, which has the same
1932 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1933 switch only affects the @code{asm} and @code{typeof} keywords, since
1934 @code{inline} is a standard keyword in ISO C99.
1937 @itemx -fno-builtin-@var{function}
1938 @opindex fno-builtin
1939 @cindex built-in functions
1940 Don't recognize built-in functions that do not begin with
1941 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1942 functions provided by GCC}, for details of the functions affected,
1943 including those which are not built-in functions when @option{-ansi} or
1944 @option{-std} options for strict ISO C conformance are used because they
1945 do not have an ISO standard meaning.
1947 GCC normally generates special code to handle certain built-in functions
1948 more efficiently; for instance, calls to @code{alloca} may become single
1949 instructions which adjust the stack directly, and calls to @code{memcpy}
1950 may become inline copy loops. The resulting code is often both smaller
1951 and faster, but since the function calls no longer appear as such, you
1952 cannot set a breakpoint on those calls, nor can you change the behavior
1953 of the functions by linking with a different library. In addition,
1954 when a function is recognized as a built-in function, GCC may use
1955 information about that function to warn about problems with calls to
1956 that function, or to generate more efficient code, even if the
1957 resulting code still contains calls to that function. For example,
1958 warnings are given with @option{-Wformat} for bad calls to
1959 @code{printf} when @code{printf} is built in and @code{strlen} is
1960 known not to modify global memory.
1962 With the @option{-fno-builtin-@var{function}} option
1963 only the built-in function @var{function} is
1964 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1965 function is named that is not built-in in this version of GCC, this
1966 option is ignored. There is no corresponding
1967 @option{-fbuiltin-@var{function}} option; if you wish to enable
1968 built-in functions selectively when using @option{-fno-builtin} or
1969 @option{-ffreestanding}, you may define macros such as:
1972 #define abs(n) __builtin_abs ((n))
1973 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1979 Enable parsing of function definitions marked with @code{__GIMPLE}.
1980 This is an experimental feature that allows unit testing of GIMPLE
1985 @cindex hosted environment
1987 Assert that compilation targets a hosted environment. This implies
1988 @option{-fbuiltin}. A hosted environment is one in which the
1989 entire standard library is available, and in which @code{main} has a return
1990 type of @code{int}. Examples are nearly everything except a kernel.
1991 This is equivalent to @option{-fno-freestanding}.
1993 @item -ffreestanding
1994 @opindex ffreestanding
1995 @cindex hosted environment
1997 Assert that compilation targets a freestanding environment. This
1998 implies @option{-fno-builtin}. A freestanding environment
1999 is one in which the standard library may not exist, and program startup may
2000 not necessarily be at @code{main}. The most obvious example is an OS kernel.
2001 This is equivalent to @option{-fno-hosted}.
2003 @xref{Standards,,Language Standards Supported by GCC}, for details of
2004 freestanding and hosted environments.
2008 @cindex OpenACC accelerator programming
2009 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
2010 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
2011 compiler generates accelerated code according to the OpenACC Application
2012 Programming Interface v2.0 @w{@uref{http://www.openacc.org/}}. This option
2013 implies @option{-pthread}, and thus is only supported on targets that
2014 have support for @option{-pthread}.
2016 @item -fopenacc-dim=@var{geom}
2017 @opindex fopenacc-dim
2018 @cindex OpenACC accelerator programming
2019 Specify default compute dimensions for parallel offload regions that do
2020 not explicitly specify. The @var{geom} value is a triple of
2021 ':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A size
2022 can be omitted, to use a target-specific default value.
2026 @cindex OpenMP parallel
2027 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
2028 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
2029 compiler generates parallel code according to the OpenMP Application
2030 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
2031 implies @option{-pthread}, and thus is only supported on targets that
2032 have support for @option{-pthread}. @option{-fopenmp} implies
2033 @option{-fopenmp-simd}.
2036 @opindex fopenmp-simd
2039 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
2040 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
2045 @cindex Enable Cilk Plus
2046 Enable the usage of Cilk Plus language extension features for C/C++.
2047 When the option @option{-fcilkplus} is specified, enable the usage of
2048 the Cilk Plus Language extension features for C/C++. The present
2049 implementation follows ABI version 1.2. This is an experimental
2050 feature that is only partially complete, and whose interface may
2051 change in future versions of GCC as the official specification
2052 changes. Currently, all features but @code{_Cilk_for} have been
2057 When the option @option{-fgnu-tm} is specified, the compiler
2058 generates code for the Linux variant of Intel's current Transactional
2059 Memory ABI specification document (Revision 1.1, May 6 2009). This is
2060 an experimental feature whose interface may change in future versions
2061 of GCC, as the official specification changes. Please note that not
2062 all architectures are supported for this feature.
2064 For more information on GCC's support for transactional memory,
2065 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
2066 Transactional Memory Library}.
2068 Note that the transactional memory feature is not supported with
2069 non-call exceptions (@option{-fnon-call-exceptions}).
2071 @item -fms-extensions
2072 @opindex fms-extensions
2073 Accept some non-standard constructs used in Microsoft header files.
2075 In C++ code, this allows member names in structures to be similar
2076 to previous types declarations.
2085 Some cases of unnamed fields in structures and unions are only
2086 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2087 fields within structs/unions}, for details.
2089 Note that this option is off for all targets but x86
2090 targets using ms-abi.
2092 @item -fplan9-extensions
2093 @opindex fplan9-extensions
2094 Accept some non-standard constructs used in Plan 9 code.
2096 This enables @option{-fms-extensions}, permits passing pointers to
2097 structures with anonymous fields to functions that expect pointers to
2098 elements of the type of the field, and permits referring to anonymous
2099 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2100 struct/union fields within structs/unions}, for details. This is only
2101 supported for C, not C++.
2105 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
2106 options for strict ISO C conformance) implies @option{-trigraphs}.
2108 @cindex traditional C language
2109 @cindex C language, traditional
2111 @itemx -traditional-cpp
2112 @opindex traditional-cpp
2113 @opindex traditional
2114 Formerly, these options caused GCC to attempt to emulate a pre-standard
2115 C compiler. They are now only supported with the @option{-E} switch.
2116 The preprocessor continues to support a pre-standard mode. See the GNU
2117 CPP manual for details.
2119 @item -fcond-mismatch
2120 @opindex fcond-mismatch
2121 Allow conditional expressions with mismatched types in the second and
2122 third arguments. The value of such an expression is void. This option
2123 is not supported for C++.
2125 @item -flax-vector-conversions
2126 @opindex flax-vector-conversions
2127 Allow implicit conversions between vectors with differing numbers of
2128 elements and/or incompatible element types. This option should not be
2131 @item -funsigned-char
2132 @opindex funsigned-char
2133 Let the type @code{char} be unsigned, like @code{unsigned char}.
2135 Each kind of machine has a default for what @code{char} should
2136 be. It is either like @code{unsigned char} by default or like
2137 @code{signed char} by default.
2139 Ideally, a portable program should always use @code{signed char} or
2140 @code{unsigned char} when it depends on the signedness of an object.
2141 But many programs have been written to use plain @code{char} and
2142 expect it to be signed, or expect it to be unsigned, depending on the
2143 machines they were written for. This option, and its inverse, let you
2144 make such a program work with the opposite default.
2146 The type @code{char} is always a distinct type from each of
2147 @code{signed char} or @code{unsigned char}, even though its behavior
2148 is always just like one of those two.
2151 @opindex fsigned-char
2152 Let the type @code{char} be signed, like @code{signed char}.
2154 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2155 the negative form of @option{-funsigned-char}. Likewise, the option
2156 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2158 @item -fsigned-bitfields
2159 @itemx -funsigned-bitfields
2160 @itemx -fno-signed-bitfields
2161 @itemx -fno-unsigned-bitfields
2162 @opindex fsigned-bitfields
2163 @opindex funsigned-bitfields
2164 @opindex fno-signed-bitfields
2165 @opindex fno-unsigned-bitfields
2166 These options control whether a bit-field is signed or unsigned, when the
2167 declaration does not use either @code{signed} or @code{unsigned}. By
2168 default, such a bit-field is signed, because this is consistent: the
2169 basic integer types such as @code{int} are signed types.
2171 @item -fsso-struct=@var{endianness}
2172 @opindex fsso-struct
2173 Set the default scalar storage order of structures and unions to the
2174 specified endianness. The accepted values are @samp{big-endian} and
2175 @samp{little-endian}. If the option is not passed, the compiler uses
2176 the native endianness of the target. This option is not supported for C++.
2178 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2179 code that is not binary compatible with code generated without it if the
2180 specified endianness is not the native endianness of the target.
2183 @node C++ Dialect Options
2184 @section Options Controlling C++ Dialect
2186 @cindex compiler options, C++
2187 @cindex C++ options, command-line
2188 @cindex options, C++
2189 This section describes the command-line options that are only meaningful
2190 for C++ programs. You can also use most of the GNU compiler options
2191 regardless of what language your program is in. For example, you
2192 might compile a file @file{firstClass.C} like this:
2195 g++ -g -fstrict-enums -O -c firstClass.C
2199 In this example, only @option{-fstrict-enums} is an option meant
2200 only for C++ programs; you can use the other options with any
2201 language supported by GCC@.
2203 Some options for compiling C programs, such as @option{-std}, are also
2204 relevant for C++ programs.
2205 @xref{C Dialect Options,,Options Controlling C Dialect}.
2207 Here is a list of options that are @emph{only} for compiling C++ programs:
2211 @item -fabi-version=@var{n}
2212 @opindex fabi-version
2213 Use version @var{n} of the C++ ABI@. The default is version 0.
2215 Version 0 refers to the version conforming most closely to
2216 the C++ ABI specification. Therefore, the ABI obtained using version 0
2217 will change in different versions of G++ as ABI bugs are fixed.
2219 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2221 Version 2 is the version of the C++ ABI that first appeared in G++
2222 3.4, and was the default through G++ 4.9.
2224 Version 3 corrects an error in mangling a constant address as a
2227 Version 4, which first appeared in G++ 4.5, implements a standard
2228 mangling for vector types.
2230 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2231 attribute const/volatile on function pointer types, decltype of a
2232 plain decl, and use of a function parameter in the declaration of
2235 Version 6, which first appeared in G++ 4.7, corrects the promotion
2236 behavior of C++11 scoped enums and the mangling of template argument
2237 packs, const/static_cast, prefix ++ and --, and a class scope function
2238 used as a template argument.
2240 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2241 builtin type and corrects the mangling of lambdas in default argument
2244 Version 8, which first appeared in G++ 4.9, corrects the substitution
2245 behavior of function types with function-cv-qualifiers.
2247 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2250 Version 10, which first appeared in G++ 6.1, adds mangling of
2251 attributes that affect type identity, such as ia32 calling convention
2252 attributes (e.g. @samp{stdcall}).
2254 Version 11, which first appeared in G++ 7, corrects the mangling of
2255 sizeof... expressions. It also implies
2256 @option{-fnew-inheriting-ctors}.
2258 See also @option{-Wabi}.
2260 @item -fabi-compat-version=@var{n}
2261 @opindex fabi-compat-version
2262 On targets that support strong aliases, G++
2263 works around mangling changes by creating an alias with the correct
2264 mangled name when defining a symbol with an incorrect mangled name.
2265 This switch specifies which ABI version to use for the alias.
2267 With @option{-fabi-version=0} (the default), this defaults to 8 (GCC 5
2268 compatibility). If another ABI version is explicitly selected, this
2269 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2270 use @option{-fabi-compat-version=2}.
2272 If this option is not provided but @option{-Wabi=@var{n}} is, that
2273 version is used for compatibility aliases. If this option is provided
2274 along with @option{-Wabi} (without the version), the version from this
2275 option is used for the warning.
2277 @item -fno-access-control
2278 @opindex fno-access-control
2279 Turn off all access checking. This switch is mainly useful for working
2280 around bugs in the access control code.
2283 @opindex faligned-new
2284 Enable support for C++17 @code{new} of types that require more
2285 alignment than @code{void* ::operator new(std::size_t)} provides. A
2286 numeric argument such as @code{-faligned-new=32} can be used to
2287 specify how much alignment (in bytes) is provided by that function,
2288 but few users will need to override the default of
2289 @code{alignof(std::max_align_t)}.
2293 Check that the pointer returned by @code{operator new} is non-null
2294 before attempting to modify the storage allocated. This check is
2295 normally unnecessary because the C++ standard specifies that
2296 @code{operator new} only returns @code{0} if it is declared
2297 @code{throw()}, in which case the compiler always checks the
2298 return value even without this option. In all other cases, when
2299 @code{operator new} has a non-empty exception specification, memory
2300 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2301 @samp{new (nothrow)}.
2305 Enable support for the C++ Extensions for Concepts Technical
2306 Specification, ISO 19217 (2015), which allows code like
2309 template <class T> concept bool Addable = requires (T t) @{ t + t; @};
2310 template <Addable T> T add (T a, T b) @{ return a + b; @}
2313 @item -fconstexpr-depth=@var{n}
2314 @opindex fconstexpr-depth
2315 Set the maximum nested evaluation depth for C++11 constexpr functions
2316 to @var{n}. A limit is needed to detect endless recursion during
2317 constant expression evaluation. The minimum specified by the standard
2320 @item -fconstexpr-loop-limit=@var{n}
2321 @opindex fconstexpr-loop-limit
2322 Set the maximum number of iterations for a loop in C++14 constexpr functions
2323 to @var{n}. A limit is needed to detect infinite loops during
2324 constant expression evaluation. The default is 262144 (1<<18).
2326 @item -fdeduce-init-list
2327 @opindex fdeduce-init-list
2328 Enable deduction of a template type parameter as
2329 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2332 template <class T> auto forward(T t) -> decltype (realfn (t))
2339 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2343 This deduction was implemented as a possible extension to the
2344 originally proposed semantics for the C++11 standard, but was not part
2345 of the final standard, so it is disabled by default. This option is
2346 deprecated, and may be removed in a future version of G++.
2348 @item -ffriend-injection
2349 @opindex ffriend-injection
2350 Inject friend functions into the enclosing namespace, so that they are
2351 visible outside the scope of the class in which they are declared.
2352 Friend functions were documented to work this way in the old Annotated
2353 C++ Reference Manual.
2354 However, in ISO C++ a friend function that is not declared
2355 in an enclosing scope can only be found using argument dependent
2356 lookup. GCC defaults to the standard behavior.
2358 This option is for compatibility, and may be removed in a future
2361 @item -fno-elide-constructors
2362 @opindex fno-elide-constructors
2363 The C++ standard allows an implementation to omit creating a temporary
2364 that is only used to initialize another object of the same type.
2365 Specifying this option disables that optimization, and forces G++ to
2366 call the copy constructor in all cases. This option also causes G++
2367 to call trivial member functions which otherwise would be expanded inline.
2369 In C++17, the compiler is required to omit these temporaries, but this
2370 option still affects trivial member functions.
2372 @item -fno-enforce-eh-specs
2373 @opindex fno-enforce-eh-specs
2374 Don't generate code to check for violation of exception specifications
2375 at run time. This option violates the C++ standard, but may be useful
2376 for reducing code size in production builds, much like defining
2377 @code{NDEBUG}. This does not give user code permission to throw
2378 exceptions in violation of the exception specifications; the compiler
2379 still optimizes based on the specifications, so throwing an
2380 unexpected exception results in undefined behavior at run time.
2382 @item -fextern-tls-init
2383 @itemx -fno-extern-tls-init
2384 @opindex fextern-tls-init
2385 @opindex fno-extern-tls-init
2386 The C++11 and OpenMP standards allow @code{thread_local} and
2387 @code{threadprivate} variables to have dynamic (runtime)
2388 initialization. To support this, any use of such a variable goes
2389 through a wrapper function that performs any necessary initialization.
2390 When the use and definition of the variable are in the same
2391 translation unit, this overhead can be optimized away, but when the
2392 use is in a different translation unit there is significant overhead
2393 even if the variable doesn't actually need dynamic initialization. If
2394 the programmer can be sure that no use of the variable in a
2395 non-defining TU needs to trigger dynamic initialization (either
2396 because the variable is statically initialized, or a use of the
2397 variable in the defining TU will be executed before any uses in
2398 another TU), they can avoid this overhead with the
2399 @option{-fno-extern-tls-init} option.
2401 On targets that support symbol aliases, the default is
2402 @option{-fextern-tls-init}. On targets that do not support symbol
2403 aliases, the default is @option{-fno-extern-tls-init}.
2406 @itemx -fno-for-scope
2408 @opindex fno-for-scope
2409 If @option{-ffor-scope} is specified, the scope of variables declared in
2410 a @i{for-init-statement} is limited to the @code{for} loop itself,
2411 as specified by the C++ standard.
2412 If @option{-fno-for-scope} is specified, the scope of variables declared in
2413 a @i{for-init-statement} extends to the end of the enclosing scope,
2414 as was the case in old versions of G++, and other (traditional)
2415 implementations of C++.
2417 If neither flag is given, the default is to follow the standard,
2418 but to allow and give a warning for old-style code that would
2419 otherwise be invalid, or have different behavior.
2421 @item -fno-gnu-keywords
2422 @opindex fno-gnu-keywords
2423 Do not recognize @code{typeof} as a keyword, so that code can use this
2424 word as an identifier. You can use the keyword @code{__typeof__} instead.
2425 This option is implied by the strict ISO C++ dialects: @option{-ansi},
2426 @option{-std=c++98}, @option{-std=c++11}, etc.
2428 @item -fno-implicit-templates
2429 @opindex fno-implicit-templates
2430 Never emit code for non-inline templates that are instantiated
2431 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2432 @xref{Template Instantiation}, for more information.
2434 @item -fno-implicit-inline-templates
2435 @opindex fno-implicit-inline-templates
2436 Don't emit code for implicit instantiations of inline templates, either.
2437 The default is to handle inlines differently so that compiles with and
2438 without optimization need the same set of explicit instantiations.
2440 @item -fno-implement-inlines
2441 @opindex fno-implement-inlines
2442 To save space, do not emit out-of-line copies of inline functions
2443 controlled by @code{#pragma implementation}. This causes linker
2444 errors if these functions are not inlined everywhere they are called.
2446 @item -fms-extensions
2447 @opindex fms-extensions
2448 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2449 int and getting a pointer to member function via non-standard syntax.
2451 @item -fnew-inheriting-ctors
2452 @opindex fnew-inheriting-ctors
2453 Enable the P0136 adjustment to the semantics of C++11 constructor
2454 inheritance. This is part of C++17 but also considered to be a Defect
2455 Report against C++11 and C++14. This flag is enabled by default
2456 unless @option{-fabi-version=10} or lower is specified.
2458 @item -fno-nonansi-builtins
2459 @opindex fno-nonansi-builtins
2460 Disable built-in declarations of functions that are not mandated by
2461 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2462 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2465 @opindex fnothrow-opt
2466 Treat a @code{throw()} exception specification as if it were a
2467 @code{noexcept} specification to reduce or eliminate the text size
2468 overhead relative to a function with no exception specification. If
2469 the function has local variables of types with non-trivial
2470 destructors, the exception specification actually makes the
2471 function smaller because the EH cleanups for those variables can be
2472 optimized away. The semantic effect is that an exception thrown out of
2473 a function with such an exception specification results in a call
2474 to @code{terminate} rather than @code{unexpected}.
2476 @item -fno-operator-names
2477 @opindex fno-operator-names
2478 Do not treat the operator name keywords @code{and}, @code{bitand},
2479 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2480 synonyms as keywords.
2482 @item -fno-optional-diags
2483 @opindex fno-optional-diags
2484 Disable diagnostics that the standard says a compiler does not need to
2485 issue. Currently, the only such diagnostic issued by G++ is the one for
2486 a name having multiple meanings within a class.
2489 @opindex fpermissive
2490 Downgrade some diagnostics about nonconformant code from errors to
2491 warnings. Thus, using @option{-fpermissive} allows some
2492 nonconforming code to compile.
2494 @item -fno-pretty-templates
2495 @opindex fno-pretty-templates
2496 When an error message refers to a specialization of a function
2497 template, the compiler normally prints the signature of the
2498 template followed by the template arguments and any typedefs or
2499 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2500 rather than @code{void f(int)}) so that it's clear which template is
2501 involved. When an error message refers to a specialization of a class
2502 template, the compiler omits any template arguments that match
2503 the default template arguments for that template. If either of these
2504 behaviors make it harder to understand the error message rather than
2505 easier, you can use @option{-fno-pretty-templates} to disable them.
2509 Enable automatic template instantiation at link time. This option also
2510 implies @option{-fno-implicit-templates}. @xref{Template
2511 Instantiation}, for more information.
2515 Disable generation of information about every class with virtual
2516 functions for use by the C++ run-time type identification features
2517 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2518 of the language, you can save some space by using this flag. Note that
2519 exception handling uses the same information, but G++ generates it as
2520 needed. The @code{dynamic_cast} operator can still be used for casts that
2521 do not require run-time type information, i.e.@: casts to @code{void *} or to
2522 unambiguous base classes.
2524 @item -fsized-deallocation
2525 @opindex fsized-deallocation
2526 Enable the built-in global declarations
2528 void operator delete (void *, std::size_t) noexcept;
2529 void operator delete[] (void *, std::size_t) noexcept;
2531 as introduced in C++14. This is useful for user-defined replacement
2532 deallocation functions that, for example, use the size of the object
2533 to make deallocation faster. Enabled by default under
2534 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2535 warns about places that might want to add a definition.
2537 @item -fstrict-enums
2538 @opindex fstrict-enums
2539 Allow the compiler to optimize using the assumption that a value of
2540 enumerated type can only be one of the values of the enumeration (as
2541 defined in the C++ standard; basically, a value that can be
2542 represented in the minimum number of bits needed to represent all the
2543 enumerators). This assumption may not be valid if the program uses a
2544 cast to convert an arbitrary integer value to the enumerated type.
2546 @item -fstrong-eval-order
2547 @opindex fstrong-eval-order
2548 Evaluate member access, array subscripting, and shift expressions in
2549 left-to-right order, and evaluate assignment in right-to-left order,
2550 as adopted for C++17. Enabled by default with @option{-std=c++1z}.
2551 @option{-fstrong-eval-order=some} enables just the ordering of member
2552 access and shift expressions, and is the default without
2553 @option{-std=c++1z}.
2555 @item -ftemplate-backtrace-limit=@var{n}
2556 @opindex ftemplate-backtrace-limit
2557 Set the maximum number of template instantiation notes for a single
2558 warning or error to @var{n}. The default value is 10.
2560 @item -ftemplate-depth=@var{n}
2561 @opindex ftemplate-depth
2562 Set the maximum instantiation depth for template classes to @var{n}.
2563 A limit on the template instantiation depth is needed to detect
2564 endless recursions during template class instantiation. ANSI/ISO C++
2565 conforming programs must not rely on a maximum depth greater than 17
2566 (changed to 1024 in C++11). The default value is 900, as the compiler
2567 can run out of stack space before hitting 1024 in some situations.
2569 @item -fno-threadsafe-statics
2570 @opindex fno-threadsafe-statics
2571 Do not emit the extra code to use the routines specified in the C++
2572 ABI for thread-safe initialization of local statics. You can use this
2573 option to reduce code size slightly in code that doesn't need to be
2576 @item -fuse-cxa-atexit
2577 @opindex fuse-cxa-atexit
2578 Register destructors for objects with static storage duration with the
2579 @code{__cxa_atexit} function rather than the @code{atexit} function.
2580 This option is required for fully standards-compliant handling of static
2581 destructors, but only works if your C library supports
2582 @code{__cxa_atexit}.
2584 @item -fno-use-cxa-get-exception-ptr
2585 @opindex fno-use-cxa-get-exception-ptr
2586 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2587 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2588 if the runtime routine is not available.
2590 @item -fvisibility-inlines-hidden
2591 @opindex fvisibility-inlines-hidden
2592 This switch declares that the user does not attempt to compare
2593 pointers to inline functions or methods where the addresses of the two functions
2594 are taken in different shared objects.
2596 The effect of this is that GCC may, effectively, mark inline methods with
2597 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2598 appear in the export table of a DSO and do not require a PLT indirection
2599 when used within the DSO@. Enabling this option can have a dramatic effect
2600 on load and link times of a DSO as it massively reduces the size of the
2601 dynamic export table when the library makes heavy use of templates.
2603 The behavior of this switch is not quite the same as marking the
2604 methods as hidden directly, because it does not affect static variables
2605 local to the function or cause the compiler to deduce that
2606 the function is defined in only one shared object.
2608 You may mark a method as having a visibility explicitly to negate the
2609 effect of the switch for that method. For example, if you do want to
2610 compare pointers to a particular inline method, you might mark it as
2611 having default visibility. Marking the enclosing class with explicit
2612 visibility has no effect.
2614 Explicitly instantiated inline methods are unaffected by this option
2615 as their linkage might otherwise cross a shared library boundary.
2616 @xref{Template Instantiation}.
2618 @item -fvisibility-ms-compat
2619 @opindex fvisibility-ms-compat
2620 This flag attempts to use visibility settings to make GCC's C++
2621 linkage model compatible with that of Microsoft Visual Studio.
2623 The flag makes these changes to GCC's linkage model:
2627 It sets the default visibility to @code{hidden}, like
2628 @option{-fvisibility=hidden}.
2631 Types, but not their members, are not hidden by default.
2634 The One Definition Rule is relaxed for types without explicit
2635 visibility specifications that are defined in more than one
2636 shared object: those declarations are permitted if they are
2637 permitted when this option is not used.
2640 In new code it is better to use @option{-fvisibility=hidden} and
2641 export those classes that are intended to be externally visible.
2642 Unfortunately it is possible for code to rely, perhaps accidentally,
2643 on the Visual Studio behavior.
2645 Among the consequences of these changes are that static data members
2646 of the same type with the same name but defined in different shared
2647 objects are different, so changing one does not change the other;
2648 and that pointers to function members defined in different shared
2649 objects may not compare equal. When this flag is given, it is a
2650 violation of the ODR to define types with the same name differently.
2654 Do not use weak symbol support, even if it is provided by the linker.
2655 By default, G++ uses weak symbols if they are available. This
2656 option exists only for testing, and should not be used by end-users;
2657 it results in inferior code and has no benefits. This option may
2658 be removed in a future release of G++.
2662 Do not search for header files in the standard directories specific to
2663 C++, but do still search the other standard directories. (This option
2664 is used when building the C++ library.)
2667 In addition, these optimization, warning, and code generation options
2668 have meanings only for C++ programs:
2671 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2674 Warn when G++ it generates code that is probably not compatible with
2675 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2676 ABI with each major release, normally @option{-Wabi} will warn only if
2677 there is a check added later in a release series for an ABI issue
2678 discovered since the initial release. @option{-Wabi} will warn about
2679 more things if an older ABI version is selected (with
2680 @option{-fabi-version=@var{n}}).
2682 @option{-Wabi} can also be used with an explicit version number to
2683 warn about compatibility with a particular @option{-fabi-version}
2684 level, e.g. @option{-Wabi=2} to warn about changes relative to
2685 @option{-fabi-version=2}.
2687 If an explicit version number is provided and
2688 @option{-fabi-compat-version} is not specified, the version number
2689 from this option is used for compatibility aliases. If no explicit
2690 version number is provided with this option, but
2691 @option{-fabi-compat-version} is specified, that version number is
2692 used for ABI warnings.
2694 Although an effort has been made to warn about
2695 all such cases, there are probably some cases that are not warned about,
2696 even though G++ is generating incompatible code. There may also be
2697 cases where warnings are emitted even though the code that is generated
2700 You should rewrite your code to avoid these warnings if you are
2701 concerned about the fact that code generated by G++ may not be binary
2702 compatible with code generated by other compilers.
2704 Known incompatibilities in @option{-fabi-version=2} (which was the
2705 default from GCC 3.4 to 4.9) include:
2710 A template with a non-type template parameter of reference type was
2711 mangled incorrectly:
2714 template <int &> struct S @{@};
2718 This was fixed in @option{-fabi-version=3}.
2721 SIMD vector types declared using @code{__attribute ((vector_size))} were
2722 mangled in a non-standard way that does not allow for overloading of
2723 functions taking vectors of different sizes.
2725 The mangling was changed in @option{-fabi-version=4}.
2728 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2729 qualifiers, and @code{decltype} of a plain declaration was folded away.
2731 These mangling issues were fixed in @option{-fabi-version=5}.
2734 Scoped enumerators passed as arguments to a variadic function are
2735 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2736 On most targets this does not actually affect the parameter passing
2737 ABI, as there is no way to pass an argument smaller than @code{int}.
2739 Also, the ABI changed the mangling of template argument packs,
2740 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2741 a class scope function used as a template argument.
2743 These issues were corrected in @option{-fabi-version=6}.
2746 Lambdas in default argument scope were mangled incorrectly, and the
2747 ABI changed the mangling of @code{nullptr_t}.
2749 These issues were corrected in @option{-fabi-version=7}.
2752 When mangling a function type with function-cv-qualifiers, the
2753 un-qualified function type was incorrectly treated as a substitution
2756 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2759 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2760 unaligned accesses. Note that this did not affect the ABI of a
2761 function with a @code{nullptr_t} parameter, as parameters have a
2764 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2767 Target-specific attributes that affect the identity of a type, such as
2768 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2769 did not affect the mangled name, leading to name collisions when
2770 function pointers were used as template arguments.
2772 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2776 It also warns about psABI-related changes. The known psABI changes at this
2782 For SysV/x86-64, unions with @code{long double} members are
2783 passed in memory as specified in psABI. For example:
2793 @code{union U} is always passed in memory.
2797 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2800 Warn when a type with an ABI tag is used in a context that does not
2801 have that ABI tag. See @ref{C++ Attributes} for more information
2804 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2805 @opindex Wctor-dtor-privacy
2806 @opindex Wno-ctor-dtor-privacy
2807 Warn when a class seems unusable because all the constructors or
2808 destructors in that class are private, and it has neither friends nor
2809 public static member functions. Also warn if there are no non-private
2810 methods, and there's at least one private member function that isn't
2811 a constructor or destructor.
2813 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2814 @opindex Wdelete-non-virtual-dtor
2815 @opindex Wno-delete-non-virtual-dtor
2816 Warn when @code{delete} is used to destroy an instance of a class that
2817 has virtual functions and non-virtual destructor. It is unsafe to delete
2818 an instance of a derived class through a pointer to a base class if the
2819 base class does not have a virtual destructor. This warning is enabled
2822 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2823 @opindex Wliteral-suffix
2824 @opindex Wno-literal-suffix
2825 Warn when a string or character literal is followed by a ud-suffix which does
2826 not begin with an underscore. As a conforming extension, GCC treats such
2827 suffixes as separate preprocessing tokens in order to maintain backwards
2828 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2832 #define __STDC_FORMAT_MACROS
2833 #include <inttypes.h>
2838 printf("My int64: %" PRId64"\n", i64);
2842 In this case, @code{PRId64} is treated as a separate preprocessing token.
2844 This warning is enabled by default.
2846 @item -Wlto-type-mismatch
2847 @opindex Wlto-type-mismatch
2848 @opindex Wno-lto-type-mismatch
2850 During the link-time optimization warn about type mismatches in
2851 global declarations from different compilation units.
2852 Requires @option{-flto} to be enabled. Enabled by default.
2854 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2856 @opindex Wno-narrowing
2857 With @option{-std=gnu++98} or @option{-std=c++98}, warn when a narrowing
2858 conversion prohibited by C++11 occurs within
2862 int i = @{ 2.2 @}; // error: narrowing from double to int
2865 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2867 When a later standard is in effect, e.g. when using @option{-std=c++11},
2868 narrowing conversions are diagnosed by default, as required by the standard.
2869 A narrowing conversion from a constant produces an error,
2870 and a narrowing conversion from a non-constant produces a warning,
2871 but @option{-Wno-narrowing} suppresses the diagnostic.
2872 Note that this does not affect the meaning of well-formed code;
2873 narrowing conversions are still considered ill-formed in SFINAE contexts.
2875 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2877 @opindex Wno-noexcept
2878 Warn when a noexcept-expression evaluates to false because of a call
2879 to a function that does not have a non-throwing exception
2880 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2881 the compiler to never throw an exception.
2883 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2884 @opindex Wnon-virtual-dtor
2885 @opindex Wno-non-virtual-dtor
2886 Warn when a class has virtual functions and an accessible non-virtual
2887 destructor itself or in an accessible polymorphic base class, in which
2888 case it is possible but unsafe to delete an instance of a derived
2889 class through a pointer to the class itself or base class. This
2890 warning is automatically enabled if @option{-Weffc++} is specified.
2892 @item -Wregister @r{(C++ and Objective-C++ only)}
2894 @opindex Wno-register
2895 Warn on uses of the @code{register} storage class specifier, except
2896 when it is part of the GNU @ref{Explicit Register Variables} extension.
2897 The use of the @code{register} keyword as storage class specifier has
2898 been deprecated in C++11 and removed in C++17.
2899 Enabled by default with @option{-std=c++1z}.
2901 @item -Wreorder @r{(C++ and Objective-C++ only)}
2903 @opindex Wno-reorder
2904 @cindex reordering, warning
2905 @cindex warning for reordering of member initializers
2906 Warn when the order of member initializers given in the code does not
2907 match the order in which they must be executed. For instance:
2913 A(): j (0), i (1) @{ @}
2918 The compiler rearranges the member initializers for @code{i}
2919 and @code{j} to match the declaration order of the members, emitting
2920 a warning to that effect. This warning is enabled by @option{-Wall}.
2922 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2923 @opindex fext-numeric-literals
2924 @opindex fno-ext-numeric-literals
2925 Accept imaginary, fixed-point, or machine-defined
2926 literal number suffixes as GNU extensions.
2927 When this option is turned off these suffixes are treated
2928 as C++11 user-defined literal numeric suffixes.
2929 This is on by default for all pre-C++11 dialects and all GNU dialects:
2930 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2931 @option{-std=gnu++14}.
2932 This option is off by default
2933 for ISO C++11 onwards (@option{-std=c++11}, ...).
2936 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2939 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2942 Warn about violations of the following style guidelines from Scott Meyers'
2943 @cite{Effective C++} series of books:
2947 Define a copy constructor and an assignment operator for classes
2948 with dynamically-allocated memory.
2951 Prefer initialization to assignment in constructors.
2954 Have @code{operator=} return a reference to @code{*this}.
2957 Don't try to return a reference when you must return an object.
2960 Distinguish between prefix and postfix forms of increment and
2961 decrement operators.
2964 Never overload @code{&&}, @code{||}, or @code{,}.
2968 This option also enables @option{-Wnon-virtual-dtor}, which is also
2969 one of the effective C++ recommendations. However, the check is
2970 extended to warn about the lack of virtual destructor in accessible
2971 non-polymorphic bases classes too.
2973 When selecting this option, be aware that the standard library
2974 headers do not obey all of these guidelines; use @samp{grep -v}
2975 to filter out those warnings.
2977 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2978 @opindex Wstrict-null-sentinel
2979 @opindex Wno-strict-null-sentinel
2980 Warn about the use of an uncasted @code{NULL} as sentinel. When
2981 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2982 to @code{__null}. Although it is a null pointer constant rather than a
2983 null pointer, it is guaranteed to be of the same size as a pointer.
2984 But this use is not portable across different compilers.
2986 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2987 @opindex Wno-non-template-friend
2988 @opindex Wnon-template-friend
2989 Disable warnings when non-templatized friend functions are declared
2990 within a template. Since the advent of explicit template specification
2991 support in G++, if the name of the friend is an unqualified-id (i.e.,
2992 @samp{friend foo(int)}), the C++ language specification demands that the
2993 friend declare or define an ordinary, nontemplate function. (Section
2994 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2995 could be interpreted as a particular specialization of a templatized
2996 function. Because this non-conforming behavior is no longer the default
2997 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2998 check existing code for potential trouble spots and is on by default.
2999 This new compiler behavior can be turned off with
3000 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
3001 but disables the helpful warning.
3003 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
3004 @opindex Wold-style-cast
3005 @opindex Wno-old-style-cast
3006 Warn if an old-style (C-style) cast to a non-void type is used within
3007 a C++ program. The new-style casts (@code{dynamic_cast},
3008 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
3009 less vulnerable to unintended effects and much easier to search for.
3011 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
3012 @opindex Woverloaded-virtual
3013 @opindex Wno-overloaded-virtual
3014 @cindex overloaded virtual function, warning
3015 @cindex warning for overloaded virtual function
3016 Warn when a function declaration hides virtual functions from a
3017 base class. For example, in:
3024 struct B: public A @{
3029 the @code{A} class version of @code{f} is hidden in @code{B}, and code
3040 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
3041 @opindex Wno-pmf-conversions
3042 @opindex Wpmf-conversions
3043 Disable the diagnostic for converting a bound pointer to member function
3046 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
3047 @opindex Wsign-promo
3048 @opindex Wno-sign-promo
3049 Warn when overload resolution chooses a promotion from unsigned or
3050 enumerated type to a signed type, over a conversion to an unsigned type of
3051 the same size. Previous versions of G++ tried to preserve
3052 unsignedness, but the standard mandates the current behavior.
3054 @item -Wtemplates @r{(C++ and Objective-C++ only)}
3056 Warn when a primary template declaration is encountered. Some coding
3057 rules disallow templates, and this may be used to enforce that rule.
3058 The warning is inactive inside a system header file, such as the STL, so
3059 one can still use the STL. One may also instantiate or specialize
3062 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
3063 @opindex Wmultiple-inheritance
3064 Warn when a class is defined with multiple direct base classes. Some
3065 coding rules disallow multiple inheritance, and this may be used to
3066 enforce that rule. The warning is inactive inside a system header file,
3067 such as the STL, so one can still use the STL. One may also define
3068 classes that indirectly use multiple inheritance.
3070 @item -Wvirtual-inheritance
3071 @opindex Wvirtual-inheritance
3072 Warn when a class is defined with a virtual direct base classe. Some
3073 coding rules disallow multiple inheritance, and this may be used to
3074 enforce that rule. The warning is inactive inside a system header file,
3075 such as the STL, so one can still use the STL. One may also define
3076 classes that indirectly use virtual inheritance.
3079 @opindex Wnamespaces
3080 Warn when a namespace definition is opened. Some coding rules disallow
3081 namespaces, and this may be used to enforce that rule. The warning is
3082 inactive inside a system header file, such as the STL, so one can still
3083 use the STL. One may also use using directives and qualified names.
3085 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
3087 @opindex Wno-terminate
3088 Disable the warning about a throw-expression that will immediately
3089 result in a call to @code{terminate}.
3092 @node Objective-C and Objective-C++ Dialect Options
3093 @section Options Controlling Objective-C and Objective-C++ Dialects
3095 @cindex compiler options, Objective-C and Objective-C++
3096 @cindex Objective-C and Objective-C++ options, command-line
3097 @cindex options, Objective-C and Objective-C++
3098 (NOTE: This manual does not describe the Objective-C and Objective-C++
3099 languages themselves. @xref{Standards,,Language Standards
3100 Supported by GCC}, for references.)
3102 This section describes the command-line options that are only meaningful
3103 for Objective-C and Objective-C++ programs. You can also use most of
3104 the language-independent GNU compiler options.
3105 For example, you might compile a file @file{some_class.m} like this:
3108 gcc -g -fgnu-runtime -O -c some_class.m
3112 In this example, @option{-fgnu-runtime} is an option meant only for
3113 Objective-C and Objective-C++ programs; you can use the other options with
3114 any language supported by GCC@.
3116 Note that since Objective-C is an extension of the C language, Objective-C
3117 compilations may also use options specific to the C front-end (e.g.,
3118 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3119 C++-specific options (e.g., @option{-Wabi}).
3121 Here is a list of options that are @emph{only} for compiling Objective-C
3122 and Objective-C++ programs:
3125 @item -fconstant-string-class=@var{class-name}
3126 @opindex fconstant-string-class
3127 Use @var{class-name} as the name of the class to instantiate for each
3128 literal string specified with the syntax @code{@@"@dots{}"}. The default
3129 class name is @code{NXConstantString} if the GNU runtime is being used, and
3130 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3131 @option{-fconstant-cfstrings} option, if also present, overrides the
3132 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3133 to be laid out as constant CoreFoundation strings.
3136 @opindex fgnu-runtime
3137 Generate object code compatible with the standard GNU Objective-C
3138 runtime. This is the default for most types of systems.
3140 @item -fnext-runtime
3141 @opindex fnext-runtime
3142 Generate output compatible with the NeXT runtime. This is the default
3143 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3144 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3147 @item -fno-nil-receivers
3148 @opindex fno-nil-receivers
3149 Assume that all Objective-C message dispatches (@code{[receiver
3150 message:arg]}) in this translation unit ensure that the receiver is
3151 not @code{nil}. This allows for more efficient entry points in the
3152 runtime to be used. This option is only available in conjunction with
3153 the NeXT runtime and ABI version 0 or 1.
3155 @item -fobjc-abi-version=@var{n}
3156 @opindex fobjc-abi-version
3157 Use version @var{n} of the Objective-C ABI for the selected runtime.
3158 This option is currently supported only for the NeXT runtime. In that
3159 case, Version 0 is the traditional (32-bit) ABI without support for
3160 properties and other Objective-C 2.0 additions. Version 1 is the
3161 traditional (32-bit) ABI with support for properties and other
3162 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3163 nothing is specified, the default is Version 0 on 32-bit target
3164 machines, and Version 2 on 64-bit target machines.
3166 @item -fobjc-call-cxx-cdtors
3167 @opindex fobjc-call-cxx-cdtors
3168 For each Objective-C class, check if any of its instance variables is a
3169 C++ object with a non-trivial default constructor. If so, synthesize a
3170 special @code{- (id) .cxx_construct} instance method which runs
3171 non-trivial default constructors on any such instance variables, in order,
3172 and then return @code{self}. Similarly, check if any instance variable
3173 is a C++ object with a non-trivial destructor, and if so, synthesize a
3174 special @code{- (void) .cxx_destruct} method which runs
3175 all such default destructors, in reverse order.
3177 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3178 methods thusly generated only operate on instance variables
3179 declared in the current Objective-C class, and not those inherited
3180 from superclasses. It is the responsibility of the Objective-C
3181 runtime to invoke all such methods in an object's inheritance
3182 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3183 by the runtime immediately after a new object instance is allocated;
3184 the @code{- (void) .cxx_destruct} methods are invoked immediately
3185 before the runtime deallocates an object instance.
3187 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3188 support for invoking the @code{- (id) .cxx_construct} and
3189 @code{- (void) .cxx_destruct} methods.
3191 @item -fobjc-direct-dispatch
3192 @opindex fobjc-direct-dispatch
3193 Allow fast jumps to the message dispatcher. On Darwin this is
3194 accomplished via the comm page.
3196 @item -fobjc-exceptions
3197 @opindex fobjc-exceptions
3198 Enable syntactic support for structured exception handling in
3199 Objective-C, similar to what is offered by C++. This option
3200 is required to use the Objective-C keywords @code{@@try},
3201 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3202 @code{@@synchronized}. This option is available with both the GNU
3203 runtime and the NeXT runtime (but not available in conjunction with
3204 the NeXT runtime on Mac OS X 10.2 and earlier).
3208 Enable garbage collection (GC) in Objective-C and Objective-C++
3209 programs. This option is only available with the NeXT runtime; the
3210 GNU runtime has a different garbage collection implementation that
3211 does not require special compiler flags.
3213 @item -fobjc-nilcheck
3214 @opindex fobjc-nilcheck
3215 For the NeXT runtime with version 2 of the ABI, check for a nil
3216 receiver in method invocations before doing the actual method call.
3217 This is the default and can be disabled using
3218 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3219 checked for nil in this way no matter what this flag is set to.
3220 Currently this flag does nothing when the GNU runtime, or an older
3221 version of the NeXT runtime ABI, is used.
3223 @item -fobjc-std=objc1
3225 Conform to the language syntax of Objective-C 1.0, the language
3226 recognized by GCC 4.0. This only affects the Objective-C additions to
3227 the C/C++ language; it does not affect conformance to C/C++ standards,
3228 which is controlled by the separate C/C++ dialect option flags. When
3229 this option is used with the Objective-C or Objective-C++ compiler,
3230 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3231 This is useful if you need to make sure that your Objective-C code can
3232 be compiled with older versions of GCC@.
3234 @item -freplace-objc-classes
3235 @opindex freplace-objc-classes
3236 Emit a special marker instructing @command{ld(1)} not to statically link in
3237 the resulting object file, and allow @command{dyld(1)} to load it in at
3238 run time instead. This is used in conjunction with the Fix-and-Continue
3239 debugging mode, where the object file in question may be recompiled and
3240 dynamically reloaded in the course of program execution, without the need
3241 to restart the program itself. Currently, Fix-and-Continue functionality
3242 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3247 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3248 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3249 compile time) with static class references that get initialized at load time,
3250 which improves run-time performance. Specifying the @option{-fzero-link} flag
3251 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3252 to be retained. This is useful in Zero-Link debugging mode, since it allows
3253 for individual class implementations to be modified during program execution.
3254 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3255 regardless of command-line options.
3257 @item -fno-local-ivars
3258 @opindex fno-local-ivars
3259 @opindex flocal-ivars
3260 By default instance variables in Objective-C can be accessed as if
3261 they were local variables from within the methods of the class they're
3262 declared in. This can lead to shadowing between instance variables
3263 and other variables declared either locally inside a class method or
3264 globally with the same name. Specifying the @option{-fno-local-ivars}
3265 flag disables this behavior thus avoiding variable shadowing issues.
3267 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3268 @opindex fivar-visibility
3269 Set the default instance variable visibility to the specified option
3270 so that instance variables declared outside the scope of any access
3271 modifier directives default to the specified visibility.
3275 Dump interface declarations for all classes seen in the source file to a
3276 file named @file{@var{sourcename}.decl}.
3278 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3279 @opindex Wassign-intercept
3280 @opindex Wno-assign-intercept
3281 Warn whenever an Objective-C assignment is being intercepted by the
3284 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3285 @opindex Wno-protocol
3287 If a class is declared to implement a protocol, a warning is issued for
3288 every method in the protocol that is not implemented by the class. The
3289 default behavior is to issue a warning for every method not explicitly
3290 implemented in the class, even if a method implementation is inherited
3291 from the superclass. If you use the @option{-Wno-protocol} option, then
3292 methods inherited from the superclass are considered to be implemented,
3293 and no warning is issued for them.
3295 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3297 @opindex Wno-selector
3298 Warn if multiple methods of different types for the same selector are
3299 found during compilation. The check is performed on the list of methods
3300 in the final stage of compilation. Additionally, a check is performed
3301 for each selector appearing in a @code{@@selector(@dots{})}
3302 expression, and a corresponding method for that selector has been found
3303 during compilation. Because these checks scan the method table only at
3304 the end of compilation, these warnings are not produced if the final
3305 stage of compilation is not reached, for example because an error is
3306 found during compilation, or because the @option{-fsyntax-only} option is
3309 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3310 @opindex Wstrict-selector-match
3311 @opindex Wno-strict-selector-match
3312 Warn if multiple methods with differing argument and/or return types are
3313 found for a given selector when attempting to send a message using this
3314 selector to a receiver of type @code{id} or @code{Class}. When this flag
3315 is off (which is the default behavior), the compiler omits such warnings
3316 if any differences found are confined to types that share the same size
3319 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3320 @opindex Wundeclared-selector
3321 @opindex Wno-undeclared-selector
3322 Warn if a @code{@@selector(@dots{})} expression referring to an
3323 undeclared selector is found. A selector is considered undeclared if no
3324 method with that name has been declared before the
3325 @code{@@selector(@dots{})} expression, either explicitly in an
3326 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3327 an @code{@@implementation} section. This option always performs its
3328 checks as soon as a @code{@@selector(@dots{})} expression is found,
3329 while @option{-Wselector} only performs its checks in the final stage of
3330 compilation. This also enforces the coding style convention
3331 that methods and selectors must be declared before being used.
3333 @item -print-objc-runtime-info
3334 @opindex print-objc-runtime-info
3335 Generate C header describing the largest structure that is passed by
3340 @node Diagnostic Message Formatting Options
3341 @section Options to Control Diagnostic Messages Formatting
3342 @cindex options to control diagnostics formatting
3343 @cindex diagnostic messages
3344 @cindex message formatting
3346 Traditionally, diagnostic messages have been formatted irrespective of
3347 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3348 options described below
3349 to control the formatting algorithm for diagnostic messages,
3350 e.g.@: how many characters per line, how often source location
3351 information should be reported. Note that some language front ends may not
3352 honor these options.
3355 @item -fmessage-length=@var{n}
3356 @opindex fmessage-length
3357 Try to format error messages so that they fit on lines of about
3358 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3359 done; each error message appears on a single line. This is the
3360 default for all front ends.
3362 @item -fdiagnostics-show-location=once
3363 @opindex fdiagnostics-show-location
3364 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3365 reporter to emit source location information @emph{once}; that is, in
3366 case the message is too long to fit on a single physical line and has to
3367 be wrapped, the source location won't be emitted (as prefix) again,
3368 over and over, in subsequent continuation lines. This is the default
3371 @item -fdiagnostics-show-location=every-line
3372 Only meaningful in line-wrapping mode. Instructs the diagnostic
3373 messages reporter to emit the same source location information (as
3374 prefix) for physical lines that result from the process of breaking
3375 a message which is too long to fit on a single line.
3377 @item -fdiagnostics-color[=@var{WHEN}]
3378 @itemx -fno-diagnostics-color
3379 @opindex fdiagnostics-color
3380 @cindex highlight, color
3381 @vindex GCC_COLORS @r{environment variable}
3382 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3383 or @samp{auto}. The default depends on how the compiler has been configured,
3384 it can be any of the above @var{WHEN} options or also @samp{never}
3385 if @env{GCC_COLORS} environment variable isn't present in the environment,
3386 and @samp{auto} otherwise.
3387 @samp{auto} means to use color only when the standard error is a terminal.
3388 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3389 aliases for @option{-fdiagnostics-color=always} and
3390 @option{-fdiagnostics-color=never}, respectively.
3392 The colors are defined by the environment variable @env{GCC_COLORS}.
3393 Its value is a colon-separated list of capabilities and Select Graphic
3394 Rendition (SGR) substrings. SGR commands are interpreted by the
3395 terminal or terminal emulator. (See the section in the documentation
3396 of your text terminal for permitted values and their meanings as
3397 character attributes.) These substring values are integers in decimal
3398 representation and can be concatenated with semicolons.
3399 Common values to concatenate include
3401 @samp{4} for underline,
3403 @samp{7} for inverse,
3404 @samp{39} for default foreground color,
3405 @samp{30} to @samp{37} for foreground colors,
3406 @samp{90} to @samp{97} for 16-color mode foreground colors,
3407 @samp{38;5;0} to @samp{38;5;255}
3408 for 88-color and 256-color modes foreground colors,
3409 @samp{49} for default background color,
3410 @samp{40} to @samp{47} for background colors,
3411 @samp{100} to @samp{107} for 16-color mode background colors,
3412 and @samp{48;5;0} to @samp{48;5;255}
3413 for 88-color and 256-color modes background colors.
3415 The default @env{GCC_COLORS} is
3417 error=01;31:warning=01;35:note=01;36:range1=32:range2=34:locus=01:quote=01:\
3418 fixit-insert=32:fixit-delete=31:\
3419 diff-filename=01:diff-hunk=32:diff-delete=31:diff-insert=32
3422 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3423 @samp{01;36} is bold cyan, @samp{32} is green, @samp{34} is blue,
3424 @samp{01} is bold, and @samp{31} is red.
3425 Setting @env{GCC_COLORS} to the empty string disables colors.
3426 Supported capabilities are as follows.
3430 @vindex error GCC_COLORS @r{capability}
3431 SGR substring for error: markers.
3434 @vindex warning GCC_COLORS @r{capability}
3435 SGR substring for warning: markers.
3438 @vindex note GCC_COLORS @r{capability}
3439 SGR substring for note: markers.
3442 @vindex range1 GCC_COLORS @r{capability}
3443 SGR substring for first additional range.
3446 @vindex range2 GCC_COLORS @r{capability}
3447 SGR substring for second additional range.
3450 @vindex locus GCC_COLORS @r{capability}
3451 SGR substring for location information, @samp{file:line} or
3452 @samp{file:line:column} etc.
3455 @vindex quote GCC_COLORS @r{capability}
3456 SGR substring for information printed within quotes.
3459 @vindex fixit-insert GCC_COLORS @r{capability}
3460 SGR substring for fix-it hints suggesting text to
3461 be inserted or replaced.
3464 @vindex fixit-delete GCC_COLORS @r{capability}
3465 SGR substring for fix-it hints suggesting text to
3468 @item diff-filename=
3469 @vindex diff-filename GCC_COLORS @r{capability}
3470 SGR substring for filename headers within generated patches.
3473 @vindex diff-hunk GCC_COLORS @r{capability}
3474 SGR substring for the starts of hunks within generated patches.
3477 @vindex diff-delete GCC_COLORS @r{capability}
3478 SGR substring for deleted lines within generated patches.
3481 @vindex diff-insert GCC_COLORS @r{capability}
3482 SGR substring for inserted lines within generated patches.
3485 @item -fno-diagnostics-show-option
3486 @opindex fno-diagnostics-show-option
3487 @opindex fdiagnostics-show-option
3488 By default, each diagnostic emitted includes text indicating the
3489 command-line option that directly controls the diagnostic (if such an
3490 option is known to the diagnostic machinery). Specifying the
3491 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3493 @item -fno-diagnostics-show-caret
3494 @opindex fno-diagnostics-show-caret
3495 @opindex fdiagnostics-show-caret
3496 By default, each diagnostic emitted includes the original source line
3497 and a caret @samp{^} indicating the column. This option suppresses this
3498 information. The source line is truncated to @var{n} characters, if
3499 the @option{-fmessage-length=n} option is given. When the output is done
3500 to the terminal, the width is limited to the width given by the
3501 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3503 @item -fdiagnostics-parseable-fixits
3504 @opindex fdiagnostics-parseable-fixits
3505 Emit fix-it hints in a machine-parseable format, suitable for consumption
3506 by IDEs. For each fix-it, a line will be printed after the relevant
3507 diagnostic, starting with the string ``fix-it:''. For example:
3510 fix-it:"test.c":@{45:3-45:21@}:"gtk_widget_show_all"
3513 The location is expressed as a half-open range, expressed as a count of
3514 bytes, starting at byte 1 for the initial column. In the above example,
3515 bytes 3 through 20 of line 45 of ``test.c'' are to be replaced with the
3519 00000000011111111112222222222
3520 12345678901234567890123456789
3521 gtk_widget_showall (dlg);
3526 The filename and replacement string escape backslash as ``\\", tab as ``\t'',
3527 newline as ``\n'', double quotes as ``\"'', non-printable characters as octal
3528 (e.g. vertical tab as ``\013'').
3530 An empty replacement string indicates that the given range is to be removed.
3531 An empty range (e.g. ``45:3-45:3'') indicates that the string is to
3532 be inserted at the given position.
3534 @item -fdiagnostics-generate-patch
3535 @opindex fdiagnostics-generate-patch
3536 Print fix-it hints to stderr in unified diff format, after any diagnostics
3537 are printed. For example:
3544 void show_cb(GtkDialog *dlg)
3546 - gtk_widget_showall(dlg);
3547 + gtk_widget_show_all(dlg);
3552 The diff may or may not be colorized, following the same rules
3553 as for diagnostics (see @option{-fdiagnostics-color}).
3557 @node Warning Options
3558 @section Options to Request or Suppress Warnings
3559 @cindex options to control warnings
3560 @cindex warning messages
3561 @cindex messages, warning
3562 @cindex suppressing warnings
3564 Warnings are diagnostic messages that report constructions that
3565 are not inherently erroneous but that are risky or suggest there
3566 may have been an error.
3568 The following language-independent options do not enable specific
3569 warnings but control the kinds of diagnostics produced by GCC@.
3572 @cindex syntax checking
3574 @opindex fsyntax-only
3575 Check the code for syntax errors, but don't do anything beyond that.
3577 @item -fmax-errors=@var{n}
3578 @opindex fmax-errors
3579 Limits the maximum number of error messages to @var{n}, at which point
3580 GCC bails out rather than attempting to continue processing the source
3581 code. If @var{n} is 0 (the default), there is no limit on the number
3582 of error messages produced. If @option{-Wfatal-errors} is also
3583 specified, then @option{-Wfatal-errors} takes precedence over this
3588 Inhibit all warning messages.
3593 Make all warnings into errors.
3598 Make the specified warning into an error. The specifier for a warning
3599 is appended; for example @option{-Werror=switch} turns the warnings
3600 controlled by @option{-Wswitch} into errors. This switch takes a
3601 negative form, to be used to negate @option{-Werror} for specific
3602 warnings; for example @option{-Wno-error=switch} makes
3603 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3606 The warning message for each controllable warning includes the
3607 option that controls the warning. That option can then be used with
3608 @option{-Werror=} and @option{-Wno-error=} as described above.
3609 (Printing of the option in the warning message can be disabled using the
3610 @option{-fno-diagnostics-show-option} flag.)
3612 Note that specifying @option{-Werror=}@var{foo} automatically implies
3613 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3616 @item -Wfatal-errors
3617 @opindex Wfatal-errors
3618 @opindex Wno-fatal-errors
3619 This option causes the compiler to abort compilation on the first error
3620 occurred rather than trying to keep going and printing further error
3625 You can request many specific warnings with options beginning with
3626 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3627 implicit declarations. Each of these specific warning options also
3628 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3629 example, @option{-Wno-implicit}. This manual lists only one of the
3630 two forms, whichever is not the default. For further
3631 language-specific options also refer to @ref{C++ Dialect Options} and
3632 @ref{Objective-C and Objective-C++ Dialect Options}.
3634 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3635 options, such as @option{-Wunused}, which may turn on further options,
3636 such as @option{-Wunused-value}. The combined effect of positive and
3637 negative forms is that more specific options have priority over less
3638 specific ones, independently of their position in the command-line. For
3639 options of the same specificity, the last one takes effect. Options
3640 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3641 as if they appeared at the end of the command-line.
3643 When an unrecognized warning option is requested (e.g.,
3644 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3645 that the option is not recognized. However, if the @option{-Wno-} form
3646 is used, the behavior is slightly different: no diagnostic is
3647 produced for @option{-Wno-unknown-warning} unless other diagnostics
3648 are being produced. This allows the use of new @option{-Wno-} options
3649 with old compilers, but if something goes wrong, the compiler
3650 warns that an unrecognized option is present.
3657 Issue all the warnings demanded by strict ISO C and ISO C++;
3658 reject all programs that use forbidden extensions, and some other
3659 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3660 version of the ISO C standard specified by any @option{-std} option used.
3662 Valid ISO C and ISO C++ programs should compile properly with or without
3663 this option (though a rare few require @option{-ansi} or a
3664 @option{-std} option specifying the required version of ISO C)@. However,
3665 without this option, certain GNU extensions and traditional C and C++
3666 features are supported as well. With this option, they are rejected.
3668 @option{-Wpedantic} does not cause warning messages for use of the
3669 alternate keywords whose names begin and end with @samp{__}. Pedantic
3670 warnings are also disabled in the expression that follows
3671 @code{__extension__}. However, only system header files should use
3672 these escape routes; application programs should avoid them.
3673 @xref{Alternate Keywords}.
3675 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3676 C conformance. They soon find that it does not do quite what they want:
3677 it finds some non-ISO practices, but not all---only those for which
3678 ISO C @emph{requires} a diagnostic, and some others for which
3679 diagnostics have been added.
3681 A feature to report any failure to conform to ISO C might be useful in
3682 some instances, but would require considerable additional work and would
3683 be quite different from @option{-Wpedantic}. We don't have plans to
3684 support such a feature in the near future.
3686 Where the standard specified with @option{-std} represents a GNU
3687 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3688 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3689 extended dialect is based. Warnings from @option{-Wpedantic} are given
3690 where they are required by the base standard. (It does not make sense
3691 for such warnings to be given only for features not in the specified GNU
3692 C dialect, since by definition the GNU dialects of C include all
3693 features the compiler supports with the given option, and there would be
3694 nothing to warn about.)
3696 @item -pedantic-errors
3697 @opindex pedantic-errors
3698 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3699 requires a diagnostic, in some cases where there is undefined behavior
3700 at compile-time and in some other cases that do not prevent compilation
3701 of programs that are valid according to the standard. This is not
3702 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3703 by this option and not enabled by the latter and vice versa.
3708 This enables all the warnings about constructions that some users
3709 consider questionable, and that are easy to avoid (or modify to
3710 prevent the warning), even in conjunction with macros. This also
3711 enables some language-specific warnings described in @ref{C++ Dialect
3712 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3714 @option{-Wall} turns on the following warning flags:
3716 @gccoptlist{-Waddress @gol
3717 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3719 -Wbool-operation @gol
3720 -Wc++11-compat -Wc++14-compat@gol
3721 -Wchar-subscripts @gol
3723 -Wduplicate-decl-specifier @r{(C and Objective-C only)} @gol
3724 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3726 -Wint-in-bool-context @gol
3727 -Wimplicit @r{(C and Objective-C only)} @gol
3728 -Wimplicit-int @r{(C and Objective-C only)} @gol
3729 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3730 -Winit-self @r{(only for C++)} @gol
3731 -Wlogical-not-parentheses
3732 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3733 -Wmaybe-uninitialized @gol
3734 -Wmemset-elt-size @gol
3735 -Wmemset-transposed-args @gol
3736 -Wmisleading-indentation @r{(only for C/C++)} @gol
3737 -Wmissing-braces @r{(only for C/ObjC)} @gol
3738 -Wnarrowing @r{(only for C++)} @gol
3740 -Wnonnull-compare @gol
3746 -Wsequence-point @gol
3747 -Wsign-compare @r{(only in C++)} @gol
3748 -Wsizeof-pointer-memaccess @gol
3749 -Wstrict-aliasing @gol
3750 -Wstrict-overflow=1 @gol
3752 -Wtautological-compare @gol
3754 -Wuninitialized @gol
3755 -Wunknown-pragmas @gol
3756 -Wunused-function @gol
3759 -Wunused-variable @gol
3760 -Wvolatile-register-var @gol
3763 Note that some warning flags are not implied by @option{-Wall}. Some of
3764 them warn about constructions that users generally do not consider
3765 questionable, but which occasionally you might wish to check for;
3766 others warn about constructions that are necessary or hard to avoid in
3767 some cases, and there is no simple way to modify the code to suppress
3768 the warning. Some of them are enabled by @option{-Wextra} but many of
3769 them must be enabled individually.
3775 This enables some extra warning flags that are not enabled by
3776 @option{-Wall}. (This option used to be called @option{-W}. The older
3777 name is still supported, but the newer name is more descriptive.)
3779 @gccoptlist{-Wclobbered @gol
3781 -Wignored-qualifiers @gol
3782 -Wimplicit-fallthrough=3 @gol
3783 -Wmissing-field-initializers @gol
3784 -Wmissing-parameter-type @r{(C only)} @gol
3785 -Wold-style-declaration @r{(C only)} @gol
3786 -Woverride-init @gol
3787 -Wsign-compare @r{(C only)} @gol
3789 -Wuninitialized @gol
3790 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
3791 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3792 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3795 The option @option{-Wextra} also prints warning messages for the
3801 A pointer is compared against integer zero with @code{<}, @code{<=},
3802 @code{>}, or @code{>=}.
3805 (C++ only) An enumerator and a non-enumerator both appear in a
3806 conditional expression.
3809 (C++ only) Ambiguous virtual bases.
3812 (C++ only) Subscripting an array that has been declared @code{register}.
3815 (C++ only) Taking the address of a variable that has been declared
3819 (C++ only) A base class is not initialized in the copy constructor
3824 @item -Wchar-subscripts
3825 @opindex Wchar-subscripts
3826 @opindex Wno-char-subscripts
3827 Warn if an array subscript has type @code{char}. This is a common cause
3828 of error, as programmers often forget that this type is signed on some
3830 This warning is enabled by @option{-Wall}.
3834 @opindex Wno-comment
3835 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3836 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3837 This warning is enabled by @option{-Wall}.
3839 @item -Wno-coverage-mismatch
3840 @opindex Wno-coverage-mismatch
3841 Warn if feedback profiles do not match when using the
3842 @option{-fprofile-use} option.
3843 If a source file is changed between compiling with @option{-fprofile-gen} and
3844 with @option{-fprofile-use}, the files with the profile feedback can fail
3845 to match the source file and GCC cannot use the profile feedback
3846 information. By default, this warning is enabled and is treated as an
3847 error. @option{-Wno-coverage-mismatch} can be used to disable the
3848 warning or @option{-Wno-error=coverage-mismatch} can be used to
3849 disable the error. Disabling the error for this warning can result in
3850 poorly optimized code and is useful only in the
3851 case of very minor changes such as bug fixes to an existing code-base.
3852 Completely disabling the warning is not recommended.
3855 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3857 Suppress warning messages emitted by @code{#warning} directives.
3859 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3860 @opindex Wdouble-promotion
3861 @opindex Wno-double-promotion
3862 Give a warning when a value of type @code{float} is implicitly
3863 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3864 floating-point unit implement @code{float} in hardware, but emulate
3865 @code{double} in software. On such a machine, doing computations
3866 using @code{double} values is much more expensive because of the
3867 overhead required for software emulation.
3869 It is easy to accidentally do computations with @code{double} because
3870 floating-point literals are implicitly of type @code{double}. For
3874 float area(float radius)
3876 return 3.14159 * radius * radius;
3880 the compiler performs the entire computation with @code{double}
3881 because the floating-point literal is a @code{double}.
3883 @item -Wduplicate-decl-specifier @r{(C and Objective-C only)}
3884 @opindex Wduplicate-decl-specifier
3885 @opindex Wno-duplicate-decl-specifier
3886 Warn if a declaration has duplicate @code{const}, @code{volatile},
3887 @code{restrict} or @code{_Atomic} specifier. This warning is enabled by
3891 @itemx -Wformat=@var{n}
3894 @opindex ffreestanding
3895 @opindex fno-builtin
3897 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3898 the arguments supplied have types appropriate to the format string
3899 specified, and that the conversions specified in the format string make
3900 sense. This includes standard functions, and others specified by format
3901 attributes (@pxref{Function Attributes}), in the @code{printf},
3902 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3903 not in the C standard) families (or other target-specific families).
3904 Which functions are checked without format attributes having been
3905 specified depends on the standard version selected, and such checks of
3906 functions without the attribute specified are disabled by
3907 @option{-ffreestanding} or @option{-fno-builtin}.
3909 The formats are checked against the format features supported by GNU
3910 libc version 2.2. These include all ISO C90 and C99 features, as well
3911 as features from the Single Unix Specification and some BSD and GNU
3912 extensions. Other library implementations may not support all these
3913 features; GCC does not support warning about features that go beyond a
3914 particular library's limitations. However, if @option{-Wpedantic} is used
3915 with @option{-Wformat}, warnings are given about format features not
3916 in the selected standard version (but not for @code{strfmon} formats,
3917 since those are not in any version of the C standard). @xref{C Dialect
3918 Options,,Options Controlling C Dialect}.
3925 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3926 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3927 @option{-Wformat} also checks for null format arguments for several
3928 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3929 aspects of this level of format checking can be disabled by the
3930 options: @option{-Wno-format-contains-nul},
3931 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3932 @option{-Wformat} is enabled by @option{-Wall}.
3934 @item -Wno-format-contains-nul
3935 @opindex Wno-format-contains-nul
3936 @opindex Wformat-contains-nul
3937 If @option{-Wformat} is specified, do not warn about format strings that
3940 @item -Wno-format-extra-args
3941 @opindex Wno-format-extra-args
3942 @opindex Wformat-extra-args
3943 If @option{-Wformat} is specified, do not warn about excess arguments to a
3944 @code{printf} or @code{scanf} format function. The C standard specifies
3945 that such arguments are ignored.
3947 Where the unused arguments lie between used arguments that are
3948 specified with @samp{$} operand number specifications, normally
3949 warnings are still given, since the implementation could not know what
3950 type to pass to @code{va_arg} to skip the unused arguments. However,
3951 in the case of @code{scanf} formats, this option suppresses the
3952 warning if the unused arguments are all pointers, since the Single
3953 Unix Specification says that such unused arguments are allowed.
3955 @item -Wformat-length
3956 @itemx -Wformat-length=@var{level}
3957 @opindex Wformat-length
3958 @opindex Wno-format-length
3959 Warn about calls to formatted input/output functions such as @code{sprintf}
3960 that might overflow the destination buffer, or about bounded functions such
3961 as @code{snprintf} that might result in output truncation. When the exact
3962 number of bytes written by a format directive cannot be determined at
3963 compile-time it is estimated based on heuristics that depend on the
3964 @var{level} argument and on optimization. While enabling optimization
3965 will in most cases improve the accuracy of the warning, it may also
3966 result in false positives.
3969 @item -Wformat-length
3970 @item -Wformat-length=1
3971 @opindex Wformat-length
3972 @opindex Wno-format-length
3973 Level @var{1} of @option{-Wformat-length} enabled by @option{-Wformat}
3974 employs a conservative approach that warns only about calls that most
3975 likely overflow the buffer or result in output truncation. At this
3976 level, numeric arguments to format directives with unknown values are
3977 assumed to have the value of one, and strings of unknown length to be
3978 empty. Numeric arguments that are known to be bounded to a subrange
3979 of their type, or string arguments whose output is bounded either by
3980 their directive's precision or by a finite set of string literals, are
3981 assumed to take on the value within the range that results in the most
3982 bytes on output. For example, the call to @code{sprintf} below is
3983 diagnosed because even with both @var{a} and @var{b} equal to zero,
3984 the terminating NUL character (@code{'\0'}) appended by the function
3985 to the destination buffer will be written past its end. Increasing
3986 the size of the buffer by a single byte is sufficient to avoid the
3987 warning, though it may not be sufficient to avoid the overflow.
3990 void f (int a, int b)
3993 sprintf (buf, "a = %i, b = %i\n", a, b);
3997 @item -Wformat-length=2
3998 Level @var{2} warns also about calls that might overflow the destination
3999 buffer or result in truncation given an argument of sufficient length
4000 or magnitude. At level @var{2}, unknown numeric arguments are assumed
4001 to have the minimum representable value for signed types with a precision
4002 greater than 1, and the maximum representable value otherwise. Unknown
4003 string arguments whose length cannot be assumed to be bounded either by
4004 the directive's precision, or by a finite set of string literals they
4005 may evaluate to, or the character array they may point to, are assumed
4006 to be 1 character long.
4008 At level @var{2}, the call in the example above is again diagnosed, but
4009 this time because with @var{a} equal to a 32-bit @code{INT_MIN} the first
4010 @code{%i} directive will write some of its digits beyond the end of
4011 the destination buffer. To make the call safe regardless of the values
4012 of the two variables, the size of the destination buffer must be increased
4013 to at least 34 bytes. GCC includes the minimum size of the buffer in
4014 an informational note following the warning.
4016 An alternative to increasing the size of the destination buffer is to
4017 constrain the range of formatted values. The maximum length of string
4018 arguments can be bounded by specifying the precision in the format
4019 directive. When numeric arguments of format directives can be assumed
4020 to be bounded by less than the precision of their type, choosing
4021 an appropriate length modifier to the format specifier will reduce
4022 the required buffer size. For example, if @var{a} and @var{b} in the
4023 example above can be assumed to be within the precision of
4024 the @code{short int} type then using either the @code{%hi} format
4025 directive or casting the argument to @code{short} reduces the maximum
4026 required size of the buffer to 24 bytes.
4029 void f (int a, int b)
4032 sprintf (buf, "a = %hi, b = %i\n", a, (short)b);
4037 @item -Wno-format-zero-length
4038 @opindex Wno-format-zero-length
4039 @opindex Wformat-zero-length
4040 If @option{-Wformat} is specified, do not warn about zero-length formats.
4041 The C standard specifies that zero-length formats are allowed.
4046 Enable @option{-Wformat} plus additional format checks. Currently
4047 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
4050 @item -Wformat-nonliteral
4051 @opindex Wformat-nonliteral
4052 @opindex Wno-format-nonliteral
4053 If @option{-Wformat} is specified, also warn if the format string is not a
4054 string literal and so cannot be checked, unless the format function
4055 takes its format arguments as a @code{va_list}.
4057 @item -Wformat-security
4058 @opindex Wformat-security
4059 @opindex Wno-format-security
4060 If @option{-Wformat} is specified, also warn about uses of format
4061 functions that represent possible security problems. At present, this
4062 warns about calls to @code{printf} and @code{scanf} functions where the
4063 format string is not a string literal and there are no format arguments,
4064 as in @code{printf (foo);}. This may be a security hole if the format
4065 string came from untrusted input and contains @samp{%n}. (This is
4066 currently a subset of what @option{-Wformat-nonliteral} warns about, but
4067 in future warnings may be added to @option{-Wformat-security} that are not
4068 included in @option{-Wformat-nonliteral}.)
4070 @item -Wformat-signedness
4071 @opindex Wformat-signedness
4072 @opindex Wno-format-signedness
4073 If @option{-Wformat} is specified, also warn if the format string
4074 requires an unsigned argument and the argument is signed and vice versa.
4077 @opindex Wformat-y2k
4078 @opindex Wno-format-y2k
4079 If @option{-Wformat} is specified, also warn about @code{strftime}
4080 formats that may yield only a two-digit year.
4085 @opindex Wno-nonnull
4086 Warn about passing a null pointer for arguments marked as
4087 requiring a non-null value by the @code{nonnull} function attribute.
4089 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
4090 can be disabled with the @option{-Wno-nonnull} option.
4092 @item -Wnonnull-compare
4093 @opindex Wnonnull-compare
4094 @opindex Wno-nonnull-compare
4095 Warn when comparing an argument marked with the @code{nonnull}
4096 function attribute against null inside the function.
4098 @option{-Wnonnull-compare} is included in @option{-Wall}. It
4099 can be disabled with the @option{-Wno-nonnull-compare} option.
4101 @item -Wnull-dereference
4102 @opindex Wnull-dereference
4103 @opindex Wno-null-dereference
4104 Warn if the compiler detects paths that trigger erroneous or
4105 undefined behavior due to dereferencing a null pointer. This option
4106 is only active when @option{-fdelete-null-pointer-checks} is active,
4107 which is enabled by optimizations in most targets. The precision of
4108 the warnings depends on the optimization options used.
4110 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
4112 @opindex Wno-init-self
4113 Warn about uninitialized variables that are initialized with themselves.
4114 Note this option can only be used with the @option{-Wuninitialized} option.
4116 For example, GCC warns about @code{i} being uninitialized in the
4117 following snippet only when @option{-Winit-self} has been specified:
4128 This warning is enabled by @option{-Wall} in C++.
4130 @item -Wimplicit-int @r{(C and Objective-C only)}
4131 @opindex Wimplicit-int
4132 @opindex Wno-implicit-int
4133 Warn when a declaration does not specify a type.
4134 This warning is enabled by @option{-Wall}.
4136 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
4137 @opindex Wimplicit-function-declaration
4138 @opindex Wno-implicit-function-declaration
4139 Give a warning whenever a function is used before being declared. In
4140 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
4141 enabled by default and it is made into an error by
4142 @option{-pedantic-errors}. This warning is also enabled by
4145 @item -Wimplicit @r{(C and Objective-C only)}
4147 @opindex Wno-implicit
4148 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
4149 This warning is enabled by @option{-Wall}.
4151 @item -Wimplicit-fallthrough
4152 @opindex Wimplicit-fallthrough
4153 @opindex Wno-implicit-fallthrough
4154 @option{-Wimplicit-fallthrough} is the same as @option{-Wimplicit-fallthrough=3}
4155 and @option{-Wno-implicit-fallthrough} is the same as
4156 @option{-Wimplicit-fallthrough=0}.
4158 @item -Wimplicit-fallthrough=@var{n}
4159 @opindex Wimplicit-fallthrough=
4160 Warn when a switch case falls through. For example:
4178 This warning does not warn when the last statement of a case cannot
4179 fall through, e.g. when there is a return statement or a call to function
4180 declared with the noreturn attribute. @option{-Wimplicit-fallthrough=}
4181 also takes into account control flow statements, such as ifs, and only
4182 warns when appropriate. E.g.@:
4192 @} else if (i < 1) @{
4202 Since there are occasions where a switch case fall through is desirable,
4203 GCC provides an attribute, @code{__attribute__ ((fallthrough))}, that is
4204 to be used along with a null statement to suppress this warning that
4205 would normally occur:
4213 __attribute__ ((fallthrough));
4220 C++17 provides a standard way to suppress the @option{-Wimplicit-fallthrough}
4221 warning using @code{[[fallthrough]];} instead of the GNU attribute. In C++11
4222 or C++14 users can use @code{[[gnu::fallthrough]];}, which is a GNU extension.
4223 Instead of the these attributes, it is also possible to add a fallthrough
4224 comment to silence the warning. The whole body of the C or C++ style comment
4225 should match the given regular expressions listed below. The option argument
4226 @var{n} specifies what kind of comments are accepted:
4230 @item @option{-Wimplicit-fallthrough=0} disables the warning altogether.
4232 @item @option{-Wimplicit-fallthrough=1} matches @code{.*} regular
4233 expression, any comment is used as fallthrough comment.
4235 @item @option{-Wimplicit-fallthrough=2} case insensitively matches
4236 @code{.*falls?[ \t-]*thr(ough|u).*} regular expression.
4238 @item @option{-Wimplicit-fallthrough=3} case sensitively matches one of the
4239 following regular expressions:
4243 @item @code{-fallthrough}
4245 @item @code{@@fallthrough@@}
4247 @item @code{lint -fallthrough[ \t]*}
4249 @item @code{[ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?@*FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)?}
4251 @item @code{[ \t.!]*(Else,? |Intentional(ly)? )?@*Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4253 @item @code{[ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?@*fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4257 @item @option{-Wimplicit-fallthrough=4} case sensitively matches one of the
4258 following regular expressions:
4262 @item @code{-fallthrough}
4264 @item @code{@@fallthrough@@}
4266 @item @code{lint -fallthrough[ \t]*}
4268 @item @code{[ \t]*FALLTHR(OUGH|U)[ \t]*}
4272 @item @option{-Wimplicit-fallthrough=5} doesn't recognize any comments as
4273 fallthrough comments, only attributes disable the warning.
4277 The comment needs to be followed after optional whitespace and other comments
4278 by @code{case} or @code{default} keywords or by a user label that preceeds some
4279 @code{case} or @code{default} label.
4294 The @option{-Wimplicit-fallthrough=3} warning is enabled by @option{-Wextra}.
4296 @item -Wignored-qualifiers @r{(C and C++ only)}
4297 @opindex Wignored-qualifiers
4298 @opindex Wno-ignored-qualifiers
4299 Warn if the return type of a function has a type qualifier
4300 such as @code{const}. For ISO C such a type qualifier has no effect,
4301 since the value returned by a function is not an lvalue.
4302 For C++, the warning is only emitted for scalar types or @code{void}.
4303 ISO C prohibits qualified @code{void} return types on function
4304 definitions, so such return types always receive a warning
4305 even without this option.
4307 This warning is also enabled by @option{-Wextra}.
4309 @item -Wignored-attributes @r{(C and C++ only)}
4310 @opindex Wignored-attributes
4311 @opindex Wno-ignored-attributes
4312 Warn when an attribute is ignored. This is different from the
4313 @option{-Wattributes} option in that it warns whenever the compiler decides
4314 to drop an attribute, not that the attribute is either unknown, used in a
4315 wrong place, etc. This warning is enabled by default.
4320 Warn if the type of @code{main} is suspicious. @code{main} should be
4321 a function with external linkage, returning int, taking either zero
4322 arguments, two, or three arguments of appropriate types. This warning
4323 is enabled by default in C++ and is enabled by either @option{-Wall}
4324 or @option{-Wpedantic}.
4326 @item -Wmisleading-indentation @r{(C and C++ only)}
4327 @opindex Wmisleading-indentation
4328 @opindex Wno-misleading-indentation
4329 Warn when the indentation of the code does not reflect the block structure.
4330 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
4331 @code{for} clauses with a guarded statement that does not use braces,
4332 followed by an unguarded statement with the same indentation.
4334 In the following example, the call to ``bar'' is misleadingly indented as
4335 if it were guarded by the ``if'' conditional.
4338 if (some_condition ())
4340 bar (); /* Gotcha: this is not guarded by the "if". */
4343 In the case of mixed tabs and spaces, the warning uses the
4344 @option{-ftabstop=} option to determine if the statements line up
4347 The warning is not issued for code involving multiline preprocessor logic
4348 such as the following example.
4353 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
4359 The warning is not issued after a @code{#line} directive, since this
4360 typically indicates autogenerated code, and no assumptions can be made
4361 about the layout of the file that the directive references.
4363 This warning is enabled by @option{-Wall} in C and C++.
4365 @item -Wmissing-braces
4366 @opindex Wmissing-braces
4367 @opindex Wno-missing-braces
4368 Warn if an aggregate or union initializer is not fully bracketed. In
4369 the following example, the initializer for @code{a} is not fully
4370 bracketed, but that for @code{b} is fully bracketed. This warning is
4371 enabled by @option{-Wall} in C.
4374 int a[2][2] = @{ 0, 1, 2, 3 @};
4375 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
4378 This warning is enabled by @option{-Wall}.
4380 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
4381 @opindex Wmissing-include-dirs
4382 @opindex Wno-missing-include-dirs
4383 Warn if a user-supplied include directory does not exist.
4386 @opindex Wparentheses
4387 @opindex Wno-parentheses
4388 Warn if parentheses are omitted in certain contexts, such
4389 as when there is an assignment in a context where a truth value
4390 is expected, or when operators are nested whose precedence people
4391 often get confused about.
4393 Also warn if a comparison like @code{x<=y<=z} appears; this is
4394 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
4395 interpretation from that of ordinary mathematical notation.
4397 Also warn for dangerous uses of the GNU extension to
4398 @code{?:} with omitted middle operand. When the condition
4399 in the @code{?}: operator is a boolean expression, the omitted value is
4400 always 1. Often programmers expect it to be a value computed
4401 inside the conditional expression instead.
4403 This warning is enabled by @option{-Wall}.
4405 @item -Wsequence-point
4406 @opindex Wsequence-point
4407 @opindex Wno-sequence-point
4408 Warn about code that may have undefined semantics because of violations
4409 of sequence point rules in the C and C++ standards.
4411 The C and C++ standards define the order in which expressions in a C/C++
4412 program are evaluated in terms of @dfn{sequence points}, which represent
4413 a partial ordering between the execution of parts of the program: those
4414 executed before the sequence point, and those executed after it. These
4415 occur after the evaluation of a full expression (one which is not part
4416 of a larger expression), after the evaluation of the first operand of a
4417 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
4418 function is called (but after the evaluation of its arguments and the
4419 expression denoting the called function), and in certain other places.
4420 Other than as expressed by the sequence point rules, the order of
4421 evaluation of subexpressions of an expression is not specified. All
4422 these rules describe only a partial order rather than a total order,
4423 since, for example, if two functions are called within one expression
4424 with no sequence point between them, the order in which the functions
4425 are called is not specified. However, the standards committee have
4426 ruled that function calls do not overlap.
4428 It is not specified when between sequence points modifications to the
4429 values of objects take effect. Programs whose behavior depends on this
4430 have undefined behavior; the C and C++ standards specify that ``Between
4431 the previous and next sequence point an object shall have its stored
4432 value modified at most once by the evaluation of an expression.
4433 Furthermore, the prior value shall be read only to determine the value
4434 to be stored.''. If a program breaks these rules, the results on any
4435 particular implementation are entirely unpredictable.
4437 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
4438 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
4439 diagnosed by this option, and it may give an occasional false positive
4440 result, but in general it has been found fairly effective at detecting
4441 this sort of problem in programs.
4443 The C++17 standard will define the order of evaluation of operands in
4444 more cases: in particular it requires that the right-hand side of an
4445 assignment be evaluated before the left-hand side, so the above
4446 examples are no longer undefined. But this warning will still warn
4447 about them, to help people avoid writing code that is undefined in C
4448 and earlier revisions of C++.
4450 The standard is worded confusingly, therefore there is some debate
4451 over the precise meaning of the sequence point rules in subtle cases.
4452 Links to discussions of the problem, including proposed formal
4453 definitions, may be found on the GCC readings page, at
4454 @uref{http://gcc.gnu.org/@/readings.html}.
4456 This warning is enabled by @option{-Wall} for C and C++.
4458 @item -Wno-return-local-addr
4459 @opindex Wno-return-local-addr
4460 @opindex Wreturn-local-addr
4461 Do not warn about returning a pointer (or in C++, a reference) to a
4462 variable that goes out of scope after the function returns.
4465 @opindex Wreturn-type
4466 @opindex Wno-return-type
4467 Warn whenever a function is defined with a return type that defaults
4468 to @code{int}. Also warn about any @code{return} statement with no
4469 return value in a function whose return type is not @code{void}
4470 (falling off the end of the function body is considered returning
4473 For C only, warn about a @code{return} statement with an expression in a
4474 function whose return type is @code{void}, unless the expression type is
4475 also @code{void}. As a GNU extension, the latter case is accepted
4476 without a warning unless @option{-Wpedantic} is used.
4478 For C++, a function without return type always produces a diagnostic
4479 message, even when @option{-Wno-return-type} is specified. The only
4480 exceptions are @code{main} and functions defined in system headers.
4482 This warning is enabled by @option{-Wall}.
4484 @item -Wshift-count-negative
4485 @opindex Wshift-count-negative
4486 @opindex Wno-shift-count-negative
4487 Warn if shift count is negative. This warning is enabled by default.
4489 @item -Wshift-count-overflow
4490 @opindex Wshift-count-overflow
4491 @opindex Wno-shift-count-overflow
4492 Warn if shift count >= width of type. This warning is enabled by default.
4494 @item -Wshift-negative-value
4495 @opindex Wshift-negative-value
4496 @opindex Wno-shift-negative-value
4497 Warn if left shifting a negative value. This warning is enabled by
4498 @option{-Wextra} in C99 and C++11 modes (and newer).
4500 @item -Wshift-overflow
4501 @itemx -Wshift-overflow=@var{n}
4502 @opindex Wshift-overflow
4503 @opindex Wno-shift-overflow
4504 Warn about left shift overflows. This warning is enabled by
4505 default in C99 and C++11 modes (and newer).
4508 @item -Wshift-overflow=1
4509 This is the warning level of @option{-Wshift-overflow} and is enabled
4510 by default in C99 and C++11 modes (and newer). This warning level does
4511 not warn about left-shifting 1 into the sign bit. (However, in C, such
4512 an overflow is still rejected in contexts where an integer constant expression
4515 @item -Wshift-overflow=2
4516 This warning level also warns about left-shifting 1 into the sign bit,
4517 unless C++14 mode is active.
4523 Warn whenever a @code{switch} statement has an index of enumerated type
4524 and lacks a @code{case} for one or more of the named codes of that
4525 enumeration. (The presence of a @code{default} label prevents this
4526 warning.) @code{case} labels outside the enumeration range also
4527 provoke warnings when this option is used (even if there is a
4528 @code{default} label).
4529 This warning is enabled by @option{-Wall}.
4531 @item -Wswitch-default
4532 @opindex Wswitch-default
4533 @opindex Wno-switch-default
4534 Warn whenever a @code{switch} statement does not have a @code{default}
4538 @opindex Wswitch-enum
4539 @opindex Wno-switch-enum
4540 Warn whenever a @code{switch} statement has an index of enumerated type
4541 and lacks a @code{case} for one or more of the named codes of that
4542 enumeration. @code{case} labels outside the enumeration range also
4543 provoke warnings when this option is used. The only difference
4544 between @option{-Wswitch} and this option is that this option gives a
4545 warning about an omitted enumeration code even if there is a
4546 @code{default} label.
4549 @opindex Wswitch-bool
4550 @opindex Wno-switch-bool
4551 Warn whenever a @code{switch} statement has an index of boolean type
4552 and the case values are outside the range of a boolean type.
4553 It is possible to suppress this warning by casting the controlling
4554 expression to a type other than @code{bool}. For example:
4557 switch ((int) (a == 4))
4563 This warning is enabled by default for C and C++ programs.
4565 @item -Wswitch-unreachable
4566 @opindex Wswitch-unreachable
4567 @opindex Wno-switch-unreachable
4568 Warn whenever a @code{switch} statement contains statements between the
4569 controlling expression and the first case label, which will never be
4570 executed. For example:
4582 @option{-Wswitch-unreachable} does not warn if the statement between the
4583 controlling expression and the first case label is just a declaration:
4596 This warning is enabled by default for C and C++ programs.
4598 @item -Wsync-nand @r{(C and C++ only)}
4600 @opindex Wno-sync-nand
4601 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
4602 built-in functions are used. These functions changed semantics in GCC 4.4.
4606 @opindex Wno-trigraphs
4607 Warn if any trigraphs are encountered that might change the meaning of
4608 the program (trigraphs within comments are not warned about).
4609 This warning is enabled by @option{-Wall}.
4611 @item -Wunused-but-set-parameter
4612 @opindex Wunused-but-set-parameter
4613 @opindex Wno-unused-but-set-parameter
4614 Warn whenever a function parameter is assigned to, but otherwise unused
4615 (aside from its declaration).
4617 To suppress this warning use the @code{unused} attribute
4618 (@pxref{Variable Attributes}).
4620 This warning is also enabled by @option{-Wunused} together with
4623 @item -Wunused-but-set-variable
4624 @opindex Wunused-but-set-variable
4625 @opindex Wno-unused-but-set-variable
4626 Warn whenever a local variable is assigned to, but otherwise unused
4627 (aside from its declaration).
4628 This warning is enabled by @option{-Wall}.
4630 To suppress this warning use the @code{unused} attribute
4631 (@pxref{Variable Attributes}).
4633 This warning is also enabled by @option{-Wunused}, which is enabled
4636 @item -Wunused-function
4637 @opindex Wunused-function
4638 @opindex Wno-unused-function
4639 Warn whenever a static function is declared but not defined or a
4640 non-inline static function is unused.
4641 This warning is enabled by @option{-Wall}.
4643 @item -Wunused-label
4644 @opindex Wunused-label
4645 @opindex Wno-unused-label
4646 Warn whenever a label is declared but not used.
4647 This warning is enabled by @option{-Wall}.
4649 To suppress this warning use the @code{unused} attribute
4650 (@pxref{Variable Attributes}).
4652 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4653 @opindex Wunused-local-typedefs
4654 Warn when a typedef locally defined in a function is not used.
4655 This warning is enabled by @option{-Wall}.
4657 @item -Wunused-parameter
4658 @opindex Wunused-parameter
4659 @opindex Wno-unused-parameter
4660 Warn whenever a function parameter is unused aside from its declaration.
4662 To suppress this warning use the @code{unused} attribute
4663 (@pxref{Variable Attributes}).
4665 @item -Wno-unused-result
4666 @opindex Wunused-result
4667 @opindex Wno-unused-result
4668 Do not warn if a caller of a function marked with attribute
4669 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4670 its return value. The default is @option{-Wunused-result}.
4672 @item -Wunused-variable
4673 @opindex Wunused-variable
4674 @opindex Wno-unused-variable
4675 Warn whenever a local or static variable is unused aside from its
4676 declaration. This option implies @option{-Wunused-const-variable=1} for C,
4677 but not for C++. This warning is enabled by @option{-Wall}.
4679 To suppress this warning use the @code{unused} attribute
4680 (@pxref{Variable Attributes}).
4682 @item -Wunused-const-variable
4683 @itemx -Wunused-const-variable=@var{n}
4684 @opindex Wunused-const-variable
4685 @opindex Wno-unused-const-variable
4686 Warn whenever a constant static variable is unused aside from its declaration.
4687 @option{-Wunused-const-variable=1} is enabled by @option{-Wunused-variable}
4688 for C, but not for C++. In C this declares variable storage, but in C++ this
4689 is not an error since const variables take the place of @code{#define}s.
4691 To suppress this warning use the @code{unused} attribute
4692 (@pxref{Variable Attributes}).
4695 @item -Wunused-const-variable=1
4696 This is the warning level that is enabled by @option{-Wunused-variable} for
4697 C. It warns only about unused static const variables defined in the main
4698 compilation unit, but not about static const variables declared in any
4701 @item -Wunused-const-variable=2
4702 This warning level also warns for unused constant static variables in
4703 headers (excluding system headers). This is the warning level of
4704 @option{-Wunused-const-variable} and must be explicitly requested since
4705 in C++ this isn't an error and in C it might be harder to clean up all
4709 @item -Wunused-value
4710 @opindex Wunused-value
4711 @opindex Wno-unused-value
4712 Warn whenever a statement computes a result that is explicitly not
4713 used. To suppress this warning cast the unused expression to
4714 @code{void}. This includes an expression-statement or the left-hand
4715 side of a comma expression that contains no side effects. For example,
4716 an expression such as @code{x[i,j]} causes a warning, while
4717 @code{x[(void)i,j]} does not.
4719 This warning is enabled by @option{-Wall}.
4724 All the above @option{-Wunused} options combined.
4726 In order to get a warning about an unused function parameter, you must
4727 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4728 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4730 @item -Wuninitialized
4731 @opindex Wuninitialized
4732 @opindex Wno-uninitialized
4733 Warn if an automatic variable is used without first being initialized
4734 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4735 warn if a non-static reference or non-static @code{const} member
4736 appears in a class without constructors.
4738 If you want to warn about code that uses the uninitialized value of the
4739 variable in its own initializer, use the @option{-Winit-self} option.
4741 These warnings occur for individual uninitialized or clobbered
4742 elements of structure, union or array variables as well as for
4743 variables that are uninitialized or clobbered as a whole. They do
4744 not occur for variables or elements declared @code{volatile}. Because
4745 these warnings depend on optimization, the exact variables or elements
4746 for which there are warnings depends on the precise optimization
4747 options and version of GCC used.
4749 Note that there may be no warning about a variable that is used only
4750 to compute a value that itself is never used, because such
4751 computations may be deleted by data flow analysis before the warnings
4754 @item -Winvalid-memory-model
4755 @opindex Winvalid-memory-model
4756 @opindex Wno-invalid-memory-model
4757 Warn for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins},
4758 and the C11 atomic generic functions with a memory consistency argument
4759 that is either invalid for the operation or outside the range of values
4760 of the @code{memory_order} enumeration. For example, since the
4761 @code{__atomic_store} and @code{__atomic_store_n} built-ins are only
4762 defined for the relaxed, release, and sequentially consistent memory
4763 orders the following code is diagnosed:
4768 __atomic_store_n (i, 0, memory_order_consume);
4772 @option{-Winvalid-memory-model} is enabled by default.
4774 @item -Wmaybe-uninitialized
4775 @opindex Wmaybe-uninitialized
4776 @opindex Wno-maybe-uninitialized
4777 For an automatic variable, if there exists a path from the function
4778 entry to a use of the variable that is initialized, but there exist
4779 some other paths for which the variable is not initialized, the compiler
4780 emits a warning if it cannot prove the uninitialized paths are not
4781 executed at run time. These warnings are made optional because GCC is
4782 not smart enough to see all the reasons why the code might be correct
4783 in spite of appearing to have an error. Here is one example of how
4804 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4805 always initialized, but GCC doesn't know this. To suppress the
4806 warning, you need to provide a default case with assert(0) or
4809 @cindex @code{longjmp} warnings
4810 This option also warns when a non-volatile automatic variable might be
4811 changed by a call to @code{longjmp}. These warnings as well are possible
4812 only in optimizing compilation.
4814 The compiler sees only the calls to @code{setjmp}. It cannot know
4815 where @code{longjmp} will be called; in fact, a signal handler could
4816 call it at any point in the code. As a result, you may get a warning
4817 even when there is in fact no problem because @code{longjmp} cannot
4818 in fact be called at the place that would cause a problem.
4820 Some spurious warnings can be avoided if you declare all the functions
4821 you use that never return as @code{noreturn}. @xref{Function
4824 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4826 @item -Wunknown-pragmas
4827 @opindex Wunknown-pragmas
4828 @opindex Wno-unknown-pragmas
4829 @cindex warning for unknown pragmas
4830 @cindex unknown pragmas, warning
4831 @cindex pragmas, warning of unknown
4832 Warn when a @code{#pragma} directive is encountered that is not understood by
4833 GCC@. If this command-line option is used, warnings are even issued
4834 for unknown pragmas in system header files. This is not the case if
4835 the warnings are only enabled by the @option{-Wall} command-line option.
4838 @opindex Wno-pragmas
4840 Do not warn about misuses of pragmas, such as incorrect parameters,
4841 invalid syntax, or conflicts between pragmas. See also
4842 @option{-Wunknown-pragmas}.
4844 @item -Wstrict-aliasing
4845 @opindex Wstrict-aliasing
4846 @opindex Wno-strict-aliasing
4847 This option is only active when @option{-fstrict-aliasing} is active.
4848 It warns about code that might break the strict aliasing rules that the
4849 compiler is using for optimization. The warning does not catch all
4850 cases, but does attempt to catch the more common pitfalls. It is
4851 included in @option{-Wall}.
4852 It is equivalent to @option{-Wstrict-aliasing=3}
4854 @item -Wstrict-aliasing=n
4855 @opindex Wstrict-aliasing=n
4856 This option is only active when @option{-fstrict-aliasing} is active.
4857 It warns about code that might break the strict aliasing rules that the
4858 compiler is using for optimization.
4859 Higher levels correspond to higher accuracy (fewer false positives).
4860 Higher levels also correspond to more effort, similar to the way @option{-O}
4862 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4864 Level 1: Most aggressive, quick, least accurate.
4865 Possibly useful when higher levels
4866 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4867 false negatives. However, it has many false positives.
4868 Warns for all pointer conversions between possibly incompatible types,
4869 even if never dereferenced. Runs in the front end only.
4871 Level 2: Aggressive, quick, not too precise.
4872 May still have many false positives (not as many as level 1 though),
4873 and few false negatives (but possibly more than level 1).
4874 Unlike level 1, it only warns when an address is taken. Warns about
4875 incomplete types. Runs in the front end only.
4877 Level 3 (default for @option{-Wstrict-aliasing}):
4878 Should have very few false positives and few false
4879 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4880 Takes care of the common pun+dereference pattern in the front end:
4881 @code{*(int*)&some_float}.
4882 If optimization is enabled, it also runs in the back end, where it deals
4883 with multiple statement cases using flow-sensitive points-to information.
