1 @c Copyright (C) 1988-2014 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{} 1988-2014 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. @samp{g++} accepts mostly the same options as @samp{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.
76 Other options are passed on to one stage of processing. Some options
77 control the preprocessor and others the compiler itself. Yet other
78 options control the assembler and linker; most of these are not
79 documented here, since you rarely need to use any of them.
81 @cindex C compilation options
82 Most of the command-line options that you can use with GCC are useful
83 for C programs; when an option is only useful with another language
84 (usually C++), the explanation says so explicitly. If the description
85 for a particular option does not mention a source language, you can use
86 that option with all supported languages.
88 @cindex C++ compilation options
89 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
90 options for compiling C++ programs.
92 @cindex grouping options
93 @cindex options, grouping
94 The @command{gcc} program accepts options and file names as operands. Many
95 options have multi-letter names; therefore multiple single-letter options
96 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
99 @cindex order of options
100 @cindex options, order
101 You can mix options and other arguments. For the most part, the order
102 you use doesn't matter. Order does matter when you use several
103 options of the same kind; for example, if you specify @option{-L} more
104 than once, the directories are searched in the order specified. Also,
105 the placement of the @option{-l} option is significant.
107 Many options have long names starting with @samp{-f} or with
108 @samp{-W}---for example,
109 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
110 these have both positive and negative forms; the negative form of
111 @option{-ffoo} is @option{-fno-foo}. This manual documents
112 only one of these two forms, whichever one is not the default.
116 @xref{Option Index}, for an index to GCC's options.
119 * Option Summary:: Brief list of all options, without explanations.
120 * Overall Options:: Controlling the kind of output:
121 an executable, object files, assembler files,
122 or preprocessed source.
123 * Invoking G++:: Compiling C++ programs.
124 * C Dialect Options:: Controlling the variant of C language compiled.
125 * C++ Dialect Options:: Variations on C++.
126 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
128 * Language Independent Options:: Controlling how diagnostics should be
130 * Warning Options:: How picky should the compiler be?
131 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
132 * Optimize Options:: How much optimization?
133 * Preprocessor Options:: Controlling header files and macro definitions.
134 Also, getting dependency information for Make.
135 * Assembler Options:: Passing options to the assembler.
136 * Link Options:: Specifying libraries and so on.
137 * Directory Options:: Where to find header files and libraries.
138 Where to find the compiler executable files.
139 * Spec Files:: How to pass switches to sub-processes.
140 * Target Options:: Running a cross-compiler, or an old version of GCC.
141 * Submodel Options:: Specifying minor hardware or convention variations,
142 such as 68010 vs 68020.
143 * Code Gen Options:: Specifying conventions for function calls, data layout
145 * Environment Variables:: Env vars that affect GCC.
146 * Precompiled Headers:: Compiling a header once, and using it many times.
152 @section Option Summary
154 Here is a summary of all the options, grouped by type. Explanations are
155 in the following sections.
158 @item Overall Options
159 @xref{Overall Options,,Options Controlling the Kind of Output}.
160 @gccoptlist{-c -S -E -o @var{file} -no-canonical-prefixes @gol
161 -pipe -pass-exit-codes @gol
162 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
163 --version -wrapper @@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
164 -fdump-ada-spec@r{[}-slim@r{]} -fada-spec-parent=@var{unit} -fdump-go-spec=@var{file}}
166 @item C Language Options
167 @xref{C Dialect Options,,Options Controlling C Dialect}.
168 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
169 -aux-info @var{filename} -fallow-parameterless-variadic-functions @gol
170 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
171 -fhosted -ffreestanding -fopenmp -fopenmp-simd -fms-extensions @gol
172 -fplan9-extensions -trigraphs -traditional -traditional-cpp @gol
173 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
174 -fsigned-bitfields -fsigned-char @gol
175 -funsigned-bitfields -funsigned-char}
177 @item C++ Language Options
178 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
179 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
180 -fconstexpr-depth=@var{n} -ffriend-injection @gol
181 -fno-elide-constructors @gol
182 -fno-enforce-eh-specs @gol
183 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
184 -fno-implicit-templates @gol
185 -fno-implicit-inline-templates @gol
186 -fno-implement-inlines -fms-extensions @gol
187 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
188 -fno-optional-diags -fpermissive @gol
189 -fno-pretty-templates @gol
190 -frepo -fno-rtti -fstats -ftemplate-backtrace-limit=@var{n} @gol
191 -ftemplate-depth=@var{n} @gol
192 -fno-threadsafe-statics -fuse-cxa-atexit @gol
193 -fno-weak -nostdinc++ @gol
194 -fvisibility-inlines-hidden @gol
195 -fvtable-verify=@var{std|preinit|none} @gol
196 -fvtv-counts -fvtv-debug @gol
197 -fvisibility-ms-compat @gol
198 -fext-numeric-literals @gol
199 -Wabi=@var{n} -Wconversion-null -Wctor-dtor-privacy @gol
200 -Wdelete-non-virtual-dtor -Wliteral-suffix -Wnarrowing @gol
201 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
202 -Weffc++ -Wstrict-null-sentinel @gol
203 -Wno-non-template-friend -Wold-style-cast @gol
204 -Woverloaded-virtual -Wno-pmf-conversions @gol
207 @item Objective-C and Objective-C++ Language Options
208 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
209 Objective-C and Objective-C++ Dialects}.
210 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
211 -fgnu-runtime -fnext-runtime @gol
212 -fno-nil-receivers @gol
213 -fobjc-abi-version=@var{n} @gol
214 -fobjc-call-cxx-cdtors @gol
215 -fobjc-direct-dispatch @gol
216 -fobjc-exceptions @gol
219 -fobjc-std=objc1 @gol
220 -fno-local-ivars @gol
221 -fivar-visibility=@var{public|protected|private|package} @gol
222 -freplace-objc-classes @gol
225 -Wassign-intercept @gol
226 -Wno-protocol -Wselector @gol
227 -Wstrict-selector-match @gol
228 -Wundeclared-selector}
230 @item Language Independent Options
231 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
232 @gccoptlist{-fmessage-length=@var{n} @gol
233 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
234 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
235 -fno-diagnostics-show-option -fno-diagnostics-show-caret}
237 @item Warning Options
238 @xref{Warning Options,,Options to Request or Suppress Warnings}.
239 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
240 -pedantic-errors @gol
241 -w -Wextra -Wall -Waddress -Waggregate-return @gol
242 -Waggressive-loop-optimizations -Warray-bounds @gol
244 -Wno-attributes -Wno-builtin-macro-redefined @gol
245 -Wc90-c99-compat -Wc99-c11-compat @gol
246 -Wc++-compat -Wc++11-compat -Wcast-align -Wcast-qual @gol
247 -Wchar-subscripts -Wclobbered -Wcomment -Wconditionally-supported @gol
248 -Wconversion -Wcoverage-mismatch -Wdate-time -Wdelete-incomplete -Wno-cpp @gol
249 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
250 -Wdisabled-optimization -Wno-discarded-qualifiers @gol
251 -Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol
252 -Wno-endif-labels -Werror -Werror=* @gol
253 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
254 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
255 -Wformat-security -Wformat-signedness -Wformat-y2k @gol
256 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
257 -Wignored-qualifiers -Wincompatible-pointer-types @gol
258 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
259 -Winit-self -Winline -Wno-int-conversion @gol
260 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
261 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
262 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
263 -Wmain -Wmaybe-uninitialized -Wmemset-transposed-args -Wmissing-braces @gol
264 -Wmissing-field-initializers -Wmissing-include-dirs @gol
265 -Wno-multichar -Wnonnull -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
266 -Wodr -Wno-overflow -Wopenmp-simd @gol
267 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
268 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
269 -Wpointer-arith -Wno-pointer-to-int-cast @gol
270 -Wredundant-decls -Wno-return-local-addr @gol
271 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
272 -Wshift-count-negative -Wshift-count-overflow @gol
273 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
274 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
275 -Wstack-protector -Wstack-usage=@var{len} -Wstrict-aliasing @gol
276 -Wstrict-aliasing=n @gol -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
277 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]} @gol
278 -Wsuggest-final-types @gol -Wsuggest-final-methods @gol
279 -Wmissing-format-attribute @gol
280 -Wswitch -Wswitch-default -Wswitch-enum -Wswitch-bool -Wsync-nand @gol
281 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
282 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
283 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
284 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
285 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
286 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
287 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
288 -Wvla -Wvolatile-register-var -Wwrite-strings -Wzero-as-null-pointer-constant}
290 @item C and Objective-C-only Warning Options
291 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
292 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
293 -Wold-style-declaration -Wold-style-definition @gol
294 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
295 -Wdeclaration-after-statement -Wpointer-sign}
297 @item Debugging Options
298 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
299 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
300 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
301 -fasan-shadow-offset=@var{number} -fsanitize-undefined-trap-on-error @gol
302 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
303 -fdisable-ipa-@var{pass_name} @gol
304 -fdisable-rtl-@var{pass_name} @gol
305 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
306 -fdisable-tree-@var{pass_name} @gol
307 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
308 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
309 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
310 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
311 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
313 -fdump-statistics @gol
315 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
316 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
317 -fdump-tree-cfg -fdump-tree-alias @gol
319 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
320 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
321 -fdump-tree-gimple@r{[}-raw@r{]} @gol
322 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
323 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
324 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
325 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
326 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
327 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
328 -fdump-tree-nrv -fdump-tree-vect @gol
329 -fdump-tree-sink @gol
330 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
331 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
332 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
333 -fdump-tree-vtable-verify @gol
334 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
335 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
336 -fdump-final-insns=@var{file} @gol
337 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
338 -feliminate-dwarf2-dups -fno-eliminate-unused-debug-types @gol
339 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
340 -fenable-@var{kind}-@var{pass} @gol
341 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
342 -fdebug-types-section -fmem-report-wpa @gol
343 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
345 -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
346 -frandom-seed=@var{number} -fsched-verbose=@var{n} @gol
347 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
348 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
349 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
350 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
351 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
352 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
353 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
354 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
355 -fdebug-prefix-map=@var{old}=@var{new} @gol
356 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
357 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
358 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
359 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
360 -print-prog-name=@var{program} -print-search-dirs -Q @gol
361 -print-sysroot -print-sysroot-headers-suffix @gol
362 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
364 @item Optimization Options
365 @xref{Optimize Options,,Options that Control Optimization}.
366 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
367 -falign-jumps[=@var{n}] @gol
368 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
369 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
370 -fauto-inc-dec -fbranch-probabilities @gol
371 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
372 -fbtr-bb-exclusive -fcaller-saves @gol
373 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
374 -fcompare-elim -fcprop-registers -fcrossjumping @gol
375 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
376 -fcx-limited-range @gol
377 -fdata-sections -fdce -fdelayed-branch @gol
378 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively -fdse @gol
379 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
380 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
381 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
382 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
383 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
384 -fif-conversion2 -findirect-inlining @gol
385 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
386 -finline-small-functions -fipa-cp -fipa-cp-clone @gol
387 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference -fipa-icf @gol
388 -fira-algorithm=@var{algorithm} @gol
389 -fira-region=@var{region} -fira-hoist-pressure @gol
390 -fira-loop-pressure -fno-ira-share-save-slots @gol
391 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
392 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
393 -fivopts -fkeep-inline-functions -fkeep-static-consts -flive-range-shrinkage @gol
394 -floop-block -floop-interchange -floop-strip-mine -floop-nest-optimize @gol
395 -floop-parallelize-all -flto -flto-compression-level @gol
396 -flto-partition=@var{alg} -flto-report -flto-report-wpa -fmerge-all-constants @gol
397 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
398 -fmove-loop-invariants -fno-branch-count-reg @gol
399 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
400 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
401 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
402 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
403 -fomit-frame-pointer -foptimize-sibling-calls @gol
404 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
405 -fprefetch-loop-arrays -fprofile-report @gol
406 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
407 -fprofile-generate=@var{path} @gol
408 -fprofile-use -fprofile-use=@var{path} -fprofile-values -fprofile-reorder-functions @gol
409 -freciprocal-math -free -frename-registers -freorder-blocks @gol
410 -freorder-blocks-and-partition -freorder-functions @gol
411 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
412 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
413 -fsched-spec-load -fsched-spec-load-dangerous @gol
414 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
415 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
416 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
417 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
418 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
419 -fselective-scheduling -fselective-scheduling2 @gol
420 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
421 -fsemantic-interposition @gol
422 -fshrink-wrap -fsignaling-nans -fsingle-precision-constant @gol
423 -fsplit-ivs-in-unroller -fsplit-wide-types -fssa-phiopt -fstack-protector @gol
424 -fstack-protector-all -fstack-protector-strong -fstrict-aliasing @gol
425 -fstrict-overflow -fthread-jumps -ftracer -ftree-bit-ccp @gol
426 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
427 -ftree-coalesce-inline-vars -ftree-coalesce-vars -ftree-copy-prop @gol
428 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
429 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
430 -ftree-loop-if-convert-stores -ftree-loop-im @gol
431 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
432 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
433 -ftree-loop-vectorize @gol
434 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
435 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
436 -ftree-switch-conversion -ftree-tail-merge -ftree-ter @gol
437 -ftree-vectorize -ftree-vrp @gol
438 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
439 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
440 -fuse-caller-save -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
441 -fweb -fwhole-program -fwpa -fuse-ld=@var{linker} -fuse-linker-plugin @gol
442 --param @var{name}=@var{value}
443 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
445 @item Preprocessor Options
446 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
447 @gccoptlist{-A@var{question}=@var{answer} @gol
448 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
449 -C -dD -dI -dM -dN @gol
450 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
451 -idirafter @var{dir} @gol
452 -include @var{file} -imacros @var{file} @gol
453 -iprefix @var{file} -iwithprefix @var{dir} @gol
454 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
455 -imultilib @var{dir} -isysroot @var{dir} @gol
456 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
457 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
458 -remap -trigraphs -undef -U@var{macro} @gol
459 -Wp,@var{option} -Xpreprocessor @var{option} -no-integrated-cpp}
461 @item Assembler Option
462 @xref{Assembler Options,,Passing Options to the Assembler}.
463 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
466 @xref{Link Options,,Options for Linking}.
467 @gccoptlist{@var{object-file-name} -l@var{library} @gol
468 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
469 -s -static -static-libgcc -static-libstdc++ @gol
470 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
471 -shared -shared-libgcc -symbolic @gol
472 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
473 -u @var{symbol} -z @var{keyword}}
475 @item Directory Options
476 @xref{Directory Options,,Options for Directory Search}.
477 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
478 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
479 --sysroot=@var{dir} --no-sysroot-suffix}
481 @item Machine Dependent Options
482 @xref{Submodel Options,,Hardware Models and Configurations}.
483 @c This list is ordered alphanumerically by subsection name.
484 @c Try and put the significant identifier (CPU or system) first,
485 @c so users have a clue at guessing where the ones they want will be.