4884 Only warns when the converted pointer is dereferenced.
4885 Does not warn about incomplete types.
4887 @item -Wstrict-overflow
4888 @itemx -Wstrict-overflow=@var{n}
4889 @opindex Wstrict-overflow
4890 @opindex Wno-strict-overflow
4891 This option is only active when @option{-fstrict-overflow} is active.
4892 It warns about cases where the compiler optimizes based on the
4893 assumption that signed overflow does not occur. Note that it does not
4894 warn about all cases where the code might overflow: it only warns
4895 about cases where the compiler implements some optimization. Thus
4896 this warning depends on the optimization level.
4898 An optimization that assumes that signed overflow does not occur is
4899 perfectly safe if the values of the variables involved are such that
4900 overflow never does, in fact, occur. Therefore this warning can
4901 easily give a false positive: a warning about code that is not
4902 actually a problem. To help focus on important issues, several
4903 warning levels are defined. No warnings are issued for the use of
4904 undefined signed overflow when estimating how many iterations a loop
4905 requires, in particular when determining whether a loop will be
4909 @item -Wstrict-overflow=1
4910 Warn about cases that are both questionable and easy to avoid. For
4911 example, with @option{-fstrict-overflow}, the compiler simplifies
4912 @code{x + 1 > x} to @code{1}. This level of
4913 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4914 are not, and must be explicitly requested.
4916 @item -Wstrict-overflow=2
4917 Also warn about other cases where a comparison is simplified to a
4918 constant. For example: @code{abs (x) >= 0}. This can only be
4919 simplified when @option{-fstrict-overflow} is in effect, because
4920 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4921 zero. @option{-Wstrict-overflow} (with no level) is the same as
4922 @option{-Wstrict-overflow=2}.
4924 @item -Wstrict-overflow=3
4925 Also warn about other cases where a comparison is simplified. For
4926 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4928 @item -Wstrict-overflow=4
4929 Also warn about other simplifications not covered by the above cases.
4930 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4932 @item -Wstrict-overflow=5
4933 Also warn about cases where the compiler reduces the magnitude of a
4934 constant involved in a comparison. For example: @code{x + 2 > y} is
4935 simplified to @code{x + 1 >= y}. This is reported only at the
4936 highest warning level because this simplification applies to many
4937 comparisons, so this warning level gives a very large number of
4941 @item -Wstringop-overflow
4942 @itemx -Wstringop-overflow=@var{type}
4943 @opindex Wstringop-overflow
4944 @opindex Wno-stringop-overflow
4945 Warn for calls to string manipulation functions such as @code{memcpy} and
4946 @code{strcpy} that are determined to overflow the destination buffer. The
4947 optional argument is one greater than the type of Object Size Checking to
4948 perform to determine the size of the destination. @xref{Object Size Checking}.
4949 The argument is meaningful only for functions that operate on character arrays
4950 but not for raw memory functions like @code{memcpy} which always make use
4951 of Object Size type-0. The option also warns for calls that specify a size
4952 in excess of the largest possible object or at most @code{SIZE_MAX / 2} bytes.
4953 The option produces the best results with optimization enabled but can detect
4954 a small subset of simple buffer overflows even without optimization in
4955 calls to the GCC built-in functions like @code{__builtin_memcpy} that
4956 correspond to the standard functions. In any case, the option warns about
4957 just a subset of buffer overflows detected by the corresponding overflow
4958 checking built-ins. For example, the option will issue a warning for
4959 the @code{strcpy} call below because it copies at least 5 characters
4960 (the string @code{"blue"} including the terminating NUL) into the buffer
4964 enum Color @{ blue, purple, yellow @};
4965 const char* f (enum Color clr)
4967 static char buf [4];
4971 case blue: str = "blue"; break;
4972 case purple: str = "purple"; break;
4973 case yellow: str = "yellow"; break;
4976 return strcpy (buf, str); // warning here
4980 Option @option{-Wstringop-overflow=2} is enabled by default.
4983 @item -Wstringop-overflow
4984 @item -Wstringop-overflow=1
4985 @opindex Wstringop-overflow
4986 @opindex Wno-stringop-overflow
4987 The @option{-Wstringop-overflow=1} option uses type-zero Object Size Checking
4988 to determine the sizes of destination objects. This is the default setting
4989 of the option. At this setting the option will not warn for writes past
4990 the end of subobjects of larger objects accessed by pointers unless the
4991 size of the largest surrounding object is known. When the destination may
4992 be one of several objects it is assumed to be the largest one of them. On
4993 Linux systems, when optimization is enabled at this setting the option warns
4994 for the same code as when the @code{_FORTIFY_SOURCE} macro is defined to
4997 @item -Wstringop-overflow=2
4998 The @option{-Wstringop-overflow=2} option uses type-one Object Size Checking
4999 to determine the sizes of destination objects. At this setting the option
5000 will warn about overflows when writing to members of the largest complete
5001 objects whose exact size is known. It will, however, not warn for excessive
5002 writes to the same members of unknown objects referenced by pointers since
5003 they may point to arrays containing unknown numbers of elements.
5005 @item -Wstringop-overflow=3
5006 The @option{-Wstringop-overflow=3} option uses type-two Object Size Checking
5007 to determine the sizes of destination objects. At this setting the option
5008 warns about overflowing the smallest object or data member. This is the
5009 most restrictive setting of the option that may result in warnings for safe
5012 @item -Wstringop-overflow=4
5013 The @option{-Wstringop-overflow=4} option uses type-three Object Size Checking
5014 to determine the sizes of destination objects. At this setting the option
5015 will warn about overflowing any data members, and when the destination is
5016 one of several objects it uses the size of the largest of them to decide
5017 whether to issue a warning. Similarly to @option{-Wstringop-overflow=3} this
5018 setting of the option may result in warnings for benign code.
5021 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
5022 @opindex Wsuggest-attribute=
5023 @opindex Wno-suggest-attribute=
5024 Warn for cases where adding an attribute may be beneficial. The
5025 attributes currently supported are listed below.
5028 @item -Wsuggest-attribute=pure
5029 @itemx -Wsuggest-attribute=const
5030 @itemx -Wsuggest-attribute=noreturn
5031 @opindex Wsuggest-attribute=pure
5032 @opindex Wno-suggest-attribute=pure
5033 @opindex Wsuggest-attribute=const
5034 @opindex Wno-suggest-attribute=const
5035 @opindex Wsuggest-attribute=noreturn
5036 @opindex Wno-suggest-attribute=noreturn
5038 Warn about functions that might be candidates for attributes
5039 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
5040 functions visible in other compilation units or (in the case of @code{pure} and
5041 @code{const}) if it cannot prove that the function returns normally. A function
5042 returns normally if it doesn't contain an infinite loop or return abnormally
5043 by throwing, calling @code{abort} or trapping. This analysis requires option
5044 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
5045 higher. Higher optimization levels improve the accuracy of the analysis.
5047 @item -Wsuggest-attribute=format
5048 @itemx -Wmissing-format-attribute
5049 @opindex Wsuggest-attribute=format
5050 @opindex Wmissing-format-attribute
5051 @opindex Wno-suggest-attribute=format
5052 @opindex Wno-missing-format-attribute
5056 Warn about function pointers that might be candidates for @code{format}
5057 attributes. Note these are only possible candidates, not absolute ones.
5058 GCC guesses that function pointers with @code{format} attributes that
5059 are used in assignment, initialization, parameter passing or return
5060 statements should have a corresponding @code{format} attribute in the
5061 resulting type. I.e.@: the left-hand side of the assignment or
5062 initialization, the type of the parameter variable, or the return type
5063 of the containing function respectively should also have a @code{format}
5064 attribute to avoid the warning.
5066 GCC also warns about function definitions that might be
5067 candidates for @code{format} attributes. Again, these are only
5068 possible candidates. GCC guesses that @code{format} attributes
5069 might be appropriate for any function that calls a function like
5070 @code{vprintf} or @code{vscanf}, but this might not always be the
5071 case, and some functions for which @code{format} attributes are
5072 appropriate may not be detected.
5075 @item -Wsuggest-final-types
5076 @opindex Wno-suggest-final-types
5077 @opindex Wsuggest-final-types
5078 Warn about types with virtual methods where code quality would be improved
5079 if the type were declared with the C++11 @code{final} specifier,
5081 declared in an anonymous namespace. This allows GCC to more aggressively
5082 devirtualize the polymorphic calls. This warning is more effective with link
5083 time optimization, where the information about the class hierarchy graph is
5086 @item -Wsuggest-final-methods
5087 @opindex Wno-suggest-final-methods
5088 @opindex Wsuggest-final-methods
5089 Warn about virtual methods where code quality would be improved if the method
5090 were declared with the C++11 @code{final} specifier,
5091 or, if possible, its type were
5092 declared in an anonymous namespace or with the @code{final} specifier.
5094 more effective with link time optimization, where the information about the
5095 class hierarchy graph is more complete. It is recommended to first consider
5096 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
5099 @item -Wsuggest-override
5100 Warn about overriding virtual functions that are not marked with the override
5104 @opindex Wno-alloc-zero
5105 @opindex Walloc-zero
5106 Warn about calls to allocation functions decorated with attribute
5107 @code{alloc_size} that specify zero bytes, including those to the built-in
5108 forms of the functions @code{aligned_alloc}, @code{alloca}, @code{calloc},
5109 @code{malloc}, and @code{realloc}. Because the behavior of these functions
5110 when called with a zero size differs among implementations (and in the case
5111 of @code{realloc} has been deprecated) relying on it may result in subtle
5112 portability bugs and should be avoided.
5114 @item -Walloc-size-larger-than=@var{n}
5115 Warn about calls to functions decorated with attribute @code{alloc_size}
5116 that attempt to allocate objects larger than the specified number of bytes,
5117 or where the result of the size computation in an integer type with infinite
5118 precision would exceed @code{SIZE_MAX / 2}. The option argument @var{n}
5119 may end in one of the standard suffixes designating a multiple of bytes
5120 such as @code{kB} and @code{KiB} for kilobyte and kibibyte, respectively,
5121 @code{MB} and @code{MiB} for magabyte and mebibyte, and so on.
5122 @xref{Function Attributes}.
5127 This option warns on all uses of @code{alloca} in the source.
5129 @item -Walloca-larger-than=@var{n}
5130 This option warns on calls to @code{alloca} that are not bounded by a
5131 controlling predicate limiting its argument of integer type to at most
5132 @var{n} bytes, or calls to @code{alloca} where the bound is unknown.
5133 Arguments of non-integer types are considered unbounded even if they
5134 appear to be constrained to the expected range.
5136 For example, a bounded case of @code{alloca} could be:
5139 void func (size_t n)
5150 In the above example, passing @code{-Walloca-larger-than=1000} would not
5151 issue a warning because the call to @code{alloca} is known to be at most
5152 1000 bytes. However, if @code{-Walloca-larger-than=500} were passed,
5153 the compiler would emit a warning.
5155 Unbounded uses, on the other hand, are uses of @code{alloca} with no
5156 controlling predicate constraining its integer argument. For example:
5161 void *p = alloca (n);
5166 If @code{-Walloca-larger-than=500} were passed, the above would trigger
5167 a warning, but this time because of the lack of bounds checking.
5169 Note, that even seemingly correct code involving signed integers could
5173 void func (signed int n)
5183 In the above example, @var{n} could be negative, causing a larger than
5184 expected argument to be implicitly cast into the @code{alloca} call.
5186 This option also warns when @code{alloca} is used in a loop.
5188 This warning is not enabled by @option{-Wall}, and is only active when
5189 @option{-ftree-vrp} is active (default for @option{-O2} and above).
5191 See also @option{-Wvla-larger-than=@var{n}}.
5193 @item -Warray-bounds
5194 @itemx -Warray-bounds=@var{n}
5195 @opindex Wno-array-bounds
5196 @opindex Warray-bounds
5197 This option is only active when @option{-ftree-vrp} is active
5198 (default for @option{-O2} and above). It warns about subscripts to arrays
5199 that are always out of bounds. This warning is enabled by @option{-Wall}.
5202 @item -Warray-bounds=1
5203 This is the warning level of @option{-Warray-bounds} and is enabled
5204 by @option{-Wall}; higher levels are not, and must be explicitly requested.
5206 @item -Warray-bounds=2
5207 This warning level also warns about out of bounds access for
5208 arrays at the end of a struct and for arrays accessed through
5209 pointers. This warning level may give a larger number of
5210 false positives and is deactivated by default.
5213 @item -Wbool-compare
5214 @opindex Wno-bool-compare
5215 @opindex Wbool-compare
5216 Warn about boolean expression compared with an integer value different from
5217 @code{true}/@code{false}. For instance, the following comparison is
5222 if ((n > 1) == 2) @{ @dots{} @}
5224 This warning is enabled by @option{-Wall}.
5226 @item -Wbool-operation
5227 @opindex Wno-bool-operation
5228 @opindex Wbool-operation
5229 Warn about suspicious operations on expressions of a boolean type. For
5230 instance, bitwise negation of a boolean is very likely a bug in the program.
5231 For C, this warning also warns about incrementing or decrementing a boolean,
5232 which rarely makes sense. (In C++, decrementing a boolean is always invalid.
5233 Incrementing a boolean is invalid in C++1z, and deprecated otherwise.)
5235 This warning is enabled by @option{-Wall}.
5237 @item -Wduplicated-cond
5238 @opindex Wno-duplicated-cond
5239 @opindex Wduplicated-cond
5240 Warn about duplicated conditions in an if-else-if chain. For instance,
5241 warn for the following code:
5243 if (p->q != NULL) @{ @dots{} @}
5244 else if (p->q != NULL) @{ @dots{} @}
5247 @item -Wframe-address
5248 @opindex Wno-frame-address
5249 @opindex Wframe-address
5250 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
5251 is called with an argument greater than 0. Such calls may return indeterminate
5252 values or crash the program. The warning is included in @option{-Wall}.
5254 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
5255 @opindex Wno-discarded-qualifiers
5256 @opindex Wdiscarded-qualifiers
5257 Do not warn if type qualifiers on pointers are being discarded.
5258 Typically, the compiler warns if a @code{const char *} variable is
5259 passed to a function that takes a @code{char *} parameter. This option
5260 can be used to suppress such a warning.
5262 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
5263 @opindex Wno-discarded-array-qualifiers
5264 @opindex Wdiscarded-array-qualifiers
5265 Do not warn if type qualifiers on arrays which are pointer targets
5266 are being discarded. Typically, the compiler warns if a
5267 @code{const int (*)[]} variable is passed to a function that
5268 takes a @code{int (*)[]} parameter. This option can be used to
5269 suppress such a warning.
5271 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
5272 @opindex Wno-incompatible-pointer-types
5273 @opindex Wincompatible-pointer-types
5274 Do not warn when there is a conversion between pointers that have incompatible
5275 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
5276 which warns for pointer argument passing or assignment with different
5279 @item -Wno-int-conversion @r{(C and Objective-C only)}
5280 @opindex Wno-int-conversion
5281 @opindex Wint-conversion
5282 Do not warn about incompatible integer to pointer and pointer to integer
5283 conversions. This warning is about implicit conversions; for explicit
5284 conversions the warnings @option{-Wno-int-to-pointer-cast} and
5285 @option{-Wno-pointer-to-int-cast} may be used.
5287 @item -Wno-div-by-zero
5288 @opindex Wno-div-by-zero
5289 @opindex Wdiv-by-zero
5290 Do not warn about compile-time integer division by zero. Floating-point
5291 division by zero is not warned about, as it can be a legitimate way of
5292 obtaining infinities and NaNs.
5294 @item -Wsystem-headers
5295 @opindex Wsystem-headers
5296 @opindex Wno-system-headers
5297 @cindex warnings from system headers
5298 @cindex system headers, warnings from
5299 Print warning messages for constructs found in system header files.
5300 Warnings from system headers are normally suppressed, on the assumption
5301 that they usually do not indicate real problems and would only make the
5302 compiler output harder to read. Using this command-line option tells
5303 GCC to emit warnings from system headers as if they occurred in user
5304 code. However, note that using @option{-Wall} in conjunction with this
5305 option does @emph{not} warn about unknown pragmas in system
5306 headers---for that, @option{-Wunknown-pragmas} must also be used.
5308 @item -Wtautological-compare
5309 @opindex Wtautological-compare
5310 @opindex Wno-tautological-compare
5311 Warn if a self-comparison always evaluates to true or false. This
5312 warning detects various mistakes such as:
5316 if (i > i) @{ @dots{} @}
5318 This warning is enabled by @option{-Wall}.
5321 @opindex Wtrampolines
5322 @opindex Wno-trampolines
5323 Warn about trampolines generated for pointers to nested functions.
5324 A trampoline is a small piece of data or code that is created at run
5325 time on the stack when the address of a nested function is taken, and is
5326 used to call the nested function indirectly. For some targets, it is
5327 made up of data only and thus requires no special treatment. But, for
5328 most targets, it is made up of code and thus requires the stack to be
5329 made executable in order for the program to work properly.
5332 @opindex Wfloat-equal
5333 @opindex Wno-float-equal
5334 Warn if floating-point values are used in equality comparisons.
5336 The idea behind this is that sometimes it is convenient (for the
5337 programmer) to consider floating-point values as approximations to
5338 infinitely precise real numbers. If you are doing this, then you need
5339 to compute (by analyzing the code, or in some other way) the maximum or
5340 likely maximum error that the computation introduces, and allow for it
5341 when performing comparisons (and when producing output, but that's a
5342 different problem). In particular, instead of testing for equality, you
5343 should check to see whether the two values have ranges that overlap; and
5344 this is done with the relational operators, so equality comparisons are
5347 @item -Wtraditional @r{(C and Objective-C only)}
5348 @opindex Wtraditional
5349 @opindex Wno-traditional
5350 Warn about certain constructs that behave differently in traditional and
5351 ISO C@. Also warn about ISO C constructs that have no traditional C
5352 equivalent, and/or problematic constructs that should be avoided.
5356 Macro parameters that appear within string literals in the macro body.
5357 In traditional C macro replacement takes place within string literals,
5358 but in ISO C it does not.
5361 In traditional C, some preprocessor directives did not exist.
5362 Traditional preprocessors only considered a line to be a directive
5363 if the @samp{#} appeared in column 1 on the line. Therefore
5364 @option{-Wtraditional} warns about directives that traditional C
5365 understands but ignores because the @samp{#} does not appear as the
5366 first character on the line. It also suggests you hide directives like
5367 @code{#pragma} not understood by traditional C by indenting them. Some
5368 traditional implementations do not recognize @code{#elif}, so this option
5369 suggests avoiding it altogether.
5372 A function-like macro that appears without arguments.
5375 The unary plus operator.
5378 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
5379 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
5380 constants.) Note, these suffixes appear in macros defined in the system
5381 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
5382 Use of these macros in user code might normally lead to spurious
5383 warnings, however GCC's integrated preprocessor has enough context to
5384 avoid warning in these cases.
5387 A function declared external in one block and then used after the end of
5391 A @code{switch} statement has an operand of type @code{long}.
5394 A non-@code{static} function declaration follows a @code{static} one.
5395 This construct is not accepted by some traditional C compilers.
5398 The ISO type of an integer constant has a different width or
5399 signedness from its traditional type. This warning is only issued if
5400 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
5401 typically represent bit patterns, are not warned about.
5404 Usage of ISO string concatenation is detected.
5407 Initialization of automatic aggregates.
5410 Identifier conflicts with labels. Traditional C lacks a separate
5411 namespace for labels.
5414 Initialization of unions. If the initializer is zero, the warning is
5415 omitted. This is done under the assumption that the zero initializer in
5416 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
5417 initializer warnings and relies on default initialization to zero in the
5421 Conversions by prototypes between fixed/floating-point values and vice
5422 versa. The absence of these prototypes when compiling with traditional
5423 C causes serious problems. This is a subset of the possible
5424 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
5427 Use of ISO C style function definitions. This warning intentionally is
5428 @emph{not} issued for prototype declarations or variadic functions
5429 because these ISO C features appear in your code when using
5430 libiberty's traditional C compatibility macros, @code{PARAMS} and
5431 @code{VPARAMS}. This warning is also bypassed for nested functions
5432 because that feature is already a GCC extension and thus not relevant to
5433 traditional C compatibility.
5436 @item -Wtraditional-conversion @r{(C and Objective-C only)}
5437 @opindex Wtraditional-conversion
5438 @opindex Wno-traditional-conversion
5439 Warn if a prototype causes a type conversion that is different from what
5440 would happen to the same argument in the absence of a prototype. This
5441 includes conversions of fixed point to floating and vice versa, and
5442 conversions changing the width or signedness of a fixed-point argument
5443 except when the same as the default promotion.
5445 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
5446 @opindex Wdeclaration-after-statement
5447 @opindex Wno-declaration-after-statement
5448 Warn when a declaration is found after a statement in a block. This
5449 construct, known from C++, was introduced with ISO C99 and is by default
5450 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
5452 @item -Wexpansion-to-defined
5453 @opindex Wexpansion-to-defined
5454 Warn whenever @samp{defined} is encountered in the expansion of a macro.
5455 This warning is also enabled by @option{-Wpedantic} and @option{-Wextra}.
5460 Warn if an undefined identifier is evaluated in an @code{#if} directive.
5462 @item -Wno-endif-labels
5463 @opindex Wno-endif-labels
5464 @opindex Wendif-labels
5465 Do not warn whenever an @code{#else} or an @code{#endif} are followed by text.
5470 Warn whenever a local variable or type declaration shadows another
5471 variable, parameter, type, class member (in C++), or instance variable
5472 (in Objective-C) or whenever a built-in function is shadowed. Note
5473 that in C++, the compiler warns if a local variable shadows an
5474 explicit typedef, but not if it shadows a struct/class/enum.
5475 Same as @option{-Wshadow=global}.
5477 @item -Wno-shadow-ivar @r{(Objective-C only)}
5478 @opindex Wno-shadow-ivar
5479 @opindex Wshadow-ivar
5480 Do not warn whenever a local variable shadows an instance variable in an
5483 @item -Wshadow=global
5484 @opindex Wshadow=local
5485 The default for @option{-Wshadow}. Warns for any (global) shadowing.
5487 @item -Wshadow=local
5488 @opindex Wshadow=local
5489 Warn when a local variable shadows another local variable or parameter.
5490 This warning is enabled by @option{-Wshadow=global}.
5492 @item -Wshadow=compatible-local
5493 @opindex Wshadow=compatible-local
5494 Warn when a local variable shadows another local variable or parameter
5495 whose type is compatible with that of the shadowing variable. In C++,
5496 type compatibility here means the type of the shadowing variable can be
5497 converted to that of the shadowed variable. The creation of this flag
5498 (in addition to @option{-Wshadow=local}) is based on the idea that when
5499 a local variable shadows another one of incompatible type, it is most
5500 likely intentional, not a bug or typo, as shown in the following example:
5504 for (SomeIterator i = SomeObj.begin(); i != SomeObj.end(); ++i)
5506 for (int i = 0; i < N; ++i)
5515 Since the two variable @code{i} in the example above have incompatible types,
5516 enabling only @option{-Wshadow=compatible-local} will not emit a warning.
5517 Because their types are incompatible, if a programmer accidentally uses one
5518 in place of the other, type checking will catch that and emit an error or
5519 warning. So not warning (about shadowing) in this case will not lead to
5520 undetected bugs. Use of this flag instead of @option{-Wshadow=local} can
5521 possibly reduce the number of warnings triggered by intentional shadowing.
5523 This warning is enabled by @option{-Wshadow=local}.
5525 @item -Wlarger-than=@var{len}
5526 @opindex Wlarger-than=@var{len}
5527 @opindex Wlarger-than-@var{len}
5528 Warn whenever an object of larger than @var{len} bytes is defined.
5530 @item -Wframe-larger-than=@var{len}
5531 @opindex Wframe-larger-than
5532 Warn if the size of a function frame is larger than @var{len} bytes.
5533 The computation done to determine the stack frame size is approximate
5534 and not conservative.
5535 The actual requirements may be somewhat greater than @var{len}
5536 even if you do not get a warning. In addition, any space allocated
5537 via @code{alloca}, variable-length arrays, or related constructs
5538 is not included by the compiler when determining
5539 whether or not to issue a warning.
5541 @item -Wno-free-nonheap-object
5542 @opindex Wno-free-nonheap-object
5543 @opindex Wfree-nonheap-object
5544 Do not warn when attempting to free an object that was not allocated
5547 @item -Wstack-usage=@var{len}
5548 @opindex Wstack-usage
5549 Warn if the stack usage of a function might be larger than @var{len} bytes.
5550 The computation done to determine the stack usage is conservative.
5551 Any space allocated via @code{alloca}, variable-length arrays, or related
5552 constructs is included by the compiler when determining whether or not to
5555 The message is in keeping with the output of @option{-fstack-usage}.
5559 If the stack usage is fully static but exceeds the specified amount, it's:
5562 warning: stack usage is 1120 bytes
5565 If the stack usage is (partly) dynamic but bounded, it's:
5568 warning: stack usage might be 1648 bytes
5571 If the stack usage is (partly) dynamic and not bounded, it's:
5574 warning: stack usage might be unbounded
5578 @item -Wunsafe-loop-optimizations
5579 @opindex Wunsafe-loop-optimizations
5580 @opindex Wno-unsafe-loop-optimizations
5581 Warn if the loop cannot be optimized because the compiler cannot
5582 assume anything on the bounds of the loop indices. With
5583 @option{-funsafe-loop-optimizations} warn if the compiler makes
5586 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
5587 @opindex Wno-pedantic-ms-format
5588 @opindex Wpedantic-ms-format
5589 When used in combination with @option{-Wformat}
5590 and @option{-pedantic} without GNU extensions, this option
5591 disables the warnings about non-ISO @code{printf} / @code{scanf} format
5592 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
5593 which depend on the MS runtime.
5596 @opindex Waligned-new
5597 @opindex Wno-aligned-new
5598 Warn about a new-expression of a type that requires greater alignment
5599 than the @code{alignof(std::max_align_t)} but uses an allocation
5600 function without an explicit alignment parameter. This option is
5601 enabled by @option{-Wall}.
5603 Normally this only warns about global allocation functions, but
5604 @option{-Waligned-new=all} also warns about class member allocation
5607 @item -Wplacement-new
5608 @itemx -Wplacement-new=@var{n}
5609 @opindex Wplacement-new
5610 @opindex Wno-placement-new
5611 Warn about placement new expressions with undefined behavior, such as
5612 constructing an object in a buffer that is smaller than the type of
5613 the object. For example, the placement new expression below is diagnosed
5614 because it attempts to construct an array of 64 integers in a buffer only
5620 This warning is enabled by default.
5623 @item -Wplacement-new=1
5624 This is the default warning level of @option{-Wplacement-new}. At this
5625 level the warning is not issued for some strictly undefined constructs that
5626 GCC allows as extensions for compatibility with legacy code. For example,
5627 the following @code{new} expression is not diagnosed at this level even
5628 though it has undefined behavior according to the C++ standard because
5629 it writes past the end of the one-element array.
5631 struct S @{ int n, a[1]; @};
5632 S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]);
5633 new (s->a)int [32]();
5636 @item -Wplacement-new=2
5637 At this level, in addition to diagnosing all the same constructs as at level
5638 1, a diagnostic is also issued for placement new expressions that construct
5639 an object in the last member of structure whose type is an array of a single
5640 element and whose size is less than the size of the object being constructed.
5641 While the previous example would be diagnosed, the following construct makes
5642 use of the flexible member array extension to avoid the warning at level 2.
5644 struct S @{ int n, a[]; @};
5645 S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]);
5646 new (s->a)int [32]();
5651 @item -Wpointer-arith
5652 @opindex Wpointer-arith
5653 @opindex Wno-pointer-arith
5654 Warn about anything that depends on the ``size of'' a function type or
5655 of @code{void}. GNU C assigns these types a size of 1, for
5656 convenience in calculations with @code{void *} pointers and pointers
5657 to functions. In C++, warn also when an arithmetic operation involves
5658 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
5661 @opindex Wtype-limits
5662 @opindex Wno-type-limits
5663 Warn if a comparison is always true or always false due to the limited
5664 range of the data type, but do not warn for constant expressions. For
5665 example, warn if an unsigned variable is compared against zero with
5666 @code{<} or @code{>=}. This warning is also enabled by
5669 @item -Wbad-function-cast @r{(C and Objective-C only)}
5670 @opindex Wbad-function-cast
5671 @opindex Wno-bad-function-cast
5672 Warn when a function call is cast to a non-matching type.
5673 For example, warn if a call to a function returning an integer type
5674 is cast to a pointer type.
5676 @item -Wc90-c99-compat @r{(C and Objective-C only)}
5677 @opindex Wc90-c99-compat
5678 @opindex Wno-c90-c99-compat
5679 Warn about features not present in ISO C90, but present in ISO C99.
5680 For instance, warn about use of variable length arrays, @code{long long}
5681 type, @code{bool} type, compound literals, designated initializers, and so
5682 on. This option is independent of the standards mode. Warnings are disabled
5683 in the expression that follows @code{__extension__}.
5685 @item -Wc99-c11-compat @r{(C and Objective-C only)}
5686 @opindex Wc99-c11-compat
5687 @opindex Wno-c99-c11-compat
5688 Warn about features not present in ISO C99, but present in ISO C11.
5689 For instance, warn about use of anonymous structures and unions,
5690 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
5691 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
5692 and so on. This option is independent of the standards mode. Warnings are
5693 disabled in the expression that follows @code{__extension__}.
5695 @item -Wc++-compat @r{(C and Objective-C only)}
5696 @opindex Wc++-compat
5697 Warn about ISO C constructs that are outside of the common subset of
5698 ISO C and ISO C++, e.g.@: request for implicit conversion from
5699 @code{void *} to a pointer to non-@code{void} type.
5701 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
5702 @opindex Wc++11-compat
5703 Warn about C++ constructs whose meaning differs between ISO C++ 1998
5704 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
5705 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
5706 enabled by @option{-Wall}.
5708 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
5709 @opindex Wc++14-compat
5710 Warn about C++ constructs whose meaning differs between ISO C++ 2011
5711 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
5713 @item -Wc++1z-compat @r{(C++ and Objective-C++ only)}
5714 @opindex Wc++1z-compat
5715 Warn about C++ constructs whose meaning differs between ISO C++ 2014
5716 and the forthoming ISO C++ 2017(?). This warning is enabled by @option{-Wall}.
5720 @opindex Wno-cast-qual
5721 Warn whenever a pointer is cast so as to remove a type qualifier from
5722 the target type. For example, warn if a @code{const char *} is cast
5723 to an ordinary @code{char *}.
5725 Also warn when making a cast that introduces a type qualifier in an
5726 unsafe way. For example, casting @code{char **} to @code{const char **}
5727 is unsafe, as in this example:
5730 /* p is char ** value. */
5731 const char **q = (const char **) p;
5732 /* Assignment of readonly string to const char * is OK. */
5734 /* Now char** pointer points to read-only memory. */
5739 @opindex Wcast-align
5740 @opindex Wno-cast-align
5741 Warn whenever a pointer is cast such that the required alignment of the
5742 target is increased. For example, warn if a @code{char *} is cast to
5743 an @code{int *} on machines where integers can only be accessed at
5744 two- or four-byte boundaries.
5746 @item -Wwrite-strings
5747 @opindex Wwrite-strings
5748 @opindex Wno-write-strings
5749 When compiling C, give string constants the type @code{const
5750 char[@var{length}]} so that copying the address of one into a
5751 non-@code{const} @code{char *} pointer produces a warning. These
5752 warnings help you find at compile time code that can try to write
5753 into a string constant, but only if you have been very careful about
5754 using @code{const} in declarations and prototypes. Otherwise, it is
5755 just a nuisance. This is why we did not make @option{-Wall} request
5758 When compiling C++, warn about the deprecated conversion from string
5759 literals to @code{char *}. This warning is enabled by default for C++
5764 @opindex Wno-clobbered
5765 Warn for variables that might be changed by @code{longjmp} or
5766 @code{vfork}. This warning is also enabled by @option{-Wextra}.
5768 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
5769 @opindex Wconditionally-supported
5770 @opindex Wno-conditionally-supported
5771 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
5774 @opindex Wconversion
5775 @opindex Wno-conversion
5776 Warn for implicit conversions that may alter a value. This includes
5777 conversions between real and integer, like @code{abs (x)} when
5778 @code{x} is @code{double}; conversions between signed and unsigned,
5779 like @code{unsigned ui = -1}; and conversions to smaller types, like
5780 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
5781 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
5782 changed by the conversion like in @code{abs (2.0)}. Warnings about
5783 conversions between signed and unsigned integers can be disabled by
5784 using @option{-Wno-sign-conversion}.
5786 For C++, also warn for confusing overload resolution for user-defined
5787 conversions; and conversions that never use a type conversion
5788 operator: conversions to @code{void}, the same type, a base class or a
5789 reference to them. Warnings about conversions between signed and
5790 unsigned integers are disabled by default in C++ unless
5791 @option{-Wsign-conversion} is explicitly enabled.
5793 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
5794 @opindex Wconversion-null
5795 @opindex Wno-conversion-null
5796 Do not warn for conversions between @code{NULL} and non-pointer
5797 types. @option{-Wconversion-null} is enabled by default.
5799 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
5800 @opindex Wzero-as-null-pointer-constant
5801 @opindex Wno-zero-as-null-pointer-constant
5802 Warn when a literal @samp{0} is used as null pointer constant. This can
5803 be useful to facilitate the conversion to @code{nullptr} in C++11.
5805 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
5806 @opindex Wsubobject-linkage
5807 @opindex Wno-subobject-linkage
5808 Warn if a class type has a base or a field whose type uses the anonymous
5809 namespace or depends on a type with no linkage. If a type A depends on
5810 a type B with no or internal linkage, defining it in multiple
5811 translation units would be an ODR violation because the meaning of B
5812 is different in each translation unit. If A only appears in a single
5813 translation unit, the best way to silence the warning is to give it
5814 internal linkage by putting it in an anonymous namespace as well. The
5815 compiler doesn't give this warning for types defined in the main .C
5816 file, as those are unlikely to have multiple definitions.
5817 @option{-Wsubobject-linkage} is enabled by default.
5819 @item -Wdangling-else
5820 @opindex Wdangling-else
5821 @opindex Wno-dangling-else
5822 Warn about constructions where there may be confusion to which
5823 @code{if} statement an @code{else} branch belongs. Here is an example of
5838 In C/C++, every @code{else} branch belongs to the innermost possible
5839 @code{if} statement, which in this example is @code{if (b)}. This is
5840 often not what the programmer expected, as illustrated in the above
5841 example by indentation the programmer chose. When there is the
5842 potential for this confusion, GCC issues a warning when this flag
5843 is specified. To eliminate the warning, add explicit braces around
5844 the innermost @code{if} statement so there is no way the @code{else}
5845 can belong to the enclosing @code{if}. The resulting code
5862 This warning is enabled by @option{-Wparentheses}.
5866 @opindex Wno-date-time
5867 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
5868 are encountered as they might prevent bit-wise-identical reproducible
5871 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
5872 @opindex Wdelete-incomplete
5873 @opindex Wno-delete-incomplete
5874 Warn when deleting a pointer to incomplete type, which may cause
5875 undefined behavior at runtime. This warning is enabled by default.
5877 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
5878 @opindex Wuseless-cast
5879 @opindex Wno-useless-cast
5880 Warn when an expression is casted to its own type.
5883 @opindex Wempty-body
5884 @opindex Wno-empty-body
5885 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
5886 while} statement. This warning is also enabled by @option{-Wextra}.
5888 @item -Wenum-compare
5889 @opindex Wenum-compare
5890 @opindex Wno-enum-compare
5891 Warn about a comparison between values of different enumerated types.
5892 In C++ enumeral mismatches in conditional expressions are also
5893 diagnosed and the warning is enabled by default. In C this warning is
5894 enabled by @option{-Wall}.
5896 @item -Wjump-misses-init @r{(C, Objective-C only)}
5897 @opindex Wjump-misses-init
5898 @opindex Wno-jump-misses-init
5899 Warn if a @code{goto} statement or a @code{switch} statement jumps
5900 forward across the initialization of a variable, or jumps backward to a
5901 label after the variable has been initialized. This only warns about
5902 variables that are initialized when they are declared. This warning is
5903 only supported for C and Objective-C; in C++ this sort of branch is an
5906 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
5907 can be disabled with the @option{-Wno-jump-misses-init} option.
5909 @item -Wsign-compare
5910 @opindex Wsign-compare
5911 @opindex Wno-sign-compare
5912 @cindex warning for comparison of signed and unsigned values
5913 @cindex comparison of signed and unsigned values, warning
5914 @cindex signed and unsigned values, comparison warning
5915 Warn when a comparison between signed and unsigned values could produce
5916 an incorrect result when the signed value is converted to unsigned.
5917 In C++, this warning is also enabled by @option{-Wall}. In C, it is
5918 also enabled by @option{-Wextra}.
5920 @item -Wsign-conversion
5921 @opindex Wsign-conversion
5922 @opindex Wno-sign-conversion
5923 Warn for implicit conversions that may change the sign of an integer
5924 value, like assigning a signed integer expression to an unsigned
5925 integer variable. An explicit cast silences the warning. In C, this
5926 option is enabled also by @option{-Wconversion}.
5928 @item -Wfloat-conversion
5929 @opindex Wfloat-conversion
5930 @opindex Wno-float-conversion
5931 Warn for implicit conversions that reduce the precision of a real value.
5932 This includes conversions from real to integer, and from higher precision
5933 real to lower precision real values. This option is also enabled by
5934 @option{-Wconversion}.
5936 @item -Wno-scalar-storage-order
5937 @opindex -Wno-scalar-storage-order
5938 @opindex -Wscalar-storage-order
5939 Do not warn on suspicious constructs involving reverse scalar storage order.
5941 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
5942 @opindex Wsized-deallocation
5943 @opindex Wno-sized-deallocation
5944 Warn about a definition of an unsized deallocation function
5946 void operator delete (void *) noexcept;
5947 void operator delete[] (void *) noexcept;
5949 without a definition of the corresponding sized deallocation function
5951 void operator delete (void *, std::size_t) noexcept;
5952 void operator delete[] (void *, std::size_t) noexcept;
5954 or vice versa. Enabled by @option{-Wextra} along with
5955 @option{-fsized-deallocation}.
5957 @item -Wsizeof-pointer-memaccess
5958 @opindex Wsizeof-pointer-memaccess
5959 @opindex Wno-sizeof-pointer-memaccess
5960 Warn for suspicious length parameters to certain string and memory built-in
5961 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
5962 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
5963 but a pointer, and suggests a possible fix, or about
5964 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
5967 @item -Wsizeof-array-argument
5968 @opindex Wsizeof-array-argument
5969 @opindex Wno-sizeof-array-argument
5970 Warn when the @code{sizeof} operator is applied to a parameter that is
5971 declared as an array in a function definition. This warning is enabled by
5972 default for C and C++ programs.
5974 @item -Wmemset-elt-size
5975 @opindex Wmemset-elt-size
5976 @opindex Wno-memset-elt-size
5977 Warn for suspicious calls to the @code{memset} built-in function, if the
5978 first argument references an array, and the third argument is a number
5979 equal to the number of elements, but not equal to the size of the array
5980 in memory. This indicates that the user has omitted a multiplication by
5981 the element size. This warning is enabled by @option{-Wall}.
5983 @item -Wmemset-transposed-args
5984 @opindex Wmemset-transposed-args
5985 @opindex Wno-memset-transposed-args
5986 Warn for suspicious calls to the @code{memset} built-in function, if the
5987 second argument is not zero and the third argument is zero. This warns e.g.@
5988 about @code{memset (buf, sizeof buf, 0)} where most probably
5989 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
5990 is only emitted if the third argument is literal zero. If it is some
5991 expression that is folded to zero, a cast of zero to some type, etc.,
5992 it is far less likely that the user has mistakenly exchanged the arguments
5993 and no warning is emitted. This warning is enabled by @option{-Wall}.
5997 @opindex Wno-address
5998 Warn about suspicious uses of memory addresses. These include using
5999 the address of a function in a conditional expression, such as
6000 @code{void func(void); if (func)}, and comparisons against the memory
6001 address of a string literal, such as @code{if (x == "abc")}. Such
6002 uses typically indicate a programmer error: the address of a function
6003 always evaluates to true, so their use in a conditional usually
6004 indicate that the programmer forgot the parentheses in a function
6005 call; and comparisons against string literals result in unspecified
6006 behavior and are not portable in C, so they usually indicate that the
6007 programmer intended to use @code{strcmp}. This warning is enabled by
6011 @opindex Wlogical-op
6012 @opindex Wno-logical-op
6013 Warn about suspicious uses of logical operators in expressions.
6014 This includes using logical operators in contexts where a
6015 bit-wise operator is likely to be expected. Also warns when
6016 the operands of a logical operator are the same:
6019 if (a < 0 && a < 0) @{ @dots{} @}
6022 @item -Wlogical-not-parentheses
6023 @opindex Wlogical-not-parentheses
6024 @opindex Wno-logical-not-parentheses
6025 Warn about logical not used on the left hand side operand of a comparison.
6026 This option does not warn if the right operand is considered to be a boolean
6027 expression. Its purpose is to detect suspicious code like the following:
6031 if (!a > 1) @{ @dots{} @}
6034 It is possible to suppress the warning by wrapping the LHS into
6037 if ((!a) > 1) @{ @dots{} @}
6040 This warning is enabled by @option{-Wall}.
6042 @item -Waggregate-return
6043 @opindex Waggregate-return
6044 @opindex Wno-aggregate-return
6045 Warn if any functions that return structures or unions are defined or
6046 called. (In languages where you can return an array, this also elicits
6049 @item -Wno-aggressive-loop-optimizations
6050 @opindex Wno-aggressive-loop-optimizations
6051 @opindex Waggressive-loop-optimizations
6052 Warn if in a loop with constant number of iterations the compiler detects
6053 undefined behavior in some statement during one or more of the iterations.
6055 @item -Wno-attributes
6056 @opindex Wno-attributes
6057 @opindex Wattributes
6058 Do not warn if an unexpected @code{__attribute__} is used, such as
6059 unrecognized attributes, function attributes applied to variables,
6060 etc. This does not stop errors for incorrect use of supported
6063 @item -Wno-builtin-declaration-mismatch
6064 @opindex Wno-builtin-declaration-mismatch
6065 @opindex Wbuiltin-declaration-mismatch
6066 Warn if a built-in function is declared with the wrong signature.
6067 This warning is enabled by default.
6069 @item -Wno-builtin-macro-redefined
6070 @opindex Wno-builtin-macro-redefined
6071 @opindex Wbuiltin-macro-redefined
6072 Do not warn if certain built-in macros are redefined. This suppresses
6073 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
6074 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
6076 @item -Wstrict-prototypes @r{(C and Objective-C only)}
6077 @opindex Wstrict-prototypes
6078 @opindex Wno-strict-prototypes
6079 Warn if a function is declared or defined without specifying the
6080 argument types. (An old-style function definition is permitted without
6081 a warning if preceded by a declaration that specifies the argument
6084 @item -Wold-style-declaration @r{(C and Objective-C only)}
6085 @opindex Wold-style-declaration
6086 @opindex Wno-old-style-declaration
6087 Warn for obsolescent usages, according to the C Standard, in a
6088 declaration. For example, warn if storage-class specifiers like
6089 @code{static} are not the first things in a declaration. This warning
6090 is also enabled by @option{-Wextra}.
6092 @item -Wold-style-definition @r{(C and Objective-C only)}
6093 @opindex Wold-style-definition
6094 @opindex Wno-old-style-definition
6095 Warn if an old-style function definition is used. A warning is given
6096 even if there is a previous prototype.
6098 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
6099 @opindex Wmissing-parameter-type
6100 @opindex Wno-missing-parameter-type
6101 A function parameter is declared without a type specifier in K&R-style
6108 This warning is also enabled by @option{-Wextra}.
6110 @item -Wmissing-prototypes @r{(C and Objective-C only)}
6111 @opindex Wmissing-prototypes
6112 @opindex Wno-missing-prototypes
6113 Warn if a global function is defined without a previous prototype
6114 declaration. This warning is issued even if the definition itself
6115 provides a prototype. Use this option to detect global functions
6116 that do not have a matching prototype declaration in a header file.
6117 This option is not valid for C++ because all function declarations
6118 provide prototypes and a non-matching declaration declares an
6119 overload rather than conflict with an earlier declaration.
6120 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
6122 @item -Wmissing-declarations
6123 @opindex Wmissing-declarations
6124 @opindex Wno-missing-declarations
6125 Warn if a global function is defined without a previous declaration.
6126 Do so even if the definition itself provides a prototype.
6127 Use this option to detect global functions that are not declared in
6128 header files. In C, no warnings are issued for functions with previous
6129 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
6130 missing prototypes. In C++, no warnings are issued for function templates,
6131 or for inline functions, or for functions in anonymous namespaces.
6133 @item -Wmissing-field-initializers
6134 @opindex Wmissing-field-initializers
6135 @opindex Wno-missing-field-initializers
6139 Warn if a structure's initializer has some fields missing. For
6140 example, the following code causes such a warning, because
6141 @code{x.h} is implicitly zero:
6144 struct s @{ int f, g, h; @};
6145 struct s x = @{ 3, 4 @};
6148 This option does not warn about designated initializers, so the following
6149 modification does not trigger a warning:
6152 struct s @{ int f, g, h; @};
6153 struct s x = @{ .f = 3, .g = 4 @};
6156 In C++ this option does not warn either about the empty @{ @}
6157 initializer, for example:
6160 struct s @{ int f, g, h; @};
6164 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
6165 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
6167 @item -Wno-multichar
6168 @opindex Wno-multichar
6170 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
6171 Usually they indicate a typo in the user's code, as they have
6172 implementation-defined values, and should not be used in portable code.
6174 @item -Wnormalized@r{[}=@r{<}none@r{|}id@r{|}nfc@r{|}nfkc@r{>]}
6175 @opindex Wnormalized=
6176 @opindex Wnormalized
6177 @opindex Wno-normalized
6180 @cindex character set, input normalization
6181 In ISO C and ISO C++, two identifiers are different if they are
6182 different sequences of characters. However, sometimes when characters
6183 outside the basic ASCII character set are used, you can have two
6184 different character sequences that look the same. To avoid confusion,
6185 the ISO 10646 standard sets out some @dfn{normalization rules} which
6186 when applied ensure that two sequences that look the same are turned into
6187 the same sequence. GCC can warn you if you are using identifiers that
6188 have not been normalized; this option controls that warning.
6190 There are four levels of warning supported by GCC@. The default is
6191 @option{-Wnormalized=nfc}, which warns about any identifier that is
6192 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
6193 recommended form for most uses. It is equivalent to
6194 @option{-Wnormalized}.
6196 Unfortunately, there are some characters allowed in identifiers by
6197 ISO C and ISO C++ that, when turned into NFC, are not allowed in
6198 identifiers. That is, there's no way to use these symbols in portable
6199 ISO C or C++ and have all your identifiers in NFC@.
6200 @option{-Wnormalized=id} suppresses the warning for these characters.
6201 It is hoped that future versions of the standards involved will correct
6202 this, which is why this option is not the default.
6204 You can switch the warning off for all characters by writing
6205 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
6206 only do this if you are using some other normalization scheme (like
6207 ``D''), because otherwise you can easily create bugs that are
6208 literally impossible to see.
6210 Some characters in ISO 10646 have distinct meanings but look identical
6211 in some fonts or display methodologies, especially once formatting has
6212 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
6213 LETTER N'', displays just like a regular @code{n} that has been
6214 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
6215 normalization scheme to convert all these into a standard form as
6216 well, and GCC warns if your code is not in NFKC if you use
6217 @option{-Wnormalized=nfkc}. This warning is comparable to warning
6218 about every identifier that contains the letter O because it might be
6219 confused with the digit 0, and so is not the default, but may be
6220 useful as a local coding convention if the programming environment
6221 cannot be fixed to display these characters distinctly.
6223 @item -Wno-deprecated
6224 @opindex Wno-deprecated
6225 @opindex Wdeprecated
6226 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
6228 @item -Wno-deprecated-declarations
6229 @opindex Wno-deprecated-declarations
6230 @opindex Wdeprecated-declarations
6231 Do not warn about uses of functions (@pxref{Function Attributes}),
6232 variables (@pxref{Variable Attributes}), and types (@pxref{Type
6233 Attributes}) marked as deprecated by using the @code{deprecated}
6237 @opindex Wno-overflow
6239 Do not warn about compile-time overflow in constant expressions.
6244 Warn about One Definition Rule violations during link-time optimization.
6245 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
6248 @opindex Wopenm-simd
6249 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
6250 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
6251 option can be used to relax the cost model.
6253 @item -Woverride-init @r{(C and Objective-C only)}
6254 @opindex Woverride-init
6255 @opindex Wno-override-init
6259 Warn if an initialized field without side effects is overridden when
6260 using designated initializers (@pxref{Designated Inits, , Designated
6263 This warning is included in @option{-Wextra}. To get other
6264 @option{-Wextra} warnings without this one, use @option{-Wextra
6265 -Wno-override-init}.
6267 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
6268 @opindex Woverride-init-side-effects
6269 @opindex Wno-override-init-side-effects
6270 Warn if an initialized field with side effects is overridden when
6271 using designated initializers (@pxref{Designated Inits, , Designated
6272 Initializers}). This warning is enabled by default.
6277 Warn if a structure is given the packed attribute, but the packed
6278 attribute has no effect on the layout or size of the structure.
6279 Such structures may be mis-aligned for little benefit. For
6280 instance, in this code, the variable @code{f.x} in @code{struct bar}
6281 is misaligned even though @code{struct bar} does not itself
6282 have the packed attribute:
6289 @} __attribute__((packed));
6297 @item -Wpacked-bitfield-compat
6298 @opindex Wpacked-bitfield-compat
6299 @opindex Wno-packed-bitfield-compat
6300 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
6301 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
6302 the change can lead to differences in the structure layout. GCC
6303 informs you when the offset of such a field has changed in GCC 4.4.
6304 For example there is no longer a 4-bit padding between field @code{a}
6305 and @code{b} in this structure:
6312 @} __attribute__ ((packed));
6315 This warning is enabled by default. Use
6316 @option{-Wno-packed-bitfield-compat} to disable this warning.
6321 Warn if padding is included in a structure, either to align an element
6322 of the structure or to align the whole structure. Sometimes when this
6323 happens it is possible to rearrange the fields of the structure to
6324 reduce the padding and so make the structure smaller.
6326 @item -Wredundant-decls
6327 @opindex Wredundant-decls
6328 @opindex Wno-redundant-decls
6329 Warn if anything is declared more than once in the same scope, even in
6330 cases where multiple declaration is valid and changes nothing.
6334 @opindex Wno-restrict
6335 Warn when an argument passed to a restrict-qualified parameter
6336 aliases with another argument.
6338 @item -Wnested-externs @r{(C and Objective-C only)}
6339 @opindex Wnested-externs
6340 @opindex Wno-nested-externs
6341 Warn if an @code{extern} declaration is encountered within a function.
6343 @item -Wno-inherited-variadic-ctor
6344 @opindex Winherited-variadic-ctor
6345 @opindex Wno-inherited-variadic-ctor
6346 Suppress warnings about use of C++11 inheriting constructors when the
6347 base class inherited from has a C variadic constructor; the warning is
6348 on by default because the ellipsis is not inherited.
6353 Warn if a function that is declared as inline cannot be inlined.
6354 Even with this option, the compiler does not warn about failures to
6355 inline functions declared in system headers.
6357 The compiler uses a variety of heuristics to determine whether or not
6358 to inline a function. For example, the compiler takes into account
6359 the size of the function being inlined and the amount of inlining
6360 that has already been done in the current function. Therefore,
6361 seemingly insignificant changes in the source program can cause the
6362 warnings produced by @option{-Winline} to appear or disappear.
6364 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
6365 @opindex Wno-invalid-offsetof
6366 @opindex Winvalid-offsetof
6367 Suppress warnings from applying the @code{offsetof} macro to a non-POD
6368 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
6369 to a non-standard-layout type is undefined. In existing C++ implementations,
6370 however, @code{offsetof} typically gives meaningful results.
6371 This flag is for users who are aware that they are
6372 writing nonportable code and who have deliberately chosen to ignore the
6375 The restrictions on @code{offsetof} may be relaxed in a future version
6376 of the C++ standard.
6378 @item -Wint-in-bool-context
6379 @opindex Wint-in-bool-context
6380 @opindex Wno-int-in-bool-context
6381 Warn for suspicious use of integer values where boolean values are expected,
6382 such as conditional expressions (?:) using non-boolean integer constants in
6383 boolean context, like @code{if (a <= b ? 2 : 3)}. Or left shifting of signed
6384 integers in boolean context, like @code{for (a = 0; 1 << a; a++);}. Likewise
6385 for all kinds of multiplications regardless of the data type.
6386 This warning is enabled by @option{-Wall}.
6388 @item -Wno-int-to-pointer-cast
6389 @opindex Wno-int-to-pointer-cast
6390 @opindex Wint-to-pointer-cast
6391 Suppress warnings from casts to pointer type of an integer of a
6392 different size. In C++, casting to a pointer type of smaller size is
6393 an error. @option{Wint-to-pointer-cast} is enabled by default.
6396 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
6397 @opindex Wno-pointer-to-int-cast
6398 @opindex Wpointer-to-int-cast
6399 Suppress warnings from casts from a pointer to an integer type of a
6403 @opindex Winvalid-pch
6404 @opindex Wno-invalid-pch
6405 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
6406 the search path but can't be used.
6410 @opindex Wno-long-long
6411 Warn if @code{long long} type is used. This is enabled by either
6412 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
6413 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
6415 @item -Wvariadic-macros
6416 @opindex Wvariadic-macros
6417 @opindex Wno-variadic-macros
6418 Warn if variadic macros are used in ISO C90 mode, or if the GNU
6419 alternate syntax is used in ISO C99 mode. This is enabled by either
6420 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
6421 messages, use @option{-Wno-variadic-macros}.
6425 @opindex Wno-varargs
6426 Warn upon questionable usage of the macros used to handle variable
6427 arguments like @code{va_start}. This is default. To inhibit the
6428 warning messages, use @option{-Wno-varargs}.
6430 @item -Wvector-operation-performance
6431 @opindex Wvector-operation-performance
6432 @opindex Wno-vector-operation-performance
6433 Warn if vector operation is not implemented via SIMD capabilities of the
6434 architecture. Mainly useful for the performance tuning.
6435 Vector operation can be implemented @code{piecewise}, which means that the
6436 scalar operation is performed on every vector element;
6437 @code{in parallel}, which means that the vector operation is implemented
6438 using scalars of wider type, which normally is more performance efficient;
6439 and @code{as a single scalar}, which means that vector fits into a
6442 @item -Wno-virtual-move-assign
6443 @opindex Wvirtual-move-assign
6444 @opindex Wno-virtual-move-assign
6445 Suppress warnings about inheriting from a virtual base with a
6446 non-trivial C++11 move assignment operator. This is dangerous because
6447 if the virtual base is reachable along more than one path, it is
6448 moved multiple times, which can mean both objects end up in the
6449 moved-from state. If the move assignment operator is written to avoid
6450 moving from a moved-from object, this warning can be disabled.
6455 Warn if a variable-length array is used in the code.
6456 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
6457 the variable-length array.
6459 @item -Wvla-larger-than=@var{n}
6460 If this option is used, the compiler will warn on uses of
6461 variable-length arrays where the size is either unbounded, or bounded
6462 by an argument that can be larger than @var{n} bytes. This is similar
6463 to how @option{-Walloca-larger-than=@var{n}} works, but with
6464 variable-length arrays.
6466 Note that GCC may optimize small variable-length arrays of a known
6467 value into plain arrays, so this warning may not get triggered for
6470 This warning is not enabled by @option{-Wall}, and is only active when
6471 @option{-ftree-vrp} is active (default for @option{-O2} and above).
6473 See also @option{-Walloca-larger-than=@var{n}}.
6475 @item -Wvolatile-register-var
6476 @opindex Wvolatile-register-var
6477 @opindex Wno-volatile-register-var
6478 Warn if a register variable is declared volatile. The volatile
6479 modifier does not inhibit all optimizations that may eliminate reads
6480 and/or writes to register variables. This warning is enabled by
6483 @item -Wdisabled-optimization
6484 @opindex Wdisabled-optimization
6485 @opindex Wno-disabled-optimization
6486 Warn if a requested optimization pass is disabled. This warning does
6487 not generally indicate that there is anything wrong with your code; it
6488 merely indicates that GCC's optimizers are unable to handle the code
6489 effectively. Often, the problem is that your code is too big or too
6490 complex; GCC refuses to optimize programs when the optimization
6491 itself is likely to take inordinate amounts of time.
6493 @item -Wpointer-sign @r{(C and Objective-C only)}
6494 @opindex Wpointer-sign
6495 @opindex Wno-pointer-sign
6496 Warn for pointer argument passing or assignment with different signedness.
6497 This option is only supported for C and Objective-C@. It is implied by
6498 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
6499 @option{-Wno-pointer-sign}.
6501 @item -Wstack-protector
6502 @opindex Wstack-protector
6503 @opindex Wno-stack-protector
6504 This option is only active when @option{-fstack-protector} is active. It
6505 warns about functions that are not protected against stack smashing.
6507 @item -Woverlength-strings
6508 @opindex Woverlength-strings
6509 @opindex Wno-overlength-strings
6510 Warn about string constants that are longer than the ``minimum
6511 maximum'' length specified in the C standard. Modern compilers
6512 generally allow string constants that are much longer than the
6513 standard's minimum limit, but very portable programs should avoid
6514 using longer strings.
6516 The limit applies @emph{after} string constant concatenation, and does
6517 not count the trailing NUL@. In C90, the limit was 509 characters; in
6518 C99, it was raised to 4095. C++98 does not specify a normative
6519 minimum maximum, so we do not diagnose overlength strings in C++@.
6521 This option is implied by @option{-Wpedantic}, and can be disabled with
6522 @option{-Wno-overlength-strings}.
6524 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
6525 @opindex Wunsuffixed-float-constants
6527 Issue a warning for any floating constant that does not have
6528 a suffix. When used together with @option{-Wsystem-headers} it
6529 warns about such constants in system header files. This can be useful
6530 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
6531 from the decimal floating-point extension to C99.
6533 @item -Wno-designated-init @r{(C and Objective-C only)}
6534 Suppress warnings when a positional initializer is used to initialize
6535 a structure that has been marked with the @code{designated_init}
6539 Issue a warning when HSAIL cannot be emitted for the compiled function or
6544 @node Debugging Options
6545 @section Options for Debugging Your Program
6546 @cindex options, debugging
6547 @cindex debugging information options
6549 To tell GCC to emit extra information for use by a debugger, in almost
6550 all cases you need only to add @option{-g} to your other options.
6552 GCC allows you to use @option{-g} with
6553 @option{-O}. The shortcuts taken by optimized code may occasionally
6554 be surprising: some variables you declared may not exist
6555 at all; flow of control may briefly move where you did not expect it;
6556 some statements may not be executed because they compute constant
6557 results or their values are already at hand; some statements may
6558 execute in different places because they have been moved out of loops.
6559 Nevertheless it is possible to debug optimized output. This makes
6560 it reasonable to use the optimizer for programs that might have bugs.
6562 If you are not using some other optimization option, consider
6563 using @option{-Og} (@pxref{Optimize Options}) with @option{-g}.
6564 With no @option{-O} option at all, some compiler passes that collect
6565 information useful for debugging do not run at all, so that
6566 @option{-Og} may result in a better debugging experience.
6571 Produce debugging information in the operating system's native format
6572 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
6575 On most systems that use stabs format, @option{-g} enables use of extra
6576 debugging information that only GDB can use; this extra information
6577 makes debugging work better in GDB but probably makes other debuggers
6579 refuse to read the program. If you want to control for certain whether
6580 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
6581 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
6585 Produce debugging information for use by GDB@. This means to use the
6586 most expressive format available (DWARF, stabs, or the native format
6587 if neither of those are supported), including GDB extensions if at all
6591 @itemx -gdwarf-@var{version}
6593 Produce debugging information in DWARF format (if that is supported).
6594 The value of @var{version} may be either 2, 3, 4 or 5; the default version
6595 for most targets is 4. DWARF Version 5 is only experimental.
6597 Note that with DWARF Version 2, some ports require and always
6598 use some non-conflicting DWARF 3 extensions in the unwind tables.
6600 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
6601 for maximum benefit.
6603 GCC no longer supports DWARF Version 1, which is substantially
6604 different than Version 2 and later. For historical reasons, some
6605 other DWARF-related options (including @option{-feliminate-dwarf2-dups}
6606 and @option{-fno-dwarf2-cfi-asm}) retain a reference to DWARF Version 2
6607 in their names, but apply to all currently-supported versions of DWARF.
6611 Produce debugging information in stabs format (if that is supported),
6612 without GDB extensions. This is the format used by DBX on most BSD
6613 systems. On MIPS, Alpha and System V Release 4 systems this option
6614 produces stabs debugging output that is not understood by DBX or SDB@.
6615 On System V Release 4 systems this option requires the GNU assembler.
6619 Produce debugging information in stabs format (if that is supported),
6620 using GNU extensions understood only by the GNU debugger (GDB)@. The
6621 use of these extensions is likely to make other debuggers crash or
6622 refuse to read the program.
6626 Produce debugging information in COFF format (if that is supported).
6627 This is the format used by SDB on most System V systems prior to
6632 Produce debugging information in XCOFF format (if that is supported).
6633 This is the format used by the DBX debugger on IBM RS/6000 systems.
6637 Produce debugging information in XCOFF format (if that is supported),
6638 using GNU extensions understood only by the GNU debugger (GDB)@. The
6639 use of these extensions is likely to make other debuggers crash or
6640 refuse to read the program, and may cause assemblers other than the GNU
6641 assembler (GAS) to fail with an error.
6645 Produce debugging information in Alpha/VMS debug format (if that is
6646 supported). This is the format used by DEBUG on Alpha/VMS systems.
6649 @itemx -ggdb@var{level}
6650 @itemx -gstabs@var{level}
6651 @itemx -gcoff@var{level}
6652 @itemx -gxcoff@var{level}
6653 @itemx -gvms@var{level}
6654 Request debugging information and also use @var{level} to specify how
6655 much information. The default level is 2.
6657 Level 0 produces no debug information at all. Thus, @option{-g0} negates
6660 Level 1 produces minimal information, enough for making backtraces in
6661 parts of the program that you don't plan to debug. This includes
6662 descriptions of functions and external variables, and line number
6663 tables, but no information about local variables.
6665 Level 3 includes extra information, such as all the macro definitions
6666 present in the program. Some debuggers support macro expansion when
6667 you use @option{-g3}.
6669 @option{-gdwarf} does not accept a concatenated debug level, to avoid
6670 confusion with @option{-gdwarf-@var{level}}.
6671 Instead use an additional @option{-g@var{level}} option to change the
6672 debug level for DWARF.
6674 @item -feliminate-unused-debug-symbols
6675 @opindex feliminate-unused-debug-symbols
6676 Produce debugging information in stabs format (if that is supported),
6677 for only symbols that are actually used.
6679 @item -femit-class-debug-always
6680 @opindex femit-class-debug-always
6681 Instead of emitting debugging information for a C++ class in only one
6682 object file, emit it in all object files using the class. This option
6683 should be used only with debuggers that are unable to handle the way GCC
6684 normally emits debugging information for classes because using this
6685 option increases the size of debugging information by as much as a
6688 @item -fno-merge-debug-strings
6689 @opindex fmerge-debug-strings
6690 @opindex fno-merge-debug-strings
6691 Direct the linker to not merge together strings in the debugging
6692 information that are identical in different object files. Merging is
6693 not supported by all assemblers or linkers. Merging decreases the size
6694 of the debug information in the output file at the cost of increasing
6695 link processing time. Merging is enabled by default.
6697 @item -fdebug-prefix-map=@var{old}=@var{new}
6698 @opindex fdebug-prefix-map
6699 When compiling files in directory @file{@var{old}}, record debugging
6700 information describing them as in @file{@var{new}} instead.
6702 @item -fvar-tracking
6703 @opindex fvar-tracking
6704 Run variable tracking pass. It computes where variables are stored at each
6705 position in code. Better debugging information is then generated
6706 (if the debugging information format supports this information).
6708 It is enabled by default when compiling with optimization (@option{-Os},
6709 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
6710 the debug info format supports it.
6712 @item -fvar-tracking-assignments
6713 @opindex fvar-tracking-assignments
6714 @opindex fno-var-tracking-assignments
6715 Annotate assignments to user variables early in the compilation and
6716 attempt to carry the annotations over throughout the compilation all the
6717 way to the end, in an attempt to improve debug information while
6718 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
6720 It can be enabled even if var-tracking is disabled, in which case
6721 annotations are created and maintained, but discarded at the end.
6722 By default, this flag is enabled together with @option{-fvar-tracking},
6723 except when selective scheduling is enabled.
6726 @opindex gsplit-dwarf
6727 Separate as much DWARF debugging information as possible into a
6728 separate output file with the extension @file{.dwo}. This option allows
6729 the build system to avoid linking files with debug information. To
6730 be useful, this option requires a debugger capable of reading @file{.dwo}
6735 Generate DWARF @code{.debug_pubnames} and @code{.debug_pubtypes} sections.
6737 @item -ggnu-pubnames
6738 @opindex ggnu-pubnames
6739 Generate @code{.debug_pubnames} and @code{.debug_pubtypes} sections in a format
6740 suitable for conversion into a GDB@ index. This option is only useful
6741 with a linker that can produce GDB@ index version 7.
6743 @item -fdebug-types-section
6744 @opindex fdebug-types-section
6745 @opindex fno-debug-types-section
6746 When using DWARF Version 4 or higher, type DIEs can be put into
6747 their own @code{.debug_types} section instead of making them part of the
6748 @code{.debug_info} section. It is more efficient to put them in a separate
6749 comdat sections since the linker can then remove duplicates.
6750 But not all DWARF consumers support @code{.debug_types} sections yet
6751 and on some objects @code{.debug_types} produces larger instead of smaller
6752 debugging information.
6754 @item -grecord-gcc-switches
6755 @item -gno-record-gcc-switches
6756 @opindex grecord-gcc-switches
6757 @opindex gno-record-gcc-switches
6758 This switch causes the command-line options used to invoke the
6759 compiler that may affect code generation to be appended to the
6760 DW_AT_producer attribute in DWARF debugging information. The options
6761 are concatenated with spaces separating them from each other and from
6762 the compiler version.
6763 It is enabled by default.
6764 See also @option{-frecord-gcc-switches} for another
6765 way of storing compiler options into the object file.
6767 @item -gstrict-dwarf
6768 @opindex gstrict-dwarf
6769 Disallow using extensions of later DWARF standard version than selected
6770 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
6771 DWARF extensions from later standard versions is allowed.
6773 @item -gno-strict-dwarf
6774 @opindex gno-strict-dwarf
6775 Allow using extensions of later DWARF standard version than selected with
6776 @option{-gdwarf-@var{version}}.
6778 @item -gz@r{[}=@var{type}@r{]}
6780 Produce compressed debug sections in DWARF format, if that is supported.
6781 If @var{type} is not given, the default type depends on the capabilities
6782 of the assembler and linker used. @var{type} may be one of
6783 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
6784 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
6785 compression in traditional GNU format). If the linker doesn't support
6786 writing compressed debug sections, the option is rejected. Otherwise,
6787 if the assembler does not support them, @option{-gz} is silently ignored
6788 when producing object files.
6790 @item -feliminate-dwarf2-dups
6791 @opindex feliminate-dwarf2-dups
6792 Compress DWARF debugging information by eliminating duplicated
6793 information about each symbol. This option only makes sense when
6794 generating DWARF debugging information.
6796 @item -femit-struct-debug-baseonly
6797 @opindex femit-struct-debug-baseonly
6798 Emit debug information for struct-like types
6799 only when the base name of the compilation source file
6800 matches the base name of file in which the struct is defined.
6802 This option substantially reduces the size of debugging information,
6803 but at significant potential loss in type information to the debugger.
6804 See @option{-femit-struct-debug-reduced} for a less aggressive option.
6805 See @option{-femit-struct-debug-detailed} for more detailed control.
6807 This option works only with DWARF debug output.
6809 @item -femit-struct-debug-reduced
6810 @opindex femit-struct-debug-reduced
6811 Emit debug information for struct-like types
6812 only when the base name of the compilation source file
6813 matches the base name of file in which the type is defined,
6814 unless the struct is a template or defined in a system header.
6816 This option significantly reduces the size of debugging information,
6817 with some potential loss in type information to the debugger.
6818 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
6819 See @option{-femit-struct-debug-detailed} for more detailed control.
6821 This option works only with DWARF debug output.
6823 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
6824 @opindex femit-struct-debug-detailed
6825 Specify the struct-like types
6826 for which the compiler generates debug information.
6827 The intent is to reduce duplicate struct debug information
6828 between different object files within the same program.
6830 This option is a detailed version of
6831 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
6832 which serves for most needs.
6834 A specification has the syntax@*
6835 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
6837 The optional first word limits the specification to
6838 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
6839 A struct type is used directly when it is the type of a variable, member.
6840 Indirect uses arise through pointers to structs.
6841 That is, when use of an incomplete struct is valid, the use is indirect.
6843 @samp{struct one direct; struct two * indirect;}.
6845 The optional second word limits the specification to
6846 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
6847 Generic structs are a bit complicated to explain.
6848 For C++, these are non-explicit specializations of template classes,
6849 or non-template classes within the above.
6850 Other programming languages have generics,
6851 but @option{-femit-struct-debug-detailed} does not yet implement them.
6853 The third word specifies the source files for those
6854 structs for which the compiler should emit debug information.
6855 The values @samp{none} and @samp{any} have the normal meaning.
6856 The value @samp{base} means that
6857 the base of name of the file in which the type declaration appears
6858 must match the base of the name of the main compilation file.
6859 In practice, this means that when compiling @file{foo.c}, debug information
6860 is generated for types declared in that file and @file{foo.h},
6861 but not other header files.
6862 The value @samp{sys} means those types satisfying @samp{base}
6863 or declared in system or compiler headers.
6865 You may need to experiment to determine the best settings for your application.
6867 The default is @option{-femit-struct-debug-detailed=all}.
6869 This option works only with DWARF debug output.
6871 @item -fno-dwarf2-cfi-asm
6872 @opindex fdwarf2-cfi-asm
6873 @opindex fno-dwarf2-cfi-asm
6874 Emit DWARF unwind info as compiler generated @code{.eh_frame} section
6875 instead of using GAS @code{.cfi_*} directives.
6877 @item -fno-eliminate-unused-debug-types
6878 @opindex feliminate-unused-debug-types
6879 @opindex fno-eliminate-unused-debug-types
6880 Normally, when producing DWARF output, GCC avoids producing debug symbol
6881 output for types that are nowhere used in the source file being compiled.
6882 Sometimes it is useful to have GCC emit debugging
6883 information for all types declared in a compilation
6884 unit, regardless of whether or not they are actually used
6885 in that compilation unit, for example
6886 if, in the debugger, you want to cast a value to a type that is
6887 not actually used in your program (but is declared). More often,
6888 however, this results in a significant amount of wasted space.
6891 @node Optimize Options
6892 @section Options That Control Optimization
6893 @cindex optimize options
6894 @cindex options, optimization
6896 These options control various sorts of optimizations.
6898 Without any optimization option, the compiler's goal is to reduce the
6899 cost of compilation and to make debugging produce the expected
6900 results. Statements are independent: if you stop the program with a
6901 breakpoint between statements, you can then assign a new value to any
6902 variable or change the program counter to any other statement in the
6903 function and get exactly the results you expect from the source
6906 Turning on optimization flags makes the compiler attempt to improve
6907 the performance and/or code size at the expense of compilation time
6908 and possibly the ability to debug the program.
6910 The compiler performs optimization based on the knowledge it has of the
6911 program. Compiling multiple files at once to a single output file mode allows
6912 the compiler to use information gained from all of the files when compiling
6915 Not all optimizations are controlled directly by a flag. Only
6916 optimizations that have a flag are listed in this section.
6918 Most optimizations are only enabled if an @option{-O} level is set on
6919 the command line. Otherwise they are disabled, even if individual
6920 optimization flags are specified.
6922 Depending on the target and how GCC was configured, a slightly different
6923 set of optimizations may be enabled at each @option{-O} level than
6924 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
6925 to find out the exact set of optimizations that are enabled at each level.
6926 @xref{Overall Options}, for examples.
6933 Optimize. Optimizing compilation takes somewhat more time, and a lot
6934 more memory for a large function.
6936 With @option{-O}, the compiler tries to reduce code size and execution
6937 time, without performing any optimizations that take a great deal of
6940 @option{-O} turns on the following optimization flags:
6943 -fbranch-count-reg @gol
6944 -fcombine-stack-adjustments @gol
6946 -fcprop-registers @gol
6949 -fdelayed-branch @gol
6951 -fforward-propagate @gol
6952 -fguess-branch-probability @gol
6953 -fif-conversion2 @gol
6954 -fif-conversion @gol
6955 -finline-functions-called-once @gol
6956 -fipa-pure-const @gol
6958 -fipa-reference @gol
6959 -fmerge-constants @gol
6960 -fmove-loop-invariants @gol
6961 -freorder-blocks @gol
6963 -fshrink-wrap-separate @gol
6964 -fsplit-wide-types @gol
6967 -fstore-merging @gol
6971 -ftree-coalesce-vars @gol
6972 -ftree-copy-prop @gol
6974 -ftree-dominator-opts @gol
6976 -ftree-forwprop @gol
6986 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6987 where doing so does not interfere with debugging.
6991 Optimize even more. GCC performs nearly all supported optimizations
6992 that do not involve a space-speed tradeoff.
6993 As compared to @option{-O}, this option increases both compilation time
6994 and the performance of the generated code.
6996 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6997 also turns on the following optimization flags:
6998 @gccoptlist{-fthread-jumps @gol
6999 -falign-functions -falign-jumps @gol
7000 -falign-loops -falign-labels @gol
7003 -fcse-follow-jumps -fcse-skip-blocks @gol
7004 -fdelete-null-pointer-checks @gol
7005 -fdevirtualize -fdevirtualize-speculatively @gol
7006 -fexpensive-optimizations @gol
7007 -fgcse -fgcse-lm @gol
7008 -fhoist-adjacent-loads @gol
7009 -finline-small-functions @gol
7010 -findirect-inlining @gol
7012 -fipa-cp-alignment @gol
7016 -fisolate-erroneous-paths-dereference @gol
7018 -foptimize-sibling-calls @gol
7019 -foptimize-strlen @gol
7020 -fpartial-inlining @gol
7022 -freorder-blocks-algorithm=stc @gol
7023 -freorder-blocks-and-partition -freorder-functions @gol
7024 -frerun-cse-after-loop @gol
7025 -fsched-interblock -fsched-spec @gol
7026 -fschedule-insns -fschedule-insns2 @gol
7027 -fstrict-aliasing -fstrict-overflow @gol
7028 -ftree-builtin-call-dce @gol
7029 -ftree-switch-conversion -ftree-tail-merge @gol
7030 -fcode-hoisting @gol
7035 Please note the warning under @option{-fgcse} about
7036 invoking @option{-O2} on programs that use computed gotos.
7040 Optimize yet more. @option{-O3} turns on all optimizations specified
7041 by @option{-O2} and also turns on the @option{-finline-functions},
7042 @option{-funswitch-loops}, @option{-fpredictive-commoning},
7043 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
7044 @option{-ftree-loop-distribute-patterns}, @option{-fsplit-paths}
7045 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
7046 @option{-ftree-partial-pre}, @option{-fpeel-loops}
7047 and @option{-fipa-cp-clone} options.
7051 Reduce compilation time and make debugging produce the expected
7052 results. This is the default.
7056 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7057 do not typically increase code size. It also performs further
7058 optimizations designed to reduce code size.
7060 @option{-Os} disables the following optimization flags:
7061 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7062 -falign-labels -freorder-blocks -freorder-blocks-algorithm=stc @gol
7063 -freorder-blocks-and-partition -fprefetch-loop-arrays}
7067 Disregard strict standards compliance. @option{-Ofast} enables all
7068 @option{-O3} optimizations. It also enables optimizations that are not
7069 valid for all standard-compliant programs.
7070 It turns on @option{-ffast-math} and the Fortran-specific
7071 @option{-fno-protect-parens} and @option{-fstack-arrays}.
7075 Optimize debugging experience. @option{-Og} enables optimizations
7076 that do not interfere with debugging. It should be the optimization
7077 level of choice for the standard edit-compile-debug cycle, offering
7078 a reasonable level of optimization while maintaining fast compilation
7079 and a good debugging experience.
7082 If you use multiple @option{-O} options, with or without level numbers,
7083 the last such option is the one that is effective.
7085 Options of the form @option{-f@var{flag}} specify machine-independent
7086 flags. Most flags have both positive and negative forms; the negative
7087 form of @option{-ffoo} is @option{-fno-foo}. In the table
7088 below, only one of the forms is listed---the one you typically
7089 use. You can figure out the other form by either removing @samp{no-}
7092 The following options control specific optimizations. They are either
7093 activated by @option{-O} options or are related to ones that are. You
7094 can use the following flags in the rare cases when ``fine-tuning'' of
7095 optimizations to be performed is desired.
7098 @item -fno-defer-pop
7099 @opindex fno-defer-pop
7100 Always pop the arguments to each function call as soon as that function
7101 returns. For machines that must pop arguments after a function call,
7102 the compiler normally lets arguments accumulate on the stack for several
7103 function calls and pops them all at once.
7105 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7107 @item -fforward-propagate
7108 @opindex fforward-propagate
7109 Perform a forward propagation pass on RTL@. The pass tries to combine two
7110 instructions and checks if the result can be simplified. If loop unrolling
7111 is active, two passes are performed and the second is scheduled after
7114 This option is enabled by default at optimization levels @option{-O},
7115 @option{-O2}, @option{-O3}, @option{-Os}.
7117 @item -ffp-contract=@var{style}
7118 @opindex ffp-contract
7119 @option{-ffp-contract=off} disables floating-point expression contraction.
7120 @option{-ffp-contract=fast} enables floating-point expression contraction
7121 such as forming of fused multiply-add operations if the target has
7122 native support for them.
7123 @option{-ffp-contract=on} enables floating-point expression contraction
7124 if allowed by the language standard. This is currently not implemented
7125 and treated equal to @option{-ffp-contract=off}.
7127 The default is @option{-ffp-contract=fast}.
7129 @item -fomit-frame-pointer
7130 @opindex fomit-frame-pointer
7131 Don't keep the frame pointer in a register for functions that
7132 don't need one. This avoids the instructions to save, set up and
7133 restore frame pointers; it also makes an extra register available
7134 in many functions. @strong{It also makes debugging impossible on
7137 On some machines, such as the VAX, this flag has no effect, because
7138 the standard calling sequence automatically handles the frame pointer
7139 and nothing is saved by pretending it doesn't exist. The
7140 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
7141 whether a target machine supports this flag. @xref{Registers,,Register
7142 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
7144 The default setting (when not optimizing for
7145 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets is
7146 @option{-fomit-frame-pointer}. You can configure GCC with the
7147 @option{--enable-frame-pointer} configure option to change the default.
7149 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7151 @item -foptimize-sibling-calls
7152 @opindex foptimize-sibling-calls
7153 Optimize sibling and tail recursive calls.
7155 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7157 @item -foptimize-strlen
7158 @opindex foptimize-strlen
7159 Optimize various standard C string functions (e.g. @code{strlen},
7160 @code{strchr} or @code{strcpy}) and
7161 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7163 Enabled at levels @option{-O2}, @option{-O3}.
7167 Do not expand any functions inline apart from those marked with
7168 the @code{always_inline} attribute. This is the default when not
7171 Single functions can be exempted from inlining by marking them
7172 with the @code{noinline} attribute.
7174 @item -finline-small-functions
7175 @opindex finline-small-functions
7176 Integrate functions into their callers when their body is smaller than expected
7177 function call code (so overall size of program gets smaller). The compiler
7178 heuristically decides which functions are simple enough to be worth integrating
7179 in this way. This inlining applies to all functions, even those not declared
7182 Enabled at level @option{-O2}.
7184 @item -findirect-inlining
7185 @opindex findirect-inlining
7186 Inline also indirect calls that are discovered to be known at compile
7187 time thanks to previous inlining. This option has any effect only
7188 when inlining itself is turned on by the @option{-finline-functions}
7189 or @option{-finline-small-functions} options.
7191 Enabled at level @option{-O2}.
7193 @item -finline-functions
7194 @opindex finline-functions
7195 Consider all functions for inlining, even if they are not declared inline.
7196 The compiler heuristically decides which functions are worth integrating
7199 If all calls to a given function are integrated, and the function is
7200 declared @code{static}, then the function is normally not output as
7201 assembler code in its own right.
7203 Enabled at level @option{-O3}.
7205 @item -finline-functions-called-once
7206 @opindex finline-functions-called-once
7207 Consider all @code{static} functions called once for inlining into their
7208 caller even if they are not marked @code{inline}. If a call to a given
7209 function is integrated, then the function is not output as assembler code
7212 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7214 @item -fearly-inlining
7215 @opindex fearly-inlining
7216 Inline functions marked by @code{always_inline} and functions whose body seems
7217 smaller than the function call overhead early before doing
7218 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7219 makes profiling significantly cheaper and usually inlining faster on programs
7220 having large chains of nested wrapper functions.
7226 Perform interprocedural scalar replacement of aggregates, removal of
7227 unused parameters and replacement of parameters passed by reference
7228 by parameters passed by value.
7230 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7232 @item -finline-limit=@var{n}
7233 @opindex finline-limit
7234 By default, GCC limits the size of functions that can be inlined. This flag
7235 allows coarse control of this limit. @var{n} is the size of functions that
7236 can be inlined in number of pseudo instructions.
7238 Inlining is actually controlled by a number of parameters, which may be
7239 specified individually by using @option{--param @var{name}=@var{value}}.
7240 The @option{-finline-limit=@var{n}} option sets some of these parameters
7244 @item max-inline-insns-single
7245 is set to @var{n}/2.
7246 @item max-inline-insns-auto
7247 is set to @var{n}/2.
7250 See below for a documentation of the individual
7251 parameters controlling inlining and for the defaults of these parameters.
7253 @emph{Note:} there may be no value to @option{-finline-limit} that results
7254 in default behavior.
7256 @emph{Note:} pseudo instruction represents, in this particular context, an
7257 abstract measurement of function's size. In no way does it represent a count
7258 of assembly instructions and as such its exact meaning might change from one
7259 release to an another.
7261 @item -fno-keep-inline-dllexport
7262 @opindex fno-keep-inline-dllexport
7263 This is a more fine-grained version of @option{-fkeep-inline-functions},
7264 which applies only to functions that are declared using the @code{dllexport}
7265 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
7268 @item -fkeep-inline-functions
7269 @opindex fkeep-inline-functions
7270 In C, emit @code{static} functions that are declared @code{inline}
7271 into the object file, even if the function has been inlined into all
7272 of its callers. This switch does not affect functions using the
7273 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7274 inline functions into the object file.
7276 @item -fkeep-static-functions
7277 @opindex fkeep-static-functions
7278 Emit @code{static} functions into the object file, even if the function
7281 @item -fkeep-static-consts
7282 @opindex fkeep-static-consts
7283 Emit variables declared @code{static const} when optimization isn't turned
7284 on, even if the variables aren't referenced.
7286 GCC enables this option by default. If you want to force the compiler to
7287 check if a variable is referenced, regardless of whether or not
7288 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7290 @item -fmerge-constants
7291 @opindex fmerge-constants
7292 Attempt to merge identical constants (string constants and floating-point
7293 constants) across compilation units.
7295 This option is the default for optimized compilation if the assembler and
7296 linker support it. Use @option{-fno-merge-constants} to inhibit this
7299 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7301 @item -fmerge-all-constants
7302 @opindex fmerge-all-constants
7303 Attempt to merge identical constants and identical variables.
7305 This option implies @option{-fmerge-constants}. In addition to
7306 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7307 arrays or initialized constant variables with integral or floating-point
7308 types. Languages like C or C++ require each variable, including multiple
7309 instances of the same variable in recursive calls, to have distinct locations,
7310 so using this option results in non-conforming
7313 @item -fmodulo-sched
7314 @opindex fmodulo-sched
7315 Perform swing modulo scheduling immediately before the first scheduling
7316 pass. This pass looks at innermost loops and reorders their
7317 instructions by overlapping different iterations.
7319 @item -fmodulo-sched-allow-regmoves
7320 @opindex fmodulo-sched-allow-regmoves
7321 Perform more aggressive SMS-based modulo scheduling with register moves
7322 allowed. By setting this flag certain anti-dependences edges are
7323 deleted, which triggers the generation of reg-moves based on the
7324 life-range analysis. This option is effective only with
7325 @option{-fmodulo-sched} enabled.
7327 @item -fno-branch-count-reg
7328 @opindex fno-branch-count-reg
7329 Avoid running a pass scanning for opportunities to use ``decrement and
7330 branch'' instructions on a count register instead of generating sequences
7331 of instructions that decrement a register, compare it against zero, and
7332 then branch based upon the result. This option is only meaningful on
7333 architectures that support such instructions, which include x86, PowerPC,
7334 IA-64 and S/390. Note that the @option{-fno-branch-count-reg} option
7335 doesn't remove the decrement and branch instructions from the generated
7336 instruction stream introduced by other optimization passes.
7338 Enabled by default at @option{-O1} and higher.
7340 The default is @option{-fbranch-count-reg}.
7342 @item -fno-function-cse
7343 @opindex fno-function-cse
7344 Do not put function addresses in registers; make each instruction that
7345 calls a constant function contain the function's address explicitly.
7347 This option results in less efficient code, but some strange hacks
7348 that alter the assembler output may be confused by the optimizations
7349 performed when this option is not used.
7351 The default is @option{-ffunction-cse}
7353 @item -fno-zero-initialized-in-bss
7354 @opindex fno-zero-initialized-in-bss
7355 If the target supports a BSS section, GCC by default puts variables that
7356 are initialized to zero into BSS@. This can save space in the resulting
7359 This option turns off this behavior because some programs explicitly
7360 rely on variables going to the data section---e.g., so that the
7361 resulting executable can find the beginning of that section and/or make
7362 assumptions based on that.
7364 The default is @option{-fzero-initialized-in-bss}.
7366 @item -fthread-jumps
7367 @opindex fthread-jumps
7368 Perform optimizations that check to see if a jump branches to a
7369 location where another comparison subsumed by the first is found. If
7370 so, the first branch is redirected to either the destination of the
7371 second branch or a point immediately following it, depending on whether
7372 the condition is known to be true or false.
7374 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7376 @item -fsplit-wide-types
7377 @opindex fsplit-wide-types
7378 When using a type that occupies multiple registers, such as @code{long
7379 long} on a 32-bit system, split the registers apart and allocate them
7380 independently. This normally generates better code for those types,
7381 but may make debugging more difficult.
7383 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7386 @item -fcse-follow-jumps
7387 @opindex fcse-follow-jumps
7388 In common subexpression elimination (CSE), scan through jump instructions
7389 when the target of the jump is not reached by any other path. For
7390 example, when CSE encounters an @code{if} statement with an
7391 @code{else} clause, CSE follows the jump when the condition
7394 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7396 @item -fcse-skip-blocks
7397 @opindex fcse-skip-blocks
7398 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
7399 follow jumps that conditionally skip over blocks. When CSE
7400 encounters a simple @code{if} statement with no else clause,
7401 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
7402 body of the @code{if}.
7404 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7406 @item -frerun-cse-after-loop
7407 @opindex frerun-cse-after-loop
7408 Re-run common subexpression elimination after loop optimizations are
7411 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7415 Perform a global common subexpression elimination pass.
7416 This pass also performs global constant and copy propagation.
7418 @emph{Note:} When compiling a program using computed gotos, a GCC
7419 extension, you may get better run-time performance if you disable
7420 the global common subexpression elimination pass by adding
7421 @option{-fno-gcse} to the command line.
7423 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7427 When @option{-fgcse-lm} is enabled, global common subexpression elimination
7428 attempts to move loads that are only killed by stores into themselves. This
7429 allows a loop containing a load/store sequence to be changed to a load outside
7430 the loop, and a copy/store within the loop.
7432 Enabled by default when @option{-fgcse} is enabled.
7436 When @option{-fgcse-sm} is enabled, a store motion pass is run after
7437 global common subexpression elimination. This pass attempts to move
7438 stores out of loops. When used in conjunction with @option{-fgcse-lm},
7439 loops containing a load/store sequence can be changed to a load before
7440 the loop and a store after the loop.
7442 Not enabled at any optimization level.
7446 When @option{-fgcse-las} is enabled, the global common subexpression
7447 elimination pass eliminates redundant loads that come after stores to the
7448 same memory location (both partial and full redundancies).
7450 Not enabled at any optimization level.
7452 @item -fgcse-after-reload
7453 @opindex fgcse-after-reload
7454 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
7455 pass is performed after reload. The purpose of this pass is to clean up
7458 @item -faggressive-loop-optimizations
7459 @opindex faggressive-loop-optimizations
7460 This option tells the loop optimizer to use language constraints to
7461 derive bounds for the number of iterations of a loop. This assumes that
7462 loop code does not invoke undefined behavior by for example causing signed
7463 integer overflows or out-of-bound array accesses. The bounds for the
7464 number of iterations of a loop are used to guide loop unrolling and peeling
7465 and loop exit test optimizations.
7466 This option is enabled by default.
7468 @item -funconstrained-commons
7469 @opindex funconstrained-commons
7470 This option tells the compiler that variables declared in common blocks
7471 (e.g. Fortran) may later be overridden with longer trailing arrays. This
7472 prevents certain optimizations that depend on knowing the array bounds.
7474 @item -fcrossjumping
7475 @opindex fcrossjumping
7476 Perform cross-jumping transformation.
7477 This transformation unifies equivalent code and saves code size. The
7478 resulting code may or may not perform better than without cross-jumping.
7480 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7482 @item -fauto-inc-dec
7483 @opindex fauto-inc-dec
7484 Combine increments or decrements of addresses with memory accesses.
7485 This pass is always skipped on architectures that do not have
7486 instructions to support this. Enabled by default at @option{-O} and
7487 higher on architectures that support this.
7491 Perform dead code elimination (DCE) on RTL@.
7492 Enabled by default at @option{-O} and higher.
7496 Perform dead store elimination (DSE) on RTL@.
7497 Enabled by default at @option{-O} and higher.
7499 @item -fif-conversion
7500 @opindex fif-conversion
7501 Attempt to transform conditional jumps into branch-less equivalents. This
7502 includes use of conditional moves, min, max, set flags and abs instructions, and
7503 some tricks doable by standard arithmetics. The use of conditional execution
7504 on chips where it is available is controlled by @option{-fif-conversion2}.
7506 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7508 @item -fif-conversion2
7509 @opindex fif-conversion2
7510 Use conditional execution (where available) to transform conditional jumps into
7511 branch-less equivalents.
7513 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7515 @item -fdeclone-ctor-dtor
7516 @opindex fdeclone-ctor-dtor
7517 The C++ ABI requires multiple entry points for constructors and
7518 destructors: one for a base subobject, one for a complete object, and
7519 one for a virtual destructor that calls operator delete afterwards.
7520 For a hierarchy with virtual bases, the base and complete variants are
7521 clones, which means two copies of the function. With this option, the
7522 base and complete variants are changed to be thunks that call a common
7525 Enabled by @option{-Os}.
7527 @item -fdelete-null-pointer-checks
7528 @opindex fdelete-null-pointer-checks
7529 Assume that programs cannot safely dereference null pointers, and that
7530 no code or data element resides at address zero.
7531 This option enables simple constant
7532 folding optimizations at all optimization levels. In addition, other
7533 optimization passes in GCC use this flag to control global dataflow
7534 analyses that eliminate useless checks for null pointers; these assume
7535 that a memory access to address zero always results in a trap, so
7536 that if a pointer is checked after it has already been dereferenced,
7539 Note however that in some environments this assumption is not true.
7540 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
7541 for programs that depend on that behavior.
7543 This option is enabled by default on most targets. On Nios II ELF, it
7544 defaults to off. On AVR and CR16, this option is completely disabled.
7546 Passes that use the dataflow information
7547 are enabled independently at different optimization levels.
7549 @item -fdevirtualize
7550 @opindex fdevirtualize
7551 Attempt to convert calls to virtual functions to direct calls. This
7552 is done both within a procedure and interprocedurally as part of
7553 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
7554 propagation (@option{-fipa-cp}).
7555 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7557 @item -fdevirtualize-speculatively
7558 @opindex fdevirtualize-speculatively
7559 Attempt to convert calls to virtual functions to speculative direct calls.
7560 Based on the analysis of the type inheritance graph, determine for a given call
7561 the set of likely targets. If the set is small, preferably of size 1, change
7562 the call into a conditional deciding between direct and indirect calls. The
7563 speculative calls enable more optimizations, such as inlining. When they seem
7564 useless after further optimization, they are converted back into original form.
7566 @item -fdevirtualize-at-ltrans
7567 @opindex fdevirtualize-at-ltrans
7568 Stream extra information needed for aggressive devirtualization when running
7569 the link-time optimizer in local transformation mode.
7570 This option enables more devirtualization but
7571 significantly increases the size of streamed data. For this reason it is
7572 disabled by default.
7574 @item -fexpensive-optimizations
7575 @opindex fexpensive-optimizations
7576 Perform a number of minor optimizations that are relatively expensive.
7578 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7582 Attempt to remove redundant extension instructions. This is especially
7583 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
7584 registers after writing to their lower 32-bit half.
7586 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
7587 @option{-O3}, @option{-Os}.
7589 @item -fno-lifetime-dse
7590 @opindex fno-lifetime-dse
7591 In C++ the value of an object is only affected by changes within its
7592 lifetime: when the constructor begins, the object has an indeterminate
7593 value, and any changes during the lifetime of the object are dead when
7594 the object is destroyed. Normally dead store elimination will take
7595 advantage of this; if your code relies on the value of the object
7596 storage persisting beyond the lifetime of the object, you can use this
7597 flag to disable this optimization. To preserve stores before the
7598 constructor starts (e.g. because your operator new clears the object
7599 storage) but still treat the object as dead after the destructor you,
7600 can use @option{-flifetime-dse=1}. The default behavior can be
7601 explicitly selected with @option{-flifetime-dse=2}.
7602 @option{-flifetime-dse=0} is equivalent to @option{-fno-lifetime-dse}.
7604 @item -flive-range-shrinkage
7605 @opindex flive-range-shrinkage
7606 Attempt to decrease register pressure through register live range
7607 shrinkage. This is helpful for fast processors with small or moderate
7610 @item -fira-algorithm=@var{algorithm}
7611 @opindex fira-algorithm
7612 Use the specified coloring algorithm for the integrated register
7613 allocator. The @var{algorithm} argument can be @samp{priority}, which
7614 specifies Chow's priority coloring, or @samp{CB}, which specifies
7615 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
7616 for all architectures, but for those targets that do support it, it is
7617 the default because it generates better code.
7619 @item -fira-region=@var{region}
7620 @opindex fira-region
7621 Use specified regions for the integrated register allocator. The
7622 @var{region} argument should be one of the following:
7627 Use all loops as register allocation regions.
7628 This can give the best results for machines with a small and/or
7629 irregular register set.
7632 Use all loops except for loops with small register pressure
7633 as the regions. This value usually gives
7634 the best results in most cases and for most architectures,
7635 and is enabled by default when compiling with optimization for speed
7636 (@option{-O}, @option{-O2}, @dots{}).
7639 Use all functions as a single region.
7640 This typically results in the smallest code size, and is enabled by default for
7641 @option{-Os} or @option{-O0}.
7645 @item -fira-hoist-pressure
7646 @opindex fira-hoist-pressure
7647 Use IRA to evaluate register pressure in the code hoisting pass for
7648 decisions to hoist expressions. This option usually results in smaller
7649 code, but it can slow the compiler down.
7651 This option is enabled at level @option{-Os} for all targets.
7653 @item -fira-loop-pressure
7654 @opindex fira-loop-pressure
7655 Use IRA to evaluate register pressure in loops for decisions to move
7656 loop invariants. This option usually results in generation
7657 of faster and smaller code on machines with large register files (>= 32
7658 registers), but it can slow the compiler down.
7660 This option is enabled at level @option{-O3} for some targets.
7662 @item -fno-ira-share-save-slots
7663 @opindex fno-ira-share-save-slots
7664 Disable sharing of stack slots used for saving call-used hard
7665 registers living through a call. Each hard register gets a
7666 separate stack slot, and as a result function stack frames are
7669 @item -fno-ira-share-spill-slots
7670 @opindex fno-ira-share-spill-slots
7671 Disable sharing of stack slots allocated for pseudo-registers. Each
7672 pseudo-register that does not get a hard register gets a separate
7673 stack slot, and as a result function stack frames are larger.
7677 Enable CFG-sensitive rematerialization in LRA. Instead of loading
7678 values of spilled pseudos, LRA tries to rematerialize (recalculate)
7679 values if it is profitable.
7681 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7683 @item -fdelayed-branch
7684 @opindex fdelayed-branch
7685 If supported for the target machine, attempt to reorder instructions
7686 to exploit instruction slots available after delayed branch
7689 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7691 @item -fschedule-insns
7692 @opindex fschedule-insns
7693 If supported for the target machine, attempt to reorder instructions to
7694 eliminate execution stalls due to required data being unavailable. This
7695 helps machines that have slow floating point or memory load instructions
7696 by allowing other instructions to be issued until the result of the load
7697 or floating-point instruction is required.
7699 Enabled at levels @option{-O2}, @option{-O3}.
7701 @item -fschedule-insns2
7702 @opindex fschedule-insns2
7703 Similar to @option{-fschedule-insns}, but requests an additional pass of
7704 instruction scheduling after register allocation has been done. This is
7705 especially useful on machines with a relatively small number of
7706 registers and where memory load instructions take more than one cycle.
7708 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7710 @item -fno-sched-interblock
7711 @opindex fno-sched-interblock
7712 Don't schedule instructions across basic blocks. This is normally
7713 enabled by default when scheduling before register allocation, i.e.@:
7714 with @option{-fschedule-insns} or at @option{-O2} or higher.
7716 @item -fno-sched-spec
7717 @opindex fno-sched-spec
7718 Don't allow speculative motion of non-load instructions. This is normally
7719 enabled by default when scheduling before register allocation, i.e.@:
7720 with @option{-fschedule-insns} or at @option{-O2} or higher.
7722 @item -fsched-pressure
7723 @opindex fsched-pressure
7724 Enable register pressure sensitive insn scheduling before register
7725 allocation. This only makes sense when scheduling before register
7726 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
7727 @option{-O2} or higher. Usage of this option can improve the
7728 generated code and decrease its size by preventing register pressure
7729 increase above the number of available hard registers and subsequent
7730 spills in register allocation.