487 @emph{AArch64 Options}
488 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
489 -mgeneral-regs-only @gol
490 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
492 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
493 -mtls-dialect=desc -mtls-dialect=traditional @gol
494 -mfix-cortex-a53-835769 -mno-fix-cortex-a53-835769 @gol
495 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
497 @emph{Adapteva Epiphany Options}
498 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
499 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
500 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
501 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
502 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
503 -msplit-vecmove-early -m1reg-@var{reg}}
506 @gccoptlist{-mbarrel-shifter @gol
507 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
508 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
509 -mea -mno-mpy -mmul32x16 -mmul64 @gol
510 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
511 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
512 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
513 -mepilogue-cfi -mlong-calls -mmedium-calls -msdata @gol
514 -mucb-mcount -mvolatile-cache @gol
515 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
516 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
517 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
518 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
519 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
520 -mtune=@var{cpu} -mmultcost=@var{num} -munalign-prob-threshold=@var{probability}}
523 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
524 -mabi=@var{name} @gol
525 -mapcs-stack-check -mno-apcs-stack-check @gol
526 -mapcs-float -mno-apcs-float @gol
527 -mapcs-reentrant -mno-apcs-reentrant @gol
528 -msched-prolog -mno-sched-prolog @gol
529 -mlittle-endian -mbig-endian @gol
530 -mfloat-abi=@var{name} @gol
531 -mfp16-format=@var{name}
532 -mthumb-interwork -mno-thumb-interwork @gol
533 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
534 -mstructure-size-boundary=@var{n} @gol
535 -mabort-on-noreturn @gol
536 -mlong-calls -mno-long-calls @gol
537 -msingle-pic-base -mno-single-pic-base @gol
538 -mpic-register=@var{reg} @gol
539 -mnop-fun-dllimport @gol
540 -mpoke-function-name @gol
542 -mtpcs-frame -mtpcs-leaf-frame @gol
543 -mcaller-super-interworking -mcallee-super-interworking @gol
544 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
545 -mword-relocations @gol
546 -mfix-cortex-m3-ldrd @gol
547 -munaligned-access @gol
548 -mneon-for-64bits @gol
549 -mslow-flash-data @gol
550 -masm-syntax-unified @gol
554 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
555 -mcall-prologues -mint8 -mno-interrupts -mrelax @gol
556 -mstrict-X -mtiny-stack -Waddr-space-convert}
558 @emph{Blackfin Options}
559 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
560 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
561 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
562 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
563 -mno-id-shared-library -mshared-library-id=@var{n} @gol
564 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
565 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
566 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
570 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
571 -msim -msdata=@var{sdata-type}}
574 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
575 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
576 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
577 -mstack-align -mdata-align -mconst-align @gol
578 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
579 -melf -maout -melinux -mlinux -sim -sim2 @gol
580 -mmul-bug-workaround -mno-mul-bug-workaround}
583 @gccoptlist{-mmac @gol
584 -mcr16cplus -mcr16c @gol
585 -msim -mint32 -mbit-ops
586 -mdata-model=@var{model}}
588 @emph{Darwin Options}
589 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
590 -arch_only -bind_at_load -bundle -bundle_loader @gol
591 -client_name -compatibility_version -current_version @gol
593 -dependency-file -dylib_file -dylinker_install_name @gol
594 -dynamic -dynamiclib -exported_symbols_list @gol
595 -filelist -flat_namespace -force_cpusubtype_ALL @gol
596 -force_flat_namespace -headerpad_max_install_names @gol
598 -image_base -init -install_name -keep_private_externs @gol
599 -multi_module -multiply_defined -multiply_defined_unused @gol
600 -noall_load -no_dead_strip_inits_and_terms @gol
601 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
602 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
603 -private_bundle -read_only_relocs -sectalign @gol
604 -sectobjectsymbols -whyload -seg1addr @gol
605 -sectcreate -sectobjectsymbols -sectorder @gol
606 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
607 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
608 -segprot -segs_read_only_addr -segs_read_write_addr @gol
609 -single_module -static -sub_library -sub_umbrella @gol
610 -twolevel_namespace -umbrella -undefined @gol
611 -unexported_symbols_list -weak_reference_mismatches @gol
612 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
613 -mkernel -mone-byte-bool}
615 @emph{DEC Alpha Options}
616 @gccoptlist{-mno-fp-regs -msoft-float @gol
617 -mieee -mieee-with-inexact -mieee-conformant @gol
618 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
619 -mtrap-precision=@var{mode} -mbuild-constants @gol
620 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
621 -mbwx -mmax -mfix -mcix @gol
622 -mfloat-vax -mfloat-ieee @gol
623 -mexplicit-relocs -msmall-data -mlarge-data @gol
624 -msmall-text -mlarge-text @gol
625 -mmemory-latency=@var{time}}
628 @gccoptlist{-msmall-model -mno-lsim}
631 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
632 -mhard-float -msoft-float @gol
633 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
634 -mdouble -mno-double @gol
635 -mmedia -mno-media -mmuladd -mno-muladd @gol
636 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
637 -mlinked-fp -mlong-calls -malign-labels @gol
638 -mlibrary-pic -macc-4 -macc-8 @gol
639 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
640 -moptimize-membar -mno-optimize-membar @gol
641 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
642 -mvliw-branch -mno-vliw-branch @gol
643 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
644 -mno-nested-cond-exec -mtomcat-stats @gol
648 @emph{GNU/Linux Options}
649 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
650 -tno-android-cc -tno-android-ld}
652 @emph{H8/300 Options}
653 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
656 @gccoptlist{-march=@var{architecture-type} @gol
657 -mdisable-fpregs -mdisable-indexing @gol
658 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
659 -mfixed-range=@var{register-range} @gol
660 -mjump-in-delay -mlinker-opt -mlong-calls @gol
661 -mlong-load-store -mno-disable-fpregs @gol
662 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
663 -mno-jump-in-delay -mno-long-load-store @gol
664 -mno-portable-runtime -mno-soft-float @gol
665 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
666 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
667 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
668 -munix=@var{unix-std} -nolibdld -static -threads}
670 @emph{i386 and x86-64 Options}
671 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
672 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
673 -mfpmath=@var{unit} @gol
674 -masm=@var{dialect} -mno-fancy-math-387 @gol
675 -mno-fp-ret-in-387 -msoft-float @gol
676 -mno-wide-multiply -mrtd -malign-double @gol
677 -mpreferred-stack-boundary=@var{num} @gol
678 -mincoming-stack-boundary=@var{num} @gol
679 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
680 -mrecip -mrecip=@var{opt} @gol
681 -mvzeroupper -mprefer-avx128 @gol
682 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
683 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -msha @gol
684 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma -mprefetchwt1 @gol
685 -mclflushopt -mxsavec -mxsaves @gol
686 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
687 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx -mthreads @gol
688 -mno-align-stringops -minline-all-stringops @gol
689 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
690 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy}
691 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
692 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
693 -mregparm=@var{num} -msseregparm @gol
694 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
695 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
696 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
697 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
698 -m32 -m64 -mx32 -m16 -mlarge-data-threshold=@var{num} @gol
699 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
700 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
701 -mstack-protector-guard=@var{guard}}
703 @emph{i386 and x86-64 Windows Options}
704 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
705 -mnop-fun-dllimport -mthread @gol
706 -municode -mwin32 -mwindows -fno-set-stack-executable}
709 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
710 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
711 -mconstant-gp -mauto-pic -mfused-madd @gol
712 -minline-float-divide-min-latency @gol
713 -minline-float-divide-max-throughput @gol
714 -mno-inline-float-divide @gol
715 -minline-int-divide-min-latency @gol
716 -minline-int-divide-max-throughput @gol
717 -mno-inline-int-divide @gol
718 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
719 -mno-inline-sqrt @gol
720 -mdwarf2-asm -mearly-stop-bits @gol
721 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
722 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
723 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
724 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
725 -msched-spec-ldc -msched-spec-control-ldc @gol
726 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
727 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
728 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
729 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
732 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
733 -msign-extend-enabled -muser-enabled}
735 @emph{M32R/D Options}
736 @gccoptlist{-m32r2 -m32rx -m32r @gol
738 -malign-loops -mno-align-loops @gol
739 -missue-rate=@var{number} @gol
740 -mbranch-cost=@var{number} @gol
741 -mmodel=@var{code-size-model-type} @gol
742 -msdata=@var{sdata-type} @gol
743 -mno-flush-func -mflush-func=@var{name} @gol
744 -mno-flush-trap -mflush-trap=@var{number} @gol
748 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
750 @emph{M680x0 Options}
751 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
752 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
753 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
754 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
755 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
756 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
757 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
758 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
762 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
763 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
764 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
765 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
766 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
769 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
770 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
771 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
772 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
775 @emph{MicroBlaze Options}
776 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
777 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
778 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
779 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
780 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
783 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
784 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
785 -mips64 -mips64r2 -mips64r3 -mips64r5 @gol
786 -mips16 -mno-mips16 -mflip-mips16 @gol
787 -minterlink-compressed -mno-interlink-compressed @gol
788 -minterlink-mips16 -mno-interlink-mips16 @gol
789 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
790 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
791 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
792 -mno-float -msingle-float -mdouble-float @gol
793 -modd-spreg -mno-odd-spreg @gol
794 -mabs=@var{mode} -mnan=@var{encoding} @gol
795 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
798 -mvirt -mno-virt @gol
800 -mmicromips -mno-micromips @gol
801 -mfpu=@var{fpu-type} @gol
802 -msmartmips -mno-smartmips @gol
803 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
804 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
805 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
806 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
807 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
808 -membedded-data -mno-embedded-data @gol
809 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
810 -mcode-readable=@var{setting} @gol
811 -msplit-addresses -mno-split-addresses @gol
812 -mexplicit-relocs -mno-explicit-relocs @gol
813 -mcheck-zero-division -mno-check-zero-division @gol
814 -mdivide-traps -mdivide-breaks @gol
815 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
816 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
817 -mfix-24k -mno-fix-24k @gol
818 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
819 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
820 -mfix-vr4120 -mno-fix-vr4120 @gol
821 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
822 -mflush-func=@var{func} -mno-flush-func @gol
823 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
824 -mfp-exceptions -mno-fp-exceptions @gol
825 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
826 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
829 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
830 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
831 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
832 -mno-base-addresses -msingle-exit -mno-single-exit}
834 @emph{MN10300 Options}
835 @gccoptlist{-mmult-bug -mno-mult-bug @gol
836 -mno-am33 -mam33 -mam33-2 -mam34 @gol
837 -mtune=@var{cpu-type} @gol
838 -mreturn-pointer-on-d0 @gol
839 -mno-crt0 -mrelax -mliw -msetlb}
842 @gccoptlist{-meb -mel -mno-crt0}
844 @emph{MSP430 Options}
845 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
849 @gccoptlist{-mbig-endian -mlittle-endian @gol
850 -mreduced-regs -mfull-regs @gol
851 -mcmov -mno-cmov @gol
852 -mperf-ext -mno-perf-ext @gol
853 -mv3push -mno-v3push @gol
854 -m16bit -mno-16bit @gol
855 -mgp-direct -mno-gp-direct @gol
856 -misr-vector-size=@var{num} @gol
857 -mcache-block-size=@var{num} @gol
858 -march=@var{arch} @gol
859 -mforce-fp-as-gp -mforbid-fp-as-gp @gol
860 -mex9 -mctor-dtor -mrelax}
862 @emph{Nios II Options}
863 @gccoptlist{-G @var{num} -mgpopt -mno-gpopt -mel -meb @gol
864 -mno-bypass-cache -mbypass-cache @gol
865 -mno-cache-volatile -mcache-volatile @gol
866 -mno-fast-sw-div -mfast-sw-div @gol
867 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
868 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
869 -mcustom-fpu-cfg=@var{name} @gol
870 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name}}
872 @emph{PDP-11 Options}
873 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
874 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
875 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
876 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
877 -mbranch-expensive -mbranch-cheap @gol
878 -munix-asm -mdec-asm}
880 @emph{picoChip Options}
881 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
882 -msymbol-as-address -mno-inefficient-warnings}
884 @emph{PowerPC Options}
885 See RS/6000 and PowerPC Options.
888 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=rl78 @gol
889 -m64bit-doubles -m32bit-doubles}
891 @emph{RS/6000 and PowerPC Options}
892 @gccoptlist{-mcpu=@var{cpu-type} @gol
893 -mtune=@var{cpu-type} @gol
894 -mcmodel=@var{code-model} @gol
896 -maltivec -mno-altivec @gol
897 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
898 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
899 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
900 -mfprnd -mno-fprnd @gol
901 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
902 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
903 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
904 -malign-power -malign-natural @gol
905 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
906 -msingle-float -mdouble-float -msimple-fpu @gol
907 -mstring -mno-string -mupdate -mno-update @gol
908 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
909 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
910 -mstrict-align -mno-strict-align -mrelocatable @gol
911 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
912 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
913 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
914 -mprioritize-restricted-insns=@var{priority} @gol
915 -msched-costly-dep=@var{dependence_type} @gol
916 -minsert-sched-nops=@var{scheme} @gol
917 -mcall-sysv -mcall-netbsd @gol
918 -maix-struct-return -msvr4-struct-return @gol
919 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
920 -mblock-move-inline-limit=@var{num} @gol
921 -misel -mno-isel @gol
922 -misel=yes -misel=no @gol
924 -mspe=yes -mspe=no @gol
926 -mgen-cell-microcode -mwarn-cell-microcode @gol
927 -mvrsave -mno-vrsave @gol
928 -mmulhw -mno-mulhw @gol
929 -mdlmzb -mno-dlmzb @gol
930 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
931 -mprototype -mno-prototype @gol
932 -msim -mmvme -mads -myellowknife -memb -msdata @gol
933 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
934 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
935 -mno-recip-precision @gol
936 -mveclibabi=@var{type} -mfriz -mno-friz @gol
937 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
938 -msave-toc-indirect -mno-save-toc-indirect @gol
939 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
940 -mcrypto -mno-crypto -mdirect-move -mno-direct-move @gol
941 -mquad-memory -mno-quad-memory @gol
942 -mquad-memory-atomic -mno-quad-memory-atomic @gol
943 -mcompat-align-parm -mno-compat-align-parm}
946 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
948 -mbig-endian-data -mlittle-endian-data @gol
951 -mas100-syntax -mno-as100-syntax@gol
953 -mmax-constant-size=@gol
956 -mno-warn-multiple-fast-interrupts@gol
957 -msave-acc-in-interrupts}
959 @emph{S/390 and zSeries Options}
960 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
961 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
962 -mlong-double-64 -mlong-double-128 @gol
963 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
964 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
965 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
966 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
967 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
968 -mhotpatch[=@var{halfwords}] -mno-hotpatch}
971 @gccoptlist{-meb -mel @gol
975 -mscore5 -mscore5u -mscore7 -mscore7d}
978 @gccoptlist{-m1 -m2 -m2e @gol
979 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
981 -m4-nofpu -m4-single-only -m4-single -m4 @gol
982 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
983 -m5-64media -m5-64media-nofpu @gol
984 -m5-32media -m5-32media-nofpu @gol
985 -m5-compact -m5-compact-nofpu @gol
986 -mb -ml -mdalign -mrelax @gol
987 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
988 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
989 -mspace -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
990 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
991 -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
992 -maccumulate-outgoing-args -minvalid-symbols @gol
993 -matomic-model=@var{atomic-model} @gol
994 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
995 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
996 -mpretend-cmove -mtas}
998 @emph{Solaris 2 Options}
999 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1002 @emph{SPARC Options}
1003 @gccoptlist{-mcpu=@var{cpu-type} @gol
1004 -mtune=@var{cpu-type} @gol
1005 -mcmodel=@var{code-model} @gol
1006 -mmemory-model=@var{mem-model} @gol
1007 -m32 -m64 -mapp-regs -mno-app-regs @gol
1008 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1009 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1010 -mhard-quad-float -msoft-quad-float @gol
1011 -mstack-bias -mno-stack-bias @gol
1012 -munaligned-doubles -mno-unaligned-doubles @gol
1013 -muser-mode -mno-user-mode @gol
1014 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1015 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1016 -mcbcond -mno-cbcond @gol
1017 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
1018 -mfix-at697f -mfix-ut699}
1021 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1022 -msafe-dma -munsafe-dma @gol
1024 -msmall-mem -mlarge-mem -mstdmain @gol
1025 -mfixed-range=@var{register-range} @gol
1027 -maddress-space-conversion -mno-address-space-conversion @gol
1028 -mcache-size=@var{cache-size} @gol
1029 -matomic-updates -mno-atomic-updates}
1031 @emph{System V Options}
1032 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1034 @emph{TILE-Gx Options}
1035 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1036 -mcmodel=@var{code-model}}
1038 @emph{TILEPro Options}
1039 @gccoptlist{-mcpu=@var{cpu} -m32}
1042 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1043 -mprolog-function -mno-prolog-function -mspace @gol
1044 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1045 -mapp-regs -mno-app-regs @gol
1046 -mdisable-callt -mno-disable-callt @gol
1047 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1048 -mv850e -mv850 -mv850e3v5 @gol
1059 @gccoptlist{-mg -mgnu -munix}
1062 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1063 -mpointer-size=@var{size}}
1065 @emph{VxWorks Options}
1066 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1067 -Xbind-lazy -Xbind-now}
1069 @emph{x86-64 Options}
1070 See i386 and x86-64 Options.
1072 @emph{Xstormy16 Options}
1075 @emph{Xtensa Options}
1076 @gccoptlist{-mconst16 -mno-const16 @gol
1077 -mfused-madd -mno-fused-madd @gol
1079 -mserialize-volatile -mno-serialize-volatile @gol
1080 -mtext-section-literals -mno-text-section-literals @gol
1081 -mtarget-align -mno-target-align @gol
1082 -mlongcalls -mno-longcalls}
1084 @emph{zSeries Options}
1085 See S/390 and zSeries Options.
1087 @item Code Generation Options
1088 @xref{Code Gen Options,,Options for Code Generation Conventions}.
1089 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
1090 -ffixed-@var{reg} -fexceptions @gol
1091 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
1092 -fasynchronous-unwind-tables @gol
1093 -fno-gnu-unique @gol
1094 -finhibit-size-directive -finstrument-functions @gol
1095 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
1096 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
1097 -fno-common -fno-ident @gol
1098 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
1099 -fno-jump-tables @gol
1100 -frecord-gcc-switches @gol
1101 -freg-struct-return -fshort-enums @gol
1102 -fshort-double -fshort-wchar @gol
1103 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
1104 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
1105 -fno-stack-limit -fsplit-stack @gol
1106 -fleading-underscore -ftls-model=@var{model} @gol
1107 -fstack-reuse=@var{reuse_level} @gol
1108 -ftrapv -fwrapv -fbounds-check @gol
1109 -fvisibility -fstrict-volatile-bitfields -fsync-libcalls}
1113 @node Overall Options
1114 @section Options Controlling the Kind of Output
1116 Compilation can involve up to four stages: preprocessing, compilation
1117 proper, assembly and linking, always in that order. GCC is capable of
1118 preprocessing and compiling several files either into several
1119 assembler input files, or into one assembler input file; then each
1120 assembler input file produces an object file, and linking combines all
1121 the object files (those newly compiled, and those specified as input)
1122 into an executable file.
1124 @cindex file name suffix
1125 For any given input file, the file name suffix determines what kind of
1126 compilation is done:
1130 C source code that must be preprocessed.
1133 C source code that should not be preprocessed.
1136 C++ source code that should not be preprocessed.
1139 Objective-C source code. Note that you must link with the @file{libobjc}
1140 library to make an Objective-C program work.
1143 Objective-C source code that should not be preprocessed.
1147 Objective-C++ source code. Note that you must link with the @file{libobjc}
1148 library to make an Objective-C++ program work. Note that @samp{.M} refers
1149 to a literal capital M@.
1151 @item @var{file}.mii
1152 Objective-C++ source code that should not be preprocessed.
1155 C, C++, Objective-C or Objective-C++ header file to be turned into a
1156 precompiled header (default), or C, C++ header file to be turned into an
1157 Ada spec (via the @option{-fdump-ada-spec} switch).
1160 @itemx @var{file}.cp
1161 @itemx @var{file}.cxx
1162 @itemx @var{file}.cpp
1163 @itemx @var{file}.CPP
1164 @itemx @var{file}.c++
1166 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1167 the last two letters must both be literally @samp{x}. Likewise,
1168 @samp{.C} refers to a literal capital C@.
1172 Objective-C++ source code that must be preprocessed.
1174 @item @var{file}.mii
1175 Objective-C++ source code that should not be preprocessed.
1179 @itemx @var{file}.hp
1180 @itemx @var{file}.hxx
1181 @itemx @var{file}.hpp
1182 @itemx @var{file}.HPP
1183 @itemx @var{file}.h++
1184 @itemx @var{file}.tcc
1185 C++ header file to be turned into a precompiled header or Ada spec.
1188 @itemx @var{file}.for
1189 @itemx @var{file}.ftn
1190 Fixed form Fortran source code that should not be preprocessed.
1193 @itemx @var{file}.FOR
1194 @itemx @var{file}.fpp
1195 @itemx @var{file}.FPP
1196 @itemx @var{file}.FTN
1197 Fixed form Fortran source code that must be preprocessed (with the traditional
1200 @item @var{file}.f90
1201 @itemx @var{file}.f95
1202 @itemx @var{file}.f03
1203 @itemx @var{file}.f08
1204 Free form Fortran source code that should not be preprocessed.
1206 @item @var{file}.F90
1207 @itemx @var{file}.F95
1208 @itemx @var{file}.F03
1209 @itemx @var{file}.F08
1210 Free form Fortran source code that must be preprocessed (with the
1211 traditional preprocessor).
1216 @c FIXME: Descriptions of Java file types.
1222 @item @var{file}.ads
1223 Ada source code file that contains a library unit declaration (a
1224 declaration of a package, subprogram, or generic, or a generic
1225 instantiation), or a library unit renaming declaration (a package,
1226 generic, or subprogram renaming declaration). Such files are also
1229 @item @var{file}.adb
1230 Ada source code file containing a library unit body (a subprogram or
1231 package body). Such files are also called @dfn{bodies}.
1233 @c GCC also knows about some suffixes for languages not yet included:
1244 @itemx @var{file}.sx
1245 Assembler code that must be preprocessed.
1248 An object file to be fed straight into linking.
1249 Any file name with no recognized suffix is treated this way.
1253 You can specify the input language explicitly with the @option{-x} option:
1256 @item -x @var{language}
1257 Specify explicitly the @var{language} for the following input files
1258 (rather than letting the compiler choose a default based on the file
1259 name suffix). This option applies to all following input files until
1260 the next @option{-x} option. Possible values for @var{language} are:
1262 c c-header cpp-output
1263 c++ c++-header c++-cpp-output
1264 objective-c objective-c-header objective-c-cpp-output
1265 objective-c++ objective-c++-header objective-c++-cpp-output
1266 assembler assembler-with-cpp
1268 f77 f77-cpp-input f95 f95-cpp-input
1274 Turn off any specification of a language, so that subsequent files are
1275 handled according to their file name suffixes (as they are if @option{-x}
1276 has not been used at all).
1278 @item -pass-exit-codes
1279 @opindex pass-exit-codes
1280 Normally the @command{gcc} program exits with the code of 1 if any
1281 phase of the compiler returns a non-success return code. If you specify
1282 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1283 the numerically highest error produced by any phase returning an error
1284 indication. The C, C++, and Fortran front ends return 4 if an internal
1285 compiler error is encountered.
1288 If you only want some of the stages of compilation, you can use
1289 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1290 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1291 @command{gcc} is to stop. Note that some combinations (for example,
1292 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1297 Compile or assemble the source files, but do not link. The linking
1298 stage simply is not done. The ultimate output is in the form of an
1299 object file for each source file.
1301 By default, the object file name for a source file is made by replacing
1302 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1304 Unrecognized input files, not requiring compilation or assembly, are
1309 Stop after the stage of compilation proper; do not assemble. The output
1310 is in the form of an assembler code file for each non-assembler input
1313 By default, the assembler file name for a source file is made by
1314 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1316 Input files that don't require compilation are ignored.
1320 Stop after the preprocessing stage; do not run the compiler proper. The
1321 output is in the form of preprocessed source code, which is sent to the
1324 Input files that don't require preprocessing are ignored.
1326 @cindex output file option
1329 Place output in file @var{file}. This applies to whatever
1330 sort of output is being produced, whether it be an executable file,
1331 an object file, an assembler file or preprocessed C code.
1333 If @option{-o} is not specified, the default is to put an executable
1334 file in @file{a.out}, the object file for
1335 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1336 assembler file in @file{@var{source}.s}, a precompiled header file in
1337 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1342 Print (on standard error output) the commands executed to run the stages
1343 of compilation. Also print the version number of the compiler driver
1344 program and of the preprocessor and the compiler proper.
1348 Like @option{-v} except the commands are not executed and arguments
1349 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1350 This is useful for shell scripts to capture the driver-generated command lines.
1354 Use pipes rather than temporary files for communication between the
1355 various stages of compilation. This fails to work on some systems where
1356 the assembler is unable to read from a pipe; but the GNU assembler has
1361 Print (on the standard output) a description of the command-line options
1362 understood by @command{gcc}. If the @option{-v} option is also specified
1363 then @option{--help} is also passed on to the various processes
1364 invoked by @command{gcc}, so that they can display the command-line options
1365 they accept. If the @option{-Wextra} option has also been specified
1366 (prior to the @option{--help} option), then command-line options that
1367 have no documentation associated with them are also displayed.
1370 @opindex target-help
1371 Print (on the standard output) a description of target-specific command-line
1372 options for each tool. For some targets extra target-specific
1373 information may also be printed.
1375 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1376 Print (on the standard output) a description of the command-line
1377 options understood by the compiler that fit into all specified classes
1378 and qualifiers. These are the supported classes:
1381 @item @samp{optimizers}
1382 Display all of the optimization options supported by the
1385 @item @samp{warnings}
1386 Display all of the options controlling warning messages
1387 produced by the compiler.
1390 Display target-specific options. Unlike the
1391 @option{--target-help} option however, target-specific options of the
1392 linker and assembler are not displayed. This is because those
1393 tools do not currently support the extended @option{--help=} syntax.
1396 Display the values recognized by the @option{--param}
1399 @item @var{language}
1400 Display the options supported for @var{language}, where
1401 @var{language} is the name of one of the languages supported in this
1405 Display the options that are common to all languages.
1408 These are the supported qualifiers:
1411 @item @samp{undocumented}
1412 Display only those options that are undocumented.
1415 Display options taking an argument that appears after an equal
1416 sign in the same continuous piece of text, such as:
1417 @samp{--help=target}.
1419 @item @samp{separate}
1420 Display options taking an argument that appears as a separate word
1421 following the original option, such as: @samp{-o output-file}.
1424 Thus for example to display all the undocumented target-specific
1425 switches supported by the compiler, use:
1428 --help=target,undocumented
1431 The sense of a qualifier can be inverted by prefixing it with the
1432 @samp{^} character, so for example to display all binary warning
1433 options (i.e., ones that are either on or off and that do not take an
1434 argument) that have a description, use:
1437 --help=warnings,^joined,^undocumented
1440 The argument to @option{--help=} should not consist solely of inverted
1443 Combining several classes is possible, although this usually
1444 restricts the output so much that there is nothing to display. One
1445 case where it does work, however, is when one of the classes is
1446 @var{target}. For example, to display all the target-specific
1447 optimization options, use:
1450 --help=target,optimizers
1453 The @option{--help=} option can be repeated on the command line. Each
1454 successive use displays its requested class of options, skipping
1455 those that have already been displayed.
1457 If the @option{-Q} option appears on the command line before the
1458 @option{--help=} option, then the descriptive text displayed by
1459 @option{--help=} is changed. Instead of describing the displayed
1460 options, an indication is given as to whether the option is enabled,
1461 disabled or set to a specific value (assuming that the compiler
1462 knows this at the point where the @option{--help=} option is used).
1464 Here is a truncated example from the ARM port of @command{gcc}:
1467 % gcc -Q -mabi=2 --help=target -c
1468 The following options are target specific:
1470 -mabort-on-noreturn [disabled]
1474 The output is sensitive to the effects of previous command-line
1475 options, so for example it is possible to find out which optimizations
1476 are enabled at @option{-O2} by using:
1479 -Q -O2 --help=optimizers
1482 Alternatively you can discover which binary optimizations are enabled
1483 by @option{-O3} by using:
1486 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1487 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1488 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1491 @item -no-canonical-prefixes
1492 @opindex no-canonical-prefixes
1493 Do not expand any symbolic links, resolve references to @samp{/../}
1494 or @samp{/./}, or make the path absolute when generating a relative
1499 Display the version number and copyrights of the invoked GCC@.
1503 Invoke all subcommands under a wrapper program. The name of the
1504 wrapper program and its parameters are passed as a comma separated
1508 gcc -c t.c -wrapper gdb,--args
1512 This invokes all subprograms of @command{gcc} under
1513 @samp{gdb --args}, thus the invocation of @command{cc1} is
1514 @samp{gdb --args cc1 @dots{}}.
1516 @item -fplugin=@var{name}.so
1518 Load the plugin code in file @var{name}.so, assumed to be a
1519 shared object to be dlopen'd by the compiler. The base name of
1520 the shared object file is used to identify the plugin for the
1521 purposes of argument parsing (See
1522 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1523 Each plugin should define the callback functions specified in the
1526 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1527 @opindex fplugin-arg
1528 Define an argument called @var{key} with a value of @var{value}
1529 for the plugin called @var{name}.
1531 @item -fdump-ada-spec@r{[}-slim@r{]}
1532 @opindex fdump-ada-spec
1533 For C and C++ source and include files, generate corresponding Ada specs.
1534 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1535 GNAT User's Guide}, which provides detailed documentation on this feature.
1537 @item -fada-spec-parent=@var{unit}
1538 @opindex fada-spec-parent
1539 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1540 Ada specs as child units of parent @var{unit}.
1542 @item -fdump-go-spec=@var{file}
1543 @opindex fdump-go-spec
1544 For input files in any language, generate corresponding Go
1545 declarations in @var{file}. This generates Go @code{const},
1546 @code{type}, @code{var}, and @code{func} declarations which may be a
1547 useful way to start writing a Go interface to code written in some
1550 @include @value{srcdir}/../libiberty/at-file.texi
1554 @section Compiling C++ Programs
1556 @cindex suffixes for C++ source
1557 @cindex C++ source file suffixes
1558 C++ source files conventionally use one of the suffixes @samp{.C},
1559 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1560 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1561 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1562 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1563 files with these names and compiles them as C++ programs even if you
1564 call the compiler the same way as for compiling C programs (usually
1565 with the name @command{gcc}).