7732 @item -fsched-spec-load
7733 @opindex fsched-spec-load
7734 Allow speculative motion of some load instructions. This only makes
7735 sense when scheduling before register allocation, i.e.@: with
7736 @option{-fschedule-insns} or at @option{-O2} or higher.
7738 @item -fsched-spec-load-dangerous
7739 @opindex fsched-spec-load-dangerous
7740 Allow speculative motion of more load instructions. This only makes
7741 sense when scheduling before register allocation, i.e.@: with
7742 @option{-fschedule-insns} or at @option{-O2} or higher.
7744 @item -fsched-stalled-insns
7745 @itemx -fsched-stalled-insns=@var{n}
7746 @opindex fsched-stalled-insns
7747 Define how many insns (if any) can be moved prematurely from the queue
7748 of stalled insns into the ready list during the second scheduling pass.
7749 @option{-fno-sched-stalled-insns} means that no insns are moved
7750 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
7751 on how many queued insns can be moved prematurely.
7752 @option{-fsched-stalled-insns} without a value is equivalent to
7753 @option{-fsched-stalled-insns=1}.
7755 @item -fsched-stalled-insns-dep
7756 @itemx -fsched-stalled-insns-dep=@var{n}
7757 @opindex fsched-stalled-insns-dep
7758 Define how many insn groups (cycles) are examined for a dependency
7759 on a stalled insn that is a candidate for premature removal from the queue
7760 of stalled insns. This has an effect only during the second scheduling pass,
7761 and only if @option{-fsched-stalled-insns} is used.
7762 @option{-fno-sched-stalled-insns-dep} is equivalent to
7763 @option{-fsched-stalled-insns-dep=0}.
7764 @option{-fsched-stalled-insns-dep} without a value is equivalent to
7765 @option{-fsched-stalled-insns-dep=1}.
7767 @item -fsched2-use-superblocks
7768 @opindex fsched2-use-superblocks
7769 When scheduling after register allocation, use superblock scheduling.
7770 This allows motion across basic block boundaries,
7771 resulting in faster schedules. This option is experimental, as not all machine
7772 descriptions used by GCC model the CPU closely enough to avoid unreliable
7773 results from the algorithm.
7775 This only makes sense when scheduling after register allocation, i.e.@: with
7776 @option{-fschedule-insns2} or at @option{-O2} or higher.
7778 @item -fsched-group-heuristic
7779 @opindex fsched-group-heuristic
7780 Enable the group heuristic in the scheduler. This heuristic favors
7781 the instruction that belongs to a schedule group. This is enabled
7782 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7783 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7785 @item -fsched-critical-path-heuristic
7786 @opindex fsched-critical-path-heuristic
7787 Enable the critical-path heuristic in the scheduler. This heuristic favors
7788 instructions on the critical path. This is enabled by default when
7789 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7790 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7792 @item -fsched-spec-insn-heuristic
7793 @opindex fsched-spec-insn-heuristic
7794 Enable the speculative instruction heuristic in the scheduler. This
7795 heuristic favors speculative instructions with greater dependency weakness.
7796 This is enabled by default when scheduling is enabled, i.e.@:
7797 with @option{-fschedule-insns} or @option{-fschedule-insns2}
7798 or at @option{-O2} or higher.
7800 @item -fsched-rank-heuristic
7801 @opindex fsched-rank-heuristic
7802 Enable the rank heuristic in the scheduler. This heuristic favors
7803 the instruction belonging to a basic block with greater size or frequency.
7804 This is enabled by default when scheduling is enabled, i.e.@:
7805 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7806 at @option{-O2} or higher.
7808 @item -fsched-last-insn-heuristic
7809 @opindex fsched-last-insn-heuristic
7810 Enable the last-instruction heuristic in the scheduler. This heuristic
7811 favors the instruction that is less dependent on the last instruction
7812 scheduled. This is enabled by default when scheduling is enabled,
7813 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7814 at @option{-O2} or higher.
7816 @item -fsched-dep-count-heuristic
7817 @opindex fsched-dep-count-heuristic
7818 Enable the dependent-count heuristic in the scheduler. This heuristic
7819 favors the instruction that has more instructions depending on it.
7820 This is enabled by default when scheduling is enabled, i.e.@:
7821 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7822 at @option{-O2} or higher.
7824 @item -freschedule-modulo-scheduled-loops
7825 @opindex freschedule-modulo-scheduled-loops
7826 Modulo scheduling is performed before traditional scheduling. If a loop
7827 is modulo scheduled, later scheduling passes may change its schedule.
7828 Use this option to control that behavior.
7830 @item -fselective-scheduling
7831 @opindex fselective-scheduling
7832 Schedule instructions using selective scheduling algorithm. Selective
7833 scheduling runs instead of the first scheduler pass.
7835 @item -fselective-scheduling2
7836 @opindex fselective-scheduling2
7837 Schedule instructions using selective scheduling algorithm. Selective
7838 scheduling runs instead of the second scheduler pass.
7840 @item -fsel-sched-pipelining
7841 @opindex fsel-sched-pipelining
7842 Enable software pipelining of innermost loops during selective scheduling.
7843 This option has no effect unless one of @option{-fselective-scheduling} or
7844 @option{-fselective-scheduling2} is turned on.
7846 @item -fsel-sched-pipelining-outer-loops
7847 @opindex fsel-sched-pipelining-outer-loops
7848 When pipelining loops during selective scheduling, also pipeline outer loops.
7849 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
7851 @item -fsemantic-interposition
7852 @opindex fsemantic-interposition
7853 Some object formats, like ELF, allow interposing of symbols by the
7855 This means that for symbols exported from the DSO, the compiler cannot perform
7856 interprocedural propagation, inlining and other optimizations in anticipation
7857 that the function or variable in question may change. While this feature is
7858 useful, for example, to rewrite memory allocation functions by a debugging
7859 implementation, it is expensive in the terms of code quality.
7860 With @option{-fno-semantic-interposition} the compiler assumes that
7861 if interposition happens for functions the overwriting function will have
7862 precisely the same semantics (and side effects).
7863 Similarly if interposition happens
7864 for variables, the constructor of the variable will be the same. The flag
7865 has no effect for functions explicitly declared inline
7866 (where it is never allowed for interposition to change semantics)
7867 and for symbols explicitly declared weak.
7870 @opindex fshrink-wrap
7871 Emit function prologues only before parts of the function that need it,
7872 rather than at the top of the function. This flag is enabled by default at
7873 @option{-O} and higher.
7875 @item -fshrink-wrap-separate
7876 @opindex fshrink-wrap-separate
7877 Shrink-wrap separate parts of the prologue and epilogue separately, so that
7878 those parts are only executed when needed.
7879 This option is on by default, but has no effect unless @option{-fshrink-wrap}
7880 is also turned on and the target supports this.
7882 @item -fcaller-saves
7883 @opindex fcaller-saves
7884 Enable allocation of values to registers that are clobbered by
7885 function calls, by emitting extra instructions to save and restore the
7886 registers around such calls. Such allocation is done only when it
7887 seems to result in better code.
7889 This option is always enabled by default on certain machines, usually
7890 those which have no call-preserved registers to use instead.
7892 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7894 @item -fcombine-stack-adjustments
7895 @opindex fcombine-stack-adjustments
7896 Tracks stack adjustments (pushes and pops) and stack memory references
7897 and then tries to find ways to combine them.
7899 Enabled by default at @option{-O1} and higher.
7903 Use caller save registers for allocation if those registers are not used by
7904 any called function. In that case it is not necessary to save and restore
7905 them around calls. This is only possible if called functions are part of
7906 same compilation unit as current function and they are compiled before it.
7908 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}, however the option
7909 is disabled if generated code will be instrumented for profiling
7910 (@option{-p}, or @option{-pg}) or if callee's register usage cannot be known
7911 exactly (this happens on targets that do not expose prologues
7912 and epilogues in RTL).
7914 @item -fconserve-stack
7915 @opindex fconserve-stack
7916 Attempt to minimize stack usage. The compiler attempts to use less
7917 stack space, even if that makes the program slower. This option
7918 implies setting the @option{large-stack-frame} parameter to 100
7919 and the @option{large-stack-frame-growth} parameter to 400.
7921 @item -ftree-reassoc
7922 @opindex ftree-reassoc
7923 Perform reassociation on trees. This flag is enabled by default
7924 at @option{-O} and higher.
7926 @item -fcode-hoisting
7927 @opindex fcode-hoisting
7928 Perform code hoisting. Code hoisting tries to move the
7929 evaluation of expressions executed on all paths to the function exit
7930 as early as possible. This is especially useful as a code size
7931 optimization, but it often helps for code speed as well.
7932 This flag is enabled by default at @option{-O2} and higher.
7936 Perform partial redundancy elimination (PRE) on trees. This flag is
7937 enabled by default at @option{-O2} and @option{-O3}.
7939 @item -ftree-partial-pre
7940 @opindex ftree-partial-pre
7941 Make partial redundancy elimination (PRE) more aggressive. This flag is
7942 enabled by default at @option{-O3}.
7944 @item -ftree-forwprop
7945 @opindex ftree-forwprop
7946 Perform forward propagation on trees. This flag is enabled by default
7947 at @option{-O} and higher.
7951 Perform full redundancy elimination (FRE) on trees. The difference
7952 between FRE and PRE is that FRE only considers expressions
7953 that are computed on all paths leading to the redundant computation.
7954 This analysis is faster than PRE, though it exposes fewer redundancies.
7955 This flag is enabled by default at @option{-O} and higher.
7957 @item -ftree-phiprop
7958 @opindex ftree-phiprop
7959 Perform hoisting of loads from conditional pointers on trees. This
7960 pass is enabled by default at @option{-O} and higher.
7962 @item -fhoist-adjacent-loads
7963 @opindex fhoist-adjacent-loads
7964 Speculatively hoist loads from both branches of an if-then-else if the
7965 loads are from adjacent locations in the same structure and the target
7966 architecture has a conditional move instruction. This flag is enabled
7967 by default at @option{-O2} and higher.
7969 @item -ftree-copy-prop
7970 @opindex ftree-copy-prop
7971 Perform copy propagation on trees. This pass eliminates unnecessary
7972 copy operations. This flag is enabled by default at @option{-O} and
7975 @item -fipa-pure-const
7976 @opindex fipa-pure-const
7977 Discover which functions are pure or constant.
7978 Enabled by default at @option{-O} and higher.
7980 @item -fipa-reference
7981 @opindex fipa-reference
7982 Discover which static variables do not escape the
7984 Enabled by default at @option{-O} and higher.
7988 Perform interprocedural pointer analysis and interprocedural modification
7989 and reference analysis. This option can cause excessive memory and
7990 compile-time usage on large compilation units. It is not enabled by
7991 default at any optimization level.
7994 @opindex fipa-profile
7995 Perform interprocedural profile propagation. The functions called only from
7996 cold functions are marked as cold. Also functions executed once (such as
7997 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
7998 functions and loop less parts of functions executed once are then optimized for
8000 Enabled by default at @option{-O} and higher.
8004 Perform interprocedural constant propagation.
8005 This optimization analyzes the program to determine when values passed
8006 to functions are constants and then optimizes accordingly.
8007 This optimization can substantially increase performance
8008 if the application has constants passed to functions.
8009 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8011 @item -fipa-cp-clone
8012 @opindex fipa-cp-clone
8013 Perform function cloning to make interprocedural constant propagation stronger.
8014 When enabled, interprocedural constant propagation performs function cloning
8015 when externally visible function can be called with constant arguments.
8016 Because this optimization can create multiple copies of functions,
8017 it may significantly increase code size
8018 (see @option{--param ipcp-unit-growth=@var{value}}).
8019 This flag is enabled by default at @option{-O3}.
8021 @item -fipa-cp-alignment
8022 @opindex -fipa-cp-alignment
8023 When enabled, this optimization propagates alignment of function
8024 parameters to support better vectorization and string operations.
8026 This flag is enabled by default at @option{-O2} and @option{-Os}. It
8027 requires that @option{-fipa-cp} is enabled.
8028 @option{-fipa-cp-alignment} is obsolete, use @option{-fipa-bit-cp} instead.
8031 @opindex -fipa-bit-cp
8032 When enabled, perform ipa bitwise constant propagation. This flag is
8033 enabled by default at @option{-O2}. It requires that @option{-fipa-cp}
8038 Perform Identical Code Folding for functions and read-only variables.
8039 The optimization reduces code size and may disturb unwind stacks by replacing
8040 a function by equivalent one with a different name. The optimization works
8041 more effectively with link time optimization enabled.
8043 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8044 works on different levels and thus the optimizations are not same - there are
8045 equivalences that are found only by GCC and equivalences found only by Gold.
8047 This flag is enabled by default at @option{-O2} and @option{-Os}.
8049 @item -fisolate-erroneous-paths-dereference
8050 @opindex fisolate-erroneous-paths-dereference
8051 Detect paths that trigger erroneous or undefined behavior due to
8052 dereferencing a null pointer. Isolate those paths from the main control
8053 flow and turn the statement with erroneous or undefined behavior into a trap.
8054 This flag is enabled by default at @option{-O2} and higher and depends on
8055 @option{-fdelete-null-pointer-checks} also being enabled.
8057 @item -fisolate-erroneous-paths-attribute
8058 @opindex fisolate-erroneous-paths-attribute
8059 Detect paths that trigger erroneous or undefined behavior due a null value
8060 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
8061 attribute. Isolate those paths from the main control flow and turn the
8062 statement with erroneous or undefined behavior into a trap. This is not
8063 currently enabled, but may be enabled by @option{-O2} in the future.
8067 Perform forward store motion on trees. This flag is
8068 enabled by default at @option{-O} and higher.
8070 @item -ftree-bit-ccp
8071 @opindex ftree-bit-ccp
8072 Perform sparse conditional bit constant propagation on trees and propagate
8073 pointer alignment information.
8074 This pass only operates on local scalar variables and is enabled by default
8075 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8079 Perform sparse conditional constant propagation (CCP) on trees. This
8080 pass only operates on local scalar variables and is enabled by default
8081 at @option{-O} and higher.
8083 @item -fssa-backprop
8084 @opindex fssa-backprop
8085 Propagate information about uses of a value up the definition chain
8086 in order to simplify the definitions. For example, this pass strips
8087 sign operations if the sign of a value never matters. The flag is
8088 enabled by default at @option{-O} and higher.
8091 @opindex fssa-phiopt
8092 Perform pattern matching on SSA PHI nodes to optimize conditional
8093 code. This pass is enabled by default at @option{-O} and higher.
8095 @item -ftree-switch-conversion
8096 @opindex ftree-switch-conversion
8097 Perform conversion of simple initializations in a switch to
8098 initializations from a scalar array. This flag is enabled by default
8099 at @option{-O2} and higher.
8101 @item -ftree-tail-merge
8102 @opindex ftree-tail-merge
8103 Look for identical code sequences. When found, replace one with a jump to the
8104 other. This optimization is known as tail merging or cross jumping. This flag
8105 is enabled by default at @option{-O2} and higher. The compilation time
8107 be limited using @option{max-tail-merge-comparisons} parameter and
8108 @option{max-tail-merge-iterations} parameter.
8112 Perform dead code elimination (DCE) on trees. This flag is enabled by
8113 default at @option{-O} and higher.
8115 @item -ftree-builtin-call-dce
8116 @opindex ftree-builtin-call-dce
8117 Perform conditional dead code elimination (DCE) for calls to built-in functions
8118 that may set @code{errno} but are otherwise side-effect free. This flag is
8119 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8122 @item -ftree-dominator-opts
8123 @opindex ftree-dominator-opts
8124 Perform a variety of simple scalar cleanups (constant/copy
8125 propagation, redundancy elimination, range propagation and expression
8126 simplification) based on a dominator tree traversal. This also
8127 performs jump threading (to reduce jumps to jumps). This flag is
8128 enabled by default at @option{-O} and higher.
8132 Perform dead store elimination (DSE) on trees. A dead store is a store into
8133 a memory location that is later overwritten by another store without
8134 any intervening loads. In this case the earlier store can be deleted. This
8135 flag is enabled by default at @option{-O} and higher.
8139 Perform loop header copying on trees. This is beneficial since it increases
8140 effectiveness of code motion optimizations. It also saves one jump. This flag
8141 is enabled by default at @option{-O} and higher. It is not enabled
8142 for @option{-Os}, since it usually increases code size.
8144 @item -ftree-loop-optimize
8145 @opindex ftree-loop-optimize
8146 Perform loop optimizations on trees. This flag is enabled by default
8147 at @option{-O} and higher.
8149 @item -ftree-loop-linear
8150 @itemx -floop-interchange
8151 @itemx -floop-strip-mine
8153 @itemx -floop-unroll-and-jam
8154 @opindex ftree-loop-linear
8155 @opindex floop-interchange
8156 @opindex floop-strip-mine
8157 @opindex floop-block
8158 @opindex floop-unroll-and-jam
8159 Perform loop nest optimizations. Same as
8160 @option{-floop-nest-optimize}. To use this code transformation, GCC has
8161 to be configured with @option{--with-isl} to enable the Graphite loop
8162 transformation infrastructure.
8164 @item -fgraphite-identity
8165 @opindex fgraphite-identity
8166 Enable the identity transformation for graphite. For every SCoP we generate
8167 the polyhedral representation and transform it back to gimple. Using
8168 @option{-fgraphite-identity} we can check the costs or benefits of the
8169 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8170 are also performed by the code generator isl, like index splitting and
8171 dead code elimination in loops.
8173 @item -floop-nest-optimize
8174 @opindex floop-nest-optimize
8175 Enable the isl based loop nest optimizer. This is a generic loop nest
8176 optimizer based on the Pluto optimization algorithms. It calculates a loop
8177 structure optimized for data-locality and parallelism. This option
8180 @item -floop-parallelize-all
8181 @opindex floop-parallelize-all
8182 Use the Graphite data dependence analysis to identify loops that can
8183 be parallelized. Parallelize all the loops that can be analyzed to
8184 not contain loop carried dependences without checking that it is
8185 profitable to parallelize the loops.
8187 @item -ftree-coalesce-vars
8188 @opindex ftree-coalesce-vars
8189 While transforming the program out of the SSA representation, attempt to
8190 reduce copying by coalescing versions of different user-defined
8191 variables, instead of just compiler temporaries. This may severely
8192 limit the ability to debug an optimized program compiled with
8193 @option{-fno-var-tracking-assignments}. In the negated form, this flag
8194 prevents SSA coalescing of user variables. This option is enabled by
8195 default if optimization is enabled, and it does very little otherwise.
8197 @item -ftree-loop-if-convert
8198 @opindex ftree-loop-if-convert
8199 Attempt to transform conditional jumps in the innermost loops to
8200 branch-less equivalents. The intent is to remove control-flow from
8201 the innermost loops in order to improve the ability of the
8202 vectorization pass to handle these loops. This is enabled by default
8203 if vectorization is enabled.
8205 @item -ftree-loop-distribution
8206 @opindex ftree-loop-distribution
8207 Perform loop distribution. This flag can improve cache performance on
8208 big loop bodies and allow further loop optimizations, like
8209 parallelization or vectorization, to take place. For example, the loop
8226 @item -ftree-loop-distribute-patterns
8227 @opindex ftree-loop-distribute-patterns
8228 Perform loop distribution of patterns that can be code generated with
8229 calls to a library. This flag is enabled by default at @option{-O3}.
8231 This pass distributes the initialization loops and generates a call to
8232 memset zero. For example, the loop
8248 and the initialization loop is transformed into a call to memset zero.
8250 @item -ftree-loop-im
8251 @opindex ftree-loop-im
8252 Perform loop invariant motion on trees. This pass moves only invariants that
8253 are hard to handle at RTL level (function calls, operations that expand to
8254 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8255 operands of conditions that are invariant out of the loop, so that we can use
8256 just trivial invariantness analysis in loop unswitching. The pass also includes
8259 @item -ftree-loop-ivcanon
8260 @opindex ftree-loop-ivcanon
8261 Create a canonical counter for number of iterations in loops for which
8262 determining number of iterations requires complicated analysis. Later
8263 optimizations then may determine the number easily. Useful especially
8264 in connection with unrolling.
8268 Perform induction variable optimizations (strength reduction, induction
8269 variable merging and induction variable elimination) on trees.
8271 @item -ftree-parallelize-loops=n
8272 @opindex ftree-parallelize-loops
8273 Parallelize loops, i.e., split their iteration space to run in n threads.
8274 This is only possible for loops whose iterations are independent
8275 and can be arbitrarily reordered. The optimization is only
8276 profitable on multiprocessor machines, for loops that are CPU-intensive,
8277 rather than constrained e.g.@: by memory bandwidth. This option
8278 implies @option{-pthread}, and thus is only supported on targets
8279 that have support for @option{-pthread}.
8283 Perform function-local points-to analysis on trees. This flag is
8284 enabled by default at @option{-O} and higher.
8288 Perform scalar replacement of aggregates. This pass replaces structure
8289 references with scalars to prevent committing structures to memory too
8290 early. This flag is enabled by default at @option{-O} and higher.
8292 @item -fstore-merging
8293 @opindex fstore-merging
8294 Perform merging of narrow stores to consecutive memory addresses. This pass
8295 merges contiguous stores of immediate values narrower than a word into fewer
8296 wider stores to reduce the number of instructions. This is enabled by default
8297 at @option{-O} and higher.
8301 Perform temporary expression replacement during the SSA->normal phase. Single
8302 use/single def temporaries are replaced at their use location with their
8303 defining expression. This results in non-GIMPLE code, but gives the expanders
8304 much more complex trees to work on resulting in better RTL generation. This is
8305 enabled by default at @option{-O} and higher.
8309 Perform straight-line strength reduction on trees. This recognizes related
8310 expressions involving multiplications and replaces them by less expensive
8311 calculations when possible. This is enabled by default at @option{-O} and
8314 @item -ftree-vectorize
8315 @opindex ftree-vectorize
8316 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8317 and @option{-ftree-slp-vectorize} if not explicitly specified.
8319 @item -ftree-loop-vectorize
8320 @opindex ftree-loop-vectorize
8321 Perform loop vectorization on trees. This flag is enabled by default at
8322 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8324 @item -ftree-slp-vectorize
8325 @opindex ftree-slp-vectorize
8326 Perform basic block vectorization on trees. This flag is enabled by default at
8327 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8329 @item -fvect-cost-model=@var{model}
8330 @opindex fvect-cost-model
8331 Alter the cost model used for vectorization. The @var{model} argument
8332 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
8333 With the @samp{unlimited} model the vectorized code-path is assumed
8334 to be profitable while with the @samp{dynamic} model a runtime check
8335 guards the vectorized code-path to enable it only for iteration
8336 counts that will likely execute faster than when executing the original
8337 scalar loop. The @samp{cheap} model disables vectorization of
8338 loops where doing so would be cost prohibitive for example due to
8339 required runtime checks for data dependence or alignment but otherwise
8340 is equal to the @samp{dynamic} model.
8341 The default cost model depends on other optimization flags and is
8342 either @samp{dynamic} or @samp{cheap}.
8344 @item -fsimd-cost-model=@var{model}
8345 @opindex fsimd-cost-model
8346 Alter the cost model used for vectorization of loops marked with the OpenMP
8347 or Cilk Plus simd directive. The @var{model} argument should be one of
8348 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
8349 have the same meaning as described in @option{-fvect-cost-model} and by
8350 default a cost model defined with @option{-fvect-cost-model} is used.
8354 Perform Value Range Propagation on trees. This is similar to the
8355 constant propagation pass, but instead of values, ranges of values are
8356 propagated. This allows the optimizers to remove unnecessary range
8357 checks like array bound checks and null pointer checks. This is
8358 enabled by default at @option{-O2} and higher. Null pointer check
8359 elimination is only done if @option{-fdelete-null-pointer-checks} is
8363 @opindex fsplit-paths
8364 Split paths leading to loop backedges. This can improve dead code
8365 elimination and common subexpression elimination. This is enabled by
8366 default at @option{-O2} and above.
8368 @item -fsplit-ivs-in-unroller
8369 @opindex fsplit-ivs-in-unroller
8370 Enables expression of values of induction variables in later iterations
8371 of the unrolled loop using the value in the first iteration. This breaks
8372 long dependency chains, thus improving efficiency of the scheduling passes.
8374 A combination of @option{-fweb} and CSE is often sufficient to obtain the
8375 same effect. However, that is not reliable in cases where the loop body
8376 is more complicated than a single basic block. It also does not work at all
8377 on some architectures due to restrictions in the CSE pass.
8379 This optimization is enabled by default.
8381 @item -fvariable-expansion-in-unroller
8382 @opindex fvariable-expansion-in-unroller
8383 With this option, the compiler creates multiple copies of some
8384 local variables when unrolling a loop, which can result in superior code.
8386 @item -fpartial-inlining
8387 @opindex fpartial-inlining
8388 Inline parts of functions. This option has any effect only
8389 when inlining itself is turned on by the @option{-finline-functions}
8390 or @option{-finline-small-functions} options.
8392 Enabled at level @option{-O2}.
8394 @item -fpredictive-commoning
8395 @opindex fpredictive-commoning
8396 Perform predictive commoning optimization, i.e., reusing computations
8397 (especially memory loads and stores) performed in previous
8398 iterations of loops.
8400 This option is enabled at level @option{-O3}.
8402 @item -fprefetch-loop-arrays
8403 @opindex fprefetch-loop-arrays
8404 If supported by the target machine, generate instructions to prefetch
8405 memory to improve the performance of loops that access large arrays.
8407 This option may generate better or worse code; results are highly
8408 dependent on the structure of loops within the source code.
8410 Disabled at level @option{-Os}.
8412 @item -fno-printf-return-value
8413 @opindex fno-printf-return-value
8414 Do not substitute constants for known return value of formatted output
8415 functions such as @code{sprintf}, @code{snprintf}, @code{vsprintf}, and
8416 @code{vsnprintf} (but not @code{printf} of @code{fprintf}). This
8417 transformation allows GCC to optimize or even eliminate branches based
8418 on the known return value of these functions called with arguments that
8419 are either constant, or whose values are known to be in a range that
8420 makes determining the exact return value possible. For example, when
8421 @option{-fprintf-return-value} is in effect, both the branch and the
8422 body of the @code{if} statement (but not the call to @code{snprint})
8423 can be optimized away when @code{i} is a 32-bit or smaller integer
8424 because the return value is guaranteed to be at most 8.
8428 if (snprintf (buf, "%08x", i) >= sizeof buf)
8432 The @option{-fprintf-return-value} option relies on other optimizations
8433 and yields best results with @option{-O2}. It works in tandem with the
8434 @option{-Wformat-length} option. The @option{-fprintf-return-value}
8435 option is enabled by default.
8438 @itemx -fno-peephole2
8439 @opindex fno-peephole
8440 @opindex fno-peephole2
8441 Disable any machine-specific peephole optimizations. The difference
8442 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
8443 are implemented in the compiler; some targets use one, some use the
8444 other, a few use both.
8446 @option{-fpeephole} is enabled by default.
8447 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8449 @item -fno-guess-branch-probability
8450 @opindex fno-guess-branch-probability
8451 Do not guess branch probabilities using heuristics.
8453 GCC uses heuristics to guess branch probabilities if they are
8454 not provided by profiling feedback (@option{-fprofile-arcs}). These
8455 heuristics are based on the control flow graph. If some branch probabilities
8456 are specified by @code{__builtin_expect}, then the heuristics are
8457 used to guess branch probabilities for the rest of the control flow graph,
8458 taking the @code{__builtin_expect} info into account. The interactions
8459 between the heuristics and @code{__builtin_expect} can be complex, and in
8460 some cases, it may be useful to disable the heuristics so that the effects
8461 of @code{__builtin_expect} are easier to understand.
8463 The default is @option{-fguess-branch-probability} at levels
8464 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8466 @item -freorder-blocks
8467 @opindex freorder-blocks
8468 Reorder basic blocks in the compiled function in order to reduce number of
8469 taken branches and improve code locality.
8471 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8473 @item -freorder-blocks-algorithm=@var{algorithm}
8474 @opindex freorder-blocks-algorithm
8475 Use the specified algorithm for basic block reordering. The
8476 @var{algorithm} argument can be @samp{simple}, which does not increase
8477 code size (except sometimes due to secondary effects like alignment),
8478 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
8479 put all often executed code together, minimizing the number of branches
8480 executed by making extra copies of code.
8482 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
8483 @samp{stc} at levels @option{-O2}, @option{-O3}.
8485 @item -freorder-blocks-and-partition
8486 @opindex freorder-blocks-and-partition
8487 In addition to reordering basic blocks in the compiled function, in order
8488 to reduce number of taken branches, partitions hot and cold basic blocks
8489 into separate sections of the assembly and @file{.o} files, to improve
8490 paging and cache locality performance.
8492 This optimization is automatically turned off in the presence of
8493 exception handling, for linkonce sections, for functions with a user-defined
8494 section attribute and on any architecture that does not support named
8497 Enabled for x86 at levels @option{-O2}, @option{-O3}.
8499 @item -freorder-functions
8500 @opindex freorder-functions
8501 Reorder functions in the object file in order to
8502 improve code locality. This is implemented by using special
8503 subsections @code{.text.hot} for most frequently executed functions and
8504 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
8505 the linker so object file format must support named sections and linker must
8506 place them in a reasonable way.
8508 Also profile feedback must be available to make this option effective. See
8509 @option{-fprofile-arcs} for details.
8511 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8513 @item -fstrict-aliasing
8514 @opindex fstrict-aliasing
8515 Allow the compiler to assume the strictest aliasing rules applicable to
8516 the language being compiled. For C (and C++), this activates
8517 optimizations based on the type of expressions. In particular, an
8518 object of one type is assumed never to reside at the same address as an
8519 object of a different type, unless the types are almost the same. For
8520 example, an @code{unsigned int} can alias an @code{int}, but not a
8521 @code{void*} or a @code{double}. A character type may alias any other
8524 @anchor{Type-punning}Pay special attention to code like this:
8537 The practice of reading from a different union member than the one most
8538 recently written to (called ``type-punning'') is common. Even with
8539 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
8540 is accessed through the union type. So, the code above works as
8541 expected. @xref{Structures unions enumerations and bit-fields
8542 implementation}. However, this code might not:
8553 Similarly, access by taking the address, casting the resulting pointer
8554 and dereferencing the result has undefined behavior, even if the cast
8555 uses a union type, e.g.:
8559 return ((union a_union *) &d)->i;
8563 The @option{-fstrict-aliasing} option is enabled at levels
8564 @option{-O2}, @option{-O3}, @option{-Os}.
8566 @item -fstrict-overflow
8567 @opindex fstrict-overflow
8568 Allow the compiler to assume strict signed overflow rules, depending
8569 on the language being compiled. For C (and C++) this means that
8570 overflow when doing arithmetic with signed numbers is undefined, which
8571 means that the compiler may assume that it does not happen. This
8572 permits various optimizations. For example, the compiler assumes
8573 that an expression like @code{i + 10 > i} is always true for
8574 signed @code{i}. This assumption is only valid if signed overflow is
8575 undefined, as the expression is false if @code{i + 10} overflows when
8576 using twos complement arithmetic. When this option is in effect any
8577 attempt to determine whether an operation on signed numbers
8578 overflows must be written carefully to not actually involve overflow.
8580 This option also allows the compiler to assume strict pointer
8581 semantics: given a pointer to an object, if adding an offset to that
8582 pointer does not produce a pointer to the same object, the addition is
8583 undefined. This permits the compiler to conclude that @code{p + u >
8584 p} is always true for a pointer @code{p} and unsigned integer
8585 @code{u}. This assumption is only valid because pointer wraparound is
8586 undefined, as the expression is false if @code{p + u} overflows using
8587 twos complement arithmetic.
8589 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
8590 that integer signed overflow is fully defined: it wraps. When
8591 @option{-fwrapv} is used, there is no difference between
8592 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
8593 integers. With @option{-fwrapv} certain types of overflow are
8594 permitted. For example, if the compiler gets an overflow when doing
8595 arithmetic on constants, the overflowed value can still be used with
8596 @option{-fwrapv}, but not otherwise.
8598 The @option{-fstrict-overflow} option is enabled at levels
8599 @option{-O2}, @option{-O3}, @option{-Os}.
8601 @item -falign-functions
8602 @itemx -falign-functions=@var{n}
8603 @opindex falign-functions
8604 Align the start of functions to the next power-of-two greater than
8605 @var{n}, skipping up to @var{n} bytes. For instance,
8606 @option{-falign-functions=32} aligns functions to the next 32-byte
8607 boundary, but @option{-falign-functions=24} aligns to the next
8608 32-byte boundary only if this can be done by skipping 23 bytes or less.
8610 @option{-fno-align-functions} and @option{-falign-functions=1} are
8611 equivalent and mean that functions are not aligned.
8613 Some assemblers only support this flag when @var{n} is a power of two;
8614 in that case, it is rounded up.
8616 If @var{n} is not specified or is zero, use a machine-dependent default.
8618 Enabled at levels @option{-O2}, @option{-O3}.
8620 @item -flimit-function-alignment
8621 If this option is enabled, the compiler tries to avoid unnecessarily
8622 overaligning functions. It attempts to instruct the assembler to align
8623 by the amount specified by @option{-falign-functions}, but not to
8624 skip more bytes than the size of the function.
8626 @item -falign-labels
8627 @itemx -falign-labels=@var{n}
8628 @opindex falign-labels
8629 Align all branch targets to a power-of-two boundary, skipping up to
8630 @var{n} bytes like @option{-falign-functions}. This option can easily
8631 make code slower, because it must insert dummy operations for when the
8632 branch target is reached in the usual flow of the code.
8634 @option{-fno-align-labels} and @option{-falign-labels=1} are
8635 equivalent and mean that labels are not aligned.
8637 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
8638 are greater than this value, then their values are used instead.
8640 If @var{n} is not specified or is zero, use a machine-dependent default
8641 which is very likely to be @samp{1}, meaning no alignment.
8643 Enabled at levels @option{-O2}, @option{-O3}.
8646 @itemx -falign-loops=@var{n}
8647 @opindex falign-loops
8648 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
8649 like @option{-falign-functions}. If the loops are
8650 executed many times, this makes up for any execution of the dummy
8653 @option{-fno-align-loops} and @option{-falign-loops=1} are
8654 equivalent and mean that loops are not aligned.
8656 If @var{n} is not specified or is zero, use a machine-dependent default.
8658 Enabled at levels @option{-O2}, @option{-O3}.
8661 @itemx -falign-jumps=@var{n}
8662 @opindex falign-jumps
8663 Align branch targets to a power-of-two boundary, for branch targets
8664 where the targets can only be reached by jumping, skipping up to @var{n}
8665 bytes like @option{-falign-functions}. In this case, no dummy operations
8668 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
8669 equivalent and mean that loops are not aligned.
8671 If @var{n} is not specified or is zero, use a machine-dependent default.
8673 Enabled at levels @option{-O2}, @option{-O3}.
8675 @item -funit-at-a-time
8676 @opindex funit-at-a-time
8677 This option is left for compatibility reasons. @option{-funit-at-a-time}
8678 has no effect, while @option{-fno-unit-at-a-time} implies
8679 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
8683 @item -fno-toplevel-reorder
8684 @opindex fno-toplevel-reorder
8685 Do not reorder top-level functions, variables, and @code{asm}
8686 statements. Output them in the same order that they appear in the
8687 input file. When this option is used, unreferenced static variables
8688 are not removed. This option is intended to support existing code
8689 that relies on a particular ordering. For new code, it is better to
8690 use attributes when possible.
8692 Enabled at level @option{-O0}. When disabled explicitly, it also implies
8693 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
8698 Constructs webs as commonly used for register allocation purposes and assign
8699 each web individual pseudo register. This allows the register allocation pass
8700 to operate on pseudos directly, but also strengthens several other optimization
8701 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
8702 however, make debugging impossible, since variables no longer stay in a
8705 Enabled by default with @option{-funroll-loops}.
8707 @item -fwhole-program
8708 @opindex fwhole-program
8709 Assume that the current compilation unit represents the whole program being
8710 compiled. All public functions and variables with the exception of @code{main}
8711 and those merged by attribute @code{externally_visible} become static functions
8712 and in effect are optimized more aggressively by interprocedural optimizers.
8714 This option should not be used in combination with @option{-flto}.
8715 Instead relying on a linker plugin should provide safer and more precise
8718 @item -flto[=@var{n}]
8720 This option runs the standard link-time optimizer. When invoked
8721 with source code, it generates GIMPLE (one of GCC's internal
8722 representations) and writes it to special ELF sections in the object
8723 file. When the object files are linked together, all the function
8724 bodies are read from these ELF sections and instantiated as if they
8725 had been part of the same translation unit.
8727 To use the link-time optimizer, @option{-flto} and optimization
8728 options should be specified at compile time and during the final link.
8729 It is recommended that you compile all the files participating in the
8730 same link with the same options and also specify those options at
8735 gcc -c -O2 -flto foo.c
8736 gcc -c -O2 -flto bar.c
8737 gcc -o myprog -flto -O2 foo.o bar.o
8740 The first two invocations to GCC save a bytecode representation
8741 of GIMPLE into special ELF sections inside @file{foo.o} and
8742 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
8743 @file{foo.o} and @file{bar.o}, merges the two files into a single
8744 internal image, and compiles the result as usual. Since both
8745 @file{foo.o} and @file{bar.o} are merged into a single image, this
8746 causes all the interprocedural analyses and optimizations in GCC to
8747 work across the two files as if they were a single one. This means,
8748 for example, that the inliner is able to inline functions in
8749 @file{bar.o} into functions in @file{foo.o} and vice-versa.
8751 Another (simpler) way to enable link-time optimization is:
8754 gcc -o myprog -flto -O2 foo.c bar.c
8757 The above generates bytecode for @file{foo.c} and @file{bar.c},
8758 merges them together into a single GIMPLE representation and optimizes
8759 them as usual to produce @file{myprog}.
8761 The only important thing to keep in mind is that to enable link-time
8762 optimizations you need to use the GCC driver to perform the link step.
8763 GCC then automatically performs link-time optimization if any of the
8764 objects involved were compiled with the @option{-flto} command-line option.
8766 should specify the optimization options to be used for link-time
8767 optimization though GCC tries to be clever at guessing an
8768 optimization level to use from the options used at compile time
8769 if you fail to specify one at link time. You can always override
8770 the automatic decision to do link-time optimization at link time
8771 by passing @option{-fno-lto} to the link command.
8773 To make whole program optimization effective, it is necessary to make
8774 certain whole program assumptions. The compiler needs to know
8775 what functions and variables can be accessed by libraries and runtime
8776 outside of the link-time optimized unit. When supported by the linker,
8777 the linker plugin (see @option{-fuse-linker-plugin}) passes information
8778 to the compiler about used and externally visible symbols. When
8779 the linker plugin is not available, @option{-fwhole-program} should be
8780 used to allow the compiler to make these assumptions, which leads
8781 to more aggressive optimization decisions.
8783 When @option{-fuse-linker-plugin} is not enabled, when a file is
8784 compiled with @option{-flto}, the generated object file is larger than
8785 a regular object file because it contains GIMPLE bytecodes and the usual
8786 final code (see @option{-ffat-lto-objects}. This means that
8787 object files with LTO information can be linked as normal object
8788 files; if @option{-fno-lto} is passed to the linker, no
8789 interprocedural optimizations are applied. Note that when
8790 @option{-fno-fat-lto-objects} is enabled the compile stage is faster
8791 but you cannot perform a regular, non-LTO link on them.
8793 Additionally, the optimization flags used to compile individual files
8794 are not necessarily related to those used at link time. For instance,
8797 gcc -c -O0 -ffat-lto-objects -flto foo.c
8798 gcc -c -O0 -ffat-lto-objects -flto bar.c
8799 gcc -o myprog -O3 foo.o bar.o
8802 This produces individual object files with unoptimized assembler
8803 code, but the resulting binary @file{myprog} is optimized at
8804 @option{-O3}. If, instead, the final binary is generated with
8805 @option{-fno-lto}, then @file{myprog} is not optimized.
8807 When producing the final binary, GCC only
8808 applies link-time optimizations to those files that contain bytecode.
8809 Therefore, you can mix and match object files and libraries with
8810 GIMPLE bytecodes and final object code. GCC automatically selects
8811 which files to optimize in LTO mode and which files to link without
8814 There are some code generation flags preserved by GCC when
8815 generating bytecodes, as they need to be used during the final link
8816 stage. Generally options specified at link time override those
8817 specified at compile time.
8819 If you do not specify an optimization level option @option{-O} at
8820 link time, then GCC uses the highest optimization level
8821 used when compiling the object files.
8823 Currently, the following options and their settings are taken from
8824 the first object file that explicitly specifies them:
8825 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
8826 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
8827 and all the @option{-m} target flags.
8829 Certain ABI-changing flags are required to match in all compilation units,
8830 and trying to override this at link time with a conflicting value
8831 is ignored. This includes options such as @option{-freg-struct-return}
8832 and @option{-fpcc-struct-return}.
8834 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
8835 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
8836 are passed through to the link stage and merged conservatively for
8837 conflicting translation units. Specifically
8838 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
8839 precedence; and for example @option{-ffp-contract=off} takes precedence
8840 over @option{-ffp-contract=fast}. You can override them at link time.
8842 If LTO encounters objects with C linkage declared with incompatible
8843 types in separate translation units to be linked together (undefined
8844 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
8845 issued. The behavior is still undefined at run time. Similar
8846 diagnostics may be raised for other languages.
8848 Another feature of LTO is that it is possible to apply interprocedural
8849 optimizations on files written in different languages:
8854 gfortran -c -flto baz.f90
8855 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
8858 Notice that the final link is done with @command{g++} to get the C++
8859 runtime libraries and @option{-lgfortran} is added to get the Fortran
8860 runtime libraries. In general, when mixing languages in LTO mode, you
8861 should use the same link command options as when mixing languages in a
8862 regular (non-LTO) compilation.
8864 If object files containing GIMPLE bytecode are stored in a library archive, say
8865 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
8866 are using a linker with plugin support. To create static libraries suitable
8867 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
8868 and @command{ranlib};
8869 to show the symbols of object files with GIMPLE bytecode, use
8870 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
8871 and @command{nm} have been compiled with plugin support. At link time, use the the
8872 flag @option{-fuse-linker-plugin} to ensure that the library participates in
8873 the LTO optimization process:
8876 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
8879 With the linker plugin enabled, the linker extracts the needed
8880 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
8881 to make them part of the aggregated GIMPLE image to be optimized.
8883 If you are not using a linker with plugin support and/or do not
8884 enable the linker plugin, then the objects inside @file{libfoo.a}
8885 are extracted and linked as usual, but they do not participate
8886 in the LTO optimization process. In order to make a static library suitable
8887 for both LTO optimization and usual linkage, compile its object files with
8888 @option{-flto} @option{-ffat-lto-objects}.
8890 Link-time optimizations do not require the presence of the whole program to
8891 operate. If the program does not require any symbols to be exported, it is
8892 possible to combine @option{-flto} and @option{-fwhole-program} to allow
8893 the interprocedural optimizers to use more aggressive assumptions which may
8894 lead to improved optimization opportunities.
8895 Use of @option{-fwhole-program} is not needed when linker plugin is
8896 active (see @option{-fuse-linker-plugin}).
8898 The current implementation of LTO makes no
8899 attempt to generate bytecode that is portable between different
8900 types of hosts. The bytecode files are versioned and there is a
8901 strict version check, so bytecode files generated in one version of
8902 GCC do not work with an older or newer version of GCC.
8904 Link-time optimization does not work well with generation of debugging
8905 information. Combining @option{-flto} with
8906 @option{-g} is currently experimental and expected to produce unexpected
8909 If you specify the optional @var{n}, the optimization and code
8910 generation done at link time is executed in parallel using @var{n}
8911 parallel jobs by utilizing an installed @command{make} program. The
8912 environment variable @env{MAKE} may be used to override the program
8913 used. The default value for @var{n} is 1.
8915 You can also specify @option{-flto=jobserver} to use GNU make's
8916 job server mode to determine the number of parallel jobs. This
8917 is useful when the Makefile calling GCC is already executing in parallel.
8918 You must prepend a @samp{+} to the command recipe in the parent Makefile
8919 for this to work. This option likely only works if @env{MAKE} is
8922 @item -flto-partition=@var{alg}
8923 @opindex flto-partition
8924 Specify the partitioning algorithm used by the link-time optimizer.
8925 The value is either @samp{1to1} to specify a partitioning mirroring
8926 the original source files or @samp{balanced} to specify partitioning
8927 into equally sized chunks (whenever possible) or @samp{max} to create
8928 new partition for every symbol where possible. Specifying @samp{none}
8929 as an algorithm disables partitioning and streaming completely.
8930 The default value is @samp{balanced}. While @samp{1to1} can be used
8931 as an workaround for various code ordering issues, the @samp{max}
8932 partitioning is intended for internal testing only.
8933 The value @samp{one} specifies that exactly one partition should be
8934 used while the value @samp{none} bypasses partitioning and executes
8935 the link-time optimization step directly from the WPA phase.
8937 @item -flto-odr-type-merging
8938 @opindex flto-odr-type-merging
8939 Enable streaming of mangled types names of C++ types and their unification
8940 at link time. This increases size of LTO object files, but enables
8941 diagnostics about One Definition Rule violations.
8943 @item -flto-compression-level=@var{n}
8944 @opindex flto-compression-level
8945 This option specifies the level of compression used for intermediate
8946 language written to LTO object files, and is only meaningful in
8947 conjunction with LTO mode (@option{-flto}). Valid
8948 values are 0 (no compression) to 9 (maximum compression). Values
8949 outside this range are clamped to either 0 or 9. If the option is not
8950 given, a default balanced compression setting is used.
8952 @item -fuse-linker-plugin
8953 @opindex fuse-linker-plugin
8954 Enables the use of a linker plugin during link-time optimization. This
8955 option relies on plugin support in the linker, which is available in gold
8956 or in GNU ld 2.21 or newer.
8958 This option enables the extraction of object files with GIMPLE bytecode out
8959 of library archives. This improves the quality of optimization by exposing
8960 more code to the link-time optimizer. This information specifies what
8961 symbols can be accessed externally (by non-LTO object or during dynamic
8962 linking). Resulting code quality improvements on binaries (and shared
8963 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
8964 See @option{-flto} for a description of the effect of this flag and how to
8967 This option is enabled by default when LTO support in GCC is enabled
8968 and GCC was configured for use with
8969 a linker supporting plugins (GNU ld 2.21 or newer or gold).
8971 @item -ffat-lto-objects
8972 @opindex ffat-lto-objects
8973 Fat LTO objects are object files that contain both the intermediate language
8974 and the object code. This makes them usable for both LTO linking and normal
8975 linking. This option is effective only when compiling with @option{-flto}
8976 and is ignored at link time.
8978 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
8979 requires the complete toolchain to be aware of LTO. It requires a linker with
8980 linker plugin support for basic functionality. Additionally,
8981 @command{nm}, @command{ar} and @command{ranlib}
8982 need to support linker plugins to allow a full-featured build environment
8983 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
8984 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
8985 to these tools. With non fat LTO makefiles need to be modified to use them.
8987 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
8990 @item -fcompare-elim
8991 @opindex fcompare-elim
8992 After register allocation and post-register allocation instruction splitting,
8993 identify arithmetic instructions that compute processor flags similar to a
8994 comparison operation based on that arithmetic. If possible, eliminate the
8995 explicit comparison operation.
8997 This pass only applies to certain targets that cannot explicitly represent
8998 the comparison operation before register allocation is complete.
9000 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9002 @item -fcprop-registers
9003 @opindex fcprop-registers
9004 After register allocation and post-register allocation instruction splitting,
9005 perform a copy-propagation pass to try to reduce scheduling dependencies
9006 and occasionally eliminate the copy.
9008 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9010 @item -fprofile-correction
9011 @opindex fprofile-correction
9012 Profiles collected using an instrumented binary for multi-threaded programs may
9013 be inconsistent due to missed counter updates. When this option is specified,
9014 GCC uses heuristics to correct or smooth out such inconsistencies. By
9015 default, GCC emits an error message when an inconsistent profile is detected.
9018 @itemx -fprofile-use=@var{path}
9019 @opindex fprofile-use
9020 Enable profile feedback-directed optimizations,
9021 and the following optimizations
9022 which are generally profitable only with profile feedback available:
9023 @option{-fbranch-probabilities}, @option{-fvpt},
9024 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9025 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
9027 Before you can use this option, you must first generate profiling information.
9028 @xref{Instrumentation Options}, for information about the
9029 @option{-fprofile-generate} option.
9031 By default, GCC emits an error message if the feedback profiles do not
9032 match the source code. This error can be turned into a warning by using
9033 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9036 If @var{path} is specified, GCC looks at the @var{path} to find
9037 the profile feedback data files. See @option{-fprofile-dir}.
9039 @item -fauto-profile
9040 @itemx -fauto-profile=@var{path}
9041 @opindex fauto-profile
9042 Enable sampling-based feedback-directed optimizations,
9043 and the following optimizations
9044 which are generally profitable only with profile feedback available:
9045 @option{-fbranch-probabilities}, @option{-fvpt},
9046 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9047 @option{-ftree-vectorize},
9048 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
9049 @option{-fpredictive-commoning}, @option{-funswitch-loops},
9050 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
9052 @var{path} is the name of a file containing AutoFDO profile information.
9053 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
9055 Producing an AutoFDO profile data file requires running your program
9056 with the @command{perf} utility on a supported GNU/Linux target system.
9057 For more information, see @uref{https://perf.wiki.kernel.org/}.
9061 perf record -e br_inst_retired:near_taken -b -o perf.data \
9065 Then use the @command{create_gcov} tool to convert the raw profile data
9066 to a format that can be used by GCC.@ You must also supply the
9067 unstripped binary for your program to this tool.
9068 See @uref{https://github.com/google/autofdo}.
9072 create_gcov --binary=your_program.unstripped --profile=perf.data \
9077 The following options control compiler behavior regarding floating-point
9078 arithmetic. These options trade off between speed and
9079 correctness. All must be specifically enabled.
9083 @opindex ffloat-store
9084 Do not store floating-point variables in registers, and inhibit other
9085 options that might change whether a floating-point value is taken from a
9088 @cindex floating-point precision
9089 This option prevents undesirable excess precision on machines such as
9090 the 68000 where the floating registers (of the 68881) keep more
9091 precision than a @code{double} is supposed to have. Similarly for the
9092 x86 architecture. For most programs, the excess precision does only
9093 good, but a few programs rely on the precise definition of IEEE floating
9094 point. Use @option{-ffloat-store} for such programs, after modifying
9095 them to store all pertinent intermediate computations into variables.
9097 @item -fexcess-precision=@var{style}
9098 @opindex fexcess-precision
9099 This option allows further control over excess precision on machines
9100 where floating-point operations occur in a format with more precision or
9101 range than the IEEE standard and interchange floating-point types. By
9102 default, @option{-fexcess-precision=fast} is in effect; this means that
9103 operations may be carried out in a wider precision than the types specified
9104 in the source if that would result in faster code, and it is unpredictable
9105 when rounding to the types specified in the source code takes place.
9106 When compiling C, if @option{-fexcess-precision=standard} is specified then
9107 excess precision follows the rules specified in ISO C99; in particular,
9108 both casts and assignments cause values to be rounded to their
9109 semantic types (whereas @option{-ffloat-store} only affects
9110 assignments). This option is enabled by default for C if a strict
9111 conformance option such as @option{-std=c99} is used.
9114 @option{-fexcess-precision=standard} is not implemented for languages
9115 other than C, and has no effect if
9116 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
9117 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
9118 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9119 semantics apply without excess precision, and in the latter, rounding
9124 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9125 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9126 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
9128 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9130 This option is not turned on by any @option{-O} option besides
9131 @option{-Ofast} since it can result in incorrect output for programs
9132 that depend on an exact implementation of IEEE or ISO rules/specifications
9133 for math functions. It may, however, yield faster code for programs
9134 that do not require the guarantees of these specifications.
9136 @item -fno-math-errno
9137 @opindex fno-math-errno
9138 Do not set @code{errno} after calling math functions that are executed
9139 with a single instruction, e.g., @code{sqrt}. A program that relies on
9140 IEEE exceptions for math error handling may want to use this flag
9141 for speed while maintaining IEEE arithmetic compatibility.
9143 This option is not turned on by any @option{-O} option since
9144 it can result in incorrect output for programs that depend on
9145 an exact implementation of IEEE or ISO rules/specifications for
9146 math functions. It may, however, yield faster code for programs
9147 that do not require the guarantees of these specifications.
9149 The default is @option{-fmath-errno}.
9151 On Darwin systems, the math library never sets @code{errno}. There is
9152 therefore no reason for the compiler to consider the possibility that
9153 it might, and @option{-fno-math-errno} is the default.
9155 @item -funsafe-math-optimizations
9156 @opindex funsafe-math-optimizations
9158 Allow optimizations for floating-point arithmetic that (a) assume
9159 that arguments and results are valid and (b) may violate IEEE or
9160 ANSI standards. When used at link time, it may include libraries
9161 or startup files that change the default FPU control word or other
9162 similar optimizations.
9164 This option is not turned on by any @option{-O} option since
9165 it can result in incorrect output for programs that depend on
9166 an exact implementation of IEEE or ISO rules/specifications for
9167 math functions. It may, however, yield faster code for programs
9168 that do not require the guarantees of these specifications.
9169 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9170 @option{-fassociative-math} and @option{-freciprocal-math}.
9172 The default is @option{-fno-unsafe-math-optimizations}.
9174 @item -fassociative-math
9175 @opindex fassociative-math
9177 Allow re-association of operands in series of floating-point operations.
9178 This violates the ISO C and C++ language standard by possibly changing
9179 computation result. NOTE: re-ordering may change the sign of zero as
9180 well as ignore NaNs and inhibit or create underflow or overflow (and
9181 thus cannot be used on code that relies on rounding behavior like
9182 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9183 and thus may not be used when ordered comparisons are required.
9184 This option requires that both @option{-fno-signed-zeros} and
9185 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9186 much sense with @option{-frounding-math}. For Fortran the option
9187 is automatically enabled when both @option{-fno-signed-zeros} and
9188 @option{-fno-trapping-math} are in effect.
9190 The default is @option{-fno-associative-math}.
9192 @item -freciprocal-math
9193 @opindex freciprocal-math
9195 Allow the reciprocal of a value to be used instead of dividing by
9196 the value if this enables optimizations. For example @code{x / y}
9197 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9198 is subject to common subexpression elimination. Note that this loses
9199 precision and increases the number of flops operating on the value.
9201 The default is @option{-fno-reciprocal-math}.
9203 @item -ffinite-math-only
9204 @opindex ffinite-math-only
9205 Allow optimizations for floating-point arithmetic that assume
9206 that arguments and results are not NaNs or +-Infs.
9208 This option is not turned on by any @option{-O} option since
9209 it can result in incorrect output for programs that depend on
9210 an exact implementation of IEEE or ISO rules/specifications for
9211 math functions. It may, however, yield faster code for programs
9212 that do not require the guarantees of these specifications.
9214 The default is @option{-fno-finite-math-only}.
9216 @item -fno-signed-zeros
9217 @opindex fno-signed-zeros
9218 Allow optimizations for floating-point arithmetic that ignore the
9219 signedness of zero. IEEE arithmetic specifies the behavior of
9220 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9221 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9222 This option implies that the sign of a zero result isn't significant.
9224 The default is @option{-fsigned-zeros}.
9226 @item -fno-trapping-math
9227 @opindex fno-trapping-math
9228 Compile code assuming that floating-point operations cannot generate
9229 user-visible traps. These traps include division by zero, overflow,
9230 underflow, inexact result and invalid operation. This option requires
9231 that @option{-fno-signaling-nans} be in effect. Setting this option may
9232 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9234 This option should never be turned on by any @option{-O} option since
9235 it can result in incorrect output for programs that depend on
9236 an exact implementation of IEEE or ISO rules/specifications for
9239 The default is @option{-ftrapping-math}.
9241 @item -frounding-math
9242 @opindex frounding-math
9243 Disable transformations and optimizations that assume default floating-point
9244 rounding behavior. This is round-to-zero for all floating point
9245 to integer conversions, and round-to-nearest for all other arithmetic
9246 truncations. This option should be specified for programs that change
9247 the FP rounding mode dynamically, or that may be executed with a
9248 non-default rounding mode. This option disables constant folding of
9249 floating-point expressions at compile time (which may be affected by
9250 rounding mode) and arithmetic transformations that are unsafe in the
9251 presence of sign-dependent rounding modes.
9253 The default is @option{-fno-rounding-math}.
9255 This option is experimental and does not currently guarantee to
9256 disable all GCC optimizations that are affected by rounding mode.
9257 Future versions of GCC may provide finer control of this setting
9258 using C99's @code{FENV_ACCESS} pragma. This command-line option
9259 will be used to specify the default state for @code{FENV_ACCESS}.
9261 @item -fsignaling-nans
9262 @opindex fsignaling-nans
9263 Compile code assuming that IEEE signaling NaNs may generate user-visible
9264 traps during floating-point operations. Setting this option disables
9265 optimizations that may change the number of exceptions visible with
9266 signaling NaNs. This option implies @option{-ftrapping-math}.
9268 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9271 The default is @option{-fno-signaling-nans}.
9273 This option is experimental and does not currently guarantee to
9274 disable all GCC optimizations that affect signaling NaN behavior.
9276 @item -fno-fp-int-builtin-inexact
9277 @opindex fno-fp-int-builtin-inexact
9278 Do not allow the built-in functions @code{ceil}, @code{floor},
9279 @code{round} and @code{trunc}, and their @code{float} and @code{long
9280 double} variants, to generate code that raises the ``inexact''
9281 floating-point exception for noninteger arguments. ISO C99 and C11
9282 allow these functions to raise the ``inexact'' exception, but ISO/IEC
9283 TS 18661-1:2014, the C bindings to IEEE 754-2008, does not allow these
9286 The default is @option{-ffp-int-builtin-inexact}, allowing the
9287 exception to be raised. This option does nothing unless
9288 @option{-ftrapping-math} is in effect.
9290 Even if @option{-fno-fp-int-builtin-inexact} is used, if the functions
9291 generate a call to a library function then the ``inexact'' exception
9292 may be raised if the library implementation does not follow TS 18661.
9294 @item -fsingle-precision-constant
9295 @opindex fsingle-precision-constant
9296 Treat floating-point constants as single precision instead of
9297 implicitly converting them to double-precision constants.
9299 @item -fcx-limited-range
9300 @opindex fcx-limited-range
9301 When enabled, this option states that a range reduction step is not
9302 needed when performing complex division. Also, there is no checking
9303 whether the result of a complex multiplication or division is @code{NaN
9304 + I*NaN}, with an attempt to rescue the situation in that case. The
9305 default is @option{-fno-cx-limited-range}, but is enabled by
9306 @option{-ffast-math}.
9308 This option controls the default setting of the ISO C99
9309 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9312 @item -fcx-fortran-rules
9313 @opindex fcx-fortran-rules
9314 Complex multiplication and division follow Fortran rules. Range
9315 reduction is done as part of complex division, but there is no checking
9316 whether the result of a complex multiplication or division is @code{NaN
9317 + I*NaN}, with an attempt to rescue the situation in that case.
9319 The default is @option{-fno-cx-fortran-rules}.
9323 The following options control optimizations that may improve
9324 performance, but are not enabled by any @option{-O} options. This
9325 section includes experimental options that may produce broken code.
9328 @item -fbranch-probabilities
9329 @opindex fbranch-probabilities
9330 After running a program compiled with @option{-fprofile-arcs}
9331 (@pxref{Instrumentation Options}),
9332 you can compile it a second time using
9333 @option{-fbranch-probabilities}, to improve optimizations based on
9334 the number of times each branch was taken. When a program
9335 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9336 counts to a file called @file{@var{sourcename}.gcda} for each source
9337 file. The information in this data file is very dependent on the
9338 structure of the generated code, so you must use the same source code
9339 and the same optimization options for both compilations.
9341 With @option{-fbranch-probabilities}, GCC puts a
9342 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9343 These can be used to improve optimization. Currently, they are only
9344 used in one place: in @file{reorg.c}, instead of guessing which path a
9345 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9346 exactly determine which path is taken more often.
9348 @item -fprofile-values
9349 @opindex fprofile-values
9350 If combined with @option{-fprofile-arcs}, it adds code so that some
9351 data about values of expressions in the program is gathered.
9353 With @option{-fbranch-probabilities}, it reads back the data gathered
9354 from profiling values of expressions for usage in optimizations.
9356 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9358 @item -fprofile-reorder-functions
9359 @opindex fprofile-reorder-functions
9360 Function reordering based on profile instrumentation collects
9361 first time of execution of a function and orders these functions
9364 Enabled with @option{-fprofile-use}.
9368 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9369 to add code to gather information about values of expressions.
9371 With @option{-fbranch-probabilities}, it reads back the data gathered
9372 and actually performs the optimizations based on them.
9373 Currently the optimizations include specialization of division operations
9374 using the knowledge about the value of the denominator.
9376 @item -frename-registers
9377 @opindex frename-registers
9378 Attempt to avoid false dependencies in scheduled code by making use
9379 of registers left over after register allocation. This optimization
9380 most benefits processors with lots of registers. Depending on the
9381 debug information format adopted by the target, however, it can
9382 make debugging impossible, since variables no longer stay in
9383 a ``home register''.
9385 Enabled by default with @option{-funroll-loops}.
9387 @item -fschedule-fusion
9388 @opindex fschedule-fusion
9389 Performs a target dependent pass over the instruction stream to schedule
9390 instructions of same type together because target machine can execute them
9391 more efficiently if they are adjacent to each other in the instruction flow.
9393 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9397 Perform tail duplication to enlarge superblock size. This transformation
9398 simplifies the control flow of the function allowing other optimizations to do
9401 Enabled with @option{-fprofile-use}.
9403 @item -funroll-loops
9404 @opindex funroll-loops
9405 Unroll loops whose number of iterations can be determined at compile time or
9406 upon entry to the loop. @option{-funroll-loops} implies
9407 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
9408 It also turns on complete loop peeling (i.e.@: complete removal of loops with
9409 a small constant number of iterations). This option makes code larger, and may
9410 or may not make it run faster.
9412 Enabled with @option{-fprofile-use}.
9414 @item -funroll-all-loops
9415 @opindex funroll-all-loops
9416 Unroll all loops, even if their number of iterations is uncertain when
9417 the loop is entered. This usually makes programs run more slowly.
9418 @option{-funroll-all-loops} implies the same options as
9419 @option{-funroll-loops}.
9422 @opindex fpeel-loops
9423 Peels loops for which there is enough information that they do not
9424 roll much (from profile feedback or static analysis). It also turns on
9425 complete loop peeling (i.e.@: complete removal of loops with small constant
9426 number of iterations).
9428 Enabled with @option{-O3} and/or @option{-fprofile-use}.
9430 @item -fmove-loop-invariants
9431 @opindex fmove-loop-invariants
9432 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
9433 at level @option{-O1}
9436 @opindex fsplit-loops
9437 Split a loop into two if it contains a condition that's always true
9438 for one side of the iteration space and false for the other.
9440 @item -funswitch-loops
9441 @opindex funswitch-loops
9442 Move branches with loop invariant conditions out of the loop, with duplicates
9443 of the loop on both branches (modified according to result of the condition).
9445 @item -ffunction-sections
9446 @itemx -fdata-sections
9447 @opindex ffunction-sections
9448 @opindex fdata-sections
9449 Place each function or data item into its own section in the output
9450 file if the target supports arbitrary sections. The name of the
9451 function or the name of the data item determines the section's name
9454 Use these options on systems where the linker can perform optimizations
9455 to improve locality of reference in the instruction space. Most systems
9456 using the ELF object format and SPARC processors running Solaris 2 have
9457 linkers with such optimizations. AIX may have these optimizations in
9460 Only use these options when there are significant benefits from doing
9461 so. When you specify these options, the assembler and linker
9462 create larger object and executable files and are also slower.
9463 You cannot use @command{gprof} on all systems if you
9464 specify this option, and you may have problems with debugging if
9465 you specify both this option and @option{-g}.
9467 @item -fbranch-target-load-optimize
9468 @opindex fbranch-target-load-optimize
9469 Perform branch target register load optimization before prologue / epilogue
9471 The use of target registers can typically be exposed only during reload,
9472 thus hoisting loads out of loops and doing inter-block scheduling needs
9473 a separate optimization pass.
9475 @item -fbranch-target-load-optimize2
9476 @opindex fbranch-target-load-optimize2
9477 Perform branch target register load optimization after prologue / epilogue
9480 @item -fbtr-bb-exclusive
9481 @opindex fbtr-bb-exclusive
9482 When performing branch target register load optimization, don't reuse
9483 branch target registers within any basic block.
9486 @opindex fstdarg-opt
9487 Optimize the prologue of variadic argument functions with respect to usage of
9490 @item -fsection-anchors
9491 @opindex fsection-anchors
9492 Try to reduce the number of symbolic address calculations by using
9493 shared ``anchor'' symbols to address nearby objects. This transformation
9494 can help to reduce the number of GOT entries and GOT accesses on some
9497 For example, the implementation of the following function @code{foo}:
9501 int foo (void) @{ return a + b + c; @}
9505 usually calculates the addresses of all three variables, but if you
9506 compile it with @option{-fsection-anchors}, it accesses the variables
9507 from a common anchor point instead. The effect is similar to the
9508 following pseudocode (which isn't valid C):
9513 register int *xr = &x;
9514 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
9518 Not all targets support this option.
9520 @item --param @var{name}=@var{value}
9522 In some places, GCC uses various constants to control the amount of
9523 optimization that is done. For example, GCC does not inline functions
9524 that contain more than a certain number of instructions. You can
9525 control some of these constants on the command line using the
9526 @option{--param} option.
9528 The names of specific parameters, and the meaning of the values, are
9529 tied to the internals of the compiler, and are subject to change
9530 without notice in future releases.
9532 In each case, the @var{value} is an integer. The allowable choices for
9536 @item predictable-branch-outcome
9537 When branch is predicted to be taken with probability lower than this threshold
9538 (in percent), then it is considered well predictable. The default is 10.
9540 @item max-rtl-if-conversion-insns
9541 RTL if-conversion tries to remove conditional branches around a block and
9542 replace them with conditionally executed instructions. This parameter
9543 gives the maximum number of instructions in a block which should be
9544 considered for if-conversion. The default is 10, though the compiler will
9545 also use other heuristics to decide whether if-conversion is likely to be
9548 @item max-rtl-if-conversion-predictable-cost
9549 @item max-rtl-if-conversion-unpredictable-cost
9550 RTL if-conversion will try to remove conditional branches around a block
9551 and replace them with conditionally executed instructions. These parameters
9552 give the maximum permissible cost for the sequence that would be generated
9553 by if-conversion depending on whether the branch is statically determined
9554 to be predictable or not. The units for this parameter are the same as
9555 those for the GCC internal seq_cost metric. The compiler will try to
9556 provide a reasonable default for this parameter using the BRANCH_COST
9559 @item max-crossjump-edges
9560 The maximum number of incoming edges to consider for cross-jumping.
9561 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
9562 the number of edges incoming to each block. Increasing values mean
9563 more aggressive optimization, making the compilation time increase with
9564 probably small improvement in executable size.
9566 @item min-crossjump-insns
9567 The minimum number of instructions that must be matched at the end
9568 of two blocks before cross-jumping is performed on them. This
9569 value is ignored in the case where all instructions in the block being
9570 cross-jumped from are matched. The default value is 5.
9572 @item max-grow-copy-bb-insns
9573 The maximum code size expansion factor when copying basic blocks
9574 instead of jumping. The expansion is relative to a jump instruction.
9575 The default value is 8.
9577 @item max-goto-duplication-insns
9578 The maximum number of instructions to duplicate to a block that jumps
9579 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
9580 passes, GCC factors computed gotos early in the compilation process,
9581 and unfactors them as late as possible. Only computed jumps at the
9582 end of a basic blocks with no more than max-goto-duplication-insns are
9583 unfactored. The default value is 8.
9585 @item max-delay-slot-insn-search
9586 The maximum number of instructions to consider when looking for an
9587 instruction to fill a delay slot. If more than this arbitrary number of
9588 instructions are searched, the time savings from filling the delay slot
9589 are minimal, so stop searching. Increasing values mean more
9590 aggressive optimization, making the compilation time increase with probably
9591 small improvement in execution time.
9593 @item max-delay-slot-live-search
9594 When trying to fill delay slots, the maximum number of instructions to
9595 consider when searching for a block with valid live register
9596 information. Increasing this arbitrarily chosen value means more
9597 aggressive optimization, increasing the compilation time. This parameter
9598 should be removed when the delay slot code is rewritten to maintain the
9601 @item max-gcse-memory
9602 The approximate maximum amount of memory that can be allocated in
9603 order to perform the global common subexpression elimination
9604 optimization. If more memory than specified is required, the
9605 optimization is not done.
9607 @item max-gcse-insertion-ratio
9608 If the ratio of expression insertions to deletions is larger than this value
9609 for any expression, then RTL PRE inserts or removes the expression and thus
9610 leaves partially redundant computations in the instruction stream. The default value is 20.
9612 @item max-pending-list-length
9613 The maximum number of pending dependencies scheduling allows
9614 before flushing the current state and starting over. Large functions
9615 with few branches or calls can create excessively large lists which
9616 needlessly consume memory and resources.
9618 @item max-modulo-backtrack-attempts
9619 The maximum number of backtrack attempts the scheduler should make
9620 when modulo scheduling a loop. Larger values can exponentially increase
9623 @item max-inline-insns-single
9624 Several parameters control the tree inliner used in GCC@.
9625 This number sets the maximum number of instructions (counted in GCC's
9626 internal representation) in a single function that the tree inliner
9627 considers for inlining. This only affects functions declared
9628 inline and methods implemented in a class declaration (C++).
9629 The default value is 400.
9631 @item max-inline-insns-auto
9632 When you use @option{-finline-functions} (included in @option{-O3}),
9633 a lot of functions that would otherwise not be considered for inlining
9634 by the compiler are investigated. To those functions, a different
9635 (more restrictive) limit compared to functions declared inline can
9637 The default value is 40.
9639 @item inline-min-speedup
9640 When estimated performance improvement of caller + callee runtime exceeds this
9641 threshold (in precent), the function can be inlined regardless the limit on
9642 @option{--param max-inline-insns-single} and @option{--param
9643 max-inline-insns-auto}.
9645 @item large-function-insns
9646 The limit specifying really large functions. For functions larger than this
9647 limit after inlining, inlining is constrained by
9648 @option{--param large-function-growth}. This parameter is useful primarily
9649 to avoid extreme compilation time caused by non-linear algorithms used by the
9651 The default value is 2700.
9653 @item large-function-growth
9654 Specifies maximal growth of large function caused by inlining in percents.
9655 The default value is 100 which limits large function growth to 2.0 times
9658 @item large-unit-insns
9659 The limit specifying large translation unit. Growth caused by inlining of
9660 units larger than this limit is limited by @option{--param inline-unit-growth}.
9661 For small units this might be too tight.
9662 For example, consider a unit consisting of function A
9663 that is inline and B that just calls A three times. If B is small relative to
9664 A, the growth of unit is 300\% and yet such inlining is very sane. For very
9665 large units consisting of small inlineable functions, however, the overall unit
9666 growth limit is needed to avoid exponential explosion of code size. Thus for
9667 smaller units, the size is increased to @option{--param large-unit-insns}
9668 before applying @option{--param inline-unit-growth}. The default is 10000.
9670 @item inline-unit-growth
9671 Specifies maximal overall growth of the compilation unit caused by inlining.
9672 The default value is 20 which limits unit growth to 1.2 times the original
9673 size. Cold functions (either marked cold via an attribute or by profile
9674 feedback) are not accounted into the unit size.
9676 @item ipcp-unit-growth
9677 Specifies maximal overall growth of the compilation unit caused by
9678 interprocedural constant propagation. The default value is 10 which limits
9679 unit growth to 1.1 times the original size.
9681 @item large-stack-frame
9682 The limit specifying large stack frames. While inlining the algorithm is trying
9683 to not grow past this limit too much. The default value is 256 bytes.
9685 @item large-stack-frame-growth
9686 Specifies maximal growth of large stack frames caused by inlining in percents.
9687 The default value is 1000 which limits large stack frame growth to 11 times
9690 @item max-inline-insns-recursive
9691 @itemx max-inline-insns-recursive-auto
9692 Specifies the maximum number of instructions an out-of-line copy of a
9693 self-recursive inline
9694 function can grow into by performing recursive inlining.
9696 @option{--param max-inline-insns-recursive} applies to functions
9698 For functions not declared inline, recursive inlining
9699 happens only when @option{-finline-functions} (included in @option{-O3}) is
9700 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
9701 default value is 450.
9703 @item max-inline-recursive-depth
9704 @itemx max-inline-recursive-depth-auto
9705 Specifies the maximum recursion depth used for recursive inlining.
9707 @option{--param max-inline-recursive-depth} applies to functions
9708 declared inline. For functions not declared inline, recursive inlining
9709 happens only when @option{-finline-functions} (included in @option{-O3}) is
9710 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
9713 @item min-inline-recursive-probability
9714 Recursive inlining is profitable only for function having deep recursion
9715 in average and can hurt for function having little recursion depth by
9716 increasing the prologue size or complexity of function body to other
9719 When profile feedback is available (see @option{-fprofile-generate}) the actual
9720 recursion depth can be guessed from the probability that function recurses
9721 via a given call expression. This parameter limits inlining only to call
9722 expressions whose probability exceeds the given threshold (in percents).
9723 The default value is 10.
9725 @item early-inlining-insns
9726 Specify growth that the early inliner can make. In effect it increases
9727 the amount of inlining for code having a large abstraction penalty.
9728 The default value is 14.
9730 @item max-early-inliner-iterations
9731 Limit of iterations of the early inliner. This basically bounds
9732 the number of nested indirect calls the early inliner can resolve.
9733 Deeper chains are still handled by late inlining.
9735 @item comdat-sharing-probability
9736 Probability (in percent) that C++ inline function with comdat visibility
9737 are shared across multiple compilation units. The default value is 20.
9739 @item profile-func-internal-id
9740 A parameter to control whether to use function internal id in profile
9741 database lookup. If the value is 0, the compiler uses an id that
9742 is based on function assembler name and filename, which makes old profile
9743 data more tolerant to source changes such as function reordering etc.
9744 The default value is 0.
9746 @item min-vect-loop-bound
9747 The minimum number of iterations under which loops are not vectorized
9748 when @option{-ftree-vectorize} is used. The number of iterations after
9749 vectorization needs to be greater than the value specified by this option
9750 to allow vectorization. The default value is 0.
9752 @item gcse-cost-distance-ratio
9753 Scaling factor in calculation of maximum distance an expression
9754 can be moved by GCSE optimizations. This is currently supported only in the
9755 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
9756 is with simple expressions, i.e., the expressions that have cost
9757 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
9758 hoisting of simple expressions. The default value is 10.
9760 @item gcse-unrestricted-cost
9761 Cost, roughly measured as the cost of a single typical machine
9762 instruction, at which GCSE optimizations do not constrain
9763 the distance an expression can travel. This is currently
9764 supported only in the code hoisting pass. The lesser the cost,
9765 the more aggressive code hoisting is. Specifying 0
9766 allows all expressions to travel unrestricted distances.
9767 The default value is 3.
9769 @item max-hoist-depth
9770 The depth of search in the dominator tree for expressions to hoist.
9771 This is used to avoid quadratic behavior in hoisting algorithm.
9772 The value of 0 does not limit on the search, but may slow down compilation
9773 of huge functions. The default value is 30.
9775 @item max-tail-merge-comparisons
9776 The maximum amount of similar bbs to compare a bb with. This is used to
9777 avoid quadratic behavior in tree tail merging. The default value is 10.
9779 @item max-tail-merge-iterations
9780 The maximum amount of iterations of the pass over the function. This is used to
9781 limit compilation time in tree tail merging. The default value is 2.
9783 @item store-merging-allow-unaligned
9784 Allow the store merging pass to introduce unaligned stores if it is legal to
9785 do so. The default value is 1.
9787 @item max-stores-to-merge
9788 The maximum number of stores to attempt to merge into wider stores in the store
9789 merging pass. The minimum value is 2 and the default is 64.
9791 @item max-unrolled-insns
9792 The maximum number of instructions that a loop may have to be unrolled.
9793 If a loop is unrolled, this parameter also determines how many times
9794 the loop code is unrolled.
9796 @item max-average-unrolled-insns
9797 The maximum number of instructions biased by probabilities of their execution
9798 that a loop may have to be unrolled. If a loop is unrolled,
9799 this parameter also determines how many times the loop code is unrolled.
9801 @item max-unroll-times
9802 The maximum number of unrollings of a single loop.
9804 @item max-peeled-insns
9805 The maximum number of instructions that a loop may have to be peeled.
9806 If a loop is peeled, this parameter also determines how many times
9807 the loop code is peeled.
9809 @item max-peel-times
9810 The maximum number of peelings of a single loop.
9812 @item max-peel-branches
9813 The maximum number of branches on the hot path through the peeled sequence.
9815 @item max-completely-peeled-insns
9816 The maximum number of insns of a completely peeled loop.
9818 @item max-completely-peel-times
9819 The maximum number of iterations of a loop to be suitable for complete peeling.
9821 @item max-completely-peel-loop-nest-depth
9822 The maximum depth of a loop nest suitable for complete peeling.
9824 @item max-unswitch-insns
9825 The maximum number of insns of an unswitched loop.
9827 @item max-unswitch-level
9828 The maximum number of branches unswitched in a single loop.
9830 @item max-loop-headers-insns
9831 The maximum number of insns in loop header duplicated by he copy loop headers
9835 The minimum cost of an expensive expression in the loop invariant motion.
9837 @item iv-consider-all-candidates-bound
9838 Bound on number of candidates for induction variables, below which
9839 all candidates are considered for each use in induction variable
9840 optimizations. If there are more candidates than this,
9841 only the most relevant ones are considered to avoid quadratic time complexity.
9843 @item iv-max-considered-uses
9844 The induction variable optimizations give up on loops that contain more
9845 induction variable uses.
9847 @item iv-always-prune-cand-set-bound
9848 If the number of candidates in the set is smaller than this value,
9849 always try to remove unnecessary ivs from the set
9850 when adding a new one.
9852 @item avg-loop-niter
9853 Average number of iterations of a loop.
9855 @item scev-max-expr-size
9856 Bound on size of expressions used in the scalar evolutions analyzer.
9857 Large expressions slow the analyzer.
9859 @item scev-max-expr-complexity
9860 Bound on the complexity of the expressions in the scalar evolutions analyzer.
9861 Complex expressions slow the analyzer.
9863 @item max-tree-if-conversion-phi-args
9864 Maximum number of arguments in a PHI supported by TREE if conversion
9865 unless the loop is marked with simd pragma.
9867 @item vect-max-version-for-alignment-checks
9868 The maximum number of run-time checks that can be performed when
9869 doing loop versioning for alignment in the vectorizer.
9871 @item vect-max-version-for-alias-checks
9872 The maximum number of run-time checks that can be performed when
9873 doing loop versioning for alias in the vectorizer.
9875 @item vect-max-peeling-for-alignment
9876 The maximum number of loop peels to enhance access alignment
9877 for vectorizer. Value -1 means no limit.
9879 @item max-iterations-to-track
9880 The maximum number of iterations of a loop the brute-force algorithm
9881 for analysis of the number of iterations of the loop tries to evaluate.
9883 @item hot-bb-count-ws-permille
9884 A basic block profile count is considered hot if it contributes to
9885 the given permillage (i.e. 0...1000) of the entire profiled execution.
9887 @item hot-bb-frequency-fraction
9888 Select fraction of the entry block frequency of executions of basic block in
9889 function given basic block needs to have to be considered hot.
9891 @item max-predicted-iterations
9892 The maximum number of loop iterations we predict statically. This is useful
9893 in cases where a function contains a single loop with known bound and
9894 another loop with unknown bound.
9895 The known number of iterations is predicted correctly, while
9896 the unknown number of iterations average to roughly 10. This means that the
9897 loop without bounds appears artificially cold relative to the other one.
9899 @item builtin-expect-probability
9900 Control the probability of the expression having the specified value. This
9901 parameter takes a percentage (i.e. 0 ... 100) as input.
9902 The default probability of 90 is obtained empirically.
9904 @item align-threshold
9906 Select fraction of the maximal frequency of executions of a basic block in
9907 a function to align the basic block.
9909 @item align-loop-iterations
9911 A loop expected to iterate at least the selected number of iterations is
9914 @item tracer-dynamic-coverage
9915 @itemx tracer-dynamic-coverage-feedback
9917 This value is used to limit superblock formation once the given percentage of
9918 executed instructions is covered. This limits unnecessary code size
9921 The @option{tracer-dynamic-coverage-feedback} parameter
9922 is used only when profile
9923 feedback is available. The real profiles (as opposed to statically estimated
9924 ones) are much less balanced allowing the threshold to be larger value.
9926 @item tracer-max-code-growth
9927 Stop tail duplication once code growth has reached given percentage. This is
9928 a rather artificial limit, as most of the duplicates are eliminated later in
9929 cross jumping, so it may be set to much higher values than is the desired code
9932 @item tracer-min-branch-ratio
9934 Stop reverse growth when the reverse probability of best edge is less than this
9935 threshold (in percent).
9937 @item tracer-min-branch-probability
9938 @itemx tracer-min-branch-probability-feedback
9940 Stop forward growth if the best edge has probability lower than this
9943 Similarly to @option{tracer-dynamic-coverage} two parameters are
9944 provided. @option{tracer-min-branch-probability-feedback} is used for
9945 compilation with profile feedback and @option{tracer-min-branch-probability}
9946 compilation without. The value for compilation with profile feedback
9947 needs to be more conservative (higher) in order to make tracer
9950 @item max-cse-path-length
9952 The maximum number of basic blocks on path that CSE considers.
9956 The maximum number of instructions CSE processes before flushing.
9957 The default is 1000.
9959 @item ggc-min-expand
9961 GCC uses a garbage collector to manage its own memory allocation. This
9962 parameter specifies the minimum percentage by which the garbage
9963 collector's heap should be allowed to expand between collections.
9964 Tuning this may improve compilation speed; it has no effect on code
9967 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
9968 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
9969 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
9970 GCC is not able to calculate RAM on a particular platform, the lower
9971 bound of 30% is used. Setting this parameter and
9972 @option{ggc-min-heapsize} to zero causes a full collection to occur at
9973 every opportunity. This is extremely slow, but can be useful for
9976 @item ggc-min-heapsize
9978 Minimum size of the garbage collector's heap before it begins bothering
9979 to collect garbage. The first collection occurs after the heap expands
9980 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
9981 tuning this may improve compilation speed, and has no effect on code
9984 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
9985 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
9986 with a lower bound of 4096 (four megabytes) and an upper bound of
9987 131072 (128 megabytes). If GCC is not able to calculate RAM on a
9988 particular platform, the lower bound is used. Setting this parameter
9989 very large effectively disables garbage collection. Setting this
9990 parameter and @option{ggc-min-expand} to zero causes a full collection
9991 to occur at every opportunity.
9993 @item max-reload-search-insns
9994 The maximum number of instruction reload should look backward for equivalent
9995 register. Increasing values mean more aggressive optimization, making the
9996 compilation time increase with probably slightly better performance.
9997 The default value is 100.
9999 @item max-cselib-memory-locations
10000 The maximum number of memory locations cselib should take into account.
10001 Increasing values mean more aggressive optimization, making the compilation time
10002 increase with probably slightly better performance. The default value is 500.
10004 @item max-sched-ready-insns
10005 The maximum number of instructions ready to be issued the scheduler should
10006 consider at any given time during the first scheduling pass. Increasing
10007 values mean more thorough searches, making the compilation time increase
10008 with probably little benefit. The default value is 100.
10010 @item max-sched-region-blocks
10011 The maximum number of blocks in a region to be considered for
10012 interblock scheduling. The default value is 10.
10014 @item max-pipeline-region-blocks
10015 The maximum number of blocks in a region to be considered for
10016 pipelining in the selective scheduler. The default value is 15.
10018 @item max-sched-region-insns
10019 The maximum number of insns in a region to be considered for
10020 interblock scheduling. The default value is 100.
10022 @item max-pipeline-region-insns
10023 The maximum number of insns in a region to be considered for
10024 pipelining in the selective scheduler. The default value is 200.
10026 @item min-spec-prob
10027 The minimum probability (in percents) of reaching a source block
10028 for interblock speculative scheduling. The default value is 40.
10030 @item max-sched-extend-regions-iters
10031 The maximum number of iterations through CFG to extend regions.
10032 A value of 0 (the default) disables region extensions.
10034 @item max-sched-insn-conflict-delay
10035 The maximum conflict delay for an insn to be considered for speculative motion.
10036 The default value is 3.
10038 @item sched-spec-prob-cutoff
10039 The minimal probability of speculation success (in percents), so that
10040 speculative insns are scheduled.
10041 The default value is 40.
10043 @item sched-state-edge-prob-cutoff
10044 The minimum probability an edge must have for the scheduler to save its
10046 The default value is 10.
10048 @item sched-mem-true-dep-cost
10049 Minimal distance (in CPU cycles) between store and load targeting same
10050 memory locations. The default value is 1.
10052 @item selsched-max-lookahead
10053 The maximum size of the lookahead window of selective scheduling. It is a
10054 depth of search for available instructions.
10055 The default value is 50.
10057 @item selsched-max-sched-times
10058 The maximum number of times that an instruction is scheduled during
10059 selective scheduling. This is the limit on the number of iterations
10060 through which the instruction may be pipelined. The default value is 2.
10062 @item selsched-insns-to-rename
10063 The maximum number of best instructions in the ready list that are considered
10064 for renaming in the selective scheduler. The default value is 2.
10067 The minimum value of stage count that swing modulo scheduler
10068 generates. The default value is 2.
10070 @item max-last-value-rtl
10071 The maximum size measured as number of RTLs that can be recorded in an expression
10072 in combiner for a pseudo register as last known value of that register. The default
10075 @item max-combine-insns
10076 The maximum number of instructions the RTL combiner tries to combine.
10077 The default value is 2 at @option{-Og} and 4 otherwise.
10079 @item integer-share-limit
10080 Small integer constants can use a shared data structure, reducing the
10081 compiler's memory usage and increasing its speed. This sets the maximum
10082 value of a shared integer constant. The default value is 256.
10084 @item ssp-buffer-size
10085 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10086 protection when @option{-fstack-protection} is used.
10088 @item min-size-for-stack-sharing
10089 The minimum size of variables taking part in stack slot sharing when not
10090 optimizing. The default value is 32.
10092 @item max-jump-thread-duplication-stmts
10093 Maximum number of statements allowed in a block that needs to be
10094 duplicated when threading jumps.
10096 @item max-fields-for-field-sensitive
10097 Maximum number of fields in a structure treated in
10098 a field sensitive manner during pointer analysis. The default is zero
10099 for @option{-O0} and @option{-O1},
10100 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10102 @item prefetch-latency
10103 Estimate on average number of instructions that are executed before
10104 prefetch finishes. The distance prefetched ahead is proportional
10105 to this constant. Increasing this number may also lead to less
10106 streams being prefetched (see @option{simultaneous-prefetches}).
10108 @item simultaneous-prefetches
10109 Maximum number of prefetches that can run at the same time.
10111 @item l1-cache-line-size
10112 The size of cache line in L1 cache, in bytes.
10114 @item l1-cache-size
10115 The size of L1 cache, in kilobytes.
10117 @item l2-cache-size
10118 The size of L2 cache, in kilobytes.
10120 @item min-insn-to-prefetch-ratio
10121 The minimum ratio between the number of instructions and the
10122 number of prefetches to enable prefetching in a loop.
10124 @item prefetch-min-insn-to-mem-ratio
10125 The minimum ratio between the number of instructions and the
10126 number of memory references to enable prefetching in a loop.
10128 @item use-canonical-types
10129 Whether the compiler should use the ``canonical'' type system. By
10130 default, this should always be 1, which uses a more efficient internal
10131 mechanism for comparing types in C++ and Objective-C++. However, if
10132 bugs in the canonical type system are causing compilation failures,
10133 set this value to 0 to disable canonical types.
10135 @item switch-conversion-max-branch-ratio
10136 Switch initialization conversion refuses to create arrays that are
10137 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10138 branches in the switch.
10140 @item max-partial-antic-length
10141 Maximum length of the partial antic set computed during the tree
10142 partial redundancy elimination optimization (@option{-ftree-pre}) when
10143 optimizing at @option{-O3} and above. For some sorts of source code
10144 the enhanced partial redundancy elimination optimization can run away,
10145 consuming all of the memory available on the host machine. This
10146 parameter sets a limit on the length of the sets that are computed,
10147 which prevents the runaway behavior. Setting a value of 0 for
10148 this parameter allows an unlimited set length.
10150 @item sccvn-max-scc-size
10151 Maximum size of a strongly connected component (SCC) during SCCVN
10152 processing. If this limit is hit, SCCVN processing for the whole
10153 function is not done and optimizations depending on it are
10154 disabled. The default maximum SCC size is 10000.
10156 @item sccvn-max-alias-queries-per-access
10157 Maximum number of alias-oracle queries we perform when looking for
10158 redundancies for loads and stores. If this limit is hit the search
10159 is aborted and the load or store is not considered redundant. The
10160 number of queries is algorithmically limited to the number of
10161 stores on all paths from the load to the function entry.
10162 The default maximum number of queries is 1000.
10164 @item ira-max-loops-num
10165 IRA uses regional register allocation by default. If a function
10166 contains more loops than the number given by this parameter, only at most
10167 the given number of the most frequently-executed loops form regions
10168 for regional register allocation. The default value of the
10171 @item ira-max-conflict-table-size
10172 Although IRA uses a sophisticated algorithm to compress the conflict
10173 table, the table can still require excessive amounts of memory for
10174 huge functions. If the conflict table for a function could be more
10175 than the size in MB given by this parameter, the register allocator
10176 instead uses a faster, simpler, and lower-quality
10177 algorithm that does not require building a pseudo-register conflict table.
10178 The default value of the parameter is 2000.
10180 @item ira-loop-reserved-regs
10181 IRA can be used to evaluate more accurate register pressure in loops
10182 for decisions to move loop invariants (see @option{-O3}). The number
10183 of available registers reserved for some other purposes is given
10184 by this parameter. The default value of the parameter is 2, which is
10185 the minimal number of registers needed by typical instructions.
10186 This value is the best found from numerous experiments.
10188 @item lra-inheritance-ebb-probability-cutoff
10189 LRA tries to reuse values reloaded in registers in subsequent insns.
10190 This optimization is called inheritance. EBB is used as a region to
10191 do this optimization. The parameter defines a minimal fall-through
10192 edge probability in percentage used to add BB to inheritance EBB in
10193 LRA. The default value of the parameter is 40. The value was chosen
10194 from numerous runs of SPEC2000 on x86-64.
10196 @item loop-invariant-max-bbs-in-loop
10197 Loop invariant motion can be very expensive, both in compilation time and
10198 in amount of needed compile-time memory, with very large loops. Loops
10199 with more basic blocks than this parameter won't have loop invariant
10200 motion optimization performed on them. The default value of the
10201 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10203 @item loop-max-datarefs-for-datadeps
10204 Building data dependencies is expensive for very large loops. This
10205 parameter limits the number of data references in loops that are
10206 considered for data dependence analysis. These large loops are no
10207 handled by the optimizations using loop data dependencies.
10208 The default value is 1000.
10210 @item max-vartrack-size
10211 Sets a maximum number of hash table slots to use during variable
10212 tracking dataflow analysis of any function. If this limit is exceeded
10213 with variable tracking at assignments enabled, analysis for that
10214 function is retried without it, after removing all debug insns from
10215 the function. If the limit is exceeded even without debug insns, var
10216 tracking analysis is completely disabled for the function. Setting
10217 the parameter to zero makes it unlimited.
10219 @item max-vartrack-expr-depth
10220 Sets a maximum number of recursion levels when attempting to map
10221 variable names or debug temporaries to value expressions. This trades
10222 compilation time for more complete debug information. If this is set too
10223 low, value expressions that are available and could be represented in
10224 debug information may end up not being used; setting this higher may
10225 enable the compiler to find more complex debug expressions, but compile
10226 time and memory use may grow. The default is 12.
10228 @item min-nondebug-insn-uid
10229 Use uids starting at this parameter for nondebug insns. The range below
10230 the parameter is reserved exclusively for debug insns created by
10231 @option{-fvar-tracking-assignments}, but debug insns may get
10232 (non-overlapping) uids above it if the reserved range is exhausted.
10234 @item ipa-sra-ptr-growth-factor
10235 IPA-SRA replaces a pointer to an aggregate with one or more new
10236 parameters only when their cumulative size is less or equal to
10237 @option{ipa-sra-ptr-growth-factor} times the size of the original
10240 @item sra-max-scalarization-size-Ospeed
10241 @item sra-max-scalarization-size-Osize
10242 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10243 replace scalar parts of aggregates with uses of independent scalar
10244 variables. These parameters control the maximum size, in storage units,
10245 of aggregate which is considered for replacement when compiling for
10247 (@option{sra-max-scalarization-size-Ospeed}) or size
10248 (@option{sra-max-scalarization-size-Osize}) respectively.
10250 @item tm-max-aggregate-size
10251 When making copies of thread-local variables in a transaction, this
10252 parameter specifies the size in bytes after which variables are
10253 saved with the logging functions as opposed to save/restore code
10254 sequence pairs. This option only applies when using
10257 @item graphite-max-nb-scop-params
10258 To avoid exponential effects in the Graphite loop transforms, the
10259 number of parameters in a Static Control Part (SCoP) is bounded. The
10260 default value is 10 parameters. A variable whose value is unknown at
10261 compilation time and defined outside a SCoP is a parameter of the SCoP.
10263 @item graphite-max-bbs-per-function
10264 To avoid exponential effects in the detection of SCoPs, the size of
10265 the functions analyzed by Graphite is bounded. The default value is
10268 @item loop-block-tile-size
10269 Loop blocking or strip mining transforms, enabled with
10270 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10271 loop in the loop nest by a given number of iterations. The strip
10272 length can be changed using the @option{loop-block-tile-size}
10273 parameter. The default value is 51 iterations.
10275 @item loop-unroll-jam-size
10276 Specify the unroll factor for the @option{-floop-unroll-and-jam} option. The
10277 default value is 4.
10279 @item loop-unroll-jam-depth
10280 Specify the dimension to be unrolled (counting from the most inner loop)
10281 for the @option{-floop-unroll-and-jam}. The default value is 2.
10283 @item ipa-cp-value-list-size
10284 IPA-CP attempts to track all possible values and types passed to a function's
10285 parameter in order to propagate them and perform devirtualization.
10286 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10287 stores per one formal parameter of a function.
10289 @item ipa-cp-eval-threshold
10290 IPA-CP calculates its own score of cloning profitability heuristics
10291 and performs those cloning opportunities with scores that exceed
10292 @option{ipa-cp-eval-threshold}.
10294 @item ipa-cp-recursion-penalty
10295 Percentage penalty the recursive functions will receive when they
10296 are evaluated for cloning.
10298 @item ipa-cp-single-call-penalty
10299 Percentage penalty functions containg a single call to another
10300 function will receive when they are evaluated for cloning.
10303 @item ipa-max-agg-items
10304 IPA-CP is also capable to propagate a number of scalar values passed
10305 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10306 number of such values per one parameter.
10308 @item ipa-cp-loop-hint-bonus
10309 When IPA-CP determines that a cloning candidate would make the number
10310 of iterations of a loop known, it adds a bonus of
10311 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10314 @item ipa-cp-array-index-hint-bonus
10315 When IPA-CP determines that a cloning candidate would make the index of
10316 an array access known, it adds a bonus of
10317 @option{ipa-cp-array-index-hint-bonus} to the profitability
10318 score of the candidate.
10320 @item ipa-max-aa-steps
10321 During its analysis of function bodies, IPA-CP employs alias analysis
10322 in order to track values pointed to by function parameters. In order
10323 not spend too much time analyzing huge functions, it gives up and
10324 consider all memory clobbered after examining
10325 @option{ipa-max-aa-steps} statements modifying memory.
10327 @item lto-partitions
10328 Specify desired number of partitions produced during WHOPR compilation.
10329 The number of partitions should exceed the number of CPUs used for compilation.
10330 The default value is 32.
10332 @item lto-min-partition
10333 Size of minimal partition for WHOPR (in estimated instructions).
10334 This prevents expenses of splitting very small programs into too many
10337 @item lto-max-partition
10338 Size of max partition for WHOPR (in estimated instructions).
10339 to provide an upper bound for individual size of partition.
10340 Meant to be used only with balanced partitioning.
10342 @item cxx-max-namespaces-for-diagnostic-help
10343 The maximum number of namespaces to consult for suggestions when C++
10344 name lookup fails for an identifier. The default is 1000.
10346 @item sink-frequency-threshold
10347 The maximum relative execution frequency (in percents) of the target block
10348 relative to a statement's original block to allow statement sinking of a
10349 statement. Larger numbers result in more aggressive statement sinking.
10350 The default value is 75. A small positive adjustment is applied for
10351 statements with memory operands as those are even more profitable so sink.
10353 @item max-stores-to-sink
10354 The maximum number of conditional store pairs that can be sunk. Set to 0
10355 if either vectorization (@option{-ftree-vectorize}) or if-conversion
10356 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
10358 @item allow-store-data-races
10359 Allow optimizers to introduce new data races on stores.
10360 Set to 1 to allow, otherwise to 0. This option is enabled by default
10361 at optimization level @option{-Ofast}.
10363 @item case-values-threshold
10364 The smallest number of different values for which it is best to use a
10365 jump-table instead of a tree of conditional branches. If the value is
10366 0, use the default for the machine. The default is 0.
10368 @item tree-reassoc-width
10369 Set the maximum number of instructions executed in parallel in
10370 reassociated tree. This parameter overrides target dependent
10371 heuristics used by default if has non zero value.
10373 @item sched-pressure-algorithm
10374 Choose between the two available implementations of
10375 @option{-fsched-pressure}. Algorithm 1 is the original implementation
10376 and is the more likely to prevent instructions from being reordered.
10377 Algorithm 2 was designed to be a compromise between the relatively
10378 conservative approach taken by algorithm 1 and the rather aggressive
10379 approach taken by the default scheduler. It relies more heavily on
10380 having a regular register file and accurate register pressure classes.
10381 See @file{haifa-sched.c} in the GCC sources for more details.
10383 The default choice depends on the target.
10385 @item max-slsr-cand-scan
10386 Set the maximum number of existing candidates that are considered when
10387 seeking a basis for a new straight-line strength reduction candidate.
10390 Enable buffer overflow detection for global objects. This kind
10391 of protection is enabled by default if you are using
10392 @option{-fsanitize=address} option.
10393 To disable global objects protection use @option{--param asan-globals=0}.
10396 Enable buffer overflow detection for stack objects. This kind of
10397 protection is enabled by default when using @option{-fsanitize=address}.
10398 To disable stack protection use @option{--param asan-stack=0} option.
10400 @item asan-instrument-reads
10401 Enable buffer overflow detection for memory reads. This kind of
10402 protection is enabled by default when using @option{-fsanitize=address}.
10403 To disable memory reads protection use
10404 @option{--param asan-instrument-reads=0}.
10406 @item asan-instrument-writes
10407 Enable buffer overflow detection for memory writes. This kind of
10408 protection is enabled by default when using @option{-fsanitize=address}.
10409 To disable memory writes protection use
10410 @option{--param asan-instrument-writes=0} option.
10412 @item asan-memintrin
10413 Enable detection for built-in functions. This kind of protection
10414 is enabled by default when using @option{-fsanitize=address}.
10415 To disable built-in functions protection use
10416 @option{--param asan-memintrin=0}.
10418 @item asan-use-after-return
10419 Enable detection of use-after-return. This kind of protection
10420 is enabled by default when using @option{-fsanitize=address} option.
10421 To disable use-after-return detection use
10422 @option{--param asan-use-after-return=0}.
10424 Note: The check is disabled by default at runtime. To enable the check,
10425 you should set environment variable @env{ASAN_OPTIONS} to
10426 @code{detect_stack_use_after_return=1}.
10428 @item asan-instrumentation-with-call-threshold
10429 If number of memory accesses in function being instrumented
10430 is greater or equal to this number, use callbacks instead of inline checks.
10431 E.g. to disable inline code use
10432 @option{--param asan-instrumentation-with-call-threshold=0}.
10434 @item use-after-scope-direct-emission-threshold
10435 If size of a local variables in bytes is smaller of equal to this number,
10436 direct instruction emission is utilized to poison and unpoison local variables.
10438 @item chkp-max-ctor-size
10439 Static constructors generated by Pointer Bounds Checker may become very
10440 large and significantly increase compile time at optimization level
10441 @option{-O1} and higher. This parameter is a maximum nubmer of statements
10442 in a single generated constructor. Default value is 5000.
10444 @item max-fsm-thread-path-insns
10445 Maximum number of instructions to copy when duplicating blocks on a
10446 finite state automaton jump thread path. The default is 100.
10448 @item max-fsm-thread-length
10449 Maximum number of basic blocks on a finite state automaton jump thread
10450 path. The default is 10.
10452 @item max-fsm-thread-paths
10453 Maximum number of new jump thread paths to create for a finite state
10454 automaton. The default is 50.
10456 @item parloops-chunk-size
10457 Chunk size of omp schedule for loops parallelized by parloops. The default
10460 @item parloops-schedule
10461 Schedule type of omp schedule for loops parallelized by parloops (static,
10462 dynamic, guided, auto, runtime). The default is static.
10464 @item max-ssa-name-query-depth
10465 Maximum depth of recursion when querying properties of SSA names in things
10466 like fold routines. One level of recursion corresponds to following a
10469 @item hsa-gen-debug-stores
10470 Enable emission of special debug stores within HSA kernels which are
10471 then read and reported by libgomp plugin. Generation of these stores
10472 is disabled by default, use @option{--param hsa-gen-debug-stores=1} to
10475 @item max-speculative-devirt-maydefs
10476 The maximum number of may-defs we analyze when looking for a must-def
10477 specifying the dynamic type of an object that invokes a virtual call
10478 we may be able to devirtualize speculatively.
10480 @item max-vrp-switch-assertions
10481 The maximum number of assertions to add along the default edge of a switch
10482 statement during VRP. The default is 10.
10486 @node Instrumentation Options
10487 @section Program Instrumentation Options
10488 @cindex instrumentation options
10489 @cindex program instrumentation options
10490 @cindex run-time error checking options
10491 @cindex profiling options
10492 @cindex options, program instrumentation
10493 @cindex options, run-time error checking
10494 @cindex options, profiling
10496 GCC supports a number of command-line options that control adding
10497 run-time instrumentation to the code it normally generates.
10498 For example, one purpose of instrumentation is collect profiling
10499 statistics for use in finding program hot spots, code coverage
10500 analysis, or profile-guided optimizations.
10501 Another class of program instrumentation is adding run-time checking
10502 to detect programming errors like invalid pointer
10503 dereferences or out-of-bounds array accesses, as well as deliberately
10504 hostile attacks such as stack smashing or C++ vtable hijacking.
10505 There is also a general hook which can be used to implement other
10506 forms of tracing or function-level instrumentation for debug or
10507 program analysis purposes.
10510 @cindex @command{prof}
10513 Generate extra code to write profile information suitable for the
10514 analysis program @command{prof}. You must use this option when compiling
10515 the source files you want data about, and you must also use it when
10518 @cindex @command{gprof}
10521 Generate extra code to write profile information suitable for the
10522 analysis program @command{gprof}. You must use this option when compiling
10523 the source files you want data about, and you must also use it when
10526 @item -fprofile-arcs
10527 @opindex fprofile-arcs
10528 Add code so that program flow @dfn{arcs} are instrumented. During
10529 execution the program records how many times each branch and call is
10530 executed and how many times it is taken or returns. When the compiled
10531 program exits it saves this data to a file called
10532 @file{@var{auxname}.gcda} for each source file. The data may be used for
10533 profile-directed optimizations (@option{-fbranch-probabilities}), or for
10534 test coverage analysis (@option{-ftest-coverage}). Each object file's
10535 @var{auxname} is generated from the name of the output file, if
10536 explicitly specified and it is not the final executable, otherwise it is
10537 the basename of the source file. In both cases any suffix is removed
10538 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
10539 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
10540 @xref{Cross-profiling}.
10542 @cindex @command{gcov}
10546 This option is used to compile and link code instrumented for coverage
10547 analysis. The option is a synonym for @option{-fprofile-arcs}
10548 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
10549 linking). See the documentation for those options for more details.
10554 Compile the source files with @option{-fprofile-arcs} plus optimization
10555 and code generation options. For test coverage analysis, use the
10556 additional @option{-ftest-coverage} option. You do not need to profile
10557 every source file in a program.
10560 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
10561 (the latter implies the former).
10564 Run the program on a representative workload to generate the arc profile
10565 information. This may be repeated any number of times. You can run
10566 concurrent instances of your program, and provided that the file system
10567 supports locking, the data files will be correctly updated. Also
10568 @code{fork} calls are detected and correctly handled (double counting
10572 For profile-directed optimizations, compile the source files again with
10573 the same optimization and code generation options plus
10574 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
10575 Control Optimization}).
10578 For test coverage analysis, use @command{gcov} to produce human readable
10579 information from the @file{.gcno} and @file{.gcda} files. Refer to the
10580 @command{gcov} documentation for further information.
10584 With @option{-fprofile-arcs}, for each function of your program GCC
10585 creates a program flow graph, then finds a spanning tree for the graph.
10586 Only arcs that are not on the spanning tree have to be instrumented: the
10587 compiler adds code to count the number of times that these arcs are
10588 executed. When an arc is the only exit or only entrance to a block, the
10589 instrumentation code can be added to the block; otherwise, a new basic
10590 block must be created to hold the instrumentation code.
10593 @item -ftest-coverage
10594 @opindex ftest-coverage
10595 Produce a notes file that the @command{gcov} code-coverage utility
10596 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
10597 show program coverage. Each source file's note file is called
10598 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
10599 above for a description of @var{auxname} and instructions on how to
10600 generate test coverage data. Coverage data matches the source files
10601 more closely if you do not optimize.
10603 @item -fprofile-dir=@var{path}
10604 @opindex fprofile-dir
10606 Set the directory to search for the profile data files in to @var{path}.
10607 This option affects only the profile data generated by
10608 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
10609 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
10610 and its related options. Both absolute and relative paths can be used.
10611 By default, GCC uses the current directory as @var{path}, thus the
10612 profile data file appears in the same directory as the object file.
10614 @item -fprofile-generate
10615 @itemx -fprofile-generate=@var{path}
10616 @opindex fprofile-generate
10618 Enable options usually used for instrumenting application to produce
10619 profile useful for later recompilation with profile feedback based
10620 optimization. You must use @option{-fprofile-generate} both when
10621 compiling and when linking your program.
10623 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
10625 If @var{path} is specified, GCC looks at the @var{path} to find
10626 the profile feedback data files. See @option{-fprofile-dir}.
10628 To optimize the program based on the collected profile information, use
10629 @option{-fprofile-use}. @xref{Optimize Options}, for more information.
10631 @item -fprofile-update=@var{method}
10632 @opindex fprofile-update
10634 Alter the update method for an application instrumented for profile
10635 feedback based optimization. The @var{method} argument should be one of
10636 @samp{single}, @samp{atomic} or @samp{prefer-atomic}.
10637 The first one is useful for single-threaded applications,
10638 while the second one prevents profile corruption by emitting thread-safe code.
10640 @strong{Warning:} When an application does not properly join all threads
10641 (or creates an detached thread), a profile file can be still corrupted.
10643 Using @samp{prefer-atomic} would be transformed either to @samp{atomic},
10644 when supported by a target, or to @samp{single} otherwise. The GCC driver
10645 automatically selects @samp{prefer-atomic} when @option{-pthread}
10646 is present in the command line.
10648 @item -fsanitize=address
10649 @opindex fsanitize=address
10650 Enable AddressSanitizer, a fast memory error detector.
10651 Memory access instructions are instrumented to detect
10652 out-of-bounds and use-after-free bugs.
10653 The option enables @option{-fsanitize-address-use-after-scope}.
10654 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
10655 more details. The run-time behavior can be influenced using the
10656 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
10657 the available options are shown at startup of the instrumented program. See
10658 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
10659 for a list of supported options.
10660 The option can't be combined with @option{-fsanitize=thread}.
10662 @item -fsanitize=kernel-address
10663 @opindex fsanitize=kernel-address
10664 Enable AddressSanitizer for Linux kernel.
10665 See @uref{https://github.com/google/kasan/wiki} for more details.
10667 @item -fsanitize=thread
10668 @opindex fsanitize=thread
10669 Enable ThreadSanitizer, a fast data race detector.
10670 Memory access instructions are instrumented to detect
10671 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
10672 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
10673 environment variable; see
10674 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
10676 The option can't be combined with @option{-fsanitize=address}
10677 and/or @option{-fsanitize=leak}.
10679 @item -fsanitize=leak
10680 @opindex fsanitize=leak
10681 Enable LeakSanitizer, a memory leak detector.
10682 This option only matters for linking of executables and
10683 the executable is linked against a library that overrides @code{malloc}
10684 and other allocator functions. See
10685 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
10686 details. The run-time behavior can be influenced using the
10687 @env{LSAN_OPTIONS} environment variable.
10688 The option can't be combined with @option{-fsanitize=thread}.
10690 @item -fsanitize=undefined
10691 @opindex fsanitize=undefined
10692 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
10693 Various computations are instrumented to detect undefined behavior
10694 at runtime. Current suboptions are:
10698 @item -fsanitize=shift
10699 @opindex fsanitize=shift
10700 This option enables checking that the result of a shift operation is
10701 not undefined. Note that what exactly is considered undefined differs
10702 slightly between C and C++, as well as between ISO C90 and C99, etc.
10703 This option has two suboptions, @option{-fsanitize=shift-base} and
10704 @option{-fsanitize=shift-exponent}.
10706 @item -fsanitize=shift-exponent
10707 @opindex fsanitize=shift-exponent
10708 This option enables checking that the second argument of a shift operation
10709 is not negative and is smaller than the precision of the promoted first
10712 @item -fsanitize=shift-base
10713 @opindex fsanitize=shift-base
10714 If the second argument of a shift operation is within range, check that the
10715 result of a shift operation is not undefined. Note that what exactly is
10716 considered undefined differs slightly between C and C++, as well as between
10717 ISO C90 and C99, etc.
10719 @item -fsanitize=integer-divide-by-zero
10720 @opindex fsanitize=integer-divide-by-zero
10721 Detect integer division by zero as well as @code{INT_MIN / -1} division.
10723 @item -fsanitize=unreachable
10724 @opindex fsanitize=unreachable
10725 With this option, the compiler turns the @code{__builtin_unreachable}
10726 call into a diagnostics message call instead. When reaching the
10727 @code{__builtin_unreachable} call, the behavior is undefined.
10729 @item -fsanitize=vla-bound
10730 @opindex fsanitize=vla-bound
10731 This option instructs the compiler to check that the size of a variable
10732 length array is positive.
10734 @item -fsanitize=null
10735 @opindex fsanitize=null
10736 This option enables pointer checking. Particularly, the application
10737 built with this option turned on will issue an error message when it
10738 tries to dereference a NULL pointer, or if a reference (possibly an
10739 rvalue reference) is bound to a NULL pointer, or if a method is invoked
10740 on an object pointed by a NULL pointer.
10742 @item -fsanitize=return
10743 @opindex fsanitize=return
10744 This option enables return statement checking. Programs
10745 built with this option turned on will issue an error message
10746 when the end of a non-void function is reached without actually
10747 returning a value. This option works in C++ only.
10749 @item -fsanitize=signed-integer-overflow
10750 @opindex fsanitize=signed-integer-overflow
10751 This option enables signed integer overflow checking. We check that
10752 the result of @code{+}, @code{*}, and both unary and binary @code{-}
10753 does not overflow in the signed arithmetics. Note, integer promotion
10754 rules must be taken into account. That is, the following is not an
10757 signed char a = SCHAR_MAX;
10761 @item -fsanitize=bounds
10762 @opindex fsanitize=bounds
10763 This option enables instrumentation of array bounds. Various out of bounds
10764 accesses are detected. Flexible array members, flexible array member-like
10765 arrays, and initializers of variables with static storage are not instrumented.
10767 @item -fsanitize=bounds-strict
10768 @opindex fsanitize=bounds-strict
10769 This option enables strict instrumentation of array bounds. Most out of bounds
10770 accesses are detected, including flexible array members and flexible array
10771 member-like arrays. Initializers of variables with static storage are not
10774 @item -fsanitize=alignment
10775 @opindex fsanitize=alignment
10777 This option enables checking of alignment of pointers when they are
10778 dereferenced, or when a reference is bound to insufficiently aligned target,
10779 or when a method or constructor is invoked on insufficiently aligned object.
10781 @item -fsanitize=object-size
10782 @opindex fsanitize=object-size
10783 This option enables instrumentation of memory references using the
10784 @code{__builtin_object_size} function. Various out of bounds pointer
10785 accesses are detected.
10787 @item -fsanitize=float-divide-by-zero
10788 @opindex fsanitize=float-divide-by-zero
10789 Detect floating-point division by zero. Unlike other similar options,
10790 @option{-fsanitize=float-divide-by-zero} is not enabled by
10791 @option{-fsanitize=undefined}, since floating-point division by zero can
10792 be a legitimate way of obtaining infinities and NaNs.
10794 @item -fsanitize=float-cast-overflow
10795 @opindex fsanitize=float-cast-overflow
10796 This option enables floating-point type to integer conversion checking.
10797 We check that the result of the conversion does not overflow.
10798 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
10799 not enabled by @option{-fsanitize=undefined}.
10800 This option does not work well with @code{FE_INVALID} exceptions enabled.
10802 @item -fsanitize=nonnull-attribute
10803 @opindex fsanitize=nonnull-attribute
10805 This option enables instrumentation of calls, checking whether null values
10806 are not passed to arguments marked as requiring a non-null value by the
10807 @code{nonnull} function attribute.
10809 @item -fsanitize=returns-nonnull-attribute
10810 @opindex fsanitize=returns-nonnull-attribute
10812 This option enables instrumentation of return statements in functions
10813 marked with @code{returns_nonnull} function attribute, to detect returning
10814 of null values from such functions.
10816 @item -fsanitize=bool
10817 @opindex fsanitize=bool
10819 This option enables instrumentation of loads from bool. If a value other
10820 than 0/1 is loaded, a run-time error is issued.
10822 @item -fsanitize=enum
10823 @opindex fsanitize=enum
10825 This option enables instrumentation of loads from an enum type. If
10826 a value outside the range of values for the enum type is loaded,
10827 a run-time error is issued.
10829 @item -fsanitize=vptr
10830 @opindex fsanitize=vptr
10832 This option enables instrumentation of C++ member function calls, member
10833 accesses and some conversions between pointers to base and derived classes,
10834 to verify the referenced object has the correct dynamic type.
10838 While @option{-ftrapv} causes traps for signed overflows to be emitted,
10839 @option{-fsanitize=undefined} gives a diagnostic message.
10840 This currently works only for the C family of languages.
10842 @item -fno-sanitize=all
10843 @opindex fno-sanitize=all
10845 This option disables all previously enabled sanitizers.
10846 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
10849 @item -fasan-shadow-offset=@var{number}
10850 @opindex fasan-shadow-offset
10851 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
10852 It is useful for experimenting with different shadow memory layouts in
10853 Kernel AddressSanitizer.
10855 @item -fsanitize-sections=@var{s1},@var{s2},...
10856 @opindex fsanitize-sections
10857 Sanitize global variables in selected user-defined sections. @var{si} may
10860 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
10861 @opindex fsanitize-recover
10862 @opindex fno-sanitize-recover
10863 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
10864 mentioned in comma-separated list of @var{opts}. Enabling this option
10865 for a sanitizer component causes it to attempt to continue
10866 running the program as if no error happened. This means multiple
10867 runtime errors can be reported in a single program run, and the exit
10868 code of the program may indicate success even when errors
10869 have been reported. The @option{-fno-sanitize-recover=} option
10870 can be used to alter
10871 this behavior: only the first detected error is reported
10872 and program then exits with a non-zero exit code.
10874 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
10875 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
10876 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
10877 @option{-fsanitize=bounds-strict},
10878 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
10879 For these sanitizers error recovery is turned on by default,
10880 except @option{-fsanitize=address}, for which this feature is experimental.
10881 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
10882 accepted, the former enables recovery for all sanitizers that support it,
10883 the latter disables recovery for all sanitizers that support it.
10885 Even if a recovery mode is turned on the compiler side, it needs to be also
10886 enabled on the runtime library side, otherwise the failures are still fatal.
10887 The runtime library defaults to @code{halt_on_error=0} for
10888 ThreadSanitizer and UndefinedBehaviorSanitizer, while default value for
10889 AddressSanitizer is @code{halt_on_error=1}. This can be overridden through
10890 setting the @code{halt_on_error} flag in the corresponding environment variable.
10892 Syntax without explicit @var{opts} parameter is deprecated. It is equivalent to
10894 -fsanitize-recover=undefined,float-cast-overflow,float-divide-by-zero,bounds-strict
10897 Similarly @option{-fno-sanitize-recover} is equivalent to
10899 -fno-sanitize-recover=undefined,float-cast-overflow,float-divide-by-zero,bounds-strict
10902 @item -fsanitize-address-use-after-scope
10903 @opindex fsanitize-address-use-after-scope
10904 Enable sanitization of local variables to detect use-after-scope bugs.
10905 The option sets @option{-fstack-reuse} to @samp{none}.
10907 @item -fsanitize-undefined-trap-on-error
10908 @opindex fsanitize-undefined-trap-on-error
10909 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
10910 report undefined behavior using @code{__builtin_trap} rather than
10911 a @code{libubsan} library routine. The advantage of this is that the
10912 @code{libubsan} library is not needed and is not linked in, so this
10913 is usable even in freestanding environments.
10915 @item -fsanitize-coverage=trace-pc
10916 @opindex fsanitize-coverage=trace-pc
10917 Enable coverage-guided fuzzing code instrumentation.
10918 Inserts a call to @code{__sanitizer_cov_trace_pc} into every basic block.
10920 @item -fbounds-check
10921 @opindex fbounds-check
10922 For front ends that support it, generate additional code to check that
10923 indices used to access arrays are within the declared range. This is
10924 currently only supported by the Fortran front end, where this option
10927 @item -fcheck-pointer-bounds
10928 @opindex fcheck-pointer-bounds
10929 @opindex fno-check-pointer-bounds
10930 @cindex Pointer Bounds Checker options
10931 Enable Pointer Bounds Checker instrumentation. Each memory reference
10932 is instrumented with checks of the pointer used for memory access against
10933 bounds associated with that pointer.
10936 is only an implementation for Intel MPX available, thus x86 GNU/Linux target
10937 and @option{-mmpx} are required to enable this feature.
10938 MPX-based instrumentation requires
10939 a runtime library to enable MPX in hardware and handle bounds
10940 violation signals. By default when @option{-fcheck-pointer-bounds}
10941 and @option{-mmpx} options are used to link a program, the GCC driver
10942 links against the @file{libmpx} and @file{libmpxwrappers} libraries.
10943 Bounds checking on calls to dynamic libraries requires a linker
10944 with @option{-z bndplt} support; if GCC was configured with a linker
10945 without support for this option (including the Gold linker and older
10946 versions of ld), a warning is given if you link with @option{-mmpx}
10947 without also specifying @option{-static}, since the overall effectiveness
10948 of the bounds checking protection is reduced.
10949 See also @option{-static-libmpxwrappers}.
10951 MPX-based instrumentation
10952 may be used for debugging and also may be included in production code
10953 to increase program security. Depending on usage, you may
10954 have different requirements for the runtime library. The current version
10955 of the MPX runtime library is more oriented for use as a debugging
10956 tool. MPX runtime library usage implies @option{-lpthread}. See
10957 also @option{-static-libmpx}. The runtime library behavior can be
10958 influenced using various @env{CHKP_RT_*} environment variables. See
10959 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
10962 Generated instrumentation may be controlled by various
10963 @option{-fchkp-*} options and by the @code{bnd_variable_size}
10964 structure field attribute (@pxref{Type Attributes}) and
10965 @code{bnd_legacy}, and @code{bnd_instrument} function attributes
10966 (@pxref{Function Attributes}). GCC also provides a number of built-in
10967 functions for controlling the Pointer Bounds Checker. @xref{Pointer
10968 Bounds Checker builtins}, for more information.
10970 @item -fchkp-check-incomplete-type
10971 @opindex fchkp-check-incomplete-type
10972 @opindex fno-chkp-check-incomplete-type
10973 Generate pointer bounds checks for variables with incomplete type.
10974 Enabled by default.
10976 @item -fchkp-narrow-bounds
10977 @opindex fchkp-narrow-bounds
10978 @opindex fno-chkp-narrow-bounds
10979 Controls bounds used by Pointer Bounds Checker for pointers to object
10980 fields. If narrowing is enabled then field bounds are used. Otherwise
10981 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
10982 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
10984 @item -fchkp-first-field-has-own-bounds
10985 @opindex fchkp-first-field-has-own-bounds
10986 @opindex fno-chkp-first-field-has-own-bounds
10987 Forces Pointer Bounds Checker to use narrowed bounds for the address of the
10988 first field in the structure. By default a pointer to the first field has
10989 the same bounds as a pointer to the whole structure.
10991 @item -fchkp-narrow-to-innermost-array
10992 @opindex fchkp-narrow-to-innermost-array
10993 @opindex fno-chkp-narrow-to-innermost-array
10994 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
10995 case of nested static array access. By default this option is disabled and
10996 bounds of the outermost array are used.
10998 @item -fchkp-optimize
10999 @opindex fchkp-optimize
11000 @opindex fno-chkp-optimize
11001 Enables Pointer Bounds Checker optimizations. Enabled by default at
11002 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
11004 @item -fchkp-use-fast-string-functions
11005 @opindex fchkp-use-fast-string-functions
11006 @opindex fno-chkp-use-fast-string-functions
11007 Enables use of @code{*_nobnd} versions of string functions (not copying bounds)
11008 by Pointer Bounds Checker. Disabled by default.
11010 @item -fchkp-use-nochk-string-functions
11011 @opindex fchkp-use-nochk-string-functions
11012 @opindex fno-chkp-use-nochk-string-functions
11013 Enables use of @code{*_nochk} versions of string functions (not checking bounds)
11014 by Pointer Bounds Checker. Disabled by default.
11016 @item -fchkp-use-static-bounds
11017 @opindex fchkp-use-static-bounds
11018 @opindex fno-chkp-use-static-bounds
11019 Allow Pointer Bounds Checker to generate static bounds holding
11020 bounds of static variables. Enabled by default.
11022 @item -fchkp-use-static-const-bounds
11023 @opindex fchkp-use-static-const-bounds
11024 @opindex fno-chkp-use-static-const-bounds
11025 Use statically-initialized bounds for constant bounds instead of
11026 generating them each time they are required. By default enabled when
11027 @option{-fchkp-use-static-bounds} is enabled.
11029 @item -fchkp-treat-zero-dynamic-size-as-infinite
11030 @opindex fchkp-treat-zero-dynamic-size-as-infinite
11031 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
11032 With this option, objects with incomplete type whose
11033 dynamically-obtained size is zero are treated as having infinite size
11034 instead by Pointer Bounds
11035 Checker. This option may be helpful if a program is linked with a library
11036 missing size information for some symbols. Disabled by default.
11038 @item -fchkp-check-read
11039 @opindex fchkp-check-read
11040 @opindex fno-chkp-check-read
11041 Instructs Pointer Bounds Checker to generate checks for all read
11042 accesses to memory. Enabled by default.
11044 @item -fchkp-check-write
11045 @opindex fchkp-check-write
11046 @opindex fno-chkp-check-write
11047 Instructs Pointer Bounds Checker to generate checks for all write
11048 accesses to memory. Enabled by default.
11050 @item -fchkp-store-bounds
11051 @opindex fchkp-store-bounds
11052 @opindex fno-chkp-store-bounds
11053 Instructs Pointer Bounds Checker to generate bounds stores for
11054 pointer writes. Enabled by default.
11056 @item -fchkp-instrument-calls
11057 @opindex fchkp-instrument-calls
11058 @opindex fno-chkp-instrument-calls
11059 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
11060 Enabled by default.
11062 @item -fchkp-instrument-marked-only
11063 @opindex fchkp-instrument-marked-only
11064 @opindex fno-chkp-instrument-marked-only
11065 Instructs Pointer Bounds Checker to instrument only functions
11066 marked with the @code{bnd_instrument} attribute
11067 (@pxref{Function Attributes}). Disabled by default.
11069 @item -fchkp-use-wrappers
11070 @opindex fchkp-use-wrappers
11071 @opindex fno-chkp-use-wrappers
11072 Allows Pointer Bounds Checker to replace calls to built-in functions
11073 with calls to wrapper functions. When @option{-fchkp-use-wrappers}
11074 is used to link a program, the GCC driver automatically links
11075 against @file{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
11076 Enabled by default.
11078 @item -fstack-protector
11079 @opindex fstack-protector
11080 Emit extra code to check for buffer overflows, such as stack smashing
11081 attacks. This is done by adding a guard variable to functions with
11082 vulnerable objects. This includes functions that call @code{alloca}, and
11083 functions with buffers larger than 8 bytes. The guards are initialized
11084 when a function is entered and then checked when the function exits.
11085 If a guard check fails, an error message is printed and the program exits.
11087 @item -fstack-protector-all
11088 @opindex fstack-protector-all
11089 Like @option{-fstack-protector} except that all functions are protected.
11091 @item -fstack-protector-strong
11092 @opindex fstack-protector-strong
11093 Like @option{-fstack-protector} but includes additional functions to
11094 be protected --- those that have local array definitions, or have
11095 references to local frame addresses.
11097 @item -fstack-protector-explicit
11098 @opindex fstack-protector-explicit
11099 Like @option{-fstack-protector} but only protects those functions which
11100 have the @code{stack_protect} attribute.
11102 @item -fstack-check
11103 @opindex fstack-check
11104 Generate code to verify that you do not go beyond the boundary of the
11105 stack. You should specify this flag if you are running in an
11106 environment with multiple threads, but you only rarely need to specify it in
11107 a single-threaded environment since stack overflow is automatically
11108 detected on nearly all systems if there is only one stack.
11110 Note that this switch does not actually cause checking to be done; the
11111 operating system or the language runtime must do that. The switch causes
11112 generation of code to ensure that they see the stack being extended.
11114 You can additionally specify a string parameter: @samp{no} means no
11115 checking, @samp{generic} means force the use of old-style checking,
11116 @samp{specific} means use the best checking method and is equivalent
11117 to bare @option{-fstack-check}.
11119 Old-style checking is a generic mechanism that requires no specific
11120 target support in the compiler but comes with the following drawbacks:
11124 Modified allocation strategy for large objects: they are always
11125 allocated dynamically if their size exceeds a fixed threshold.
11128 Fixed limit on the size of the static frame of functions: when it is
11129 topped by a particular function, stack checking is not reliable and
11130 a warning is issued by the compiler.
11133 Inefficiency: because of both the modified allocation strategy and the
11134 generic implementation, code performance is hampered.
11137 Note that old-style stack checking is also the fallback method for
11138 @samp{specific} if no target support has been added in the compiler.
11140 @item -fstack-limit-register=@var{reg}
11141 @itemx -fstack-limit-symbol=@var{sym}
11142 @itemx -fno-stack-limit
11143 @opindex fstack-limit-register
11144 @opindex fstack-limit-symbol
11145 @opindex fno-stack-limit
11146 Generate code to ensure that the stack does not grow beyond a certain value,
11147 either the value of a register or the address of a symbol. If a larger
11148 stack is required, a signal is raised at run time. For most targets,
11149 the signal is raised before the stack overruns the boundary, so
11150 it is possible to catch the signal without taking special precautions.
11152 For instance, if the stack starts at absolute address @samp{0x80000000}
11153 and grows downwards, you can use the flags
11154 @option{-fstack-limit-symbol=__stack_limit} and
11155 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
11156 of 128KB@. Note that this may only work with the GNU linker.
11158 You can locally override stack limit checking by using the
11159 @code{no_stack_limit} function attribute (@pxref{Function Attributes}).
11161 @item -fsplit-stack
11162 @opindex fsplit-stack
11163 Generate code to automatically split the stack before it overflows.
11164 The resulting program has a discontiguous stack which can only
11165 overflow if the program is unable to allocate any more memory. This
11166 is most useful when running threaded programs, as it is no longer
11167 necessary to calculate a good stack size to use for each thread. This
11168 is currently only implemented for the x86 targets running
11171 When code compiled with @option{-fsplit-stack} calls code compiled
11172 without @option{-fsplit-stack}, there may not be much stack space
11173 available for the latter code to run. If compiling all code,
11174 including library code, with @option{-fsplit-stack} is not an option,
11175 then the linker can fix up these calls so that the code compiled
11176 without @option{-fsplit-stack} always has a large stack. Support for
11177 this is implemented in the gold linker in GNU binutils release 2.21
11180 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
11181 @opindex fvtable-verify
11182 This option is only available when compiling C++ code.
11183 It turns on (or off, if using @option{-fvtable-verify=none}) the security
11184 feature that verifies at run time, for every virtual call, that
11185 the vtable pointer through which the call is made is valid for the type of
11186 the object, and has not been corrupted or overwritten. If an invalid vtable
11187 pointer is detected at run time, an error is reported and execution of the
11188 program is immediately halted.
11190 This option causes run-time data structures to be built at program startup,
11191 which are used for verifying the vtable pointers.
11192 The options @samp{std} and @samp{preinit}
11193 control the timing of when these data structures are built. In both cases the
11194 data structures are built before execution reaches @code{main}. Using
11195 @option{-fvtable-verify=std} causes the data structures to be built after
11196 shared libraries have been loaded and initialized.
11197 @option{-fvtable-verify=preinit} causes them to be built before shared
11198 libraries have been loaded and initialized.
11200 If this option appears multiple times in the command line with different
11201 values specified, @samp{none} takes highest priority over both @samp{std} and
11202 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
11205 @opindex fvtv-debug
11206 When used in conjunction with @option{-fvtable-verify=std} or
11207 @option{-fvtable-verify=preinit}, causes debug versions of the
11208 runtime functions for the vtable verification feature to be called.
11209 This flag also causes the compiler to log information about which
11210 vtable pointers it finds for each class.
11211 This information is written to a file named @file{vtv_set_ptr_data.log}
11212 in the directory named by the environment variable @env{VTV_LOGS_DIR}
11213 if that is defined or the current working directory otherwise.
11215 Note: This feature @emph{appends} data to the log file. If you want a fresh log
11216 file, be sure to delete any existing one.
11219 @opindex fvtv-counts
11220 This is a debugging flag. When used in conjunction with
11221 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
11222 causes the compiler to keep track of the total number of virtual calls
11223 it encounters and the number of verifications it inserts. It also
11224 counts the number of calls to certain run-time library functions
11225 that it inserts and logs this information for each compilation unit.
11226 The compiler writes this information to a file named
11227 @file{vtv_count_data.log} in the directory named by the environment
11228 variable @env{VTV_LOGS_DIR} if that is defined or the current working
11229 directory otherwise. It also counts the size of the vtable pointer sets
11230 for each class, and writes this information to @file{vtv_class_set_sizes.log}
11231 in the same directory.
11233 Note: This feature @emph{appends} data to the log files. To get fresh log
11234 files, be sure to delete any existing ones.
11236 @item -finstrument-functions
11237 @opindex finstrument-functions
11238 Generate instrumentation calls for entry and exit to functions. Just
11239 after function entry and just before function exit, the following
11240 profiling functions are called with the address of the current
11241 function and its call site. (On some platforms,
11242 @code{__builtin_return_address} does not work beyond the current
11243 function, so the call site information may not be available to the
11244 profiling functions otherwise.)
11247 void __cyg_profile_func_enter (void *this_fn,
11249 void __cyg_profile_func_exit (void *this_fn,
11253 The first argument is the address of the start of the current function,
11254 which may be looked up exactly in the symbol table.
11256 This instrumentation is also done for functions expanded inline in other
11257 functions. The profiling calls indicate where, conceptually, the
11258 inline function is entered and exited. This means that addressable
11259 versions of such functions must be available. If all your uses of a
11260 function are expanded inline, this may mean an additional expansion of
11261 code size. If you use @code{extern inline} in your C code, an
11262 addressable version of such functions must be provided. (This is
11263 normally the case anyway, but if you get lucky and the optimizer always
11264 expands the functions inline, you might have gotten away without
11265 providing static copies.)
11267 A function may be given the attribute @code{no_instrument_function}, in
11268 which case this instrumentation is not done. This can be used, for
11269 example, for the profiling functions listed above, high-priority
11270 interrupt routines, and any functions from which the profiling functions
11271 cannot safely be called (perhaps signal handlers, if the profiling
11272 routines generate output or allocate memory).
11274 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
11275 @opindex finstrument-functions-exclude-file-list
11277 Set the list of functions that are excluded from instrumentation (see
11278 the description of @option{-finstrument-functions}). If the file that
11279 contains a function definition matches with one of @var{file}, then
11280 that function is not instrumented. The match is done on substrings:
11281 if the @var{file} parameter is a substring of the file name, it is
11282 considered to be a match.
11287 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
11291 excludes any inline function defined in files whose pathnames
11292 contain @file{/bits/stl} or @file{include/sys}.
11294 If, for some reason, you want to include letter @samp{,} in one of
11295 @var{sym}, write @samp{\,}. For example,
11296 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
11297 (note the single quote surrounding the option).
11299 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
11300 @opindex finstrument-functions-exclude-function-list
11302 This is similar to @option{-finstrument-functions-exclude-file-list},
11303 but this option sets the list of function names to be excluded from
11304 instrumentation. The function name to be matched is its user-visible
11305 name, such as @code{vector<int> blah(const vector<int> &)}, not the
11306 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
11307 match is done on substrings: if the @var{sym} parameter is a substring
11308 of the function name, it is considered to be a match. For C99 and C++
11309 extended identifiers, the function name must be given in UTF-8, not
11310 using universal character names.
11315 @node Preprocessor Options
11316 @section Options Controlling the Preprocessor
11317 @cindex preprocessor options
11318 @cindex options, preprocessor
11320 These options control the C preprocessor, which is run on each C source
11321 file before actual compilation.
11323 If you use the @option{-E} option, nothing is done except preprocessing.
11324 Some of these options make sense only together with @option{-E} because
11325 they cause the preprocessor output to be unsuitable for actual
11329 @item -Wp,@var{option}
11331 You can use @option{-Wp,@var{option}} to bypass the compiler driver
11332 and pass @var{option} directly through to the preprocessor. If
11333 @var{option} contains commas, it is split into multiple options at the
11334 commas. However, many options are modified, translated or interpreted
11335 by the compiler driver before being passed to the preprocessor, and
11336 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
11337 interface is undocumented and subject to change, so whenever possible
11338 you should avoid using @option{-Wp} and let the driver handle the
11341 @item -Xpreprocessor @var{option}
11342 @opindex Xpreprocessor
11343 Pass @var{option} as an option to the preprocessor. You can use this to
11344 supply system-specific preprocessor options that GCC does not
11347 If you want to pass an option that takes an argument, you must use
11348 @option{-Xpreprocessor} twice, once for the option and once for the argument.
11350 @item -no-integrated-cpp
11351 @opindex no-integrated-cpp
11352 Perform preprocessing as a separate pass before compilation.
11353 By default, GCC performs preprocessing as an integrated part of
11354 input tokenization and parsing.
11355 If this option is provided, the appropriate language front end
11356 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
11357 and Objective-C, respectively) is instead invoked twice,
11358 once for preprocessing only and once for actual compilation
11359 of the preprocessed input.
11360 This option may be useful in conjunction with the @option{-B} or
11361 @option{-wrapper} options to specify an alternate preprocessor or
11362 perform additional processing of the program source between
11363 normal preprocessing and compilation.
11366 @include cppopts.texi
11368 @node Assembler Options
11369 @section Passing Options to the Assembler
11371 @c prevent bad page break with this line
11372 You can pass options to the assembler.
11375 @item -Wa,@var{option}
11377 Pass @var{option} as an option to the assembler. If @var{option}
11378 contains commas, it is split into multiple options at the commas.
11380 @item -Xassembler @var{option}
11381 @opindex Xassembler
11382 Pass @var{option} as an option to the assembler. You can use this to
11383 supply system-specific assembler options that GCC does not
11386 If you want to pass an option that takes an argument, you must use
11387 @option{-Xassembler} twice, once for the option and once for the argument.
11392 @section Options for Linking
11393 @cindex link options
11394 @cindex options, linking
11396 These options come into play when the compiler links object files into
11397 an executable output file. They are meaningless if the compiler is
11398 not doing a link step.
11402 @item @var{object-file-name}
11403 A file name that does not end in a special recognized suffix is
11404 considered to name an object file or library. (Object files are
11405 distinguished from libraries by the linker according to the file
11406 contents.) If linking is done, these object files are used as input
11415 If any of these options is used, then the linker is not run, and
11416 object file names should not be used as arguments. @xref{Overall
11420 @opindex fuse-ld=bfd
11421 Use the @command{bfd} linker instead of the default linker.
11423 @item -fuse-ld=gold
11424 @opindex fuse-ld=gold
11425 Use the @command{gold} linker instead of the default linker.
11428 @item -l@var{library}
11429 @itemx -l @var{library}
11431 Search the library named @var{library} when linking. (The second
11432 alternative with the library as a separate argument is only for
11433 POSIX compliance and is not recommended.)
11435 It makes a difference where in the command you write this option; the
11436 linker searches and processes libraries and object files in the order they
11437 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
11438 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
11439 to functions in @samp{z}, those functions may not be loaded.
11441 The linker searches a standard list of directories for the library,
11442 which is actually a file named @file{lib@var{library}.a}. The linker
11443 then uses this file as if it had been specified precisely by name.
11445 The directories searched include several standard system directories
11446 plus any that you specify with @option{-L}.
11448 Normally the files found this way are library files---archive files
11449 whose members are object files. The linker handles an archive file by
11450 scanning through it for members which define symbols that have so far
11451 been referenced but not defined. But if the file that is found is an
11452 ordinary object file, it is linked in the usual fashion. The only
11453 difference between using an @option{-l} option and specifying a file name
11454 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
11455 and searches several directories.
11459 You need this special case of the @option{-l} option in order to
11460 link an Objective-C or Objective-C++ program.
11462 @item -nostartfiles
11463 @opindex nostartfiles
11464 Do not use the standard system startup files when linking.
11465 The standard system libraries are used normally, unless @option{-nostdlib}
11466 or @option{-nodefaultlibs} is used.
11468 @item -nodefaultlibs
11469 @opindex nodefaultlibs
11470 Do not use the standard system libraries when linking.
11471 Only the libraries you specify are passed to the linker, and options
11472 specifying linkage of the system libraries, such as @option{-static-libgcc}
11473 or @option{-shared-libgcc}, are ignored.
11474 The standard startup files are used normally, unless @option{-nostartfiles}
11477 The compiler may generate calls to @code{memcmp},
11478 @code{memset}, @code{memcpy} and @code{memmove}.
11479 These entries are usually resolved by entries in
11480 libc. These entry points should be supplied through some other
11481 mechanism when this option is specified.
11485 Do not use the standard system startup files or libraries when linking.
11486 No startup files and only the libraries you specify are passed to
11487 the linker, and options specifying linkage of the system libraries, such as
11488 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
11490 The compiler may generate calls to @code{memcmp}, @code{memset},
11491 @code{memcpy} and @code{memmove}.
11492 These entries are usually resolved by entries in
11493 libc. These entry points should be supplied through some other
11494 mechanism when this option is specified.
11496 @cindex @option{-lgcc}, use with @option{-nostdlib}
11497 @cindex @option{-nostdlib} and unresolved references
11498 @cindex unresolved references and @option{-nostdlib}
11499 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
11500 @cindex @option{-nodefaultlibs} and unresolved references
11501 @cindex unresolved references and @option{-nodefaultlibs}
11502 One of the standard libraries bypassed by @option{-nostdlib} and
11503 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
11504 which GCC uses to overcome shortcomings of particular machines, or special
11505 needs for some languages.
11506 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
11507 Collection (GCC) Internals},
11508 for more discussion of @file{libgcc.a}.)
11509 In most cases, you need @file{libgcc.a} even when you want to avoid
11510 other standard libraries. In other words, when you specify @option{-nostdlib}
11511 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
11512 This ensures that you have no unresolved references to internal GCC
11513 library subroutines.
11514 (An example of such an internal subroutine is @code{__main}, used to ensure C++
11515 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
11516 GNU Compiler Collection (GCC) Internals}.)
11520 Produce a position independent executable on targets that support it.
11521 For predictable results, you must also specify the same set of options
11522 used for compilation (@option{-fpie}, @option{-fPIE},
11523 or model suboptions) when you specify this linker option.
11527 Don't produce a position independent executable.
11531 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
11532 that support it. This instructs the linker to add all symbols, not
11533 only used ones, to the dynamic symbol table. This option is needed
11534 for some uses of @code{dlopen} or to allow obtaining backtraces
11535 from within a program.
11539 Remove all symbol table and relocation information from the executable.
11543 On systems that support dynamic linking, this prevents linking with the shared
11544 libraries. On other systems, this option has no effect.
11548 Produce a shared object which can then be linked with other objects to
11549 form an executable. Not all systems support this option. For predictable
11550 results, you must also specify the same set of options used for compilation
11551 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
11552 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
11553 needs to build supplementary stub code for constructors to work. On
11554 multi-libbed systems, @samp{gcc -shared} must select the correct support
11555 libraries to link against. Failing to supply the correct flags may lead
11556 to subtle defects. Supplying them in cases where they are not necessary
11559 @item -shared-libgcc
11560 @itemx -static-libgcc
11561 @opindex shared-libgcc
11562 @opindex static-libgcc
11563 On systems that provide @file{libgcc} as a shared library, these options
11564 force the use of either the shared or static version, respectively.
11565 If no shared version of @file{libgcc} was built when the compiler was
11566 configured, these options have no effect.
11568 There are several situations in which an application should use the
11569 shared @file{libgcc} instead of the static version. The most common
11570 of these is when the application wishes to throw and catch exceptions
11571 across different shared libraries. In that case, each of the libraries
11572 as well as the application itself should use the shared @file{libgcc}.
11574 Therefore, the G++ and driver automatically adds @option{-shared-libgcc}
11575 whenever you build a shared library or a main executable, because C++
11576 programs typically use exceptions, so this is the right thing to do.
11578 If, instead, you use the GCC driver to create shared libraries, you may
11579 find that they are not always linked with the shared @file{libgcc}.
11580 If GCC finds, at its configuration time, that you have a non-GNU linker
11581 or a GNU linker that does not support option @option{--eh-frame-hdr},
11582 it links the shared version of @file{libgcc} into shared libraries
11583 by default. Otherwise, it takes advantage of the linker and optimizes
11584 away the linking with the shared version of @file{libgcc}, linking with
11585 the static version of libgcc by default. This allows exceptions to
11586 propagate through such shared libraries, without incurring relocation
11587 costs at library load time.
11589 However, if a library or main executable is supposed to throw or catch
11590 exceptions, you must link it using the G++ driver, as appropriate
11591 for the languages used in the program, or using the option
11592 @option{-shared-libgcc}, such that it is linked with the shared
11595 @item -static-libasan
11596 @opindex static-libasan
11597 When the @option{-fsanitize=address} option is used to link a program,
11598 the GCC driver automatically links against @option{libasan}. If
11599 @file{libasan} is available as a shared library, and the @option{-static}
11600 option is not used, then this links against the shared version of
11601 @file{libasan}. The @option{-static-libasan} option directs the GCC
11602 driver to link @file{libasan} statically, without necessarily linking
11603 other libraries statically.
11605 @item -static-libtsan
11606 @opindex static-libtsan
11607 When the @option{-fsanitize=thread} option is used to link a program,
11608 the GCC driver automatically links against @option{libtsan}. If
11609 @file{libtsan} is available as a shared library, and the @option{-static}
11610 option is not used, then this links against the shared version of
11611 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
11612 driver to link @file{libtsan} statically, without necessarily linking
11613 other libraries statically.
11615 @item -static-liblsan
11616 @opindex static-liblsan
11617 When the @option{-fsanitize=leak} option is used to link a program,
11618 the GCC driver automatically links against @option{liblsan}. If
11619 @file{liblsan} is available as a shared library, and the @option{-static}
11620 option is not used, then this links against the shared version of
11621 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
11622 driver to link @file{liblsan} statically, without necessarily linking
11623 other libraries statically.
11625 @item -static-libubsan
11626 @opindex static-libubsan
11627 When the @option{-fsanitize=undefined} option is used to link a program,
11628 the GCC driver automatically links against @option{libubsan}. If
11629 @file{libubsan} is available as a shared library, and the @option{-static}
11630 option is not used, then this links against the shared version of
11631 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
11632 driver to link @file{libubsan} statically, without necessarily linking
11633 other libraries statically.
11635 @item -static-libmpx
11636 @opindex static-libmpx
11637 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are
11638 used to link a program, the GCC driver automatically links against
11639 @file{libmpx}. If @file{libmpx} is available as a shared library,
11640 and the @option{-static} option is not used, then this links against
11641 the shared version of @file{libmpx}. The @option{-static-libmpx}
11642 option directs the GCC driver to link @file{libmpx} statically,
11643 without necessarily linking other libraries statically.
11645 @item -static-libmpxwrappers
11646 @opindex static-libmpxwrappers
11647 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are used
11648 to link a program without also using @option{-fno-chkp-use-wrappers}, the
11649 GCC driver automatically links against @file{libmpxwrappers}. If
11650 @file{libmpxwrappers} is available as a shared library, and the
11651 @option{-static} option is not used, then this links against the shared
11652 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
11653 option directs the GCC driver to link @file{libmpxwrappers} statically,
11654 without necessarily linking other libraries statically.
11656 @item -static-libstdc++
11657 @opindex static-libstdc++
11658 When the @command{g++} program is used to link a C++ program, it
11659 normally automatically links against @option{libstdc++}. If
11660 @file{libstdc++} is available as a shared library, and the
11661 @option{-static} option is not used, then this links against the
11662 shared version of @file{libstdc++}. That is normally fine. However, it
11663 is sometimes useful to freeze the version of @file{libstdc++} used by
11664 the program without going all the way to a fully static link. The
11665 @option{-static-libstdc++} option directs the @command{g++} driver to
11666 link @file{libstdc++} statically, without necessarily linking other
11667 libraries statically.
11671 Bind references to global symbols when building a shared object. Warn
11672 about any unresolved references (unless overridden by the link editor
11673 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
11676 @item -T @var{script}
11678 @cindex linker script
11679 Use @var{script} as the linker script. This option is supported by most
11680 systems using the GNU linker. On some targets, such as bare-board
11681 targets without an operating system, the @option{-T} option may be required
11682 when linking to avoid references to undefined symbols.
11684 @item -Xlinker @var{option}
11686 Pass @var{option} as an option to the linker. You can use this to
11687 supply system-specific linker options that GCC does not recognize.
11689 If you want to pass an option that takes a separate argument, you must use
11690 @option{-Xlinker} twice, once for the option and once for the argument.
11691 For example, to pass @option{-assert definitions}, you must write
11692 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
11693 @option{-Xlinker "-assert definitions"}, because this passes the entire
11694 string as a single argument, which is not what the linker expects.
11696 When using the GNU linker, it is usually more convenient to pass
11697 arguments to linker options using the @option{@var{option}=@var{value}}
11698 syntax than as separate arguments. For example, you can specify
11699 @option{-Xlinker -Map=output.map} rather than
11700 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
11701 this syntax for command-line options.
11703 @item -Wl,@var{option}
11705 Pass @var{option} as an option to the linker. If @var{option} contains
11706 commas, it is split into multiple options at the commas. You can use this
11707 syntax to pass an argument to the option.
11708 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
11709 linker. When using the GNU linker, you can also get the same effect with
11710 @option{-Wl,-Map=output.map}.
11712 @item -u @var{symbol}
11714 Pretend the symbol @var{symbol} is undefined, to force linking of
11715 library modules to define it. You can use @option{-u} multiple times with
11716 different symbols to force loading of additional library modules.
11718 @item -z @var{keyword}
11720 @option{-z} is passed directly on to the linker along with the keyword
11721 @var{keyword}. See the section in the documentation of your linker for
11722 permitted values and their meanings.
11725 @node Directory Options
11726 @section Options for Directory Search
11727 @cindex directory options
11728 @cindex options, directory search
11729 @cindex search path
11731 These options specify directories to search for header files, for
11732 libraries and for parts of the compiler:
11737 Add the directory @var{dir} to the head of the list of directories to be
11738 searched for header files. This can be used to override a system header
11739 file, substituting your own version, since these directories are
11740 searched before the system header file directories. However, you should
11741 not use this option to add directories that contain vendor-supplied
11742 system header files (use @option{-isystem} for that). If you use more than
11743 one @option{-I} option, the directories are scanned in left-to-right
11744 order; the standard system directories come after.
11746 If a standard system include directory, or a directory specified with
11747 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
11748 option is ignored. The directory is still searched but as a
11749 system directory at its normal position in the system include chain.
11750 This is to ensure that GCC's procedure to fix buggy system headers and
11751 the ordering for the @code{include_next} directive are not inadvertently changed.
11752 If you really need to change the search order for system directories,
11753 use the @option{-nostdinc} and/or @option{-isystem} options.
11755 @item -iplugindir=@var{dir}
11756 @opindex iplugindir=
11757 Set the directory to search for plugins that are passed
11758 by @option{-fplugin=@var{name}} instead of
11759 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
11760 to be used by the user, but only passed by the driver.
11762 @item -iquote@var{dir}
11764 Add the directory @var{dir} to the head of the list of directories to
11765 be searched for header files only for the case of @code{#include
11766 "@var{file}"}; they are not searched for @code{#include <@var{file}>},
11767 otherwise just like @option{-I}.
11771 Add directory @var{dir} to the list of directories to be searched
11774 @item -B@var{prefix}
11776 This option specifies where to find the executables, libraries,
11777 include files, and data files of the compiler itself.
11779 The compiler driver program runs one or more of the subprograms
11780 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
11781 @var{prefix} as a prefix for each program it tries to run, both with and
11782 without @samp{@var{machine}/@var{version}/} for the corresponding target
11783 machine and compiler version.
11785 For each subprogram to be run, the compiler driver first tries the
11786 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
11787 is not specified, the driver tries two standard prefixes,
11788 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
11789 those results in a file name that is found, the unmodified program
11790 name is searched for using the directories specified in your
11791 @env{PATH} environment variable.
11793 The compiler checks to see if the path provided by @option{-B}
11794 refers to a directory, and if necessary it adds a directory
11795 separator character at the end of the path.
11797 @option{-B} prefixes that effectively specify directory names also apply
11798 to libraries in the linker, because the compiler translates these
11799 options into @option{-L} options for the linker. They also apply to
11800 include files in the preprocessor, because the compiler translates these
11801 options into @option{-isystem} options for the preprocessor. In this case,
11802 the compiler appends @samp{include} to the prefix.
11804 The runtime support file @file{libgcc.a} can also be searched for using
11805 the @option{-B} prefix, if needed. If it is not found there, the two
11806 standard prefixes above are tried, and that is all. The file is left
11807 out of the link if it is not found by those means.
11809 Another way to specify a prefix much like the @option{-B} prefix is to use
11810 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
11813 As a special kludge, if the path provided by @option{-B} is
11814 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
11815 9, then it is replaced by @file{[dir/]include}. This is to help
11816 with boot-strapping the compiler.
11818 @item -no-canonical-prefixes
11819 @opindex no-canonical-prefixes
11820 Do not expand any symbolic links, resolve references to @samp{/../}
11821 or @samp{/./}, or make the path absolute when generating a relative
11824 @item --sysroot=@var{dir}
11826 Use @var{dir} as the logical root directory for headers and libraries.
11827 For example, if the compiler normally searches for headers in
11828 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
11829 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
11831 If you use both this option and the @option{-isysroot} option, then
11832 the @option{--sysroot} option applies to libraries, but the
11833 @option{-isysroot} option applies to header files.
11835 The GNU linker (beginning with version 2.16) has the necessary support
11836 for this option. If your linker does not support this option, the
11837 header file aspect of @option{--sysroot} still works, but the
11838 library aspect does not.
11840 @item --no-sysroot-suffix
11841 @opindex no-sysroot-suffix
11842 For some targets, a suffix is added to the root directory specified
11843 with @option{--sysroot}, depending on the other options used, so that
11844 headers may for example be found in
11845 @file{@var{dir}/@var{suffix}/usr/include} instead of
11846 @file{@var{dir}/usr/include}. This option disables the addition of
11851 This option has been deprecated. Please use @option{-iquote} instead for
11852 @option{-I} directories before the @option{-I-} and remove the @option{-I-}
11854 Any directories you specify with @option{-I} options before the @option{-I-}
11855 option are searched only for the case of @code{#include "@var{file}"};
11856 they are not searched for @code{#include <@var{file}>}.
11858 If additional directories are specified with @option{-I} options after
11859 the @option{-I-} option, these directories are searched for all @code{#include}
11860 directives. (Ordinarily @emph{all} @option{-I} directories are used
11863 In addition, the @option{-I-} option inhibits the use of the current
11864 directory (where the current input file came from) as the first search
11865 directory for @code{#include "@var{file}"}. There is no way to
11866 override this effect of @option{-I-}. With @option{-I.} you can specify
11867 searching the directory that is current when the compiler is
11868 invoked. That is not exactly the same as what the preprocessor does
11869 by default, but it is often satisfactory.
11871 @option{-I-} does not inhibit the use of the standard system directories
11872 for header files. Thus, @option{-I-} and @option{-nostdinc} are
11876 @node Code Gen Options
11877 @section Options for Code Generation Conventions
11878 @cindex code generation conventions
11879 @cindex options, code generation
11880 @cindex run-time options
11882 These machine-independent options control the interface conventions
11883 used in code generation.
11885 Most of them have both positive and negative forms; the negative form
11886 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
11887 one of the forms is listed---the one that is not the default. You
11888 can figure out the other form by either removing @samp{no-} or adding
11892 @item -fstack-reuse=@var{reuse-level}
11893 @opindex fstack_reuse
11894 This option controls stack space reuse for user declared local/auto variables
11895 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
11896 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
11897 local variables and temporaries, @samp{named_vars} enables the reuse only for
11898 user defined local variables with names, and @samp{none} disables stack reuse
11899 completely. The default value is @samp{all}. The option is needed when the
11900 program extends the lifetime of a scoped local variable or a compiler generated
11901 temporary beyond the end point defined by the language. When a lifetime of
11902 a variable ends, and if the variable lives in memory, the optimizing compiler
11903 has the freedom to reuse its stack space with other temporaries or scoped
11904 local variables whose live range does not overlap with it. Legacy code extending
11905 local lifetime is likely to break with the stack reuse optimization.
11924 if (*p == 10) // out of scope use of local1
11935 A(int k) : i(k), j(k) @{ @}
11942 void foo(const A& ar)
11949 foo(A(10)); // temp object's lifetime ends when foo returns
11955 ap->i+= 10; // ap references out of scope temp whose space
11956 // is reused with a. What is the value of ap->i?
11961 The lifetime of a compiler generated temporary is well defined by the C++
11962 standard. When a lifetime of a temporary ends, and if the temporary lives
11963 in memory, the optimizing compiler has the freedom to reuse its stack
11964 space with other temporaries or scoped local variables whose live range
11965 does not overlap with it. However some of the legacy code relies on
11966 the behavior of older compilers in which temporaries' stack space is
11967 not reused, the aggressive stack reuse can lead to runtime errors. This
11968 option is used to control the temporary stack reuse optimization.
11972 This option generates traps for signed overflow on addition, subtraction,
11973 multiplication operations.
11974 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
11975 @option{-ftrapv} @option{-fwrapv} on the command-line results in
11976 @option{-fwrapv} being effective. Note that only active options override, so
11977 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
11978 results in @option{-ftrapv} being effective.
11982 This option instructs the compiler to assume that signed arithmetic
11983 overflow of addition, subtraction and multiplication wraps around
11984 using twos-complement representation. This flag enables some optimizations
11985 and disables others.
11986 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
11987 @option{-ftrapv} @option{-fwrapv} on the command-line results in
11988 @option{-fwrapv} being effective. Note that only active options override, so
11989 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
11990 results in @option{-ftrapv} being effective.
11993 @opindex fexceptions
11994 Enable exception handling. Generates extra code needed to propagate
11995 exceptions. For some targets, this implies GCC generates frame
11996 unwind information for all functions, which can produce significant data
11997 size overhead, although it does not affect execution. If you do not
11998 specify this option, GCC enables it by default for languages like
11999 C++ that normally require exception handling, and disables it for
12000 languages like C that do not normally require it. However, you may need
12001 to enable this option when compiling C code that needs to interoperate
12002 properly with exception handlers written in C++. You may also wish to
12003 disable this option if you are compiling older C++ programs that don't
12004 use exception handling.
12006 @item -fnon-call-exceptions
12007 @opindex fnon-call-exceptions
12008 Generate code that allows trapping instructions to throw exceptions.
12009 Note that this requires platform-specific runtime support that does
12010 not exist everywhere. Moreover, it only allows @emph{trapping}
12011 instructions to throw exceptions, i.e.@: memory references or floating-point
12012 instructions. It does not allow exceptions to be thrown from
12013 arbitrary signal handlers such as @code{SIGALRM}.
12015 @item -fdelete-dead-exceptions
12016 @opindex fdelete-dead-exceptions
12017 Consider that instructions that may throw exceptions but don't otherwise
12018 contribute to the execution of the program can be optimized away.
12019 This option is enabled by default for the Ada front end, as permitted by
12020 the Ada language specification.
12021 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
12023 @item -funwind-tables
12024 @opindex funwind-tables
12025 Similar to @option{-fexceptions}, except that it just generates any needed
12026 static data, but does not affect the generated code in any other way.
12027 You normally do not need to enable this option; instead, a language processor
12028 that needs this handling enables it on your behalf.
12030 @item -fasynchronous-unwind-tables
12031 @opindex fasynchronous-unwind-tables
12032 Generate unwind table in DWARF format, if supported by target machine. The
12033 table is exact at each instruction boundary, so it can be used for stack
12034 unwinding from asynchronous events (such as debugger or garbage collector).
12036 @item -fno-gnu-unique
12037 @opindex fno-gnu-unique
12038 On systems with recent GNU assembler and C library, the C++ compiler
12039 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
12040 of template static data members and static local variables in inline
12041 functions are unique even in the presence of @code{RTLD_LOCAL}; this
12042 is necessary to avoid problems with a library used by two different
12043 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
12044 therefore disagreeing with the other one about the binding of the
12045 symbol. But this causes @code{dlclose} to be ignored for affected
12046 DSOs; if your program relies on reinitialization of a DSO via
12047 @code{dlclose} and @code{dlopen}, you can use
12048 @option{-fno-gnu-unique}.
12050 @item -fpcc-struct-return
12051 @opindex fpcc-struct-return
12052 Return ``short'' @code{struct} and @code{union} values in memory like
12053 longer ones, rather than in registers. This convention is less
12054 efficient, but it has the advantage of allowing intercallability between
12055 GCC-compiled files and files compiled with other compilers, particularly
12056 the Portable C Compiler (pcc).
12058 The precise convention for returning structures in memory depends
12059 on the target configuration macros.
12061 Short structures and unions are those whose size and alignment match
12062 that of some integer type.
12064 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
12065 switch is not binary compatible with code compiled with the
12066 @option{-freg-struct-return} switch.
12067 Use it to conform to a non-default application binary interface.
12069 @item -freg-struct-return
12070 @opindex freg-struct-return
12071 Return @code{struct} and @code{union} values in registers when possible.
12072 This is more efficient for small structures than
12073 @option{-fpcc-struct-return}.
12075 If you specify neither @option{-fpcc-struct-return} nor
12076 @option{-freg-struct-return}, GCC defaults to whichever convention is
12077 standard for the target. If there is no standard convention, GCC
12078 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
12079 the principal compiler. In those cases, we can choose the standard, and
12080 we chose the more efficient register return alternative.
12082 @strong{Warning:} code compiled with the @option{-freg-struct-return}
12083 switch is not binary compatible with code compiled with the
12084 @option{-fpcc-struct-return} switch.
12085 Use it to conform to a non-default application binary interface.
12087 @item -fshort-enums
12088 @opindex fshort-enums
12089 Allocate to an @code{enum} type only as many bytes as it needs for the
12090 declared range of possible values. Specifically, the @code{enum} type
12091 is equivalent to the smallest integer type that has enough room.
12093 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
12094 code that is not binary compatible with code generated without that switch.
12095 Use it to conform to a non-default application binary interface.
12097 @item -fshort-wchar
12098 @opindex fshort-wchar
12099 Override the underlying type for @code{wchar_t} to be @code{short
12100 unsigned int} instead of the default for the target. This option is
12101 useful for building programs to run under WINE@.
12103 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
12104 code that is not binary compatible with code generated without that switch.
12105 Use it to conform to a non-default application binary interface.
12108 @opindex fno-common
12109 @cindex tentative definitions
12110 In C code, this option controls the placement of global variables
12111 defined without an initializer, known as @dfn{tentative definitions}
12112 in the C standard. Tentative definitions are distinct from declarations
12113 of a variable with the @code{extern} keyword, which do not allocate storage.
12115 Unix C compilers have traditionally allocated storage for
12116 uninitialized global variables in a common block. This allows the
12117 linker to resolve all tentative definitions of the same variable
12118 in different compilation units to the same object, or to a non-tentative
12120 This is the behavior specified by @option{-fcommon}, and is the default for
12121 GCC on most targets.
12122 On the other hand, this behavior is not required by ISO
12123 C, and on some targets may carry a speed or code size penalty on
12124 variable references.
12126 The @option{-fno-common} option specifies that the compiler should instead
12127 place uninitialized global variables in the data section of the object file.
12128 This inhibits the merging of tentative definitions by the linker so
12129 you get a multiple-definition error if the same
12130 variable is defined in more than one compilation unit.
12131 Compiling with @option{-fno-common} is useful on targets for which
12132 it provides better performance, or if you wish to verify that the
12133 program will work on other systems that always treat uninitialized
12134 variable definitions this way.
12138 Ignore the @code{#ident} directive.
12140 @item -finhibit-size-directive
12141 @opindex finhibit-size-directive
12142 Don't output a @code{.size} assembler directive, or anything else that
12143 would cause trouble if the function is split in the middle, and the
12144 two halves are placed at locations far apart in memory. This option is
12145 used when compiling @file{crtstuff.c}; you should not need to use it
12148 @item -fverbose-asm
12149 @opindex fverbose-asm
12150 Put extra commentary information in the generated assembly code to
12151 make it more readable. This option is generally only of use to those
12152 who actually need to read the generated assembly code (perhaps while
12153 debugging the compiler itself).
12155 @option{-fno-verbose-asm}, the default, causes the
12156 extra information to be omitted and is useful when comparing two assembler
12159 The added comments include:
12164 information on the compiler version and command-line options,
12167 the source code lines associated with the assembly instructions,
12168 in the form FILENAME:LINENUMBER:CONTENT OF LINE,
12171 hints on which high-level expressions correspond to
12172 the various assembly instruction operands.
12176 For example, given this C source file:
12184 for (i = 0; i < n; i++)
12191 compiling to (x86_64) assembly via @option{-S} and emitting the result
12192 direct to stdout via @option{-o} @option{-}
12195 gcc -S test.c -fverbose-asm -Os -o -
12198 gives output similar to this:
12202 # GNU C11 (GCC) version 7.0.0 20160809 (experimental) (x86_64-pc-linux-gnu)
12209 .type test, @@function
12213 # test.c:4: int total = 0;
12214 xorl %eax, %eax # <retval>
12215 # test.c:6: for (i = 0; i < n; i++)
12216 xorl %edx, %edx # i
12218 # test.c:6: for (i = 0; i < n; i++)
12219 cmpl %edi, %edx # n, i
12221 # test.c:7: total += i * i;
12222 movl %edx, %ecx # i, tmp92
12223 imull %edx, %ecx # i, tmp92
12224 # test.c:6: for (i = 0; i < n; i++)
12226 # test.c:7: total += i * i;
12227 addl %ecx, %eax # tmp92, <retval>
12235 .ident "GCC: (GNU) 7.0.0 20160809 (experimental)"
12236 .section .note.GNU-stack,"",@@progbits
12239 The comments are intended for humans rather than machines and hence the
12240 precise format of the comments is subject to change.
12242 @item -frecord-gcc-switches
12243 @opindex frecord-gcc-switches
12244 This switch causes the command line used to invoke the
12245 compiler to be recorded into the object file that is being created.
12246 This switch is only implemented on some targets and the exact format
12247 of the recording is target and binary file format dependent, but it
12248 usually takes the form of a section containing ASCII text. This
12249 switch is related to the @option{-fverbose-asm} switch, but that
12250 switch only records information in the assembler output file as
12251 comments, so it never reaches the object file.
12252 See also @option{-grecord-gcc-switches} for another
12253 way of storing compiler options into the object file.
12257 @cindex global offset table
12259 Generate position-independent code (PIC) suitable for use in a shared
12260 library, if supported for the target machine. Such code accesses all
12261 constant addresses through a global offset table (GOT)@. The dynamic
12262 loader resolves the GOT entries when the program starts (the dynamic
12263 loader is not part of GCC; it is part of the operating system). If
12264 the GOT size for the linked executable exceeds a machine-specific
12265 maximum size, you get an error message from the linker indicating that
12266 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12267 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
12268 on the m68k and RS/6000. The x86 has no such limit.)
12270 Position-independent code requires special support, and therefore works
12271 only on certain machines. For the x86, GCC supports PIC for System V
12272 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12273 position-independent.
12275 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12280 If supported for the target machine, emit position-independent code,
12281 suitable for dynamic linking and avoiding any limit on the size of the
12282 global offset table. This option makes a difference on AArch64, m68k,
12283 PowerPC and SPARC@.
12285 Position-independent code requires special support, and therefore works
12286 only on certain machines.
12288 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12295 These options are similar to @option{-fpic} and @option{-fPIC}, but
12296 generated position independent code can be only linked into executables.
12297 Usually these options are used when @option{-pie} GCC option is
12298 used during linking.
12300 @option{-fpie} and @option{-fPIE} both define the macros
12301 @code{__pie__} and @code{__PIE__}. The macros have the value 1
12302 for @option{-fpie} and 2 for @option{-fPIE}.
12306 Do not use the PLT for external function calls in position-independent code.
12307 Instead, load the callee address at call sites from the GOT and branch to it.
12308 This leads to more efficient code by eliminating PLT stubs and exposing
12309 GOT loads to optimizations. On architectures such as 32-bit x86 where
12310 PLT stubs expect the GOT pointer in a specific register, this gives more
12311 register allocation freedom to the compiler.
12312 Lazy binding requires use of the PLT;
12313 with @option{-fno-plt} all external symbols are resolved at load time.
12315 Alternatively, the function attribute @code{noplt} can be used to avoid calls
12316 through the PLT for specific external functions.
12318 In position-dependent code, a few targets also convert calls to
12319 functions that are marked to not use the PLT to use the GOT instead.
12321 @item -fno-jump-tables
12322 @opindex fno-jump-tables
12323 Do not use jump tables for switch statements even where it would be
12324 more efficient than other code generation strategies. This option is
12325 of use in conjunction with @option{-fpic} or @option{-fPIC} for
12326 building code that forms part of a dynamic linker and cannot
12327 reference the address of a jump table. On some targets, jump tables
12328 do not require a GOT and this option is not needed.
12330 @item -ffixed-@var{reg}
12332 Treat the register named @var{reg} as a fixed register; generated code
12333 should never refer to it (except perhaps as a stack pointer, frame
12334 pointer or in some other fixed role).
12336 @var{reg} must be the name of a register. The register names accepted
12337 are machine-specific and are defined in the @code{REGISTER_NAMES}
12338 macro in the machine description macro file.
12340 This flag does not have a negative form, because it specifies a
12343 @item -fcall-used-@var{reg}
12344 @opindex fcall-used
12345 Treat the register named @var{reg} as an allocable register that is
12346 clobbered by function calls. It may be allocated for temporaries or
12347 variables that do not live across a call. Functions compiled this way
12348 do not save and restore the register @var{reg}.
12350 It is an error to use this flag with the frame pointer or stack pointer.
12351 Use of this flag for other registers that have fixed pervasive roles in
12352 the machine's execution model produces disastrous results.
12354 This flag does not have a negative form, because it specifies a
12357 @item -fcall-saved-@var{reg}
12358 @opindex fcall-saved
12359 Treat the register named @var{reg} as an allocable register saved by
12360 functions. It may be allocated even for temporaries or variables that
12361 live across a call. Functions compiled this way save and restore
12362 the register @var{reg} if they use it.
12364 It is an error to use this flag with the frame pointer or stack pointer.
12365 Use of this flag for other registers that have fixed pervasive roles in
12366 the machine's execution model produces disastrous results.
12368 A different sort of disaster results from the use of this flag for
12369 a register in which function values may be returned.
12371 This flag does not have a negative form, because it specifies a
12374 @item -fpack-struct[=@var{n}]
12375 @opindex fpack-struct
12376 Without a value specified, pack all structure members together without
12377 holes. When a value is specified (which must be a small power of two), pack
12378 structure members according to this value, representing the maximum
12379 alignment (that is, objects with default alignment requirements larger than
12380 this are output potentially unaligned at the next fitting location.
12382 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
12383 code that is not binary compatible with code generated without that switch.
12384 Additionally, it makes the code suboptimal.
12385 Use it to conform to a non-default application binary interface.
12387 @item -fleading-underscore
12388 @opindex fleading-underscore
12389 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
12390 change the way C symbols are represented in the object file. One use
12391 is to help link with legacy assembly code.
12393 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
12394 generate code that is not binary compatible with code generated without that
12395 switch. Use it to conform to a non-default application binary interface.
12396 Not all targets provide complete support for this switch.
12398 @item -ftls-model=@var{model}
12399 @opindex ftls-model
12400 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
12401 The @var{model} argument should be one of @samp{global-dynamic},
12402 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
12403 Note that the choice is subject to optimization: the compiler may use
12404 a more efficient model for symbols not visible outside of the translation
12405 unit, or if @option{-fpic} is not given on the command line.
12407 The default without @option{-fpic} is @samp{initial-exec}; with
12408 @option{-fpic} the default is @samp{global-dynamic}.
12410 @item -ftrampolines
12411 @opindex ftrampolines
12412 For targets that normally need trampolines for nested functions, always
12413 generate them instead of using descriptors. Otherwise, for targets that
12414 do not need them, like for example HP-PA or IA-64, do nothing.
12416 A trampoline is a small piece of code that is created at run time on the
12417 stack when the address of a nested function is taken, and is used to call
12418 the nested function indirectly. Therefore, it requires the stack to be
12419 made executable in order for the program to work properly.
12421 @option{-fno-trampolines} is enabled by default on a language by language
12422 basis to let the compiler avoid generating them, if it computes that this
12423 is safe, and replace them with descriptors. Descriptors are made up of data
12424 only, but the generated code must be prepared to deal with them. As of this
12425 writing, @option{-fno-trampolines} is enabled by default only for Ada.
12427 Moreover, code compiled with @option{-ftrampolines} and code compiled with
12428 @option{-fno-trampolines} are not binary compatible if nested functions are
12429 present. This option must therefore be used on a program-wide basis and be
12430 manipulated with extreme care.
12432 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
12433 @opindex fvisibility
12434 Set the default ELF image symbol visibility to the specified option---all
12435 symbols are marked with this unless overridden within the code.
12436 Using this feature can very substantially improve linking and
12437 load times of shared object libraries, produce more optimized
12438 code, provide near-perfect API export and prevent symbol clashes.
12439 It is @strong{strongly} recommended that you use this in any shared objects
12442 Despite the nomenclature, @samp{default} always means public; i.e.,
12443 available to be linked against from outside the shared object.
12444 @samp{protected} and @samp{internal} are pretty useless in real-world
12445 usage so the only other commonly used option is @samp{hidden}.
12446 The default if @option{-fvisibility} isn't specified is
12447 @samp{default}, i.e., make every symbol public.
12449 A good explanation of the benefits offered by ensuring ELF
12450 symbols have the correct visibility is given by ``How To Write
12451 Shared Libraries'' by Ulrich Drepper (which can be found at
12452 @w{@uref{http://www.akkadia.org/drepper/}})---however a superior
12453 solution made possible by this option to marking things hidden when
12454 the default is public is to make the default hidden and mark things
12455 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
12456 and @code{__attribute__ ((visibility("default")))} instead of
12457 @code{__declspec(dllexport)} you get almost identical semantics with
12458 identical syntax. This is a great boon to those working with
12459 cross-platform projects.
12461 For those adding visibility support to existing code, you may find
12462 @code{#pragma GCC visibility} of use. This works by you enclosing
12463 the declarations you wish to set visibility for with (for example)
12464 @code{#pragma GCC visibility push(hidden)} and
12465 @code{#pragma GCC visibility pop}.
12466 Bear in mind that symbol visibility should be viewed @strong{as
12467 part of the API interface contract} and thus all new code should
12468 always specify visibility when it is not the default; i.e., declarations
12469 only for use within the local DSO should @strong{always} be marked explicitly
12470 as hidden as so to avoid PLT indirection overheads---making this
12471 abundantly clear also aids readability and self-documentation of the code.
12472 Note that due to ISO C++ specification requirements, @code{operator new} and
12473 @code{operator delete} must always be of default visibility.
12475 Be aware that headers from outside your project, in particular system
12476 headers and headers from any other library you use, may not be
12477 expecting to be compiled with visibility other than the default. You
12478 may need to explicitly say @code{#pragma GCC visibility push(default)}
12479 before including any such headers.
12481 @code{extern} declarations are not affected by @option{-fvisibility}, so
12482 a lot of code can be recompiled with @option{-fvisibility=hidden} with
12483 no modifications. However, this means that calls to @code{extern}
12484 functions with no explicit visibility use the PLT, so it is more
12485 effective to use @code{__attribute ((visibility))} and/or
12486 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
12487 declarations should be treated as hidden.
12489 Note that @option{-fvisibility} does affect C++ vague linkage
12490 entities. This means that, for instance, an exception class that is
12491 be thrown between DSOs must be explicitly marked with default
12492 visibility so that the @samp{type_info} nodes are unified between
12495 An overview of these techniques, their benefits and how to use them
12496 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
12498 @item -fstrict-volatile-bitfields
12499 @opindex fstrict-volatile-bitfields
12500 This option should be used if accesses to volatile bit-fields (or other
12501 structure fields, although the compiler usually honors those types
12502 anyway) should use a single access of the width of the
12503 field's type, aligned to a natural alignment if possible. For
12504 example, targets with memory-mapped peripheral registers might require
12505 all such accesses to be 16 bits wide; with this flag you can
12506 declare all peripheral bit-fields as @code{unsigned short} (assuming short
12507 is 16 bits on these targets) to force GCC to use 16-bit accesses
12508 instead of, perhaps, a more efficient 32-bit access.
12510 If this option is disabled, the compiler uses the most efficient
12511 instruction. In the previous example, that might be a 32-bit load
12512 instruction, even though that accesses bytes that do not contain
12513 any portion of the bit-field, or memory-mapped registers unrelated to
12514 the one being updated.
12516 In some cases, such as when the @code{packed} attribute is applied to a
12517 structure field, it may not be possible to access the field with a single
12518 read or write that is correctly aligned for the target machine. In this
12519 case GCC falls back to generating multiple accesses rather than code that
12520 will fault or truncate the result at run time.
12522 Note: Due to restrictions of the C/C++11 memory model, write accesses are
12523 not allowed to touch non bit-field members. It is therefore recommended
12524 to define all bits of the field's type as bit-field members.
12526 The default value of this option is determined by the application binary
12527 interface for the target processor.
12529 @item -fsync-libcalls
12530 @opindex fsync-libcalls
12531 This option controls whether any out-of-line instance of the @code{__sync}
12532 family of functions may be used to implement the C++11 @code{__atomic}
12533 family of functions.
12535 The default value of this option is enabled, thus the only useful form
12536 of the option is @option{-fno-sync-libcalls}. This option is used in
12537 the implementation of the @file{libatomic} runtime library.
12541 @node Developer Options
12542 @section GCC Developer Options
12543 @cindex developer options
12544 @cindex debugging GCC
12545 @cindex debug dump options
12546 @cindex dump options
12547 @cindex compilation statistics
12549 This section describes command-line options that are primarily of
12550 interest to GCC developers, including options to support compiler
12551 testing and investigation of compiler bugs and compile-time
12552 performance problems. This includes options that produce debug dumps
12553 at various points in the compilation; that print statistics such as
12554 memory use and execution time; and that print information about GCC's
12555 configuration, such as where it searches for libraries. You should
12556 rarely need to use any of these options for ordinary compilation and
12561 @item -d@var{letters}
12562 @itemx -fdump-rtl-@var{pass}
12563 @itemx -fdump-rtl-@var{pass}=@var{filename}
12565 @opindex fdump-rtl-@var{pass}
12566 Says to make debugging dumps during compilation at times specified by
12567 @var{letters}. This is used for debugging the RTL-based passes of the
12568 compiler. The file names for most of the dumps are made by appending
12569 a pass number and a word to the @var{dumpname}, and the files are
12570 created in the directory of the output file. In case of
12571 @option{=@var{filename}} option, the dump is output on the given file
12572 instead of the pass numbered dump files. Note that the pass number is
12573 assigned as passes are registered into the pass manager. Most passes
12574 are registered in the order that they will execute and for these passes
12575 the number corresponds to the pass execution order. However, passes
12576 registered by plugins, passes specific to compilation targets, or
12577 passes that are otherwise registered after all the other passes are
12578 numbered higher than a pass named "final", even if they are executed
12579 earlier. @var{dumpname} is generated from the name of the output
12580 file if explicitly specified and not an executable, otherwise it is
12581 the basename of the source file. These switches may have different
12582 effects when @option{-E} is used for preprocessing.
12584 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
12585 @option{-d} option @var{letters}. Here are the possible
12586 letters for use in @var{pass} and @var{letters}, and their meanings:
12590 @item -fdump-rtl-alignments
12591 @opindex fdump-rtl-alignments
12592 Dump after branch alignments have been computed.
12594 @item -fdump-rtl-asmcons
12595 @opindex fdump-rtl-asmcons
12596 Dump after fixing rtl statements that have unsatisfied in/out constraints.
12598 @item -fdump-rtl-auto_inc_dec
12599 @opindex fdump-rtl-auto_inc_dec
12600 Dump after auto-inc-dec discovery. This pass is only run on
12601 architectures that have auto inc or auto dec instructions.
12603 @item -fdump-rtl-barriers
12604 @opindex fdump-rtl-barriers
12605 Dump after cleaning up the barrier instructions.
12607 @item -fdump-rtl-bbpart
12608 @opindex fdump-rtl-bbpart
12609 Dump after partitioning hot and cold basic blocks.
12611 @item -fdump-rtl-bbro
12612 @opindex fdump-rtl-bbro
12613 Dump after block reordering.
12615 @item -fdump-rtl-btl1
12616 @itemx -fdump-rtl-btl2
12617 @opindex fdump-rtl-btl2
12618 @opindex fdump-rtl-btl2
12619 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
12620 after the two branch
12621 target load optimization passes.
12623 @item -fdump-rtl-bypass
12624 @opindex fdump-rtl-bypass
12625 Dump after jump bypassing and control flow optimizations.
12627 @item -fdump-rtl-combine
12628 @opindex fdump-rtl-combine
12629 Dump after the RTL instruction combination pass.
12631 @item -fdump-rtl-compgotos
12632 @opindex fdump-rtl-compgotos
12633 Dump after duplicating the computed gotos.
12635 @item -fdump-rtl-ce1
12636 @itemx -fdump-rtl-ce2
12637 @itemx -fdump-rtl-ce3
12638 @opindex fdump-rtl-ce1
12639 @opindex fdump-rtl-ce2
12640 @opindex fdump-rtl-ce3
12641 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
12642 @option{-fdump-rtl-ce3} enable dumping after the three
12643 if conversion passes.
12645 @item -fdump-rtl-cprop_hardreg
12646 @opindex fdump-rtl-cprop_hardreg
12647 Dump after hard register copy propagation.
12649 @item -fdump-rtl-csa
12650 @opindex fdump-rtl-csa
12651 Dump after combining stack adjustments.
12653 @item -fdump-rtl-cse1
12654 @itemx -fdump-rtl-cse2
12655 @opindex fdump-rtl-cse1
12656 @opindex fdump-rtl-cse2
12657 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
12658 the two common subexpression elimination passes.
12660 @item -fdump-rtl-dce
12661 @opindex fdump-rtl-dce
12662 Dump after the standalone dead code elimination passes.
12664 @item -fdump-rtl-dbr
12665 @opindex fdump-rtl-dbr
12666 Dump after delayed branch scheduling.
12668 @item -fdump-rtl-dce1
12669 @itemx -fdump-rtl-dce2
12670 @opindex fdump-rtl-dce1
12671 @opindex fdump-rtl-dce2
12672 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
12673 the two dead store elimination passes.
12675 @item -fdump-rtl-eh
12676 @opindex fdump-rtl-eh
12677 Dump after finalization of EH handling code.
12679 @item -fdump-rtl-eh_ranges
12680 @opindex fdump-rtl-eh_ranges
12681 Dump after conversion of EH handling range regions.
12683 @item -fdump-rtl-expand
12684 @opindex fdump-rtl-expand
12685 Dump after RTL generation.
12687 @item -fdump-rtl-fwprop1
12688 @itemx -fdump-rtl-fwprop2
12689 @opindex fdump-rtl-fwprop1
12690 @opindex fdump-rtl-fwprop2
12691 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
12692 dumping after the two forward propagation passes.
12694 @item -fdump-rtl-gcse1
12695 @itemx -fdump-rtl-gcse2
12696 @opindex fdump-rtl-gcse1
12697 @opindex fdump-rtl-gcse2
12698 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
12699 after global common subexpression elimination.
12701 @item -fdump-rtl-init-regs
12702 @opindex fdump-rtl-init-regs
12703 Dump after the initialization of the registers.
12705 @item -fdump-rtl-initvals
12706 @opindex fdump-rtl-initvals
12707 Dump after the computation of the initial value sets.
12709 @item -fdump-rtl-into_cfglayout
12710 @opindex fdump-rtl-into_cfglayout
12711 Dump after converting to cfglayout mode.
12713 @item -fdump-rtl-ira
12714 @opindex fdump-rtl-ira
12715 Dump after iterated register allocation.
12717 @item -fdump-rtl-jump
12718 @opindex fdump-rtl-jump
12719 Dump after the second jump optimization.
12721 @item -fdump-rtl-loop2
12722 @opindex fdump-rtl-loop2
12723 @option{-fdump-rtl-loop2} enables dumping after the rtl
12724 loop optimization passes.
12726 @item -fdump-rtl-mach
12727 @opindex fdump-rtl-mach
12728 Dump after performing the machine dependent reorganization pass, if that
12731 @item -fdump-rtl-mode_sw
12732 @opindex fdump-rtl-mode_sw
12733 Dump after removing redundant mode switches.
12735 @item -fdump-rtl-rnreg
12736 @opindex fdump-rtl-rnreg
12737 Dump after register renumbering.
12739 @item -fdump-rtl-outof_cfglayout
12740 @opindex fdump-rtl-outof_cfglayout
12741 Dump after converting from cfglayout mode.
12743 @item -fdump-rtl-peephole2
12744 @opindex fdump-rtl-peephole2
12745 Dump after the peephole pass.
12747 @item -fdump-rtl-postreload
12748 @opindex fdump-rtl-postreload
12749 Dump after post-reload optimizations.
12751 @item -fdump-rtl-pro_and_epilogue
12752 @opindex fdump-rtl-pro_and_epilogue
12753 Dump after generating the function prologues and epilogues.
12755 @item -fdump-rtl-sched1
12756 @itemx -fdump-rtl-sched2
12757 @opindex fdump-rtl-sched1
12758 @opindex fdump-rtl-sched2
12759 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
12760 after the basic block scheduling passes.
12762 @item -fdump-rtl-ree
12763 @opindex fdump-rtl-ree
12764 Dump after sign/zero extension elimination.
12766 @item -fdump-rtl-seqabstr
12767 @opindex fdump-rtl-seqabstr
12768 Dump after common sequence discovery.
12770 @item -fdump-rtl-shorten
12771 @opindex fdump-rtl-shorten
12772 Dump after shortening branches.
12774 @item -fdump-rtl-sibling
12775 @opindex fdump-rtl-sibling
12776 Dump after sibling call optimizations.
12778 @item -fdump-rtl-split1
12779 @itemx -fdump-rtl-split2
12780 @itemx -fdump-rtl-split3
12781 @itemx -fdump-rtl-split4
12782 @itemx -fdump-rtl-split5
12783 @opindex fdump-rtl-split1
12784 @opindex fdump-rtl-split2
12785 @opindex fdump-rtl-split3
12786 @opindex fdump-rtl-split4
12787 @opindex fdump-rtl-split5
12788 These options enable dumping after five rounds of
12789 instruction splitting.
12791 @item -fdump-rtl-sms
12792 @opindex fdump-rtl-sms
12793 Dump after modulo scheduling. This pass is only run on some
12796 @item -fdump-rtl-stack
12797 @opindex fdump-rtl-stack
12798 Dump after conversion from GCC's ``flat register file'' registers to the
12799 x87's stack-like registers. This pass is only run on x86 variants.
12801 @item -fdump-rtl-subreg1
12802 @itemx -fdump-rtl-subreg2
12803 @opindex fdump-rtl-subreg1
12804 @opindex fdump-rtl-subreg2
12805 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
12806 the two subreg expansion passes.
12808 @item -fdump-rtl-unshare
12809 @opindex fdump-rtl-unshare
12810 Dump after all rtl has been unshared.
12812 @item -fdump-rtl-vartrack
12813 @opindex fdump-rtl-vartrack
12814 Dump after variable tracking.
12816 @item -fdump-rtl-vregs
12817 @opindex fdump-rtl-vregs
12818 Dump after converting virtual registers to hard registers.
12820 @item -fdump-rtl-web
12821 @opindex fdump-rtl-web
12822 Dump after live range splitting.
12824 @item -fdump-rtl-regclass
12825 @itemx -fdump-rtl-subregs_of_mode_init
12826 @itemx -fdump-rtl-subregs_of_mode_finish
12827 @itemx -fdump-rtl-dfinit
12828 @itemx -fdump-rtl-dfinish
12829 @opindex fdump-rtl-regclass
12830 @opindex fdump-rtl-subregs_of_mode_init
12831 @opindex fdump-rtl-subregs_of_mode_finish
12832 @opindex fdump-rtl-dfinit
12833 @opindex fdump-rtl-dfinish
12834 These dumps are defined but always produce empty files.
12837 @itemx -fdump-rtl-all
12839 @opindex fdump-rtl-all
12840 Produce all the dumps listed above.
12844 Annotate the assembler output with miscellaneous debugging information.
12848 Dump all macro definitions, at the end of preprocessing, in addition to
12853 Produce a core dump whenever an error occurs.
12857 Annotate the assembler output with a comment indicating which
12858 pattern and alternative is used. The length of each instruction is
12863 Dump the RTL in the assembler output as a comment before each instruction.
12864 Also turns on @option{-dp} annotation.
12868 Just generate RTL for a function instead of compiling it. Usually used
12869 with @option{-fdump-rtl-expand}.
12872 @item -fdump-noaddr
12873 @opindex fdump-noaddr
12874 When doing debugging dumps, suppress address output. This makes it more
12875 feasible to use diff on debugging dumps for compiler invocations with
12876 different compiler binaries and/or different
12877 text / bss / data / heap / stack / dso start locations.
12880 @opindex freport-bug
12881 Collect and dump debug information into a temporary file if an
12882 internal compiler error (ICE) occurs.
12884 @item -fdump-unnumbered
12885 @opindex fdump-unnumbered
12886 When doing debugging dumps, suppress instruction numbers and address output.
12887 This makes it more feasible to use diff on debugging dumps for compiler
12888 invocations with different options, in particular with and without
12891 @item -fdump-unnumbered-links
12892 @opindex fdump-unnumbered-links
12893 When doing debugging dumps (see @option{-d} option above), suppress
12894 instruction numbers for the links to the previous and next instructions
12897 @item -fdump-translation-unit @r{(C++ only)}
12898 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
12899 @opindex fdump-translation-unit
12900 Dump a representation of the tree structure for the entire translation
12901 unit to a file. The file name is made by appending @file{.tu} to the
12902 source file name, and the file is created in the same directory as the
12903 output file. If the @samp{-@var{options}} form is used, @var{options}
12904 controls the details of the dump as described for the
12905 @option{-fdump-tree} options.
12907 @item -fdump-class-hierarchy @r{(C++ only)}
12908 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
12909 @opindex fdump-class-hierarchy
12910 Dump a representation of each class's hierarchy and virtual function
12911 table layout to a file. The file name is made by appending
12912 @file{.class} to the source file name, and the file is created in the
12913 same directory as the output file. If the @samp{-@var{options}} form
12914 is used, @var{options} controls the details of the dump as described
12915 for the @option{-fdump-tree} options.
12917 @item -fdump-ipa-@var{switch}
12919 Control the dumping at various stages of inter-procedural analysis
12920 language tree to a file. The file name is generated by appending a
12921 switch specific suffix to the source file name, and the file is created
12922 in the same directory as the output file. The following dumps are
12927 Enables all inter-procedural analysis dumps.
12930 Dumps information about call-graph optimization, unused function removal,
12931 and inlining decisions.
12934 Dump after function inlining.
12938 @item -fdump-passes
12939 @opindex fdump-passes
12940 Dump the list of optimization passes that are turned on and off by
12941 the current command-line options.
12943 @item -fdump-statistics-@var{option}
12944 @opindex fdump-statistics
12945 Enable and control dumping of pass statistics in a separate file. The
12946 file name is generated by appending a suffix ending in
12947 @samp{.statistics} to the source file name, and the file is created in
12948 the same directory as the output file. If the @samp{-@var{option}}
12949 form is used, @samp{-stats} causes counters to be summed over the
12950 whole compilation unit while @samp{-details} dumps every event as
12951 the passes generate them. The default with no option is to sum
12952 counters for each function compiled.
12954 @item -fdump-tree-@var{switch}
12955 @itemx -fdump-tree-@var{switch}-@var{options}
12956 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
12957 @opindex fdump-tree
12958 Control the dumping at various stages of processing the intermediate
12959 language tree to a file. The file name is generated by appending a
12960 switch-specific suffix to the source file name, and the file is
12961 created in the same directory as the output file. In case of
12962 @option{=@var{filename}} option, the dump is output on the given file
12963 instead of the auto named dump files. If the @samp{-@var{options}}
12964 form is used, @var{options} is a list of @samp{-} separated options
12965 which control the details of the dump. Not all options are applicable
12966 to all dumps; those that are not meaningful are ignored. The
12967 following options are available
12971 Print the address of each node. Usually this is not meaningful as it
12972 changes according to the environment and source file. Its primary use
12973 is for tying up a dump file with a debug environment.
12975 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
12976 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
12977 use working backward from mangled names in the assembly file.
12979 When dumping front-end intermediate representations, inhibit dumping
12980 of members of a scope or body of a function merely because that scope
12981 has been reached. Only dump such items when they are directly reachable
12982 by some other path.
12984 When dumping pretty-printed trees, this option inhibits dumping the
12985 bodies of control structures.
12987 When dumping RTL, print the RTL in slim (condensed) form instead of
12988 the default LISP-like representation.
12990 Print a raw representation of the tree. By default, trees are
12991 pretty-printed into a C-like representation.
12993 Enable more detailed dumps (not honored by every dump option). Also
12994 include information from the optimization passes.
12996 Enable dumping various statistics about the pass (not honored by every dump
12999 Enable showing basic block boundaries (disabled in raw dumps).
13001 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
13002 dump a representation of the control flow graph suitable for viewing with
13003 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
13004 the file is pretty-printed as a subgraph, so that GraphViz can render them
13005 all in a single plot.
13007 This option currently only works for RTL dumps, and the RTL is always
13008 dumped in slim form.
13010 Enable showing virtual operands for every statement.
13012 Enable showing line numbers for statements.
13014 Enable showing the unique ID (@code{DECL_UID}) for each variable.
13016 Enable showing the tree dump for each statement.
13018 Enable showing the EH region number holding each statement.
13020 Enable showing scalar evolution analysis details.
13022 Enable showing optimization information (only available in certain
13025 Enable showing missed optimization information (only available in certain
13028 Enable other detailed optimization information (only available in
13030 @item =@var{filename}
13031 Instead of an auto named dump file, output into the given file
13032 name. The file names @file{stdout} and @file{stderr} are treated
13033 specially and are considered already open standard streams. For
13037 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
13038 -fdump-tree-pre=stderr file.c
13041 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
13042 output on to @file{stderr}. If two conflicting dump filenames are
13043 given for the same pass, then the latter option overrides the earlier
13047 @opindex fdump-tree-split-paths
13048 Dump each function after splitting paths to loop backedges. The file
13049 name is made by appending @file{.split-paths} to the source file name.
13052 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
13053 and @option{lineno}.
13056 Turn on all optimization options, i.e., @option{optimized},
13057 @option{missed}, and @option{note}.
13060 The following tree dumps are possible:
13064 @opindex fdump-tree-original
13065 Dump before any tree based optimization, to @file{@var{file}.original}.
13068 @opindex fdump-tree-optimized
13069 Dump after all tree based optimization, to @file{@var{file}.optimized}.
13072 @opindex fdump-tree-gimple
13073 Dump each function before and after the gimplification pass to a file. The
13074 file name is made by appending @file{.gimple} to the source file name.
13077 @opindex fdump-tree-cfg
13078 Dump the control flow graph of each function to a file. The file name is
13079 made by appending @file{.cfg} to the source file name.
13082 @opindex fdump-tree-ch
13083 Dump each function after copying loop headers. The file name is made by
13084 appending @file{.ch} to the source file name.
13087 @opindex fdump-tree-ssa
13088 Dump SSA related information to a file. The file name is made by appending
13089 @file{.ssa} to the source file name.
13092 @opindex fdump-tree-alias
13093 Dump aliasing information for each function. The file name is made by
13094 appending @file{.alias} to the source file name.
13097 @opindex fdump-tree-ccp
13098 Dump each function after CCP@. The file name is made by appending
13099 @file{.ccp} to the source file name.
13102 @opindex fdump-tree-storeccp
13103 Dump each function after STORE-CCP@. The file name is made by appending
13104 @file{.storeccp} to the source file name.
13107 @opindex fdump-tree-pre
13108 Dump trees after partial redundancy elimination and/or code hoisting.
13109 The file name is made by appending @file{.pre} to the source file name.
13112 @opindex fdump-tree-fre
13113 Dump trees after full redundancy elimination. The file name is made
13114 by appending @file{.fre} to the source file name.
13117 @opindex fdump-tree-copyprop
13118 Dump trees after copy propagation. The file name is made
13119 by appending @file{.copyprop} to the source file name.
13121 @item store_copyprop
13122 @opindex fdump-tree-store_copyprop
13123 Dump trees after store copy-propagation. The file name is made
13124 by appending @file{.store_copyprop} to the source file name.
13127 @opindex fdump-tree-dce
13128 Dump each function after dead code elimination. The file name is made by
13129 appending @file{.dce} to the source file name.
13132 @opindex fdump-tree-sra
13133 Dump each function after performing scalar replacement of aggregates. The
13134 file name is made by appending @file{.sra} to the source file name.
13137 @opindex fdump-tree-sink
13138 Dump each function after performing code sinking. The file name is made
13139 by appending @file{.sink} to the source file name.
13142 @opindex fdump-tree-dom
13143 Dump each function after applying dominator tree optimizations. The file
13144 name is made by appending @file{.dom} to the source file name.
13147 @opindex fdump-tree-dse
13148 Dump each function after applying dead store elimination. The file
13149 name is made by appending @file{.dse} to the source file name.
13152 @opindex fdump-tree-phiopt
13153 Dump each function after optimizing PHI nodes into straightline code. The file
13154 name is made by appending @file{.phiopt} to the source file name.
13157 @opindex fdump-tree-backprop
13158 Dump each function after back-propagating use information up the definition
13159 chain. The file name is made by appending @file{.backprop} to the
13163 @opindex fdump-tree-forwprop
13164 Dump each function after forward propagating single use variables. The file
13165 name is made by appending @file{.forwprop} to the source file name.