1569 However, the use of @command{gcc} does not add the C++ library.
1570 @command{g++} is a program that calls GCC and automatically specifies linking
1571 against the C++ library. It treats @samp{.c},
1572 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1573 files unless @option{-x} is used. This program is also useful when
1574 precompiling a C header file with a @samp{.h} extension for use in C++
1575 compilations. On many systems, @command{g++} is also installed with
1576 the name @command{c++}.
1578 @cindex invoking @command{g++}
1579 When you compile C++ programs, you may specify many of the same
1580 command-line options that you use for compiling programs in any
1581 language; or command-line options meaningful for C and related
1582 languages; or options that are meaningful only for C++ programs.
1583 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1584 explanations of options for languages related to C@.
1585 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1586 explanations of options that are meaningful only for C++ programs.
1588 @node C Dialect Options
1589 @section Options Controlling C Dialect
1590 @cindex dialect options
1591 @cindex language dialect options
1592 @cindex options, dialect
1594 The following options control the dialect of C (or languages derived
1595 from C, such as C++, Objective-C and Objective-C++) that the compiler
1599 @cindex ANSI support
1603 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1604 equivalent to @option{-std=c++98}.
1606 This turns off certain features of GCC that are incompatible with ISO
1607 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1608 such as the @code{asm} and @code{typeof} keywords, and
1609 predefined macros such as @code{unix} and @code{vax} that identify the
1610 type of system you are using. It also enables the undesirable and
1611 rarely used ISO trigraph feature. For the C compiler,
1612 it disables recognition of C++ style @samp{//} comments as well as
1613 the @code{inline} keyword.
1615 The alternate keywords @code{__asm__}, @code{__extension__},
1616 @code{__inline__} and @code{__typeof__} continue to work despite
1617 @option{-ansi}. You would not want to use them in an ISO C program, of
1618 course, but it is useful to put them in header files that might be included
1619 in compilations done with @option{-ansi}. Alternate predefined macros
1620 such as @code{__unix__} and @code{__vax__} are also available, with or
1621 without @option{-ansi}.
1623 The @option{-ansi} option does not cause non-ISO programs to be
1624 rejected gratuitously. For that, @option{-Wpedantic} is required in
1625 addition to @option{-ansi}. @xref{Warning Options}.
1627 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1628 option is used. Some header files may notice this macro and refrain
1629 from declaring certain functions or defining certain macros that the
1630 ISO standard doesn't call for; this is to avoid interfering with any
1631 programs that might use these names for other things.
1633 Functions that are normally built in but do not have semantics
1634 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1635 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1636 built-in functions provided by GCC}, for details of the functions
1641 Determine the language standard. @xref{Standards,,Language Standards
1642 Supported by GCC}, for details of these standard versions. This option
1643 is currently only supported when compiling C or C++.
1645 The compiler can accept several base standards, such as @samp{c90} or
1646 @samp{c++98}, and GNU dialects of those standards, such as
1647 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1648 compiler accepts all programs following that standard plus those
1649 using GNU extensions that do not contradict it. For example,
1650 @option{-std=c90} turns off certain features of GCC that are
1651 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1652 keywords, but not other GNU extensions that do not have a meaning in
1653 ISO C90, such as omitting the middle term of a @code{?:}
1654 expression. On the other hand, when a GNU dialect of a standard is
1655 specified, all features supported by the compiler are enabled, even when
1656 those features change the meaning of the base standard. As a result, some
1657 strict-conforming programs may be rejected. The particular standard
1658 is used by @option{-Wpedantic} to identify which features are GNU
1659 extensions given that version of the standard. For example
1660 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1661 comments, while @option{-std=gnu99 -Wpedantic} does not.
1663 A value for this option must be provided; possible values are
1669 Support all ISO C90 programs (certain GNU extensions that conflict
1670 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1672 @item iso9899:199409
1673 ISO C90 as modified in amendment 1.
1679 ISO C99. This standard is substantially completely supported, modulo
1680 bugs and floating-point issues
1681 (mainly but not entirely relating to optional C99 features from
1682 Annexes F and G). See
1683 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1684 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1689 ISO C11, the 2011 revision of the ISO C standard. This standard is
1690 substantially completely supported, modulo bugs, floating-point issues
1691 (mainly but not entirely relating to optional C11 features from
1692 Annexes F and G) and the optional Annexes K (Bounds-checking
1693 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1697 GNU dialect of ISO C90 (including some C99 features).
1701 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1705 GNU dialect of ISO C11. This is the default for C code.
1706 The name @samp{gnu1x} is deprecated.
1710 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1711 additional defect reports. Same as @option{-ansi} for C++ code.
1715 GNU dialect of @option{-std=c++98}. This is the default for
1720 The 2011 ISO C++ standard plus amendments.
1721 The name @samp{c++0x} is deprecated.
1725 GNU dialect of @option{-std=c++11}.
1726 The name @samp{gnu++0x} is deprecated.
1730 The 2014 ISO C++ standard plus amendments.
1731 The name @samp{c++1y} is deprecated.
1735 GNU dialect of @option{-std=c++14}.
1736 The name @samp{gnu++1y} is deprecated.
1739 The next revision of the ISO C++ standard, tentatively planned for
1740 2017. Support is highly experimental, and will almost certainly
1741 change in incompatible ways in future releases.
1744 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1745 and will almost certainly change in incompatible ways in future
1749 @item -fgnu89-inline
1750 @opindex fgnu89-inline
1751 The option @option{-fgnu89-inline} tells GCC to use the traditional
1752 GNU semantics for @code{inline} functions when in C99 mode.
1753 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1754 is accepted and ignored by GCC versions 4.1.3 up to but not including
1755 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1756 C99 mode. Using this option is roughly equivalent to adding the
1757 @code{gnu_inline} function attribute to all inline functions
1758 (@pxref{Function Attributes}).
1760 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1761 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1762 specifies the default behavior). This option was first supported in
1763 GCC 4.3. This option is not supported in @option{-std=c90} or
1764 @option{-std=gnu90} mode.
1766 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1767 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1768 in effect for @code{inline} functions. @xref{Common Predefined
1769 Macros,,,cpp,The C Preprocessor}.
1771 @item -aux-info @var{filename}
1773 Output to the given filename prototyped declarations for all functions
1774 declared and/or defined in a translation unit, including those in header
1775 files. This option is silently ignored in any language other than C@.
1777 Besides declarations, the file indicates, in comments, the origin of
1778 each declaration (source file and line), whether the declaration was
1779 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1780 @samp{O} for old, respectively, in the first character after the line
1781 number and the colon), and whether it came from a declaration or a
1782 definition (@samp{C} or @samp{F}, respectively, in the following
1783 character). In the case of function definitions, a K&R-style list of
1784 arguments followed by their declarations is also provided, inside
1785 comments, after the declaration.
1787 @item -fallow-parameterless-variadic-functions
1788 @opindex fallow-parameterless-variadic-functions
1789 Accept variadic functions without named parameters.
1791 Although it is possible to define such a function, this is not very
1792 useful as it is not possible to read the arguments. This is only
1793 supported for C as this construct is allowed by C++.
1797 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1798 keyword, so that code can use these words as identifiers. You can use
1799 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1800 instead. @option{-ansi} implies @option{-fno-asm}.
1802 In C++, this switch only affects the @code{typeof} keyword, since
1803 @code{asm} and @code{inline} are standard keywords. You may want to
1804 use the @option{-fno-gnu-keywords} flag instead, which has the same
1805 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1806 switch only affects the @code{asm} and @code{typeof} keywords, since
1807 @code{inline} is a standard keyword in ISO C99.
1810 @itemx -fno-builtin-@var{function}
1811 @opindex fno-builtin
1812 @cindex built-in functions
1813 Don't recognize built-in functions that do not begin with
1814 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1815 functions provided by GCC}, for details of the functions affected,
1816 including those which are not built-in functions when @option{-ansi} or
1817 @option{-std} options for strict ISO C conformance are used because they
1818 do not have an ISO standard meaning.
1820 GCC normally generates special code to handle certain built-in functions
1821 more efficiently; for instance, calls to @code{alloca} may become single
1822 instructions which adjust the stack directly, and calls to @code{memcpy}
1823 may become inline copy loops. The resulting code is often both smaller
1824 and faster, but since the function calls no longer appear as such, you
1825 cannot set a breakpoint on those calls, nor can you change the behavior
1826 of the functions by linking with a different library. In addition,
1827 when a function is recognized as a built-in function, GCC may use
1828 information about that function to warn about problems with calls to
1829 that function, or to generate more efficient code, even if the
1830 resulting code still contains calls to that function. For example,
1831 warnings are given with @option{-Wformat} for bad calls to
1832 @code{printf} when @code{printf} is built in and @code{strlen} is
1833 known not to modify global memory.
1835 With the @option{-fno-builtin-@var{function}} option
1836 only the built-in function @var{function} is
1837 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1838 function is named that is not built-in in this version of GCC, this
1839 option is ignored. There is no corresponding
1840 @option{-fbuiltin-@var{function}} option; if you wish to enable
1841 built-in functions selectively when using @option{-fno-builtin} or
1842 @option{-ffreestanding}, you may define macros such as:
1845 #define abs(n) __builtin_abs ((n))
1846 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1851 @cindex hosted environment
1853 Assert that compilation targets a hosted environment. This implies
1854 @option{-fbuiltin}. A hosted environment is one in which the
1855 entire standard library is available, and in which @code{main} has a return
1856 type of @code{int}. Examples are nearly everything except a kernel.
1857 This is equivalent to @option{-fno-freestanding}.
1859 @item -ffreestanding
1860 @opindex ffreestanding
1861 @cindex hosted environment
1863 Assert that compilation targets a freestanding environment. This
1864 implies @option{-fno-builtin}. A freestanding environment
1865 is one in which the standard library may not exist, and program startup may
1866 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1867 This is equivalent to @option{-fno-hosted}.
1869 @xref{Standards,,Language Standards Supported by GCC}, for details of
1870 freestanding and hosted environments.
1874 @cindex OpenMP parallel
1875 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1876 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1877 compiler generates parallel code according to the OpenMP Application
1878 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
1879 implies @option{-pthread}, and thus is only supported on targets that
1880 have support for @option{-pthread}. @option{-fopenmp} implies
1881 @option{-fopenmp-simd}.
1884 @opindex fopenmp-simd
1887 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
1888 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
1893 @cindex Enable Cilk Plus
1894 Enable the usage of Cilk Plus language extension features for C/C++.
1895 When the option @option{-fcilkplus} is specified, enable the usage of
1896 the Cilk Plus Language extension features for C/C++. The present
1897 implementation follows ABI version 1.2. This is an experimental
1898 feature that is only partially complete, and whose interface may
1899 change in future versions of GCC as the official specification
1900 changes. Currently, all features but @code{_Cilk_for} have been
1905 When the option @option{-fgnu-tm} is specified, the compiler
1906 generates code for the Linux variant of Intel's current Transactional
1907 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1908 an experimental feature whose interface may change in future versions
1909 of GCC, as the official specification changes. Please note that not
1910 all architectures are supported for this feature.
1912 For more information on GCC's support for transactional memory,
1913 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1914 Transactional Memory Library}.
1916 Note that the transactional memory feature is not supported with
1917 non-call exceptions (@option{-fnon-call-exceptions}).
1919 @item -fms-extensions
1920 @opindex fms-extensions
1921 Accept some non-standard constructs used in Microsoft header files.
1923 In C++ code, this allows member names in structures to be similar
1924 to previous types declarations.
1933 Some cases of unnamed fields in structures and unions are only
1934 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1935 fields within structs/unions}, for details.
1937 Note that this option is off for all targets but i?86 and x86_64
1938 targets using ms-abi.
1939 @item -fplan9-extensions
1940 Accept some non-standard constructs used in Plan 9 code.
1942 This enables @option{-fms-extensions}, permits passing pointers to
1943 structures with anonymous fields to functions that expect pointers to
1944 elements of the type of the field, and permits referring to anonymous
1945 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1946 struct/union fields within structs/unions}, for details. This is only
1947 supported for C, not C++.
1951 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1952 options for strict ISO C conformance) implies @option{-trigraphs}.
1954 @cindex traditional C language
1955 @cindex C language, traditional
1957 @itemx -traditional-cpp
1958 @opindex traditional-cpp
1959 @opindex traditional
1960 Formerly, these options caused GCC to attempt to emulate a pre-standard
1961 C compiler. They are now only supported with the @option{-E} switch.
1962 The preprocessor continues to support a pre-standard mode. See the GNU
1963 CPP manual for details.
1965 @item -fcond-mismatch
1966 @opindex fcond-mismatch
1967 Allow conditional expressions with mismatched types in the second and
1968 third arguments. The value of such an expression is void. This option
1969 is not supported for C++.
1971 @item -flax-vector-conversions
1972 @opindex flax-vector-conversions
1973 Allow implicit conversions between vectors with differing numbers of
1974 elements and/or incompatible element types. This option should not be
1977 @item -funsigned-char
1978 @opindex funsigned-char
1979 Let the type @code{char} be unsigned, like @code{unsigned char}.
1981 Each kind of machine has a default for what @code{char} should
1982 be. It is either like @code{unsigned char} by default or like
1983 @code{signed char} by default.
1985 Ideally, a portable program should always use @code{signed char} or
1986 @code{unsigned char} when it depends on the signedness of an object.
1987 But many programs have been written to use plain @code{char} and
1988 expect it to be signed, or expect it to be unsigned, depending on the
1989 machines they were written for. This option, and its inverse, let you
1990 make such a program work with the opposite default.
1992 The type @code{char} is always a distinct type from each of
1993 @code{signed char} or @code{unsigned char}, even though its behavior
1994 is always just like one of those two.
1997 @opindex fsigned-char
1998 Let the type @code{char} be signed, like @code{signed char}.
2000 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2001 the negative form of @option{-funsigned-char}. Likewise, the option
2002 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2004 @item -fsigned-bitfields
2005 @itemx -funsigned-bitfields
2006 @itemx -fno-signed-bitfields
2007 @itemx -fno-unsigned-bitfields
2008 @opindex fsigned-bitfields
2009 @opindex funsigned-bitfields
2010 @opindex fno-signed-bitfields
2011 @opindex fno-unsigned-bitfields
2012 These options control whether a bit-field is signed or unsigned, when the
2013 declaration does not use either @code{signed} or @code{unsigned}. By
2014 default, such a bit-field is signed, because this is consistent: the
2015 basic integer types such as @code{int} are signed types.
2018 @node C++ Dialect Options
2019 @section Options Controlling C++ Dialect
2021 @cindex compiler options, C++
2022 @cindex C++ options, command-line
2023 @cindex options, C++
2024 This section describes the command-line options that are only meaningful
2025 for C++ programs. You can also use most of the GNU compiler options
2026 regardless of what language your program is in. For example, you
2027 might compile a file @code{firstClass.C} like this:
2030 g++ -g -frepo -O -c firstClass.C
2034 In this example, only @option{-frepo} is an option meant
2035 only for C++ programs; you can use the other options with any
2036 language supported by GCC@.