13168 @opindex fdump-tree-nrv
13169 Dump each function after applying the named return value optimization on
13170 generic trees. The file name is made by appending @file{.nrv} to the source
13174 @opindex fdump-tree-vect
13175 Dump each function after applying vectorization of loops. The file name is
13176 made by appending @file{.vect} to the source file name.
13179 @opindex fdump-tree-slp
13180 Dump each function after applying vectorization of basic blocks. The file name
13181 is made by appending @file{.slp} to the source file name.
13184 @opindex fdump-tree-vrp
13185 Dump each function after Value Range Propagation (VRP). The file name
13186 is made by appending @file{.vrp} to the source file name.
13189 @opindex fdump-tree-evrp
13190 Dump each function after Early Value Range Propagation (EVRP). The file name
13191 is made by appending @file{.evrp} to the source file name.
13194 @opindex fdump-tree-oaccdevlow
13195 Dump each function after applying device-specific OpenACC transformations.
13196 The file name is made by appending @file{.oaccdevlow} to the source file name.
13199 @opindex fdump-tree-all
13200 Enable all the available tree dumps with the flags provided in this option.
13204 @itemx -fopt-info-@var{options}
13205 @itemx -fopt-info-@var{options}=@var{filename}
13207 Controls optimization dumps from various optimization passes. If the
13208 @samp{-@var{options}} form is used, @var{options} is a list of
13209 @samp{-} separated option keywords to select the dump details and
13212 The @var{options} can be divided into two groups: options describing the
13213 verbosity of the dump, and options describing which optimizations
13214 should be included. The options from both the groups can be freely
13215 mixed as they are non-overlapping. However, in case of any conflicts,
13216 the later options override the earlier options on the command
13219 The following options control the dump verbosity:
13223 Print information when an optimization is successfully applied. It is
13224 up to a pass to decide which information is relevant. For example, the
13225 vectorizer passes print the source location of loops which are
13226 successfully vectorized.
13228 Print information about missed optimizations. Individual passes
13229 control which information to include in the output.
13231 Print verbose information about optimizations, such as certain
13232 transformations, more detailed messages about decisions etc.
13234 Print detailed optimization information. This includes
13235 @samp{optimized}, @samp{missed}, and @samp{note}.
13238 One or more of the following option keywords can be used to describe a
13239 group of optimizations:
13243 Enable dumps from all interprocedural optimizations.
13245 Enable dumps from all loop optimizations.
13247 Enable dumps from all inlining optimizations.
13249 Enable dumps from all vectorization optimizations.
13251 Enable dumps from all optimizations. This is a superset of
13252 the optimization groups listed above.
13255 If @var{options} is
13256 omitted, it defaults to @samp{optimized-optall}, which means to dump all
13257 info about successful optimizations from all the passes.
13259 If the @var{filename} is provided, then the dumps from all the
13260 applicable optimizations are concatenated into the @var{filename}.
13261 Otherwise the dump is output onto @file{stderr}. Though multiple
13262 @option{-fopt-info} options are accepted, only one of them can include
13263 a @var{filename}. If other filenames are provided then all but the
13264 first such option are ignored.
13266 Note that the output @var{filename} is overwritten
13267 in case of multiple translation units. If a combined output from
13268 multiple translation units is desired, @file{stderr} should be used
13271 In the following example, the optimization info is output to
13280 gcc -O3 -fopt-info-missed=missed.all
13284 outputs missed optimization report from all the passes into
13285 @file{missed.all}, and this one:
13288 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
13292 prints information about missed optimization opportunities from
13293 vectorization passes on @file{stderr}.
13294 Note that @option{-fopt-info-vec-missed} is equivalent to
13295 @option{-fopt-info-missed-vec}.
13297 As another example,
13299 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
13303 outputs information about missed optimizations as well as
13304 optimized locations from all the inlining passes into
13310 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
13314 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
13315 in conflict since only one output file is allowed. In this case, only
13316 the first option takes effect and the subsequent options are
13317 ignored. Thus only @file{vec.miss} is produced which contains
13318 dumps from the vectorizer about missed opportunities.
13320 @item -fsched-verbose=@var{n}
13321 @opindex fsched-verbose
13322 On targets that use instruction scheduling, this option controls the
13323 amount of debugging output the scheduler prints to the dump files.
13325 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
13326 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
13327 For @var{n} greater than one, it also output basic block probabilities,
13328 detailed ready list information and unit/insn info. For @var{n} greater
13329 than two, it includes RTL at abort point, control-flow and regions info.
13330 And for @var{n} over four, @option{-fsched-verbose} also includes
13335 @item -fenable-@var{kind}-@var{pass}
13336 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
13340 This is a set of options that are used to explicitly disable/enable
13341 optimization passes. These options are intended for use for debugging GCC.
13342 Compiler users should use regular options for enabling/disabling
13347 @item -fdisable-ipa-@var{pass}
13348 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
13349 statically invoked in the compiler multiple times, the pass name should be
13350 appended with a sequential number starting from 1.
13352 @item -fdisable-rtl-@var{pass}
13353 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
13354 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
13355 statically invoked in the compiler multiple times, the pass name should be
13356 appended with a sequential number starting from 1. @var{range-list} is a
13357 comma-separated list of function ranges or assembler names. Each range is a number
13358 pair separated by a colon. The range is inclusive in both ends. If the range
13359 is trivial, the number pair can be simplified as a single number. If the
13360 function's call graph node's @var{uid} falls within one of the specified ranges,
13361 the @var{pass} is disabled for that function. The @var{uid} is shown in the
13362 function header of a dump file, and the pass names can be dumped by using
13363 option @option{-fdump-passes}.
13365 @item -fdisable-tree-@var{pass}
13366 @itemx -fdisable-tree-@var{pass}=@var{range-list}
13367 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
13370 @item -fenable-ipa-@var{pass}
13371 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
13372 statically invoked in the compiler multiple times, the pass name should be
13373 appended with a sequential number starting from 1.
13375 @item -fenable-rtl-@var{pass}
13376 @itemx -fenable-rtl-@var{pass}=@var{range-list}
13377 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
13378 description and examples.
13380 @item -fenable-tree-@var{pass}
13381 @itemx -fenable-tree-@var{pass}=@var{range-list}
13382 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
13383 of option arguments.
13387 Here are some examples showing uses of these options.
13391 # disable ccp1 for all functions
13392 -fdisable-tree-ccp1
13393 # disable complete unroll for function whose cgraph node uid is 1
13394 -fenable-tree-cunroll=1
13395 # disable gcse2 for functions at the following ranges [1,1],
13396 # [300,400], and [400,1000]
13397 # disable gcse2 for functions foo and foo2
13398 -fdisable-rtl-gcse2=foo,foo2
13399 # disable early inlining
13400 -fdisable-tree-einline
13401 # disable ipa inlining
13402 -fdisable-ipa-inline
13403 # enable tree full unroll
13404 -fenable-tree-unroll
13409 @itemx -fchecking=@var{n}
13411 @opindex fno-checking
13412 Enable internal consistency checking. The default depends on
13413 the compiler configuration. @option{-fchecking=2} enables further
13414 internal consistency checking that might affect code generation.
13416 @item -frandom-seed=@var{string}
13417 @opindex frandom-seed
13418 This option provides a seed that GCC uses in place of
13419 random numbers in generating certain symbol names
13420 that have to be different in every compiled file. It is also used to
13421 place unique stamps in coverage data files and the object files that
13422 produce them. You can use the @option{-frandom-seed} option to produce
13423 reproducibly identical object files.
13425 The @var{string} can either be a number (decimal, octal or hex) or an
13426 arbitrary string (in which case it's converted to a number by
13429 The @var{string} should be different for every file you compile.
13432 @itemx -save-temps=cwd
13433 @opindex save-temps
13434 Store the usual ``temporary'' intermediate files permanently; place them
13435 in the current directory and name them based on the source file. Thus,
13436 compiling @file{foo.c} with @option{-c -save-temps} produces files
13437 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
13438 preprocessed @file{foo.i} output file even though the compiler now
13439 normally uses an integrated preprocessor.
13441 When used in combination with the @option{-x} command-line option,
13442 @option{-save-temps} is sensible enough to avoid over writing an
13443 input source file with the same extension as an intermediate file.
13444 The corresponding intermediate file may be obtained by renaming the
13445 source file before using @option{-save-temps}.
13447 If you invoke GCC in parallel, compiling several different source
13448 files that share a common base name in different subdirectories or the
13449 same source file compiled for multiple output destinations, it is
13450 likely that the different parallel compilers will interfere with each
13451 other, and overwrite the temporary files. For instance:
13454 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
13455 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
13458 may result in @file{foo.i} and @file{foo.o} being written to
13459 simultaneously by both compilers.
13461 @item -save-temps=obj
13462 @opindex save-temps=obj
13463 Store the usual ``temporary'' intermediate files permanently. If the
13464 @option{-o} option is used, the temporary files are based on the
13465 object file. If the @option{-o} option is not used, the
13466 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
13471 gcc -save-temps=obj -c foo.c
13472 gcc -save-temps=obj -c bar.c -o dir/xbar.o
13473 gcc -save-temps=obj foobar.c -o dir2/yfoobar
13477 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
13478 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
13479 @file{dir2/yfoobar.o}.
13481 @item -time@r{[}=@var{file}@r{]}
13483 Report the CPU time taken by each subprocess in the compilation
13484 sequence. For C source files, this is the compiler proper and assembler
13485 (plus the linker if linking is done).
13487 Without the specification of an output file, the output looks like this:
13494 The first number on each line is the ``user time'', that is time spent
13495 executing the program itself. The second number is ``system time'',
13496 time spent executing operating system routines on behalf of the program.
13497 Both numbers are in seconds.
13499 With the specification of an output file, the output is appended to the
13500 named file, and it looks like this:
13503 0.12 0.01 cc1 @var{options}
13504 0.00 0.01 as @var{options}
13507 The ``user time'' and the ``system time'' are moved before the program
13508 name, and the options passed to the program are displayed, so that one
13509 can later tell what file was being compiled, and with which options.
13511 @item -fdump-final-insns@r{[}=@var{file}@r{]}
13512 @opindex fdump-final-insns
13513 Dump the final internal representation (RTL) to @var{file}. If the
13514 optional argument is omitted (or if @var{file} is @code{.}), the name
13515 of the dump file is determined by appending @code{.gkd} to the
13516 compilation output file name.
13518 @item -fcompare-debug@r{[}=@var{opts}@r{]}
13519 @opindex fcompare-debug
13520 @opindex fno-compare-debug
13521 If no error occurs during compilation, run the compiler a second time,
13522 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
13523 passed to the second compilation. Dump the final internal
13524 representation in both compilations, and print an error if they differ.
13526 If the equal sign is omitted, the default @option{-gtoggle} is used.
13528 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
13529 and nonzero, implicitly enables @option{-fcompare-debug}. If
13530 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
13531 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
13534 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
13535 is equivalent to @option{-fno-compare-debug}, which disables the dumping
13536 of the final representation and the second compilation, preventing even
13537 @env{GCC_COMPARE_DEBUG} from taking effect.
13539 To verify full coverage during @option{-fcompare-debug} testing, set
13540 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
13541 which GCC rejects as an invalid option in any actual compilation
13542 (rather than preprocessing, assembly or linking). To get just a
13543 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
13544 not overridden} will do.
13546 @item -fcompare-debug-second
13547 @opindex fcompare-debug-second
13548 This option is implicitly passed to the compiler for the second
13549 compilation requested by @option{-fcompare-debug}, along with options to
13550 silence warnings, and omitting other options that would cause
13551 side-effect compiler outputs to files or to the standard output. Dump
13552 files and preserved temporary files are renamed so as to contain the
13553 @code{.gk} additional extension during the second compilation, to avoid
13554 overwriting those generated by the first.
13556 When this option is passed to the compiler driver, it causes the
13557 @emph{first} compilation to be skipped, which makes it useful for little
13558 other than debugging the compiler proper.
13562 Turn off generation of debug info, if leaving out this option
13563 generates it, or turn it on at level 2 otherwise. The position of this
13564 argument in the command line does not matter; it takes effect after all
13565 other options are processed, and it does so only once, no matter how
13566 many times it is given. This is mainly intended to be used with
13567 @option{-fcompare-debug}.
13569 @item -fvar-tracking-assignments-toggle
13570 @opindex fvar-tracking-assignments-toggle
13571 @opindex fno-var-tracking-assignments-toggle
13572 Toggle @option{-fvar-tracking-assignments}, in the same way that
13573 @option{-gtoggle} toggles @option{-g}.
13577 Makes the compiler print out each function name as it is compiled, and
13578 print some statistics about each pass when it finishes.
13580 @item -ftime-report
13581 @opindex ftime-report
13582 Makes the compiler print some statistics about the time consumed by each
13583 pass when it finishes.
13585 @item -ftime-report-details
13586 @opindex ftime-report-details
13587 Record the time consumed by infrastructure parts separately for each pass.
13589 @item -fira-verbose=@var{n}
13590 @opindex fira-verbose
13591 Control the verbosity of the dump file for the integrated register allocator.
13592 The default value is 5. If the value @var{n} is greater or equal to 10,
13593 the dump output is sent to stderr using the same format as @var{n} minus 10.
13596 @opindex flto-report
13597 Prints a report with internal details on the workings of the link-time
13598 optimizer. The contents of this report vary from version to version.
13599 It is meant to be useful to GCC developers when processing object
13600 files in LTO mode (via @option{-flto}).
13602 Disabled by default.
13604 @item -flto-report-wpa
13605 @opindex flto-report-wpa
13606 Like @option{-flto-report}, but only print for the WPA phase of Link
13610 @opindex fmem-report
13611 Makes the compiler print some statistics about permanent memory
13612 allocation when it finishes.
13614 @item -fmem-report-wpa
13615 @opindex fmem-report-wpa
13616 Makes the compiler print some statistics about permanent memory
13617 allocation for the WPA phase only.
13619 @item -fpre-ipa-mem-report
13620 @opindex fpre-ipa-mem-report
13621 @item -fpost-ipa-mem-report
13622 @opindex fpost-ipa-mem-report
13623 Makes the compiler print some statistics about permanent memory
13624 allocation before or after interprocedural optimization.
13626 @item -fprofile-report
13627 @opindex fprofile-report
13628 Makes the compiler print some statistics about consistency of the
13629 (estimated) profile and effect of individual passes.
13631 @item -fstack-usage
13632 @opindex fstack-usage
13633 Makes the compiler output stack usage information for the program, on a
13634 per-function basis. The filename for the dump is made by appending
13635 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
13636 the output file, if explicitly specified and it is not an executable,
13637 otherwise it is the basename of the source file. An entry is made up
13642 The name of the function.
13646 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
13649 The qualifier @code{static} means that the function manipulates the stack
13650 statically: a fixed number of bytes are allocated for the frame on function
13651 entry and released on function exit; no stack adjustments are otherwise made
13652 in the function. The second field is this fixed number of bytes.
13654 The qualifier @code{dynamic} means that the function manipulates the stack
13655 dynamically: in addition to the static allocation described above, stack
13656 adjustments are made in the body of the function, for example to push/pop
13657 arguments around function calls. If the qualifier @code{bounded} is also
13658 present, the amount of these adjustments is bounded at compile time and
13659 the second field is an upper bound of the total amount of stack used by
13660 the function. If it is not present, the amount of these adjustments is
13661 not bounded at compile time and the second field only represents the
13666 Emit statistics about front-end processing at the end of the compilation.
13667 This option is supported only by the C++ front end, and
13668 the information is generally only useful to the G++ development team.
13670 @item -fdbg-cnt-list
13671 @opindex fdbg-cnt-list
13672 Print the name and the counter upper bound for all debug counters.
13675 @item -fdbg-cnt=@var{counter-value-list}
13677 Set the internal debug counter upper bound. @var{counter-value-list}
13678 is a comma-separated list of @var{name}:@var{value} pairs
13679 which sets the upper bound of each debug counter @var{name} to @var{value}.
13680 All debug counters have the initial upper bound of @code{UINT_MAX};
13681 thus @code{dbg_cnt} returns true always unless the upper bound
13682 is set by this option.
13683 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
13684 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
13686 @item -print-file-name=@var{library}
13687 @opindex print-file-name
13688 Print the full absolute name of the library file @var{library} that
13689 would be used when linking---and don't do anything else. With this
13690 option, GCC does not compile or link anything; it just prints the
13693 @item -print-multi-directory
13694 @opindex print-multi-directory
13695 Print the directory name corresponding to the multilib selected by any
13696 other switches present in the command line. This directory is supposed
13697 to exist in @env{GCC_EXEC_PREFIX}.
13699 @item -print-multi-lib
13700 @opindex print-multi-lib
13701 Print the mapping from multilib directory names to compiler switches
13702 that enable them. The directory name is separated from the switches by
13703 @samp{;}, and each switch starts with an @samp{@@} instead of the
13704 @samp{-}, without spaces between multiple switches. This is supposed to
13705 ease shell processing.
13707 @item -print-multi-os-directory
13708 @opindex print-multi-os-directory
13709 Print the path to OS libraries for the selected
13710 multilib, relative to some @file{lib} subdirectory. If OS libraries are
13711 present in the @file{lib} subdirectory and no multilibs are used, this is
13712 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
13713 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
13714 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
13715 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
13717 @item -print-multiarch
13718 @opindex print-multiarch
13719 Print the path to OS libraries for the selected multiarch,
13720 relative to some @file{lib} subdirectory.
13722 @item -print-prog-name=@var{program}
13723 @opindex print-prog-name
13724 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
13726 @item -print-libgcc-file-name
13727 @opindex print-libgcc-file-name
13728 Same as @option{-print-file-name=libgcc.a}.
13730 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
13731 but you do want to link with @file{libgcc.a}. You can do:
13734 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
13737 @item -print-search-dirs
13738 @opindex print-search-dirs
13739 Print the name of the configured installation directory and a list of
13740 program and library directories @command{gcc} searches---and don't do anything else.
13742 This is useful when @command{gcc} prints the error message
13743 @samp{installation problem, cannot exec cpp0: No such file or directory}.
13744 To resolve this you either need to put @file{cpp0} and the other compiler
13745 components where @command{gcc} expects to find them, or you can set the environment
13746 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
13747 Don't forget the trailing @samp{/}.
13748 @xref{Environment Variables}.
13750 @item -print-sysroot
13751 @opindex print-sysroot
13752 Print the target sysroot directory that is used during
13753 compilation. This is the target sysroot specified either at configure
13754 time or using the @option{--sysroot} option, possibly with an extra
13755 suffix that depends on compilation options. If no target sysroot is
13756 specified, the option prints nothing.
13758 @item -print-sysroot-headers-suffix
13759 @opindex print-sysroot-headers-suffix
13760 Print the suffix added to the target sysroot when searching for
13761 headers, or give an error if the compiler is not configured with such
13762 a suffix---and don't do anything else.
13765 @opindex dumpmachine
13766 Print the compiler's target machine (for example,
13767 @samp{i686-pc-linux-gnu})---and don't do anything else.
13770 @opindex dumpversion
13771 Print the compiler version (for example, @code{3.0})---and don't do
13776 Print the compiler's built-in specs---and don't do anything else. (This
13777 is used when GCC itself is being built.) @xref{Spec Files}.
13780 @node Submodel Options
13781 @section Machine-Dependent Options
13782 @cindex submodel options
13783 @cindex specifying hardware config
13784 @cindex hardware models and configurations, specifying
13785 @cindex target-dependent options
13786 @cindex machine-dependent options
13788 Each target machine supported by GCC can have its own options---for
13789 example, to allow you to compile for a particular processor variant or
13790 ABI, or to control optimizations specific to that machine. By
13791 convention, the names of machine-specific options start with
13794 Some configurations of the compiler also support additional target-specific
13795 options, usually for compatibility with other compilers on the same
13798 @c This list is ordered alphanumerically by subsection name.
13799 @c It should be the same order and spelling as these options are listed
13800 @c in Machine Dependent Options
13803 * AArch64 Options::
13804 * Adapteva Epiphany Options::
13808 * Blackfin Options::
13813 * DEC Alpha Options::
13817 * GNU/Linux Options::
13827 * MicroBlaze Options::
13830 * MN10300 Options::
13834 * Nios II Options::
13835 * Nvidia PTX Options::
13837 * picoChip Options::
13838 * PowerPC Options::
13840 * RS/6000 and PowerPC Options::
13842 * S/390 and zSeries Options::
13845 * Solaris 2 Options::
13848 * System V Options::
13849 * TILE-Gx Options::
13850 * TILEPro Options::
13855 * VxWorks Options::
13857 * x86 Windows Options::
13858 * Xstormy16 Options::
13860 * zSeries Options::
13863 @node AArch64 Options
13864 @subsection AArch64 Options
13865 @cindex AArch64 Options
13867 These options are defined for AArch64 implementations:
13871 @item -mabi=@var{name}
13873 Generate code for the specified data model. Permissible values
13874 are @samp{ilp32} for SysV-like data model where int, long int and pointers
13875 are 32 bits, and @samp{lp64} for SysV-like data model where int is 32 bits,
13876 but long int and pointers are 64 bits.
13878 The default depends on the specific target configuration. Note that
13879 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
13880 entire program with the same ABI, and link with a compatible set of libraries.
13883 @opindex mbig-endian
13884 Generate big-endian code. This is the default when GCC is configured for an
13885 @samp{aarch64_be-*-*} target.
13887 @item -mgeneral-regs-only
13888 @opindex mgeneral-regs-only
13889 Generate code which uses only the general-purpose registers. This will prevent
13890 the compiler from using floating-point and Advanced SIMD registers but will not
13891 impose any restrictions on the assembler.
13893 @item -mlittle-endian
13894 @opindex mlittle-endian
13895 Generate little-endian code. This is the default when GCC is configured for an
13896 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
13898 @item -mcmodel=tiny
13899 @opindex mcmodel=tiny
13900 Generate code for the tiny code model. The program and its statically defined
13901 symbols must be within 1MB of each other. Programs can be statically or
13902 dynamically linked.
13904 @item -mcmodel=small
13905 @opindex mcmodel=small
13906 Generate code for the small code model. The program and its statically defined
13907 symbols must be within 4GB of each other. Programs can be statically or
13908 dynamically linked. This is the default code model.
13910 @item -mcmodel=large
13911 @opindex mcmodel=large
13912 Generate code for the large code model. This makes no assumptions about
13913 addresses and sizes of sections. Programs can be statically linked only.
13915 @item -mstrict-align
13916 @opindex mstrict-align
13917 Avoid generating memory accesses that may not be aligned on a natural object
13918 boundary as described in the architecture specification.
13920 @item -momit-leaf-frame-pointer
13921 @itemx -mno-omit-leaf-frame-pointer
13922 @opindex momit-leaf-frame-pointer
13923 @opindex mno-omit-leaf-frame-pointer
13924 Omit or keep the frame pointer in leaf functions. The former behavior is the
13927 @item -mtls-dialect=desc
13928 @opindex mtls-dialect=desc
13929 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
13930 of TLS variables. This is the default.
13932 @item -mtls-dialect=traditional
13933 @opindex mtls-dialect=traditional
13934 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
13937 @item -mtls-size=@var{size}
13939 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
13940 This option requires binutils 2.26 or newer.
13942 @item -mfix-cortex-a53-835769
13943 @itemx -mno-fix-cortex-a53-835769
13944 @opindex mfix-cortex-a53-835769
13945 @opindex mno-fix-cortex-a53-835769
13946 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
13947 This involves inserting a NOP instruction between memory instructions and
13948 64-bit integer multiply-accumulate instructions.
13950 @item -mfix-cortex-a53-843419
13951 @itemx -mno-fix-cortex-a53-843419
13952 @opindex mfix-cortex-a53-843419
13953 @opindex mno-fix-cortex-a53-843419
13954 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
13955 This erratum workaround is made at link time and this will only pass the
13956 corresponding flag to the linker.
13958 @item -mlow-precision-recip-sqrt
13959 @item -mno-low-precision-recip-sqrt
13960 @opindex mlow-precision-recip-sqrt
13961 @opindex mno-low-precision-recip-sqrt
13962 Enable or disable the reciprocal square root approximation.
13963 This option only has an effect if @option{-ffast-math} or
13964 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
13965 precision of reciprocal square root results to about 16 bits for
13966 single precision and to 32 bits for double precision.
13968 @item -mlow-precision-sqrt
13969 @item -mno-low-precision-sqrt
13970 @opindex -mlow-precision-sqrt
13971 @opindex -mno-low-precision-sqrt
13972 Enable or disable the square root approximation.
13973 This option only has an effect if @option{-ffast-math} or
13974 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
13975 precision of square root results to about 16 bits for
13976 single precision and to 32 bits for double precision.
13977 If enabled, it implies @option{-mlow-precision-recip-sqrt}.
13979 @item -mlow-precision-div
13980 @item -mno-low-precision-div
13981 @opindex -mlow-precision-div
13982 @opindex -mno-low-precision-div
13983 Enable or disable the division approximation.
13984 This option only has an effect if @option{-ffast-math} or
13985 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
13986 precision of division results to about 16 bits for
13987 single precision and to 32 bits for double precision.
13989 @item -march=@var{name}
13991 Specify the name of the target architecture and, optionally, one or
13992 more feature modifiers. This option has the form
13993 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
13995 The permissible values for @var{arch} are @samp{armv8-a},
13996 @samp{armv8.1-a}, @samp{armv8.2-a} or @var{native}.
13998 The value @samp{armv8.2-a} implies @samp{armv8.1-a} and enables compiler
13999 support for the ARMv8.2-A architecture extensions.
14001 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
14002 support for the ARMv8.1-A architecture extension. In particular, it
14003 enables the @samp{+crc} and @samp{+lse} features.
14005 The value @samp{native} is available on native AArch64 GNU/Linux and
14006 causes the compiler to pick the architecture of the host system. This
14007 option has no effect if the compiler is unable to recognize the
14008 architecture of the host system,
14010 The permissible values for @var{feature} are listed in the sub-section
14011 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
14012 Feature Modifiers}. Where conflicting feature modifiers are
14013 specified, the right-most feature is used.
14015 GCC uses @var{name} to determine what kind of instructions it can emit
14016 when generating assembly code. If @option{-march} is specified
14017 without either of @option{-mtune} or @option{-mcpu} also being
14018 specified, the code is tuned to perform well across a range of target
14019 processors implementing the target architecture.
14021 @item -mtune=@var{name}
14023 Specify the name of the target processor for which GCC should tune the
14024 performance of the code. Permissible values for this option are:
14025 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57},
14026 @samp{cortex-a72}, @samp{cortex-a73}, @samp{exynos-m1}, @samp{falkor},
14027 @samp{qdf24xx}, @samp{thunderx}, @samp{xgene1}, @samp{vulcan},
14028 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
14029 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53}, @samp{native}.
14031 The values @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
14032 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53}
14033 specify that GCC should tune for a big.LITTLE system.
14035 Additionally on native AArch64 GNU/Linux systems the value
14036 @samp{native} tunes performance to the host system. This option has no effect
14037 if the compiler is unable to recognize the processor of the host system.
14039 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
14040 are specified, the code is tuned to perform well across a range
14041 of target processors.
14043 This option cannot be suffixed by feature modifiers.
14045 @item -mcpu=@var{name}
14047 Specify the name of the target processor, optionally suffixed by one
14048 or more feature modifiers. This option has the form
14049 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
14050 the permissible values for @var{cpu} are the same as those available
14051 for @option{-mtune}. The permissible values for @var{feature} are
14052 documented in the sub-section on
14053 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
14054 Feature Modifiers}. Where conflicting feature modifiers are
14055 specified, the right-most feature is used.
14057 GCC uses @var{name} to determine what kind of instructions it can emit when
14058 generating assembly code (as if by @option{-march}) and to determine
14059 the target processor for which to tune for performance (as if
14060 by @option{-mtune}). Where this option is used in conjunction
14061 with @option{-march} or @option{-mtune}, those options take precedence
14062 over the appropriate part of this option.
14064 @item -moverride=@var{string}
14066 Override tuning decisions made by the back-end in response to a
14067 @option{-mtune=} switch. The syntax, semantics, and accepted values
14068 for @var{string} in this option are not guaranteed to be consistent
14071 This option is only intended to be useful when developing GCC.
14073 @item -mpc-relative-literal-loads
14074 @opindex mpc-relative-literal-loads
14075 Enable PC-relative literal loads. With this option literal pools are
14076 accessed using a single instruction and emitted after each function. This
14077 limits the maximum size of functions to 1MB. This is enabled by default for
14078 @option{-mcmodel=tiny}.
14082 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
14083 @anchor{aarch64-feature-modifiers}
14084 @cindex @option{-march} feature modifiers
14085 @cindex @option{-mcpu} feature modifiers
14086 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
14087 the following and their inverses @option{no@var{feature}}:
14091 Enable CRC extension. This is on by default for
14092 @option{-march=armv8.1-a}.
14094 Enable Crypto extension. This also enables Advanced SIMD and floating-point
14097 Enable floating-point instructions. This is on by default for all possible
14098 values for options @option{-march} and @option{-mcpu}.
14100 Enable Advanced SIMD instructions. This also enables floating-point
14101 instructions. This is on by default for all possible values for options
14102 @option{-march} and @option{-mcpu}.
14104 Enable Large System Extension instructions. This is on by default for
14105 @option{-march=armv8.1-a}.
14107 Enable FP16 extension. This also enables floating-point instructions.
14111 Feature @option{crypto} implies @option{simd}, which implies @option{fp}.
14112 Conversely, @option{nofp} implies @option{nosimd}, which implies
14115 @node Adapteva Epiphany Options
14116 @subsection Adapteva Epiphany Options
14118 These @samp{-m} options are defined for Adapteva Epiphany:
14121 @item -mhalf-reg-file
14122 @opindex mhalf-reg-file
14123 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
14124 That allows code to run on hardware variants that lack these registers.
14126 @item -mprefer-short-insn-regs
14127 @opindex mprefer-short-insn-regs
14128 Preferentially allocate registers that allow short instruction generation.
14129 This can result in increased instruction count, so this may either reduce or
14130 increase overall code size.
14132 @item -mbranch-cost=@var{num}
14133 @opindex mbranch-cost
14134 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14135 This cost is only a heuristic and is not guaranteed to produce
14136 consistent results across releases.
14140 Enable the generation of conditional moves.
14142 @item -mnops=@var{num}
14144 Emit @var{num} NOPs before every other generated instruction.
14146 @item -mno-soft-cmpsf
14147 @opindex mno-soft-cmpsf
14148 For single-precision floating-point comparisons, emit an @code{fsub} instruction
14149 and test the flags. This is faster than a software comparison, but can
14150 get incorrect results in the presence of NaNs, or when two different small
14151 numbers are compared such that their difference is calculated as zero.
14152 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
14153 software comparisons.
14155 @item -mstack-offset=@var{num}
14156 @opindex mstack-offset
14157 Set the offset between the top of the stack and the stack pointer.
14158 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
14159 can be used by leaf functions without stack allocation.
14160 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
14161 Note also that this option changes the ABI; compiling a program with a
14162 different stack offset than the libraries have been compiled with
14163 generally does not work.
14164 This option can be useful if you want to evaluate if a different stack
14165 offset would give you better code, but to actually use a different stack
14166 offset to build working programs, it is recommended to configure the
14167 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
14169 @item -mno-round-nearest
14170 @opindex mno-round-nearest
14171 Make the scheduler assume that the rounding mode has been set to
14172 truncating. The default is @option{-mround-nearest}.
14175 @opindex mlong-calls
14176 If not otherwise specified by an attribute, assume all calls might be beyond
14177 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
14178 function address into a register before performing a (otherwise direct) call.
14179 This is the default.
14181 @item -mshort-calls
14182 @opindex short-calls
14183 If not otherwise specified by an attribute, assume all direct calls are
14184 in the range of the @code{b} / @code{bl} instructions, so use these instructions
14185 for direct calls. The default is @option{-mlong-calls}.
14189 Assume addresses can be loaded as 16-bit unsigned values. This does not
14190 apply to function addresses for which @option{-mlong-calls} semantics
14193 @item -mfp-mode=@var{mode}
14195 Set the prevailing mode of the floating-point unit.
14196 This determines the floating-point mode that is provided and expected
14197 at function call and return time. Making this mode match the mode you
14198 predominantly need at function start can make your programs smaller and
14199 faster by avoiding unnecessary mode switches.
14201 @var{mode} can be set to one the following values:
14205 Any mode at function entry is valid, and retained or restored when
14206 the function returns, and when it calls other functions.
14207 This mode is useful for compiling libraries or other compilation units
14208 you might want to incorporate into different programs with different
14209 prevailing FPU modes, and the convenience of being able to use a single
14210 object file outweighs the size and speed overhead for any extra
14211 mode switching that might be needed, compared with what would be needed
14212 with a more specific choice of prevailing FPU mode.
14215 This is the mode used for floating-point calculations with
14216 truncating (i.e.@: round towards zero) rounding mode. That includes
14217 conversion from floating point to integer.
14219 @item round-nearest
14220 This is the mode used for floating-point calculations with
14221 round-to-nearest-or-even rounding mode.
14224 This is the mode used to perform integer calculations in the FPU, e.g.@:
14225 integer multiply, or integer multiply-and-accumulate.
14228 The default is @option{-mfp-mode=caller}
14230 @item -mnosplit-lohi
14231 @itemx -mno-postinc
14232 @itemx -mno-postmodify
14233 @opindex mnosplit-lohi
14234 @opindex mno-postinc
14235 @opindex mno-postmodify
14236 Code generation tweaks that disable, respectively, splitting of 32-bit
14237 loads, generation of post-increment addresses, and generation of
14238 post-modify addresses. The defaults are @option{msplit-lohi},
14239 @option{-mpost-inc}, and @option{-mpost-modify}.
14241 @item -mnovect-double
14242 @opindex mno-vect-double
14243 Change the preferred SIMD mode to SImode. The default is
14244 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
14246 @item -max-vect-align=@var{num}
14247 @opindex max-vect-align
14248 The maximum alignment for SIMD vector mode types.
14249 @var{num} may be 4 or 8. The default is 8.
14250 Note that this is an ABI change, even though many library function
14251 interfaces are unaffected if they don't use SIMD vector modes
14252 in places that affect size and/or alignment of relevant types.
14254 @item -msplit-vecmove-early
14255 @opindex msplit-vecmove-early
14256 Split vector moves into single word moves before reload. In theory this
14257 can give better register allocation, but so far the reverse seems to be
14258 generally the case.
14260 @item -m1reg-@var{reg}
14262 Specify a register to hold the constant @minus{}1, which makes loading small negative
14263 constants and certain bitmasks faster.
14264 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
14265 which specify use of that register as a fixed register,
14266 and @samp{none}, which means that no register is used for this
14267 purpose. The default is @option{-m1reg-none}.
14272 @subsection ARC Options
14273 @cindex ARC options
14275 The following options control the architecture variant for which code
14278 @c architecture variants
14281 @item -mbarrel-shifter
14282 @opindex mbarrel-shifter
14283 Generate instructions supported by barrel shifter. This is the default
14284 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
14286 @item -mcpu=@var{cpu}
14288 Set architecture type, register usage, and instruction scheduling
14289 parameters for @var{cpu}. There are also shortcut alias options
14290 available for backward compatibility and convenience. Supported
14291 values for @var{cpu} are
14297 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
14301 Compile for ARC601. Alias: @option{-mARC601}.
14306 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
14307 This is the default when configured with @option{--with-cpu=arc700}@.
14310 Compile for ARC EM.
14313 Compile for ARC HS.
14317 Compile for ARC EM cpu with no hardware extension.
14321 Compile for ARC EM4 cpu.
14325 Compile for ARC EM4 DMIPS cpu.
14329 Compile for ARC EM4 DMIPS cpu with single precision floating point
14334 Compile for ARC EM4 DMIPS cpu with single precision floating point and
14335 double assists instructions.
14339 Compile for ARC HS cpu with no hardware extension, except the atomic
14344 Compile for ARC HS34 cpu.
14348 Compile for ARC HS38 cpu.
14351 @opindex hs38_linux
14352 Compile for ARC HS38 cpu with all hardware extensions on.
14355 @opindex arc600_norm
14356 Compile for ARC 600 cpu with norm instruction enabled.
14358 @item arc600_mul32x16
14359 @opindex arc600_mul32x16
14360 Compile for ARC 600 cpu with norm and mul32x16 instructions enabled.
14363 @opindex arc600_mul64
14364 Compile for ARC 600 cpu with norm and mul64 instructions enabled.
14367 @opindex arc601_norm
14368 Compile for ARC 601 cpu with norm instruction enabled.
14370 @item arc601_mul32x16
14371 @opindex arc601_mul32x16
14372 Compile for ARC 601 cpu with norm and mul32x16 instructions enabled.
14375 @opindex arc601_mul64
14376 Compile for ARC 601 cpu with norm and mul64 instructions enabled.
14380 Compile for ARC 700 on NPS400 chip.
14386 @itemx -mdpfp-compact
14387 @opindex mdpfp-compact
14388 FPX: Generate Double Precision FPX instructions, tuned for the compact
14392 @opindex mdpfp-fast
14393 FPX: Generate Double Precision FPX instructions, tuned for the fast
14396 @item -mno-dpfp-lrsr
14397 @opindex mno-dpfp-lrsr
14398 Disable LR and SR instructions from using FPX extension aux registers.
14402 Generate Extended arithmetic instructions. Currently only
14403 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
14404 supported. This is always enabled for @option{-mcpu=ARC700}.
14408 Do not generate mpy instructions for ARC700. This instruction is
14413 Generate 32x16 bit multiply and mac instructions.
14417 Generate mul64 and mulu64 instructions. Only valid for @option{-mcpu=ARC600}.
14421 Generate norm instruction. This is the default if @option{-mcpu=ARC700}
14426 @itemx -mspfp-compact
14427 @opindex mspfp-compact
14428 FPX: Generate Single Precision FPX instructions, tuned for the compact
14432 @opindex mspfp-fast
14433 FPX: Generate Single Precision FPX instructions, tuned for the fast
14438 Enable generation of ARC SIMD instructions via target-specific
14439 builtins. Only valid for @option{-mcpu=ARC700}.
14442 @opindex msoft-float
14443 This option ignored; it is provided for compatibility purposes only.
14444 Software floating point code is emitted by default, and this default
14445 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
14446 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
14447 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
14451 Generate swap instructions.
14455 This enables Locked Load/Store Conditional extension to implement
14456 atomic memopry built-in functions. Not available for ARC 6xx or ARC
14461 Enable DIV/REM instructions for ARCv2 cores.
14463 @item -mcode-density
14464 @opindex mcode-density
14465 Enable code density instructions for ARC EM, default on for ARC HS.
14469 Enable double load/store operations for ARC HS cores.
14471 @item -mtp-regno=@var{regno}
14473 Specify thread pointer register number.
14475 @item -mmpy-option=@var{multo}
14476 @opindex mmpy-option
14477 Compile ARCv2 code with a multiplier design option. @samp{wlh1} is
14478 the default value. The recognized values for @var{multo} are:
14482 No multiplier available.
14486 The multiply option is set to w: 16x16 multiplier, fully pipelined.
14487 The following instructions are enabled: MPYW, and MPYUW.
14491 The multiply option is set to wlh1: 32x32 multiplier, fully
14492 pipelined (1 stage). The following instructions are additionally
14493 enabled: MPY, MPYU, MPYM, MPYMU, and MPY_S.
14497 The multiply option is set to wlh2: 32x32 multiplier, fully pipelined
14498 (2 stages). The following instructions are additionally enabled: MPY,
14499 MPYU, MPYM, MPYMU, and MPY_S.
14503 The multiply option is set to wlh3: Two 16x16 multiplier, blocking,
14504 sequential. The following instructions are additionally enabled: MPY,
14505 MPYU, MPYM, MPYMU, and MPY_S.
14509 The multiply option is set to wlh4: One 16x16 multiplier, blocking,
14510 sequential. The following instructions are additionally enabled: MPY,
14511 MPYU, MPYM, MPYMU, and MPY_S.
14515 The multiply option is set to wlh5: One 32x4 multiplier, blocking,
14516 sequential. The following instructions are additionally enabled: MPY,
14517 MPYU, MPYM, MPYMU, and MPY_S.
14521 This option is only available for ARCv2 cores@.
14523 @item -mfpu=@var{fpu}
14525 Enables specific floating-point hardware extension for ARCv2
14526 core. Supported values for @var{fpu} are:
14532 Enables support for single precision floating point hardware
14537 Enables support for double precision floating point hardware
14538 extensions. The single precision floating point extension is also
14539 enabled. Not available for ARC EM@.
14543 Enables support for double precision floating point hardware
14544 extensions using double precision assist instructions. The single
14545 precision floating point extension is also enabled. This option is
14546 only available for ARC EM@.
14550 Enables support for double precision floating point hardware
14551 extensions using double precision assist instructions, and simple
14552 precision square-root and divide hardware extensions. The single
14553 precision floating point extension is also enabled. This option is
14554 only available for ARC EM@.
14558 Enables support for double precision floating point hardware
14559 extensions using double precision assist instructions, and simple
14560 precision fused multiple and add hardware extension. The single
14561 precision floating point extension is also enabled. This option is
14562 only available for ARC EM@.
14566 Enables support for double precision floating point hardware
14567 extensions using double precision assist instructions, and all simple
14568 precision hardware extensions. The single precision floating point
14569 extension is also enabled. This option is only available for ARC EM@.
14573 Enables support for single precision floating point, and single
14574 precision square-root and divide hardware extensions@.
14578 Enables support for double precision floating point, and double
14579 precision square-root and divide hardware extensions. This option
14580 includes option @samp{fpus_div}. Not available for ARC EM@.
14584 Enables support for single precision floating point, and single
14585 precision fused multiple and add hardware extensions@.
14589 Enables support for double precision floating point, and double
14590 precision fused multiple and add hardware extensions. This option
14591 includes option @samp{fpus_fma}. Not available for ARC EM@.
14595 Enables support for all single precision floating point hardware
14600 Enables support for all single and double precision floating point
14601 hardware extensions. Not available for ARC EM@.
14607 The following options are passed through to the assembler, and also
14608 define preprocessor macro symbols.
14610 @c Flags used by the assembler, but for which we define preprocessor
14611 @c macro symbols as well.
14614 @opindex mdsp-packa
14615 Passed down to the assembler to enable the DSP Pack A extensions.
14616 Also sets the preprocessor symbol @code{__Xdsp_packa}. This option is
14621 Passed down to the assembler to enable the dual viterbi butterfly
14622 extension. Also sets the preprocessor symbol @code{__Xdvbf}. This
14623 option is deprecated.
14625 @c ARC700 4.10 extension instruction
14628 Passed down to the assembler to enable the Locked Load/Store
14629 Conditional extension. Also sets the preprocessor symbol
14634 Passed down to the assembler. Also sets the preprocessor symbol
14635 @code{__Xxmac_d16}. This option is deprecated.
14639 Passed down to the assembler. Also sets the preprocessor symbol
14640 @code{__Xxmac_24}. This option is deprecated.
14642 @c ARC700 4.10 extension instruction
14645 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
14646 extension instruction. Also sets the preprocessor symbol
14647 @code{__Xrtsc}. This option is deprecated.
14649 @c ARC700 4.10 extension instruction
14652 Passed down to the assembler to enable the swap byte ordering
14653 extension instruction. Also sets the preprocessor symbol
14657 @opindex mtelephony
14658 Passed down to the assembler to enable dual and single operand
14659 instructions for telephony. Also sets the preprocessor symbol
14660 @code{__Xtelephony}. This option is deprecated.
14664 Passed down to the assembler to enable the XY Memory extension. Also
14665 sets the preprocessor symbol @code{__Xxy}.
14669 The following options control how the assembly code is annotated:
14671 @c Assembly annotation options
14675 Annotate assembler instructions with estimated addresses.
14677 @item -mannotate-align
14678 @opindex mannotate-align
14679 Explain what alignment considerations lead to the decision to make an
14680 instruction short or long.
14684 The following options are passed through to the linker:
14686 @c options passed through to the linker
14690 Passed through to the linker, to specify use of the @code{arclinux} emulation.
14691 This option is enabled by default in tool chains built for
14692 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
14693 when profiling is not requested.
14695 @item -marclinux_prof
14696 @opindex marclinux_prof
14697 Passed through to the linker, to specify use of the
14698 @code{arclinux_prof} emulation. This option is enabled by default in
14699 tool chains built for @w{@code{arc-linux-uclibc}} and
14700 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
14704 The following options control the semantics of generated code:
14706 @c semantically relevant code generation options
14709 @opindex mlong-calls
14710 Generate call insns as register indirect calls, thus providing access
14711 to the full 32-bit address range.
14713 @item -mmedium-calls
14714 @opindex mmedium-calls
14715 Don't use less than 25 bit addressing range for calls, which is the
14716 offset available for an unconditional branch-and-link
14717 instruction. Conditional execution of function calls is suppressed, to
14718 allow use of the 25-bit range, rather than the 21-bit range with
14719 conditional branch-and-link. This is the default for tool chains built
14720 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
14724 Do not generate sdata references. This is the default for tool chains
14725 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
14729 @opindex mucb-mcount
14730 Instrument with mcount calls as used in UCB code. I.e. do the
14731 counting in the callee, not the caller. By default ARC instrumentation
14732 counts in the caller.
14734 @item -mvolatile-cache
14735 @opindex mvolatile-cache
14736 Use ordinarily cached memory accesses for volatile references. This is the
14739 @item -mno-volatile-cache
14740 @opindex mno-volatile-cache
14741 Enable cache bypass for volatile references.
14745 The following options fine tune code generation:
14746 @c code generation tuning options
14749 @opindex malign-call
14750 Do alignment optimizations for call instructions.
14752 @item -mauto-modify-reg
14753 @opindex mauto-modify-reg
14754 Enable the use of pre/post modify with register displacement.
14756 @item -mbbit-peephole
14757 @opindex mbbit-peephole
14758 Enable bbit peephole2.
14762 This option disables a target-specific pass in @file{arc_reorg} to
14763 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
14764 generation driven by the combiner pass.
14766 @item -mcase-vector-pcrel
14767 @opindex mcase-vector-pcrel
14768 Use pc-relative switch case tables - this enables case table shortening.
14769 This is the default for @option{-Os}.
14771 @item -mcompact-casesi
14772 @opindex mcompact-casesi
14773 Enable compact casesi pattern. This is the default for @option{-Os},
14774 and only available for ARCv1 cores.
14776 @item -mno-cond-exec
14777 @opindex mno-cond-exec
14778 Disable ARCompact specific pass to generate conditional execution instructions.
14779 Due to delay slot scheduling and interactions between operand numbers,
14780 literal sizes, instruction lengths, and the support for conditional execution,
14781 the target-independent pass to generate conditional execution is often lacking,
14782 so the ARC port has kept a special pass around that tries to find more
14783 conditional execution generating opportunities after register allocation,
14784 branch shortening, and delay slot scheduling have been done. This pass
14785 generally, but not always, improves performance and code size, at the cost of
14786 extra compilation time, which is why there is an option to switch it off.
14787 If you have a problem with call instructions exceeding their allowable
14788 offset range because they are conditionalized, you should consider using
14789 @option{-mmedium-calls} instead.
14791 @item -mearly-cbranchsi
14792 @opindex mearly-cbranchsi
14793 Enable pre-reload use of the cbranchsi pattern.
14795 @item -mexpand-adddi
14796 @opindex mexpand-adddi
14797 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
14798 @code{add.f}, @code{adc} etc.
14800 @item -mindexed-loads
14801 @opindex mindexed-loads
14802 Enable the use of indexed loads. This can be problematic because some
14803 optimizers then assume that indexed stores exist, which is not
14807 Enable Local Register Allocation. This is still experimental for ARC,
14808 so by default the compiler uses standard reload
14809 (i.e. @option{-mno-lra}).
14811 @item -mlra-priority-none
14812 @opindex mlra-priority-none
14813 Don't indicate any priority for target registers.
14815 @item -mlra-priority-compact
14816 @opindex mlra-priority-compact
14817 Indicate target register priority for r0..r3 / r12..r15.
14819 @item -mlra-priority-noncompact
14820 @opindex mlra-priority-noncompact
14821 Reduce target register priority for r0..r3 / r12..r15.
14823 @item -mno-millicode
14824 @opindex mno-millicode
14825 When optimizing for size (using @option{-Os}), prologues and epilogues
14826 that have to save or restore a large number of registers are often
14827 shortened by using call to a special function in libgcc; this is
14828 referred to as a @emph{millicode} call. As these calls can pose
14829 performance issues, and/or cause linking issues when linking in a
14830 nonstandard way, this option is provided to turn off millicode call
14834 @opindex mmixed-code
14835 Tweak register allocation to help 16-bit instruction generation.
14836 This generally has the effect of decreasing the average instruction size
14837 while increasing the instruction count.
14841 Enable 'q' instruction alternatives.
14842 This is the default for @option{-Os}.
14846 Enable Rcq constraint handling - most short code generation depends on this.
14847 This is the default.
14851 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
14852 This is the default.
14854 @item -msize-level=@var{level}
14855 @opindex msize-level
14856 Fine-tune size optimization with regards to instruction lengths and alignment.
14857 The recognized values for @var{level} are:
14860 No size optimization. This level is deprecated and treated like @samp{1}.
14863 Short instructions are used opportunistically.
14866 In addition, alignment of loops and of code after barriers are dropped.
14869 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
14873 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
14874 the behavior when this is not set is equivalent to level @samp{1}.
14876 @item -mtune=@var{cpu}
14878 Set instruction scheduling parameters for @var{cpu}, overriding any implied
14879 by @option{-mcpu=}.
14881 Supported values for @var{cpu} are
14885 Tune for ARC600 cpu.
14888 Tune for ARC601 cpu.
14891 Tune for ARC700 cpu with standard multiplier block.
14894 Tune for ARC700 cpu with XMAC block.
14897 Tune for ARC725D cpu.
14900 Tune for ARC750D cpu.
14904 @item -mmultcost=@var{num}
14906 Cost to assume for a multiply instruction, with @samp{4} being equal to a
14907 normal instruction.
14909 @item -munalign-prob-threshold=@var{probability}
14910 @opindex munalign-prob-threshold
14911 Set probability threshold for unaligning branches.
14912 When tuning for @samp{ARC700} and optimizing for speed, branches without
14913 filled delay slot are preferably emitted unaligned and long, unless
14914 profiling indicates that the probability for the branch to be taken
14915 is below @var{probability}. @xref{Cross-profiling}.
14916 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
14920 The following options are maintained for backward compatibility, but
14921 are now deprecated and will be removed in a future release:
14923 @c Deprecated options
14931 @opindex mbig-endian
14934 Compile code for big endian targets. Use of these options is now
14935 deprecated. Users wanting big-endian code, should use the
14936 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
14937 building the tool chain, for which big-endian is the default.
14939 @item -mlittle-endian
14940 @opindex mlittle-endian
14943 Compile code for little endian targets. Use of these options is now
14944 deprecated. Users wanting little-endian code should use the
14945 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
14946 building the tool chain, for which little-endian is the default.
14948 @item -mbarrel_shifter
14949 @opindex mbarrel_shifter
14950 Replaced by @option{-mbarrel-shifter}.
14952 @item -mdpfp_compact
14953 @opindex mdpfp_compact
14954 Replaced by @option{-mdpfp-compact}.
14957 @opindex mdpfp_fast
14958 Replaced by @option{-mdpfp-fast}.
14961 @opindex mdsp_packa
14962 Replaced by @option{-mdsp-packa}.
14966 Replaced by @option{-mea}.
14970 Replaced by @option{-mmac-24}.
14974 Replaced by @option{-mmac-d16}.
14976 @item -mspfp_compact
14977 @opindex mspfp_compact
14978 Replaced by @option{-mspfp-compact}.
14981 @opindex mspfp_fast
14982 Replaced by @option{-mspfp-fast}.
14984 @item -mtune=@var{cpu}
14986 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
14987 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
14988 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
14990 @item -multcost=@var{num}
14992 Replaced by @option{-mmultcost}.
14997 @subsection ARM Options
14998 @cindex ARM options
15000 These @samp{-m} options are defined for the ARM port:
15003 @item -mabi=@var{name}
15005 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
15006 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
15009 @opindex mapcs-frame
15010 Generate a stack frame that is compliant with the ARM Procedure Call
15011 Standard for all functions, even if this is not strictly necessary for
15012 correct execution of the code. Specifying @option{-fomit-frame-pointer}
15013 with this option causes the stack frames not to be generated for
15014 leaf functions. The default is @option{-mno-apcs-frame}.
15015 This option is deprecated.
15019 This is a synonym for @option{-mapcs-frame} and is deprecated.
15022 @c not currently implemented
15023 @item -mapcs-stack-check
15024 @opindex mapcs-stack-check
15025 Generate code to check the amount of stack space available upon entry to
15026 every function (that actually uses some stack space). If there is
15027 insufficient space available then either the function
15028 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
15029 called, depending upon the amount of stack space required. The runtime
15030 system is required to provide these functions. The default is
15031 @option{-mno-apcs-stack-check}, since this produces smaller code.
15033 @c not currently implemented
15034 @item -mapcs-reentrant
15035 @opindex mapcs-reentrant
15036 Generate reentrant, position-independent code. The default is
15037 @option{-mno-apcs-reentrant}.
15040 @item -mthumb-interwork
15041 @opindex mthumb-interwork
15042 Generate code that supports calling between the ARM and Thumb
15043 instruction sets. Without this option, on pre-v5 architectures, the
15044 two instruction sets cannot be reliably used inside one program. The
15045 default is @option{-mno-thumb-interwork}, since slightly larger code
15046 is generated when @option{-mthumb-interwork} is specified. In AAPCS
15047 configurations this option is meaningless.
15049 @item -mno-sched-prolog
15050 @opindex mno-sched-prolog
15051 Prevent the reordering of instructions in the function prologue, or the
15052 merging of those instruction with the instructions in the function's
15053 body. This means that all functions start with a recognizable set
15054 of instructions (or in fact one of a choice from a small set of
15055 different function prologues), and this information can be used to
15056 locate the start of functions inside an executable piece of code. The
15057 default is @option{-msched-prolog}.
15059 @item -mfloat-abi=@var{name}
15060 @opindex mfloat-abi
15061 Specifies which floating-point ABI to use. Permissible values
15062 are: @samp{soft}, @samp{softfp} and @samp{hard}.
15064 Specifying @samp{soft} causes GCC to generate output containing
15065 library calls for floating-point operations.
15066 @samp{softfp} allows the generation of code using hardware floating-point
15067 instructions, but still uses the soft-float calling conventions.
15068 @samp{hard} allows generation of floating-point instructions
15069 and uses FPU-specific calling conventions.
15071 The default depends on the specific target configuration. Note that
15072 the hard-float and soft-float ABIs are not link-compatible; you must
15073 compile your entire program with the same ABI, and link with a
15074 compatible set of libraries.
15076 @item -mlittle-endian
15077 @opindex mlittle-endian
15078 Generate code for a processor running in little-endian mode. This is
15079 the default for all standard configurations.
15082 @opindex mbig-endian
15083 Generate code for a processor running in big-endian mode; the default is
15084 to compile code for a little-endian processor.
15086 @item -march=@var{name}
15088 This specifies the name of the target ARM architecture. GCC uses this
15089 name to determine what kind of instructions it can emit when generating
15090 assembly code. This option can be used in conjunction with or instead
15091 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
15092 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
15093 @samp{armv5}, @samp{armv5e}, @samp{armv5t}, @samp{armv5te},
15094 @samp{armv6}, @samp{armv6-m}, @samp{armv6j}, @samp{armv6k},
15095 @samp{armv6kz}, @samp{armv6s-m},
15096 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk},
15097 @samp{armv7}, @samp{armv7-a}, @samp{armv7-m}, @samp{armv7-r}, @samp{armv7e-m},
15098 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc}, @samp{armv8.1-a},
15099 @samp{armv8.1-a+crc}, @samp{armv8-m.base}, @samp{armv8-m.main},
15100 @samp{armv8-m.main+dsp}, @samp{iwmmxt}, @samp{iwmmxt2}.
15102 Architecture revisions older than @samp{armv4t} are deprecated.
15104 @option{-march=armv6s-m} is the @samp{armv6-m} architecture with support for
15105 the (now mandatory) SVC instruction.
15107 @option{-march=armv6zk} is an alias for @samp{armv6kz}, existing for backwards
15110 @option{-march=armv7ve} is the @samp{armv7-a} architecture with virtualization
15113 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
15114 architecture together with the optional CRC32 extensions.
15116 @option{-march=armv8.1-a} enables compiler support for the ARMv8.1-A
15117 architecture. This also enables the features provided by
15118 @option{-march=armv8-a+crc}.
15120 @option{-march=armv8.2-a} enables compiler support for the ARMv8.2-A
15121 architecture. This also enables the features provided by
15122 @option{-march=armv8.1-a}.
15124 @option{-march=armv8.2-a+fp16} enables compiler support for the
15125 ARMv8.2-A architecture with the optional FP16 instructions extension.
15126 This also enables the features provided by @option{-march=armv8.1-a}
15127 and implies @option{-mfp16-format=ieee}.
15129 @option{-march=native} causes the compiler to auto-detect the architecture
15130 of the build computer. At present, this feature is only supported on
15131 GNU/Linux, and not all architectures are recognized. If the auto-detect
15132 is unsuccessful the option has no effect.
15134 @item -mtune=@var{name}
15136 This option specifies the name of the target ARM processor for
15137 which GCC should tune the performance of the code.
15138 For some ARM implementations better performance can be obtained by using
15140 Permissible names are: @samp{arm2}, @samp{arm250},
15141 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
15142 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
15143 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
15144 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
15146 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
15147 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
15148 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
15149 @samp{strongarm1110},
15150 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
15151 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
15152 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
15153 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
15154 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
15155 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
15156 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
15157 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
15158 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
15159 @samp{cortex-a32}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57},
15160 @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-r4},
15161 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8},
15169 @samp{cortex-m0plus},
15170 @samp{cortex-m1.small-multiply},
15171 @samp{cortex-m0.small-multiply},
15172 @samp{cortex-m0plus.small-multiply},
15176 @samp{marvell-pj4},
15177 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
15178 @samp{fa526}, @samp{fa626},
15179 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
15182 Additionally, this option can specify that GCC should tune the performance
15183 of the code for a big.LITTLE system. Permissible names are:
15184 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
15185 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15186 @samp{cortex-a72.cortex-a35}, @samp{cortex-a73.cortex-a53}.
15188 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
15189 performance for a blend of processors within architecture @var{arch}.
15190 The aim is to generate code that run well on the current most popular
15191 processors, balancing between optimizations that benefit some CPUs in the
15192 range, and avoiding performance pitfalls of other CPUs. The effects of
15193 this option may change in future GCC versions as CPU models come and go.
15195 @option{-mtune=native} causes the compiler to auto-detect the CPU
15196 of the build computer. At present, this feature is only supported on
15197 GNU/Linux, and not all architectures are recognized. If the auto-detect is
15198 unsuccessful the option has no effect.
15200 @item -mcpu=@var{name}
15202 This specifies the name of the target ARM processor. GCC uses this name
15203 to derive the name of the target ARM architecture (as if specified
15204 by @option{-march}) and the ARM processor type for which to tune for
15205 performance (as if specified by @option{-mtune}). Where this option
15206 is used in conjunction with @option{-march} or @option{-mtune},
15207 those options take precedence over the appropriate part of this option.
15209 Permissible names for this option are the same as those for
15212 @option{-mcpu=generic-@var{arch}} is also permissible, and is
15213 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
15214 See @option{-mtune} for more information.
15216 @option{-mcpu=native} causes the compiler to auto-detect the CPU
15217 of the build computer. At present, this feature is only supported on
15218 GNU/Linux, and not all architectures are recognized. If the auto-detect
15219 is unsuccessful the option has no effect.
15221 @item -mfpu=@var{name}
15223 This specifies what floating-point hardware (or hardware emulation) is
15224 available on the target. Permissible names are: @samp{vfpv2}, @samp{vfpv3},
15225 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
15226 @samp{vfpv3xd-fp16}, @samp{neon-vfpv3}, @samp{neon-fp16}, @samp{vfpv4},
15227 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
15228 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
15229 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
15230 Note that @samp{neon} is an alias for @samp{neon-vfpv3} and @samp{vfp}
15231 is an alias for @samp{vfpv2}.
15233 If @option{-msoft-float} is specified this specifies the format of
15234 floating-point values.
15236 If the selected floating-point hardware includes the NEON extension
15237 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
15238 operations are not generated by GCC's auto-vectorization pass unless
15239 @option{-funsafe-math-optimizations} is also specified. This is
15240 because NEON hardware does not fully implement the IEEE 754 standard for
15241 floating-point arithmetic (in particular denormal values are treated as
15242 zero), so the use of NEON instructions may lead to a loss of precision.
15244 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}).
15246 @item -mfp16-format=@var{name}
15247 @opindex mfp16-format
15248 Specify the format of the @code{__fp16} half-precision floating-point type.
15249 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
15250 the default is @samp{none}, in which case the @code{__fp16} type is not
15251 defined. @xref{Half-Precision}, for more information.
15253 @item -mstructure-size-boundary=@var{n}
15254 @opindex mstructure-size-boundary
15255 The sizes of all structures and unions are rounded up to a multiple
15256 of the number of bits set by this option. Permissible values are 8, 32
15257 and 64. The default value varies for different toolchains. For the COFF
15258 targeted toolchain the default value is 8. A value of 64 is only allowed
15259 if the underlying ABI supports it.
15261 Specifying a larger number can produce faster, more efficient code, but
15262 can also increase the size of the program. Different values are potentially
15263 incompatible. Code compiled with one value cannot necessarily expect to
15264 work with code or libraries compiled with another value, if they exchange
15265 information using structures or unions.
15267 @item -mabort-on-noreturn
15268 @opindex mabort-on-noreturn
15269 Generate a call to the function @code{abort} at the end of a
15270 @code{noreturn} function. It is executed if the function tries to
15274 @itemx -mno-long-calls
15275 @opindex mlong-calls
15276 @opindex mno-long-calls
15277 Tells the compiler to perform function calls by first loading the
15278 address of the function into a register and then performing a subroutine
15279 call on this register. This switch is needed if the target function
15280 lies outside of the 64-megabyte addressing range of the offset-based
15281 version of subroutine call instruction.
15283 Even if this switch is enabled, not all function calls are turned
15284 into long calls. The heuristic is that static functions, functions
15285 that have the @code{short_call} attribute, functions that are inside
15286 the scope of a @code{#pragma no_long_calls} directive, and functions whose
15287 definitions have already been compiled within the current compilation
15288 unit are not turned into long calls. The exceptions to this rule are
15289 that weak function definitions, functions with the @code{long_call}
15290 attribute or the @code{section} attribute, and functions that are within
15291 the scope of a @code{#pragma long_calls} directive are always
15292 turned into long calls.
15294 This feature is not enabled by default. Specifying
15295 @option{-mno-long-calls} restores the default behavior, as does
15296 placing the function calls within the scope of a @code{#pragma
15297 long_calls_off} directive. Note these switches have no effect on how
15298 the compiler generates code to handle function calls via function
15301 @item -msingle-pic-base
15302 @opindex msingle-pic-base
15303 Treat the register used for PIC addressing as read-only, rather than
15304 loading it in the prologue for each function. The runtime system is
15305 responsible for initializing this register with an appropriate value
15306 before execution begins.
15308 @item -mpic-register=@var{reg}
15309 @opindex mpic-register
15310 Specify the register to be used for PIC addressing.
15311 For standard PIC base case, the default is any suitable register
15312 determined by compiler. For single PIC base case, the default is
15313 @samp{R9} if target is EABI based or stack-checking is enabled,
15314 otherwise the default is @samp{R10}.
15316 @item -mpic-data-is-text-relative
15317 @opindex mpic-data-is-text-relative
15318 Assume that the displacement between the text and data segments is fixed
15319 at static link time. This permits using PC-relative addressing
15320 operations to access data known to be in the data segment. For
15321 non-VxWorks RTP targets, this option is enabled by default. When
15322 disabled on such targets, it will enable @option{-msingle-pic-base} by
15325 @item -mpoke-function-name
15326 @opindex mpoke-function-name
15327 Write the name of each function into the text section, directly
15328 preceding the function prologue. The generated code is similar to this:
15332 .ascii "arm_poke_function_name", 0
15335 .word 0xff000000 + (t1 - t0)
15336 arm_poke_function_name
15338 stmfd sp!, @{fp, ip, lr, pc@}
15342 When performing a stack backtrace, code can inspect the value of
15343 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
15344 location @code{pc - 12} and the top 8 bits are set, then we know that
15345 there is a function name embedded immediately preceding this location
15346 and has length @code{((pc[-3]) & 0xff000000)}.
15353 Select between generating code that executes in ARM and Thumb
15354 states. The default for most configurations is to generate code
15355 that executes in ARM state, but the default can be changed by
15356 configuring GCC with the @option{--with-mode=}@var{state}
15359 You can also override the ARM and Thumb mode for each function
15360 by using the @code{target("thumb")} and @code{target("arm")} function attributes
15361 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
15364 @opindex mtpcs-frame
15365 Generate a stack frame that is compliant with the Thumb Procedure Call
15366 Standard for all non-leaf functions. (A leaf function is one that does
15367 not call any other functions.) The default is @option{-mno-tpcs-frame}.
15369 @item -mtpcs-leaf-frame
15370 @opindex mtpcs-leaf-frame
15371 Generate a stack frame that is compliant with the Thumb Procedure Call
15372 Standard for all leaf functions. (A leaf function is one that does
15373 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
15375 @item -mcallee-super-interworking
15376 @opindex mcallee-super-interworking
15377 Gives all externally visible functions in the file being compiled an ARM
15378 instruction set header which switches to Thumb mode before executing the
15379 rest of the function. This allows these functions to be called from
15380 non-interworking code. This option is not valid in AAPCS configurations
15381 because interworking is enabled by default.
15383 @item -mcaller-super-interworking
15384 @opindex mcaller-super-interworking
15385 Allows calls via function pointers (including virtual functions) to
15386 execute correctly regardless of whether the target code has been
15387 compiled for interworking or not. There is a small overhead in the cost
15388 of executing a function pointer if this option is enabled. This option
15389 is not valid in AAPCS configurations because interworking is enabled
15392 @item -mtp=@var{name}
15394 Specify the access model for the thread local storage pointer. The valid
15395 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
15396 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
15397 (supported in the arm6k architecture), and @samp{auto}, which uses the
15398 best available method for the selected processor. The default setting is
15401 @item -mtls-dialect=@var{dialect}
15402 @opindex mtls-dialect
15403 Specify the dialect to use for accessing thread local storage. Two
15404 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
15405 @samp{gnu} dialect selects the original GNU scheme for supporting
15406 local and global dynamic TLS models. The @samp{gnu2} dialect
15407 selects the GNU descriptor scheme, which provides better performance
15408 for shared libraries. The GNU descriptor scheme is compatible with
15409 the original scheme, but does require new assembler, linker and
15410 library support. Initial and local exec TLS models are unaffected by
15411 this option and always use the original scheme.
15413 @item -mword-relocations
15414 @opindex mword-relocations
15415 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
15416 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
15417 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
15420 @item -mfix-cortex-m3-ldrd
15421 @opindex mfix-cortex-m3-ldrd
15422 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
15423 with overlapping destination and base registers are used. This option avoids
15424 generating these instructions. This option is enabled by default when
15425 @option{-mcpu=cortex-m3} is specified.
15427 @item -munaligned-access
15428 @itemx -mno-unaligned-access
15429 @opindex munaligned-access
15430 @opindex mno-unaligned-access
15431 Enables (or disables) reading and writing of 16- and 32- bit values
15432 from addresses that are not 16- or 32- bit aligned. By default
15433 unaligned access is disabled for all pre-ARMv6, all ARMv6-M and for
15434 ARMv8-M Baseline architectures, and enabled for all other
15435 architectures. If unaligned access is not enabled then words in packed
15436 data structures are accessed a byte at a time.
15438 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
15439 generated object file to either true or false, depending upon the
15440 setting of this option. If unaligned access is enabled then the
15441 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
15444 @item -mneon-for-64bits
15445 @opindex mneon-for-64bits
15446 Enables using Neon to handle scalar 64-bits operations. This is
15447 disabled by default since the cost of moving data from core registers
15450 @item -mslow-flash-data
15451 @opindex mslow-flash-data
15452 Assume loading data from flash is slower than fetching instruction.
15453 Therefore literal load is minimized for better performance.
15454 This option is only supported when compiling for ARMv7 M-profile and
15457 @item -masm-syntax-unified
15458 @opindex masm-syntax-unified
15459 Assume inline assembler is using unified asm syntax. The default is
15460 currently off which implies divided syntax. This option has no impact
15461 on Thumb2. However, this may change in future releases of GCC.
15462 Divided syntax should be considered deprecated.
15464 @item -mrestrict-it
15465 @opindex mrestrict-it
15466 Restricts generation of IT blocks to conform to the rules of ARMv8.
15467 IT blocks can only contain a single 16-bit instruction from a select
15468 set of instructions. This option is on by default for ARMv8 Thumb mode.
15470 @item -mprint-tune-info
15471 @opindex mprint-tune-info
15472 Print CPU tuning information as comment in assembler file. This is
15473 an option used only for regression testing of the compiler and not
15474 intended for ordinary use in compiling code. This option is disabled
15478 @opindex mpure-code
15479 Do not allow constant data to be placed in code sections.
15480 Additionally, when compiling for ELF object format give all text sections the
15481 ELF processor-specific section attribute @code{SHF_ARM_PURECODE}. This option
15482 is only available when generating non-pic code for ARMv7-M targets.
15486 Generate secure code as per the "ARMv8-M Security Extensions: Requirements on
15487 Development Tools Engineering Specification", which can be found on
15488 @url{http://infocenter.arm.com/help/topic/com.arm.doc.ecm0359818/ECM0359818_armv8m_security_extensions_reqs_on_dev_tools_1_0.pdf}.
15492 @subsection AVR Options
15493 @cindex AVR Options
15495 These options are defined for AVR implementations:
15498 @item -mmcu=@var{mcu}
15500 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
15502 The default for this option is@tie{}@samp{avr2}.
15504 GCC supports the following AVR devices and ISAs:
15506 @include avr-mmcu.texi
15511 Assume that all data in static storage can be accessed by LDS / STS
15512 instructions. This option has only an effect on reduced Tiny devices like
15513 ATtiny40. See also the @code{absdata}
15514 @ref{AVR Variable Attributes,variable attribute}.
15516 @item -maccumulate-args
15517 @opindex maccumulate-args
15518 Accumulate outgoing function arguments and acquire/release the needed
15519 stack space for outgoing function arguments once in function
15520 prologue/epilogue. Without this option, outgoing arguments are pushed
15521 before calling a function and popped afterwards.
15523 Popping the arguments after the function call can be expensive on
15524 AVR so that accumulating the stack space might lead to smaller
15525 executables because arguments need not to be removed from the
15526 stack after such a function call.
15528 This option can lead to reduced code size for functions that perform
15529 several calls to functions that get their arguments on the stack like
15530 calls to printf-like functions.
15532 @item -mbranch-cost=@var{cost}
15533 @opindex mbranch-cost
15534 Set the branch costs for conditional branch instructions to
15535 @var{cost}. Reasonable values for @var{cost} are small, non-negative
15536 integers. The default branch cost is 0.
15538 @item -mcall-prologues
15539 @opindex mcall-prologues
15540 Functions prologues/epilogues are expanded as calls to appropriate
15541 subroutines. Code size is smaller.
15545 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
15546 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
15547 and @code{long long} is 4 bytes. Please note that this option does not
15548 conform to the C standards, but it results in smaller code
15551 @item -mn-flash=@var{num}
15553 Assume that the flash memory has a size of
15554 @var{num} times 64@tie{}KiB.
15556 @item -mno-interrupts
15557 @opindex mno-interrupts
15558 Generated code is not compatible with hardware interrupts.
15559 Code size is smaller.
15563 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
15564 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
15565 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
15566 the assembler's command line and the @option{--relax} option to the
15567 linker's command line.
15569 Jump relaxing is performed by the linker because jump offsets are not
15570 known before code is located. Therefore, the assembler code generated by the
15571 compiler is the same, but the instructions in the executable may
15572 differ from instructions in the assembler code.
15574 Relaxing must be turned on if linker stubs are needed, see the
15575 section on @code{EIND} and linker stubs below.
15579 Assume that the device supports the Read-Modify-Write
15580 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
15584 Treat the stack pointer register as an 8-bit register,
15585 i.e.@: assume the high byte of the stack pointer is zero.
15586 In general, you don't need to set this option by hand.
15588 This option is used internally by the compiler to select and
15589 build multilibs for architectures @code{avr2} and @code{avr25}.
15590 These architectures mix devices with and without @code{SPH}.
15591 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
15592 the compiler driver adds or removes this option from the compiler
15593 proper's command line, because the compiler then knows if the device
15594 or architecture has an 8-bit stack pointer and thus no @code{SPH}
15599 Use address register @code{X} in a way proposed by the hardware. This means
15600 that @code{X} is only used in indirect, post-increment or
15601 pre-decrement addressing.
15603 Without this option, the @code{X} register may be used in the same way
15604 as @code{Y} or @code{Z} which then is emulated by additional
15606 For example, loading a value with @code{X+const} addressing with a
15607 small non-negative @code{const < 64} to a register @var{Rn} is
15611 adiw r26, const ; X += const
15612 ld @var{Rn}, X ; @var{Rn} = *X
15613 sbiw r26, const ; X -= const
15617 @opindex mtiny-stack
15618 Only change the lower 8@tie{}bits of the stack pointer.
15620 @item -mfract-convert-truncate
15621 @opindex mfract-convert-truncate
15622 Allow to use truncation instead of rounding towards zero for fractional fixed-point types.
15625 @opindex nodevicelib
15626 Don't link against AVR-LibC's device specific library @code{lib<mcu>.a}.
15628 @item -Waddr-space-convert
15629 @opindex Waddr-space-convert
15630 Warn about conversions between address spaces in the case where the
15631 resulting address space is not contained in the incoming address space.
15633 @item -Wmisspelled-isr
15634 @opindex Wmisspelled-isr
15635 Warn if the ISR is misspelled, i.e. without __vector prefix.
15636 Enabled by default.
15639 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
15640 @cindex @code{EIND}
15641 Pointers in the implementation are 16@tie{}bits wide.
15642 The address of a function or label is represented as word address so
15643 that indirect jumps and calls can target any code address in the
15644 range of 64@tie{}Ki words.
15646 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
15647 bytes of program memory space, there is a special function register called
15648 @code{EIND} that serves as most significant part of the target address
15649 when @code{EICALL} or @code{EIJMP} instructions are used.
15651 Indirect jumps and calls on these devices are handled as follows by
15652 the compiler and are subject to some limitations:
15657 The compiler never sets @code{EIND}.
15660 The compiler uses @code{EIND} implicitly in @code{EICALL}/@code{EIJMP}
15661 instructions or might read @code{EIND} directly in order to emulate an
15662 indirect call/jump by means of a @code{RET} instruction.
15665 The compiler assumes that @code{EIND} never changes during the startup
15666 code or during the application. In particular, @code{EIND} is not
15667 saved/restored in function or interrupt service routine
15671 For indirect calls to functions and computed goto, the linker
15672 generates @emph{stubs}. Stubs are jump pads sometimes also called
15673 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
15674 The stub contains a direct jump to the desired address.
15677 Linker relaxation must be turned on so that the linker generates
15678 the stubs correctly in all situations. See the compiler option
15679 @option{-mrelax} and the linker option @option{--relax}.
15680 There are corner cases where the linker is supposed to generate stubs
15681 but aborts without relaxation and without a helpful error message.
15684 The default linker script is arranged for code with @code{EIND = 0}.
15685 If code is supposed to work for a setup with @code{EIND != 0}, a custom
15686 linker script has to be used in order to place the sections whose
15687 name start with @code{.trampolines} into the segment where @code{EIND}
15691 The startup code from libgcc never sets @code{EIND}.
15692 Notice that startup code is a blend of code from libgcc and AVR-LibC.
15693 For the impact of AVR-LibC on @code{EIND}, see the
15694 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
15697 It is legitimate for user-specific startup code to set up @code{EIND}
15698 early, for example by means of initialization code located in
15699 section @code{.init3}. Such code runs prior to general startup code
15700 that initializes RAM and calls constructors, but after the bit
15701 of startup code from AVR-LibC that sets @code{EIND} to the segment
15702 where the vector table is located.
15704 #include <avr/io.h>
15707 __attribute__((section(".init3"),naked,used,no_instrument_function))
15708 init3_set_eind (void)
15710 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
15711 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
15716 The @code{__trampolines_start} symbol is defined in the linker script.
15719 Stubs are generated automatically by the linker if
15720 the following two conditions are met:
15723 @item The address of a label is taken by means of the @code{gs} modifier
15724 (short for @emph{generate stubs}) like so:
15726 LDI r24, lo8(gs(@var{func}))
15727 LDI r25, hi8(gs(@var{func}))
15729 @item The final location of that label is in a code segment
15730 @emph{outside} the segment where the stubs are located.
15734 The compiler emits such @code{gs} modifiers for code labels in the
15735 following situations:
15737 @item Taking address of a function or code label.
15738 @item Computed goto.
15739 @item If prologue-save function is used, see @option{-mcall-prologues}
15740 command-line option.
15741 @item Switch/case dispatch tables. If you do not want such dispatch
15742 tables you can specify the @option{-fno-jump-tables} command-line option.
15743 @item C and C++ constructors/destructors called during startup/shutdown.
15744 @item If the tools hit a @code{gs()} modifier explained above.
15748 Jumping to non-symbolic addresses like so is @emph{not} supported:
15753 /* Call function at word address 0x2 */
15754 return ((int(*)(void)) 0x2)();
15758 Instead, a stub has to be set up, i.e.@: the function has to be called
15759 through a symbol (@code{func_4} in the example):
15764 extern int func_4 (void);
15766 /* Call function at byte address 0x4 */
15771 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
15772 Alternatively, @code{func_4} can be defined in the linker script.
15775 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
15776 @cindex @code{RAMPD}
15777 @cindex @code{RAMPX}
15778 @cindex @code{RAMPY}
15779 @cindex @code{RAMPZ}
15780 Some AVR devices support memories larger than the 64@tie{}KiB range
15781 that can be accessed with 16-bit pointers. To access memory locations
15782 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
15783 register is used as high part of the address:
15784 The @code{X}, @code{Y}, @code{Z} address register is concatenated
15785 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
15786 register, respectively, to get a wide address. Similarly,
15787 @code{RAMPD} is used together with direct addressing.
15791 The startup code initializes the @code{RAMP} special function
15792 registers with zero.
15795 If a @ref{AVR Named Address Spaces,named address space} other than
15796 generic or @code{__flash} is used, then @code{RAMPZ} is set
15797 as needed before the operation.
15800 If the device supports RAM larger than 64@tie{}KiB and the compiler
15801 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
15802 is reset to zero after the operation.
15805 If the device comes with a specific @code{RAMP} register, the ISR
15806 prologue/epilogue saves/restores that SFR and initializes it with
15807 zero in case the ISR code might (implicitly) use it.
15810 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
15811 If you use inline assembler to read from locations outside the
15812 16-bit address range and change one of the @code{RAMP} registers,
15813 you must reset it to zero after the access.
15817 @subsubsection AVR Built-in Macros
15819 GCC defines several built-in macros so that the user code can test
15820 for the presence or absence of features. Almost any of the following
15821 built-in macros are deduced from device capabilities and thus
15822 triggered by the @option{-mmcu=} command-line option.
15824 For even more AVR-specific built-in macros see
15825 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
15830 Build-in macro that resolves to a decimal number that identifies the
15831 architecture and depends on the @option{-mmcu=@var{mcu}} option.
15832 Possible values are:
15834 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
15835 @code{4}, @code{5}, @code{51}, @code{6}
15837 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
15838 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
15842 @code{100}, @code{102}, @code{104},
15843 @code{105}, @code{106}, @code{107}
15845 for @var{mcu}=@code{avrtiny}, @code{avrxmega2}, @code{avrxmega4},
15846 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
15847 If @var{mcu} specifies a device, this built-in macro is set
15848 accordingly. For example, with @option{-mmcu=atmega8} the macro is
15849 defined to @code{4}.
15851 @item __AVR_@var{Device}__
15852 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
15853 the device's name. For example, @option{-mmcu=atmega8} defines the
15854 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
15855 @code{__AVR_ATtiny261A__}, etc.
15857 The built-in macros' names follow
15858 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
15859 the device name as from the AVR user manual. The difference between
15860 @var{Device} in the built-in macro and @var{device} in
15861 @option{-mmcu=@var{device}} is that the latter is always lowercase.
15863 If @var{device} is not a device but only a core architecture like
15864 @samp{avr51}, this macro is not defined.
15866 @item __AVR_DEVICE_NAME__
15867 Setting @option{-mmcu=@var{device}} defines this built-in macro to
15868 the device's name. For example, with @option{-mmcu=atmega8} the macro
15869 is defined to @code{atmega8}.
15871 If @var{device} is not a device but only a core architecture like
15872 @samp{avr51}, this macro is not defined.
15874 @item __AVR_XMEGA__
15875 The device / architecture belongs to the XMEGA family of devices.
15877 @item __AVR_HAVE_ELPM__
15878 The device has the @code{ELPM} instruction.
15880 @item __AVR_HAVE_ELPMX__
15881 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
15882 R@var{n},Z+} instructions.
15884 @item __AVR_HAVE_MOVW__
15885 The device has the @code{MOVW} instruction to perform 16-bit
15886 register-register moves.
15888 @item __AVR_HAVE_LPMX__
15889 The device has the @code{LPM R@var{n},Z} and
15890 @code{LPM R@var{n},Z+} instructions.
15892 @item __AVR_HAVE_MUL__
15893 The device has a hardware multiplier.
15895 @item __AVR_HAVE_JMP_CALL__
15896 The device has the @code{JMP} and @code{CALL} instructions.
15897 This is the case for devices with at least 16@tie{}KiB of program
15900 @item __AVR_HAVE_EIJMP_EICALL__
15901 @itemx __AVR_3_BYTE_PC__
15902 The device has the @code{EIJMP} and @code{EICALL} instructions.
15903 This is the case for devices with more than 128@tie{}KiB of program memory.
15904 This also means that the program counter
15905 (PC) is 3@tie{}bytes wide.
15907 @item __AVR_2_BYTE_PC__
15908 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
15909 with up to 128@tie{}KiB of program memory.
15911 @item __AVR_HAVE_8BIT_SP__
15912 @itemx __AVR_HAVE_16BIT_SP__
15913 The stack pointer (SP) register is treated as 8-bit respectively
15914 16-bit register by the compiler.
15915 The definition of these macros is affected by @option{-mtiny-stack}.
15917 @item __AVR_HAVE_SPH__
15919 The device has the SPH (high part of stack pointer) special function
15920 register or has an 8-bit stack pointer, respectively.
15921 The definition of these macros is affected by @option{-mmcu=} and
15922 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
15925 @item __AVR_HAVE_RAMPD__
15926 @itemx __AVR_HAVE_RAMPX__
15927 @itemx __AVR_HAVE_RAMPY__
15928 @itemx __AVR_HAVE_RAMPZ__
15929 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
15930 @code{RAMPZ} special function register, respectively.
15932 @item __NO_INTERRUPTS__
15933 This macro reflects the @option{-mno-interrupts} command-line option.
15935 @item __AVR_ERRATA_SKIP__
15936 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
15937 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
15938 instructions because of a hardware erratum. Skip instructions are
15939 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
15940 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
15943 @item __AVR_ISA_RMW__
15944 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
15946 @item __AVR_SFR_OFFSET__=@var{offset}
15947 Instructions that can address I/O special function registers directly
15948 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
15949 address as if addressed by an instruction to access RAM like @code{LD}
15950 or @code{STS}. This offset depends on the device architecture and has
15951 to be subtracted from the RAM address in order to get the
15952 respective I/O@tie{}address.
15954 @item __WITH_AVRLIBC__
15955 The compiler is configured to be used together with AVR-Libc.
15956 See the @option{--with-avrlibc} configure option.
15960 @node Blackfin Options
15961 @subsection Blackfin Options
15962 @cindex Blackfin Options
15965 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
15967 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
15968 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
15969 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
15970 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
15971 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
15972 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
15973 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
15974 @samp{bf561}, @samp{bf592}.
15976 The optional @var{sirevision} specifies the silicon revision of the target
15977 Blackfin processor. Any workarounds available for the targeted silicon revision
15978 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
15979 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
15980 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
15981 hexadecimal digits representing the major and minor numbers in the silicon
15982 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
15983 is not defined. If @var{sirevision} is @samp{any}, the
15984 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
15985 If this optional @var{sirevision} is not used, GCC assumes the latest known
15986 silicon revision of the targeted Blackfin processor.
15988 GCC defines a preprocessor macro for the specified @var{cpu}.
15989 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
15990 provided by libgloss to be linked in if @option{-msim} is not given.
15992 Without this option, @samp{bf532} is used as the processor by default.
15994 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
15995 only the preprocessor macro is defined.
15999 Specifies that the program will be run on the simulator. This causes
16000 the simulator BSP provided by libgloss to be linked in. This option
16001 has effect only for @samp{bfin-elf} toolchain.
16002 Certain other options, such as @option{-mid-shared-library} and
16003 @option{-mfdpic}, imply @option{-msim}.
16005 @item -momit-leaf-frame-pointer
16006 @opindex momit-leaf-frame-pointer
16007 Don't keep the frame pointer in a register for leaf functions. This
16008 avoids the instructions to save, set up and restore frame pointers and
16009 makes an extra register available in leaf functions. The option
16010 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
16011 which might make debugging harder.
16013 @item -mspecld-anomaly
16014 @opindex mspecld-anomaly
16015 When enabled, the compiler ensures that the generated code does not
16016 contain speculative loads after jump instructions. If this option is used,
16017 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
16019 @item -mno-specld-anomaly
16020 @opindex mno-specld-anomaly
16021 Don't generate extra code to prevent speculative loads from occurring.
16023 @item -mcsync-anomaly
16024 @opindex mcsync-anomaly
16025 When enabled, the compiler ensures that the generated code does not
16026 contain CSYNC or SSYNC instructions too soon after conditional branches.
16027 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
16029 @item -mno-csync-anomaly
16030 @opindex mno-csync-anomaly
16031 Don't generate extra code to prevent CSYNC or SSYNC instructions from
16032 occurring too soon after a conditional branch.
16036 When enabled, the compiler is free to take advantage of the knowledge that
16037 the entire program fits into the low 64k of memory.
16040 @opindex mno-low-64k
16041 Assume that the program is arbitrarily large. This is the default.
16043 @item -mstack-check-l1
16044 @opindex mstack-check-l1
16045 Do stack checking using information placed into L1 scratchpad memory by the
16048 @item -mid-shared-library
16049 @opindex mid-shared-library
16050 Generate code that supports shared libraries via the library ID method.
16051 This allows for execute in place and shared libraries in an environment
16052 without virtual memory management. This option implies @option{-fPIC}.
16053 With a @samp{bfin-elf} target, this option implies @option{-msim}.
16055 @item -mno-id-shared-library
16056 @opindex mno-id-shared-library
16057 Generate code that doesn't assume ID-based shared libraries are being used.
16058 This is the default.
16060 @item -mleaf-id-shared-library
16061 @opindex mleaf-id-shared-library
16062 Generate code that supports shared libraries via the library ID method,
16063 but assumes that this library or executable won't link against any other
16064 ID shared libraries. That allows the compiler to use faster code for jumps
16067 @item -mno-leaf-id-shared-library
16068 @opindex mno-leaf-id-shared-library
16069 Do not assume that the code being compiled won't link against any ID shared
16070 libraries. Slower code is generated for jump and call insns.
16072 @item -mshared-library-id=n
16073 @opindex mshared-library-id
16074 Specifies the identification number of the ID-based shared library being
16075 compiled. Specifying a value of 0 generates more compact code; specifying
16076 other values forces the allocation of that number to the current
16077 library but is no more space- or time-efficient than omitting this option.
16081 Generate code that allows the data segment to be located in a different
16082 area of memory from the text segment. This allows for execute in place in
16083 an environment without virtual memory management by eliminating relocations
16084 against the text section.
16086 @item -mno-sep-data
16087 @opindex mno-sep-data
16088 Generate code that assumes that the data segment follows the text segment.
16089 This is the default.
16092 @itemx -mno-long-calls
16093 @opindex mlong-calls
16094 @opindex mno-long-calls
16095 Tells the compiler to perform function calls by first loading the
16096 address of the function into a register and then performing a subroutine
16097 call on this register. This switch is needed if the target function
16098 lies outside of the 24-bit addressing range of the offset-based
16099 version of subroutine call instruction.
16101 This feature is not enabled by default. Specifying
16102 @option{-mno-long-calls} restores the default behavior. Note these
16103 switches have no effect on how the compiler generates code to handle
16104 function calls via function pointers.
16108 Link with the fast floating-point library. This library relaxes some of
16109 the IEEE floating-point standard's rules for checking inputs against
16110 Not-a-Number (NAN), in the interest of performance.
16113 @opindex minline-plt
16114 Enable inlining of PLT entries in function calls to functions that are
16115 not known to bind locally. It has no effect without @option{-mfdpic}.
16118 @opindex mmulticore
16119 Build a standalone application for multicore Blackfin processors.
16120 This option causes proper start files and link scripts supporting
16121 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
16122 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
16124 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
16125 selects the one-application-per-core programming model. Without
16126 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
16127 programming model is used. In this model, the main function of Core B
16128 should be named as @code{coreb_main}.
16130 If this option is not used, the single-core application programming
16135 Build a standalone application for Core A of BF561 when using
16136 the one-application-per-core programming model. Proper start files
16137 and link scripts are used to support Core A, and the macro
16138 @code{__BFIN_COREA} is defined.
16139 This option can only be used in conjunction with @option{-mmulticore}.
16143 Build a standalone application for Core B of BF561 when using
16144 the one-application-per-core programming model. Proper start files
16145 and link scripts are used to support Core B, and the macro
16146 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
16147 should be used instead of @code{main}.
16148 This option can only be used in conjunction with @option{-mmulticore}.
16152 Build a standalone application for SDRAM. Proper start files and
16153 link scripts are used to put the application into SDRAM, and the macro
16154 @code{__BFIN_SDRAM} is defined.
16155 The loader should initialize SDRAM before loading the application.
16159 Assume that ICPLBs are enabled at run time. This has an effect on certain
16160 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
16161 are enabled; for standalone applications the default is off.
16165 @subsection C6X Options
16166 @cindex C6X Options
16169 @item -march=@var{name}
16171 This specifies the name of the target architecture. GCC uses this
16172 name to determine what kind of instructions it can emit when generating
16173 assembly code. Permissible names are: @samp{c62x},
16174 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
16177 @opindex mbig-endian
16178 Generate code for a big-endian target.
16180 @item -mlittle-endian
16181 @opindex mlittle-endian
16182 Generate code for a little-endian target. This is the default.
16186 Choose startup files and linker script suitable for the simulator.
16188 @item -msdata=default
16189 @opindex msdata=default
16190 Put small global and static data in the @code{.neardata} section,
16191 which is pointed to by register @code{B14}. Put small uninitialized
16192 global and static data in the @code{.bss} section, which is adjacent
16193 to the @code{.neardata} section. Put small read-only data into the
16194 @code{.rodata} section. The corresponding sections used for large
16195 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
16198 @opindex msdata=all
16199 Put all data, not just small objects, into the sections reserved for
16200 small data, and use addressing relative to the @code{B14} register to
16204 @opindex msdata=none
16205 Make no use of the sections reserved for small data, and use absolute
16206 addresses to access all data. Put all initialized global and static
16207 data in the @code{.fardata} section, and all uninitialized data in the
16208 @code{.far} section. Put all constant data into the @code{.const}
16213 @subsection CRIS Options
16214 @cindex CRIS Options
16216 These options are defined specifically for the CRIS ports.
16219 @item -march=@var{architecture-type}
16220 @itemx -mcpu=@var{architecture-type}
16223 Generate code for the specified architecture. The choices for
16224 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
16225 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
16226 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
16229 @item -mtune=@var{architecture-type}
16231 Tune to @var{architecture-type} everything applicable about the generated
16232 code, except for the ABI and the set of available instructions. The
16233 choices for @var{architecture-type} are the same as for
16234 @option{-march=@var{architecture-type}}.
16236 @item -mmax-stack-frame=@var{n}
16237 @opindex mmax-stack-frame
16238 Warn when the stack frame of a function exceeds @var{n} bytes.
16244 The options @option{-metrax4} and @option{-metrax100} are synonyms for
16245 @option{-march=v3} and @option{-march=v8} respectively.
16247 @item -mmul-bug-workaround
16248 @itemx -mno-mul-bug-workaround
16249 @opindex mmul-bug-workaround
16250 @opindex mno-mul-bug-workaround
16251 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
16252 models where it applies. This option is active by default.
16256 Enable CRIS-specific verbose debug-related information in the assembly
16257 code. This option also has the effect of turning off the @samp{#NO_APP}
16258 formatted-code indicator to the assembler at the beginning of the
16263 Do not use condition-code results from previous instruction; always emit
16264 compare and test instructions before use of condition codes.
16266 @item -mno-side-effects
16267 @opindex mno-side-effects
16268 Do not emit instructions with side effects in addressing modes other than
16271 @item -mstack-align
16272 @itemx -mno-stack-align
16273 @itemx -mdata-align
16274 @itemx -mno-data-align
16275 @itemx -mconst-align
16276 @itemx -mno-const-align
16277 @opindex mstack-align
16278 @opindex mno-stack-align
16279 @opindex mdata-align
16280 @opindex mno-data-align
16281 @opindex mconst-align
16282 @opindex mno-const-align
16283 These options (@samp{no-} options) arrange (eliminate arrangements) for the
16284 stack frame, individual data and constants to be aligned for the maximum
16285 single data access size for the chosen CPU model. The default is to
16286 arrange for 32-bit alignment. ABI details such as structure layout are
16287 not affected by these options.
16295 Similar to the stack- data- and const-align options above, these options
16296 arrange for stack frame, writable data and constants to all be 32-bit,
16297 16-bit or 8-bit aligned. The default is 32-bit alignment.
16299 @item -mno-prologue-epilogue
16300 @itemx -mprologue-epilogue
16301 @opindex mno-prologue-epilogue
16302 @opindex mprologue-epilogue
16303 With @option{-mno-prologue-epilogue}, the normal function prologue and
16304 epilogue which set up the stack frame are omitted and no return
16305 instructions or return sequences are generated in the code. Use this
16306 option only together with visual inspection of the compiled code: no
16307 warnings or errors are generated when call-saved registers must be saved,
16308 or storage for local variables needs to be allocated.
16312 @opindex mno-gotplt
16314 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
16315 instruction sequences that load addresses for functions from the PLT part
16316 of the GOT rather than (traditional on other architectures) calls to the
16317 PLT@. The default is @option{-mgotplt}.
16321 Legacy no-op option only recognized with the cris-axis-elf and
16322 cris-axis-linux-gnu targets.
16326 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
16330 This option, recognized for the cris-axis-elf, arranges
16331 to link with input-output functions from a simulator library. Code,
16332 initialized data and zero-initialized data are allocated consecutively.
16336 Like @option{-sim}, but pass linker options to locate initialized data at
16337 0x40000000 and zero-initialized data at 0x80000000.
16341 @subsection CR16 Options
16342 @cindex CR16 Options
16344 These options are defined specifically for the CR16 ports.
16350 Enable the use of multiply-accumulate instructions. Disabled by default.
16354 @opindex mcr16cplus
16356 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
16361 Links the library libsim.a which is in compatible with simulator. Applicable
16362 to ELF compiler only.
16366 Choose integer type as 32-bit wide.
16370 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
16372 @item -mdata-model=@var{model}
16373 @opindex mdata-model
16374 Choose a data model. The choices for @var{model} are @samp{near},
16375 @samp{far} or @samp{medium}. @samp{medium} is default.
16376 However, @samp{far} is not valid with @option{-mcr16c}, as the
16377 CR16C architecture does not support the far data model.
16380 @node Darwin Options
16381 @subsection Darwin Options
16382 @cindex Darwin options
16384 These options are defined for all architectures running the Darwin operating
16387 FSF GCC on Darwin does not create ``fat'' object files; it creates
16388 an object file for the single architecture that GCC was built to
16389 target. Apple's GCC on Darwin does create ``fat'' files if multiple
16390 @option{-arch} options are used; it does so by running the compiler or
16391 linker multiple times and joining the results together with
16394 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
16395 @samp{i686}) is determined by the flags that specify the ISA
16396 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
16397 @option{-force_cpusubtype_ALL} option can be used to override this.
16399 The Darwin tools vary in their behavior when presented with an ISA
16400 mismatch. The assembler, @file{as}, only permits instructions to
16401 be used that are valid for the subtype of the file it is generating,
16402 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
16403 The linker for shared libraries, @file{/usr/bin/libtool}, fails
16404 and prints an error if asked to create a shared library with a less
16405 restrictive subtype than its input files (for instance, trying to put
16406 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
16407 for executables, @command{ld}, quietly gives the executable the most
16408 restrictive subtype of any of its input files.
16413 Add the framework directory @var{dir} to the head of the list of
16414 directories to be searched for header files. These directories are
16415 interleaved with those specified by @option{-I} options and are
16416 scanned in a left-to-right order.
16418 A framework directory is a directory with frameworks in it. A
16419 framework is a directory with a @file{Headers} and/or
16420 @file{PrivateHeaders} directory contained directly in it that ends
16421 in @file{.framework}. The name of a framework is the name of this
16422 directory excluding the @file{.framework}. Headers associated with
16423 the framework are found in one of those two directories, with
16424 @file{Headers} being searched first. A subframework is a framework
16425 directory that is in a framework's @file{Frameworks} directory.
16426 Includes of subframework headers can only appear in a header of a
16427 framework that contains the subframework, or in a sibling subframework
16428 header. Two subframeworks are siblings if they occur in the same
16429 framework. A subframework should not have the same name as a
16430 framework; a warning is issued if this is violated. Currently a
16431 subframework cannot have subframeworks; in the future, the mechanism
16432 may be extended to support this. The standard frameworks can be found
16433 in @file{/System/Library/Frameworks} and
16434 @file{/Library/Frameworks}. An example include looks like
16435 @code{#include <Framework/header.h>}, where @file{Framework} denotes
16436 the name of the framework and @file{header.h} is found in the
16437 @file{PrivateHeaders} or @file{Headers} directory.
16439 @item -iframework@var{dir}
16440 @opindex iframework
16441 Like @option{-F} except the directory is a treated as a system
16442 directory. The main difference between this @option{-iframework} and
16443 @option{-F} is that with @option{-iframework} the compiler does not
16444 warn about constructs contained within header files found via
16445 @var{dir}. This option is valid only for the C family of languages.
16449 Emit debugging information for symbols that are used. For stabs
16450 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
16451 This is by default ON@.
16455 Emit debugging information for all symbols and types.
16457 @item -mmacosx-version-min=@var{version}
16458 The earliest version of MacOS X that this executable will run on
16459 is @var{version}. Typical values of @var{version} include @code{10.1},
16460 @code{10.2}, and @code{10.3.9}.
16462 If the compiler was built to use the system's headers by default,
16463 then the default for this option is the system version on which the
16464 compiler is running, otherwise the default is to make choices that
16465 are compatible with as many systems and code bases as possible.
16469 Enable kernel development mode. The @option{-mkernel} option sets
16470 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
16471 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
16472 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
16473 applicable. This mode also sets @option{-mno-altivec},
16474 @option{-msoft-float}, @option{-fno-builtin} and
16475 @option{-mlong-branch} for PowerPC targets.
16477 @item -mone-byte-bool
16478 @opindex mone-byte-bool
16479 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
16480 By default @code{sizeof(bool)} is @code{4} when compiling for
16481 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
16482 option has no effect on x86.
16484 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
16485 to generate code that is not binary compatible with code generated
16486 without that switch. Using this switch may require recompiling all
16487 other modules in a program, including system libraries. Use this
16488 switch to conform to a non-default data model.
16490 @item -mfix-and-continue
16491 @itemx -ffix-and-continue
16492 @itemx -findirect-data
16493 @opindex mfix-and-continue
16494 @opindex ffix-and-continue
16495 @opindex findirect-data
16496 Generate code suitable for fast turnaround development, such as to
16497 allow GDB to dynamically load @file{.o} files into already-running
16498 programs. @option{-findirect-data} and @option{-ffix-and-continue}
16499 are provided for backwards compatibility.
16503 Loads all members of static archive libraries.
16504 See man ld(1) for more information.
16506 @item -arch_errors_fatal
16507 @opindex arch_errors_fatal
16508 Cause the errors having to do with files that have the wrong architecture
16511 @item -bind_at_load
16512 @opindex bind_at_load
16513 Causes the output file to be marked such that the dynamic linker will
16514 bind all undefined references when the file is loaded or launched.
16518 Produce a Mach-o bundle format file.
16519 See man ld(1) for more information.
16521 @item -bundle_loader @var{executable}
16522 @opindex bundle_loader
16523 This option specifies the @var{executable} that will load the build
16524 output file being linked. See man ld(1) for more information.
16527 @opindex dynamiclib
16528 When passed this option, GCC produces a dynamic library instead of
16529 an executable when linking, using the Darwin @file{libtool} command.
16531 @item -force_cpusubtype_ALL
16532 @opindex force_cpusubtype_ALL
16533 This causes GCC's output file to have the @samp{ALL} subtype, instead of
16534 one controlled by the @option{-mcpu} or @option{-march} option.
16536 @item -allowable_client @var{client_name}
16537 @itemx -client_name
16538 @itemx -compatibility_version
16539 @itemx -current_version
16541 @itemx -dependency-file
16543 @itemx -dylinker_install_name
16545 @itemx -exported_symbols_list
16548 @itemx -flat_namespace
16549 @itemx -force_flat_namespace
16550 @itemx -headerpad_max_install_names
16553 @itemx -install_name
16554 @itemx -keep_private_externs
16555 @itemx -multi_module
16556 @itemx -multiply_defined
16557 @itemx -multiply_defined_unused
16560 @itemx -no_dead_strip_inits_and_terms
16561 @itemx -nofixprebinding
16562 @itemx -nomultidefs
16564 @itemx -noseglinkedit
16565 @itemx -pagezero_size
16567 @itemx -prebind_all_twolevel_modules
16568 @itemx -private_bundle
16570 @itemx -read_only_relocs
16572 @itemx -sectobjectsymbols
16576 @itemx -sectobjectsymbols
16579 @itemx -segs_read_only_addr
16581 @itemx -segs_read_write_addr
16582 @itemx -seg_addr_table
16583 @itemx -seg_addr_table_filename
16584 @itemx -seglinkedit
16586 @itemx -segs_read_only_addr
16587 @itemx -segs_read_write_addr
16588 @itemx -single_module
16590 @itemx -sub_library
16592 @itemx -sub_umbrella
16593 @itemx -twolevel_namespace
16596 @itemx -unexported_symbols_list
16597 @itemx -weak_reference_mismatches
16598 @itemx -whatsloaded
16599 @opindex allowable_client
16600 @opindex client_name
16601 @opindex compatibility_version
16602 @opindex current_version
16603 @opindex dead_strip
16604 @opindex dependency-file
16605 @opindex dylib_file
16606 @opindex dylinker_install_name
16608 @opindex exported_symbols_list
16610 @opindex flat_namespace
16611 @opindex force_flat_namespace
16612 @opindex headerpad_max_install_names
16613 @opindex image_base
16615 @opindex install_name
16616 @opindex keep_private_externs
16617 @opindex multi_module
16618 @opindex multiply_defined
16619 @opindex multiply_defined_unused
16620 @opindex noall_load
16621 @opindex no_dead_strip_inits_and_terms
16622 @opindex nofixprebinding
16623 @opindex nomultidefs
16625 @opindex noseglinkedit
16626 @opindex pagezero_size
16628 @opindex prebind_all_twolevel_modules
16629 @opindex private_bundle
16630 @opindex read_only_relocs
16632 @opindex sectobjectsymbols
16635 @opindex sectcreate
16636 @opindex sectobjectsymbols
16639 @opindex segs_read_only_addr
16640 @opindex segs_read_write_addr
16641 @opindex seg_addr_table
16642 @opindex seg_addr_table_filename
16643 @opindex seglinkedit
16645 @opindex segs_read_only_addr
16646 @opindex segs_read_write_addr
16647 @opindex single_module
16649 @opindex sub_library
16650 @opindex sub_umbrella
16651 @opindex twolevel_namespace
16654 @opindex unexported_symbols_list
16655 @opindex weak_reference_mismatches
16656 @opindex whatsloaded
16657 These options are passed to the Darwin linker. The Darwin linker man page
16658 describes them in detail.
16661 @node DEC Alpha Options
16662 @subsection DEC Alpha Options
16664 These @samp{-m} options are defined for the DEC Alpha implementations:
16667 @item -mno-soft-float
16668 @itemx -msoft-float
16669 @opindex mno-soft-float
16670 @opindex msoft-float
16671 Use (do not use) the hardware floating-point instructions for
16672 floating-point operations. When @option{-msoft-float} is specified,
16673 functions in @file{libgcc.a} are used to perform floating-point
16674 operations. Unless they are replaced by routines that emulate the
16675 floating-point operations, or compiled in such a way as to call such
16676 emulations routines, these routines issue floating-point
16677 operations. If you are compiling for an Alpha without floating-point
16678 operations, you must ensure that the library is built so as not to call
16681 Note that Alpha implementations without floating-point operations are
16682 required to have floating-point registers.
16685 @itemx -mno-fp-regs
16687 @opindex mno-fp-regs
16688 Generate code that uses (does not use) the floating-point register set.
16689 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
16690 register set is not used, floating-point operands are passed in integer
16691 registers as if they were integers and floating-point results are passed
16692 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
16693 so any function with a floating-point argument or return value called by code
16694 compiled with @option{-mno-fp-regs} must also be compiled with that
16697 A typical use of this option is building a kernel that does not use,
16698 and hence need not save and restore, any floating-point registers.
16702 The Alpha architecture implements floating-point hardware optimized for
16703 maximum performance. It is mostly compliant with the IEEE floating-point
16704 standard. However, for full compliance, software assistance is
16705 required. This option generates code fully IEEE-compliant code
16706 @emph{except} that the @var{inexact-flag} is not maintained (see below).
16707 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
16708 defined during compilation. The resulting code is less efficient but is
16709 able to correctly support denormalized numbers and exceptional IEEE
16710 values such as not-a-number and plus/minus infinity. Other Alpha
16711 compilers call this option @option{-ieee_with_no_inexact}.
16713 @item -mieee-with-inexact
16714 @opindex mieee-with-inexact
16715 This is like @option{-mieee} except the generated code also maintains
16716 the IEEE @var{inexact-flag}. Turning on this option causes the
16717 generated code to implement fully-compliant IEEE math. In addition to
16718 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
16719 macro. On some Alpha implementations the resulting code may execute
16720 significantly slower than the code generated by default. Since there is
16721 very little code that depends on the @var{inexact-flag}, you should
16722 normally not specify this option. Other Alpha compilers call this
16723 option @option{-ieee_with_inexact}.
16725 @item -mfp-trap-mode=@var{trap-mode}
16726 @opindex mfp-trap-mode
16727 This option controls what floating-point related traps are enabled.
16728 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
16729 The trap mode can be set to one of four values:
16733 This is the default (normal) setting. The only traps that are enabled
16734 are the ones that cannot be disabled in software (e.g., division by zero
16738 In addition to the traps enabled by @samp{n}, underflow traps are enabled
16742 Like @samp{u}, but the instructions are marked to be safe for software
16743 completion (see Alpha architecture manual for details).
16746 Like @samp{su}, but inexact traps are enabled as well.
16749 @item -mfp-rounding-mode=@var{rounding-mode}
16750 @opindex mfp-rounding-mode
16751 Selects the IEEE rounding mode. Other Alpha compilers call this option
16752 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
16757 Normal IEEE rounding mode. Floating-point numbers are rounded towards
16758 the nearest machine number or towards the even machine number in case
16762 Round towards minus infinity.
16765 Chopped rounding mode. Floating-point numbers are rounded towards zero.
16768 Dynamic rounding mode. A field in the floating-point control register
16769 (@var{fpcr}, see Alpha architecture reference manual) controls the
16770 rounding mode in effect. The C library initializes this register for
16771 rounding towards plus infinity. Thus, unless your program modifies the
16772 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
16775 @item -mtrap-precision=@var{trap-precision}
16776 @opindex mtrap-precision
16777 In the Alpha architecture, floating-point traps are imprecise. This
16778 means without software assistance it is impossible to recover from a
16779 floating trap and program execution normally needs to be terminated.
16780 GCC can generate code that can assist operating system trap handlers
16781 in determining the exact location that caused a floating-point trap.
16782 Depending on the requirements of an application, different levels of
16783 precisions can be selected:
16787 Program precision. This option is the default and means a trap handler
16788 can only identify which program caused a floating-point exception.
16791 Function precision. The trap handler can determine the function that
16792 caused a floating-point exception.
16795 Instruction precision. The trap handler can determine the exact
16796 instruction that caused a floating-point exception.
16799 Other Alpha compilers provide the equivalent options called
16800 @option{-scope_safe} and @option{-resumption_safe}.
16802 @item -mieee-conformant
16803 @opindex mieee-conformant
16804 This option marks the generated code as IEEE conformant. You must not
16805 use this option unless you also specify @option{-mtrap-precision=i} and either
16806 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
16807 is to emit the line @samp{.eflag 48} in the function prologue of the
16808 generated assembly file.
16810 @item -mbuild-constants
16811 @opindex mbuild-constants
16812 Normally GCC examines a 32- or 64-bit integer constant to
16813 see if it can construct it from smaller constants in two or three
16814 instructions. If it cannot, it outputs the constant as a literal and
16815 generates code to load it from the data segment at run time.
16817 Use this option to require GCC to construct @emph{all} integer constants
16818 using code, even if it takes more instructions (the maximum is six).
16820 You typically use this option to build a shared library dynamic
16821 loader. Itself a shared library, it must relocate itself in memory
16822 before it can find the variables and constants in its own data segment.
16840 Indicate whether GCC should generate code to use the optional BWX,
16841 CIX, FIX and MAX instruction sets. The default is to use the instruction
16842 sets supported by the CPU type specified via @option{-mcpu=} option or that
16843 of the CPU on which GCC was built if none is specified.
16846 @itemx -mfloat-ieee
16847 @opindex mfloat-vax
16848 @opindex mfloat-ieee
16849 Generate code that uses (does not use) VAX F and G floating-point
16850 arithmetic instead of IEEE single and double precision.
16852 @item -mexplicit-relocs
16853 @itemx -mno-explicit-relocs
16854 @opindex mexplicit-relocs
16855 @opindex mno-explicit-relocs
16856 Older Alpha assemblers provided no way to generate symbol relocations
16857 except via assembler macros. Use of these macros does not allow
16858 optimal instruction scheduling. GNU binutils as of version 2.12
16859 supports a new syntax that allows the compiler to explicitly mark
16860 which relocations should apply to which instructions. This option
16861 is mostly useful for debugging, as GCC detects the capabilities of
16862 the assembler when it is built and sets the default accordingly.
16865 @itemx -mlarge-data
16866 @opindex msmall-data
16867 @opindex mlarge-data
16868 When @option{-mexplicit-relocs} is in effect, static data is
16869 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
16870 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
16871 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
16872 16-bit relocations off of the @code{$gp} register. This limits the
16873 size of the small data area to 64KB, but allows the variables to be
16874 directly accessed via a single instruction.
16876 The default is @option{-mlarge-data}. With this option the data area
16877 is limited to just below 2GB@. Programs that require more than 2GB of
16878 data must use @code{malloc} or @code{mmap} to allocate the data in the
16879 heap instead of in the program's data segment.
16881 When generating code for shared libraries, @option{-fpic} implies
16882 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
16885 @itemx -mlarge-text
16886 @opindex msmall-text
16887 @opindex mlarge-text
16888 When @option{-msmall-text} is used, the compiler assumes that the
16889 code of the entire program (or shared library) fits in 4MB, and is
16890 thus reachable with a branch instruction. When @option{-msmall-data}
16891 is used, the compiler can assume that all local symbols share the
16892 same @code{$gp} value, and thus reduce the number of instructions
16893 required for a function call from 4 to 1.
16895 The default is @option{-mlarge-text}.
16897 @item -mcpu=@var{cpu_type}
16899 Set the instruction set and instruction scheduling parameters for
16900 machine type @var{cpu_type}. You can specify either the @samp{EV}
16901 style name or the corresponding chip number. GCC supports scheduling
16902 parameters for the EV4, EV5 and EV6 family of processors and
16903 chooses the default values for the instruction set from the processor
16904 you specify. If you do not specify a processor type, GCC defaults
16905 to the processor on which the compiler was built.
16907 Supported values for @var{cpu_type} are
16913 Schedules as an EV4 and has no instruction set extensions.
16917 Schedules as an EV5 and has no instruction set extensions.
16921 Schedules as an EV5 and supports the BWX extension.
16926 Schedules as an EV5 and supports the BWX and MAX extensions.
16930 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
16934 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
16937 Native toolchains also support the value @samp{native},
16938 which selects the best architecture option for the host processor.
16939 @option{-mcpu=native} has no effect if GCC does not recognize
16942 @item -mtune=@var{cpu_type}
16944 Set only the instruction scheduling parameters for machine type
16945 @var{cpu_type}. The instruction set is not changed.
16947 Native toolchains also support the value @samp{native},
16948 which selects the best architecture option for the host processor.
16949 @option{-mtune=native} has no effect if GCC does not recognize
16952 @item -mmemory-latency=@var{time}
16953 @opindex mmemory-latency
16954 Sets the latency the scheduler should assume for typical memory
16955 references as seen by the application. This number is highly
16956 dependent on the memory access patterns used by the application
16957 and the size of the external cache on the machine.
16959 Valid options for @var{time} are
16963 A decimal number representing clock cycles.
16969 The compiler contains estimates of the number of clock cycles for
16970 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
16971 (also called Dcache, Scache, and Bcache), as well as to main memory.
16972 Note that L3 is only valid for EV5.
16978 @subsection FR30 Options
16979 @cindex FR30 Options
16981 These options are defined specifically for the FR30 port.
16985 @item -msmall-model
16986 @opindex msmall-model
16987 Use the small address space model. This can produce smaller code, but
16988 it does assume that all symbolic values and addresses fit into a
16993 Assume that runtime support has been provided and so there is no need
16994 to include the simulator library (@file{libsim.a}) on the linker
17000 @subsection FT32 Options
17001 @cindex FT32 Options
17003 These options are defined specifically for the FT32 port.
17009 Specifies that the program will be run on the simulator. This causes
17010 an alternate runtime startup and library to be linked.
17011 You must not use this option when generating programs that will run on
17012 real hardware; you must provide your own runtime library for whatever
17013 I/O functions are needed.
17017 Enable Local Register Allocation. This is still experimental for FT32,
17018 so by default the compiler uses standard reload.
17022 Do not use div and mod instructions.
17027 @subsection FRV Options
17028 @cindex FRV Options
17034 Only use the first 32 general-purpose registers.
17039 Use all 64 general-purpose registers.
17044 Use only the first 32 floating-point registers.
17049 Use all 64 floating-point registers.
17052 @opindex mhard-float
17054 Use hardware instructions for floating-point operations.
17057 @opindex msoft-float
17059 Use library routines for floating-point operations.
17064 Dynamically allocate condition code registers.
17069 Do not try to dynamically allocate condition code registers, only
17070 use @code{icc0} and @code{fcc0}.
17075 Change ABI to use double word insns.
17080 Do not use double word instructions.
17085 Use floating-point double instructions.
17088 @opindex mno-double
17090 Do not use floating-point double instructions.
17095 Use media instructions.
17100 Do not use media instructions.
17105 Use multiply and add/subtract instructions.
17108 @opindex mno-muladd
17110 Do not use multiply and add/subtract instructions.
17115 Select the FDPIC ABI, which uses function descriptors to represent
17116 pointers to functions. Without any PIC/PIE-related options, it
17117 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
17118 assumes GOT entries and small data are within a 12-bit range from the
17119 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
17120 are computed with 32 bits.
17121 With a @samp{bfin-elf} target, this option implies @option{-msim}.
17124 @opindex minline-plt
17126 Enable inlining of PLT entries in function calls to functions that are
17127 not known to bind locally. It has no effect without @option{-mfdpic}.
17128 It's enabled by default if optimizing for speed and compiling for
17129 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
17130 optimization option such as @option{-O3} or above is present in the
17136 Assume a large TLS segment when generating thread-local code.
17141 Do not assume a large TLS segment when generating thread-local code.
17146 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
17147 that is known to be in read-only sections. It's enabled by default,
17148 except for @option{-fpic} or @option{-fpie}: even though it may help
17149 make the global offset table smaller, it trades 1 instruction for 4.
17150 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
17151 one of which may be shared by multiple symbols, and it avoids the need
17152 for a GOT entry for the referenced symbol, so it's more likely to be a
17153 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
17155 @item -multilib-library-pic
17156 @opindex multilib-library-pic
17158 Link with the (library, not FD) pic libraries. It's implied by
17159 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
17160 @option{-fpic} without @option{-mfdpic}. You should never have to use
17164 @opindex mlinked-fp
17166 Follow the EABI requirement of always creating a frame pointer whenever
17167 a stack frame is allocated. This option is enabled by default and can
17168 be disabled with @option{-mno-linked-fp}.
17171 @opindex mlong-calls
17173 Use indirect addressing to call functions outside the current
17174 compilation unit. This allows the functions to be placed anywhere
17175 within the 32-bit address space.
17177 @item -malign-labels
17178 @opindex malign-labels
17180 Try to align labels to an 8-byte boundary by inserting NOPs into the
17181 previous packet. This option only has an effect when VLIW packing
17182 is enabled. It doesn't create new packets; it merely adds NOPs to
17185 @item -mlibrary-pic
17186 @opindex mlibrary-pic
17188 Generate position-independent EABI code.
17193 Use only the first four media accumulator registers.
17198 Use all eight media accumulator registers.
17203 Pack VLIW instructions.
17208 Do not pack VLIW instructions.
17211 @opindex mno-eflags
17213 Do not mark ABI switches in e_flags.
17216 @opindex mcond-move
17218 Enable the use of conditional-move instructions (default).
17220 This switch is mainly for debugging the compiler and will likely be removed
17221 in a future version.
17223 @item -mno-cond-move
17224 @opindex mno-cond-move
17226 Disable the use of conditional-move instructions.
17228 This switch is mainly for debugging the compiler and will likely be removed
17229 in a future version.
17234 Enable the use of conditional set instructions (default).
17236 This switch is mainly for debugging the compiler and will likely be removed
17237 in a future version.
17242 Disable the use of conditional set instructions.
17244 This switch is mainly for debugging the compiler and will likely be removed
17245 in a future version.
17248 @opindex mcond-exec
17250 Enable the use of conditional execution (default).
17252 This switch is mainly for debugging the compiler and will likely be removed
17253 in a future version.
17255 @item -mno-cond-exec
17256 @opindex mno-cond-exec
17258 Disable the use of conditional execution.
17260 This switch is mainly for debugging the compiler and will likely be removed
17261 in a future version.
17263 @item -mvliw-branch
17264 @opindex mvliw-branch
17266 Run a pass to pack branches into VLIW instructions (default).
17268 This switch is mainly for debugging the compiler and will likely be removed
17269 in a future version.
17271 @item -mno-vliw-branch
17272 @opindex mno-vliw-branch
17274 Do not run a pass to pack branches into VLIW instructions.
17276 This switch is mainly for debugging the compiler and will likely be removed
17277 in a future version.
17279 @item -mmulti-cond-exec
17280 @opindex mmulti-cond-exec
17282 Enable optimization of @code{&&} and @code{||} in conditional execution
17285 This switch is mainly for debugging the compiler and will likely be removed
17286 in a future version.
17288 @item -mno-multi-cond-exec
17289 @opindex mno-multi-cond-exec
17291 Disable optimization of @code{&&} and @code{||} in conditional execution.
17293 This switch is mainly for debugging the compiler and will likely be removed
17294 in a future version.
17296 @item -mnested-cond-exec
17297 @opindex mnested-cond-exec
17299 Enable nested conditional execution optimizations (default).
17301 This switch is mainly for debugging the compiler and will likely be removed
17302 in a future version.
17304 @item -mno-nested-cond-exec
17305 @opindex mno-nested-cond-exec
17307 Disable nested conditional execution optimizations.
17309 This switch is mainly for debugging the compiler and will likely be removed
17310 in a future version.
17312 @item -moptimize-membar
17313 @opindex moptimize-membar
17315 This switch removes redundant @code{membar} instructions from the
17316 compiler-generated code. It is enabled by default.
17318 @item -mno-optimize-membar
17319 @opindex mno-optimize-membar
17321 This switch disables the automatic removal of redundant @code{membar}
17322 instructions from the generated code.
17324 @item -mtomcat-stats
17325 @opindex mtomcat-stats
17327 Cause gas to print out tomcat statistics.
17329 @item -mcpu=@var{cpu}
17332 Select the processor type for which to generate code. Possible values are
17333 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
17334 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
17338 @node GNU/Linux Options
17339 @subsection GNU/Linux Options
17341 These @samp{-m} options are defined for GNU/Linux targets:
17346 Use the GNU C library. This is the default except
17347 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
17348 @samp{*-*-linux-*android*} targets.
17352 Use uClibc C library. This is the default on
17353 @samp{*-*-linux-*uclibc*} targets.
17357 Use the musl C library. This is the default on
17358 @samp{*-*-linux-*musl*} targets.
17362 Use Bionic C library. This is the default on
17363 @samp{*-*-linux-*android*} targets.
17367 Compile code compatible with Android platform. This is the default on
17368 @samp{*-*-linux-*android*} targets.
17370 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
17371 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
17372 this option makes the GCC driver pass Android-specific options to the linker.
17373 Finally, this option causes the preprocessor macro @code{__ANDROID__}
17376 @item -tno-android-cc
17377 @opindex tno-android-cc
17378 Disable compilation effects of @option{-mandroid}, i.e., do not enable
17379 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
17380 @option{-fno-rtti} by default.
17382 @item -tno-android-ld
17383 @opindex tno-android-ld
17384 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
17385 linking options to the linker.
17389 @node H8/300 Options
17390 @subsection H8/300 Options
17392 These @samp{-m} options are defined for the H8/300 implementations:
17397 Shorten some address references at link time, when possible; uses the
17398 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
17399 ld, Using ld}, for a fuller description.
17403 Generate code for the H8/300H@.
17407 Generate code for the H8S@.
17411 Generate code for the H8S and H8/300H in the normal mode. This switch
17412 must be used either with @option{-mh} or @option{-ms}.
17416 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
17420 Extended registers are stored on stack before execution of function
17421 with monitor attribute. Default option is @option{-mexr}.
17422 This option is valid only for H8S targets.
17426 Extended registers are not stored on stack before execution of function
17427 with monitor attribute. Default option is @option{-mno-exr}.
17428 This option is valid only for H8S targets.
17432 Make @code{int} data 32 bits by default.
17435 @opindex malign-300
17436 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
17437 The default for the H8/300H and H8S is to align longs and floats on
17439 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
17440 This option has no effect on the H8/300.
17444 @subsection HPPA Options
17445 @cindex HPPA Options
17447 These @samp{-m} options are defined for the HPPA family of computers:
17450 @item -march=@var{architecture-type}
17452 Generate code for the specified architecture. The choices for
17453 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
17454 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
17455 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
17456 architecture option for your machine. Code compiled for lower numbered
17457 architectures runs on higher numbered architectures, but not the
17460 @item -mpa-risc-1-0
17461 @itemx -mpa-risc-1-1
17462 @itemx -mpa-risc-2-0
17463 @opindex mpa-risc-1-0
17464 @opindex mpa-risc-1-1
17465 @opindex mpa-risc-2-0
17466 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
17468 @item -mcaller-copies
17469 @opindex mcaller-copies
17470 The caller copies function arguments passed by hidden reference. This
17471 option should be used with care as it is not compatible with the default
17472 32-bit runtime. However, only aggregates larger than eight bytes are
17473 passed by hidden reference and the option provides better compatibility
17476 @item -mjump-in-delay
17477 @opindex mjump-in-delay
17478 This option is ignored and provided for compatibility purposes only.
17480 @item -mdisable-fpregs
17481 @opindex mdisable-fpregs
17482 Prevent floating-point registers from being used in any manner. This is
17483 necessary for compiling kernels that perform lazy context switching of
17484 floating-point registers. If you use this option and attempt to perform
17485 floating-point operations, the compiler aborts.
17487 @item -mdisable-indexing
17488 @opindex mdisable-indexing
17489 Prevent the compiler from using indexing address modes. This avoids some
17490 rather obscure problems when compiling MIG generated code under MACH@.
17492 @item -mno-space-regs
17493 @opindex mno-space-regs
17494 Generate code that assumes the target has no space registers. This allows
17495 GCC to generate faster indirect calls and use unscaled index address modes.
17497 Such code is suitable for level 0 PA systems and kernels.
17499 @item -mfast-indirect-calls
17500 @opindex mfast-indirect-calls
17501 Generate code that assumes calls never cross space boundaries. This
17502 allows GCC to emit code that performs faster indirect calls.
17504 This option does not work in the presence of shared libraries or nested
17507 @item -mfixed-range=@var{register-range}
17508 @opindex mfixed-range
17509 Generate code treating the given register range as fixed registers.
17510 A fixed register is one that the register allocator cannot use. This is
17511 useful when compiling kernel code. A register range is specified as
17512 two registers separated by a dash. Multiple register ranges can be
17513 specified separated by a comma.
17515 @item -mlong-load-store
17516 @opindex mlong-load-store
17517 Generate 3-instruction load and store sequences as sometimes required by
17518 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
17521 @item -mportable-runtime
17522 @opindex mportable-runtime
17523 Use the portable calling conventions proposed by HP for ELF systems.
17527 Enable the use of assembler directives only GAS understands.
17529 @item -mschedule=@var{cpu-type}
17531 Schedule code according to the constraints for the machine type
17532 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
17533 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
17534 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
17535 proper scheduling option for your machine. The default scheduling is
17539 @opindex mlinker-opt
17540 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
17541 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
17542 linkers in which they give bogus error messages when linking some programs.
17545 @opindex msoft-float
17546 Generate output containing library calls for floating point.
17547 @strong{Warning:} the requisite libraries are not available for all HPPA
17548 targets. Normally the facilities of the machine's usual C compiler are
17549 used, but this cannot be done directly in cross-compilation. You must make
17550 your own arrangements to provide suitable library functions for
17553 @option{-msoft-float} changes the calling convention in the output file;
17554 therefore, it is only useful if you compile @emph{all} of a program with
17555 this option. In particular, you need to compile @file{libgcc.a}, the
17556 library that comes with GCC, with @option{-msoft-float} in order for
17561 Generate the predefine, @code{_SIO}, for server IO@. The default is
17562 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
17563 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
17564 options are available under HP-UX and HI-UX@.
17568 Use options specific to GNU @command{ld}.
17569 This passes @option{-shared} to @command{ld} when
17570 building a shared library. It is the default when GCC is configured,
17571 explicitly or implicitly, with the GNU linker. This option does not
17572 affect which @command{ld} is called; it only changes what parameters
17573 are passed to that @command{ld}.
17574 The @command{ld} that is called is determined by the
17575 @option{--with-ld} configure option, GCC's program search path, and
17576 finally by the user's @env{PATH}. The linker used by GCC can be printed
17577 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
17578 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
17582 Use options specific to HP @command{ld}.
17583 This passes @option{-b} to @command{ld} when building
17584 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
17585 links. It is the default when GCC is configured, explicitly or
17586 implicitly, with the HP linker. This option does not affect
17587 which @command{ld} is called; it only changes what parameters are passed to that
17589 The @command{ld} that is called is determined by the @option{--with-ld}
17590 configure option, GCC's program search path, and finally by the user's
17591 @env{PATH}. The linker used by GCC can be printed using @samp{which
17592 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
17593 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
17596 @opindex mno-long-calls
17597 Generate code that uses long call sequences. This ensures that a call
17598 is always able to reach linker generated stubs. The default is to generate
17599 long calls only when the distance from the call site to the beginning
17600 of the function or translation unit, as the case may be, exceeds a
17601 predefined limit set by the branch type being used. The limits for
17602 normal calls are 7,600,000 and 240,000 bytes, respectively for the
17603 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
17606 Distances are measured from the beginning of functions when using the
17607 @option{-ffunction-sections} option, or when using the @option{-mgas}
17608 and @option{-mno-portable-runtime} options together under HP-UX with
17611 It is normally not desirable to use this option as it degrades
17612 performance. However, it may be useful in large applications,
17613 particularly when partial linking is used to build the application.
17615 The types of long calls used depends on the capabilities of the
17616 assembler and linker, and the type of code being generated. The
17617 impact on systems that support long absolute calls, and long pic
17618 symbol-difference or pc-relative calls should be relatively small.
17619 However, an indirect call is used on 32-bit ELF systems in pic code
17620 and it is quite long.
17622 @item -munix=@var{unix-std}
17624 Generate compiler predefines and select a startfile for the specified
17625 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
17626 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
17627 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
17628 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
17629 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
17632 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
17633 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
17634 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
17635 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
17636 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
17637 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
17639 It is @emph{important} to note that this option changes the interfaces
17640 for various library routines. It also affects the operational behavior
17641 of the C library. Thus, @emph{extreme} care is needed in using this
17644 Library code that is intended to operate with more than one UNIX
17645 standard must test, set and restore the variable @code{__xpg4_extended_mask}
17646 as appropriate. Most GNU software doesn't provide this capability.
17650 Suppress the generation of link options to search libdld.sl when the
17651 @option{-static} option is specified on HP-UX 10 and later.
17655 The HP-UX implementation of setlocale in libc has a dependency on
17656 libdld.sl. There isn't an archive version of libdld.sl. Thus,
17657 when the @option{-static} option is specified, special link options
17658 are needed to resolve this dependency.
17660 On HP-UX 10 and later, the GCC driver adds the necessary options to
17661 link with libdld.sl when the @option{-static} option is specified.
17662 This causes the resulting binary to be dynamic. On the 64-bit port,
17663 the linkers generate dynamic binaries by default in any case. The
17664 @option{-nolibdld} option can be used to prevent the GCC driver from
17665 adding these link options.
17669 Add support for multithreading with the @dfn{dce thread} library
17670 under HP-UX@. This option sets flags for both the preprocessor and
17674 @node IA-64 Options
17675 @subsection IA-64 Options
17676 @cindex IA-64 Options
17678 These are the @samp{-m} options defined for the Intel IA-64 architecture.
17682 @opindex mbig-endian
17683 Generate code for a big-endian target. This is the default for HP-UX@.
17685 @item -mlittle-endian
17686 @opindex mlittle-endian
17687 Generate code for a little-endian target. This is the default for AIX5
17693 @opindex mno-gnu-as
17694 Generate (or don't) code for the GNU assembler. This is the default.
17695 @c Also, this is the default if the configure option @option{--with-gnu-as}
17701 @opindex mno-gnu-ld
17702 Generate (or don't) code for the GNU linker. This is the default.
17703 @c Also, this is the default if the configure option @option{--with-gnu-ld}
17708 Generate code that does not use a global pointer register. The result
17709 is not position independent code, and violates the IA-64 ABI@.
17711 @item -mvolatile-asm-stop
17712 @itemx -mno-volatile-asm-stop
17713 @opindex mvolatile-asm-stop
17714 @opindex mno-volatile-asm-stop
17715 Generate (or don't) a stop bit immediately before and after volatile asm
17718 @item -mregister-names
17719 @itemx -mno-register-names
17720 @opindex mregister-names
17721 @opindex mno-register-names
17722 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
17723 the stacked registers. This may make assembler output more readable.
17729 Disable (or enable) optimizations that use the small data section. This may
17730 be useful for working around optimizer bugs.
17732 @item -mconstant-gp
17733 @opindex mconstant-gp
17734 Generate code that uses a single constant global pointer value. This is
17735 useful when compiling kernel code.
17739 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
17740 This is useful when compiling firmware code.
17742 @item -minline-float-divide-min-latency
17743 @opindex minline-float-divide-min-latency
17744 Generate code for inline divides of floating-point values
17745 using the minimum latency algorithm.
17747 @item -minline-float-divide-max-throughput
17748 @opindex minline-float-divide-max-throughput
17749 Generate code for inline divides of floating-point values
17750 using the maximum throughput algorithm.
17752 @item -mno-inline-float-divide
17753 @opindex mno-inline-float-divide
17754 Do not generate inline code for divides of floating-point values.
17756 @item -minline-int-divide-min-latency
17757 @opindex minline-int-divide-min-latency
17758 Generate code for inline divides of integer values
17759 using the minimum latency algorithm.
17761 @item -minline-int-divide-max-throughput
17762 @opindex minline-int-divide-max-throughput
17763 Generate code for inline divides of integer values
17764 using the maximum throughput algorithm.
17766 @item -mno-inline-int-divide
17767 @opindex mno-inline-int-divide
17768 Do not generate inline code for divides of integer values.
17770 @item -minline-sqrt-min-latency
17771 @opindex minline-sqrt-min-latency
17772 Generate code for inline square roots
17773 using the minimum latency algorithm.
17775 @item -minline-sqrt-max-throughput
17776 @opindex minline-sqrt-max-throughput
17777 Generate code for inline square roots
17778 using the maximum throughput algorithm.
17780 @item -mno-inline-sqrt
17781 @opindex mno-inline-sqrt
17782 Do not generate inline code for @code{sqrt}.
17785 @itemx -mno-fused-madd
17786 @opindex mfused-madd
17787 @opindex mno-fused-madd
17788 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
17789 instructions. The default is to use these instructions.
17791 @item -mno-dwarf2-asm
17792 @itemx -mdwarf2-asm
17793 @opindex mno-dwarf2-asm
17794 @opindex mdwarf2-asm
17795 Don't (or do) generate assembler code for the DWARF line number debugging
17796 info. This may be useful when not using the GNU assembler.
17798 @item -mearly-stop-bits
17799 @itemx -mno-early-stop-bits
17800 @opindex mearly-stop-bits
17801 @opindex mno-early-stop-bits
17802 Allow stop bits to be placed earlier than immediately preceding the
17803 instruction that triggered the stop bit. This can improve instruction
17804 scheduling, but does not always do so.
17806 @item -mfixed-range=@var{register-range}
17807 @opindex mfixed-range
17808 Generate code treating the given register range as fixed registers.
17809 A fixed register is one that the register allocator cannot use. This is
17810 useful when compiling kernel code. A register range is specified as
17811 two registers separated by a dash. Multiple register ranges can be
17812 specified separated by a comma.
17814 @item -mtls-size=@var{tls-size}
17816 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
17819 @item -mtune=@var{cpu-type}
17821 Tune the instruction scheduling for a particular CPU, Valid values are
17822 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
17823 and @samp{mckinley}.
17829 Generate code for a 32-bit or 64-bit environment.
17830 The 32-bit environment sets int, long and pointer to 32 bits.
17831 The 64-bit environment sets int to 32 bits and long and pointer
17832 to 64 bits. These are HP-UX specific flags.
17834 @item -mno-sched-br-data-spec
17835 @itemx -msched-br-data-spec
17836 @opindex mno-sched-br-data-spec
17837 @opindex msched-br-data-spec
17838 (Dis/En)able data speculative scheduling before reload.
17839 This results in generation of @code{ld.a} instructions and
17840 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
17841 The default setting is disabled.
17843 @item -msched-ar-data-spec
17844 @itemx -mno-sched-ar-data-spec
17845 @opindex msched-ar-data-spec
17846 @opindex mno-sched-ar-data-spec
17847 (En/Dis)able data speculative scheduling after reload.
17848 This results in generation of @code{ld.a} instructions and
17849 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
17850 The default setting is enabled.
17852 @item -mno-sched-control-spec
17853 @itemx -msched-control-spec
17854 @opindex mno-sched-control-spec
17855 @opindex msched-control-spec
17856 (Dis/En)able control speculative scheduling. This feature is
17857 available only during region scheduling (i.e.@: before reload).
17858 This results in generation of the @code{ld.s} instructions and
17859 the corresponding check instructions @code{chk.s}.
17860 The default setting is disabled.
17862 @item -msched-br-in-data-spec
17863 @itemx -mno-sched-br-in-data-spec
17864 @opindex msched-br-in-data-spec
17865 @opindex mno-sched-br-in-data-spec
17866 (En/Dis)able speculative scheduling of the instructions that
17867 are dependent on the data speculative loads before reload.
17868 This is effective only with @option{-msched-br-data-spec} enabled.
17869 The default setting is enabled.
17871 @item -msched-ar-in-data-spec
17872 @itemx -mno-sched-ar-in-data-spec
17873 @opindex msched-ar-in-data-spec
17874 @opindex mno-sched-ar-in-data-spec
17875 (En/Dis)able speculative scheduling of the instructions that
17876 are dependent on the data speculative loads after reload.
17877 This is effective only with @option{-msched-ar-data-spec} enabled.
17878 The default setting is enabled.
17880 @item -msched-in-control-spec
17881 @itemx -mno-sched-in-control-spec
17882 @opindex msched-in-control-spec
17883 @opindex mno-sched-in-control-spec
17884 (En/Dis)able speculative scheduling of the instructions that
17885 are dependent on the control speculative loads.
17886 This is effective only with @option{-msched-control-spec} enabled.
17887 The default setting is enabled.
17889 @item -mno-sched-prefer-non-data-spec-insns
17890 @itemx -msched-prefer-non-data-spec-insns
17891 @opindex mno-sched-prefer-non-data-spec-insns
17892 @opindex msched-prefer-non-data-spec-insns
17893 If enabled, data-speculative instructions are chosen for schedule
17894 only if there are no other choices at the moment. This makes
17895 the use of the data speculation much more conservative.
17896 The default setting is disabled.
17898 @item -mno-sched-prefer-non-control-spec-insns
17899 @itemx -msched-prefer-non-control-spec-insns
17900 @opindex mno-sched-prefer-non-control-spec-insns
17901 @opindex msched-prefer-non-control-spec-insns
17902 If enabled, control-speculative instructions are chosen for schedule
17903 only if there are no other choices at the moment. This makes
17904 the use of the control speculation much more conservative.
17905 The default setting is disabled.
17907 @item -mno-sched-count-spec-in-critical-path
17908 @itemx -msched-count-spec-in-critical-path
17909 @opindex mno-sched-count-spec-in-critical-path
17910 @opindex msched-count-spec-in-critical-path
17911 If enabled, speculative dependencies are considered during
17912 computation of the instructions priorities. This makes the use of the
17913 speculation a bit more conservative.
17914 The default setting is disabled.
17916 @item -msched-spec-ldc
17917 @opindex msched-spec-ldc
17918 Use a simple data speculation check. This option is on by default.
17920 @item -msched-control-spec-ldc
17921 @opindex msched-spec-ldc
17922 Use a simple check for control speculation. This option is on by default.
17924 @item -msched-stop-bits-after-every-cycle
17925 @opindex msched-stop-bits-after-every-cycle
17926 Place a stop bit after every cycle when scheduling. This option is on
17929 @item -msched-fp-mem-deps-zero-cost
17930 @opindex msched-fp-mem-deps-zero-cost
17931 Assume that floating-point stores and loads are not likely to cause a conflict
17932 when placed into the same instruction group. This option is disabled by
17935 @item -msel-sched-dont-check-control-spec
17936 @opindex msel-sched-dont-check-control-spec
17937 Generate checks for control speculation in selective scheduling.
17938 This flag is disabled by default.
17940 @item -msched-max-memory-insns=@var{max-insns}
17941 @opindex msched-max-memory-insns
17942 Limit on the number of memory insns per instruction group, giving lower
17943 priority to subsequent memory insns attempting to schedule in the same
17944 instruction group. Frequently useful to prevent cache bank conflicts.
17945 The default value is 1.
17947 @item -msched-max-memory-insns-hard-limit
17948 @opindex msched-max-memory-insns-hard-limit
17949 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
17950 disallowing more than that number in an instruction group.
17951 Otherwise, the limit is ``soft'', meaning that non-memory operations
17952 are preferred when the limit is reached, but memory operations may still
17958 @subsection LM32 Options
17959 @cindex LM32 options
17961 These @option{-m} options are defined for the LatticeMico32 architecture:
17964 @item -mbarrel-shift-enabled
17965 @opindex mbarrel-shift-enabled
17966 Enable barrel-shift instructions.
17968 @item -mdivide-enabled
17969 @opindex mdivide-enabled
17970 Enable divide and modulus instructions.
17972 @item -mmultiply-enabled
17973 @opindex multiply-enabled
17974 Enable multiply instructions.
17976 @item -msign-extend-enabled
17977 @opindex msign-extend-enabled
17978 Enable sign extend instructions.
17980 @item -muser-enabled
17981 @opindex muser-enabled
17982 Enable user-defined instructions.
17987 @subsection M32C Options
17988 @cindex M32C options
17991 @item -mcpu=@var{name}
17993 Select the CPU for which code is generated. @var{name} may be one of
17994 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
17995 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
17996 the M32C/80 series.
18000 Specifies that the program will be run on the simulator. This causes
18001 an alternate runtime library to be linked in which supports, for
18002 example, file I/O@. You must not use this option when generating
18003 programs that will run on real hardware; you must provide your own
18004 runtime library for whatever I/O functions are needed.
18006 @item -memregs=@var{number}
18008 Specifies the number of memory-based pseudo-registers GCC uses
18009 during code generation. These pseudo-registers are used like real
18010 registers, so there is a tradeoff between GCC's ability to fit the
18011 code into available registers, and the performance penalty of using
18012 memory instead of registers. Note that all modules in a program must
18013 be compiled with the same value for this option. Because of that, you
18014 must not use this option with GCC's default runtime libraries.
18018 @node M32R/D Options
18019 @subsection M32R/D Options
18020 @cindex M32R/D options
18022 These @option{-m} options are defined for Renesas M32R/D architectures:
18027 Generate code for the M32R/2@.
18031 Generate code for the M32R/X@.
18035 Generate code for the M32R@. This is the default.
18037 @item -mmodel=small
18038 @opindex mmodel=small
18039 Assume all objects live in the lower 16MB of memory (so that their addresses
18040 can be loaded with the @code{ld24} instruction), and assume all subroutines
18041 are reachable with the @code{bl} instruction.
18042 This is the default.
18044 The addressability of a particular object can be set with the
18045 @code{model} attribute.
18047 @item -mmodel=medium
18048 @opindex mmodel=medium
18049 Assume objects may be anywhere in the 32-bit address space (the compiler
18050 generates @code{seth/add3} instructions to load their addresses), and
18051 assume all subroutines are reachable with the @code{bl} instruction.
18053 @item -mmodel=large
18054 @opindex mmodel=large
18055 Assume objects may be anywhere in the 32-bit address space (the compiler
18056 generates @code{seth/add3} instructions to load their addresses), and
18057 assume subroutines may not be reachable with the @code{bl} instruction
18058 (the compiler generates the much slower @code{seth/add3/jl}
18059 instruction sequence).
18062 @opindex msdata=none
18063 Disable use of the small data area. Variables are put into
18064 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
18065 @code{section} attribute has been specified).
18066 This is the default.
18068 The small data area consists of sections @code{.sdata} and @code{.sbss}.
18069 Objects may be explicitly put in the small data area with the
18070 @code{section} attribute using one of these sections.
18072 @item -msdata=sdata
18073 @opindex msdata=sdata
18074 Put small global and static data in the small data area, but do not
18075 generate special code to reference them.
18078 @opindex msdata=use
18079 Put small global and static data in the small data area, and generate
18080 special instructions to reference them.
18084 @cindex smaller data references
18085 Put global and static objects less than or equal to @var{num} bytes
18086 into the small data or BSS sections instead of the normal data or BSS
18087 sections. The default value of @var{num} is 8.
18088 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
18089 for this option to have any effect.
18091 All modules should be compiled with the same @option{-G @var{num}} value.
18092 Compiling with different values of @var{num} may or may not work; if it
18093 doesn't the linker gives an error message---incorrect code is not
18098 Makes the M32R-specific code in the compiler display some statistics
18099 that might help in debugging programs.
18101 @item -malign-loops
18102 @opindex malign-loops
18103 Align all loops to a 32-byte boundary.
18105 @item -mno-align-loops
18106 @opindex mno-align-loops
18107 Do not enforce a 32-byte alignment for loops. This is the default.
18109 @item -missue-rate=@var{number}
18110 @opindex missue-rate=@var{number}
18111 Issue @var{number} instructions per cycle. @var{number} can only be 1
18114 @item -mbranch-cost=@var{number}
18115 @opindex mbranch-cost=@var{number}
18116 @var{number} can only be 1 or 2. If it is 1 then branches are
18117 preferred over conditional code, if it is 2, then the opposite applies.
18119 @item -mflush-trap=@var{number}
18120 @opindex mflush-trap=@var{number}
18121 Specifies the trap number to use to flush the cache. The default is
18122 12. Valid numbers are between 0 and 15 inclusive.
18124 @item -mno-flush-trap
18125 @opindex mno-flush-trap
18126 Specifies that the cache cannot be flushed by using a trap.
18128 @item -mflush-func=@var{name}
18129 @opindex mflush-func=@var{name}
18130 Specifies the name of the operating system function to call to flush
18131 the cache. The default is @samp{_flush_cache}, but a function call
18132 is only used if a trap is not available.
18134 @item -mno-flush-func
18135 @opindex mno-flush-func
18136 Indicates that there is no OS function for flushing the cache.
18140 @node M680x0 Options
18141 @subsection M680x0 Options
18142 @cindex M680x0 options
18144 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
18145 The default settings depend on which architecture was selected when
18146 the compiler was configured; the defaults for the most common choices
18150 @item -march=@var{arch}
18152 Generate code for a specific M680x0 or ColdFire instruction set
18153 architecture. Permissible values of @var{arch} for M680x0
18154 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
18155 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
18156 architectures are selected according to Freescale's ISA classification
18157 and the permissible values are: @samp{isaa}, @samp{isaaplus},
18158 @samp{isab} and @samp{isac}.
18160 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
18161 code for a ColdFire target. The @var{arch} in this macro is one of the
18162 @option{-march} arguments given above.
18164 When used together, @option{-march} and @option{-mtune} select code
18165 that runs on a family of similar processors but that is optimized
18166 for a particular microarchitecture.
18168 @item -mcpu=@var{cpu}
18170 Generate code for a specific M680x0 or ColdFire processor.
18171 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
18172 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
18173 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
18174 below, which also classifies the CPUs into families:
18176 @multitable @columnfractions 0.20 0.80
18177 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
18178 @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}
18179 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
18180 @item @samp{5206e} @tab @samp{5206e}
18181 @item @samp{5208} @tab @samp{5207} @samp{5208}
18182 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
18183 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
18184 @item @samp{5216} @tab @samp{5214} @samp{5216}
18185 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
18186 @item @samp{5225} @tab @samp{5224} @samp{5225}
18187 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
18188 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
18189 @item @samp{5249} @tab @samp{5249}
18190 @item @samp{5250} @tab @samp{5250}
18191 @item @samp{5271} @tab @samp{5270} @samp{5271}
18192 @item @samp{5272} @tab @samp{5272}
18193 @item @samp{5275} @tab @samp{5274} @samp{5275}
18194 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
18195 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
18196 @item @samp{5307} @tab @samp{5307}
18197 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
18198 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
18199 @item @samp{5407} @tab @samp{5407}
18200 @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}
18203 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
18204 @var{arch} is compatible with @var{cpu}. Other combinations of
18205 @option{-mcpu} and @option{-march} are rejected.
18207 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
18208 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
18209 where the value of @var{family} is given by the table above.
18211 @item -mtune=@var{tune}
18213 Tune the code for a particular microarchitecture within the
18214 constraints set by @option{-march} and @option{-mcpu}.
18215 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
18216 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
18217 and @samp{cpu32}. The ColdFire microarchitectures
18218 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
18220 You can also use @option{-mtune=68020-40} for code that needs
18221 to run relatively well on 68020, 68030 and 68040 targets.
18222 @option{-mtune=68020-60} is similar but includes 68060 targets
18223 as well. These two options select the same tuning decisions as
18224 @option{-m68020-40} and @option{-m68020-60} respectively.
18226 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
18227 when tuning for 680x0 architecture @var{arch}. It also defines
18228 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
18229 option is used. If GCC is tuning for a range of architectures,
18230 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
18231 it defines the macros for every architecture in the range.
18233 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
18234 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
18235 of the arguments given above.
18241 Generate output for a 68000. This is the default
18242 when the compiler is configured for 68000-based systems.
18243 It is equivalent to @option{-march=68000}.
18245 Use this option for microcontrollers with a 68000 or EC000 core,
18246 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
18250 Generate output for a 68010. This is the default
18251 when the compiler is configured for 68010-based systems.
18252 It is equivalent to @option{-march=68010}.
18258 Generate output for a 68020. This is the default
18259 when the compiler is configured for 68020-based systems.
18260 It is equivalent to @option{-march=68020}.
18264 Generate output for a 68030. This is the default when the compiler is
18265 configured for 68030-based systems. It is equivalent to
18266 @option{-march=68030}.
18270 Generate output for a 68040. This is the default when the compiler is
18271 configured for 68040-based systems. It is equivalent to
18272 @option{-march=68040}.
18274 This option inhibits the use of 68881/68882 instructions that have to be
18275 emulated by software on the 68040. Use this option if your 68040 does not
18276 have code to emulate those instructions.
18280 Generate output for a 68060. This is the default when the compiler is
18281 configured for 68060-based systems. It is equivalent to
18282 @option{-march=68060}.
18284 This option inhibits the use of 68020 and 68881/68882 instructions that
18285 have to be emulated by software on the 68060. Use this option if your 68060
18286 does not have code to emulate those instructions.
18290 Generate output for a CPU32. This is the default
18291 when the compiler is configured for CPU32-based systems.
18292 It is equivalent to @option{-march=cpu32}.
18294 Use this option for microcontrollers with a
18295 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
18296 68336, 68340, 68341, 68349 and 68360.
18300 Generate output for a 520X ColdFire CPU@. This is the default
18301 when the compiler is configured for 520X-based systems.
18302 It is equivalent to @option{-mcpu=5206}, and is now deprecated
18303 in favor of that option.
18305 Use this option for microcontroller with a 5200 core, including
18306 the MCF5202, MCF5203, MCF5204 and MCF5206.
18310 Generate output for a 5206e ColdFire CPU@. The option is now
18311 deprecated in favor of the equivalent @option{-mcpu=5206e}.
18315 Generate output for a member of the ColdFire 528X family.
18316 The option is now deprecated in favor of the equivalent
18317 @option{-mcpu=528x}.
18321 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
18322 in favor of the equivalent @option{-mcpu=5307}.
18326 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
18327 in favor of the equivalent @option{-mcpu=5407}.
18331 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
18332 This includes use of hardware floating-point instructions.
18333 The option is equivalent to @option{-mcpu=547x}, and is now
18334 deprecated in favor of that option.
18338 Generate output for a 68040, without using any of the new instructions.
18339 This results in code that can run relatively efficiently on either a
18340 68020/68881 or a 68030 or a 68040. The generated code does use the
18341 68881 instructions that are emulated on the 68040.
18343 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
18347 Generate output for a 68060, without using any of the new instructions.
18348 This results in code that can run relatively efficiently on either a
18349 68020/68881 or a 68030 or a 68040. The generated code does use the
18350 68881 instructions that are emulated on the 68060.
18352 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
18356 @opindex mhard-float
18358 Generate floating-point instructions. This is the default for 68020
18359 and above, and for ColdFire devices that have an FPU@. It defines the
18360 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
18361 on ColdFire targets.
18364 @opindex msoft-float
18365 Do not generate floating-point instructions; use library calls instead.
18366 This is the default for 68000, 68010, and 68832 targets. It is also
18367 the default for ColdFire devices that have no FPU.
18373 Generate (do not generate) ColdFire hardware divide and remainder
18374 instructions. If @option{-march} is used without @option{-mcpu},
18375 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
18376 architectures. Otherwise, the default is taken from the target CPU
18377 (either the default CPU, or the one specified by @option{-mcpu}). For
18378 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
18379 @option{-mcpu=5206e}.
18381 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
18385 Consider type @code{int} to be 16 bits wide, like @code{short int}.
18386 Additionally, parameters passed on the stack are also aligned to a
18387 16-bit boundary even on targets whose API mandates promotion to 32-bit.
18391 Do not consider type @code{int} to be 16 bits wide. This is the default.
18394 @itemx -mno-bitfield
18395 @opindex mnobitfield
18396 @opindex mno-bitfield
18397 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
18398 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
18402 Do use the bit-field instructions. The @option{-m68020} option implies
18403 @option{-mbitfield}. This is the default if you use a configuration
18404 designed for a 68020.
18408 Use a different function-calling convention, in which functions
18409 that take a fixed number of arguments return with the @code{rtd}
18410 instruction, which pops their arguments while returning. This
18411 saves one instruction in the caller since there is no need to pop
18412 the arguments there.
18414 This calling convention is incompatible with the one normally
18415 used on Unix, so you cannot use it if you need to call libraries
18416 compiled with the Unix compiler.
18418 Also, you must provide function prototypes for all functions that
18419 take variable numbers of arguments (including @code{printf});
18420 otherwise incorrect code is generated for calls to those
18423 In addition, seriously incorrect code results if you call a
18424 function with too many arguments. (Normally, extra arguments are
18425 harmlessly ignored.)
18427 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
18428 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
18432 Do not use the calling conventions selected by @option{-mrtd}.
18433 This is the default.
18436 @itemx -mno-align-int
18437 @opindex malign-int
18438 @opindex mno-align-int
18439 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
18440 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
18441 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
18442 Aligning variables on 32-bit boundaries produces code that runs somewhat
18443 faster on processors with 32-bit busses at the expense of more memory.
18445 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
18446 aligns structures containing the above types differently than
18447 most published application binary interface specifications for the m68k.
18451 Use the pc-relative addressing mode of the 68000 directly, instead of
18452 using a global offset table. At present, this option implies @option{-fpic},
18453 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
18454 not presently supported with @option{-mpcrel}, though this could be supported for
18455 68020 and higher processors.
18457 @item -mno-strict-align
18458 @itemx -mstrict-align
18459 @opindex mno-strict-align
18460 @opindex mstrict-align
18461 Do not (do) assume that unaligned memory references are handled by
18465 Generate code that allows the data segment to be located in a different
18466 area of memory from the text segment. This allows for execute-in-place in
18467 an environment without virtual memory management. This option implies
18470 @item -mno-sep-data
18471 Generate code that assumes that the data segment follows the text segment.
18472 This is the default.
18474 @item -mid-shared-library
18475 Generate code that supports shared libraries via the library ID method.
18476 This allows for execute-in-place and shared libraries in an environment
18477 without virtual memory management. This option implies @option{-fPIC}.
18479 @item -mno-id-shared-library
18480 Generate code that doesn't assume ID-based shared libraries are being used.
18481 This is the default.
18483 @item -mshared-library-id=n
18484 Specifies the identification number of the ID-based shared library being
18485 compiled. Specifying a value of 0 generates more compact code; specifying
18486 other values forces the allocation of that number to the current
18487 library, but is no more space- or time-efficient than omitting this option.
18493 When generating position-independent code for ColdFire, generate code
18494 that works if the GOT has more than 8192 entries. This code is
18495 larger and slower than code generated without this option. On M680x0
18496 processors, this option is not needed; @option{-fPIC} suffices.
18498 GCC normally uses a single instruction to load values from the GOT@.
18499 While this is relatively efficient, it only works if the GOT
18500 is smaller than about 64k. Anything larger causes the linker
18501 to report an error such as:
18503 @cindex relocation truncated to fit (ColdFire)
18505 relocation truncated to fit: R_68K_GOT16O foobar
18508 If this happens, you should recompile your code with @option{-mxgot}.
18509 It should then work with very large GOTs. However, code generated with
18510 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
18511 the value of a global symbol.
18513 Note that some linkers, including newer versions of the GNU linker,
18514 can create multiple GOTs and sort GOT entries. If you have such a linker,
18515 you should only need to use @option{-mxgot} when compiling a single
18516 object file that accesses more than 8192 GOT entries. Very few do.
18518 These options have no effect unless GCC is generating
18519 position-independent code.
18523 @node MCore Options
18524 @subsection MCore Options
18525 @cindex MCore options
18527 These are the @samp{-m} options defined for the Motorola M*Core
18533 @itemx -mno-hardlit
18535 @opindex mno-hardlit
18536 Inline constants into the code stream if it can be done in two
18537 instructions or less.
18543 Use the divide instruction. (Enabled by default).
18545 @item -mrelax-immediate
18546 @itemx -mno-relax-immediate
18547 @opindex mrelax-immediate
18548 @opindex mno-relax-immediate
18549 Allow arbitrary-sized immediates in bit operations.
18551 @item -mwide-bitfields
18552 @itemx -mno-wide-bitfields
18553 @opindex mwide-bitfields
18554 @opindex mno-wide-bitfields
18555 Always treat bit-fields as @code{int}-sized.
18557 @item -m4byte-functions
18558 @itemx -mno-4byte-functions
18559 @opindex m4byte-functions
18560 @opindex mno-4byte-functions
18561 Force all functions to be aligned to a 4-byte boundary.
18563 @item -mcallgraph-data
18564 @itemx -mno-callgraph-data
18565 @opindex mcallgraph-data
18566 @opindex mno-callgraph-data
18567 Emit callgraph information.
18570 @itemx -mno-slow-bytes
18571 @opindex mslow-bytes
18572 @opindex mno-slow-bytes
18573 Prefer word access when reading byte quantities.
18575 @item -mlittle-endian
18576 @itemx -mbig-endian
18577 @opindex mlittle-endian
18578 @opindex mbig-endian
18579 Generate code for a little-endian target.
18585 Generate code for the 210 processor.
18589 Assume that runtime support has been provided and so omit the
18590 simulator library (@file{libsim.a)} from the linker command line.
18592 @item -mstack-increment=@var{size}
18593 @opindex mstack-increment
18594 Set the maximum amount for a single stack increment operation. Large
18595 values can increase the speed of programs that contain functions
18596 that need a large amount of stack space, but they can also trigger a
18597 segmentation fault if the stack is extended too much. The default
18603 @subsection MeP Options
18604 @cindex MeP options
18610 Enables the @code{abs} instruction, which is the absolute difference
18611 between two registers.
18615 Enables all the optional instructions---average, multiply, divide, bit
18616 operations, leading zero, absolute difference, min/max, clip, and
18622 Enables the @code{ave} instruction, which computes the average of two
18625 @item -mbased=@var{n}
18627 Variables of size @var{n} bytes or smaller are placed in the
18628 @code{.based} section by default. Based variables use the @code{$tp}
18629 register as a base register, and there is a 128-byte limit to the
18630 @code{.based} section.
18634 Enables the bit operation instructions---bit test (@code{btstm}), set
18635 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
18636 test-and-set (@code{tas}).
18638 @item -mc=@var{name}
18640 Selects which section constant data is placed in. @var{name} may
18641 be @samp{tiny}, @samp{near}, or @samp{far}.
18645 Enables the @code{clip} instruction. Note that @option{-mclip} is not
18646 useful unless you also provide @option{-mminmax}.
18648 @item -mconfig=@var{name}
18650 Selects one of the built-in core configurations. Each MeP chip has
18651 one or more modules in it; each module has a core CPU and a variety of
18652 coprocessors, optional instructions, and peripherals. The
18653 @code{MeP-Integrator} tool, not part of GCC, provides these
18654 configurations through this option; using this option is the same as
18655 using all the corresponding command-line options. The default
18656 configuration is @samp{default}.
18660 Enables the coprocessor instructions. By default, this is a 32-bit
18661 coprocessor. Note that the coprocessor is normally enabled via the
18662 @option{-mconfig=} option.
18666 Enables the 32-bit coprocessor's instructions.
18670 Enables the 64-bit coprocessor's instructions.
18674 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
18678 Causes constant variables to be placed in the @code{.near} section.
18682 Enables the @code{div} and @code{divu} instructions.
18686 Generate big-endian code.
18690 Generate little-endian code.
18692 @item -mio-volatile
18693 @opindex mio-volatile
18694 Tells the compiler that any variable marked with the @code{io}
18695 attribute is to be considered volatile.
18699 Causes variables to be assigned to the @code{.far} section by default.
18703 Enables the @code{leadz} (leading zero) instruction.
18707 Causes variables to be assigned to the @code{.near} section by default.
18711 Enables the @code{min} and @code{max} instructions.
18715 Enables the multiplication and multiply-accumulate instructions.
18719 Disables all the optional instructions enabled by @option{-mall-opts}.
18723 Enables the @code{repeat} and @code{erepeat} instructions, used for
18724 low-overhead looping.
18728 Causes all variables to default to the @code{.tiny} section. Note
18729 that there is a 65536-byte limit to this section. Accesses to these
18730 variables use the @code{%gp} base register.
18734 Enables the saturation instructions. Note that the compiler does not
18735 currently generate these itself, but this option is included for
18736 compatibility with other tools, like @code{as}.
18740 Link the SDRAM-based runtime instead of the default ROM-based runtime.
18744 Link the simulator run-time libraries.
18748 Link the simulator runtime libraries, excluding built-in support
18749 for reset and exception vectors and tables.
18753 Causes all functions to default to the @code{.far} section. Without
18754 this option, functions default to the @code{.near} section.
18756 @item -mtiny=@var{n}
18758 Variables that are @var{n} bytes or smaller are allocated to the
18759 @code{.tiny} section. These variables use the @code{$gp} base
18760 register. The default for this option is 4, but note that there's a
18761 65536-byte limit to the @code{.tiny} section.
18765 @node MicroBlaze Options
18766 @subsection MicroBlaze Options
18767 @cindex MicroBlaze Options
18772 @opindex msoft-float
18773 Use software emulation for floating point (default).
18776 @opindex mhard-float
18777 Use hardware floating-point instructions.
18781 Do not optimize block moves, use @code{memcpy}.
18783 @item -mno-clearbss
18784 @opindex mno-clearbss
18785 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
18787 @item -mcpu=@var{cpu-type}
18789 Use features of, and schedule code for, the given CPU.
18790 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
18791 where @var{X} is a major version, @var{YY} is the minor version, and
18792 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
18793 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
18795 @item -mxl-soft-mul
18796 @opindex mxl-soft-mul
18797 Use software multiply emulation (default).
18799 @item -mxl-soft-div
18800 @opindex mxl-soft-div
18801 Use software emulation for divides (default).
18803 @item -mxl-barrel-shift
18804 @opindex mxl-barrel-shift
18805 Use the hardware barrel shifter.
18807 @item -mxl-pattern-compare
18808 @opindex mxl-pattern-compare
18809 Use pattern compare instructions.
18811 @item -msmall-divides
18812 @opindex msmall-divides
18813 Use table lookup optimization for small signed integer divisions.
18815 @item -mxl-stack-check
18816 @opindex mxl-stack-check
18817 This option is deprecated. Use @option{-fstack-check} instead.
18820 @opindex mxl-gp-opt
18821 Use GP-relative @code{.sdata}/@code{.sbss} sections.
18823 @item -mxl-multiply-high
18824 @opindex mxl-multiply-high
18825 Use multiply high instructions for high part of 32x32 multiply.
18827 @item -mxl-float-convert
18828 @opindex mxl-float-convert
18829 Use hardware floating-point conversion instructions.
18831 @item -mxl-float-sqrt
18832 @opindex mxl-float-sqrt
18833 Use hardware floating-point square root instruction.
18836 @opindex mbig-endian
18837 Generate code for a big-endian target.
18839 @item -mlittle-endian
18840 @opindex mlittle-endian
18841 Generate code for a little-endian target.
18844 @opindex mxl-reorder
18845 Use reorder instructions (swap and byte reversed load/store).
18847 @item -mxl-mode-@var{app-model}
18848 Select application model @var{app-model}. Valid models are
18851 normal executable (default), uses startup code @file{crt0.o}.
18854 for use with Xilinx Microprocessor Debugger (XMD) based
18855 software intrusive debug agent called xmdstub. This uses startup file
18856 @file{crt1.o} and sets the start address of the program to 0x800.
18859 for applications that are loaded using a bootloader.
18860 This model uses startup file @file{crt2.o} which does not contain a processor
18861 reset vector handler. This is suitable for transferring control on a
18862 processor reset to the bootloader rather than the application.
18865 for applications that do not require any of the
18866 MicroBlaze vectors. This option may be useful for applications running
18867 within a monitoring application. This model uses @file{crt3.o} as a startup file.
18870 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
18871 @option{-mxl-mode-@var{app-model}}.
18876 @subsection MIPS Options
18877 @cindex MIPS options
18883 Generate big-endian code.
18887 Generate little-endian code. This is the default for @samp{mips*el-*-*}
18890 @item -march=@var{arch}
18892 Generate code that runs on @var{arch}, which can be the name of a
18893 generic MIPS ISA, or the name of a particular processor.
18895 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
18896 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
18897 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
18898 @samp{mips64r5} and @samp{mips64r6}.
18899 The processor names are:
18900 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
18901 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
18902 @samp{5kc}, @samp{5kf},
18904 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
18905 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
18906 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
18907 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
18908 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
18911 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
18913 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
18914 @samp{m5100}, @samp{m5101},
18915 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
18918 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
18919 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
18920 @samp{rm7000}, @samp{rm9000},
18921 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
18924 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
18925 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
18926 @samp{xlr} and @samp{xlp}.
18927 The special value @samp{from-abi} selects the
18928 most compatible architecture for the selected ABI (that is,
18929 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
18931 The native Linux/GNU toolchain also supports the value @samp{native},
18932 which selects the best architecture option for the host processor.
18933 @option{-march=native} has no effect if GCC does not recognize
18936 In processor names, a final @samp{000} can be abbreviated as @samp{k}
18937 (for example, @option{-march=r2k}). Prefixes are optional, and
18938 @samp{vr} may be written @samp{r}.
18940 Names of the form @samp{@var{n}f2_1} refer to processors with
18941 FPUs clocked at half the rate of the core, names of the form
18942 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
18943 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
18944 processors with FPUs clocked a ratio of 3:2 with respect to the core.
18945 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
18946 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
18947 accepted as synonyms for @samp{@var{n}f1_1}.
18949 GCC defines two macros based on the value of this option. The first
18950 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
18951 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
18952 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
18953 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
18954 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
18956 Note that the @code{_MIPS_ARCH} macro uses the processor names given
18957 above. In other words, it has the full prefix and does not
18958 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
18959 the macro names the resolved architecture (either @code{"mips1"} or
18960 @code{"mips3"}). It names the default architecture when no
18961 @option{-march} option is given.
18963 @item -mtune=@var{arch}
18965 Optimize for @var{arch}. Among other things, this option controls
18966 the way instructions are scheduled, and the perceived cost of arithmetic
18967 operations. The list of @var{arch} values is the same as for
18970 When this option is not used, GCC optimizes for the processor
18971 specified by @option{-march}. By using @option{-march} and
18972 @option{-mtune} together, it is possible to generate code that
18973 runs on a family of processors, but optimize the code for one
18974 particular member of that family.
18976 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
18977 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
18978 @option{-march} ones described above.
18982 Equivalent to @option{-march=mips1}.
18986 Equivalent to @option{-march=mips2}.
18990 Equivalent to @option{-march=mips3}.
18994 Equivalent to @option{-march=mips4}.
18998 Equivalent to @option{-march=mips32}.
19002 Equivalent to @option{-march=mips32r3}.
19006 Equivalent to @option{-march=mips32r5}.
19010 Equivalent to @option{-march=mips32r6}.
19014 Equivalent to @option{-march=mips64}.
19018 Equivalent to @option{-march=mips64r2}.
19022 Equivalent to @option{-march=mips64r3}.
19026 Equivalent to @option{-march=mips64r5}.
19030 Equivalent to @option{-march=mips64r6}.
19035 @opindex mno-mips16
19036 Generate (do not generate) MIPS16 code. If GCC is targeting a
19037 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
19039 MIPS16 code generation can also be controlled on a per-function basis
19040 by means of @code{mips16} and @code{nomips16} attributes.
19041 @xref{Function Attributes}, for more information.
19043 @item -mflip-mips16
19044 @opindex mflip-mips16
19045 Generate MIPS16 code on alternating functions. This option is provided
19046 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
19047 not intended for ordinary use in compiling user code.
19049 @item -minterlink-compressed
19050 @item -mno-interlink-compressed
19051 @opindex minterlink-compressed
19052 @opindex mno-interlink-compressed
19053 Require (do not require) that code using the standard (uncompressed) MIPS ISA
19054 be link-compatible with MIPS16 and microMIPS code, and vice versa.
19056 For example, code using the standard ISA encoding cannot jump directly
19057 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
19058 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
19059 knows that the target of the jump is not compressed.
19061 @item -minterlink-mips16
19062 @itemx -mno-interlink-mips16
19063 @opindex minterlink-mips16
19064 @opindex mno-interlink-mips16
19065 Aliases of @option{-minterlink-compressed} and
19066 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
19067 and are retained for backwards compatibility.
19079 Generate code for the given ABI@.
19081 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
19082 generates 64-bit code when you select a 64-bit architecture, but you
19083 can use @option{-mgp32} to get 32-bit code instead.
19085 For information about the O64 ABI, see
19086 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
19088 GCC supports a variant of the o32 ABI in which floating-point registers
19089 are 64 rather than 32 bits wide. You can select this combination with
19090 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
19091 and @code{mfhc1} instructions and is therefore only supported for
19092 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
19094 The register assignments for arguments and return values remain the
19095 same, but each scalar value is passed in a single 64-bit register
19096 rather than a pair of 32-bit registers. For example, scalar
19097 floating-point values are returned in @samp{$f0} only, not a
19098 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
19099 remains the same in that the even-numbered double-precision registers
19102 Two additional variants of the o32 ABI are supported to enable
19103 a transition from 32-bit to 64-bit registers. These are FPXX
19104 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
19105 The FPXX extension mandates that all code must execute correctly
19106 when run using 32-bit or 64-bit registers. The code can be interlinked
19107 with either FP32 or FP64, but not both.
19108 The FP64A extension is similar to the FP64 extension but forbids the
19109 use of odd-numbered single-precision registers. This can be used
19110 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
19111 processors and allows both FP32 and FP64A code to interlink and
19112 run in the same process without changing FPU modes.
19115 @itemx -mno-abicalls
19117 @opindex mno-abicalls
19118 Generate (do not generate) code that is suitable for SVR4-style
19119 dynamic objects. @option{-mabicalls} is the default for SVR4-based
19124 Generate (do not generate) code that is fully position-independent,
19125 and that can therefore be linked into shared libraries. This option
19126 only affects @option{-mabicalls}.
19128 All @option{-mabicalls} code has traditionally been position-independent,
19129 regardless of options like @option{-fPIC} and @option{-fpic}. However,
19130 as an extension, the GNU toolchain allows executables to use absolute
19131 accesses for locally-binding symbols. It can also use shorter GP
19132 initialization sequences and generate direct calls to locally-defined
19133 functions. This mode is selected by @option{-mno-shared}.
19135 @option{-mno-shared} depends on binutils 2.16 or higher and generates
19136 objects that can only be linked by the GNU linker. However, the option
19137 does not affect the ABI of the final executable; it only affects the ABI
19138 of relocatable objects. Using @option{-mno-shared} generally makes
19139 executables both smaller and quicker.
19141 @option{-mshared} is the default.
19147 Assume (do not assume) that the static and dynamic linkers
19148 support PLTs and copy relocations. This option only affects
19149 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
19150 has no effect without @option{-msym32}.
19152 You can make @option{-mplt} the default by configuring
19153 GCC with @option{--with-mips-plt}. The default is
19154 @option{-mno-plt} otherwise.
19160 Lift (do not lift) the usual restrictions on the size of the global
19163 GCC normally uses a single instruction to load values from the GOT@.
19164 While this is relatively efficient, it only works if the GOT
19165 is smaller than about 64k. Anything larger causes the linker
19166 to report an error such as:
19168 @cindex relocation truncated to fit (MIPS)
19170 relocation truncated to fit: R_MIPS_GOT16 foobar
19173 If this happens, you should recompile your code with @option{-mxgot}.
19174 This works with very large GOTs, although the code is also
19175 less efficient, since it takes three instructions to fetch the
19176 value of a global symbol.
19178 Note that some linkers can create multiple GOTs. If you have such a
19179 linker, you should only need to use @option{-mxgot} when a single object
19180 file accesses more than 64k's worth of GOT entries. Very few do.
19182 These options have no effect unless GCC is generating position
19187 Assume that general-purpose registers are 32 bits wide.
19191 Assume that general-purpose registers are 64 bits wide.
19195 Assume that floating-point registers are 32 bits wide.
19199 Assume that floating-point registers are 64 bits wide.
19203 Do not assume the width of floating-point registers.
19206 @opindex mhard-float
19207 Use floating-point coprocessor instructions.
19210 @opindex msoft-float
19211 Do not use floating-point coprocessor instructions. Implement
19212 floating-point calculations using library calls instead.
19216 Equivalent to @option{-msoft-float}, but additionally asserts that the
19217 program being compiled does not perform any floating-point operations.
19218 This option is presently supported only by some bare-metal MIPS
19219 configurations, where it may select a special set of libraries
19220 that lack all floating-point support (including, for example, the
19221 floating-point @code{printf} formats).
19222 If code compiled with @option{-mno-float} accidentally contains
19223 floating-point operations, it is likely to suffer a link-time
19224 or run-time failure.
19226 @item -msingle-float
19227 @opindex msingle-float
19228 Assume that the floating-point coprocessor only supports single-precision
19231 @item -mdouble-float
19232 @opindex mdouble-float
19233 Assume that the floating-point coprocessor supports double-precision
19234 operations. This is the default.
19237 @itemx -mno-odd-spreg
19238 @opindex modd-spreg
19239 @opindex mno-odd-spreg
19240 Enable the use of odd-numbered single-precision floating-point registers
19241 for the o32 ABI. This is the default for processors that are known to
19242 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
19246 @itemx -mabs=legacy
19248 @opindex mabs=legacy
19249 These options control the treatment of the special not-a-number (NaN)
19250 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
19251 @code{neg.@i{fmt}} machine instructions.
19253 By default or when @option{-mabs=legacy} is used the legacy
19254 treatment is selected. In this case these instructions are considered
19255 arithmetic and avoided where correct operation is required and the
19256 input operand might be a NaN. A longer sequence of instructions that
19257 manipulate the sign bit of floating-point datum manually is used
19258 instead unless the @option{-ffinite-math-only} option has also been
19261 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
19262 this case these instructions are considered non-arithmetic and therefore
19263 operating correctly in all cases, including in particular where the
19264 input operand is a NaN. These instructions are therefore always used
19265 for the respective operations.
19268 @itemx -mnan=legacy
19270 @opindex mnan=legacy
19271 These options control the encoding of the special not-a-number (NaN)
19272 IEEE 754 floating-point data.
19274 The @option{-mnan=legacy} option selects the legacy encoding. In this
19275 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
19276 significand field being 0, whereas signaling NaNs (sNaNs) are denoted
19277 by the first bit of their trailing significand field being 1.
19279 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
19280 this case qNaNs are denoted by the first bit of their trailing
19281 significand field being 1, whereas sNaNs are denoted by the first bit of
19282 their trailing significand field being 0.
19284 The default is @option{-mnan=legacy} unless GCC has been configured with
19285 @option{--with-nan=2008}.
19291 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
19292 implement atomic memory built-in functions. When neither option is
19293 specified, GCC uses the instructions if the target architecture
19296 @option{-mllsc} is useful if the runtime environment can emulate the
19297 instructions and @option{-mno-llsc} can be useful when compiling for
19298 nonstandard ISAs. You can make either option the default by
19299 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
19300 respectively. @option{--with-llsc} is the default for some
19301 configurations; see the installation documentation for details.
19307 Use (do not use) revision 1 of the MIPS DSP ASE@.
19308 @xref{MIPS DSP Built-in Functions}. This option defines the
19309 preprocessor macro @code{__mips_dsp}. It also defines
19310 @code{__mips_dsp_rev} to 1.
19316 Use (do not use) revision 2 of the MIPS DSP ASE@.
19317 @xref{MIPS DSP Built-in Functions}. This option defines the
19318 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
19319 It also defines @code{__mips_dsp_rev} to 2.
19322 @itemx -mno-smartmips
19323 @opindex msmartmips
19324 @opindex mno-smartmips
19325 Use (do not use) the MIPS SmartMIPS ASE.
19327 @item -mpaired-single
19328 @itemx -mno-paired-single
19329 @opindex mpaired-single
19330 @opindex mno-paired-single
19331 Use (do not use) paired-single floating-point instructions.
19332 @xref{MIPS Paired-Single Support}. This option requires
19333 hardware floating-point support to be enabled.
19339 Use (do not use) MIPS Digital Media Extension instructions.
19340 This option can only be used when generating 64-bit code and requires
19341 hardware floating-point support to be enabled.
19346 @opindex mno-mips3d
19347 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
19348 The option @option{-mips3d} implies @option{-mpaired-single}.
19351 @itemx -mno-micromips
19352 @opindex mmicromips
19353 @opindex mno-mmicromips
19354 Generate (do not generate) microMIPS code.
19356 MicroMIPS code generation can also be controlled on a per-function basis
19357 by means of @code{micromips} and @code{nomicromips} attributes.
19358 @xref{Function Attributes}, for more information.
19364 Use (do not use) MT Multithreading instructions.
19370 Use (do not use) the MIPS MCU ASE instructions.
19376 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
19382 Use (do not use) the MIPS Virtualization Application Specific instructions.
19388 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
19392 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
19393 an explanation of the default and the way that the pointer size is
19398 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
19400 The default size of @code{int}s, @code{long}s and pointers depends on
19401 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
19402 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
19403 32-bit @code{long}s. Pointers are the same size as @code{long}s,
19404 or the same size as integer registers, whichever is smaller.
19410 Assume (do not assume) that all symbols have 32-bit values, regardless
19411 of the selected ABI@. This option is useful in combination with
19412 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
19413 to generate shorter and faster references to symbolic addresses.
19417 Put definitions of externally-visible data in a small data section
19418 if that data is no bigger than @var{num} bytes. GCC can then generate
19419 more efficient accesses to the data; see @option{-mgpopt} for details.
19421 The default @option{-G} option depends on the configuration.
19423 @item -mlocal-sdata
19424 @itemx -mno-local-sdata
19425 @opindex mlocal-sdata
19426 @opindex mno-local-sdata
19427 Extend (do not extend) the @option{-G} behavior to local data too,
19428 such as to static variables in C@. @option{-mlocal-sdata} is the
19429 default for all configurations.
19431 If the linker complains that an application is using too much small data,
19432 you might want to try rebuilding the less performance-critical parts with
19433 @option{-mno-local-sdata}. You might also want to build large
19434 libraries with @option{-mno-local-sdata}, so that the libraries leave
19435 more room for the main program.
19437 @item -mextern-sdata
19438 @itemx -mno-extern-sdata
19439 @opindex mextern-sdata
19440 @opindex mno-extern-sdata
19441 Assume (do not assume) that externally-defined data is in
19442 a small data section if the size of that data is within the @option{-G} limit.
19443 @option{-mextern-sdata} is the default for all configurations.
19445 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
19446 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
19447 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
19448 is placed in a small data section. If @var{Var} is defined by another
19449 module, you must either compile that module with a high-enough
19450 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
19451 definition. If @var{Var} is common, you must link the application
19452 with a high-enough @option{-G} setting.
19454 The easiest way of satisfying these restrictions is to compile
19455 and link every module with the same @option{-G} option. However,
19456 you may wish to build a library that supports several different
19457 small data limits. You can do this by compiling the library with
19458 the highest supported @option{-G} setting and additionally using
19459 @option{-mno-extern-sdata} to stop the library from making assumptions
19460 about externally-defined data.
19466 Use (do not use) GP-relative accesses for symbols that are known to be
19467 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
19468 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
19471 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
19472 might not hold the value of @code{_gp}. For example, if the code is
19473 part of a library that might be used in a boot monitor, programs that
19474 call boot monitor routines pass an unknown value in @code{$gp}.
19475 (In such situations, the boot monitor itself is usually compiled
19476 with @option{-G0}.)
19478 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
19479 @option{-mno-extern-sdata}.
19481 @item -membedded-data
19482 @itemx -mno-embedded-data
19483 @opindex membedded-data
19484 @opindex mno-embedded-data
19485 Allocate variables to the read-only data section first if possible, then
19486 next in the small data section if possible, otherwise in data. This gives
19487 slightly slower code than the default, but reduces the amount of RAM required
19488 when executing, and thus may be preferred for some embedded systems.
19490 @item -muninit-const-in-rodata
19491 @itemx -mno-uninit-const-in-rodata
19492 @opindex muninit-const-in-rodata
19493 @opindex mno-uninit-const-in-rodata
19494 Put uninitialized @code{const} variables in the read-only data section.
19495 This option is only meaningful in conjunction with @option{-membedded-data}.
19497 @item -mcode-readable=@var{setting}
19498 @opindex mcode-readable
19499 Specify whether GCC may generate code that reads from executable sections.
19500 There are three possible settings:
19503 @item -mcode-readable=yes
19504 Instructions may freely access executable sections. This is the
19507 @item -mcode-readable=pcrel
19508 MIPS16 PC-relative load instructions can access executable sections,
19509 but other instructions must not do so. This option is useful on 4KSc
19510 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
19511 It is also useful on processors that can be configured to have a dual
19512 instruction/data SRAM interface and that, like the M4K, automatically
19513 redirect PC-relative loads to the instruction RAM.
19515 @item -mcode-readable=no
19516 Instructions must not access executable sections. This option can be
19517 useful on targets that are configured to have a dual instruction/data
19518 SRAM interface but that (unlike the M4K) do not automatically redirect
19519 PC-relative loads to the instruction RAM.
19522 @item -msplit-addresses
19523 @itemx -mno-split-addresses
19524 @opindex msplit-addresses
19525 @opindex mno-split-addresses
19526 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
19527 relocation operators. This option has been superseded by
19528 @option{-mexplicit-relocs} but is retained for backwards compatibility.
19530 @item -mexplicit-relocs
19531 @itemx -mno-explicit-relocs
19532 @opindex mexplicit-relocs
19533 @opindex mno-explicit-relocs
19534 Use (do not use) assembler relocation operators when dealing with symbolic
19535 addresses. The alternative, selected by @option{-mno-explicit-relocs},
19536 is to use assembler macros instead.
19538 @option{-mexplicit-relocs} is the default if GCC was configured
19539 to use an assembler that supports relocation operators.
19541 @item -mcheck-zero-division
19542 @itemx -mno-check-zero-division
19543 @opindex mcheck-zero-division
19544 @opindex mno-check-zero-division
19545 Trap (do not trap) on integer division by zero.
19547 The default is @option{-mcheck-zero-division}.
19549 @item -mdivide-traps
19550 @itemx -mdivide-breaks
19551 @opindex mdivide-traps
19552 @opindex mdivide-breaks
19553 MIPS systems check for division by zero by generating either a
19554 conditional trap or a break instruction. Using traps results in
19555 smaller code, but is only supported on MIPS II and later. Also, some
19556 versions of the Linux kernel have a bug that prevents trap from
19557 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
19558 allow conditional traps on architectures that support them and
19559 @option{-mdivide-breaks} to force the use of breaks.
19561 The default is usually @option{-mdivide-traps}, but this can be
19562 overridden at configure time using @option{--with-divide=breaks}.
19563 Divide-by-zero checks can be completely disabled using
19564 @option{-mno-check-zero-division}.
19569 @opindex mno-memcpy
19570 Force (do not force) the use of @code{memcpy} for non-trivial block
19571 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
19572 most constant-sized copies.
19575 @itemx -mno-long-calls
19576 @opindex mlong-calls
19577 @opindex mno-long-calls
19578 Disable (do not disable) use of the @code{jal} instruction. Calling
19579 functions using @code{jal} is more efficient but requires the caller
19580 and callee to be in the same 256 megabyte segment.
19582 This option has no effect on abicalls code. The default is
19583 @option{-mno-long-calls}.
19589 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
19590 instructions, as provided by the R4650 ISA@.
19596 Enable (disable) use of the @code{madd} and @code{msub} integer
19597 instructions. The default is @option{-mimadd} on architectures
19598 that support @code{madd} and @code{msub} except for the 74k
19599 architecture where it was found to generate slower code.
19602 @itemx -mno-fused-madd
19603 @opindex mfused-madd
19604 @opindex mno-fused-madd
19605 Enable (disable) use of the floating-point multiply-accumulate
19606 instructions, when they are available. The default is
19607 @option{-mfused-madd}.
19609 On the R8000 CPU when multiply-accumulate instructions are used,
19610 the intermediate product is calculated to infinite precision
19611 and is not subject to the FCSR Flush to Zero bit. This may be
19612 undesirable in some circumstances. On other processors the result
19613 is numerically identical to the equivalent computation using
19614 separate multiply, add, subtract and negate instructions.
19618 Tell the MIPS assembler to not run its preprocessor over user
19619 assembler files (with a @samp{.s} suffix) when assembling them.
19624 @opindex mno-fix-24k
19625 Work around the 24K E48 (lost data on stores during refill) errata.
19626 The workarounds are implemented by the assembler rather than by GCC@.
19629 @itemx -mno-fix-r4000
19630 @opindex mfix-r4000
19631 @opindex mno-fix-r4000
19632 Work around certain R4000 CPU errata:
19635 A double-word or a variable shift may give an incorrect result if executed
19636 immediately after starting an integer division.
19638 A double-word or a variable shift may give an incorrect result if executed
19639 while an integer multiplication is in progress.
19641 An integer division may give an incorrect result if started in a delay slot
19642 of a taken branch or a jump.
19646 @itemx -mno-fix-r4400
19647 @opindex mfix-r4400
19648 @opindex mno-fix-r4400
19649 Work around certain R4400 CPU errata:
19652 A double-word or a variable shift may give an incorrect result if executed
19653 immediately after starting an integer division.
19657 @itemx -mno-fix-r10000
19658 @opindex mfix-r10000
19659 @opindex mno-fix-r10000
19660 Work around certain R10000 errata:
19663 @code{ll}/@code{sc} sequences may not behave atomically on revisions
19664 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
19667 This option can only be used if the target architecture supports
19668 branch-likely instructions. @option{-mfix-r10000} is the default when
19669 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
19673 @itemx -mno-fix-rm7000
19674 @opindex mfix-rm7000
19675 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
19676 workarounds are implemented by the assembler rather than by GCC@.
19679 @itemx -mno-fix-vr4120
19680 @opindex mfix-vr4120
19681 Work around certain VR4120 errata:
19684 @code{dmultu} does not always produce the correct result.
19686 @code{div} and @code{ddiv} do not always produce the correct result if one
19687 of the operands is negative.
19689 The workarounds for the division errata rely on special functions in
19690 @file{libgcc.a}. At present, these functions are only provided by
19691 the @code{mips64vr*-elf} configurations.
19693 Other VR4120 errata require a NOP to be inserted between certain pairs of
19694 instructions. These errata are handled by the assembler, not by GCC itself.
19697 @opindex mfix-vr4130
19698 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
19699 workarounds are implemented by the assembler rather than by GCC,
19700 although GCC avoids using @code{mflo} and @code{mfhi} if the
19701 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
19702 instructions are available instead.
19705 @itemx -mno-fix-sb1
19707 Work around certain SB-1 CPU core errata.
19708 (This flag currently works around the SB-1 revision 2
19709 ``F1'' and ``F2'' floating-point errata.)
19711 @item -mr10k-cache-barrier=@var{setting}
19712 @opindex mr10k-cache-barrier
19713 Specify whether GCC should insert cache barriers to avoid the
19714 side-effects of speculation on R10K processors.
19716 In common with many processors, the R10K tries to predict the outcome
19717 of a conditional branch and speculatively executes instructions from
19718 the ``taken'' branch. It later aborts these instructions if the
19719 predicted outcome is wrong. However, on the R10K, even aborted
19720 instructions can have side effects.
19722 This problem only affects kernel stores and, depending on the system,
19723 kernel loads. As an example, a speculatively-executed store may load
19724 the target memory into cache and mark the cache line as dirty, even if
19725 the store itself is later aborted. If a DMA operation writes to the
19726 same area of memory before the ``dirty'' line is flushed, the cached
19727 data overwrites the DMA-ed data. See the R10K processor manual
19728 for a full description, including other potential problems.
19730 One workaround is to insert cache barrier instructions before every memory
19731 access that might be speculatively executed and that might have side
19732 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
19733 controls GCC's implementation of this workaround. It assumes that
19734 aborted accesses to any byte in the following regions does not have
19739 the memory occupied by the current function's stack frame;
19742 the memory occupied by an incoming stack argument;
19745 the memory occupied by an object with a link-time-constant address.
19748 It is the kernel's responsibility to ensure that speculative
19749 accesses to these regions are indeed safe.
19751 If the input program contains a function declaration such as:
19757 then the implementation of @code{foo} must allow @code{j foo} and
19758 @code{jal foo} to be executed speculatively. GCC honors this
19759 restriction for functions it compiles itself. It expects non-GCC
19760 functions (such as hand-written assembly code) to do the same.
19762 The option has three forms:
19765 @item -mr10k-cache-barrier=load-store
19766 Insert a cache barrier before a load or store that might be
19767 speculatively executed and that might have side effects even
19770 @item -mr10k-cache-barrier=store
19771 Insert a cache barrier before a store that might be speculatively
19772 executed and that might have side effects even if aborted.
19774 @item -mr10k-cache-barrier=none
19775 Disable the insertion of cache barriers. This is the default setting.
19778 @item -mflush-func=@var{func}
19779 @itemx -mno-flush-func
19780 @opindex mflush-func
19781 Specifies the function to call to flush the I and D caches, or to not
19782 call any such function. If called, the function must take the same
19783 arguments as the common @code{_flush_func}, that is, the address of the
19784 memory range for which the cache is being flushed, the size of the
19785 memory range, and the number 3 (to flush both caches). The default
19786 depends on the target GCC was configured for, but commonly is either
19787 @code{_flush_func} or @code{__cpu_flush}.
19789 @item mbranch-cost=@var{num}
19790 @opindex mbranch-cost
19791 Set the cost of branches to roughly @var{num} ``simple'' instructions.
19792 This cost is only a heuristic and is not guaranteed to produce
19793 consistent results across releases. A zero cost redundantly selects
19794 the default, which is based on the @option{-mtune} setting.
19796 @item -mbranch-likely
19797 @itemx -mno-branch-likely
19798 @opindex mbranch-likely
19799 @opindex mno-branch-likely
19800 Enable or disable use of Branch Likely instructions, regardless of the
19801 default for the selected architecture. By default, Branch Likely
19802 instructions may be generated if they are supported by the selected
19803 architecture. An exception is for the MIPS32 and MIPS64 architectures
19804 and processors that implement those architectures; for those, Branch
19805 Likely instructions are not be generated by default because the MIPS32
19806 and MIPS64 architectures specifically deprecate their use.
19808 @item -mcompact-branches=never
19809 @itemx -mcompact-branches=optimal
19810 @itemx -mcompact-branches=always
19811 @opindex mcompact-branches=never
19812 @opindex mcompact-branches=optimal
19813 @opindex mcompact-branches=always
19814 These options control which form of branches will be generated. The
19815 default is @option{-mcompact-branches=optimal}.
19817 The @option{-mcompact-branches=never} option ensures that compact branch
19818 instructions will never be generated.
19820 The @option{-mcompact-branches=always} option ensures that a compact
19821 branch instruction will be generated if available. If a compact branch
19822 instruction is not available, a delay slot form of the branch will be
19825 This option is supported from MIPS Release 6 onwards.
19827 The @option{-mcompact-branches=optimal} option will cause a delay slot
19828 branch to be used if one is available in the current ISA and the delay
19829 slot is successfully filled. If the delay slot is not filled, a compact
19830 branch will be chosen if one is available.
19832 @item -mfp-exceptions
19833 @itemx -mno-fp-exceptions
19834 @opindex mfp-exceptions
19835 Specifies whether FP exceptions are enabled. This affects how
19836 FP instructions are scheduled for some processors.
19837 The default is that FP exceptions are
19840 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
19841 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
19844 @item -mvr4130-align
19845 @itemx -mno-vr4130-align
19846 @opindex mvr4130-align
19847 The VR4130 pipeline is two-way superscalar, but can only issue two
19848 instructions together if the first one is 8-byte aligned. When this
19849 option is enabled, GCC aligns pairs of instructions that it
19850 thinks should execute in parallel.
19852 This option only has an effect when optimizing for the VR4130.
19853 It normally makes code faster, but at the expense of making it bigger.
19854 It is enabled by default at optimization level @option{-O3}.
19859 Enable (disable) generation of @code{synci} instructions on
19860 architectures that support it. The @code{synci} instructions (if
19861 enabled) are generated when @code{__builtin___clear_cache} is
19864 This option defaults to @option{-mno-synci}, but the default can be
19865 overridden by configuring GCC with @option{--with-synci}.
19867 When compiling code for single processor systems, it is generally safe
19868 to use @code{synci}. However, on many multi-core (SMP) systems, it
19869 does not invalidate the instruction caches on all cores and may lead
19870 to undefined behavior.
19872 @item -mrelax-pic-calls
19873 @itemx -mno-relax-pic-calls
19874 @opindex mrelax-pic-calls
19875 Try to turn PIC calls that are normally dispatched via register
19876 @code{$25} into direct calls. This is only possible if the linker can
19877 resolve the destination at link time and if the destination is within
19878 range for a direct call.
19880 @option{-mrelax-pic-calls} is the default if GCC was configured to use
19881 an assembler and a linker that support the @code{.reloc} assembly
19882 directive and @option{-mexplicit-relocs} is in effect. With
19883 @option{-mno-explicit-relocs}, this optimization can be performed by the
19884 assembler and the linker alone without help from the compiler.
19886 @item -mmcount-ra-address
19887 @itemx -mno-mcount-ra-address
19888 @opindex mmcount-ra-address
19889 @opindex mno-mcount-ra-address
19890 Emit (do not emit) code that allows @code{_mcount} to modify the
19891 calling function's return address. When enabled, this option extends
19892 the usual @code{_mcount} interface with a new @var{ra-address}
19893 parameter, which has type @code{intptr_t *} and is passed in register
19894 @code{$12}. @code{_mcount} can then modify the return address by
19895 doing both of the following:
19898 Returning the new address in register @code{$31}.
19900 Storing the new address in @code{*@var{ra-address}},
19901 if @var{ra-address} is nonnull.
19904 The default is @option{-mno-mcount-ra-address}.
19906 @item -mframe-header-opt
19907 @itemx -mno-frame-header-opt
19908 @opindex mframe-header-opt
19909 Enable (disable) frame header optimization in the o32 ABI. When using the
19910 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
19911 function to write out register arguments. When enabled, this optimization
19912 will suppress the allocation of the frame header if it can be determined that
19915 This optimization is off by default at all optimization levels.
19920 @subsection MMIX Options
19921 @cindex MMIX Options
19923 These options are defined for the MMIX:
19927 @itemx -mno-libfuncs
19929 @opindex mno-libfuncs
19930 Specify that intrinsic library functions are being compiled, passing all
19931 values in registers, no matter the size.
19934 @itemx -mno-epsilon
19936 @opindex mno-epsilon
19937 Generate floating-point comparison instructions that compare with respect
19938 to the @code{rE} epsilon register.
19940 @item -mabi=mmixware
19942 @opindex mabi=mmixware
19944 Generate code that passes function parameters and return values that (in
19945 the called function) are seen as registers @code{$0} and up, as opposed to
19946 the GNU ABI which uses global registers @code{$231} and up.
19948 @item -mzero-extend
19949 @itemx -mno-zero-extend
19950 @opindex mzero-extend
19951 @opindex mno-zero-extend
19952 When reading data from memory in sizes shorter than 64 bits, use (do not
19953 use) zero-extending load instructions by default, rather than
19954 sign-extending ones.
19957 @itemx -mno-knuthdiv
19959 @opindex mno-knuthdiv
19960 Make the result of a division yielding a remainder have the same sign as
19961 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
19962 remainder follows the sign of the dividend. Both methods are
19963 arithmetically valid, the latter being almost exclusively used.
19965 @item -mtoplevel-symbols
19966 @itemx -mno-toplevel-symbols
19967 @opindex mtoplevel-symbols
19968 @opindex mno-toplevel-symbols
19969 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
19970 code can be used with the @code{PREFIX} assembly directive.
19974 Generate an executable in the ELF format, rather than the default
19975 @samp{mmo} format used by the @command{mmix} simulator.
19977 @item -mbranch-predict
19978 @itemx -mno-branch-predict
19979 @opindex mbranch-predict
19980 @opindex mno-branch-predict
19981 Use (do not use) the probable-branch instructions, when static branch
19982 prediction indicates a probable branch.