2038 Here is a list of options that are @emph{only} for compiling C++ programs:
2042 @item -fabi-version=@var{n}
2043 @opindex fabi-version
2044 Use version @var{n} of the C++ ABI@. The default is version 0.
2046 Version 0 refers to the version conforming most closely to
2047 the C++ ABI specification. Therefore, the ABI obtained using version 0
2048 will change in different versions of G++ as ABI bugs are fixed.
2050 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2052 Version 2 is the version of the C++ ABI that first appeared in G++
2053 3.4, and was the default through G++ 4.9.
2055 Version 3 corrects an error in mangling a constant address as a
2058 Version 4, which first appeared in G++ 4.5, implements a standard
2059 mangling for vector types.
2061 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2062 attribute const/volatile on function pointer types, decltype of a
2063 plain decl, and use of a function parameter in the declaration of
2066 Version 6, which first appeared in G++ 4.7, corrects the promotion
2067 behavior of C++11 scoped enums and the mangling of template argument
2068 packs, const/static_cast, prefix ++ and --, and a class scope function
2069 used as a template argument.
2071 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2072 builtin type and corrects the mangling of lambdas in default argument
2075 Version 8, which first appeared in G++ 4.9, corrects the substitution
2076 behavior of function types with function-cv-qualifiers.
2078 See also @option{-Wabi}.
2080 @item -fabi-compat-version=@var{n}
2081 @opindex fabi-compat-version
2082 Starting with GCC 4.5, on targets that support strong aliases, G++
2083 works around mangling changes by creating an alias with the correct
2084 mangled name when defining a symbol with an incorrect mangled name.
2085 This switch specifies which ABI version to use for the alias.
2087 With @option{-fabi-version=0} (the default), this defaults to 2. If
2088 another ABI version is explicitly selected, this defaults to 0.
2090 The compatibility version is also set by @option{-Wabi=@var{n}}.
2092 @item -fno-access-control
2093 @opindex fno-access-control
2094 Turn off all access checking. This switch is mainly useful for working
2095 around bugs in the access control code.
2099 Check that the pointer returned by @code{operator new} is non-null
2100 before attempting to modify the storage allocated. This check is
2101 normally unnecessary because the C++ standard specifies that
2102 @code{operator new} only returns @code{0} if it is declared
2103 @samp{throw()}, in which case the compiler always checks the
2104 return value even without this option. In all other cases, when
2105 @code{operator new} has a non-empty exception specification, memory
2106 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2107 @samp{new (nothrow)}.
2109 @item -fconstexpr-depth=@var{n}
2110 @opindex fconstexpr-depth
2111 Set the maximum nested evaluation depth for C++11 constexpr functions
2112 to @var{n}. A limit is needed to detect endless recursion during
2113 constant expression evaluation. The minimum specified by the standard
2116 @item -fdeduce-init-list
2117 @opindex fdeduce-init-list
2118 Enable deduction of a template type parameter as
2119 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2122 template <class T> auto forward(T t) -> decltype (realfn (t))
2129 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2133 This deduction was implemented as a possible extension to the
2134 originally proposed semantics for the C++11 standard, but was not part
2135 of the final standard, so it is disabled by default. This option is
2136 deprecated, and may be removed in a future version of G++.
2138 @item -ffriend-injection
2139 @opindex ffriend-injection
2140 Inject friend functions into the enclosing namespace, so that they are
2141 visible outside the scope of the class in which they are declared.
2142 Friend functions were documented to work this way in the old Annotated
2143 C++ Reference Manual, and versions of G++ before 4.1 always worked
2144 that way. However, in ISO C++ a friend function that is not declared
2145 in an enclosing scope can only be found using argument dependent
2146 lookup. This option causes friends to be injected as they were in
2149 This option is for compatibility, and may be removed in a future
2152 @item -fno-elide-constructors
2153 @opindex fno-elide-constructors
2154 The C++ standard allows an implementation to omit creating a temporary
2155 that is only used to initialize another object of the same type.
2156 Specifying this option disables that optimization, and forces G++ to
2157 call the copy constructor in all cases.
2159 @item -fno-enforce-eh-specs
2160 @opindex fno-enforce-eh-specs
2161 Don't generate code to check for violation of exception specifications
2162 at run time. This option violates the C++ standard, but may be useful
2163 for reducing code size in production builds, much like defining
2164 @samp{NDEBUG}. This does not give user code permission to throw
2165 exceptions in violation of the exception specifications; the compiler
2166 still optimizes based on the specifications, so throwing an
2167 unexpected exception results in undefined behavior at run time.
2169 @item -fextern-tls-init
2170 @itemx -fno-extern-tls-init
2171 @opindex fextern-tls-init
2172 @opindex fno-extern-tls-init
2173 The C++11 and OpenMP standards allow @samp{thread_local} and
2174 @samp{threadprivate} variables to have dynamic (runtime)
2175 initialization. To support this, any use of such a variable goes
2176 through a wrapper function that performs any necessary initialization.
2177 When the use and definition of the variable are in the same
2178 translation unit, this overhead can be optimized away, but when the
2179 use is in a different translation unit there is significant overhead
2180 even if the variable doesn't actually need dynamic initialization. If
2181 the programmer can be sure that no use of the variable in a
2182 non-defining TU needs to trigger dynamic initialization (either
2183 because the variable is statically initialized, or a use of the
2184 variable in the defining TU will be executed before any uses in
2185 another TU), they can avoid this overhead with the
2186 @option{-fno-extern-tls-init} option.
2188 On targets that support symbol aliases, the default is
2189 @option{-fextern-tls-init}. On targets that do not support symbol
2190 aliases, the default is @option{-fno-extern-tls-init}.
2193 @itemx -fno-for-scope
2195 @opindex fno-for-scope
2196 If @option{-ffor-scope} is specified, the scope of variables declared in
2197 a @i{for-init-statement} is limited to the @samp{for} loop itself,
2198 as specified by the C++ standard.
2199 If @option{-fno-for-scope} is specified, the scope of variables declared in
2200 a @i{for-init-statement} extends to the end of the enclosing scope,
2201 as was the case in old versions of G++, and other (traditional)
2202 implementations of C++.
2204 If neither flag is given, the default is to follow the standard,
2205 but to allow and give a warning for old-style code that would
2206 otherwise be invalid, or have different behavior.
2208 @item -fno-gnu-keywords
2209 @opindex fno-gnu-keywords
2210 Do not recognize @code{typeof} as a keyword, so that code can use this
2211 word as an identifier. You can use the keyword @code{__typeof__} instead.
2212 @option{-ansi} implies @option{-fno-gnu-keywords}.
2214 @item -fno-implicit-templates
2215 @opindex fno-implicit-templates
2216 Never emit code for non-inline templates that are instantiated
2217 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2218 @xref{Template Instantiation}, for more information.
2220 @item -fno-implicit-inline-templates
2221 @opindex fno-implicit-inline-templates
2222 Don't emit code for implicit instantiations of inline templates, either.
2223 The default is to handle inlines differently so that compiles with and
2224 without optimization need the same set of explicit instantiations.
2226 @item -fno-implement-inlines
2227 @opindex fno-implement-inlines
2228 To save space, do not emit out-of-line copies of inline functions
2229 controlled by @samp{#pragma implementation}. This causes linker
2230 errors if these functions are not inlined everywhere they are called.
2232 @item -fms-extensions
2233 @opindex fms-extensions
2234 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2235 int and getting a pointer to member function via non-standard syntax.
2237 @item -fno-nonansi-builtins
2238 @opindex fno-nonansi-builtins
2239 Disable built-in declarations of functions that are not mandated by
2240 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2241 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2244 @opindex fnothrow-opt
2245 Treat a @code{throw()} exception specification as if it were a
2246 @code{noexcept} specification to reduce or eliminate the text size
2247 overhead relative to a function with no exception specification. If
2248 the function has local variables of types with non-trivial
2249 destructors, the exception specification actually makes the
2250 function smaller because the EH cleanups for those variables can be
2251 optimized away. The semantic effect is that an exception thrown out of
2252 a function with such an exception specification results in a call
2253 to @code{terminate} rather than @code{unexpected}.
2255 @item -fno-operator-names
2256 @opindex fno-operator-names
2257 Do not treat the operator name keywords @code{and}, @code{bitand},
2258 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2259 synonyms as keywords.
2261 @item -fno-optional-diags
2262 @opindex fno-optional-diags
2263 Disable diagnostics that the standard says a compiler does not need to
2264 issue. Currently, the only such diagnostic issued by G++ is the one for
2265 a name having multiple meanings within a class.
2268 @opindex fpermissive
2269 Downgrade some diagnostics about nonconformant code from errors to
2270 warnings. Thus, using @option{-fpermissive} allows some
2271 nonconforming code to compile.
2273 @item -fno-pretty-templates
2274 @opindex fno-pretty-templates
2275 When an error message refers to a specialization of a function
2276 template, the compiler normally prints the signature of the
2277 template followed by the template arguments and any typedefs or
2278 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2279 rather than @code{void f(int)}) so that it's clear which template is
2280 involved. When an error message refers to a specialization of a class
2281 template, the compiler omits any template arguments that match
2282 the default template arguments for that template. If either of these
2283 behaviors make it harder to understand the error message rather than
2284 easier, you can use @option{-fno-pretty-templates} to disable them.
2288 Enable automatic template instantiation at link time. This option also
2289 implies @option{-fno-implicit-templates}. @xref{Template
2290 Instantiation}, for more information.
2294 Disable generation of information about every class with virtual
2295 functions for use by the C++ run-time type identification features
2296 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2297 of the language, you can save some space by using this flag. Note that
2298 exception handling uses the same information, but G++ generates it as
2299 needed. The @samp{dynamic_cast} operator can still be used for casts that
2300 do not require run-time type information, i.e.@: casts to @code{void *} or to
2301 unambiguous base classes.
2305 Emit statistics about front-end processing at the end of the compilation.
2306 This information is generally only useful to the G++ development team.
2308 @item -fstrict-enums
2309 @opindex fstrict-enums
2310 Allow the compiler to optimize using the assumption that a value of
2311 enumerated type can only be one of the values of the enumeration (as
2312 defined in the C++ standard; basically, a value that can be
2313 represented in the minimum number of bits needed to represent all the
2314 enumerators). This assumption may not be valid if the program uses a
2315 cast to convert an arbitrary integer value to the enumerated type.
2317 @item -ftemplate-backtrace-limit=@var{n}
2318 @opindex ftemplate-backtrace-limit
2319 Set the maximum number of template instantiation notes for a single
2320 warning or error to @var{n}. The default value is 10.
2322 @item -ftemplate-depth=@var{n}
2323 @opindex ftemplate-depth
2324 Set the maximum instantiation depth for template classes to @var{n}.
2325 A limit on the template instantiation depth is needed to detect
2326 endless recursions during template class instantiation. ANSI/ISO C++
2327 conforming programs must not rely on a maximum depth greater than 17
2328 (changed to 1024 in C++11). The default value is 900, as the compiler
2329 can run out of stack space before hitting 1024 in some situations.
2331 @item -fno-threadsafe-statics
2332 @opindex fno-threadsafe-statics
2333 Do not emit the extra code to use the routines specified in the C++
2334 ABI for thread-safe initialization of local statics. You can use this
2335 option to reduce code size slightly in code that doesn't need to be
2338 @item -fuse-cxa-atexit
2339 @opindex fuse-cxa-atexit
2340 Register destructors for objects with static storage duration with the
2341 @code{__cxa_atexit} function rather than the @code{atexit} function.
2342 This option is required for fully standards-compliant handling of static
2343 destructors, but only works if your C library supports
2344 @code{__cxa_atexit}.
2346 @item -fno-use-cxa-get-exception-ptr
2347 @opindex fno-use-cxa-get-exception-ptr
2348 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2349 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2350 if the runtime routine is not available.
2352 @item -fvisibility-inlines-hidden
2353 @opindex fvisibility-inlines-hidden
2354 This switch declares that the user does not attempt to compare
2355 pointers to inline functions or methods where the addresses of the two functions
2356 are taken in different shared objects.
2358 The effect of this is that GCC may, effectively, mark inline methods with
2359 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2360 appear in the export table of a DSO and do not require a PLT indirection
2361 when used within the DSO@. Enabling this option can have a dramatic effect
2362 on load and link times of a DSO as it massively reduces the size of the
2363 dynamic export table when the library makes heavy use of templates.
2365 The behavior of this switch is not quite the same as marking the
2366 methods as hidden directly, because it does not affect static variables
2367 local to the function or cause the compiler to deduce that
2368 the function is defined in only one shared object.
2370 You may mark a method as having a visibility explicitly to negate the
2371 effect of the switch for that method. For example, if you do want to
2372 compare pointers to a particular inline method, you might mark it as
2373 having default visibility. Marking the enclosing class with explicit
2374 visibility has no effect.
2376 Explicitly instantiated inline methods are unaffected by this option
2377 as their linkage might otherwise cross a shared library boundary.
2378 @xref{Template Instantiation}.
2380 @item -fvisibility-ms-compat
2381 @opindex fvisibility-ms-compat
2382 This flag attempts to use visibility settings to make GCC's C++
2383 linkage model compatible with that of Microsoft Visual Studio.
2385 The flag makes these changes to GCC's linkage model:
2389 It sets the default visibility to @code{hidden}, like
2390 @option{-fvisibility=hidden}.
2393 Types, but not their members, are not hidden by default.
2396 The One Definition Rule is relaxed for types without explicit
2397 visibility specifications that are defined in more than one
2398 shared object: those declarations are permitted if they are
2399 permitted when this option is not used.
2402 In new code it is better to use @option{-fvisibility=hidden} and
2403 export those classes that are intended to be externally visible.
2404 Unfortunately it is possible for code to rely, perhaps accidentally,
2405 on the Visual Studio behavior.
2407 Among the consequences of these changes are that static data members
2408 of the same type with the same name but defined in different shared
2409 objects are different, so changing one does not change the other;
2410 and that pointers to function members defined in different shared
2411 objects may not compare equal. When this flag is given, it is a
2412 violation of the ODR to define types with the same name differently.
2414 @item -fvtable-verify=@var{std|preinit|none}
2415 @opindex fvtable-verify
2416 Turn on (or off, if using @option{-fvtable-verify=none}) the security
2417 feature that verifies at runtime, for every virtual call that is made, that
2418 the vtable pointer through which the call is made is valid for the type of
2419 the object, and has not been corrupted or overwritten. If an invalid vtable
2420 pointer is detected (at runtime), an error is reported and execution of the
2421 program is immediately halted.
2423 This option causes runtime data structures to be built, at program start up,
2424 for verifying the vtable pointers. The options @code{std} and @code{preinit}
2425 control the timing of when these data structures are built. In both cases the
2426 data structures are built before execution reaches 'main'. The
2427 @option{-fvtable-verify=std} causes these data structure to be built after the
2428 shared libraries have been loaded and initialized.
2429 @option{-fvtable-verify=preinit} causes them to be built before the shared
2430 libraries have been loaded and initialized.
2432 If this option appears multiple times in the compiler line, with different
2433 values specified, 'none' will take highest priority over both 'std' and
2434 'preinit'; 'preinit' will take priority over 'std'.
2437 @opindex (fvtv-debug)
2438 Causes debug versions of the runtime functions for the vtable verification
2439 feature to be called. This assumes the @option{-fvtable-verify=std} or
2440 @option{-fvtable-verify=preinit} has been used. This flag will also cause the
2441 compiler to keep track of which vtable pointers it found for each class, and
2442 record that information in the file ``vtv_set_ptr_data.log'', in the dump
2443 file directory on the user's machine.
2445 Note: This feature APPENDS data to the log file. If you want a fresh log
2446 file, be sure to delete any existing one.
2449 @opindex fvtv-counts
2450 This is a debugging flag. When used in conjunction with
2451 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
2452 causes the compiler to keep track of the total number of virtual calls
2453 it encountered and the number of verifications it inserted. It also
2454 counts the number of calls to certain runtime library functions
2455 that it inserts. This information, for each compilation unit, is written
2456 to a file named ``vtv_count_data.log'', in the dump_file directory on
2457 the user's machine. It also counts the size of the vtable pointer sets
2458 for each class, and writes this information to ``vtv_class_set_sizes.log''
2459 in the same directory.
2461 Note: This feature APPENDS data to the log files. To get a fresh log
2462 files, be sure to delete any existing ones.
2466 Do not use weak symbol support, even if it is provided by the linker.
2467 By default, G++ uses weak symbols if they are available. This
2468 option exists only for testing, and should not be used by end-users;
2469 it results in inferior code and has no benefits. This option may
2470 be removed in a future release of G++.
2474 Do not search for header files in the standard directories specific to
2475 C++, but do still search the other standard directories. (This option
2476 is used when building the C++ library.)
2479 In addition, these optimization, warning, and code generation options
2480 have meanings only for C++ programs:
2483 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2486 When an explicit @option{-fabi-version=@var{n}} option is used, causes
2487 G++ to warn when it generates code that is probably not compatible with the
2488 vendor-neutral C++ ABI@. Since G++ now defaults to
2489 @option{-fabi-version=0}, @option{-Wabi} has no effect unless either
2490 an older ABI version is selected (with @option{-fabi-version=@var{n}})
2491 or an older compatibility version is selected (with
2492 @option{-Wabi=@var{n}} or @option{-fabi-compat-version=@var{n}}).
2494 Although an effort has been made to warn about
2495 all such cases, there are probably some cases that are not warned about,
2496 even though G++ is generating incompatible code. There may also be
2497 cases where warnings are emitted even though the code that is generated
2500 You should rewrite your code to avoid these warnings if you are
2501 concerned about the fact that code generated by G++ may not be binary
2502 compatible with code generated by other compilers.
2504 @option{-Wabi} can also be used with an explicit version number to
2505 warn about compatibility with a particular @option{-fabi-version}
2506 level, e.g. @option{-Wabi=2} to warn about changes relative to
2507 @option{-fabi-version=2}. Specifying a version number also sets
2508 @option{-fabi-compat-version=@var{n}}.
2510 The known incompatibilities in @option{-fabi-version=2} (which was the
2511 default from GCC 3.4 to 4.9) include:
2516 A template with a non-type template parameter of reference type was
2517 mangled incorrectly:
2520 template <int &> struct S @{@};
2524 This was fixed in @option{-fabi-version=3}.
2527 SIMD vector types declared using @code{__attribute ((vector_size))} were
2528 mangled in a non-standard way that does not allow for overloading of
2529 functions taking vectors of different sizes.
2531 The mangling was changed in @option{-fabi-version=4}.
2534 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2535 qualifiers, and @code{decltype} of a plain declaration was folded away.
2537 These mangling issues were fixed in @option{-fabi-version=5}.
2540 Scoped enumerators passed as arguments to a variadic function are
2541 promoted like unscoped enumerators, causing @samp{va_arg} to complain.
2542 On most targets this does not actually affect the parameter passing
2543 ABI, as there is no way to pass an argument smaller than @samp{int}.
2545 Also, the ABI changed the mangling of template argument packs,
2546 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2547 a class scope function used as a template argument.
2549 These issues were corrected in @option{-fabi-version=6}.
2552 Lambdas in default argument scope were mangled incorrectly, and the
2553 ABI changed the mangling of nullptr_t.
2555 These issues were corrected in @option{-fabi-version=7}.
2558 When mangling a function type with function-cv-qualifiers, the
2559 un-qualified function type was incorrectly treated as a substitution
2562 This was fixed in @option{-fabi-version=8}.
2565 It also warns about psABI-related changes. The known psABI changes at this
2571 For SysV/x86-64, unions with @code{long double} members are
2572 passed in memory as specified in psABI. For example:
2582 @code{union U} is always passed in memory.
2586 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2587 @opindex Wctor-dtor-privacy
2588 @opindex Wno-ctor-dtor-privacy
2589 Warn when a class seems unusable because all the constructors or
2590 destructors in that class are private, and it has neither friends nor
2591 public static member functions. Also warn if there are no non-private
2592 methods, and there's at least one private member function that isn't
2593 a constructor or destructor.
2595 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2596 @opindex Wdelete-non-virtual-dtor
2597 @opindex Wno-delete-non-virtual-dtor
2598 Warn when @samp{delete} is used to destroy an instance of a class that
2599 has virtual functions and non-virtual destructor. It is unsafe to delete
2600 an instance of a derived class through a pointer to a base class if the
2601 base class does not have a virtual destructor. This warning is enabled
2604 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2605 @opindex Wliteral-suffix
2606 @opindex Wno-literal-suffix
2607 Warn when a string or character literal is followed by a ud-suffix which does
2608 not begin with an underscore. As a conforming extension, GCC treats such
2609 suffixes as separate preprocessing tokens in order to maintain backwards
2610 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2614 #define __STDC_FORMAT_MACROS
2615 #include <inttypes.h>
2620 printf("My int64: %"PRId64"\n", i64);
2624 In this case, @code{PRId64} is treated as a separate preprocessing token.
2626 This warning is enabled by default.
2628 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2630 @opindex Wno-narrowing
2631 Warn when a narrowing conversion prohibited by C++11 occurs within
2635 int i = @{ 2.2 @}; // error: narrowing from double to int
2638 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2640 With @option{-std=c++11}, @option{-Wno-narrowing} suppresses for
2641 non-constants the diagnostic required by the standard. Note that this
2642 does not affect the meaning of well-formed code; narrowing conversions
2643 are still considered ill-formed in SFINAE context.
2645 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2647 @opindex Wno-noexcept
2648 Warn when a noexcept-expression evaluates to false because of a call
2649 to a function that does not have a non-throwing exception
2650 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2651 the compiler to never throw an exception.
2653 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2654 @opindex Wnon-virtual-dtor
2655 @opindex Wno-non-virtual-dtor
2656 Warn when a class has virtual functions and an accessible non-virtual
2657 destructor itself or in an accessible polymorphic base class, in which
2658 case it is possible but unsafe to delete an instance of a derived
2659 class through a pointer to the class itself or base class. This
2660 warning is automatically enabled if @option{-Weffc++} is specified.
2662 @item -Wreorder @r{(C++ and Objective-C++ only)}
2664 @opindex Wno-reorder
2665 @cindex reordering, warning
2666 @cindex warning for reordering of member initializers
2667 Warn when the order of member initializers given in the code does not
2668 match the order in which they must be executed. For instance:
2674 A(): j (0), i (1) @{ @}
2679 The compiler rearranges the member initializers for @samp{i}
2680 and @samp{j} to match the declaration order of the members, emitting
2681 a warning to that effect. This warning is enabled by @option{-Wall}.
2683 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2684 @opindex fext-numeric-literals
2685 @opindex fno-ext-numeric-literals
2686 Accept imaginary, fixed-point, or machine-defined
2687 literal number suffixes as GNU extensions.
2688 When this option is turned off these suffixes are treated
2689 as C++11 user-defined literal numeric suffixes.
2690 This is on by default for all pre-C++11 dialects and all GNU dialects:
2691 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2692 @option{-std=gnu++14}.
2693 This option is off by default
2694 for ISO C++11 onwards (@option{-std=c++11}, ...).
2697 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2700 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2703 Warn about violations of the following style guidelines from Scott Meyers'
2704 @cite{Effective C++} series of books:
2708 Define a copy constructor and an assignment operator for classes
2709 with dynamically-allocated memory.
2712 Prefer initialization to assignment in constructors.
2715 Have @code{operator=} return a reference to @code{*this}.
2718 Don't try to return a reference when you must return an object.
2721 Distinguish between prefix and postfix forms of increment and
2722 decrement operators.
2725 Never overload @code{&&}, @code{||}, or @code{,}.
2729 This option also enables @option{-Wnon-virtual-dtor}, which is also
2730 one of the effective C++ recommendations. However, the check is
2731 extended to warn about the lack of virtual destructor in accessible
2732 non-polymorphic bases classes too.
2734 When selecting this option, be aware that the standard library
2735 headers do not obey all of these guidelines; use @samp{grep -v}
2736 to filter out those warnings.
2738 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2739 @opindex Wstrict-null-sentinel
2740 @opindex Wno-strict-null-sentinel
2741 Warn about the use of an uncasted @code{NULL} as sentinel. When
2742 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2743 to @code{__null}. Although it is a null pointer constant rather than a
2744 null pointer, it is guaranteed to be of the same size as a pointer.
2745 But this use is not portable across different compilers.
2747 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2748 @opindex Wno-non-template-friend
2749 @opindex Wnon-template-friend
2750 Disable warnings when non-templatized friend functions are declared
2751 within a template. Since the advent of explicit template specification
2752 support in G++, if the name of the friend is an unqualified-id (i.e.,
2753 @samp{friend foo(int)}), the C++ language specification demands that the
2754 friend declare or define an ordinary, nontemplate function. (Section
2755 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2756 could be interpreted as a particular specialization of a templatized
2757 function. Because this non-conforming behavior is no longer the default
2758 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2759 check existing code for potential trouble spots and is on by default.
2760 This new compiler behavior can be turned off with
2761 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2762 but disables the helpful warning.
2764 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2765 @opindex Wold-style-cast
2766 @opindex Wno-old-style-cast
2767 Warn if an old-style (C-style) cast to a non-void type is used within
2768 a C++ program. The new-style casts (@samp{dynamic_cast},
2769 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2770 less vulnerable to unintended effects and much easier to search for.
2772 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2773 @opindex Woverloaded-virtual
2774 @opindex Wno-overloaded-virtual
2775 @cindex overloaded virtual function, warning
2776 @cindex warning for overloaded virtual function
2777 Warn when a function declaration hides virtual functions from a
2778 base class. For example, in:
2785 struct B: public A @{
2790 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2801 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2802 @opindex Wno-pmf-conversions
2803 @opindex Wpmf-conversions
2804 Disable the diagnostic for converting a bound pointer to member function
2807 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2808 @opindex Wsign-promo
2809 @opindex Wno-sign-promo
2810 Warn when overload resolution chooses a promotion from unsigned or
2811 enumerated type to a signed type, over a conversion to an unsigned type of
2812 the same size. Previous versions of G++ tried to preserve
2813 unsignedness, but the standard mandates the current behavior.
2816 @node Objective-C and Objective-C++ Dialect Options
2817 @section Options Controlling Objective-C and Objective-C++ Dialects
2819 @cindex compiler options, Objective-C and Objective-C++
2820 @cindex Objective-C and Objective-C++ options, command-line
2821 @cindex options, Objective-C and Objective-C++
2822 (NOTE: This manual does not describe the Objective-C and Objective-C++
2823 languages themselves. @xref{Standards,,Language Standards
2824 Supported by GCC}, for references.)
2826 This section describes the command-line options that are only meaningful
2827 for Objective-C and Objective-C++ programs. You can also use most of
2828 the language-independent GNU compiler options.
2829 For example, you might compile a file @code{some_class.m} like this:
2832 gcc -g -fgnu-runtime -O -c some_class.m
2836 In this example, @option{-fgnu-runtime} is an option meant only for
2837 Objective-C and Objective-C++ programs; you can use the other options with
2838 any language supported by GCC@.
2840 Note that since Objective-C is an extension of the C language, Objective-C
2841 compilations may also use options specific to the C front-end (e.g.,
2842 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2843 C++-specific options (e.g., @option{-Wabi}).
2845 Here is a list of options that are @emph{only} for compiling Objective-C
2846 and Objective-C++ programs:
2849 @item -fconstant-string-class=@var{class-name}
2850 @opindex fconstant-string-class
2851 Use @var{class-name} as the name of the class to instantiate for each
2852 literal string specified with the syntax @code{@@"@dots{}"}. The default
2853 class name is @code{NXConstantString} if the GNU runtime is being used, and
2854 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2855 @option{-fconstant-cfstrings} option, if also present, overrides the
2856 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2857 to be laid out as constant CoreFoundation strings.
2860 @opindex fgnu-runtime
2861 Generate object code compatible with the standard GNU Objective-C
2862 runtime. This is the default for most types of systems.
2864 @item -fnext-runtime
2865 @opindex fnext-runtime
2866 Generate output compatible with the NeXT runtime. This is the default
2867 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2868 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2871 @item -fno-nil-receivers
2872 @opindex fno-nil-receivers
2873 Assume that all Objective-C message dispatches (@code{[receiver
2874 message:arg]}) in this translation unit ensure that the receiver is
2875 not @code{nil}. This allows for more efficient entry points in the
2876 runtime to be used. This option is only available in conjunction with
2877 the NeXT runtime and ABI version 0 or 1.
2879 @item -fobjc-abi-version=@var{n}
2880 @opindex fobjc-abi-version
2881 Use version @var{n} of the Objective-C ABI for the selected runtime.
2882 This option is currently supported only for the NeXT runtime. In that
2883 case, Version 0 is the traditional (32-bit) ABI without support for
2884 properties and other Objective-C 2.0 additions. Version 1 is the
2885 traditional (32-bit) ABI with support for properties and other
2886 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2887 nothing is specified, the default is Version 0 on 32-bit target
2888 machines, and Version 2 on 64-bit target machines.
2890 @item -fobjc-call-cxx-cdtors
2891 @opindex fobjc-call-cxx-cdtors
2892 For each Objective-C class, check if any of its instance variables is a
2893 C++ object with a non-trivial default constructor. If so, synthesize a
2894 special @code{- (id) .cxx_construct} instance method which runs
2895 non-trivial default constructors on any such instance variables, in order,
2896 and then return @code{self}. Similarly, check if any instance variable
2897 is a C++ object with a non-trivial destructor, and if so, synthesize a
2898 special @code{- (void) .cxx_destruct} method which runs
2899 all such default destructors, in reverse order.
2901 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2902 methods thusly generated only operate on instance variables
2903 declared in the current Objective-C class, and not those inherited
2904 from superclasses. It is the responsibility of the Objective-C
2905 runtime to invoke all such methods in an object's inheritance
2906 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
2907 by the runtime immediately after a new object instance is allocated;
2908 the @code{- (void) .cxx_destruct} methods are invoked immediately
2909 before the runtime deallocates an object instance.
2911 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2912 support for invoking the @code{- (id) .cxx_construct} and
2913 @code{- (void) .cxx_destruct} methods.
2915 @item -fobjc-direct-dispatch
2916 @opindex fobjc-direct-dispatch
2917 Allow fast jumps to the message dispatcher. On Darwin this is
2918 accomplished via the comm page.