19984 @item -mbase-addresses
19985 @itemx -mno-base-addresses
19986 @opindex mbase-addresses
19987 @opindex mno-base-addresses
19988 Generate (do not generate) code that uses @emph{base addresses}. Using a
19989 base address automatically generates a request (handled by the assembler
19990 and the linker) for a constant to be set up in a global register. The
19991 register is used for one or more base address requests within the range 0
19992 to 255 from the value held in the register. The generally leads to short
19993 and fast code, but the number of different data items that can be
19994 addressed is limited. This means that a program that uses lots of static
19995 data may require @option{-mno-base-addresses}.
19997 @item -msingle-exit
19998 @itemx -mno-single-exit
19999 @opindex msingle-exit
20000 @opindex mno-single-exit
20001 Force (do not force) generated code to have a single exit point in each
20005 @node MN10300 Options
20006 @subsection MN10300 Options
20007 @cindex MN10300 options
20009 These @option{-m} options are defined for Matsushita MN10300 architectures:
20014 Generate code to avoid bugs in the multiply instructions for the MN10300
20015 processors. This is the default.
20017 @item -mno-mult-bug
20018 @opindex mno-mult-bug
20019 Do not generate code to avoid bugs in the multiply instructions for the
20020 MN10300 processors.
20024 Generate code using features specific to the AM33 processor.
20028 Do not generate code using features specific to the AM33 processor. This
20033 Generate code using features specific to the AM33/2.0 processor.
20037 Generate code using features specific to the AM34 processor.
20039 @item -mtune=@var{cpu-type}
20041 Use the timing characteristics of the indicated CPU type when
20042 scheduling instructions. This does not change the targeted processor
20043 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
20044 @samp{am33-2} or @samp{am34}.
20046 @item -mreturn-pointer-on-d0
20047 @opindex mreturn-pointer-on-d0
20048 When generating a function that returns a pointer, return the pointer
20049 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
20050 only in @code{a0}, and attempts to call such functions without a prototype
20051 result in errors. Note that this option is on by default; use
20052 @option{-mno-return-pointer-on-d0} to disable it.
20056 Do not link in the C run-time initialization object file.
20060 Indicate to the linker that it should perform a relaxation optimization pass
20061 to shorten branches, calls and absolute memory addresses. This option only
20062 has an effect when used on the command line for the final link step.
20064 This option makes symbolic debugging impossible.
20068 Allow the compiler to generate @emph{Long Instruction Word}
20069 instructions if the target is the @samp{AM33} or later. This is the
20070 default. This option defines the preprocessor macro @code{__LIW__}.
20074 Do not allow the compiler to generate @emph{Long Instruction Word}
20075 instructions. This option defines the preprocessor macro
20080 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
20081 instructions if the target is the @samp{AM33} or later. This is the
20082 default. This option defines the preprocessor macro @code{__SETLB__}.
20086 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
20087 instructions. This option defines the preprocessor macro
20088 @code{__NO_SETLB__}.
20092 @node Moxie Options
20093 @subsection Moxie Options
20094 @cindex Moxie Options
20100 Generate big-endian code. This is the default for @samp{moxie-*-*}
20105 Generate little-endian code.
20109 Generate mul.x and umul.x instructions. This is the default for
20110 @samp{moxiebox-*-*} configurations.
20114 Do not link in the C run-time initialization object file.
20118 @node MSP430 Options
20119 @subsection MSP430 Options
20120 @cindex MSP430 Options
20122 These options are defined for the MSP430:
20128 Force assembly output to always use hex constants. Normally such
20129 constants are signed decimals, but this option is available for
20130 testsuite and/or aesthetic purposes.
20134 Select the MCU to target. This is used to create a C preprocessor
20135 symbol based upon the MCU name, converted to upper case and pre- and
20136 post-fixed with @samp{__}. This in turn is used by the
20137 @file{msp430.h} header file to select an MCU-specific supplementary
20140 The option also sets the ISA to use. If the MCU name is one that is
20141 known to only support the 430 ISA then that is selected, otherwise the
20142 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
20143 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
20144 name selects the 430X ISA.
20146 In addition an MCU-specific linker script is added to the linker
20147 command line. The script's name is the name of the MCU with
20148 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
20149 command line defines the C preprocessor symbol @code{__XXX__} and
20150 cause the linker to search for a script called @file{xxx.ld}.
20152 This option is also passed on to the assembler.
20155 @itemx -mno-warn-mcu
20157 @opindex mno-warn-mcu
20158 This option enables or disables warnings about conflicts between the
20159 MCU name specified by the @option{-mmcu} option and the ISA set by the
20160 @option{-mcpu} option and/or the hardware multiply support set by the
20161 @option{-mhwmult} option. It also toggles warnings about unrecognized
20162 MCU names. This option is on by default.
20166 Specifies the ISA to use. Accepted values are @samp{msp430},
20167 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
20168 @option{-mmcu=} option should be used to select the ISA.
20172 Link to the simulator runtime libraries and linker script. Overrides
20173 any scripts that would be selected by the @option{-mmcu=} option.
20177 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
20181 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
20185 This option is passed to the assembler and linker, and allows the
20186 linker to perform certain optimizations that cannot be done until
20191 Describes the type of hardware multiply supported by the target.
20192 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
20193 for the original 16-bit-only multiply supported by early MCUs.
20194 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
20195 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
20196 A value of @samp{auto} can also be given. This tells GCC to deduce
20197 the hardware multiply support based upon the MCU name provided by the
20198 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
20199 the MCU name is not recognized then no hardware multiply support is
20200 assumed. @code{auto} is the default setting.
20202 Hardware multiplies are normally performed by calling a library
20203 routine. This saves space in the generated code. When compiling at
20204 @option{-O3} or higher however the hardware multiplier is invoked
20205 inline. This makes for bigger, but faster code.
20207 The hardware multiply routines disable interrupts whilst running and
20208 restore the previous interrupt state when they finish. This makes
20209 them safe to use inside interrupt handlers as well as in normal code.
20213 Enable the use of a minimum runtime environment - no static
20214 initializers or constructors. This is intended for memory-constrained
20215 devices. The compiler includes special symbols in some objects
20216 that tell the linker and runtime which code fragments are required.
20218 @item -mcode-region=
20219 @itemx -mdata-region=
20220 @opindex mcode-region
20221 @opindex mdata-region
20222 These options tell the compiler where to place functions and data that
20223 do not have one of the @code{lower}, @code{upper}, @code{either} or
20224 @code{section} attributes. Possible values are @code{lower},
20225 @code{upper}, @code{either} or @code{any}. The first three behave
20226 like the corresponding attribute. The fourth possible value -
20227 @code{any} - is the default. It leaves placement entirely up to the
20228 linker script and how it assigns the standard sections
20229 (@code{.text}, @code{.data}, etc) to the memory regions.
20231 @item -msilicon-errata=
20232 @opindex msilicon-errata
20233 This option passes on a request to assembler to enable the fixes for
20234 the named silicon errata.
20236 @item -msilicon-errata-warn=
20237 @opindex msilicon-errata-warn
20238 This option passes on a request to the assembler to enable warning
20239 messages when a silicon errata might need to be applied.
20243 @node NDS32 Options
20244 @subsection NDS32 Options
20245 @cindex NDS32 Options
20247 These options are defined for NDS32 implementations:
20252 @opindex mbig-endian
20253 Generate code in big-endian mode.
20255 @item -mlittle-endian
20256 @opindex mlittle-endian
20257 Generate code in little-endian mode.
20259 @item -mreduced-regs
20260 @opindex mreduced-regs
20261 Use reduced-set registers for register allocation.
20264 @opindex mfull-regs
20265 Use full-set registers for register allocation.
20269 Generate conditional move instructions.
20273 Do not generate conditional move instructions.
20277 Generate performance extension instructions.
20279 @item -mno-perf-ext
20280 @opindex mno-perf-ext
20281 Do not generate performance extension instructions.
20285 Generate v3 push25/pop25 instructions.
20288 @opindex mno-v3push
20289 Do not generate v3 push25/pop25 instructions.
20293 Generate 16-bit instructions.
20296 @opindex mno-16-bit
20297 Do not generate 16-bit instructions.
20299 @item -misr-vector-size=@var{num}
20300 @opindex misr-vector-size
20301 Specify the size of each interrupt vector, which must be 4 or 16.
20303 @item -mcache-block-size=@var{num}
20304 @opindex mcache-block-size
20305 Specify the size of each cache block,
20306 which must be a power of 2 between 4 and 512.
20308 @item -march=@var{arch}
20310 Specify the name of the target architecture.
20312 @item -mcmodel=@var{code-model}
20314 Set the code model to one of
20317 All the data and read-only data segments must be within 512KB addressing space.
20318 The text segment must be within 16MB addressing space.
20319 @item @samp{medium}
20320 The data segment must be within 512KB while the read-only data segment can be
20321 within 4GB addressing space. The text segment should be still within 16MB
20324 All the text and data segments can be within 4GB addressing space.
20328 @opindex mctor-dtor
20329 Enable constructor/destructor feature.
20333 Guide linker to relax instructions.
20337 @node Nios II Options
20338 @subsection Nios II Options
20339 @cindex Nios II options
20340 @cindex Altera Nios II options
20342 These are the options defined for the Altera Nios II processor.
20348 @cindex smaller data references
20349 Put global and static objects less than or equal to @var{num} bytes
20350 into the small data or BSS sections instead of the normal data or BSS
20351 sections. The default value of @var{num} is 8.
20353 @item -mgpopt=@var{option}
20358 Generate (do not generate) GP-relative accesses. The following
20359 @var{option} names are recognized:
20364 Do not generate GP-relative accesses.
20367 Generate GP-relative accesses for small data objects that are not
20368 external, weak, or uninitialized common symbols.
20369 Also use GP-relative addressing for objects that
20370 have been explicitly placed in a small data section via a @code{section}
20374 As for @samp{local}, but also generate GP-relative accesses for
20375 small data objects that are external, weak, or common. If you use this option,
20376 you must ensure that all parts of your program (including libraries) are
20377 compiled with the same @option{-G} setting.
20380 Generate GP-relative accesses for all data objects in the program. If you
20381 use this option, the entire data and BSS segments
20382 of your program must fit in 64K of memory and you must use an appropriate
20383 linker script to allocate them within the addressable range of the
20387 Generate GP-relative addresses for function pointers as well as data
20388 pointers. If you use this option, the entire text, data, and BSS segments
20389 of your program must fit in 64K of memory and you must use an appropriate
20390 linker script to allocate them within the addressable range of the
20395 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
20396 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
20398 The default is @option{-mgpopt} except when @option{-fpic} or
20399 @option{-fPIC} is specified to generate position-independent code.
20400 Note that the Nios II ABI does not permit GP-relative accesses from
20403 You may need to specify @option{-mno-gpopt} explicitly when building
20404 programs that include large amounts of small data, including large
20405 GOT data sections. In this case, the 16-bit offset for GP-relative
20406 addressing may not be large enough to allow access to the entire
20407 small data section.
20413 Generate little-endian (default) or big-endian (experimental) code,
20416 @item -march=@var{arch}
20418 This specifies the name of the target Nios II architecture. GCC uses this
20419 name to determine what kind of instructions it can emit when generating
20420 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
20422 The preprocessor macro @code{__nios2_arch__} is available to programs,
20423 with value 1 or 2, indicating the targeted ISA level.
20425 @item -mbypass-cache
20426 @itemx -mno-bypass-cache
20427 @opindex mno-bypass-cache
20428 @opindex mbypass-cache
20429 Force all load and store instructions to always bypass cache by
20430 using I/O variants of the instructions. The default is not to
20433 @item -mno-cache-volatile
20434 @itemx -mcache-volatile
20435 @opindex mcache-volatile
20436 @opindex mno-cache-volatile
20437 Volatile memory access bypass the cache using the I/O variants of
20438 the load and store instructions. The default is not to bypass the cache.
20440 @item -mno-fast-sw-div
20441 @itemx -mfast-sw-div
20442 @opindex mno-fast-sw-div
20443 @opindex mfast-sw-div
20444 Do not use table-based fast divide for small numbers. The default
20445 is to use the fast divide at @option{-O3} and above.
20449 @itemx -mno-hw-mulx
20453 @opindex mno-hw-mul
20455 @opindex mno-hw-mulx
20457 @opindex mno-hw-div
20459 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
20460 instructions by the compiler. The default is to emit @code{mul}
20461 and not emit @code{div} and @code{mulx}.
20467 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
20468 CDX (code density) instructions. Enabling these instructions also
20469 requires @option{-march=r2}. Since these instructions are optional
20470 extensions to the R2 architecture, the default is not to emit them.
20472 @item -mcustom-@var{insn}=@var{N}
20473 @itemx -mno-custom-@var{insn}
20474 @opindex mcustom-@var{insn}
20475 @opindex mno-custom-@var{insn}
20476 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
20477 custom instruction with encoding @var{N} when generating code that uses
20478 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
20479 instruction 253 for single-precision floating-point add operations instead
20480 of the default behavior of using a library call.
20482 The following values of @var{insn} are supported. Except as otherwise
20483 noted, floating-point operations are expected to be implemented with
20484 normal IEEE 754 semantics and correspond directly to the C operators or the
20485 equivalent GCC built-in functions (@pxref{Other Builtins}).
20487 Single-precision floating point:
20490 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
20491 Binary arithmetic operations.
20497 Unary absolute value.
20499 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
20500 Comparison operations.
20502 @item @samp{fmins}, @samp{fmaxs}
20503 Floating-point minimum and maximum. These instructions are only
20504 generated if @option{-ffinite-math-only} is specified.
20506 @item @samp{fsqrts}
20507 Unary square root operation.
20509 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
20510 Floating-point trigonometric and exponential functions. These instructions
20511 are only generated if @option{-funsafe-math-optimizations} is also specified.
20515 Double-precision floating point:
20518 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
20519 Binary arithmetic operations.
20525 Unary absolute value.
20527 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
20528 Comparison operations.
20530 @item @samp{fmind}, @samp{fmaxd}
20531 Double-precision minimum and maximum. These instructions are only
20532 generated if @option{-ffinite-math-only} is specified.
20534 @item @samp{fsqrtd}
20535 Unary square root operation.
20537 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
20538 Double-precision trigonometric and exponential functions. These instructions
20539 are only generated if @option{-funsafe-math-optimizations} is also specified.
20545 @item @samp{fextsd}
20546 Conversion from single precision to double precision.
20548 @item @samp{ftruncds}
20549 Conversion from double precision to single precision.
20551 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
20552 Conversion from floating point to signed or unsigned integer types, with
20553 truncation towards zero.
20556 Conversion from single-precision floating point to signed integer,
20557 rounding to the nearest integer and ties away from zero.
20558 This corresponds to the @code{__builtin_lroundf} function when
20559 @option{-fno-math-errno} is used.
20561 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
20562 Conversion from signed or unsigned integer types to floating-point types.
20566 In addition, all of the following transfer instructions for internal
20567 registers X and Y must be provided to use any of the double-precision
20568 floating-point instructions. Custom instructions taking two
20569 double-precision source operands expect the first operand in the
20570 64-bit register X. The other operand (or only operand of a unary
20571 operation) is given to the custom arithmetic instruction with the
20572 least significant half in source register @var{src1} and the most
20573 significant half in @var{src2}. A custom instruction that returns a
20574 double-precision result returns the most significant 32 bits in the
20575 destination register and the other half in 32-bit register Y.
20576 GCC automatically generates the necessary code sequences to write
20577 register X and/or read register Y when double-precision floating-point
20578 instructions are used.
20583 Write @var{src1} into the least significant half of X and @var{src2} into
20584 the most significant half of X.
20587 Write @var{src1} into Y.
20589 @item @samp{frdxhi}, @samp{frdxlo}
20590 Read the most or least (respectively) significant half of X and store it in
20594 Read the value of Y and store it into @var{dest}.
20597 Note that you can gain more local control over generation of Nios II custom
20598 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
20599 and @code{target("no-custom-@var{insn}")} function attributes
20600 (@pxref{Function Attributes})
20601 or pragmas (@pxref{Function Specific Option Pragmas}).
20603 @item -mcustom-fpu-cfg=@var{name}
20604 @opindex mcustom-fpu-cfg
20606 This option enables a predefined, named set of custom instruction encodings
20607 (see @option{-mcustom-@var{insn}} above).
20608 Currently, the following sets are defined:
20610 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
20611 @gccoptlist{-mcustom-fmuls=252 @gol
20612 -mcustom-fadds=253 @gol
20613 -mcustom-fsubs=254 @gol
20614 -fsingle-precision-constant}
20616 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
20617 @gccoptlist{-mcustom-fmuls=252 @gol
20618 -mcustom-fadds=253 @gol
20619 -mcustom-fsubs=254 @gol
20620 -mcustom-fdivs=255 @gol
20621 -fsingle-precision-constant}
20623 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
20624 @gccoptlist{-mcustom-floatus=243 @gol
20625 -mcustom-fixsi=244 @gol
20626 -mcustom-floatis=245 @gol
20627 -mcustom-fcmpgts=246 @gol
20628 -mcustom-fcmples=249 @gol
20629 -mcustom-fcmpeqs=250 @gol
20630 -mcustom-fcmpnes=251 @gol
20631 -mcustom-fmuls=252 @gol
20632 -mcustom-fadds=253 @gol
20633 -mcustom-fsubs=254 @gol
20634 -mcustom-fdivs=255 @gol
20635 -fsingle-precision-constant}
20637 Custom instruction assignments given by individual
20638 @option{-mcustom-@var{insn}=} options override those given by
20639 @option{-mcustom-fpu-cfg=}, regardless of the
20640 order of the options on the command line.
20642 Note that you can gain more local control over selection of a FPU
20643 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
20644 function attribute (@pxref{Function Attributes})
20645 or pragma (@pxref{Function Specific Option Pragmas}).
20649 These additional @samp{-m} options are available for the Altera Nios II
20650 ELF (bare-metal) target:
20656 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
20657 startup and termination code, and is typically used in conjunction with
20658 @option{-msys-crt0=} to specify the location of the alternate startup code
20659 provided by the HAL BSP.
20663 Link with a limited version of the C library, @option{-lsmallc}, rather than
20666 @item -msys-crt0=@var{startfile}
20668 @var{startfile} is the file name of the startfile (crt0) to use
20669 when linking. This option is only useful in conjunction with @option{-mhal}.
20671 @item -msys-lib=@var{systemlib}
20673 @var{systemlib} is the library name of the library that provides
20674 low-level system calls required by the C library,
20675 e.g. @code{read} and @code{write}.
20676 This option is typically used to link with a library provided by a HAL BSP.
20680 @node Nvidia PTX Options
20681 @subsection Nvidia PTX Options
20682 @cindex Nvidia PTX options
20683 @cindex nvptx options
20685 These options are defined for Nvidia PTX:
20693 Generate code for 32-bit or 64-bit ABI.
20696 @opindex mmainkernel
20697 Link in code for a __main kernel. This is for stand-alone instead of
20698 offloading execution.
20702 Apply partitioned execution optimizations. This is the default when any
20703 level of optimization is selected.
20706 @opindex msoft-stack
20707 Generate code that does not use @code{.local} memory
20708 directly for stack storage. Instead, a per-warp stack pointer is
20709 maintained explicitly. This enables variable-length stack allocation (with
20710 variable-length arrays or @code{alloca}), and when global memory is used for
20711 underlying storage, makes it possible to access automatic variables from other
20712 threads, or with atomic instructions. This code generation variant is used
20713 for OpenMP offloading, but the option is exposed on its own for the purpose
20714 of testing the compiler; to generate code suitable for linking into programs
20715 using OpenMP offloading, use option @option{-mgomp}.
20717 @item -muniform-simt
20718 @opindex muniform-simt
20719 Switch to code generation variant that allows to execute all threads in each
20720 warp, while maintaining memory state and side effects as if only one thread
20721 in each warp was active outside of OpenMP SIMD regions. All atomic operations
20722 and calls to runtime (malloc, free, vprintf) are conditionally executed (iff
20723 current lane index equals the master lane index), and the register being
20724 assigned is copied via a shuffle instruction from the master lane. Outside of
20725 SIMD regions lane 0 is the master; inside, each thread sees itself as the
20726 master. Shared memory array @code{int __nvptx_uni[]} stores all-zeros or
20727 all-ones bitmasks for each warp, indicating current mode (0 outside of SIMD
20728 regions). Each thread can bitwise-and the bitmask at position @code{tid.y}
20729 with current lane index to compute the master lane index.
20733 Generate code for use in OpenMP offloading: enables @option{-msoft-stack} and
20734 @option{-muniform-simt} options, and selects corresponding multilib variant.
20738 @node PDP-11 Options
20739 @subsection PDP-11 Options
20740 @cindex PDP-11 Options
20742 These options are defined for the PDP-11:
20747 Use hardware FPP floating point. This is the default. (FIS floating
20748 point on the PDP-11/40 is not supported.)
20751 @opindex msoft-float
20752 Do not use hardware floating point.
20756 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
20760 Return floating-point results in memory. This is the default.
20764 Generate code for a PDP-11/40.
20768 Generate code for a PDP-11/45. This is the default.
20772 Generate code for a PDP-11/10.
20774 @item -mbcopy-builtin
20775 @opindex mbcopy-builtin
20776 Use inline @code{movmemhi} patterns for copying memory. This is the
20781 Do not use inline @code{movmemhi} patterns for copying memory.
20787 Use 16-bit @code{int}. This is the default.
20793 Use 32-bit @code{int}.
20796 @itemx -mno-float32
20798 @opindex mno-float32
20799 Use 64-bit @code{float}. This is the default.
20802 @itemx -mno-float64
20804 @opindex mno-float64
20805 Use 32-bit @code{float}.
20809 Use @code{abshi2} pattern. This is the default.
20813 Do not use @code{abshi2} pattern.
20815 @item -mbranch-expensive
20816 @opindex mbranch-expensive
20817 Pretend that branches are expensive. This is for experimenting with
20818 code generation only.
20820 @item -mbranch-cheap
20821 @opindex mbranch-cheap
20822 Do not pretend that branches are expensive. This is the default.
20826 Use Unix assembler syntax. This is the default when configured for
20827 @samp{pdp11-*-bsd}.
20831 Use DEC assembler syntax. This is the default when configured for any
20832 PDP-11 target other than @samp{pdp11-*-bsd}.
20835 @node picoChip Options
20836 @subsection picoChip Options
20837 @cindex picoChip options
20839 These @samp{-m} options are defined for picoChip implementations:
20843 @item -mae=@var{ae_type}
20845 Set the instruction set, register set, and instruction scheduling
20846 parameters for array element type @var{ae_type}. Supported values
20847 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
20849 @option{-mae=ANY} selects a completely generic AE type. Code
20850 generated with this option runs on any of the other AE types. The
20851 code is not as efficient as it would be if compiled for a specific
20852 AE type, and some types of operation (e.g., multiplication) do not
20853 work properly on all types of AE.
20855 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
20856 for compiled code, and is the default.
20858 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
20859 option may suffer from poor performance of byte (char) manipulation,
20860 since the DSP AE does not provide hardware support for byte load/stores.
20862 @item -msymbol-as-address
20863 Enable the compiler to directly use a symbol name as an address in a
20864 load/store instruction, without first loading it into a
20865 register. Typically, the use of this option generates larger
20866 programs, which run faster than when the option isn't used. However, the
20867 results vary from program to program, so it is left as a user option,
20868 rather than being permanently enabled.
20870 @item -mno-inefficient-warnings
20871 Disables warnings about the generation of inefficient code. These
20872 warnings can be generated, for example, when compiling code that
20873 performs byte-level memory operations on the MAC AE type. The MAC AE has
20874 no hardware support for byte-level memory operations, so all byte
20875 load/stores must be synthesized from word load/store operations. This is
20876 inefficient and a warning is generated to indicate
20877 that you should rewrite the code to avoid byte operations, or to target
20878 an AE type that has the necessary hardware support. This option disables
20883 @node PowerPC Options
20884 @subsection PowerPC Options
20885 @cindex PowerPC options
20887 These are listed under @xref{RS/6000 and PowerPC Options}.
20890 @subsection RL78 Options
20891 @cindex RL78 Options
20897 Links in additional target libraries to support operation within a
20906 Specifies the type of hardware multiplication and division support to
20907 be used. The simplest is @code{none}, which uses software for both
20908 multiplication and division. This is the default. The @code{g13}
20909 value is for the hardware multiply/divide peripheral found on the
20910 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
20911 the multiplication and division instructions supported by the RL78/G14
20912 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
20913 the value @code{mg10} is an alias for @code{none}.
20915 In addition a C preprocessor macro is defined, based upon the setting
20916 of this option. Possible values are: @code{__RL78_MUL_NONE__},
20917 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
20924 Specifies the RL78 core to target. The default is the G14 core, also
20925 known as an S3 core or just RL78. The G13 or S2 core does not have
20926 multiply or divide instructions, instead it uses a hardware peripheral
20927 for these operations. The G10 or S1 core does not have register
20928 banks, so it uses a different calling convention.
20930 If this option is set it also selects the type of hardware multiply
20931 support to use, unless this is overridden by an explicit
20932 @option{-mmul=none} option on the command line. Thus specifying
20933 @option{-mcpu=g13} enables the use of the G13 hardware multiply
20934 peripheral and specifying @option{-mcpu=g10} disables the use of
20935 hardware multiplications altogether.
20937 Note, although the RL78/G14 core is the default target, specifying
20938 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
20939 change the behavior of the toolchain since it also enables G14
20940 hardware multiply support. If these options are not specified on the
20941 command line then software multiplication routines will be used even
20942 though the code targets the RL78 core. This is for backwards
20943 compatibility with older toolchains which did not have hardware
20944 multiply and divide support.
20946 In addition a C preprocessor macro is defined, based upon the setting
20947 of this option. Possible values are: @code{__RL78_G10__},
20948 @code{__RL78_G13__} or @code{__RL78_G14__}.
20958 These are aliases for the corresponding @option{-mcpu=} option. They
20959 are provided for backwards compatibility.
20963 Allow the compiler to use all of the available registers. By default
20964 registers @code{r24..r31} are reserved for use in interrupt handlers.
20965 With this option enabled these registers can be used in ordinary
20968 @item -m64bit-doubles
20969 @itemx -m32bit-doubles
20970 @opindex m64bit-doubles
20971 @opindex m32bit-doubles
20972 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
20973 or 32 bits (@option{-m32bit-doubles}) in size. The default is
20974 @option{-m32bit-doubles}.
20976 @item -msave-mduc-in-interrupts
20977 @item -mno-save-mduc-in-interrupts
20978 @opindex msave-mduc-in-interrupts
20979 @opindex mno-save-mduc-in-interrupts
20980 Specifies that interrupt handler functions should preserve the
20981 MDUC registers. This is only necessary if normal code might use
20982 the MDUC registers, for example because it performs multiplication
20983 and division operations. The default is to ignore the MDUC registers
20984 as this makes the interrupt handlers faster. The target option -mg13
20985 needs to be passed for this to work as this feature is only available
20986 on the G13 target (S2 core). The MDUC registers will only be saved
20987 if the interrupt handler performs a multiplication or division
20988 operation or it calls another function.
20992 @node RS/6000 and PowerPC Options
20993 @subsection IBM RS/6000 and PowerPC Options
20994 @cindex RS/6000 and PowerPC Options
20995 @cindex IBM RS/6000 and PowerPC Options
20997 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
20999 @item -mpowerpc-gpopt
21000 @itemx -mno-powerpc-gpopt
21001 @itemx -mpowerpc-gfxopt
21002 @itemx -mno-powerpc-gfxopt
21005 @itemx -mno-powerpc64
21009 @itemx -mno-popcntb
21011 @itemx -mno-popcntd
21020 @itemx -mno-hard-dfp
21021 @opindex mpowerpc-gpopt
21022 @opindex mno-powerpc-gpopt
21023 @opindex mpowerpc-gfxopt
21024 @opindex mno-powerpc-gfxopt
21025 @opindex mpowerpc64
21026 @opindex mno-powerpc64
21030 @opindex mno-popcntb
21032 @opindex mno-popcntd
21038 @opindex mno-mfpgpr
21040 @opindex mno-hard-dfp
21041 You use these options to specify which instructions are available on the
21042 processor you are using. The default value of these options is
21043 determined when configuring GCC@. Specifying the
21044 @option{-mcpu=@var{cpu_type}} overrides the specification of these
21045 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
21046 rather than the options listed above.
21048 Specifying @option{-mpowerpc-gpopt} allows
21049 GCC to use the optional PowerPC architecture instructions in the
21050 General Purpose group, including floating-point square root. Specifying
21051 @option{-mpowerpc-gfxopt} allows GCC to
21052 use the optional PowerPC architecture instructions in the Graphics
21053 group, including floating-point select.
21055 The @option{-mmfcrf} option allows GCC to generate the move from
21056 condition register field instruction implemented on the POWER4
21057 processor and other processors that support the PowerPC V2.01
21059 The @option{-mpopcntb} option allows GCC to generate the popcount and
21060 double-precision FP reciprocal estimate instruction implemented on the
21061 POWER5 processor and other processors that support the PowerPC V2.02
21063 The @option{-mpopcntd} option allows GCC to generate the popcount
21064 instruction implemented on the POWER7 processor and other processors
21065 that support the PowerPC V2.06 architecture.
21066 The @option{-mfprnd} option allows GCC to generate the FP round to
21067 integer instructions implemented on the POWER5+ processor and other
21068 processors that support the PowerPC V2.03 architecture.
21069 The @option{-mcmpb} option allows GCC to generate the compare bytes
21070 instruction implemented on the POWER6 processor and other processors
21071 that support the PowerPC V2.05 architecture.
21072 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
21073 general-purpose register instructions implemented on the POWER6X
21074 processor and other processors that support the extended PowerPC V2.05
21076 The @option{-mhard-dfp} option allows GCC to generate the decimal
21077 floating-point instructions implemented on some POWER processors.
21079 The @option{-mpowerpc64} option allows GCC to generate the additional
21080 64-bit instructions that are found in the full PowerPC64 architecture
21081 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
21082 @option{-mno-powerpc64}.
21084 @item -mcpu=@var{cpu_type}
21086 Set architecture type, register usage, and
21087 instruction scheduling parameters for machine type @var{cpu_type}.
21088 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
21089 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
21090 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
21091 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
21092 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
21093 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
21094 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
21095 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
21096 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
21097 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
21098 @samp{power9}, @samp{powerpc}, @samp{powerpc64}, @samp{powerpc64le},
21101 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
21102 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
21103 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
21104 architecture machine types, with an appropriate, generic processor
21105 model assumed for scheduling purposes.
21107 The other options specify a specific processor. Code generated under
21108 those options runs best on that processor, and may not run at all on
21111 The @option{-mcpu} options automatically enable or disable the
21114 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
21115 -mpopcntb -mpopcntd -mpowerpc64 @gol
21116 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
21117 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
21118 -mcrypto -mdirect-move -mhtm -mpower8-fusion -mpower8-vector @gol
21119 -mquad-memory -mquad-memory-atomic -mfloat128 -mfloat128-hardware}
21121 The particular options set for any particular CPU varies between
21122 compiler versions, depending on what setting seems to produce optimal
21123 code for that CPU; it doesn't necessarily reflect the actual hardware's
21124 capabilities. If you wish to set an individual option to a particular
21125 value, you may specify it after the @option{-mcpu} option, like
21126 @option{-mcpu=970 -mno-altivec}.
21128 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
21129 not enabled or disabled by the @option{-mcpu} option at present because
21130 AIX does not have full support for these options. You may still
21131 enable or disable them individually if you're sure it'll work in your
21134 @item -mtune=@var{cpu_type}
21136 Set the instruction scheduling parameters for machine type
21137 @var{cpu_type}, but do not set the architecture type or register usage,
21138 as @option{-mcpu=@var{cpu_type}} does. The same
21139 values for @var{cpu_type} are used for @option{-mtune} as for
21140 @option{-mcpu}. If both are specified, the code generated uses the
21141 architecture and registers set by @option{-mcpu}, but the
21142 scheduling parameters set by @option{-mtune}.
21144 @item -mcmodel=small
21145 @opindex mcmodel=small
21146 Generate PowerPC64 code for the small model: The TOC is limited to
21149 @item -mcmodel=medium
21150 @opindex mcmodel=medium
21151 Generate PowerPC64 code for the medium model: The TOC and other static
21152 data may be up to a total of 4G in size.
21154 @item -mcmodel=large
21155 @opindex mcmodel=large
21156 Generate PowerPC64 code for the large model: The TOC may be up to 4G
21157 in size. Other data and code is only limited by the 64-bit address
21161 @itemx -mno-altivec
21163 @opindex mno-altivec
21164 Generate code that uses (does not use) AltiVec instructions, and also
21165 enable the use of built-in functions that allow more direct access to
21166 the AltiVec instruction set. You may also need to set
21167 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
21170 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
21171 @option{-maltivec=be}, the element order for AltiVec intrinsics such
21172 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
21173 match array element order corresponding to the endianness of the
21174 target. That is, element zero identifies the leftmost element in a
21175 vector register when targeting a big-endian platform, and identifies
21176 the rightmost element in a vector register when targeting a
21177 little-endian platform.
21180 @opindex maltivec=be
21181 Generate AltiVec instructions using big-endian element order,
21182 regardless of whether the target is big- or little-endian. This is
21183 the default when targeting a big-endian platform.
21185 The element order is used to interpret element numbers in AltiVec
21186 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
21187 @code{vec_insert}. By default, these match array element order
21188 corresponding to the endianness for the target.
21191 @opindex maltivec=le
21192 Generate AltiVec instructions using little-endian element order,
21193 regardless of whether the target is big- or little-endian. This is
21194 the default when targeting a little-endian platform. This option is
21195 currently ignored when targeting a big-endian platform.
21197 The element order is used to interpret element numbers in AltiVec
21198 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
21199 @code{vec_insert}. By default, these match array element order
21200 corresponding to the endianness for the target.
21205 @opindex mno-vrsave
21206 Generate VRSAVE instructions when generating AltiVec code.
21208 @item -mgen-cell-microcode
21209 @opindex mgen-cell-microcode
21210 Generate Cell microcode instructions.
21212 @item -mwarn-cell-microcode
21213 @opindex mwarn-cell-microcode
21214 Warn when a Cell microcode instruction is emitted. An example
21215 of a Cell microcode instruction is a variable shift.
21218 @opindex msecure-plt
21219 Generate code that allows @command{ld} and @command{ld.so}
21220 to build executables and shared
21221 libraries with non-executable @code{.plt} and @code{.got} sections.
21223 32-bit SYSV ABI option.
21227 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
21229 requires @code{.plt} and @code{.got}
21230 sections that are both writable and executable.
21231 This is a PowerPC 32-bit SYSV ABI option.
21237 This switch enables or disables the generation of ISEL instructions.
21239 @item -misel=@var{yes/no}
21240 This switch has been deprecated. Use @option{-misel} and
21241 @option{-mno-isel} instead.
21245 Enable Local Register Allocation. This is still experimental for PowerPC,
21246 so by default the compiler uses standard reload
21247 (i.e. @option{-mno-lra}).
21253 This switch enables or disables the generation of SPE simd
21259 @opindex mno-paired
21260 This switch enables or disables the generation of PAIRED simd
21263 @item -mspe=@var{yes/no}
21264 This option has been deprecated. Use @option{-mspe} and
21265 @option{-mno-spe} instead.
21271 Generate code that uses (does not use) vector/scalar (VSX)
21272 instructions, and also enable the use of built-in functions that allow
21273 more direct access to the VSX instruction set.
21278 @opindex mno-crypto
21279 Enable the use (disable) of the built-in functions that allow direct
21280 access to the cryptographic instructions that were added in version
21281 2.07 of the PowerPC ISA.
21283 @item -mdirect-move
21284 @itemx -mno-direct-move
21285 @opindex mdirect-move
21286 @opindex mno-direct-move
21287 Generate code that uses (does not use) the instructions to move data
21288 between the general purpose registers and the vector/scalar (VSX)
21289 registers that were added in version 2.07 of the PowerPC ISA.
21295 Enable (disable) the use of the built-in functions that allow direct
21296 access to the Hardware Transactional Memory (HTM) instructions that
21297 were added in version 2.07 of the PowerPC ISA.
21299 @item -mpower8-fusion
21300 @itemx -mno-power8-fusion
21301 @opindex mpower8-fusion
21302 @opindex mno-power8-fusion
21303 Generate code that keeps (does not keeps) some integer operations
21304 adjacent so that the instructions can be fused together on power8 and
21307 @item -mpower8-vector
21308 @itemx -mno-power8-vector
21309 @opindex mpower8-vector
21310 @opindex mno-power8-vector
21311 Generate code that uses (does not use) the vector and scalar
21312 instructions that were added in version 2.07 of the PowerPC ISA. Also
21313 enable the use of built-in functions that allow more direct access to
21314 the vector instructions.
21316 @item -mquad-memory
21317 @itemx -mno-quad-memory
21318 @opindex mquad-memory
21319 @opindex mno-quad-memory
21320 Generate code that uses (does not use) the non-atomic quad word memory
21321 instructions. The @option{-mquad-memory} option requires use of
21324 @item -mquad-memory-atomic
21325 @itemx -mno-quad-memory-atomic
21326 @opindex mquad-memory-atomic
21327 @opindex mno-quad-memory-atomic
21328 Generate code that uses (does not use) the atomic quad word memory
21329 instructions. The @option{-mquad-memory-atomic} option requires use of
21332 @item -mupper-regs-di
21333 @itemx -mno-upper-regs-di
21334 @opindex mupper-regs-di
21335 @opindex mno-upper-regs-di
21336 Generate code that uses (does not use) the scalar instructions that
21337 target all 64 registers in the vector/scalar floating point register
21338 set that were added in version 2.06 of the PowerPC ISA when processing
21339 integers. @option{-mupper-regs-di} is turned on by default if you use
21340 any of the @option{-mcpu=power7}, @option{-mcpu=power8},
21341 @option{-mcpu=power9}, or @option{-mvsx} options.
21343 @item -mupper-regs-df
21344 @itemx -mno-upper-regs-df
21345 @opindex mupper-regs-df
21346 @opindex mno-upper-regs-df
21347 Generate code that uses (does not use) the scalar double precision
21348 instructions that target all 64 registers in the vector/scalar
21349 floating point register set that were added in version 2.06 of the
21350 PowerPC ISA. @option{-mupper-regs-df} is turned on by default if you
21351 use any of the @option{-mcpu=power7}, @option{-mcpu=power8},
21352 @option{-mcpu=power9}, or @option{-mvsx} options.
21354 @item -mupper-regs-sf
21355 @itemx -mno-upper-regs-sf
21356 @opindex mupper-regs-sf
21357 @opindex mno-upper-regs-sf
21358 Generate code that uses (does not use) the scalar single precision
21359 instructions that target all 64 registers in the vector/scalar
21360 floating point register set that were added in version 2.07 of the
21361 PowerPC ISA. @option{-mupper-regs-sf} is turned on by default if you
21362 use either of the @option{-mcpu=power8}, @option{-mpower8-vector}, or
21363 @option{-mcpu=power9} options.
21366 @itemx -mno-upper-regs
21367 @opindex mupper-regs
21368 @opindex mno-upper-regs
21369 Generate code that uses (does not use) the scalar
21370 instructions that target all 64 registers in the vector/scalar
21371 floating point register set, depending on the model of the machine.
21373 If the @option{-mno-upper-regs} option is used, it turns off both
21374 @option{-mupper-regs-sf} and @option{-mupper-regs-df} options.
21377 @itemx -mno-float128
21379 @opindex mno-float128
21380 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
21381 and use either software emulation for IEEE 128-bit floating point or
21382 hardware instructions.
21384 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7}, or
21385 @option{-mcpu=power8}) must be enabled to use the @option{-mfloat128}
21386 option. The @option{-mfloat128} option only works on PowerPC 64-bit
21389 If you use the ISA 3.0 instruction set (@option{-mcpu=power9}), the
21390 @option{-mfloat128} option will also enable the generation of ISA 3.0
21391 IEEE 128-bit floating point instructions. Otherwise, IEEE 128-bit
21392 floating point will be done with software emulation.
21394 @item -mfloat128-hardware
21395 @itemx -mno-float128-hardware
21396 @opindex mfloat128-hardware
21397 @opindex mno-float128-hardware
21398 Enable/disable using ISA 3.0 hardware instructions to support the
21399 @var{__float128} data type.
21401 If you use @option{-mfloat128-hardware}, it will enable the option
21402 @option{-mfloat128} as well.
21404 If you select ISA 3.0 instructions with @option{-mcpu=power9}, but do
21405 not use either @option{-mfloat128} or @option{-mfloat128-hardware},
21406 the IEEE 128-bit floating point support will not be enabled.
21408 @item -mfloat-gprs=@var{yes/single/double/no}
21409 @itemx -mfloat-gprs
21410 @opindex mfloat-gprs
21411 This switch enables or disables the generation of floating-point
21412 operations on the general-purpose registers for architectures that
21415 The argument @samp{yes} or @samp{single} enables the use of
21416 single-precision floating-point operations.
21418 The argument @samp{double} enables the use of single and
21419 double-precision floating-point operations.
21421 The argument @samp{no} disables floating-point operations on the
21422 general-purpose registers.
21424 This option is currently only available on the MPC854x.
21430 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
21431 targets (including GNU/Linux). The 32-bit environment sets int, long
21432 and pointer to 32 bits and generates code that runs on any PowerPC
21433 variant. The 64-bit environment sets int to 32 bits and long and
21434 pointer to 64 bits, and generates code for PowerPC64, as for
21435 @option{-mpowerpc64}.
21438 @itemx -mno-fp-in-toc
21439 @itemx -mno-sum-in-toc
21440 @itemx -mminimal-toc
21442 @opindex mno-fp-in-toc
21443 @opindex mno-sum-in-toc
21444 @opindex mminimal-toc
21445 Modify generation of the TOC (Table Of Contents), which is created for
21446 every executable file. The @option{-mfull-toc} option is selected by
21447 default. In that case, GCC allocates at least one TOC entry for
21448 each unique non-automatic variable reference in your program. GCC
21449 also places floating-point constants in the TOC@. However, only
21450 16,384 entries are available in the TOC@.
21452 If you receive a linker error message that saying you have overflowed
21453 the available TOC space, you can reduce the amount of TOC space used
21454 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
21455 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
21456 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
21457 generate code to calculate the sum of an address and a constant at
21458 run time instead of putting that sum into the TOC@. You may specify one
21459 or both of these options. Each causes GCC to produce very slightly
21460 slower and larger code at the expense of conserving TOC space.
21462 If you still run out of space in the TOC even when you specify both of
21463 these options, specify @option{-mminimal-toc} instead. This option causes
21464 GCC to make only one TOC entry for every file. When you specify this
21465 option, GCC produces code that is slower and larger but which
21466 uses extremely little TOC space. You may wish to use this option
21467 only on files that contain less frequently-executed code.
21473 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
21474 @code{long} type, and the infrastructure needed to support them.
21475 Specifying @option{-maix64} implies @option{-mpowerpc64},
21476 while @option{-maix32} disables the 64-bit ABI and
21477 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
21480 @itemx -mno-xl-compat
21481 @opindex mxl-compat
21482 @opindex mno-xl-compat
21483 Produce code that conforms more closely to IBM XL compiler semantics
21484 when using AIX-compatible ABI@. Pass floating-point arguments to
21485 prototyped functions beyond the register save area (RSA) on the stack
21486 in addition to argument FPRs. Do not assume that most significant
21487 double in 128-bit long double value is properly rounded when comparing
21488 values and converting to double. Use XL symbol names for long double
21491 The AIX calling convention was extended but not initially documented to
21492 handle an obscure K&R C case of calling a function that takes the
21493 address of its arguments with fewer arguments than declared. IBM XL
21494 compilers access floating-point arguments that do not fit in the
21495 RSA from the stack when a subroutine is compiled without
21496 optimization. Because always storing floating-point arguments on the
21497 stack is inefficient and rarely needed, this option is not enabled by
21498 default and only is necessary when calling subroutines compiled by IBM
21499 XL compilers without optimization.
21503 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
21504 application written to use message passing with special startup code to
21505 enable the application to run. The system must have PE installed in the
21506 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
21507 must be overridden with the @option{-specs=} option to specify the
21508 appropriate directory location. The Parallel Environment does not
21509 support threads, so the @option{-mpe} option and the @option{-pthread}
21510 option are incompatible.
21512 @item -malign-natural
21513 @itemx -malign-power
21514 @opindex malign-natural
21515 @opindex malign-power
21516 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
21517 @option{-malign-natural} overrides the ABI-defined alignment of larger
21518 types, such as floating-point doubles, on their natural size-based boundary.
21519 The option @option{-malign-power} instructs GCC to follow the ABI-specified
21520 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
21522 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
21526 @itemx -mhard-float
21527 @opindex msoft-float
21528 @opindex mhard-float
21529 Generate code that does not use (uses) the floating-point register set.
21530 Software floating-point emulation is provided if you use the
21531 @option{-msoft-float} option, and pass the option to GCC when linking.
21533 @item -msingle-float
21534 @itemx -mdouble-float
21535 @opindex msingle-float
21536 @opindex mdouble-float
21537 Generate code for single- or double-precision floating-point operations.
21538 @option{-mdouble-float} implies @option{-msingle-float}.
21541 @opindex msimple-fpu
21542 Do not generate @code{sqrt} and @code{div} instructions for hardware
21543 floating-point unit.
21545 @item -mfpu=@var{name}
21547 Specify type of floating-point unit. Valid values for @var{name} are
21548 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
21549 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
21550 @samp{sp_full} (equivalent to @option{-msingle-float}),
21551 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
21554 @opindex mxilinx-fpu
21555 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
21558 @itemx -mno-multiple
21560 @opindex mno-multiple
21561 Generate code that uses (does not use) the load multiple word
21562 instructions and the store multiple word instructions. These
21563 instructions are generated by default on POWER systems, and not
21564 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
21565 PowerPC systems, since those instructions do not work when the
21566 processor is in little-endian mode. The exceptions are PPC740 and
21567 PPC750 which permit these instructions in little-endian mode.
21572 @opindex mno-string
21573 Generate code that uses (does not use) the load string instructions
21574 and the store string word instructions to save multiple registers and
21575 do small block moves. These instructions are generated by default on
21576 POWER systems, and not generated on PowerPC systems. Do not use
21577 @option{-mstring} on little-endian PowerPC systems, since those
21578 instructions do not work when the processor is in little-endian mode.
21579 The exceptions are PPC740 and PPC750 which permit these instructions
21580 in little-endian mode.
21585 @opindex mno-update
21586 Generate code that uses (does not use) the load or store instructions
21587 that update the base register to the address of the calculated memory
21588 location. These instructions are generated by default. If you use
21589 @option{-mno-update}, there is a small window between the time that the
21590 stack pointer is updated and the address of the previous frame is
21591 stored, which means code that walks the stack frame across interrupts or
21592 signals may get corrupted data.
21594 @item -mavoid-indexed-addresses
21595 @itemx -mno-avoid-indexed-addresses
21596 @opindex mavoid-indexed-addresses
21597 @opindex mno-avoid-indexed-addresses
21598 Generate code that tries to avoid (not avoid) the use of indexed load
21599 or store instructions. These instructions can incur a performance
21600 penalty on Power6 processors in certain situations, such as when
21601 stepping through large arrays that cross a 16M boundary. This option
21602 is enabled by default when targeting Power6 and disabled otherwise.
21605 @itemx -mno-fused-madd
21606 @opindex mfused-madd
21607 @opindex mno-fused-madd
21608 Generate code that uses (does not use) the floating-point multiply and
21609 accumulate instructions. These instructions are generated by default
21610 if hardware floating point is used. The machine-dependent
21611 @option{-mfused-madd} option is now mapped to the machine-independent
21612 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
21613 mapped to @option{-ffp-contract=off}.
21619 Generate code that uses (does not use) the half-word multiply and
21620 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
21621 These instructions are generated by default when targeting those
21628 Generate code that uses (does not use) the string-search @samp{dlmzb}
21629 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
21630 generated by default when targeting those processors.
21632 @item -mno-bit-align
21634 @opindex mno-bit-align
21635 @opindex mbit-align
21636 On System V.4 and embedded PowerPC systems do not (do) force structures
21637 and unions that contain bit-fields to be aligned to the base type of the
21640 For example, by default a structure containing nothing but 8
21641 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
21642 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
21643 the structure is aligned to a 1-byte boundary and is 1 byte in
21646 @item -mno-strict-align
21647 @itemx -mstrict-align
21648 @opindex mno-strict-align
21649 @opindex mstrict-align
21650 On System V.4 and embedded PowerPC systems do not (do) assume that
21651 unaligned memory references are handled by the system.
21653 @item -mrelocatable
21654 @itemx -mno-relocatable
21655 @opindex mrelocatable
21656 @opindex mno-relocatable
21657 Generate code that allows (does not allow) a static executable to be
21658 relocated to a different address at run time. A simple embedded
21659 PowerPC system loader should relocate the entire contents of
21660 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
21661 a table of 32-bit addresses generated by this option. For this to
21662 work, all objects linked together must be compiled with
21663 @option{-mrelocatable} or @option{-mrelocatable-lib}.
21664 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
21666 @item -mrelocatable-lib
21667 @itemx -mno-relocatable-lib
21668 @opindex mrelocatable-lib
21669 @opindex mno-relocatable-lib
21670 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
21671 @code{.fixup} section to allow static executables to be relocated at
21672 run time, but @option{-mrelocatable-lib} does not use the smaller stack
21673 alignment of @option{-mrelocatable}. Objects compiled with
21674 @option{-mrelocatable-lib} may be linked with objects compiled with
21675 any combination of the @option{-mrelocatable} options.
21681 On System V.4 and embedded PowerPC systems do not (do) assume that
21682 register 2 contains a pointer to a global area pointing to the addresses
21683 used in the program.
21686 @itemx -mlittle-endian
21688 @opindex mlittle-endian
21689 On System V.4 and embedded PowerPC systems compile code for the
21690 processor in little-endian mode. The @option{-mlittle-endian} option is
21691 the same as @option{-mlittle}.
21694 @itemx -mbig-endian
21696 @opindex mbig-endian
21697 On System V.4 and embedded PowerPC systems compile code for the
21698 processor in big-endian mode. The @option{-mbig-endian} option is
21699 the same as @option{-mbig}.
21701 @item -mdynamic-no-pic
21702 @opindex mdynamic-no-pic
21703 On Darwin and Mac OS X systems, compile code so that it is not
21704 relocatable, but that its external references are relocatable. The
21705 resulting code is suitable for applications, but not shared
21708 @item -msingle-pic-base
21709 @opindex msingle-pic-base
21710 Treat the register used for PIC addressing as read-only, rather than
21711 loading it in the prologue for each function. The runtime system is
21712 responsible for initializing this register with an appropriate value
21713 before execution begins.
21715 @item -mprioritize-restricted-insns=@var{priority}
21716 @opindex mprioritize-restricted-insns
21717 This option controls the priority that is assigned to
21718 dispatch-slot restricted instructions during the second scheduling
21719 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
21720 or @samp{2} to assign no, highest, or second-highest (respectively)
21721 priority to dispatch-slot restricted
21724 @item -msched-costly-dep=@var{dependence_type}
21725 @opindex msched-costly-dep
21726 This option controls which dependences are considered costly
21727 by the target during instruction scheduling. The argument
21728 @var{dependence_type} takes one of the following values:
21732 No dependence is costly.
21735 All dependences are costly.
21737 @item @samp{true_store_to_load}
21738 A true dependence from store to load is costly.
21740 @item @samp{store_to_load}
21741 Any dependence from store to load is costly.
21744 Any dependence for which the latency is greater than or equal to
21745 @var{number} is costly.
21748 @item -minsert-sched-nops=@var{scheme}
21749 @opindex minsert-sched-nops
21750 This option controls which NOP insertion scheme is used during
21751 the second scheduling pass. The argument @var{scheme} takes one of the
21759 Pad with NOPs any dispatch group that has vacant issue slots,
21760 according to the scheduler's grouping.
21762 @item @samp{regroup_exact}
21763 Insert NOPs to force costly dependent insns into
21764 separate groups. Insert exactly as many NOPs as needed to force an insn
21765 to a new group, according to the estimated processor grouping.
21768 Insert NOPs to force costly dependent insns into
21769 separate groups. Insert @var{number} NOPs to force an insn to a new group.
21773 @opindex mcall-sysv
21774 On System V.4 and embedded PowerPC systems compile code using calling
21775 conventions that adhere to the March 1995 draft of the System V
21776 Application Binary Interface, PowerPC processor supplement. This is the
21777 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
21779 @item -mcall-sysv-eabi
21781 @opindex mcall-sysv-eabi
21782 @opindex mcall-eabi
21783 Specify both @option{-mcall-sysv} and @option{-meabi} options.
21785 @item -mcall-sysv-noeabi
21786 @opindex mcall-sysv-noeabi
21787 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
21789 @item -mcall-aixdesc
21791 On System V.4 and embedded PowerPC systems compile code for the AIX
21795 @opindex mcall-linux
21796 On System V.4 and embedded PowerPC systems compile code for the
21797 Linux-based GNU system.
21799 @item -mcall-freebsd
21800 @opindex mcall-freebsd
21801 On System V.4 and embedded PowerPC systems compile code for the
21802 FreeBSD operating system.
21804 @item -mcall-netbsd
21805 @opindex mcall-netbsd
21806 On System V.4 and embedded PowerPC systems compile code for the
21807 NetBSD operating system.
21809 @item -mcall-openbsd
21810 @opindex mcall-netbsd
21811 On System V.4 and embedded PowerPC systems compile code for the
21812 OpenBSD operating system.
21814 @item -maix-struct-return
21815 @opindex maix-struct-return
21816 Return all structures in memory (as specified by the AIX ABI)@.
21818 @item -msvr4-struct-return
21819 @opindex msvr4-struct-return
21820 Return structures smaller than 8 bytes in registers (as specified by the
21823 @item -mabi=@var{abi-type}
21825 Extend the current ABI with a particular extension, or remove such extension.
21826 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
21827 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
21828 @samp{elfv1}, @samp{elfv2}@.
21832 Extend the current ABI with SPE ABI extensions. This does not change
21833 the default ABI, instead it adds the SPE ABI extensions to the current
21837 @opindex mabi=no-spe
21838 Disable Book-E SPE ABI extensions for the current ABI@.
21840 @item -mabi=ibmlongdouble
21841 @opindex mabi=ibmlongdouble
21842 Change the current ABI to use IBM extended-precision long double.
21843 This is a PowerPC 32-bit SYSV ABI option.
21845 @item -mabi=ieeelongdouble
21846 @opindex mabi=ieeelongdouble
21847 Change the current ABI to use IEEE extended-precision long double.
21848 This is a PowerPC 32-bit Linux ABI option.
21851 @opindex mabi=elfv1
21852 Change the current ABI to use the ELFv1 ABI.
21853 This is the default ABI for big-endian PowerPC 64-bit Linux.
21854 Overriding the default ABI requires special system support and is
21855 likely to fail in spectacular ways.
21858 @opindex mabi=elfv2
21859 Change the current ABI to use the ELFv2 ABI.
21860 This is the default ABI for little-endian PowerPC 64-bit Linux.
21861 Overriding the default ABI requires special system support and is
21862 likely to fail in spectacular ways.
21864 @item -mgnu-attribute
21865 @itemx -mno-gnu-attribute
21866 @opindex mgnu-attribute
21867 @opindex mno-gnu-attribute
21868 Emit .gnu_attribute assembly directives to set tag/value pairs in a
21869 .gnu.attributes section that specify ABI variations in function
21870 parameters or return values.
21873 @itemx -mno-prototype
21874 @opindex mprototype
21875 @opindex mno-prototype
21876 On System V.4 and embedded PowerPC systems assume that all calls to
21877 variable argument functions are properly prototyped. Otherwise, the
21878 compiler must insert an instruction before every non-prototyped call to
21879 set or clear bit 6 of the condition code register (@code{CR}) to
21880 indicate whether floating-point values are passed in the floating-point
21881 registers in case the function takes variable arguments. With
21882 @option{-mprototype}, only calls to prototyped variable argument functions
21883 set or clear the bit.
21887 On embedded PowerPC systems, assume that the startup module is called
21888 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
21889 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
21894 On embedded PowerPC systems, assume that the startup module is called
21895 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
21900 On embedded PowerPC systems, assume that the startup module is called
21901 @file{crt0.o} and the standard C libraries are @file{libads.a} and
21904 @item -myellowknife
21905 @opindex myellowknife
21906 On embedded PowerPC systems, assume that the startup module is called
21907 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
21912 On System V.4 and embedded PowerPC systems, specify that you are
21913 compiling for a VxWorks system.
21917 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
21918 header to indicate that @samp{eabi} extended relocations are used.
21924 On System V.4 and embedded PowerPC systems do (do not) adhere to the
21925 Embedded Applications Binary Interface (EABI), which is a set of
21926 modifications to the System V.4 specifications. Selecting @option{-meabi}
21927 means that the stack is aligned to an 8-byte boundary, a function
21928 @code{__eabi} is called from @code{main} to set up the EABI
21929 environment, and the @option{-msdata} option can use both @code{r2} and
21930 @code{r13} to point to two separate small data areas. Selecting
21931 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
21932 no EABI initialization function is called from @code{main}, and the
21933 @option{-msdata} option only uses @code{r13} to point to a single
21934 small data area. The @option{-meabi} option is on by default if you
21935 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
21938 @opindex msdata=eabi
21939 On System V.4 and embedded PowerPC systems, put small initialized
21940 @code{const} global and static data in the @code{.sdata2} section, which
21941 is pointed to by register @code{r2}. Put small initialized
21942 non-@code{const} global and static data in the @code{.sdata} section,
21943 which is pointed to by register @code{r13}. Put small uninitialized
21944 global and static data in the @code{.sbss} section, which is adjacent to
21945 the @code{.sdata} section. The @option{-msdata=eabi} option is
21946 incompatible with the @option{-mrelocatable} option. The
21947 @option{-msdata=eabi} option also sets the @option{-memb} option.
21950 @opindex msdata=sysv
21951 On System V.4 and embedded PowerPC systems, put small global and static
21952 data in the @code{.sdata} section, which is pointed to by register
21953 @code{r13}. Put small uninitialized global and static data in the
21954 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
21955 The @option{-msdata=sysv} option is incompatible with the
21956 @option{-mrelocatable} option.
21958 @item -msdata=default
21960 @opindex msdata=default
21962 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
21963 compile code the same as @option{-msdata=eabi}, otherwise compile code the
21964 same as @option{-msdata=sysv}.
21967 @opindex msdata=data
21968 On System V.4 and embedded PowerPC systems, put small global
21969 data in the @code{.sdata} section. Put small uninitialized global
21970 data in the @code{.sbss} section. Do not use register @code{r13}
21971 to address small data however. This is the default behavior unless
21972 other @option{-msdata} options are used.
21976 @opindex msdata=none
21978 On embedded PowerPC systems, put all initialized global and static data
21979 in the @code{.data} section, and all uninitialized data in the
21980 @code{.bss} section.
21982 @item -mblock-move-inline-limit=@var{num}
21983 @opindex mblock-move-inline-limit
21984 Inline all block moves (such as calls to @code{memcpy} or structure
21985 copies) less than or equal to @var{num} bytes. The minimum value for
21986 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
21987 targets. The default value is target-specific.
21991 @cindex smaller data references (PowerPC)
21992 @cindex .sdata/.sdata2 references (PowerPC)
21993 On embedded PowerPC systems, put global and static items less than or
21994 equal to @var{num} bytes into the small data or BSS sections instead of
21995 the normal data or BSS section. By default, @var{num} is 8. The
21996 @option{-G @var{num}} switch is also passed to the linker.
21997 All modules should be compiled with the same @option{-G @var{num}} value.
22000 @itemx -mno-regnames
22002 @opindex mno-regnames
22003 On System V.4 and embedded PowerPC systems do (do not) emit register
22004 names in the assembly language output using symbolic forms.
22007 @itemx -mno-longcall
22009 @opindex mno-longcall
22010 By default assume that all calls are far away so that a longer and more
22011 expensive calling sequence is required. This is required for calls
22012 farther than 32 megabytes (33,554,432 bytes) from the current location.
22013 A short call is generated if the compiler knows
22014 the call cannot be that far away. This setting can be overridden by
22015 the @code{shortcall} function attribute, or by @code{#pragma
22018 Some linkers are capable of detecting out-of-range calls and generating
22019 glue code on the fly. On these systems, long calls are unnecessary and
22020 generate slower code. As of this writing, the AIX linker can do this,
22021 as can the GNU linker for PowerPC/64. It is planned to add this feature
22022 to the GNU linker for 32-bit PowerPC systems as well.
22024 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
22025 callee, L42}, plus a @dfn{branch island} (glue code). The two target
22026 addresses represent the callee and the branch island. The
22027 Darwin/PPC linker prefers the first address and generates a @code{bl
22028 callee} if the PPC @code{bl} instruction reaches the callee directly;
22029 otherwise, the linker generates @code{bl L42} to call the branch
22030 island. The branch island is appended to the body of the
22031 calling function; it computes the full 32-bit address of the callee
22034 On Mach-O (Darwin) systems, this option directs the compiler emit to
22035 the glue for every direct call, and the Darwin linker decides whether
22036 to use or discard it.
22038 In the future, GCC may ignore all longcall specifications
22039 when the linker is known to generate glue.
22041 @item -mtls-markers
22042 @itemx -mno-tls-markers
22043 @opindex mtls-markers
22044 @opindex mno-tls-markers
22045 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
22046 specifying the function argument. The relocation allows the linker to
22047 reliably associate function call with argument setup instructions for
22048 TLS optimization, which in turn allows GCC to better schedule the
22053 Adds support for multithreading with the @dfn{pthreads} library.
22054 This option sets flags for both the preprocessor and linker.
22059 This option enables use of the reciprocal estimate and
22060 reciprocal square root estimate instructions with additional
22061 Newton-Raphson steps to increase precision instead of doing a divide or
22062 square root and divide for floating-point arguments. You should use
22063 the @option{-ffast-math} option when using @option{-mrecip} (or at
22064 least @option{-funsafe-math-optimizations},
22065 @option{-ffinite-math-only}, @option{-freciprocal-math} and
22066 @option{-fno-trapping-math}). Note that while the throughput of the
22067 sequence is generally higher than the throughput of the non-reciprocal
22068 instruction, the precision of the sequence can be decreased by up to 2
22069 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
22072 @item -mrecip=@var{opt}
22073 @opindex mrecip=opt
22074 This option controls which reciprocal estimate instructions
22075 may be used. @var{opt} is a comma-separated list of options, which may
22076 be preceded by a @code{!} to invert the option:
22081 Enable all estimate instructions.
22084 Enable the default instructions, equivalent to @option{-mrecip}.
22087 Disable all estimate instructions, equivalent to @option{-mno-recip}.
22090 Enable the reciprocal approximation instructions for both
22091 single and double precision.
22094 Enable the single-precision reciprocal approximation instructions.
22097 Enable the double-precision reciprocal approximation instructions.
22100 Enable the reciprocal square root approximation instructions for both
22101 single and double precision.
22104 Enable the single-precision reciprocal square root approximation instructions.
22107 Enable the double-precision reciprocal square root approximation instructions.
22111 So, for example, @option{-mrecip=all,!rsqrtd} enables
22112 all of the reciprocal estimate instructions, except for the
22113 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
22114 which handle the double-precision reciprocal square root calculations.
22116 @item -mrecip-precision
22117 @itemx -mno-recip-precision
22118 @opindex mrecip-precision
22119 Assume (do not assume) that the reciprocal estimate instructions
22120 provide higher-precision estimates than is mandated by the PowerPC
22121 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
22122 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
22123 The double-precision square root estimate instructions are not generated by
22124 default on low-precision machines, since they do not provide an
22125 estimate that converges after three steps.
22127 @item -mveclibabi=@var{type}
22128 @opindex mveclibabi
22129 Specifies the ABI type to use for vectorizing intrinsics using an
22130 external library. The only type supported at present is @samp{mass},
22131 which specifies to use IBM's Mathematical Acceleration Subsystem
22132 (MASS) libraries for vectorizing intrinsics using external libraries.
22133 GCC currently emits calls to @code{acosd2}, @code{acosf4},
22134 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
22135 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
22136 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
22137 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
22138 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
22139 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
22140 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
22141 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
22142 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
22143 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
22144 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
22145 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
22146 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
22147 for power7. Both @option{-ftree-vectorize} and
22148 @option{-funsafe-math-optimizations} must also be enabled. The MASS
22149 libraries must be specified at link time.
22154 Generate (do not generate) the @code{friz} instruction when the
22155 @option{-funsafe-math-optimizations} option is used to optimize
22156 rounding of floating-point values to 64-bit integer and back to floating
22157 point. The @code{friz} instruction does not return the same value if
22158 the floating-point number is too large to fit in an integer.
22160 @item -mpointers-to-nested-functions
22161 @itemx -mno-pointers-to-nested-functions
22162 @opindex mpointers-to-nested-functions
22163 Generate (do not generate) code to load up the static chain register
22164 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
22165 systems where a function pointer points to a 3-word descriptor giving
22166 the function address, TOC value to be loaded in register @code{r2}, and
22167 static chain value to be loaded in register @code{r11}. The
22168 @option{-mpointers-to-nested-functions} is on by default. You cannot
22169 call through pointers to nested functions or pointers
22170 to functions compiled in other languages that use the static chain if
22171 you use @option{-mno-pointers-to-nested-functions}.
22173 @item -msave-toc-indirect
22174 @itemx -mno-save-toc-indirect
22175 @opindex msave-toc-indirect
22176 Generate (do not generate) code to save the TOC value in the reserved
22177 stack location in the function prologue if the function calls through
22178 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
22179 saved in the prologue, it is saved just before the call through the
22180 pointer. The @option{-mno-save-toc-indirect} option is the default.
22182 @item -mcompat-align-parm
22183 @itemx -mno-compat-align-parm
22184 @opindex mcompat-align-parm
22185 Generate (do not generate) code to pass structure parameters with a
22186 maximum alignment of 64 bits, for compatibility with older versions
22189 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
22190 structure parameter on a 128-bit boundary when that structure contained
22191 a member requiring 128-bit alignment. This is corrected in more
22192 recent versions of GCC. This option may be used to generate code
22193 that is compatible with functions compiled with older versions of
22196 The @option{-mno-compat-align-parm} option is the default.
22200 @subsection RX Options
22203 These command-line options are defined for RX targets:
22206 @item -m64bit-doubles
22207 @itemx -m32bit-doubles
22208 @opindex m64bit-doubles
22209 @opindex m32bit-doubles
22210 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
22211 or 32 bits (@option{-m32bit-doubles}) in size. The default is
22212 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
22213 works on 32-bit values, which is why the default is
22214 @option{-m32bit-doubles}.
22220 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
22221 floating-point hardware. The default is enabled for the RX600
22222 series and disabled for the RX200 series.
22224 Floating-point instructions are only generated for 32-bit floating-point
22225 values, however, so the FPU hardware is not used for doubles if the
22226 @option{-m64bit-doubles} option is used.
22228 @emph{Note} If the @option{-fpu} option is enabled then
22229 @option{-funsafe-math-optimizations} is also enabled automatically.
22230 This is because the RX FPU instructions are themselves unsafe.
22232 @item -mcpu=@var{name}
22234 Selects the type of RX CPU to be targeted. Currently three types are
22235 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
22236 the specific @samp{RX610} CPU. The default is @samp{RX600}.
22238 The only difference between @samp{RX600} and @samp{RX610} is that the
22239 @samp{RX610} does not support the @code{MVTIPL} instruction.
22241 The @samp{RX200} series does not have a hardware floating-point unit
22242 and so @option{-nofpu} is enabled by default when this type is
22245 @item -mbig-endian-data
22246 @itemx -mlittle-endian-data
22247 @opindex mbig-endian-data
22248 @opindex mlittle-endian-data
22249 Store data (but not code) in the big-endian format. The default is
22250 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
22253 @item -msmall-data-limit=@var{N}
22254 @opindex msmall-data-limit
22255 Specifies the maximum size in bytes of global and static variables
22256 which can be placed into the small data area. Using the small data
22257 area can lead to smaller and faster code, but the size of area is
22258 limited and it is up to the programmer to ensure that the area does
22259 not overflow. Also when the small data area is used one of the RX's
22260 registers (usually @code{r13}) is reserved for use pointing to this
22261 area, so it is no longer available for use by the compiler. This
22262 could result in slower and/or larger code if variables are pushed onto
22263 the stack instead of being held in this register.
22265 Note, common variables (variables that have not been initialized) and
22266 constants are not placed into the small data area as they are assigned
22267 to other sections in the output executable.
22269 The default value is zero, which disables this feature. Note, this
22270 feature is not enabled by default with higher optimization levels
22271 (@option{-O2} etc) because of the potentially detrimental effects of
22272 reserving a register. It is up to the programmer to experiment and
22273 discover whether this feature is of benefit to their program. See the
22274 description of the @option{-mpid} option for a description of how the
22275 actual register to hold the small data area pointer is chosen.
22281 Use the simulator runtime. The default is to use the libgloss
22282 board-specific runtime.
22284 @item -mas100-syntax
22285 @itemx -mno-as100-syntax
22286 @opindex mas100-syntax
22287 @opindex mno-as100-syntax
22288 When generating assembler output use a syntax that is compatible with
22289 Renesas's AS100 assembler. This syntax can also be handled by the GAS
22290 assembler, but it has some restrictions so it is not generated by default.
22292 @item -mmax-constant-size=@var{N}
22293 @opindex mmax-constant-size
22294 Specifies the maximum size, in bytes, of a constant that can be used as
22295 an operand in a RX instruction. Although the RX instruction set does
22296 allow constants of up to 4 bytes in length to be used in instructions,
22297 a longer value equates to a longer instruction. Thus in some
22298 circumstances it can be beneficial to restrict the size of constants
22299 that are used in instructions. Constants that are too big are instead
22300 placed into a constant pool and referenced via register indirection.
22302 The value @var{N} can be between 0 and 4. A value of 0 (the default)
22303 or 4 means that constants of any size are allowed.
22307 Enable linker relaxation. Linker relaxation is a process whereby the
22308 linker attempts to reduce the size of a program by finding shorter
22309 versions of various instructions. Disabled by default.
22311 @item -mint-register=@var{N}
22312 @opindex mint-register
22313 Specify the number of registers to reserve for fast interrupt handler
22314 functions. The value @var{N} can be between 0 and 4. A value of 1
22315 means that register @code{r13} is reserved for the exclusive use
22316 of fast interrupt handlers. A value of 2 reserves @code{r13} and
22317 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
22318 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
22319 A value of 0, the default, does not reserve any registers.
22321 @item -msave-acc-in-interrupts
22322 @opindex msave-acc-in-interrupts
22323 Specifies that interrupt handler functions should preserve the
22324 accumulator register. This is only necessary if normal code might use
22325 the accumulator register, for example because it performs 64-bit
22326 multiplications. The default is to ignore the accumulator as this
22327 makes the interrupt handlers faster.
22333 Enables the generation of position independent data. When enabled any
22334 access to constant data is done via an offset from a base address
22335 held in a register. This allows the location of constant data to be
22336 determined at run time without requiring the executable to be
22337 relocated, which is a benefit to embedded applications with tight
22338 memory constraints. Data that can be modified is not affected by this
22341 Note, using this feature reserves a register, usually @code{r13}, for
22342 the constant data base address. This can result in slower and/or
22343 larger code, especially in complicated functions.
22345 The actual register chosen to hold the constant data base address
22346 depends upon whether the @option{-msmall-data-limit} and/or the
22347 @option{-mint-register} command-line options are enabled. Starting
22348 with register @code{r13} and proceeding downwards, registers are
22349 allocated first to satisfy the requirements of @option{-mint-register},
22350 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
22351 is possible for the small data area register to be @code{r8} if both
22352 @option{-mint-register=4} and @option{-mpid} are specified on the
22355 By default this feature is not enabled. The default can be restored
22356 via the @option{-mno-pid} command-line option.
22358 @item -mno-warn-multiple-fast-interrupts
22359 @itemx -mwarn-multiple-fast-interrupts
22360 @opindex mno-warn-multiple-fast-interrupts
22361 @opindex mwarn-multiple-fast-interrupts
22362 Prevents GCC from issuing a warning message if it finds more than one
22363 fast interrupt handler when it is compiling a file. The default is to
22364 issue a warning for each extra fast interrupt handler found, as the RX
22365 only supports one such interrupt.
22367 @item -mallow-string-insns
22368 @itemx -mno-allow-string-insns
22369 @opindex mallow-string-insns
22370 @opindex mno-allow-string-insns
22371 Enables or disables the use of the string manipulation instructions
22372 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
22373 @code{SWHILE} and also the @code{RMPA} instruction. These
22374 instructions may prefetch data, which is not safe to do if accessing
22375 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
22376 for more information).
22378 The default is to allow these instructions, but it is not possible for
22379 GCC to reliably detect all circumstances where a string instruction
22380 might be used to access an I/O register, so their use cannot be
22381 disabled automatically. Instead it is reliant upon the programmer to
22382 use the @option{-mno-allow-string-insns} option if their program
22383 accesses I/O space.
22385 When the instructions are enabled GCC defines the C preprocessor
22386 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
22387 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
22393 Use only (or not only) @code{JSR} instructions to access functions.
22394 This option can be used when code size exceeds the range of @code{BSR}
22395 instructions. Note that @option{-mno-jsr} does not mean to not use
22396 @code{JSR} but instead means that any type of branch may be used.
22399 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
22400 has special significance to the RX port when used with the
22401 @code{interrupt} function attribute. This attribute indicates a
22402 function intended to process fast interrupts. GCC ensures
22403 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
22404 and/or @code{r13} and only provided that the normal use of the
22405 corresponding registers have been restricted via the
22406 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
22409 @node S/390 and zSeries Options
22410 @subsection S/390 and zSeries Options
22411 @cindex S/390 and zSeries Options
22413 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
22417 @itemx -msoft-float
22418 @opindex mhard-float
22419 @opindex msoft-float
22420 Use (do not use) the hardware floating-point instructions and registers
22421 for floating-point operations. When @option{-msoft-float} is specified,
22422 functions in @file{libgcc.a} are used to perform floating-point
22423 operations. When @option{-mhard-float} is specified, the compiler
22424 generates IEEE floating-point instructions. This is the default.
22427 @itemx -mno-hard-dfp
22429 @opindex mno-hard-dfp
22430 Use (do not use) the hardware decimal-floating-point instructions for
22431 decimal-floating-point operations. When @option{-mno-hard-dfp} is
22432 specified, functions in @file{libgcc.a} are used to perform
22433 decimal-floating-point operations. When @option{-mhard-dfp} is
22434 specified, the compiler generates decimal-floating-point hardware
22435 instructions. This is the default for @option{-march=z9-ec} or higher.
22437 @item -mlong-double-64
22438 @itemx -mlong-double-128
22439 @opindex mlong-double-64
22440 @opindex mlong-double-128
22441 These switches control the size of @code{long double} type. A size
22442 of 64 bits makes the @code{long double} type equivalent to the @code{double}
22443 type. This is the default.
22446 @itemx -mno-backchain
22447 @opindex mbackchain
22448 @opindex mno-backchain
22449 Store (do not store) the address of the caller's frame as backchain pointer
22450 into the callee's stack frame.
22451 A backchain may be needed to allow debugging using tools that do not understand
22452 DWARF call frame information.
22453 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
22454 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
22455 the backchain is placed into the topmost word of the 96/160 byte register
22458 In general, code compiled with @option{-mbackchain} is call-compatible with
22459 code compiled with @option{-mmo-backchain}; however, use of the backchain
22460 for debugging purposes usually requires that the whole binary is built with
22461 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
22462 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
22463 to build a linux kernel use @option{-msoft-float}.
22465 The default is to not maintain the backchain.
22467 @item -mpacked-stack
22468 @itemx -mno-packed-stack
22469 @opindex mpacked-stack
22470 @opindex mno-packed-stack
22471 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
22472 specified, the compiler uses the all fields of the 96/160 byte register save
22473 area only for their default purpose; unused fields still take up stack space.
22474 When @option{-mpacked-stack} is specified, register save slots are densely
22475 packed at the top of the register save area; unused space is reused for other
22476 purposes, allowing for more efficient use of the available stack space.
22477 However, when @option{-mbackchain} is also in effect, the topmost word of
22478 the save area is always used to store the backchain, and the return address
22479 register is always saved two words below the backchain.
22481 As long as the stack frame backchain is not used, code generated with
22482 @option{-mpacked-stack} is call-compatible with code generated with
22483 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
22484 S/390 or zSeries generated code that uses the stack frame backchain at run
22485 time, not just for debugging purposes. Such code is not call-compatible
22486 with code compiled with @option{-mpacked-stack}. Also, note that the
22487 combination of @option{-mbackchain},
22488 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
22489 to build a linux kernel use @option{-msoft-float}.
22491 The default is to not use the packed stack layout.
22494 @itemx -mno-small-exec
22495 @opindex msmall-exec
22496 @opindex mno-small-exec
22497 Generate (or do not generate) code using the @code{bras} instruction
22498 to do subroutine calls.
22499 This only works reliably if the total executable size does not
22500 exceed 64k. The default is to use the @code{basr} instruction instead,
22501 which does not have this limitation.
22507 When @option{-m31} is specified, generate code compliant to the
22508 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
22509 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
22510 particular to generate 64-bit instructions. For the @samp{s390}
22511 targets, the default is @option{-m31}, while the @samp{s390x}
22512 targets default to @option{-m64}.
22518 When @option{-mzarch} is specified, generate code using the
22519 instructions available on z/Architecture.
22520 When @option{-mesa} is specified, generate code using the
22521 instructions available on ESA/390. Note that @option{-mesa} is
22522 not possible with @option{-m64}.
22523 When generating code compliant to the GNU/Linux for S/390 ABI,
22524 the default is @option{-mesa}. When generating code compliant
22525 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
22531 The @option{-mhtm} option enables a set of builtins making use of
22532 instructions available with the transactional execution facility
22533 introduced with the IBM zEnterprise EC12 machine generation
22534 @ref{S/390 System z Built-in Functions}.
22535 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
22541 When @option{-mvx} is specified, generate code using the instructions
22542 available with the vector extension facility introduced with the IBM
22543 z13 machine generation.
22544 This option changes the ABI for some vector type values with regard to
22545 alignment and calling conventions. In case vector type values are
22546 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
22547 command will be added to mark the resulting binary with the ABI used.
22548 @option{-mvx} is enabled by default when using @option{-march=z13}.
22551 @itemx -mno-zvector
22553 @opindex mno-zvector
22554 The @option{-mzvector} option enables vector language extensions and
22555 builtins using instructions available with the vector extension
22556 facility introduced with the IBM z13 machine generation.
22557 This option adds support for @samp{vector} to be used as a keyword to
22558 define vector type variables and arguments. @samp{vector} is only
22559 available when GNU extensions are enabled. It will not be expanded
22560 when requesting strict standard compliance e.g. with @option{-std=c99}.
22561 In addition to the GCC low-level builtins @option{-mzvector} enables
22562 a set of builtins added for compatibility with AltiVec-style
22563 implementations like Power and Cell. In order to make use of these
22564 builtins the header file @file{vecintrin.h} needs to be included.
22565 @option{-mzvector} is disabled by default.
22571 Generate (or do not generate) code using the @code{mvcle} instruction
22572 to perform block moves. When @option{-mno-mvcle} is specified,
22573 use a @code{mvc} loop instead. This is the default unless optimizing for
22580 Print (or do not print) additional debug information when compiling.
22581 The default is to not print debug information.
22583 @item -march=@var{cpu-type}
22585 Generate code that runs on @var{cpu-type}, which is the name of a
22586 system representing a certain processor type. Possible values for
22587 @var{cpu-type} are @samp{z900}/@samp{arch5}, @samp{z990}/@samp{arch6},
22588 @samp{z9-109}, @samp{z9-ec}/@samp{arch7}, @samp{z10}/@samp{arch8},
22589 @samp{z196}/@samp{arch9}, @samp{zEC12}, @samp{z13}/@samp{arch11}, and
22592 The default is @option{-march=z900}. @samp{g5}/@samp{arch3} and
22593 @samp{g6} are deprecated and will be removed with future releases.
22595 Specifying @samp{native} as cpu type can be used to select the best
22596 architecture option for the host processor.
22597 @option{-march=native} has no effect if GCC does not recognize the
22600 @item -mtune=@var{cpu-type}
22602 Tune to @var{cpu-type} everything applicable about the generated code,
22603 except for the ABI and the set of available instructions.
22604 The list of @var{cpu-type} values is the same as for @option{-march}.
22605 The default is the value used for @option{-march}.
22608 @itemx -mno-tpf-trace
22609 @opindex mtpf-trace
22610 @opindex mno-tpf-trace
22611 Generate code that adds (does not add) in TPF OS specific branches to trace
22612 routines in the operating system. This option is off by default, even
22613 when compiling for the TPF OS@.
22616 @itemx -mno-fused-madd
22617 @opindex mfused-madd
22618 @opindex mno-fused-madd
22619 Generate code that uses (does not use) the floating-point multiply and
22620 accumulate instructions. These instructions are generated by default if
22621 hardware floating point is used.
22623 @item -mwarn-framesize=@var{framesize}
22624 @opindex mwarn-framesize
22625 Emit a warning if the current function exceeds the given frame size. Because
22626 this is a compile-time check it doesn't need to be a real problem when the program
22627 runs. It is intended to identify functions that most probably cause
22628 a stack overflow. It is useful to be used in an environment with limited stack
22629 size e.g.@: the linux kernel.
22631 @item -mwarn-dynamicstack
22632 @opindex mwarn-dynamicstack
22633 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
22634 arrays. This is generally a bad idea with a limited stack size.
22636 @item -mstack-guard=@var{stack-guard}
22637 @itemx -mstack-size=@var{stack-size}
22638 @opindex mstack-guard
22639 @opindex mstack-size
22640 If these options are provided the S/390 back end emits additional instructions in
22641 the function prologue that trigger a trap if the stack size is @var{stack-guard}
22642 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
22643 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
22644 the frame size of the compiled function is chosen.
22645 These options are intended to be used to help debugging stack overflow problems.
22646 The additionally emitted code causes only little overhead and hence can also be
22647 used in production-like systems without greater performance degradation. The given
22648 values have to be exact powers of 2 and @var{stack-size} has to be greater than
22649 @var{stack-guard} without exceeding 64k.
22650 In order to be efficient the extra code makes the assumption that the stack starts
22651 at an address aligned to the value given by @var{stack-size}.
22652 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
22654 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
22656 If the hotpatch option is enabled, a ``hot-patching'' function
22657 prologue is generated for all functions in the compilation unit.
22658 The funtion label is prepended with the given number of two-byte
22659 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
22660 the label, 2 * @var{post-halfwords} bytes are appended, using the
22661 largest NOP like instructions the architecture allows (maximum
22664 If both arguments are zero, hotpatching is disabled.
22666 This option can be overridden for individual functions with the
22667 @code{hotpatch} attribute.
22670 @node Score Options
22671 @subsection Score Options
22672 @cindex Score Options
22674 These options are defined for Score implementations:
22679 Compile code for big-endian mode. This is the default.
22683 Compile code for little-endian mode.
22687 Disable generation of @code{bcnz} instructions.
22691 Enable generation of unaligned load and store instructions.
22695 Enable the use of multiply-accumulate instructions. Disabled by default.
22699 Specify the SCORE5 as the target architecture.
22703 Specify the SCORE5U of the target architecture.
22707 Specify the SCORE7 as the target architecture. This is the default.
22711 Specify the SCORE7D as the target architecture.
22715 @subsection SH Options
22717 These @samp{-m} options are defined for the SH implementations:
22722 Generate code for the SH1.
22726 Generate code for the SH2.
22729 Generate code for the SH2e.
22733 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
22734 that the floating-point unit is not used.
22736 @item -m2a-single-only
22737 @opindex m2a-single-only
22738 Generate code for the SH2a-FPU, in such a way that no double-precision
22739 floating-point operations are used.
22742 @opindex m2a-single
22743 Generate code for the SH2a-FPU assuming the floating-point unit is in
22744 single-precision mode by default.
22748 Generate code for the SH2a-FPU assuming the floating-point unit is in
22749 double-precision mode by default.
22753 Generate code for the SH3.
22757 Generate code for the SH3e.
22761 Generate code for the SH4 without a floating-point unit.
22763 @item -m4-single-only
22764 @opindex m4-single-only
22765 Generate code for the SH4 with a floating-point unit that only
22766 supports single-precision arithmetic.
22770 Generate code for the SH4 assuming the floating-point unit is in
22771 single-precision mode by default.
22775 Generate code for the SH4.
22779 Generate code for SH4-100.
22781 @item -m4-100-nofpu
22782 @opindex m4-100-nofpu
22783 Generate code for SH4-100 in such a way that the
22784 floating-point unit is not used.
22786 @item -m4-100-single
22787 @opindex m4-100-single
22788 Generate code for SH4-100 assuming the floating-point unit is in
22789 single-precision mode by default.
22791 @item -m4-100-single-only
22792 @opindex m4-100-single-only
22793 Generate code for SH4-100 in such a way that no double-precision
22794 floating-point operations are used.
22798 Generate code for SH4-200.
22800 @item -m4-200-nofpu
22801 @opindex m4-200-nofpu
22802 Generate code for SH4-200 without in such a way that the
22803 floating-point unit is not used.
22805 @item -m4-200-single
22806 @opindex m4-200-single
22807 Generate code for SH4-200 assuming the floating-point unit is in
22808 single-precision mode by default.
22810 @item -m4-200-single-only
22811 @opindex m4-200-single-only
22812 Generate code for SH4-200 in such a way that no double-precision
22813 floating-point operations are used.
22817 Generate code for SH4-300.
22819 @item -m4-300-nofpu
22820 @opindex m4-300-nofpu
22821 Generate code for SH4-300 without in such a way that the
22822 floating-point unit is not used.
22824 @item -m4-300-single
22825 @opindex m4-300-single
22826 Generate code for SH4-300 in such a way that no double-precision
22827 floating-point operations are used.
22829 @item -m4-300-single-only
22830 @opindex m4-300-single-only
22831 Generate code for SH4-300 in such a way that no double-precision
22832 floating-point operations are used.
22836 Generate code for SH4-340 (no MMU, no FPU).
22840 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
22845 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
22846 floating-point unit is not used.
22848 @item -m4a-single-only
22849 @opindex m4a-single-only
22850 Generate code for the SH4a, in such a way that no double-precision
22851 floating-point operations are used.
22854 @opindex m4a-single
22855 Generate code for the SH4a assuming the floating-point unit is in
22856 single-precision mode by default.
22860 Generate code for the SH4a.
22864 Same as @option{-m4a-nofpu}, except that it implicitly passes
22865 @option{-dsp} to the assembler. GCC doesn't generate any DSP
22866 instructions at the moment.
22870 Compile code for the processor in big-endian mode.
22874 Compile code for the processor in little-endian mode.
22878 Align doubles at 64-bit boundaries. Note that this changes the calling
22879 conventions, and thus some functions from the standard C library do
22880 not work unless you recompile it first with @option{-mdalign}.
22884 Shorten some address references at link time, when possible; uses the
22885 linker option @option{-relax}.
22889 Use 32-bit offsets in @code{switch} tables. The default is to use
22894 Enable the use of bit manipulation instructions on SH2A.
22898 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
22899 alignment constraints.
22903 Comply with the calling conventions defined by Renesas.
22906 @opindex mno-renesas
22907 Comply with the calling conventions defined for GCC before the Renesas
22908 conventions were available. This option is the default for all
22909 targets of the SH toolchain.
22912 @opindex mnomacsave
22913 Mark the @code{MAC} register as call-clobbered, even if
22914 @option{-mrenesas} is given.
22920 Control the IEEE compliance of floating-point comparisons, which affects the
22921 handling of cases where the result of a comparison is unordered. By default
22922 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
22923 enabled @option{-mno-ieee} is implicitly set, which results in faster
22924 floating-point greater-equal and less-equal comparisons. The implicit settings
22925 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
22927 @item -minline-ic_invalidate
22928 @opindex minline-ic_invalidate
22929 Inline code to invalidate instruction cache entries after setting up
22930 nested function trampolines.
22931 This option has no effect if @option{-musermode} is in effect and the selected
22932 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
22934 If the selected code generation option does not allow the use of the @code{icbi}
22935 instruction, and @option{-musermode} is not in effect, the inlined code
22936 manipulates the instruction cache address array directly with an associative
22937 write. This not only requires privileged mode at run time, but it also
22938 fails if the cache line had been mapped via the TLB and has become unmapped.
22942 Dump instruction size and location in the assembly code.
22945 @opindex mpadstruct
22946 This option is deprecated. It pads structures to multiple of 4 bytes,
22947 which is incompatible with the SH ABI@.
22949 @item -matomic-model=@var{model}
22950 @opindex matomic-model=@var{model}
22951 Sets the model of atomic operations and additional parameters as a comma
22952 separated list. For details on the atomic built-in functions see
22953 @ref{__atomic Builtins}. The following models and parameters are supported:
22958 Disable compiler generated atomic sequences and emit library calls for atomic
22959 operations. This is the default if the target is not @code{sh*-*-linux*}.
22962 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
22963 built-in functions. The generated atomic sequences require additional support
22964 from the interrupt/exception handling code of the system and are only suitable
22965 for SH3* and SH4* single-core systems. This option is enabled by default when
22966 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
22967 this option also partially utilizes the hardware atomic instructions
22968 @code{movli.l} and @code{movco.l} to create more efficient code, unless
22969 @samp{strict} is specified.
22972 Generate software atomic sequences that use a variable in the thread control
22973 block. This is a variation of the gUSA sequences which can also be used on
22974 SH1* and SH2* targets. The generated atomic sequences require additional
22975 support from the interrupt/exception handling code of the system and are only
22976 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
22977 parameter has to be specified as well.
22980 Generate software atomic sequences that temporarily disable interrupts by
22981 setting @code{SR.IMASK = 1111}. This model works only when the program runs
22982 in privileged mode and is only suitable for single-core systems. Additional
22983 support from the interrupt/exception handling code of the system is not
22984 required. This model is enabled by default when the target is
22985 @code{sh*-*-linux*} and SH1* or SH2*.
22988 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
22989 instructions only. This is only available on SH4A and is suitable for
22990 multi-core systems. Since the hardware instructions support only 32 bit atomic
22991 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
22992 Code compiled with this option is also compatible with other software
22993 atomic model interrupt/exception handling systems if executed on an SH4A
22994 system. Additional support from the interrupt/exception handling code of the
22995 system is not required for this model.
22998 This parameter specifies the offset in bytes of the variable in the thread
22999 control block structure that should be used by the generated atomic sequences
23000 when the @samp{soft-tcb} model has been selected. For other models this
23001 parameter is ignored. The specified value must be an integer multiple of four
23002 and in the range 0-1020.
23005 This parameter prevents mixed usage of multiple atomic models, even if they
23006 are compatible, and makes the compiler generate atomic sequences of the
23007 specified model only.
23013 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
23014 Notice that depending on the particular hardware and software configuration
23015 this can degrade overall performance due to the operand cache line flushes
23016 that are implied by the @code{tas.b} instruction. On multi-core SH4A
23017 processors the @code{tas.b} instruction must be used with caution since it
23018 can result in data corruption for certain cache configurations.
23021 @opindex mprefergot
23022 When generating position-independent code, emit function calls using
23023 the Global Offset Table instead of the Procedure Linkage Table.
23026 @itemx -mno-usermode
23028 @opindex mno-usermode
23029 Don't allow (allow) the compiler generating privileged mode code. Specifying
23030 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
23031 inlined code would not work in user mode. @option{-musermode} is the default
23032 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
23033 @option{-musermode} has no effect, since there is no user mode.
23035 @item -multcost=@var{number}
23036 @opindex multcost=@var{number}
23037 Set the cost to assume for a multiply insn.
23039 @item -mdiv=@var{strategy}
23040 @opindex mdiv=@var{strategy}
23041 Set the division strategy to be used for integer division operations.
23042 @var{strategy} can be one of:
23047 Calls a library function that uses the single-step division instruction
23048 @code{div1} to perform the operation. Division by zero calculates an
23049 unspecified result and does not trap. This is the default except for SH4,
23050 SH2A and SHcompact.
23053 Calls a library function that performs the operation in double precision
23054 floating point. Division by zero causes a floating-point exception. This is
23055 the default for SHcompact with FPU. Specifying this for targets that do not
23056 have a double precision FPU defaults to @code{call-div1}.
23059 Calls a library function that uses a lookup table for small divisors and
23060 the @code{div1} instruction with case distinction for larger divisors. Division
23061 by zero calculates an unspecified result and does not trap. This is the default
23062 for SH4. Specifying this for targets that do not have dynamic shift
23063 instructions defaults to @code{call-div1}.
23067 When a division strategy has not been specified the default strategy is
23068 selected based on the current target. For SH2A the default strategy is to
23069 use the @code{divs} and @code{divu} instructions instead of library function
23072 @item -maccumulate-outgoing-args
23073 @opindex maccumulate-outgoing-args
23074 Reserve space once for outgoing arguments in the function prologue rather
23075 than around each call. Generally beneficial for performance and size. Also
23076 needed for unwinding to avoid changing the stack frame around conditional code.
23078 @item -mdivsi3_libfunc=@var{name}
23079 @opindex mdivsi3_libfunc=@var{name}
23080 Set the name of the library function used for 32-bit signed division to
23082 This only affects the name used in the @samp{call} division strategies, and
23083 the compiler still expects the same sets of input/output/clobbered registers as
23084 if this option were not present.
23086 @item -mfixed-range=@var{register-range}
23087 @opindex mfixed-range
23088 Generate code treating the given register range as fixed registers.
23089 A fixed register is one that the register allocator can not use. This is
23090 useful when compiling kernel code. A register range is specified as
23091 two registers separated by a dash. Multiple register ranges can be
23092 specified separated by a comma.
23094 @item -mbranch-cost=@var{num}
23095 @opindex mbranch-cost=@var{num}
23096 Assume @var{num} to be the cost for a branch instruction. Higher numbers
23097 make the compiler try to generate more branch-free code if possible.
23098 If not specified the value is selected depending on the processor type that
23099 is being compiled for.
23102 @itemx -mno-zdcbranch
23103 @opindex mzdcbranch
23104 @opindex mno-zdcbranch
23105 Assume (do not assume) that zero displacement conditional branch instructions
23106 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
23107 compiler prefers zero displacement branch code sequences. This is
23108 enabled by default when generating code for SH4 and SH4A. It can be explicitly
23109 disabled by specifying @option{-mno-zdcbranch}.
23111 @item -mcbranch-force-delay-slot
23112 @opindex mcbranch-force-delay-slot
23113 Force the usage of delay slots for conditional branches, which stuffs the delay
23114 slot with a @code{nop} if a suitable instruction can't be found. By default
23115 this option is disabled. It can be enabled to work around hardware bugs as
23116 found in the original SH7055.
23119 @itemx -mno-fused-madd
23120 @opindex mfused-madd
23121 @opindex mno-fused-madd
23122 Generate code that uses (does not use) the floating-point multiply and
23123 accumulate instructions. These instructions are generated by default
23124 if hardware floating point is used. The machine-dependent
23125 @option{-mfused-madd} option is now mapped to the machine-independent
23126 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
23127 mapped to @option{-ffp-contract=off}.
23133 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
23134 and cosine approximations. The option @option{-mfsca} must be used in
23135 combination with @option{-funsafe-math-optimizations}. It is enabled by default
23136 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
23137 approximations even if @option{-funsafe-math-optimizations} is in effect.
23143 Allow or disallow the compiler to emit the @code{fsrra} instruction for
23144 reciprocal square root approximations. The option @option{-mfsrra} must be used
23145 in combination with @option{-funsafe-math-optimizations} and
23146 @option{-ffinite-math-only}. It is enabled by default when generating code for
23147 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
23148 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
23151 @item -mpretend-cmove
23152 @opindex mpretend-cmove
23153 Prefer zero-displacement conditional branches for conditional move instruction
23154 patterns. This can result in faster code on the SH4 processor.
23158 Generate code using the FDPIC ABI.
23162 @node Solaris 2 Options
23163 @subsection Solaris 2 Options
23164 @cindex Solaris 2 options
23166 These @samp{-m} options are supported on Solaris 2:
23169 @item -mclear-hwcap
23170 @opindex mclear-hwcap
23171 @option{-mclear-hwcap} tells the compiler to remove the hardware
23172 capabilities generated by the Solaris assembler. This is only necessary
23173 when object files use ISA extensions not supported by the current
23174 machine, but check at runtime whether or not to use them.
23176 @item -mimpure-text
23177 @opindex mimpure-text
23178 @option{-mimpure-text}, used in addition to @option{-shared}, tells
23179 the compiler to not pass @option{-z text} to the linker when linking a
23180 shared object. Using this option, you can link position-dependent
23181 code into a shared object.
23183 @option{-mimpure-text} suppresses the ``relocations remain against
23184 allocatable but non-writable sections'' linker error message.
23185 However, the necessary relocations trigger copy-on-write, and the
23186 shared object is not actually shared across processes. Instead of
23187 using @option{-mimpure-text}, you should compile all source code with
23188 @option{-fpic} or @option{-fPIC}.
23192 These switches are supported in addition to the above on Solaris 2:
23197 Add support for multithreading using the POSIX threads library. This
23198 option sets flags for both the preprocessor and linker. This option does
23199 not affect the thread safety of object code produced by the compiler or
23200 that of libraries supplied with it.
23204 This is a synonym for @option{-pthreads}.
23207 @node SPARC Options
23208 @subsection SPARC Options
23209 @cindex SPARC options
23211 These @samp{-m} options are supported on the SPARC:
23214 @item -mno-app-regs
23216 @opindex mno-app-regs
23218 Specify @option{-mapp-regs} to generate output using the global registers
23219 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
23220 global register 1, each global register 2 through 4 is then treated as an
23221 allocable register that is clobbered by function calls. This is the default.
23223 To be fully SVR4 ABI-compliant at the cost of some performance loss,
23224 specify @option{-mno-app-regs}. You should compile libraries and system
23225 software with this option.
23231 With @option{-mflat}, the compiler does not generate save/restore instructions
23232 and uses a ``flat'' or single register window model. This model is compatible
23233 with the regular register window model. The local registers and the input
23234 registers (0--5) are still treated as ``call-saved'' registers and are
23235 saved on the stack as needed.