2920 @item -fobjc-exceptions
2921 @opindex fobjc-exceptions
2922 Enable syntactic support for structured exception handling in
2923 Objective-C, similar to what is offered by C++ and Java. This option
2924 is required to use the Objective-C keywords @code{@@try},
2925 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2926 @code{@@synchronized}. This option is available with both the GNU
2927 runtime and the NeXT runtime (but not available in conjunction with
2928 the NeXT runtime on Mac OS X 10.2 and earlier).
2932 Enable garbage collection (GC) in Objective-C and Objective-C++
2933 programs. This option is only available with the NeXT runtime; the
2934 GNU runtime has a different garbage collection implementation that
2935 does not require special compiler flags.
2937 @item -fobjc-nilcheck
2938 @opindex fobjc-nilcheck
2939 For the NeXT runtime with version 2 of the ABI, check for a nil
2940 receiver in method invocations before doing the actual method call.
2941 This is the default and can be disabled using
2942 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2943 checked for nil in this way no matter what this flag is set to.
2944 Currently this flag does nothing when the GNU runtime, or an older
2945 version of the NeXT runtime ABI, is used.
2947 @item -fobjc-std=objc1
2949 Conform to the language syntax of Objective-C 1.0, the language
2950 recognized by GCC 4.0. This only affects the Objective-C additions to
2951 the C/C++ language; it does not affect conformance to C/C++ standards,
2952 which is controlled by the separate C/C++ dialect option flags. When
2953 this option is used with the Objective-C or Objective-C++ compiler,
2954 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2955 This is useful if you need to make sure that your Objective-C code can
2956 be compiled with older versions of GCC@.
2958 @item -freplace-objc-classes
2959 @opindex freplace-objc-classes
2960 Emit a special marker instructing @command{ld(1)} not to statically link in
2961 the resulting object file, and allow @command{dyld(1)} to load it in at
2962 run time instead. This is used in conjunction with the Fix-and-Continue
2963 debugging mode, where the object file in question may be recompiled and
2964 dynamically reloaded in the course of program execution, without the need
2965 to restart the program itself. Currently, Fix-and-Continue functionality
2966 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2971 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2972 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2973 compile time) with static class references that get initialized at load time,
2974 which improves run-time performance. Specifying the @option{-fzero-link} flag
2975 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2976 to be retained. This is useful in Zero-Link debugging mode, since it allows
2977 for individual class implementations to be modified during program execution.
2978 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2979 regardless of command-line options.
2981 @item -fno-local-ivars
2982 @opindex fno-local-ivars
2983 @opindex flocal-ivars
2984 By default instance variables in Objective-C can be accessed as if
2985 they were local variables from within the methods of the class they're
2986 declared in. This can lead to shadowing between instance variables
2987 and other variables declared either locally inside a class method or
2988 globally with the same name. Specifying the @option{-fno-local-ivars}
2989 flag disables this behavior thus avoiding variable shadowing issues.
2991 @item -fivar-visibility=@var{public|protected|private|package}
2992 @opindex fivar-visibility
2993 Set the default instance variable visibility to the specified option
2994 so that instance variables declared outside the scope of any access
2995 modifier directives default to the specified visibility.
2999 Dump interface declarations for all classes seen in the source file to a
3000 file named @file{@var{sourcename}.decl}.
3002 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3003 @opindex Wassign-intercept
3004 @opindex Wno-assign-intercept
3005 Warn whenever an Objective-C assignment is being intercepted by the
3008 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3009 @opindex Wno-protocol
3011 If a class is declared to implement a protocol, a warning is issued for
3012 every method in the protocol that is not implemented by the class. The
3013 default behavior is to issue a warning for every method not explicitly
3014 implemented in the class, even if a method implementation is inherited
3015 from the superclass. If you use the @option{-Wno-protocol} option, then
3016 methods inherited from the superclass are considered to be implemented,
3017 and no warning is issued for them.
3019 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3021 @opindex Wno-selector
3022 Warn if multiple methods of different types for the same selector are
3023 found during compilation. The check is performed on the list of methods
3024 in the final stage of compilation. Additionally, a check is performed
3025 for each selector appearing in a @code{@@selector(@dots{})}
3026 expression, and a corresponding method for that selector has been found
3027 during compilation. Because these checks scan the method table only at
3028 the end of compilation, these warnings are not produced if the final
3029 stage of compilation is not reached, for example because an error is
3030 found during compilation, or because the @option{-fsyntax-only} option is
3033 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3034 @opindex Wstrict-selector-match
3035 @opindex Wno-strict-selector-match
3036 Warn if multiple methods with differing argument and/or return types are
3037 found for a given selector when attempting to send a message using this
3038 selector to a receiver of type @code{id} or @code{Class}. When this flag
3039 is off (which is the default behavior), the compiler omits such warnings
3040 if any differences found are confined to types that share the same size
3043 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3044 @opindex Wundeclared-selector
3045 @opindex Wno-undeclared-selector
3046 Warn if a @code{@@selector(@dots{})} expression referring to an
3047 undeclared selector is found. A selector is considered undeclared if no
3048 method with that name has been declared before the
3049 @code{@@selector(@dots{})} expression, either explicitly in an
3050 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3051 an @code{@@implementation} section. This option always performs its
3052 checks as soon as a @code{@@selector(@dots{})} expression is found,
3053 while @option{-Wselector} only performs its checks in the final stage of
3054 compilation. This also enforces the coding style convention
3055 that methods and selectors must be declared before being used.
3057 @item -print-objc-runtime-info
3058 @opindex print-objc-runtime-info
3059 Generate C header describing the largest structure that is passed by
3064 @node Language Independent Options
3065 @section Options to Control Diagnostic Messages Formatting
3066 @cindex options to control diagnostics formatting
3067 @cindex diagnostic messages
3068 @cindex message formatting
3070 Traditionally, diagnostic messages have been formatted irrespective of
3071 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3072 options described below
3073 to control the formatting algorithm for diagnostic messages,
3074 e.g.@: how many characters per line, how often source location
3075 information should be reported. Note that some language front ends may not
3076 honor these options.
3079 @item -fmessage-length=@var{n}
3080 @opindex fmessage-length
3081 Try to format error messages so that they fit on lines of about
3082 @var{n} characters. If @var{n} is zero, then no line-wrapping will be
3083 done; each error message will appear on a single line. This is the
3084 default for all front ends.
3086 @item -fdiagnostics-show-location=once
3087 @opindex fdiagnostics-show-location
3088 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3089 reporter to emit source location information @emph{once}; that is, in
3090 case the message is too long to fit on a single physical line and has to
3091 be wrapped, the source location won't be emitted (as prefix) again,
3092 over and over, in subsequent continuation lines. This is the default
3095 @item -fdiagnostics-show-location=every-line
3096 Only meaningful in line-wrapping mode. Instructs the diagnostic
3097 messages reporter to emit the same source location information (as
3098 prefix) for physical lines that result from the process of breaking
3099 a message which is too long to fit on a single line.
3101 @item -fdiagnostics-color[=@var{WHEN}]
3102 @itemx -fno-diagnostics-color
3103 @opindex fdiagnostics-color
3104 @cindex highlight, color, colour
3105 @vindex GCC_COLORS @r{environment variable}
3106 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3107 or @samp{auto}. The default is @samp{never} if @env{GCC_COLORS} environment
3108 variable isn't present in the environment, and @samp{auto} otherwise.
3109 @samp{auto} means to use color only when the standard error is a terminal.
3110 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3111 aliases for @option{-fdiagnostics-color=always} and
3112 @option{-fdiagnostics-color=never}, respectively.
3114 The colors are defined by the environment variable @env{GCC_COLORS}.
3115 Its value is a colon-separated list of capabilities and Select Graphic
3116 Rendition (SGR) substrings. SGR commands are interpreted by the
3117 terminal or terminal emulator. (See the section in the documentation
3118 of your text terminal for permitted values and their meanings as
3119 character attributes.) These substring values are integers in decimal
3120 representation and can be concatenated with semicolons.
3121 Common values to concatenate include
3123 @samp{4} for underline,
3125 @samp{7} for inverse,
3126 @samp{39} for default foreground color,
3127 @samp{30} to @samp{37} for foreground colors,
3128 @samp{90} to @samp{97} for 16-color mode foreground colors,
3129 @samp{38;5;0} to @samp{38;5;255}
3130 for 88-color and 256-color modes foreground colors,
3131 @samp{49} for default background color,
3132 @samp{40} to @samp{47} for background colors,
3133 @samp{100} to @samp{107} for 16-color mode background colors,
3134 and @samp{48;5;0} to @samp{48;5;255}
3135 for 88-color and 256-color modes background colors.
3137 The default @env{GCC_COLORS} is
3138 @samp{error=01;31:warning=01;35:note=01;36:caret=01;32:locus=01:quote=01}
3139 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3140 @samp{01;36} is bold cyan, @samp{01;32} is bold green and
3141 @samp{01} is bold. Setting @env{GCC_COLORS} to the empty
3142 string disables colors.
3143 Supported capabilities are as follows.
3147 @vindex error GCC_COLORS @r{capability}
3148 SGR substring for error: markers.
3151 @vindex warning GCC_COLORS @r{capability}
3152 SGR substring for warning: markers.
3155 @vindex note GCC_COLORS @r{capability}
3156 SGR substring for note: markers.
3159 @vindex caret GCC_COLORS @r{capability}
3160 SGR substring for caret line.
3163 @vindex locus GCC_COLORS @r{capability}
3164 SGR substring for location information, @samp{file:line} or
3165 @samp{file:line:column} etc.
3168 @vindex quote GCC_COLORS @r{capability}
3169 SGR substring for information printed within quotes.
3172 @item -fno-diagnostics-show-option
3173 @opindex fno-diagnostics-show-option
3174 @opindex fdiagnostics-show-option
3175 By default, each diagnostic emitted includes text indicating the
3176 command-line option that directly controls the diagnostic (if such an
3177 option is known to the diagnostic machinery). Specifying the
3178 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3180 @item -fno-diagnostics-show-caret
3181 @opindex fno-diagnostics-show-caret
3182 @opindex fdiagnostics-show-caret
3183 By default, each diagnostic emitted includes the original source line
3184 and a caret '^' indicating the column. This option suppresses this
3189 @node Warning Options
3190 @section Options to Request or Suppress Warnings
3191 @cindex options to control warnings
3192 @cindex warning messages
3193 @cindex messages, warning
3194 @cindex suppressing warnings
3196 Warnings are diagnostic messages that report constructions that
3197 are not inherently erroneous but that are risky or suggest there
3198 may have been an error.
3200 The following language-independent options do not enable specific
3201 warnings but control the kinds of diagnostics produced by GCC@.
3204 @cindex syntax checking
3206 @opindex fsyntax-only
3207 Check the code for syntax errors, but don't do anything beyond that.
3209 @item -fmax-errors=@var{n}
3210 @opindex fmax-errors
3211 Limits the maximum number of error messages to @var{n}, at which point
3212 GCC bails out rather than attempting to continue processing the source
3213 code. If @var{n} is 0 (the default), there is no limit on the number
3214 of error messages produced. If @option{-Wfatal-errors} is also
3215 specified, then @option{-Wfatal-errors} takes precedence over this
3220 Inhibit all warning messages.
3225 Make all warnings into errors.
3230 Make the specified warning into an error. The specifier for a warning
3231 is appended; for example @option{-Werror=switch} turns the warnings
3232 controlled by @option{-Wswitch} into errors. This switch takes a
3233 negative form, to be used to negate @option{-Werror} for specific
3234 warnings; for example @option{-Wno-error=switch} makes
3235 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3238 The warning message for each controllable warning includes the
3239 option that controls the warning. That option can then be used with
3240 @option{-Werror=} and @option{-Wno-error=} as described above.
3241 (Printing of the option in the warning message can be disabled using the
3242 @option{-fno-diagnostics-show-option} flag.)
3244 Note that specifying @option{-Werror=}@var{foo} automatically implies
3245 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3248 @item -Wfatal-errors
3249 @opindex Wfatal-errors
3250 @opindex Wno-fatal-errors
3251 This option causes the compiler to abort compilation on the first error
3252 occurred rather than trying to keep going and printing further error
3257 You can request many specific warnings with options beginning with
3258 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3259 implicit declarations. Each of these specific warning options also
3260 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3261 example, @option{-Wno-implicit}. This manual lists only one of the
3262 two forms, whichever is not the default. For further
3263 language-specific options also refer to @ref{C++ Dialect Options} and
3264 @ref{Objective-C and Objective-C++ Dialect Options}.
3266 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3267 options, such as @option{-Wunused}, which may turn on further options,
3268 such as @option{-Wunused-value}. The combined effect of positive and
3269 negative forms is that more specific options have priority over less
3270 specific ones, independently of their position in the command-line. For
3271 options of the same specificity, the last one takes effect. Options
3272 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3273 as if they appeared at the end of the command-line.
3275 When an unrecognized warning option is requested (e.g.,
3276 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3277 that the option is not recognized. However, if the @option{-Wno-} form
3278 is used, the behavior is slightly different: no diagnostic is
3279 produced for @option{-Wno-unknown-warning} unless other diagnostics
3280 are being produced. This allows the use of new @option{-Wno-} options
3281 with old compilers, but if something goes wrong, the compiler
3282 warns that an unrecognized option is present.
3289 Issue all the warnings demanded by strict ISO C and ISO C++;
3290 reject all programs that use forbidden extensions, and some other
3291 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3292 version of the ISO C standard specified by any @option{-std} option used.
3294 Valid ISO C and ISO C++ programs should compile properly with or without
3295 this option (though a rare few require @option{-ansi} or a
3296 @option{-std} option specifying the required version of ISO C)@. However,
3297 without this option, certain GNU extensions and traditional C and C++
3298 features are supported as well. With this option, they are rejected.
3300 @option{-Wpedantic} does not cause warning messages for use of the
3301 alternate keywords whose names begin and end with @samp{__}. Pedantic
3302 warnings are also disabled in the expression that follows
3303 @code{__extension__}. However, only system header files should use
3304 these escape routes; application programs should avoid them.
3305 @xref{Alternate Keywords}.
3307 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3308 C conformance. They soon find that it does not do quite what they want:
3309 it finds some non-ISO practices, but not all---only those for which
3310 ISO C @emph{requires} a diagnostic, and some others for which
3311 diagnostics have been added.
3313 A feature to report any failure to conform to ISO C might be useful in
3314 some instances, but would require considerable additional work and would
3315 be quite different from @option{-Wpedantic}. We don't have plans to
3316 support such a feature in the near future.
3318 Where the standard specified with @option{-std} represents a GNU
3319 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3320 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3321 extended dialect is based. Warnings from @option{-Wpedantic} are given
3322 where they are required by the base standard. (It does not make sense
3323 for such warnings to be given only for features not in the specified GNU
3324 C dialect, since by definition the GNU dialects of C include all
3325 features the compiler supports with the given option, and there would be
3326 nothing to warn about.)
3328 @item -pedantic-errors
3329 @opindex pedantic-errors
3330 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3331 requires a diagnostic, in some cases where there is undefined behavior
3332 at compile-time and in some other cases that do not prevent compilation
3333 of programs that are valid according to the standard. This is not
3334 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3335 by this option and not enabled by the latter and vice versa.