23237 With @option{-mno-flat} (the default), the compiler generates save/restore
23238 instructions (except for leaf functions). This is the normal operating mode.
23241 @itemx -mhard-float
23243 @opindex mhard-float
23244 Generate output containing floating-point instructions. This is the
23248 @itemx -msoft-float
23250 @opindex msoft-float
23251 Generate output containing library calls for floating point.
23252 @strong{Warning:} the requisite libraries are not available for all SPARC
23253 targets. Normally the facilities of the machine's usual C compiler are
23254 used, but this cannot be done directly in cross-compilation. You must make
23255 your own arrangements to provide suitable library functions for
23256 cross-compilation. The embedded targets @samp{sparc-*-aout} and
23257 @samp{sparclite-*-*} do provide software floating-point support.
23259 @option{-msoft-float} changes the calling convention in the output file;
23260 therefore, it is only useful if you compile @emph{all} of a program with
23261 this option. In particular, you need to compile @file{libgcc.a}, the
23262 library that comes with GCC, with @option{-msoft-float} in order for
23265 @item -mhard-quad-float
23266 @opindex mhard-quad-float
23267 Generate output containing quad-word (long double) floating-point
23270 @item -msoft-quad-float
23271 @opindex msoft-quad-float
23272 Generate output containing library calls for quad-word (long double)
23273 floating-point instructions. The functions called are those specified
23274 in the SPARC ABI@. This is the default.
23276 As of this writing, there are no SPARC implementations that have hardware
23277 support for the quad-word floating-point instructions. They all invoke
23278 a trap handler for one of these instructions, and then the trap handler
23279 emulates the effect of the instruction. Because of the trap handler overhead,
23280 this is much slower than calling the ABI library routines. Thus the
23281 @option{-msoft-quad-float} option is the default.
23283 @item -mno-unaligned-doubles
23284 @itemx -munaligned-doubles
23285 @opindex mno-unaligned-doubles
23286 @opindex munaligned-doubles
23287 Assume that doubles have 8-byte alignment. This is the default.
23289 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
23290 alignment only if they are contained in another type, or if they have an
23291 absolute address. Otherwise, it assumes they have 4-byte alignment.
23292 Specifying this option avoids some rare compatibility problems with code
23293 generated by other compilers. It is not the default because it results
23294 in a performance loss, especially for floating-point code.
23297 @itemx -mno-user-mode
23298 @opindex muser-mode
23299 @opindex mno-user-mode
23300 Do not generate code that can only run in supervisor mode. This is relevant
23301 only for the @code{casa} instruction emitted for the LEON3 processor. This
23304 @item -mfaster-structs
23305 @itemx -mno-faster-structs
23306 @opindex mfaster-structs
23307 @opindex mno-faster-structs
23308 With @option{-mfaster-structs}, the compiler assumes that structures
23309 should have 8-byte alignment. This enables the use of pairs of
23310 @code{ldd} and @code{std} instructions for copies in structure
23311 assignment, in place of twice as many @code{ld} and @code{st} pairs.
23312 However, the use of this changed alignment directly violates the SPARC
23313 ABI@. Thus, it's intended only for use on targets where the developer
23314 acknowledges that their resulting code is not directly in line with
23315 the rules of the ABI@.
23317 @item -mstd-struct-return
23318 @itemx -mno-std-struct-return
23319 @opindex mstd-struct-return
23320 @opindex mno-std-struct-return
23321 With @option{-mstd-struct-return}, the compiler generates checking code
23322 in functions returning structures or unions to detect size mismatches
23323 between the two sides of function calls, as per the 32-bit ABI@.
23325 The default is @option{-mno-std-struct-return}. This option has no effect
23328 @item -mcpu=@var{cpu_type}
23330 Set the instruction set, register set, and instruction scheduling parameters
23331 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
23332 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
23333 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
23334 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
23335 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
23336 @samp{niagara3}, @samp{niagara4} and @samp{niagara7}.
23338 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
23339 which selects the best architecture option for the host processor.
23340 @option{-mcpu=native} has no effect if GCC does not recognize
23343 Default instruction scheduling parameters are used for values that select
23344 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
23345 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
23347 Here is a list of each supported architecture and their supported
23355 supersparc, hypersparc, leon, leon3
23358 f930, f934, sparclite86x
23364 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4, niagara7
23367 By default (unless configured otherwise), GCC generates code for the V7
23368 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
23369 additionally optimizes it for the Cypress CY7C602 chip, as used in the
23370 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
23371 SPARCStation 1, 2, IPX etc.
23373 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
23374 architecture. The only difference from V7 code is that the compiler emits
23375 the integer multiply and integer divide instructions which exist in SPARC-V8
23376 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
23377 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
23380 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
23381 the SPARC architecture. This adds the integer multiply, integer divide step
23382 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
23383 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
23384 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
23385 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
23386 MB86934 chip, which is the more recent SPARClite with FPU@.
23388 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
23389 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
23390 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
23391 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
23392 optimizes it for the TEMIC SPARClet chip.
23394 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
23395 architecture. This adds 64-bit integer and floating-point move instructions,
23396 3 additional floating-point condition code registers and conditional move
23397 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
23398 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
23399 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
23400 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
23401 @option{-mcpu=niagara}, the compiler additionally optimizes it for
23402 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
23403 additionally optimizes it for Sun UltraSPARC T2 chips. With
23404 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
23405 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
23406 additionally optimizes it for Sun UltraSPARC T4 chips. With
23407 @option{-mcpu=niagara7}, the compiler additionally optimizes it for
23408 Oracle SPARC M7 chips.
23410 @item -mtune=@var{cpu_type}
23412 Set the instruction scheduling parameters for machine type
23413 @var{cpu_type}, but do not set the instruction set or register set that the
23414 option @option{-mcpu=@var{cpu_type}} does.
23416 The same values for @option{-mcpu=@var{cpu_type}} can be used for
23417 @option{-mtune=@var{cpu_type}}, but the only useful values are those
23418 that select a particular CPU implementation. Those are
23419 @samp{cypress}, @samp{supersparc}, @samp{hypersparc}, @samp{leon},
23420 @samp{leon3}, @samp{leon3v7}, @samp{f930}, @samp{f934},
23421 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
23422 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
23423 @samp{niagara4} and @samp{niagara7}. With native Solaris and
23424 GNU/Linux toolchains, @samp{native} can also be used.
23429 @opindex mno-v8plus
23430 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
23431 difference from the V8 ABI is that the global and out registers are
23432 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
23433 mode for all SPARC-V9 processors.
23439 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
23440 Visual Instruction Set extensions. The default is @option{-mno-vis}.
23446 With @option{-mvis2}, GCC generates code that takes advantage of
23447 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
23448 default is @option{-mvis2} when targeting a cpu that supports such
23449 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
23450 also sets @option{-mvis}.
23456 With @option{-mvis3}, GCC generates code that takes advantage of
23457 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
23458 default is @option{-mvis3} when targeting a cpu that supports such
23459 instructions, such as niagara-3 and later. Setting @option{-mvis3}
23460 also sets @option{-mvis2} and @option{-mvis}.
23466 With @option{-mvis4}, GCC generates code that takes advantage of
23467 version 4.0 of the UltraSPARC Visual Instruction Set extensions. The
23468 default is @option{-mvis4} when targeting a cpu that supports such
23469 instructions, such as niagara-7 and later. Setting @option{-mvis4}
23470 also sets @option{-mvis3}, @option{-mvis2} and @option{-mvis}.
23475 @opindex mno-cbcond
23476 With @option{-mcbcond}, GCC generates code that takes advantage of the UltraSPARC
23477 Compare-and-Branch-on-Condition instructions. The default is @option{-mcbcond}
23478 when targeting a CPU that supports such instructions, such as Niagara-4 and
23485 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
23486 Fused Multiply-Add Floating-point instructions. The default is @option{-mfmaf}
23487 when targeting a CPU that supports such instructions, such as Niagara-3 and
23494 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
23495 Population Count instruction. The default is @option{-mpopc}
23496 when targeting a CPU that supports such an instruction, such as Niagara-2 and
23503 With @option{-msubxc}, GCC generates code that takes advantage of the UltraSPARC
23504 Subtract-Extended-with-Carry instruction. The default is @option{-msubxc}
23505 when targeting a CPU that supports such an instruction, such as Niagara-7 and
23509 @opindex mfix-at697f
23510 Enable the documented workaround for the single erratum of the Atmel AT697F
23511 processor (which corresponds to erratum #13 of the AT697E processor).
23514 @opindex mfix-ut699
23515 Enable the documented workarounds for the floating-point errata and the data
23516 cache nullify errata of the UT699 processor.
23519 These @samp{-m} options are supported in addition to the above
23520 on SPARC-V9 processors in 64-bit environments:
23527 Generate code for a 32-bit or 64-bit environment.
23528 The 32-bit environment sets int, long and pointer to 32 bits.
23529 The 64-bit environment sets int to 32 bits and long and pointer
23532 @item -mcmodel=@var{which}
23534 Set the code model to one of
23538 The Medium/Low code model: 64-bit addresses, programs
23539 must be linked in the low 32 bits of memory. Programs can be statically
23540 or dynamically linked.
23543 The Medium/Middle code model: 64-bit addresses, programs
23544 must be linked in the low 44 bits of memory, the text and data segments must
23545 be less than 2GB in size and the data segment must be located within 2GB of
23549 The Medium/Anywhere code model: 64-bit addresses, programs
23550 may be linked anywhere in memory, the text and data segments must be less
23551 than 2GB in size and the data segment must be located within 2GB of the
23555 The Medium/Anywhere code model for embedded systems:
23556 64-bit addresses, the text and data segments must be less than 2GB in
23557 size, both starting anywhere in memory (determined at link time). The
23558 global register %g4 points to the base of the data segment. Programs
23559 are statically linked and PIC is not supported.
23562 @item -mmemory-model=@var{mem-model}
23563 @opindex mmemory-model
23564 Set the memory model in force on the processor to one of
23568 The default memory model for the processor and operating system.
23571 Relaxed Memory Order
23574 Partial Store Order
23580 Sequential Consistency
23583 These memory models are formally defined in Appendix D of the Sparc V9
23584 architecture manual, as set in the processor's @code{PSTATE.MM} field.
23587 @itemx -mno-stack-bias
23588 @opindex mstack-bias
23589 @opindex mno-stack-bias
23590 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
23591 frame pointer if present, are offset by @minus{}2047 which must be added back
23592 when making stack frame references. This is the default in 64-bit mode.
23593 Otherwise, assume no such offset is present.
23597 @subsection SPU Options
23598 @cindex SPU options
23600 These @samp{-m} options are supported on the SPU:
23604 @itemx -merror-reloc
23605 @opindex mwarn-reloc
23606 @opindex merror-reloc
23608 The loader for SPU does not handle dynamic relocations. By default, GCC
23609 gives an error when it generates code that requires a dynamic
23610 relocation. @option{-mno-error-reloc} disables the error,
23611 @option{-mwarn-reloc} generates a warning instead.
23614 @itemx -munsafe-dma
23616 @opindex munsafe-dma
23618 Instructions that initiate or test completion of DMA must not be
23619 reordered with respect to loads and stores of the memory that is being
23621 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
23622 memory accesses, but that can lead to inefficient code in places where the
23623 memory is known to not change. Rather than mark the memory as volatile,
23624 you can use @option{-msafe-dma} to tell the compiler to treat
23625 the DMA instructions as potentially affecting all memory.
23627 @item -mbranch-hints
23628 @opindex mbranch-hints
23630 By default, GCC generates a branch hint instruction to avoid
23631 pipeline stalls for always-taken or probably-taken branches. A hint
23632 is not generated closer than 8 instructions away from its branch.
23633 There is little reason to disable them, except for debugging purposes,
23634 or to make an object a little bit smaller.
23638 @opindex msmall-mem
23639 @opindex mlarge-mem
23641 By default, GCC generates code assuming that addresses are never larger
23642 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
23643 a full 32-bit address.
23648 By default, GCC links against startup code that assumes the SPU-style
23649 main function interface (which has an unconventional parameter list).
23650 With @option{-mstdmain}, GCC links your program against startup
23651 code that assumes a C99-style interface to @code{main}, including a
23652 local copy of @code{argv} strings.
23654 @item -mfixed-range=@var{register-range}
23655 @opindex mfixed-range
23656 Generate code treating the given register range as fixed registers.
23657 A fixed register is one that the register allocator cannot use. This is
23658 useful when compiling kernel code. A register range is specified as
23659 two registers separated by a dash. Multiple register ranges can be
23660 specified separated by a comma.
23666 Compile code assuming that pointers to the PPU address space accessed
23667 via the @code{__ea} named address space qualifier are either 32 or 64
23668 bits wide. The default is 32 bits. As this is an ABI-changing option,
23669 all object code in an executable must be compiled with the same setting.
23671 @item -maddress-space-conversion
23672 @itemx -mno-address-space-conversion
23673 @opindex maddress-space-conversion
23674 @opindex mno-address-space-conversion
23675 Allow/disallow treating the @code{__ea} address space as superset
23676 of the generic address space. This enables explicit type casts
23677 between @code{__ea} and generic pointer as well as implicit
23678 conversions of generic pointers to @code{__ea} pointers. The
23679 default is to allow address space pointer conversions.
23681 @item -mcache-size=@var{cache-size}
23682 @opindex mcache-size
23683 This option controls the version of libgcc that the compiler links to an
23684 executable and selects a software-managed cache for accessing variables
23685 in the @code{__ea} address space with a particular cache size. Possible
23686 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
23687 and @samp{128}. The default cache size is 64KB.
23689 @item -matomic-updates
23690 @itemx -mno-atomic-updates
23691 @opindex matomic-updates
23692 @opindex mno-atomic-updates
23693 This option controls the version of libgcc that the compiler links to an
23694 executable and selects whether atomic updates to the software-managed
23695 cache of PPU-side variables are used. If you use atomic updates, changes
23696 to a PPU variable from SPU code using the @code{__ea} named address space
23697 qualifier do not interfere with changes to other PPU variables residing
23698 in the same cache line from PPU code. If you do not use atomic updates,
23699 such interference may occur; however, writing back cache lines is
23700 more efficient. The default behavior is to use atomic updates.
23703 @itemx -mdual-nops=@var{n}
23704 @opindex mdual-nops
23705 By default, GCC inserts NOPs to increase dual issue when it expects
23706 it to increase performance. @var{n} can be a value from 0 to 10. A
23707 smaller @var{n} inserts fewer NOPs. 10 is the default, 0 is the
23708 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
23710 @item -mhint-max-nops=@var{n}
23711 @opindex mhint-max-nops
23712 Maximum number of NOPs to insert for a branch hint. A branch hint must
23713 be at least 8 instructions away from the branch it is affecting. GCC
23714 inserts up to @var{n} NOPs to enforce this, otherwise it does not
23715 generate the branch hint.
23717 @item -mhint-max-distance=@var{n}
23718 @opindex mhint-max-distance
23719 The encoding of the branch hint instruction limits the hint to be within
23720 256 instructions of the branch it is affecting. By default, GCC makes
23721 sure it is within 125.
23724 @opindex msafe-hints
23725 Work around a hardware bug that causes the SPU to stall indefinitely.
23726 By default, GCC inserts the @code{hbrp} instruction to make sure
23727 this stall won't happen.
23731 @node System V Options
23732 @subsection Options for System V
23734 These additional options are available on System V Release 4 for
23735 compatibility with other compilers on those systems:
23740 Create a shared object.
23741 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
23745 Identify the versions of each tool used by the compiler, in a
23746 @code{.ident} assembler directive in the output.
23750 Refrain from adding @code{.ident} directives to the output file (this is
23753 @item -YP,@var{dirs}
23755 Search the directories @var{dirs}, and no others, for libraries
23756 specified with @option{-l}.
23758 @item -Ym,@var{dir}
23760 Look in the directory @var{dir} to find the M4 preprocessor.
23761 The assembler uses this option.
23762 @c This is supposed to go with a -Yd for predefined M4 macro files, but
23763 @c the generic assembler that comes with Solaris takes just -Ym.
23766 @node TILE-Gx Options
23767 @subsection TILE-Gx Options
23768 @cindex TILE-Gx options
23770 These @samp{-m} options are supported on the TILE-Gx:
23773 @item -mcmodel=small
23774 @opindex mcmodel=small
23775 Generate code for the small model. The distance for direct calls is
23776 limited to 500M in either direction. PC-relative addresses are 32
23777 bits. Absolute addresses support the full address range.
23779 @item -mcmodel=large
23780 @opindex mcmodel=large
23781 Generate code for the large model. There is no limitation on call
23782 distance, pc-relative addresses, or absolute addresses.
23784 @item -mcpu=@var{name}
23786 Selects the type of CPU to be targeted. Currently the only supported
23787 type is @samp{tilegx}.
23793 Generate code for a 32-bit or 64-bit environment. The 32-bit
23794 environment sets int, long, and pointer to 32 bits. The 64-bit
23795 environment sets int to 32 bits and long and pointer to 64 bits.
23798 @itemx -mlittle-endian
23799 @opindex mbig-endian
23800 @opindex mlittle-endian
23801 Generate code in big/little endian mode, respectively.
23804 @node TILEPro Options
23805 @subsection TILEPro Options
23806 @cindex TILEPro options
23808 These @samp{-m} options are supported on the TILEPro:
23811 @item -mcpu=@var{name}
23813 Selects the type of CPU to be targeted. Currently the only supported
23814 type is @samp{tilepro}.
23818 Generate code for a 32-bit environment, which sets int, long, and
23819 pointer to 32 bits. This is the only supported behavior so the flag
23820 is essentially ignored.
23824 @subsection V850 Options
23825 @cindex V850 Options
23827 These @samp{-m} options are defined for V850 implementations:
23831 @itemx -mno-long-calls
23832 @opindex mlong-calls
23833 @opindex mno-long-calls
23834 Treat all calls as being far away (near). If calls are assumed to be
23835 far away, the compiler always loads the function's address into a
23836 register, and calls indirect through the pointer.
23842 Do not optimize (do optimize) basic blocks that use the same index
23843 pointer 4 or more times to copy pointer into the @code{ep} register, and
23844 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
23845 option is on by default if you optimize.
23847 @item -mno-prolog-function
23848 @itemx -mprolog-function
23849 @opindex mno-prolog-function
23850 @opindex mprolog-function
23851 Do not use (do use) external functions to save and restore registers
23852 at the prologue and epilogue of a function. The external functions
23853 are slower, but use less code space if more than one function saves
23854 the same number of registers. The @option{-mprolog-function} option
23855 is on by default if you optimize.
23859 Try to make the code as small as possible. At present, this just turns
23860 on the @option{-mep} and @option{-mprolog-function} options.
23862 @item -mtda=@var{n}
23864 Put static or global variables whose size is @var{n} bytes or less into
23865 the tiny data area that register @code{ep} points to. The tiny data
23866 area can hold up to 256 bytes in total (128 bytes for byte references).
23868 @item -msda=@var{n}
23870 Put static or global variables whose size is @var{n} bytes or less into
23871 the small data area that register @code{gp} points to. The small data
23872 area can hold up to 64 kilobytes.
23874 @item -mzda=@var{n}
23876 Put static or global variables whose size is @var{n} bytes or less into
23877 the first 32 kilobytes of memory.
23881 Specify that the target processor is the V850.
23885 Specify that the target processor is the V850E3V5. The preprocessor
23886 constant @code{__v850e3v5__} is defined if this option is used.
23890 Specify that the target processor is the V850E3V5. This is an alias for
23891 the @option{-mv850e3v5} option.
23895 Specify that the target processor is the V850E2V3. The preprocessor
23896 constant @code{__v850e2v3__} is defined if this option is used.
23900 Specify that the target processor is the V850E2. The preprocessor
23901 constant @code{__v850e2__} is defined if this option is used.
23905 Specify that the target processor is the V850E1. The preprocessor
23906 constants @code{__v850e1__} and @code{__v850e__} are defined if
23907 this option is used.
23911 Specify that the target processor is the V850ES. This is an alias for
23912 the @option{-mv850e1} option.
23916 Specify that the target processor is the V850E@. The preprocessor
23917 constant @code{__v850e__} is defined if this option is used.
23919 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
23920 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
23921 are defined then a default target processor is chosen and the
23922 relevant @samp{__v850*__} preprocessor constant is defined.
23924 The preprocessor constants @code{__v850} and @code{__v851__} are always
23925 defined, regardless of which processor variant is the target.
23927 @item -mdisable-callt
23928 @itemx -mno-disable-callt
23929 @opindex mdisable-callt
23930 @opindex mno-disable-callt
23931 This option suppresses generation of the @code{CALLT} instruction for the
23932 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
23935 This option is enabled by default when the RH850 ABI is
23936 in use (see @option{-mrh850-abi}), and disabled by default when the
23937 GCC ABI is in use. If @code{CALLT} instructions are being generated
23938 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
23944 Pass on (or do not pass on) the @option{-mrelax} command-line option
23948 @itemx -mno-long-jumps
23949 @opindex mlong-jumps
23950 @opindex mno-long-jumps
23951 Disable (or re-enable) the generation of PC-relative jump instructions.
23954 @itemx -mhard-float
23955 @opindex msoft-float
23956 @opindex mhard-float
23957 Disable (or re-enable) the generation of hardware floating point
23958 instructions. This option is only significant when the target
23959 architecture is @samp{V850E2V3} or higher. If hardware floating point
23960 instructions are being generated then the C preprocessor symbol
23961 @code{__FPU_OK__} is defined, otherwise the symbol
23962 @code{__NO_FPU__} is defined.
23966 Enables the use of the e3v5 LOOP instruction. The use of this
23967 instruction is not enabled by default when the e3v5 architecture is
23968 selected because its use is still experimental.
23972 @opindex mrh850-abi
23974 Enables support for the RH850 version of the V850 ABI. This is the
23975 default. With this version of the ABI the following rules apply:
23979 Integer sized structures and unions are returned via a memory pointer
23980 rather than a register.
23983 Large structures and unions (more than 8 bytes in size) are passed by
23987 Functions are aligned to 16-bit boundaries.
23990 The @option{-m8byte-align} command-line option is supported.
23993 The @option{-mdisable-callt} command-line option is enabled by
23994 default. The @option{-mno-disable-callt} command-line option is not
23998 When this version of the ABI is enabled the C preprocessor symbol
23999 @code{__V850_RH850_ABI__} is defined.
24003 Enables support for the old GCC version of the V850 ABI. With this
24004 version of the ABI the following rules apply:
24008 Integer sized structures and unions are returned in register @code{r10}.
24011 Large structures and unions (more than 8 bytes in size) are passed by
24015 Functions are aligned to 32-bit boundaries, unless optimizing for
24019 The @option{-m8byte-align} command-line option is not supported.
24022 The @option{-mdisable-callt} command-line option is supported but not
24023 enabled by default.
24026 When this version of the ABI is enabled the C preprocessor symbol
24027 @code{__V850_GCC_ABI__} is defined.
24029 @item -m8byte-align
24030 @itemx -mno-8byte-align
24031 @opindex m8byte-align
24032 @opindex mno-8byte-align
24033 Enables support for @code{double} and @code{long long} types to be
24034 aligned on 8-byte boundaries. The default is to restrict the
24035 alignment of all objects to at most 4-bytes. When
24036 @option{-m8byte-align} is in effect the C preprocessor symbol
24037 @code{__V850_8BYTE_ALIGN__} is defined.
24040 @opindex mbig-switch
24041 Generate code suitable for big switch tables. Use this option only if
24042 the assembler/linker complain about out of range branches within a switch
24047 This option causes r2 and r5 to be used in the code generated by
24048 the compiler. This setting is the default.
24050 @item -mno-app-regs
24051 @opindex mno-app-regs
24052 This option causes r2 and r5 to be treated as fixed registers.
24057 @subsection VAX Options
24058 @cindex VAX options
24060 These @samp{-m} options are defined for the VAX:
24065 Do not output certain jump instructions (@code{aobleq} and so on)
24066 that the Unix assembler for the VAX cannot handle across long
24071 Do output those jump instructions, on the assumption that the
24072 GNU assembler is being used.
24076 Output code for G-format floating-point numbers instead of D-format.
24079 @node Visium Options
24080 @subsection Visium Options
24081 @cindex Visium options
24087 A program which performs file I/O and is destined to run on an MCM target
24088 should be linked with this option. It causes the libraries libc.a and
24089 libdebug.a to be linked. The program should be run on the target under
24090 the control of the GDB remote debugging stub.
24094 A program which performs file I/O and is destined to run on the simulator
24095 should be linked with option. This causes libraries libc.a and libsim.a to
24099 @itemx -mhard-float
24101 @opindex mhard-float
24102 Generate code containing floating-point instructions. This is the
24106 @itemx -msoft-float
24108 @opindex msoft-float
24109 Generate code containing library calls for floating-point.
24111 @option{-msoft-float} changes the calling convention in the output file;
24112 therefore, it is only useful if you compile @emph{all} of a program with
24113 this option. In particular, you need to compile @file{libgcc.a}, the
24114 library that comes with GCC, with @option{-msoft-float} in order for
24117 @item -mcpu=@var{cpu_type}
24119 Set the instruction set, register set, and instruction scheduling parameters
24120 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
24121 @samp{mcm}, @samp{gr5} and @samp{gr6}.
24123 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
24125 By default (unless configured otherwise), GCC generates code for the GR5
24126 variant of the Visium architecture.
24128 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
24129 architecture. The only difference from GR5 code is that the compiler will
24130 generate block move instructions.
24132 @item -mtune=@var{cpu_type}
24134 Set the instruction scheduling parameters for machine type @var{cpu_type},
24135 but do not set the instruction set or register set that the option
24136 @option{-mcpu=@var{cpu_type}} would.
24140 Generate code for the supervisor mode, where there are no restrictions on
24141 the access to general registers. This is the default.
24144 @opindex muser-mode
24145 Generate code for the user mode, where the access to some general registers
24146 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
24147 mode; on the GR6, only registers r29 to r31 are affected.
24151 @subsection VMS Options
24153 These @samp{-m} options are defined for the VMS implementations:
24156 @item -mvms-return-codes
24157 @opindex mvms-return-codes
24158 Return VMS condition codes from @code{main}. The default is to return POSIX-style
24159 condition (e.g.@ error) codes.
24161 @item -mdebug-main=@var{prefix}
24162 @opindex mdebug-main=@var{prefix}
24163 Flag the first routine whose name starts with @var{prefix} as the main
24164 routine for the debugger.
24168 Default to 64-bit memory allocation routines.
24170 @item -mpointer-size=@var{size}
24171 @opindex mpointer-size=@var{size}
24172 Set the default size of pointers. Possible options for @var{size} are
24173 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
24174 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
24175 The later option disables @code{pragma pointer_size}.
24178 @node VxWorks Options
24179 @subsection VxWorks Options
24180 @cindex VxWorks Options
24182 The options in this section are defined for all VxWorks targets.
24183 Options specific to the target hardware are listed with the other
24184 options for that target.
24189 GCC can generate code for both VxWorks kernels and real time processes
24190 (RTPs). This option switches from the former to the latter. It also
24191 defines the preprocessor macro @code{__RTP__}.
24194 @opindex non-static
24195 Link an RTP executable against shared libraries rather than static
24196 libraries. The options @option{-static} and @option{-shared} can
24197 also be used for RTPs (@pxref{Link Options}); @option{-static}
24204 These options are passed down to the linker. They are defined for
24205 compatibility with Diab.
24208 @opindex Xbind-lazy
24209 Enable lazy binding of function calls. This option is equivalent to
24210 @option{-Wl,-z,now} and is defined for compatibility with Diab.
24214 Disable lazy binding of function calls. This option is the default and
24215 is defined for compatibility with Diab.
24219 @subsection x86 Options
24220 @cindex x86 Options
24222 These @samp{-m} options are defined for the x86 family of computers.
24226 @item -march=@var{cpu-type}
24228 Generate instructions for the machine type @var{cpu-type}. In contrast to
24229 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
24230 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
24231 to generate code that may not run at all on processors other than the one
24232 indicated. Specifying @option{-march=@var{cpu-type}} implies
24233 @option{-mtune=@var{cpu-type}}.
24235 The choices for @var{cpu-type} are:
24239 This selects the CPU to generate code for at compilation time by determining
24240 the processor type of the compiling machine. Using @option{-march=native}
24241 enables all instruction subsets supported by the local machine (hence
24242 the result might not run on different machines). Using @option{-mtune=native}
24243 produces code optimized for the local machine under the constraints
24244 of the selected instruction set.
24247 Original Intel i386 CPU@.
24250 Intel i486 CPU@. (No scheduling is implemented for this chip.)
24254 Intel Pentium CPU with no MMX support.
24257 Intel Lakemont MCU, based on Intel Pentium CPU.
24260 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
24263 Intel Pentium Pro CPU@.
24266 When used with @option{-march}, the Pentium Pro
24267 instruction set is used, so the code runs on all i686 family chips.
24268 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
24271 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
24276 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
24280 Intel Pentium M; low-power version of Intel Pentium III CPU
24281 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
24285 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
24288 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
24292 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
24293 SSE2 and SSE3 instruction set support.
24296 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
24297 instruction set support.
24300 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
24301 SSE4.1, SSE4.2 and POPCNT instruction set support.
24304 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
24305 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
24308 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
24309 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
24312 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
24313 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
24314 instruction set support.
24317 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24318 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24319 BMI, BMI2 and F16C instruction set support.
24322 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24323 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24324 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
24327 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24328 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24329 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
24330 XSAVES instruction set support.
24333 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
24334 instruction set support.
24337 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24338 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
24341 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
24342 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24343 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
24344 AVX512CD instruction set support.
24346 @item skylake-avx512
24347 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
24348 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24349 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
24350 AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
24353 AMD K6 CPU with MMX instruction set support.
24357 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
24360 @itemx athlon-tbird
24361 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
24367 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
24368 instruction set support.
24374 Processors based on the AMD K8 core with x86-64 instruction set support,
24375 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
24376 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
24377 instruction set extensions.)
24380 @itemx opteron-sse3
24381 @itemx athlon64-sse3
24382 Improved versions of AMD K8 cores with SSE3 instruction set support.
24386 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
24387 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
24388 instruction set extensions.)
24391 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
24392 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
24393 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
24395 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
24396 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
24397 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
24400 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
24401 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
24402 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
24403 64-bit instruction set extensions.
24405 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
24406 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
24407 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
24408 SSE4.2, ABM and 64-bit instruction set extensions.
24411 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
24412 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
24413 SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
24414 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
24415 instruction set extensions.
24418 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
24419 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
24420 instruction set extensions.)
24423 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
24424 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
24425 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
24428 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
24432 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
24433 instruction set support.
24436 VIA C3 CPU with MMX and 3DNow!@: instruction set support.
24437 (No scheduling is implemented for this chip.)
24440 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
24441 (No scheduling is implemented for this chip.)
24444 VIA C7 (Esther) CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
24445 (No scheduling is implemented for this chip.)
24448 VIA Eden Samuel 2 CPU with MMX and 3DNow!@: instruction set support.
24449 (No scheduling is implemented for this chip.)
24452 VIA Eden Nehemiah CPU with MMX and SSE instruction set support.
24453 (No scheduling is implemented for this chip.)
24456 VIA Eden Esther CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
24457 (No scheduling is implemented for this chip.)
24460 VIA Eden X2 CPU with x86-64, MMX, SSE, SSE2 and SSE3 instruction set support.
24461 (No scheduling is implemented for this chip.)
24464 VIA Eden X4 CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2,
24465 AVX and AVX2 instruction set support.
24466 (No scheduling is implemented for this chip.)
24469 Generic VIA Nano CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
24470 instruction set support.
24471 (No scheduling is implemented for this chip.)
24474 VIA Nano 1xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
24475 instruction set support.
24476 (No scheduling is implemented for this chip.)
24479 VIA Nano 2xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
24480 instruction set support.
24481 (No scheduling is implemented for this chip.)
24484 VIA Nano 3xxx CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
24485 instruction set support.
24486 (No scheduling is implemented for this chip.)
24489 VIA Nano Dual Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
24490 instruction set support.
24491 (No scheduling is implemented for this chip.)
24494 VIA Nano Quad Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
24495 instruction set support.
24496 (No scheduling is implemented for this chip.)
24499 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
24502 @item -mtune=@var{cpu-type}
24504 Tune to @var{cpu-type} everything applicable about the generated code, except
24505 for the ABI and the set of available instructions.
24506 While picking a specific @var{cpu-type} schedules things appropriately
24507 for that particular chip, the compiler does not generate any code that
24508 cannot run on the default machine type unless you use a
24509 @option{-march=@var{cpu-type}} option.
24510 For example, if GCC is configured for i686-pc-linux-gnu
24511 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
24512 but still runs on i686 machines.
24514 The choices for @var{cpu-type} are the same as for @option{-march}.
24515 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
24519 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
24520 If you know the CPU on which your code will run, then you should use
24521 the corresponding @option{-mtune} or @option{-march} option instead of
24522 @option{-mtune=generic}. But, if you do not know exactly what CPU users
24523 of your application will have, then you should use this option.
24525 As new processors are deployed in the marketplace, the behavior of this
24526 option will change. Therefore, if you upgrade to a newer version of
24527 GCC, code generation controlled by this option will change to reflect
24529 that are most common at the time that version of GCC is released.
24531 There is no @option{-march=generic} option because @option{-march}
24532 indicates the instruction set the compiler can use, and there is no
24533 generic instruction set applicable to all processors. In contrast,
24534 @option{-mtune} indicates the processor (or, in this case, collection of
24535 processors) for which the code is optimized.
24538 Produce code optimized for the most current Intel processors, which are
24539 Haswell and Silvermont for this version of GCC. If you know the CPU
24540 on which your code will run, then you should use the corresponding
24541 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
24542 But, if you want your application performs better on both Haswell and
24543 Silvermont, then you should use this option.
24545 As new Intel processors are deployed in the marketplace, the behavior of
24546 this option will change. Therefore, if you upgrade to a newer version of
24547 GCC, code generation controlled by this option will change to reflect
24548 the most current Intel processors at the time that version of GCC is
24551 There is no @option{-march=intel} option because @option{-march} indicates
24552 the instruction set the compiler can use, and there is no common
24553 instruction set applicable to all processors. In contrast,
24554 @option{-mtune} indicates the processor (or, in this case, collection of
24555 processors) for which the code is optimized.
24558 @item -mcpu=@var{cpu-type}
24560 A deprecated synonym for @option{-mtune}.
24562 @item -mfpmath=@var{unit}
24564 Generate floating-point arithmetic for selected unit @var{unit}. The choices
24565 for @var{unit} are:
24569 Use the standard 387 floating-point coprocessor present on the majority of chips and
24570 emulated otherwise. Code compiled with this option runs almost everywhere.
24571 The temporary results are computed in 80-bit precision instead of the precision
24572 specified by the type, resulting in slightly different results compared to most
24573 of other chips. See @option{-ffloat-store} for more detailed description.
24575 This is the default choice for x86-32 targets.
24578 Use scalar floating-point instructions present in the SSE instruction set.
24579 This instruction set is supported by Pentium III and newer chips,
24580 and in the AMD line
24581 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
24582 instruction set supports only single-precision arithmetic, thus the double and
24583 extended-precision arithmetic are still done using 387. A later version, present
24584 only in Pentium 4 and AMD x86-64 chips, supports double-precision
24587 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
24588 or @option{-msse2} switches to enable SSE extensions and make this option
24589 effective. For the x86-64 compiler, these extensions are enabled by default.
24591 The resulting code should be considerably faster in the majority of cases and avoid
24592 the numerical instability problems of 387 code, but may break some existing
24593 code that expects temporaries to be 80 bits.
24595 This is the default choice for the x86-64 compiler.
24600 Attempt to utilize both instruction sets at once. This effectively doubles the
24601 amount of available registers, and on chips with separate execution units for
24602 387 and SSE the execution resources too. Use this option with care, as it is
24603 still experimental, because the GCC register allocator does not model separate
24604 functional units well, resulting in unstable performance.
24607 @item -masm=@var{dialect}
24608 @opindex masm=@var{dialect}
24609 Output assembly instructions using selected @var{dialect}. Also affects
24610 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
24611 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
24612 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
24613 not support @samp{intel}.
24616 @itemx -mno-ieee-fp
24618 @opindex mno-ieee-fp
24619 Control whether or not the compiler uses IEEE floating-point
24620 comparisons. These correctly handle the case where the result of a
24621 comparison is unordered.
24626 @opindex mhard-float
24627 Generate output containing 80387 instructions for floating point.
24632 @opindex msoft-float
24633 Generate output containing library calls for floating point.
24635 @strong{Warning:} the requisite libraries are not part of GCC@.
24636 Normally the facilities of the machine's usual C compiler are used, but
24637 this can't be done directly in cross-compilation. You must make your
24638 own arrangements to provide suitable library functions for
24641 On machines where a function returns floating-point results in the 80387
24642 register stack, some floating-point opcodes may be emitted even if
24643 @option{-msoft-float} is used.
24645 @item -mno-fp-ret-in-387
24646 @opindex mno-fp-ret-in-387
24647 Do not use the FPU registers for return values of functions.
24649 The usual calling convention has functions return values of types
24650 @code{float} and @code{double} in an FPU register, even if there
24651 is no FPU@. The idea is that the operating system should emulate
24654 The option @option{-mno-fp-ret-in-387} causes such values to be returned
24655 in ordinary CPU registers instead.
24657 @item -mno-fancy-math-387
24658 @opindex mno-fancy-math-387
24659 Some 387 emulators do not support the @code{sin}, @code{cos} and
24660 @code{sqrt} instructions for the 387. Specify this option to avoid
24661 generating those instructions. This option is the default on
24662 OpenBSD and NetBSD@. This option is overridden when @option{-march}
24663 indicates that the target CPU always has an FPU and so the
24664 instruction does not need emulation. These
24665 instructions are not generated unless you also use the
24666 @option{-funsafe-math-optimizations} switch.
24668 @item -malign-double
24669 @itemx -mno-align-double
24670 @opindex malign-double
24671 @opindex mno-align-double
24672 Control whether GCC aligns @code{double}, @code{long double}, and
24673 @code{long long} variables on a two-word boundary or a one-word
24674 boundary. Aligning @code{double} variables on a two-word boundary
24675 produces code that runs somewhat faster on a Pentium at the
24676 expense of more memory.
24678 On x86-64, @option{-malign-double} is enabled by default.
24680 @strong{Warning:} if you use the @option{-malign-double} switch,
24681 structures containing the above types are aligned differently than
24682 the published application binary interface specifications for the x86-32
24683 and are not binary compatible with structures in code compiled
24684 without that switch.
24686 @item -m96bit-long-double
24687 @itemx -m128bit-long-double
24688 @opindex m96bit-long-double
24689 @opindex m128bit-long-double
24690 These switches control the size of @code{long double} type. The x86-32
24691 application binary interface specifies the size to be 96 bits,
24692 so @option{-m96bit-long-double} is the default in 32-bit mode.
24694 Modern architectures (Pentium and newer) prefer @code{long double}
24695 to be aligned to an 8- or 16-byte boundary. In arrays or structures
24696 conforming to the ABI, this is not possible. So specifying
24697 @option{-m128bit-long-double} aligns @code{long double}
24698 to a 16-byte boundary by padding the @code{long double} with an additional
24701 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
24702 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
24704 Notice that neither of these options enable any extra precision over the x87
24705 standard of 80 bits for a @code{long double}.
24707 @strong{Warning:} if you override the default value for your target ABI, this
24708 changes the size of
24709 structures and arrays containing @code{long double} variables,
24710 as well as modifying the function calling convention for functions taking
24711 @code{long double}. Hence they are not binary-compatible
24712 with code compiled without that switch.
24714 @item -mlong-double-64
24715 @itemx -mlong-double-80
24716 @itemx -mlong-double-128
24717 @opindex mlong-double-64
24718 @opindex mlong-double-80
24719 @opindex mlong-double-128
24720 These switches control the size of @code{long double} type. A size
24721 of 64 bits makes the @code{long double} type equivalent to the @code{double}
24722 type. This is the default for 32-bit Bionic C library. A size
24723 of 128 bits makes the @code{long double} type equivalent to the
24724 @code{__float128} type. This is the default for 64-bit Bionic C library.
24726 @strong{Warning:} if you override the default value for your target ABI, this
24727 changes the size of
24728 structures and arrays containing @code{long double} variables,
24729 as well as modifying the function calling convention for functions taking
24730 @code{long double}. Hence they are not binary-compatible
24731 with code compiled without that switch.
24733 @item -malign-data=@var{type}
24734 @opindex malign-data
24735 Control how GCC aligns variables. Supported values for @var{type} are
24736 @samp{compat} uses increased alignment value compatible uses GCC 4.8
24737 and earlier, @samp{abi} uses alignment value as specified by the
24738 psABI, and @samp{cacheline} uses increased alignment value to match
24739 the cache line size. @samp{compat} is the default.
24741 @item -mlarge-data-threshold=@var{threshold}
24742 @opindex mlarge-data-threshold
24743 When @option{-mcmodel=medium} is specified, data objects larger than
24744 @var{threshold} are placed in the large data section. This value must be the
24745 same across all objects linked into the binary, and defaults to 65535.
24749 Use a different function-calling convention, in which functions that
24750 take a fixed number of arguments return with the @code{ret @var{num}}
24751 instruction, which pops their arguments while returning. This saves one
24752 instruction in the caller since there is no need to pop the arguments
24755 You can specify that an individual function is called with this calling
24756 sequence with the function attribute @code{stdcall}. You can also
24757 override the @option{-mrtd} option by using the function attribute
24758 @code{cdecl}. @xref{Function Attributes}.
24760 @strong{Warning:} this calling convention is incompatible with the one
24761 normally used on Unix, so you cannot use it if you need to call
24762 libraries compiled with the Unix compiler.
24764 Also, you must provide function prototypes for all functions that
24765 take variable numbers of arguments (including @code{printf});
24766 otherwise incorrect code is generated for calls to those
24769 In addition, seriously incorrect code results if you call a
24770 function with too many arguments. (Normally, extra arguments are
24771 harmlessly ignored.)
24773 @item -mregparm=@var{num}
24775 Control how many registers are used to pass integer arguments. By
24776 default, no registers are used to pass arguments, and at most 3
24777 registers can be used. You can control this behavior for a specific
24778 function by using the function attribute @code{regparm}.
24779 @xref{Function Attributes}.
24781 @strong{Warning:} if you use this switch, and
24782 @var{num} is nonzero, then you must build all modules with the same
24783 value, including any libraries. This includes the system libraries and
24787 @opindex msseregparm
24788 Use SSE register passing conventions for float and double arguments
24789 and return values. You can control this behavior for a specific
24790 function by using the function attribute @code{sseregparm}.
24791 @xref{Function Attributes}.
24793 @strong{Warning:} if you use this switch then you must build all
24794 modules with the same value, including any libraries. This includes
24795 the system libraries and startup modules.
24797 @item -mvect8-ret-in-mem
24798 @opindex mvect8-ret-in-mem
24799 Return 8-byte vectors in memory instead of MMX registers. This is the
24800 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
24801 Studio compilers until version 12. Later compiler versions (starting
24802 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
24803 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
24804 you need to remain compatible with existing code produced by those
24805 previous compiler versions or older versions of GCC@.
24814 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
24815 is specified, the significands of results of floating-point operations are
24816 rounded to 24 bits (single precision); @option{-mpc64} rounds the
24817 significands of results of floating-point operations to 53 bits (double
24818 precision) and @option{-mpc80} rounds the significands of results of
24819 floating-point operations to 64 bits (extended double precision), which is
24820 the default. When this option is used, floating-point operations in higher
24821 precisions are not available to the programmer without setting the FPU
24822 control word explicitly.
24824 Setting the rounding of floating-point operations to less than the default
24825 80 bits can speed some programs by 2% or more. Note that some mathematical
24826 libraries assume that extended-precision (80-bit) floating-point operations
24827 are enabled by default; routines in such libraries could suffer significant
24828 loss of accuracy, typically through so-called ``catastrophic cancellation'',
24829 when this option is used to set the precision to less than extended precision.
24831 @item -mstackrealign
24832 @opindex mstackrealign
24833 Realign the stack at entry. On the x86, the @option{-mstackrealign}
24834 option generates an alternate prologue and epilogue that realigns the
24835 run-time stack if necessary. This supports mixing legacy codes that keep
24836 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
24837 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
24838 applicable to individual functions.
24840 @item -mpreferred-stack-boundary=@var{num}
24841 @opindex mpreferred-stack-boundary
24842 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
24843 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
24844 the default is 4 (16 bytes or 128 bits).
24846 @strong{Warning:} When generating code for the x86-64 architecture with
24847 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
24848 used to keep the stack boundary aligned to 8 byte boundary. Since
24849 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
24850 intended to be used in controlled environment where stack space is
24851 important limitation. This option leads to wrong code when functions
24852 compiled with 16 byte stack alignment (such as functions from a standard
24853 library) are called with misaligned stack. In this case, SSE
24854 instructions may lead to misaligned memory access traps. In addition,
24855 variable arguments are handled incorrectly for 16 byte aligned
24856 objects (including x87 long double and __int128), leading to wrong
24857 results. You must build all modules with
24858 @option{-mpreferred-stack-boundary=3}, including any libraries. This
24859 includes the system libraries and startup modules.
24861 @item -mincoming-stack-boundary=@var{num}
24862 @opindex mincoming-stack-boundary
24863 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
24864 boundary. If @option{-mincoming-stack-boundary} is not specified,
24865 the one specified by @option{-mpreferred-stack-boundary} is used.
24867 On Pentium and Pentium Pro, @code{double} and @code{long double} values
24868 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
24869 suffer significant run time performance penalties. On Pentium III, the
24870 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
24871 properly if it is not 16-byte aligned.
24873 To ensure proper alignment of this values on the stack, the stack boundary
24874 must be as aligned as that required by any value stored on the stack.
24875 Further, every function must be generated such that it keeps the stack
24876 aligned. Thus calling a function compiled with a higher preferred
24877 stack boundary from a function compiled with a lower preferred stack
24878 boundary most likely misaligns the stack. It is recommended that
24879 libraries that use callbacks always use the default setting.
24881 This extra alignment does consume extra stack space, and generally
24882 increases code size. Code that is sensitive to stack space usage, such
24883 as embedded systems and operating system kernels, may want to reduce the
24884 preferred alignment to @option{-mpreferred-stack-boundary=2}.
24941 @itemx -mavx512ifma
24942 @opindex mavx512ifma
24944 @itemx -mavx512vbmi
24945 @opindex mavx512vbmi
24957 @opindex mclfushopt
24974 @itemx -mprefetchwt1
24975 @opindex mprefetchwt1
25031 These switches enable the use of instructions in the MMX, SSE,
25032 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
25033 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
25034 AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA AVX512VBMI, BMI, BMI2, FXSR,
25035 XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX, PKU or 3DNow!@:
25036 extended instruction sets. Each has a corresponding @option{-mno-} option
25037 to disable use of these instructions.
25039 These extensions are also available as built-in functions: see
25040 @ref{x86 Built-in Functions}, for details of the functions enabled and
25041 disabled by these switches.
25043 To generate SSE/SSE2 instructions automatically from floating-point
25044 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
25046 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
25047 generates new AVX instructions or AVX equivalence for all SSEx instructions
25050 These options enable GCC to use these extended instructions in
25051 generated code, even without @option{-mfpmath=sse}. Applications that
25052 perform run-time CPU detection must compile separate files for each
25053 supported architecture, using the appropriate flags. In particular,
25054 the file containing the CPU detection code should be compiled without
25057 @item -mdump-tune-features
25058 @opindex mdump-tune-features
25059 This option instructs GCC to dump the names of the x86 performance
25060 tuning features and default settings. The names can be used in
25061 @option{-mtune-ctrl=@var{feature-list}}.
25063 @item -mtune-ctrl=@var{feature-list}
25064 @opindex mtune-ctrl=@var{feature-list}
25065 This option is used to do fine grain control of x86 code generation features.
25066 @var{feature-list} is a comma separated list of @var{feature} names. See also
25067 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
25068 on if it is not preceded with @samp{^}, otherwise, it is turned off.
25069 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
25070 developers. Using it may lead to code paths not covered by testing and can
25071 potentially result in compiler ICEs or runtime errors.
25074 @opindex mno-default
25075 This option instructs GCC to turn off all tunable features. See also
25076 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
25080 This option instructs GCC to emit a @code{cld} instruction in the prologue
25081 of functions that use string instructions. String instructions depend on
25082 the DF flag to select between autoincrement or autodecrement mode. While the
25083 ABI specifies the DF flag to be cleared on function entry, some operating
25084 systems violate this specification by not clearing the DF flag in their
25085 exception dispatchers. The exception handler can be invoked with the DF flag
25086 set, which leads to wrong direction mode when string instructions are used.
25087 This option can be enabled by default on 32-bit x86 targets by configuring
25088 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
25089 instructions can be suppressed with the @option{-mno-cld} compiler option
25093 @opindex mvzeroupper
25094 This option instructs GCC to emit a @code{vzeroupper} instruction
25095 before a transfer of control flow out of the function to minimize
25096 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
25099 @item -mprefer-avx128
25100 @opindex mprefer-avx128
25101 This option instructs GCC to use 128-bit AVX instructions instead of
25102 256-bit AVX instructions in the auto-vectorizer.
25106 This option enables GCC to generate @code{CMPXCHG16B} instructions.
25107 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
25108 (or oword) data types.
25109 This is useful for high-resolution counters that can be updated
25110 by multiple processors (or cores). This instruction is generated as part of
25111 atomic built-in functions: see @ref{__sync Builtins} or
25112 @ref{__atomic Builtins} for details.
25116 This option enables generation of @code{SAHF} instructions in 64-bit code.
25117 Early Intel Pentium 4 CPUs with Intel 64 support,
25118 prior to the introduction of Pentium 4 G1 step in December 2005,
25119 lacked the @code{LAHF} and @code{SAHF} instructions
25120 which are supported by AMD64.
25121 These are load and store instructions, respectively, for certain status flags.
25122 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
25123 @code{drem}, and @code{remainder} built-in functions;
25124 see @ref{Other Builtins} for details.
25128 This option enables use of the @code{movbe} instruction to implement
25129 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
25133 This option enables built-in functions @code{__builtin_ia32_crc32qi},
25134 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
25135 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
25139 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
25140 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
25141 with an additional Newton-Raphson step
25142 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
25143 (and their vectorized
25144 variants) for single-precision floating-point arguments. These instructions
25145 are generated only when @option{-funsafe-math-optimizations} is enabled
25146 together with @option{-ffinite-math-only} and @option{-fno-trapping-math}.
25147 Note that while the throughput of the sequence is higher than the throughput
25148 of the non-reciprocal instruction, the precision of the sequence can be
25149 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
25151 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
25152 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
25153 combination), and doesn't need @option{-mrecip}.
25155 Also note that GCC emits the above sequence with additional Newton-Raphson step
25156 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
25157 already with @option{-ffast-math} (or the above option combination), and
25158 doesn't need @option{-mrecip}.
25160 @item -mrecip=@var{opt}
25161 @opindex mrecip=opt
25162 This option controls which reciprocal estimate instructions
25163 may be used. @var{opt} is a comma-separated list of options, which may
25164 be preceded by a @samp{!} to invert the option:
25168 Enable all estimate instructions.
25171 Enable the default instructions, equivalent to @option{-mrecip}.
25174 Disable all estimate instructions, equivalent to @option{-mno-recip}.
25177 Enable the approximation for scalar division.
25180 Enable the approximation for vectorized division.
25183 Enable the approximation for scalar square root.
25186 Enable the approximation for vectorized square root.
25189 So, for example, @option{-mrecip=all,!sqrt} enables
25190 all of the reciprocal approximations, except for square root.
25192 @item -mveclibabi=@var{type}
25193 @opindex mveclibabi
25194 Specifies the ABI type to use for vectorizing intrinsics using an
25195 external library. Supported values for @var{type} are @samp{svml}
25196 for the Intel short
25197 vector math library and @samp{acml} for the AMD math core library.
25198 To use this option, both @option{-ftree-vectorize} and
25199 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
25200 ABI-compatible library must be specified at link time.
25202 GCC currently emits calls to @code{vmldExp2},
25203 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
25204 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
25205 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
25206 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
25207 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
25208 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
25209 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
25210 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
25211 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
25212 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
25213 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
25214 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
25215 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
25216 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
25217 when @option{-mveclibabi=acml} is used.
25219 @item -mabi=@var{name}
25221 Generate code for the specified calling convention. Permissible values
25222 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
25223 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
25224 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
25225 You can control this behavior for specific functions by
25226 using the function attributes @code{ms_abi} and @code{sysv_abi}.
25227 @xref{Function Attributes}.
25229 @item -mtls-dialect=@var{type}
25230 @opindex mtls-dialect
25231 Generate code to access thread-local storage using the @samp{gnu} or
25232 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
25233 @samp{gnu2} is more efficient, but it may add compile- and run-time
25234 requirements that cannot be satisfied on all systems.
25237 @itemx -mno-push-args
25238 @opindex mpush-args
25239 @opindex mno-push-args
25240 Use PUSH operations to store outgoing parameters. This method is shorter
25241 and usually equally fast as method using SUB/MOV operations and is enabled
25242 by default. In some cases disabling it may improve performance because of
25243 improved scheduling and reduced dependencies.
25245 @item -maccumulate-outgoing-args
25246 @opindex maccumulate-outgoing-args
25247 If enabled, the maximum amount of space required for outgoing arguments is
25248 computed in the function prologue. This is faster on most modern CPUs
25249 because of reduced dependencies, improved scheduling and reduced stack usage
25250 when the preferred stack boundary is not equal to 2. The drawback is a notable
25251 increase in code size. This switch implies @option{-mno-push-args}.
25255 Support thread-safe exception handling on MinGW. Programs that rely
25256 on thread-safe exception handling must compile and link all code with the
25257 @option{-mthreads} option. When compiling, @option{-mthreads} defines
25258 @option{-D_MT}; when linking, it links in a special thread helper library
25259 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
25261 @item -mms-bitfields
25262 @itemx -mno-ms-bitfields
25263 @opindex mms-bitfields
25264 @opindex mno-ms-bitfields
25266 Enable/disable bit-field layout compatible with the native Microsoft
25269 If @code{packed} is used on a structure, or if bit-fields are used,
25270 it may be that the Microsoft ABI lays out the structure differently
25271 than the way GCC normally does. Particularly when moving packed
25272 data between functions compiled with GCC and the native Microsoft compiler
25273 (either via function call or as data in a file), it may be necessary to access
25276 This option is enabled by default for Microsoft Windows
25277 targets. This behavior can also be controlled locally by use of variable
25278 or type attributes. For more information, see @ref{x86 Variable Attributes}
25279 and @ref{x86 Type Attributes}.
25281 The Microsoft structure layout algorithm is fairly simple with the exception
25282 of the bit-field packing.
25283 The padding and alignment of members of structures and whether a bit-field
25284 can straddle a storage-unit boundary are determine by these rules:
25287 @item Structure members are stored sequentially in the order in which they are
25288 declared: the first member has the lowest memory address and the last member
25291 @item Every data object has an alignment requirement. The alignment requirement
25292 for all data except structures, unions, and arrays is either the size of the
25293 object or the current packing size (specified with either the
25294 @code{aligned} attribute or the @code{pack} pragma),
25295 whichever is less. For structures, unions, and arrays,
25296 the alignment requirement is the largest alignment requirement of its members.
25297 Every object is allocated an offset so that:
25300 offset % alignment_requirement == 0
25303 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
25304 unit if the integral types are the same size and if the next bit-field fits
25305 into the current allocation unit without crossing the boundary imposed by the
25306 common alignment requirements of the bit-fields.
25309 MSVC interprets zero-length bit-fields in the following ways:
25312 @item If a zero-length bit-field is inserted between two bit-fields that
25313 are normally coalesced, the bit-fields are not coalesced.
25320 unsigned long bf_1 : 12;
25322 unsigned long bf_2 : 12;
25327 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
25328 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
25330 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
25331 alignment of the zero-length bit-field is greater than the member that follows it,
25332 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
25353 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
25354 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
25355 bit-field does not affect the alignment of @code{bar} or, as a result, the size
25358 Taking this into account, it is important to note the following:
25361 @item If a zero-length bit-field follows a normal bit-field, the type of the
25362 zero-length bit-field may affect the alignment of the structure as whole. For
25363 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
25364 normal bit-field, and is of type short.
25366 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
25367 still affect the alignment of the structure:
25378 Here, @code{t4} takes up 4 bytes.
25381 @item Zero-length bit-fields following non-bit-field members are ignored:
25393 Here, @code{t5} takes up 2 bytes.
25397 @item -mno-align-stringops
25398 @opindex mno-align-stringops
25399 Do not align the destination of inlined string operations. This switch reduces
25400 code size and improves performance in case the destination is already aligned,
25401 but GCC doesn't know about it.
25403 @item -minline-all-stringops
25404 @opindex minline-all-stringops
25405 By default GCC inlines string operations only when the destination is
25406 known to be aligned to least a 4-byte boundary.
25407 This enables more inlining and increases code
25408 size, but may improve performance of code that depends on fast
25409 @code{memcpy}, @code{strlen},
25410 and @code{memset} for short lengths.
25412 @item -minline-stringops-dynamically
25413 @opindex minline-stringops-dynamically
25414 For string operations of unknown size, use run-time checks with
25415 inline code for small blocks and a library call for large blocks.
25417 @item -mstringop-strategy=@var{alg}
25418 @opindex mstringop-strategy=@var{alg}
25419 Override the internal decision heuristic for the particular algorithm to use
25420 for inlining string operations. The allowed values for @var{alg} are:
25426 Expand using i386 @code{rep} prefix of the specified size.
25430 @itemx unrolled_loop
25431 Expand into an inline loop.
25434 Always use a library call.
25437 @item -mmemcpy-strategy=@var{strategy}
25438 @opindex mmemcpy-strategy=@var{strategy}
25439 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
25440 should be inlined and what inline algorithm to use when the expected size
25441 of the copy operation is known. @var{strategy}
25442 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
25443 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
25444 the max byte size with which inline algorithm @var{alg} is allowed. For the last
25445 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
25446 in the list must be specified in increasing order. The minimal byte size for
25447 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
25450 @item -mmemset-strategy=@var{strategy}
25451 @opindex mmemset-strategy=@var{strategy}
25452 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
25453 @code{__builtin_memset} expansion.
25455 @item -momit-leaf-frame-pointer
25456 @opindex momit-leaf-frame-pointer
25457 Don't keep the frame pointer in a register for leaf functions. This
25458 avoids the instructions to save, set up, and restore frame pointers and
25459 makes an extra register available in leaf functions. The option
25460 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
25461 which might make debugging harder.
25463 @item -mtls-direct-seg-refs
25464 @itemx -mno-tls-direct-seg-refs
25465 @opindex mtls-direct-seg-refs
25466 Controls whether TLS variables may be accessed with offsets from the
25467 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
25468 or whether the thread base pointer must be added. Whether or not this
25469 is valid depends on the operating system, and whether it maps the
25470 segment to cover the entire TLS area.
25472 For systems that use the GNU C Library, the default is on.
25475 @itemx -mno-sse2avx
25477 Specify that the assembler should encode SSE instructions with VEX
25478 prefix. The option @option{-mavx} turns this on by default.
25483 If profiling is active (@option{-pg}), put the profiling
25484 counter call before the prologue.
25485 Note: On x86 architectures the attribute @code{ms_hook_prologue}
25486 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
25488 @item -mrecord-mcount
25489 @itemx -mno-record-mcount
25490 @opindex mrecord-mcount
25491 If profiling is active (@option{-pg}), generate a __mcount_loc section
25492 that contains pointers to each profiling call. This is useful for
25493 automatically patching and out calls.
25496 @itemx -mno-nop-mcount
25497 @opindex mnop-mcount
25498 If profiling is active (@option{-pg}), generate the calls to
25499 the profiling functions as NOPs. This is useful when they
25500 should be patched in later dynamically. This is likely only
25501 useful together with @option{-mrecord-mcount}.
25503 @item -mskip-rax-setup
25504 @itemx -mno-skip-rax-setup
25505 @opindex mskip-rax-setup
25506 When generating code for the x86-64 architecture with SSE extensions
25507 disabled, @option{-mskip-rax-setup} can be used to skip setting up RAX
25508 register when there are no variable arguments passed in vector registers.
25510 @strong{Warning:} Since RAX register is used to avoid unnecessarily
25511 saving vector registers on stack when passing variable arguments, the
25512 impacts of this option are callees may waste some stack space,
25513 misbehave or jump to a random location. GCC 4.4 or newer don't have
25514 those issues, regardless the RAX register value.
25517 @itemx -mno-8bit-idiv
25518 @opindex m8bit-idiv
25519 On some processors, like Intel Atom, 8-bit unsigned integer divide is
25520 much faster than 32-bit/64-bit integer divide. This option generates a
25521 run-time check. If both dividend and divisor are within range of 0
25522 to 255, 8-bit unsigned integer divide is used instead of
25523 32-bit/64-bit integer divide.
25525 @item -mavx256-split-unaligned-load
25526 @itemx -mavx256-split-unaligned-store
25527 @opindex mavx256-split-unaligned-load
25528 @opindex mavx256-split-unaligned-store
25529 Split 32-byte AVX unaligned load and store.
25531 @item -mstack-protector-guard=@var{guard}
25532 @opindex mstack-protector-guard=@var{guard}
25533 Generate stack protection code using canary at @var{guard}. Supported
25534 locations are @samp{global} for global canary or @samp{tls} for per-thread
25535 canary in the TLS block (the default). This option has effect only when
25536 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
25538 @item -mmitigate-rop
25539 @opindex mmitigate-rop
25540 Try to avoid generating code sequences that contain unintended return
25541 opcodes, to mitigate against certain forms of attack. At the moment,
25542 this option is limited in what it can do and should not be relied
25543 on to provide serious protection.
25545 @item -mgeneral-regs-only
25546 @opindex mgeneral-regs-only
25547 Generate code that uses only the general-purpose registers. This
25548 prevents the compiler from using floating-point, vector, mask and bound
25553 These @samp{-m} switches are supported in addition to the above
25554 on x86-64 processors in 64-bit environments.
25567 Generate code for a 16-bit, 32-bit or 64-bit environment.
25568 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
25570 generates code that runs on any i386 system.
25572 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
25573 types to 64 bits, and generates code for the x86-64 architecture.
25574 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
25575 and @option{-mdynamic-no-pic} options.
25577 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
25579 generates code for the x86-64 architecture.
25581 The @option{-m16} option is the same as @option{-m32}, except for that
25582 it outputs the @code{.code16gcc} assembly directive at the beginning of
25583 the assembly output so that the binary can run in 16-bit mode.
25585 The @option{-miamcu} option generates code which conforms to Intel MCU
25586 psABI. It requires the @option{-m32} option to be turned on.
25588 @item -mno-red-zone
25589 @opindex mno-red-zone
25590 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
25591 by the x86-64 ABI; it is a 128-byte area beyond the location of the
25592 stack pointer that is not modified by signal or interrupt handlers
25593 and therefore can be used for temporary data without adjusting the stack
25594 pointer. The flag @option{-mno-red-zone} disables this red zone.
25596 @item -mcmodel=small
25597 @opindex mcmodel=small
25598 Generate code for the small code model: the program and its symbols must
25599 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
25600 Programs can be statically or dynamically linked. This is the default
25603 @item -mcmodel=kernel
25604 @opindex mcmodel=kernel
25605 Generate code for the kernel code model. The kernel runs in the
25606 negative 2 GB of the address space.
25607 This model has to be used for Linux kernel code.
25609 @item -mcmodel=medium
25610 @opindex mcmodel=medium
25611 Generate code for the medium model: the program is linked in the lower 2
25612 GB of the address space. Small symbols are also placed there. Symbols
25613 with sizes larger than @option{-mlarge-data-threshold} are put into
25614 large data or BSS sections and can be located above 2GB. Programs can
25615 be statically or dynamically linked.
25617 @item -mcmodel=large
25618 @opindex mcmodel=large
25619 Generate code for the large model. This model makes no assumptions
25620 about addresses and sizes of sections.
25622 @item -maddress-mode=long
25623 @opindex maddress-mode=long
25624 Generate code for long address mode. This is only supported for 64-bit
25625 and x32 environments. It is the default address mode for 64-bit
25628 @item -maddress-mode=short
25629 @opindex maddress-mode=short
25630 Generate code for short address mode. This is only supported for 32-bit
25631 and x32 environments. It is the default address mode for 32-bit and
25635 @node x86 Windows Options
25636 @subsection x86 Windows Options
25637 @cindex x86 Windows Options
25638 @cindex Windows Options for x86
25640 These additional options are available for Microsoft Windows targets:
25646 specifies that a console application is to be generated, by
25647 instructing the linker to set the PE header subsystem type
25648 required for console applications.
25649 This option is available for Cygwin and MinGW targets and is
25650 enabled by default on those targets.
25654 This option is available for Cygwin and MinGW targets. It
25655 specifies that a DLL---a dynamic link library---is to be
25656 generated, enabling the selection of the required runtime
25657 startup object and entry point.
25659 @item -mnop-fun-dllimport
25660 @opindex mnop-fun-dllimport
25661 This option is available for Cygwin and MinGW targets. It
25662 specifies that the @code{dllimport} attribute should be ignored.
25666 This option is available for MinGW targets. It specifies
25667 that MinGW-specific thread support is to be used.
25671 This option is available for MinGW-w64 targets. It causes
25672 the @code{UNICODE} preprocessor macro to be predefined, and
25673 chooses Unicode-capable runtime startup code.
25677 This option is available for Cygwin and MinGW targets. It
25678 specifies that the typical Microsoft Windows predefined macros are to
25679 be set in the pre-processor, but does not influence the choice
25680 of runtime library/startup code.
25684 This option is available for Cygwin and MinGW targets. It
25685 specifies that a GUI application is to be generated by
25686 instructing the linker to set the PE header subsystem type
25689 @item -fno-set-stack-executable
25690 @opindex fno-set-stack-executable
25691 This option is available for MinGW targets. It specifies that
25692 the executable flag for the stack used by nested functions isn't
25693 set. This is necessary for binaries running in kernel mode of
25694 Microsoft Windows, as there the User32 API, which is used to set executable
25695 privileges, isn't available.
25697 @item -fwritable-relocated-rdata
25698 @opindex fno-writable-relocated-rdata
25699 This option is available for MinGW and Cygwin targets. It specifies
25700 that relocated-data in read-only section is put into the @code{.data}
25701 section. This is a necessary for older runtimes not supporting
25702 modification of @code{.rdata} sections for pseudo-relocation.
25704 @item -mpe-aligned-commons
25705 @opindex mpe-aligned-commons
25706 This option is available for Cygwin and MinGW targets. It
25707 specifies that the GNU extension to the PE file format that
25708 permits the correct alignment of COMMON variables should be
25709 used when generating code. It is enabled by default if
25710 GCC detects that the target assembler found during configuration
25711 supports the feature.
25714 See also under @ref{x86 Options} for standard options.
25716 @node Xstormy16 Options
25717 @subsection Xstormy16 Options
25718 @cindex Xstormy16 Options
25720 These options are defined for Xstormy16:
25725 Choose startup files and linker script suitable for the simulator.
25728 @node Xtensa Options
25729 @subsection Xtensa Options
25730 @cindex Xtensa Options
25732 These options are supported for Xtensa targets:
25736 @itemx -mno-const16
25738 @opindex mno-const16
25739 Enable or disable use of @code{CONST16} instructions for loading
25740 constant values. The @code{CONST16} instruction is currently not a
25741 standard option from Tensilica. When enabled, @code{CONST16}
25742 instructions are always used in place of the standard @code{L32R}
25743 instructions. The use of @code{CONST16} is enabled by default only if
25744 the @code{L32R} instruction is not available.
25747 @itemx -mno-fused-madd
25748 @opindex mfused-madd
25749 @opindex mno-fused-madd
25750 Enable or disable use of fused multiply/add and multiply/subtract
25751 instructions in the floating-point option. This has no effect if the
25752 floating-point option is not also enabled. Disabling fused multiply/add
25753 and multiply/subtract instructions forces the compiler to use separate
25754 instructions for the multiply and add/subtract operations. This may be
25755 desirable in some cases where strict IEEE 754-compliant results are
25756 required: the fused multiply add/subtract instructions do not round the
25757 intermediate result, thereby producing results with @emph{more} bits of
25758 precision than specified by the IEEE standard. Disabling fused multiply
25759 add/subtract instructions also ensures that the program output is not
25760 sensitive to the compiler's ability to combine multiply and add/subtract
25763 @item -mserialize-volatile
25764 @itemx -mno-serialize-volatile
25765 @opindex mserialize-volatile
25766 @opindex mno-serialize-volatile
25767 When this option is enabled, GCC inserts @code{MEMW} instructions before
25768 @code{volatile} memory references to guarantee sequential consistency.
25769 The default is @option{-mserialize-volatile}. Use
25770 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
25772 @item -mforce-no-pic
25773 @opindex mforce-no-pic
25774 For targets, like GNU/Linux, where all user-mode Xtensa code must be
25775 position-independent code (PIC), this option disables PIC for compiling
25778 @item -mtext-section-literals
25779 @itemx -mno-text-section-literals
25780 @opindex mtext-section-literals
25781 @opindex mno-text-section-literals
25782 These options control the treatment of literal pools. The default is
25783 @option{-mno-text-section-literals}, which places literals in a separate
25784 section in the output file. This allows the literal pool to be placed
25785 in a data RAM/ROM, and it also allows the linker to combine literal
25786 pools from separate object files to remove redundant literals and
25787 improve code size. With @option{-mtext-section-literals}, the literals
25788 are interspersed in the text section in order to keep them as close as
25789 possible to their references. This may be necessary for large assembly
25790 files. Literals for each function are placed right before that function.
25792 @item -mauto-litpools
25793 @itemx -mno-auto-litpools
25794 @opindex mauto-litpools
25795 @opindex mno-auto-litpools
25796 These options control the treatment of literal pools. The default is
25797 @option{-mno-auto-litpools}, which places literals in a separate
25798 section in the output file unless @option{-mtext-section-literals} is
25799 used. With @option{-mauto-litpools} the literals are interspersed in
25800 the text section by the assembler. Compiler does not produce explicit
25801 @code{.literal} directives and loads literals into registers with
25802 @code{MOVI} instructions instead of @code{L32R} to let the assembler
25803 do relaxation and place literals as necessary. This option allows
25804 assembler to create several literal pools per function and assemble
25805 very big functions, which may not be possible with
25806 @option{-mtext-section-literals}.
25808 @item -mtarget-align
25809 @itemx -mno-target-align
25810 @opindex mtarget-align
25811 @opindex mno-target-align
25812 When this option is enabled, GCC instructs the assembler to
25813 automatically align instructions to reduce branch penalties at the
25814 expense of some code density. The assembler attempts to widen density
25815 instructions to align branch targets and the instructions following call
25816 instructions. If there are not enough preceding safe density
25817 instructions to align a target, no widening is performed. The
25818 default is @option{-mtarget-align}. These options do not affect the
25819 treatment of auto-aligned instructions like @code{LOOP}, which the
25820 assembler always aligns, either by widening density instructions or
25821 by inserting NOP instructions.
25824 @itemx -mno-longcalls
25825 @opindex mlongcalls
25826 @opindex mno-longcalls
25827 When this option is enabled, GCC instructs the assembler to translate
25828 direct calls to indirect calls unless it can determine that the target
25829 of a direct call is in the range allowed by the call instruction. This
25830 translation typically occurs for calls to functions in other source
25831 files. Specifically, the assembler translates a direct @code{CALL}
25832 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
25833 The default is @option{-mno-longcalls}. This option should be used in
25834 programs where the call target can potentially be out of range. This
25835 option is implemented in the assembler, not the compiler, so the
25836 assembly code generated by GCC still shows direct call
25837 instructions---look at the disassembled object code to see the actual
25838 instructions. Note that the assembler uses an indirect call for
25839 every cross-file call, not just those that really are out of range.
25842 @node zSeries Options
25843 @subsection zSeries Options
25844 @cindex zSeries options
25846 These are listed under @xref{S/390 and zSeries Options}.
25852 @section Specifying Subprocesses and the Switches to Pass to Them
25855 @command{gcc} is a driver program. It performs its job by invoking a
25856 sequence of other programs to do the work of compiling, assembling and
25857 linking. GCC interprets its command-line parameters and uses these to
25858 deduce which programs it should invoke, and which command-line options
25859 it ought to place on their command lines. This behavior is controlled
25860 by @dfn{spec strings}. In most cases there is one spec string for each
25861 program that GCC can invoke, but a few programs have multiple spec
25862 strings to control their behavior. The spec strings built into GCC can
25863 be overridden by using the @option{-specs=} command-line switch to specify
25866 @dfn{Spec files} are plain-text files that are used to construct spec
25867 strings. They consist of a sequence of directives separated by blank
25868 lines. The type of directive is determined by the first non-whitespace
25869 character on the line, which can be one of the following:
25872 @item %@var{command}
25873 Issues a @var{command} to the spec file processor. The commands that can
25877 @item %include <@var{file}>
25878 @cindex @code{%include}
25879 Search for @var{file} and insert its text at the current point in the
25882 @item %include_noerr <@var{file}>
25883 @cindex @code{%include_noerr}
25884 Just like @samp{%include}, but do not generate an error message if the include
25885 file cannot be found.
25887 @item %rename @var{old_name} @var{new_name}
25888 @cindex @code{%rename}
25889 Rename the spec string @var{old_name} to @var{new_name}.
25893 @item *[@var{spec_name}]:
25894 This tells the compiler to create, override or delete the named spec
25895 string. All lines after this directive up to the next directive or
25896 blank line are considered to be the text for the spec string. If this
25897 results in an empty string then the spec is deleted. (Or, if the
25898 spec did not exist, then nothing happens.) Otherwise, if the spec
25899 does not currently exist a new spec is created. If the spec does
25900 exist then its contents are overridden by the text of this
25901 directive, unless the first character of that text is the @samp{+}
25902 character, in which case the text is appended to the spec.
25904 @item [@var{suffix}]:
25905 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
25906 and up to the next directive or blank line are considered to make up the
25907 spec string for the indicated suffix. When the compiler encounters an
25908 input file with the named suffix, it processes the spec string in
25909 order to work out how to compile that file. For example:
25913 z-compile -input %i
25916 This says that any input file whose name ends in @samp{.ZZ} should be
25917 passed to the program @samp{z-compile}, which should be invoked with the
25918 command-line switch @option{-input} and with the result of performing the
25919 @samp{%i} substitution. (See below.)
25921 As an alternative to providing a spec string, the text following a
25922 suffix directive can be one of the following:
25925 @item @@@var{language}
25926 This says that the suffix is an alias for a known @var{language}. This is
25927 similar to using the @option{-x} command-line switch to GCC to specify a
25928 language explicitly. For example:
25935 Says that .ZZ files are, in fact, C++ source files.
25938 This causes an error messages saying:
25941 @var{name} compiler not installed on this system.
25945 GCC already has an extensive list of suffixes built into it.
25946 This directive adds an entry to the end of the list of suffixes, but
25947 since the list is searched from the end backwards, it is effectively
25948 possible to override earlier entries using this technique.
25952 GCC has the following spec strings built into it. Spec files can
25953 override these strings or create their own. Note that individual
25954 targets can also add their own spec strings to this list.
25957 asm Options to pass to the assembler
25958 asm_final Options to pass to the assembler post-processor
25959 cpp Options to pass to the C preprocessor
25960 cc1 Options to pass to the C compiler
25961 cc1plus Options to pass to the C++ compiler
25962 endfile Object files to include at the end of the link
25963 link Options to pass to the linker
25964 lib Libraries to include on the command line to the linker
25965 libgcc Decides which GCC support library to pass to the linker
25966 linker Sets the name of the linker
25967 predefines Defines to be passed to the C preprocessor
25968 signed_char Defines to pass to CPP to say whether @code{char} is signed
25970 startfile Object files to include at the start of the link
25973 Here is a small example of a spec file:
25976 %rename lib old_lib
25979 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
25982 This example renames the spec called @samp{lib} to @samp{old_lib} and
25983 then overrides the previous definition of @samp{lib} with a new one.
25984 The new definition adds in some extra command-line options before
25985 including the text of the old definition.
25987 @dfn{Spec strings} are a list of command-line options to be passed to their
25988 corresponding program. In addition, the spec strings can contain
25989 @samp{%}-prefixed sequences to substitute variable text or to
25990 conditionally insert text into the command line. Using these constructs
25991 it is possible to generate quite complex command lines.
25993 Here is a table of all defined @samp{%}-sequences for spec
25994 strings. Note that spaces are not generated automatically around the
25995 results of expanding these sequences. Therefore you can concatenate them
25996 together or combine them with constant text in a single argument.
26000 Substitute one @samp{%} into the program name or argument.
26003 Substitute the name of the input file being processed.
26006 Substitute the basename of the input file being processed.
26007 This is the substring up to (and not including) the last period
26008 and not including the directory.
26011 This is the same as @samp{%b}, but include the file suffix (text after
26015 Marks the argument containing or following the @samp{%d} as a
26016 temporary file name, so that that file is deleted if GCC exits
26017 successfully. Unlike @samp{%g}, this contributes no text to the
26020 @item %g@var{suffix}
26021 Substitute a file name that has suffix @var{suffix} and is chosen
26022 once per compilation, and mark the argument in the same way as
26023 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
26024 name is now chosen in a way that is hard to predict even when previously
26025 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
26026 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
26027 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
26028 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
26029 was simply substituted with a file name chosen once per compilation,
26030 without regard to any appended suffix (which was therefore treated
26031 just like ordinary text), making such attacks more likely to succeed.
26033 @item %u@var{suffix}
26034 Like @samp{%g}, but generates a new temporary file name
26035 each time it appears instead of once per compilation.
26037 @item %U@var{suffix}
26038 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
26039 new one if there is no such last file name. In the absence of any
26040 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
26041 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
26042 involves the generation of two distinct file names, one
26043 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
26044 simply substituted with a file name chosen for the previous @samp{%u},
26045 without regard to any appended suffix.
26047 @item %j@var{suffix}
26048 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
26049 writable, and if @option{-save-temps} is not used;
26050 otherwise, substitute the name
26051 of a temporary file, just like @samp{%u}. This temporary file is not
26052 meant for communication between processes, but rather as a junk
26053 disposal mechanism.
26055 @item %|@var{suffix}
26056 @itemx %m@var{suffix}
26057 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
26058 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
26059 all. These are the two most common ways to instruct a program that it
26060 should read from standard input or write to standard output. If you
26061 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
26062 construct: see for example @file{f/lang-specs.h}.
26064 @item %.@var{SUFFIX}
26065 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
26066 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
26067 terminated by the next space or %.
26070 Marks the argument containing or following the @samp{%w} as the
26071 designated output file of this compilation. This puts the argument
26072 into the sequence of arguments that @samp{%o} substitutes.
26075 Substitutes the names of all the output files, with spaces
26076 automatically placed around them. You should write spaces
26077 around the @samp{%o} as well or the results are undefined.
26078 @samp{%o} is for use in the specs for running the linker.
26079 Input files whose names have no recognized suffix are not compiled
26080 at all, but they are included among the output files, so they are
26084 Substitutes the suffix for object files. Note that this is
26085 handled specially when it immediately follows @samp{%g, %u, or %U},
26086 because of the need for those to form complete file names. The
26087 handling is such that @samp{%O} is treated exactly as if it had already
26088 been substituted, except that @samp{%g, %u, and %U} do not currently
26089 support additional @var{suffix} characters following @samp{%O} as they do
26090 following, for example, @samp{.o}.
26093 Substitutes the standard macro predefinitions for the
26094 current target machine. Use this when running @command{cpp}.
26097 Like @samp{%p}, but puts @samp{__} before and after the name of each
26098 predefined macro, except for macros that start with @samp{__} or with
26099 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
26103 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
26104 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
26105 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
26106 and @option{-imultilib} as necessary.
26109 Current argument is the name of a library or startup file of some sort.
26110 Search for that file in a standard list of directories and substitute
26111 the full name found. The current working directory is included in the
26112 list of directories scanned.
26115 Current argument is the name of a linker script. Search for that file
26116 in the current list of directories to scan for libraries. If the file
26117 is located insert a @option{--script} option into the command line
26118 followed by the full path name found. If the file is not found then
26119 generate an error message. Note: the current working directory is not
26123 Print @var{str} as an error message. @var{str} is terminated by a newline.
26124 Use this when inconsistent options are detected.
26126 @item %(@var{name})
26127 Substitute the contents of spec string @var{name} at this point.
26129 @item %x@{@var{option}@}
26130 Accumulate an option for @samp{%X}.
26133 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
26137 Output the accumulated assembler options specified by @option{-Wa}.
26140 Output the accumulated preprocessor options specified by @option{-Wp}.
26143 Process the @code{asm} spec. This is used to compute the
26144 switches to be passed to the assembler.
26147 Process the @code{asm_final} spec. This is a spec string for
26148 passing switches to an assembler post-processor, if such a program is
26152 Process the @code{link} spec. This is the spec for computing the
26153 command line passed to the linker. Typically it makes use of the
26154 @samp{%L %G %S %D and %E} sequences.
26157 Dump out a @option{-L} option for each directory that GCC believes might
26158 contain startup files. If the target supports multilibs then the
26159 current multilib directory is prepended to each of these paths.
26162 Process the @code{lib} spec. This is a spec string for deciding which
26163 libraries are included on the command line to the linker.
26166 Process the @code{libgcc} spec. This is a spec string for deciding
26167 which GCC support library is included on the command line to the linker.
26170 Process the @code{startfile} spec. This is a spec for deciding which
26171 object files are the first ones passed to the linker. Typically
26172 this might be a file named @file{crt0.o}.
26175 Process the @code{endfile} spec. This is a spec string that specifies
26176 the last object files that are passed to the linker.
26179 Process the @code{cpp} spec. This is used to construct the arguments
26180 to be passed to the C preprocessor.
26183 Process the @code{cc1} spec. This is used to construct the options to be
26184 passed to the actual C compiler (@command{cc1}).
26187 Process the @code{cc1plus} spec. This is used to construct the options to be
26188 passed to the actual C++ compiler (@command{cc1plus}).
26191 Substitute the variable part of a matched option. See below.
26192 Note that each comma in the substituted string is replaced by
26196 Remove all occurrences of @code{-S} from the command line. Note---this
26197 command is position dependent. @samp{%} commands in the spec string
26198 before this one see @code{-S}, @samp{%} commands in the spec string
26199 after this one do not.
26201 @item %:@var{function}(@var{args})
26202 Call the named function @var{function}, passing it @var{args}.
26203 @var{args} is first processed as a nested spec string, then split
26204 into an argument vector in the usual fashion. The function returns
26205 a string which is processed as if it had appeared literally as part
26206 of the current spec.
26208 The following built-in spec functions are provided:
26211 @item @code{getenv}
26212 The @code{getenv} spec function takes two arguments: an environment
26213 variable name and a string. If the environment variable is not
26214 defined, a fatal error is issued. Otherwise, the return value is the
26215 value of the environment variable concatenated with the string. For
26216 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
26219 %:getenv(TOPDIR /include)
26222 expands to @file{/path/to/top/include}.
26224 @item @code{if-exists}
26225 The @code{if-exists} spec function takes one argument, an absolute
26226 pathname to a file. If the file exists, @code{if-exists} returns the
26227 pathname. Here is a small example of its usage:
26231 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
26234 @item @code{if-exists-else}
26235 The @code{if-exists-else} spec function is similar to the @code{if-exists}
26236 spec function, except that it takes two arguments. The first argument is
26237 an absolute pathname to a file. If the file exists, @code{if-exists-else}
26238 returns the pathname. If it does not exist, it returns the second argument.
26239 This way, @code{if-exists-else} can be used to select one file or another,
26240 based on the existence of the first. Here is a small example of its usage:
26244 crt0%O%s %:if-exists(crti%O%s) \
26245 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
26248 @item @code{replace-outfile}
26249 The @code{replace-outfile} spec function takes two arguments. It looks for the
26250 first argument in the outfiles array and replaces it with the second argument. Here
26251 is a small example of its usage:
26254 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
26257 @item @code{remove-outfile}
26258 The @code{remove-outfile} spec function takes one argument. It looks for the
26259 first argument in the outfiles array and removes it. Here is a small example
26263 %:remove-outfile(-lm)
26266 @item @code{pass-through-libs}
26267 The @code{pass-through-libs} spec function takes any number of arguments. It
26268 finds any @option{-l} options and any non-options ending in @file{.a} (which it
26269 assumes are the names of linker input library archive files) and returns a
26270 result containing all the found arguments each prepended by
26271 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
26272 intended to be passed to the LTO linker plugin.
26275 %:pass-through-libs(%G %L %G)
26278 @item @code{print-asm-header}
26279 The @code{print-asm-header} function takes no arguments and simply
26280 prints a banner like:
26286 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
26289 It is used to separate compiler options from assembler options
26290 in the @option{--target-help} output.
26293 @item %@{@code{S}@}
26294 Substitutes the @code{-S} switch, if that switch is given to GCC@.
26295 If that switch is not specified, this substitutes nothing. Note that
26296 the leading dash is omitted when specifying this option, and it is
26297 automatically inserted if the substitution is performed. Thus the spec
26298 string @samp{%@{foo@}} matches the command-line option @option{-foo}
26299 and outputs the command-line option @option{-foo}.
26301 @item %W@{@code{S}@}
26302 Like %@{@code{S}@} but mark last argument supplied within as a file to be
26303 deleted on failure.
26305 @item %@{@code{S}*@}
26306 Substitutes all the switches specified to GCC whose names start
26307 with @code{-S}, but which also take an argument. This is used for
26308 switches like @option{-o}, @option{-D}, @option{-I}, etc.
26309 GCC considers @option{-o foo} as being
26310 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
26311 text, including the space. Thus two arguments are generated.
26313 @item %@{@code{S}*&@code{T}*@}
26314 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
26315 (the order of @code{S} and @code{T} in the spec is not significant).
26316 There can be any number of ampersand-separated variables; for each the
26317 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
26319 @item %@{@code{S}:@code{X}@}
26320 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
26322 @item %@{!@code{S}:@code{X}@}
26323 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
26325 @item %@{@code{S}*:@code{X}@}
26326 Substitutes @code{X} if one or more switches whose names start with
26327 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
26328 once, no matter how many such switches appeared. However, if @code{%*}
26329 appears somewhere in @code{X}, then @code{X} is substituted once
26330 for each matching switch, with the @code{%*} replaced by the part of
26331 that switch matching the @code{*}.
26333 If @code{%*} appears as the last part of a spec sequence then a space
26334 is added after the end of the last substitution. If there is more
26335 text in the sequence, however, then a space is not generated. This
26336 allows the @code{%*} substitution to be used as part of a larger
26337 string. For example, a spec string like this:
26340 %@{mcu=*:--script=%*/memory.ld@}
26344 when matching an option like @option{-mcu=newchip} produces:
26347 --script=newchip/memory.ld
26350 @item %@{.@code{S}:@code{X}@}
26351 Substitutes @code{X}, if processing a file with suffix @code{S}.
26353 @item %@{!.@code{S}:@code{X}@}
26354 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
26356 @item %@{,@code{S}:@code{X}@}
26357 Substitutes @code{X}, if processing a file for language @code{S}.
26359 @item %@{!,@code{S}:@code{X}@}
26360 Substitutes @code{X}, if not processing a file for language @code{S}.
26362 @item %@{@code{S}|@code{P}:@code{X}@}
26363 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
26364 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
26365 @code{*} sequences as well, although they have a stronger binding than
26366 the @samp{|}. If @code{%*} appears in @code{X}, all of the
26367 alternatives must be starred, and only the first matching alternative
26370 For example, a spec string like this:
26373 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
26377 outputs the following command-line options from the following input
26378 command-line options:
26383 -d fred.c -foo -baz -boggle
26384 -d jim.d -bar -baz -boggle
26387 @item %@{S:X; T:Y; :D@}
26389 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
26390 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
26391 be as many clauses as you need. This may be combined with @code{.},
26392 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
26397 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
26398 construct may contain other nested @samp{%} constructs or spaces, or
26399 even newlines. They are processed as usual, as described above.
26400 Trailing white space in @code{X} is ignored. White space may also
26401 appear anywhere on the left side of the colon in these constructs,
26402 except between @code{.} or @code{*} and the corresponding word.
26404 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
26405 handled specifically in these constructs. If another value of
26406 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
26407 @option{-W} switch is found later in the command line, the earlier
26408 switch value is ignored, except with @{@code{S}*@} where @code{S} is
26409 just one letter, which passes all matching options.
26411 The character @samp{|} at the beginning of the predicate text is used to
26412 indicate that a command should be piped to the following command, but
26413 only if @option{-pipe} is specified.
26415 It is built into GCC which switches take arguments and which do not.
26416 (You might think it would be useful to generalize this to allow each
26417 compiler's spec to say which switches take arguments. But this cannot
26418 be done in a consistent fashion. GCC cannot even decide which input
26419 files have been specified without knowing which switches take arguments,
26420 and it must know which input files to compile in order to tell which
26423 GCC also knows implicitly that arguments starting in @option{-l} are to be
26424 treated as compiler output files, and passed to the linker in their
26425 proper position among the other output files.
26427 @node Environment Variables
26428 @section Environment Variables Affecting GCC
26429 @cindex environment variables
26431 @c man begin ENVIRONMENT
26432 This section describes several environment variables that affect how GCC
26433 operates. Some of them work by specifying directories or prefixes to use
26434 when searching for various kinds of files. Some are used to specify other
26435 aspects of the compilation environment.
26437 Note that you can also specify places to search using options such as
26438 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
26439 take precedence over places specified using environment variables, which
26440 in turn take precedence over those specified by the configuration of GCC@.
26441 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
26442 GNU Compiler Collection (GCC) Internals}.
26447 @c @itemx LC_COLLATE
26449 @c @itemx LC_MONETARY
26450 @c @itemx LC_NUMERIC
26455 @c @findex LC_COLLATE
26456 @findex LC_MESSAGES
26457 @c @findex LC_MONETARY
26458 @c @findex LC_NUMERIC
26462 These environment variables control the way that GCC uses
26463 localization information which allows GCC to work with different
26464 national conventions. GCC inspects the locale categories
26465 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
26466 so. These locale categories can be set to any value supported by your
26467 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
26468 Kingdom encoded in UTF-8.
26470 The @env{LC_CTYPE} environment variable specifies character
26471 classification. GCC uses it to determine the character boundaries in
26472 a string; this is needed for some multibyte encodings that contain quote
26473 and escape characters that are otherwise interpreted as a string
26476 The @env{LC_MESSAGES} environment variable specifies the language to
26477 use in diagnostic messages.
26479 If the @env{LC_ALL} environment variable is set, it overrides the value
26480 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
26481 and @env{LC_MESSAGES} default to the value of the @env{LANG}
26482 environment variable. If none of these variables are set, GCC
26483 defaults to traditional C English behavior.
26487 If @env{TMPDIR} is set, it specifies the directory to use for temporary
26488 files. GCC uses temporary files to hold the output of one stage of
26489 compilation which is to be used as input to the next stage: for example,
26490 the output of the preprocessor, which is the input to the compiler
26493 @item GCC_COMPARE_DEBUG
26494 @findex GCC_COMPARE_DEBUG
26495 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
26496 @option{-fcompare-debug} to the compiler driver. See the documentation
26497 of this option for more details.
26499 @item GCC_EXEC_PREFIX
26500 @findex GCC_EXEC_PREFIX
26501 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
26502 names of the subprograms executed by the compiler. No slash is added
26503 when this prefix is combined with the name of a subprogram, but you can
26504 specify a prefix that ends with a slash if you wish.
26506 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
26507 an appropriate prefix to use based on the pathname it is invoked with.
26509 If GCC cannot find the subprogram using the specified prefix, it
26510 tries looking in the usual places for the subprogram.
26512 The default value of @env{GCC_EXEC_PREFIX} is
26513 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
26514 the installed compiler. In many cases @var{prefix} is the value
26515 of @code{prefix} when you ran the @file{configure} script.
26517 Other prefixes specified with @option{-B} take precedence over this prefix.
26519 This prefix is also used for finding files such as @file{crt0.o} that are
26522 In addition, the prefix is used in an unusual way in finding the
26523 directories to search for header files. For each of the standard
26524 directories whose name normally begins with @samp{/usr/local/lib/gcc}
26525 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
26526 replacing that beginning with the specified prefix to produce an
26527 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
26528 @file{foo/bar} just before it searches the standard directory
26529 @file{/usr/local/lib/bar}.
26530 If a standard directory begins with the configured
26531 @var{prefix} then the value of @var{prefix} is replaced by
26532 @env{GCC_EXEC_PREFIX} when looking for header files.
26534 @item COMPILER_PATH
26535 @findex COMPILER_PATH
26536 The value of @env{COMPILER_PATH} is a colon-separated list of
26537 directories, much like @env{PATH}. GCC tries the directories thus
26538 specified when searching for subprograms, if it can't find the
26539 subprograms using @env{GCC_EXEC_PREFIX}.
26542 @findex LIBRARY_PATH
26543 The value of @env{LIBRARY_PATH} is a colon-separated list of
26544 directories, much like @env{PATH}. When configured as a native compiler,
26545 GCC tries the directories thus specified when searching for special
26546 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
26547 using GCC also uses these directories when searching for ordinary
26548 libraries for the @option{-l} option (but directories specified with
26549 @option{-L} come first).
26553 @cindex locale definition
26554 This variable is used to pass locale information to the compiler. One way in
26555 which this information is used is to determine the character set to be used
26556 when character literals, string literals and comments are parsed in C and C++.
26557 When the compiler is configured to allow multibyte characters,
26558 the following values for @env{LANG} are recognized:
26562 Recognize JIS characters.
26564 Recognize SJIS characters.
26566 Recognize EUCJP characters.
26569 If @env{LANG} is not defined, or if it has some other value, then the
26570 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
26571 recognize and translate multibyte characters.
26575 Some additional environment variables affect the behavior of the
26578 @include cppenv.texi
26582 @node Precompiled Headers
26583 @section Using Precompiled Headers
26584 @cindex precompiled headers
26585 @cindex speed of compilation
26587 Often large projects have many header files that are included in every
26588 source file. The time the compiler takes to process these header files
26589 over and over again can account for nearly all of the time required to
26590 build the project. To make builds faster, GCC allows you to
26591 @dfn{precompile} a header file.
26593 To create a precompiled header file, simply compile it as you would any
26594 other file, if necessary using the @option{-x} option to make the driver
26595 treat it as a C or C++ header file. You may want to use a
26596 tool like @command{make} to keep the precompiled header up-to-date when
26597 the headers it contains change.
26599 A precompiled header file is searched for when @code{#include} is
26600 seen in the compilation. As it searches for the included file
26601 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
26602 compiler looks for a precompiled header in each directory just before it
26603 looks for the include file in that directory. The name searched for is
26604 the name specified in the @code{#include} with @samp{.gch} appended. If
26605 the precompiled header file can't be used, it is ignored.
26607 For instance, if you have @code{#include "all.h"}, and you have
26608 @file{all.h.gch} in the same directory as @file{all.h}, then the
26609 precompiled header file is used if possible, and the original
26610 header is used otherwise.
26612 Alternatively, you might decide to put the precompiled header file in a
26613 directory and use @option{-I} to ensure that directory is searched
26614 before (or instead of) the directory containing the original header.
26615 Then, if you want to check that the precompiled header file is always
26616 used, you can put a file of the same name as the original header in this
26617 directory containing an @code{#error} command.
26619 This also works with @option{-include}. So yet another way to use
26620 precompiled headers, good for projects not designed with precompiled
26621 header files in mind, is to simply take most of the header files used by
26622 a project, include them from another header file, precompile that header
26623 file, and @option{-include} the precompiled header. If the header files
26624 have guards against multiple inclusion, they are skipped because
26625 they've already been included (in the precompiled header).
26627 If you need to precompile the same header file for different
26628 languages, targets, or compiler options, you can instead make a
26629 @emph{directory} named like @file{all.h.gch}, and put each precompiled
26630 header in the directory, perhaps using @option{-o}. It doesn't matter
26631 what you call the files in the directory; every precompiled header in
26632 the directory is considered. The first precompiled header
26633 encountered in the directory that is valid for this compilation is
26634 used; they're searched in no particular order.
26636 There are many other possibilities, limited only by your imagination,
26637 good sense, and the constraints of your build system.
26639 A precompiled header file can be used only when these conditions apply:
26643 Only one precompiled header can be used in a particular compilation.
26646 A precompiled header can't be used once the first C token is seen. You
26647 can have preprocessor directives before a precompiled header; you cannot
26648 include a precompiled header from inside another header.
26651 The precompiled header file must be produced for the same language as
26652 the current compilation. You can't use a C precompiled header for a C++
26656 The precompiled header file must have been produced by the same compiler
26657 binary as the current compilation is using.
26660 Any macros defined before the precompiled header is included must
26661 either be defined in the same way as when the precompiled header was
26662 generated, or must not affect the precompiled header, which usually
26663 means that they don't appear in the precompiled header at all.
26665 The @option{-D} option is one way to define a macro before a
26666 precompiled header is included; using a @code{#define} can also do it.
26667 There are also some options that define macros implicitly, like
26668 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
26671 @item If debugging information is output when using the precompiled
26672 header, using @option{-g} or similar, the same kind of debugging information
26673 must have been output when building the precompiled header. However,
26674 a precompiled header built using @option{-g} can be used in a compilation
26675 when no debugging information is being output.
26677 @item The same @option{-m} options must generally be used when building
26678 and using the precompiled header. @xref{Submodel Options},
26679 for any cases where this rule is relaxed.
26681 @item Each of the following options must be the same when building and using
26682 the precompiled header:
26684 @gccoptlist{-fexceptions}
26687 Some other command-line options starting with @option{-f},
26688 @option{-p}, or @option{-O} must be defined in the same way as when
26689 the precompiled header was generated. At present, it's not clear
26690 which options are safe to change and which are not; the safest choice
26691 is to use exactly the same options when generating and using the
26692 precompiled header. The following are known to be safe:
26694 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
26695 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
26696 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
26701 For all of these except the last, the compiler automatically
26702 ignores the precompiled header if the conditions aren't met. If you
26703 find an option combination that doesn't work and doesn't cause the
26704 precompiled header to be ignored, please consider filing a bug report,
26707 If you do use differing options when generating and using the
26708 precompiled header, the actual behavior is a mixture of the
26709 behavior for the options. For instance, if you use @option{-g} to
26710 generate the precompiled header but not when using it, you may or may
26711 not get debugging information for routines in the precompiled header.