3340 This enables all the warnings about constructions that some users
3341 consider questionable, and that are easy to avoid (or modify to
3342 prevent the warning), even in conjunction with macros. This also
3343 enables some language-specific warnings described in @ref{C++ Dialect
3344 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3346 @option{-Wall} turns on the following warning flags:
3348 @gccoptlist{-Waddress @gol
3349 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
3351 -Wchar-subscripts @gol
3352 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3353 -Wimplicit-int @r{(C and Objective-C only)} @gol
3354 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3357 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3358 -Wmaybe-uninitialized @gol
3359 -Wmissing-braces @r{(only for C/ObjC)} @gol
3366 -Wsequence-point @gol
3367 -Wsign-compare @r{(only in C++)} @gol
3368 -Wstrict-aliasing @gol
3369 -Wstrict-overflow=1 @gol
3372 -Wuninitialized @gol
3373 -Wunknown-pragmas @gol
3374 -Wunused-function @gol
3377 -Wunused-variable @gol
3378 -Wvolatile-register-var @gol
3381 Note that some warning flags are not implied by @option{-Wall}. Some of
3382 them warn about constructions that users generally do not consider
3383 questionable, but which occasionally you might wish to check for;
3384 others warn about constructions that are necessary or hard to avoid in
3385 some cases, and there is no simple way to modify the code to suppress
3386 the warning. Some of them are enabled by @option{-Wextra} but many of
3387 them must be enabled individually.
3393 This enables some extra warning flags that are not enabled by
3394 @option{-Wall}. (This option used to be called @option{-W}. The older
3395 name is still supported, but the newer name is more descriptive.)
3397 @gccoptlist{-Wclobbered @gol
3399 -Wignored-qualifiers @gol
3400 -Wmissing-field-initializers @gol
3401 -Wmissing-parameter-type @r{(C only)} @gol
3402 -Wold-style-declaration @r{(C only)} @gol
3403 -Woverride-init @gol
3406 -Wuninitialized @gol
3407 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3408 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3411 The option @option{-Wextra} also prints warning messages for the
3417 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3418 @samp{>}, or @samp{>=}.
3421 (C++ only) An enumerator and a non-enumerator both appear in a
3422 conditional expression.
3425 (C++ only) Ambiguous virtual bases.
3428 (C++ only) Subscripting an array that has been declared @samp{register}.
3431 (C++ only) Taking the address of a variable that has been declared
3435 (C++ only) A base class is not initialized in a derived class's copy
3440 @item -Wchar-subscripts
3441 @opindex Wchar-subscripts
3442 @opindex Wno-char-subscripts
3443 Warn if an array subscript has type @code{char}. This is a common cause
3444 of error, as programmers often forget that this type is signed on some
3446 This warning is enabled by @option{-Wall}.
3450 @opindex Wno-comment
3451 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3452 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3453 This warning is enabled by @option{-Wall}.
3455 @item -Wno-coverage-mismatch
3456 @opindex Wno-coverage-mismatch
3457 Warn if feedback profiles do not match when using the
3458 @option{-fprofile-use} option.
3459 If a source file is changed between compiling with @option{-fprofile-gen} and
3460 with @option{-fprofile-use}, the files with the profile feedback can fail
3461 to match the source file and GCC cannot use the profile feedback
3462 information. By default, this warning is enabled and is treated as an
3463 error. @option{-Wno-coverage-mismatch} can be used to disable the
3464 warning or @option{-Wno-error=coverage-mismatch} can be used to
3465 disable the error. Disabling the error for this warning can result in
3466 poorly optimized code and is useful only in the
3467 case of very minor changes such as bug fixes to an existing code-base.
3468 Completely disabling the warning is not recommended.
3471 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3473 Suppress warning messages emitted by @code{#warning} directives.
3475 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3476 @opindex Wdouble-promotion
3477 @opindex Wno-double-promotion
3478 Give a warning when a value of type @code{float} is implicitly
3479 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3480 floating-point unit implement @code{float} in hardware, but emulate
3481 @code{double} in software. On such a machine, doing computations
3482 using @code{double} values is much more expensive because of the
3483 overhead required for software emulation.
3485 It is easy to accidentally do computations with @code{double} because
3486 floating-point literals are implicitly of type @code{double}. For
3490 float area(float radius)
3492 return 3.14159 * radius * radius;
3496 the compiler performs the entire computation with @code{double}
3497 because the floating-point literal is a @code{double}.
3500 @itemx -Wformat=@var{n}
3503 @opindex ffreestanding
3504 @opindex fno-builtin
3506 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3507 the arguments supplied have types appropriate to the format string
3508 specified, and that the conversions specified in the format string make
3509 sense. This includes standard functions, and others specified by format
3510 attributes (@pxref{Function Attributes}), in the @code{printf},
3511 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3512 not in the C standard) families (or other target-specific families).
3513 Which functions are checked without format attributes having been
3514 specified depends on the standard version selected, and such checks of
3515 functions without the attribute specified are disabled by
3516 @option{-ffreestanding} or @option{-fno-builtin}.
3518 The formats are checked against the format features supported by GNU
3519 libc version 2.2. These include all ISO C90 and C99 features, as well
3520 as features from the Single Unix Specification and some BSD and GNU
3521 extensions. Other library implementations may not support all these
3522 features; GCC does not support warning about features that go beyond a
3523 particular library's limitations. However, if @option{-Wpedantic} is used
3524 with @option{-Wformat}, warnings are given about format features not
3525 in the selected standard version (but not for @code{strfmon} formats,
3526 since those are not in any version of the C standard). @xref{C Dialect
3527 Options,,Options Controlling C Dialect}.
3534 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3535 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3536 @option{-Wformat} also checks for null format arguments for several
3537 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3538 aspects of this level of format checking can be disabled by the
3539 options: @option{-Wno-format-contains-nul},
3540 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3541 @option{-Wformat} is enabled by @option{-Wall}.
3543 @item -Wno-format-contains-nul
3544 @opindex Wno-format-contains-nul
3545 @opindex Wformat-contains-nul
3546 If @option{-Wformat} is specified, do not warn about format strings that
3549 @item -Wno-format-extra-args
3550 @opindex Wno-format-extra-args
3551 @opindex Wformat-extra-args
3552 If @option{-Wformat} is specified, do not warn about excess arguments to a
3553 @code{printf} or @code{scanf} format function. The C standard specifies
3554 that such arguments are ignored.
3556 Where the unused arguments lie between used arguments that are
3557 specified with @samp{$} operand number specifications, normally
3558 warnings are still given, since the implementation could not know what
3559 type to pass to @code{va_arg} to skip the unused arguments. However,
3560 in the case of @code{scanf} formats, this option suppresses the
3561 warning if the unused arguments are all pointers, since the Single
3562 Unix Specification says that such unused arguments are allowed.
3564 @item -Wno-format-zero-length
3565 @opindex Wno-format-zero-length
3566 @opindex Wformat-zero-length
3567 If @option{-Wformat} is specified, do not warn about zero-length formats.
3568 The C standard specifies that zero-length formats are allowed.
3573 Enable @option{-Wformat} plus additional format checks. Currently
3574 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
3575 -Wformat-signedness -Wformat-y2k}.
3577 @item -Wformat-nonliteral
3578 @opindex Wformat-nonliteral
3579 @opindex Wno-format-nonliteral
3580 If @option{-Wformat} is specified, also warn if the format string is not a
3581 string literal and so cannot be checked, unless the format function
3582 takes its format arguments as a @code{va_list}.
3584 @item -Wformat-security
3585 @opindex Wformat-security
3586 @opindex Wno-format-security
3587 If @option{-Wformat} is specified, also warn about uses of format
3588 functions that represent possible security problems. At present, this
3589 warns about calls to @code{printf} and @code{scanf} functions where the
3590 format string is not a string literal and there are no format arguments,
3591 as in @code{printf (foo);}. This may be a security hole if the format
3592 string came from untrusted input and contains @samp{%n}. (This is
3593 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3594 in future warnings may be added to @option{-Wformat-security} that are not
3595 included in @option{-Wformat-nonliteral}.)
3597 @item -Wformat-signedness
3598 @opindex Wformat-signedness
3599 @opindex Wno-format-signedness
3600 If @option{-Wformat} is specified, also warn if the format string
3601 requires an unsigned argument and the argument is signed and vice versa.
3604 @opindex Wformat-y2k
3605 @opindex Wno-format-y2k
3606 If @option{-Wformat} is specified, also warn about @code{strftime}
3607 formats that may yield only a two-digit year.
3612 @opindex Wno-nonnull
3613 Warn about passing a null pointer for arguments marked as
3614 requiring a non-null value by the @code{nonnull} function attribute.
3616 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3617 can be disabled with the @option{-Wno-nonnull} option.
3619 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3621 @opindex Wno-init-self
3622 Warn about uninitialized variables that are initialized with themselves.
3623 Note this option can only be used with the @option{-Wuninitialized} option.
3625 For example, GCC warns about @code{i} being uninitialized in the
3626 following snippet only when @option{-Winit-self} has been specified:
3637 This warning is enabled by @option{-Wall} in C++.
3639 @item -Wimplicit-int @r{(C and Objective-C only)}
3640 @opindex Wimplicit-int
3641 @opindex Wno-implicit-int
3642 Warn when a declaration does not specify a type.
3643 This warning is enabled by @option{-Wall}.
3645 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3646 @opindex Wimplicit-function-declaration
3647 @opindex Wno-implicit-function-declaration
3648 Give a warning whenever a function is used before being declared. In
3649 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3650 enabled by default and it is made into an error by
3651 @option{-pedantic-errors}. This warning is also enabled by
3654 @item -Wimplicit @r{(C and Objective-C only)}
3656 @opindex Wno-implicit
3657 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3658 This warning is enabled by @option{-Wall}.
3660 @item -Wignored-qualifiers @r{(C and C++ only)}
3661 @opindex Wignored-qualifiers
3662 @opindex Wno-ignored-qualifiers
3663 Warn if the return type of a function has a type qualifier
3664 such as @code{const}. For ISO C such a type qualifier has no effect,
3665 since the value returned by a function is not an lvalue.
3666 For C++, the warning is only emitted for scalar types or @code{void}.
3667 ISO C prohibits qualified @code{void} return types on function
3668 definitions, so such return types always receive a warning
3669 even without this option.
3671 This warning is also enabled by @option{-Wextra}.
3676 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3677 a function with external linkage, returning int, taking either zero
3678 arguments, two, or three arguments of appropriate types. This warning
3679 is enabled by default in C++ and is enabled by either @option{-Wall}
3680 or @option{-Wpedantic}.
3682 @item -Wmissing-braces
3683 @opindex Wmissing-braces
3684 @opindex Wno-missing-braces
3685 Warn if an aggregate or union initializer is not fully bracketed. In
3686 the following example, the initializer for @samp{a} is not fully
3687 bracketed, but that for @samp{b} is fully bracketed. This warning is
3688 enabled by @option{-Wall} in C.
3691 int a[2][2] = @{ 0, 1, 2, 3 @};
3692 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3695 This warning is enabled by @option{-Wall}.
3697 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3698 @opindex Wmissing-include-dirs
3699 @opindex Wno-missing-include-dirs
3700 Warn if a user-supplied include directory does not exist.
3703 @opindex Wparentheses
3704 @opindex Wno-parentheses
3705 Warn if parentheses are omitted in certain contexts, such
3706 as when there is an assignment in a context where a truth value
3707 is expected, or when operators are nested whose precedence people
3708 often get confused about.
3710 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3711 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3712 interpretation from that of ordinary mathematical notation.
3714 Also warn about constructions where there may be confusion to which
3715 @code{if} statement an @code{else} branch belongs. Here is an example of
3730 In C/C++, every @code{else} branch belongs to the innermost possible
3731 @code{if} statement, which in this example is @code{if (b)}. This is
3732 often not what the programmer expected, as illustrated in the above
3733 example by indentation the programmer chose. When there is the
3734 potential for this confusion, GCC issues a warning when this flag
3735 is specified. To eliminate the warning, add explicit braces around
3736 the innermost @code{if} statement so there is no way the @code{else}
3737 can belong to the enclosing @code{if}. The resulting code
3754 Also warn for dangerous uses of the GNU extension to
3755 @code{?:} with omitted middle operand. When the condition
3756 in the @code{?}: operator is a boolean expression, the omitted value is
3757 always 1. Often programmers expect it to be a value computed
3758 inside the conditional expression instead.
3760 This warning is enabled by @option{-Wall}.
3762 @item -Wsequence-point
3763 @opindex Wsequence-point
3764 @opindex Wno-sequence-point
3765 Warn about code that may have undefined semantics because of violations
3766 of sequence point rules in the C and C++ standards.
3768 The C and C++ standards define the order in which expressions in a C/C++
3769 program are evaluated in terms of @dfn{sequence points}, which represent
3770 a partial ordering between the execution of parts of the program: those
3771 executed before the sequence point, and those executed after it. These
3772 occur after the evaluation of a full expression (one which is not part
3773 of a larger expression), after the evaluation of the first operand of a
3774 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3775 function is called (but after the evaluation of its arguments and the
3776 expression denoting the called function), and in certain other places.
3777 Other than as expressed by the sequence point rules, the order of
3778 evaluation of subexpressions of an expression is not specified. All
3779 these rules describe only a partial order rather than a total order,
3780 since, for example, if two functions are called within one expression
3781 with no sequence point between them, the order in which the functions
3782 are called is not specified. However, the standards committee have
3783 ruled that function calls do not overlap.
3785 It is not specified when between sequence points modifications to the
3786 values of objects take effect. Programs whose behavior depends on this
3787 have undefined behavior; the C and C++ standards specify that ``Between
3788 the previous and next sequence point an object shall have its stored
3789 value modified at most once by the evaluation of an expression.
3790 Furthermore, the prior value shall be read only to determine the value
3791 to be stored.''. If a program breaks these rules, the results on any
3792 particular implementation are entirely unpredictable.
3794 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3795 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3796 diagnosed by this option, and it may give an occasional false positive
3797 result, but in general it has been found fairly effective at detecting
3798 this sort of problem in programs.
3800 The standard is worded confusingly, therefore there is some debate
3801 over the precise meaning of the sequence point rules in subtle cases.
3802 Links to discussions of the problem, including proposed formal
3803 definitions, may be found on the GCC readings page, at
3804 @uref{http://gcc.gnu.org/@/readings.html}.
3806 This warning is enabled by @option{-Wall} for C and C++.
3808 @item -Wno-return-local-addr
3809 @opindex Wno-return-local-addr
3810 @opindex Wreturn-local-addr
3811 Do not warn about returning a pointer (or in C++, a reference) to a
3812 variable that goes out of scope after the function returns.
3815 @opindex Wreturn-type
3816 @opindex Wno-return-type
3817 Warn whenever a function is defined with a return type that defaults
3818 to @code{int}. Also warn about any @code{return} statement with no
3819 return value in a function whose return type is not @code{void}
3820 (falling off the end of the function body is considered returning
3821 without a value), and about a @code{return} statement with an
3822 expression in a function whose return type is @code{void}.
3824 For C++, a function without return type always produces a diagnostic
3825 message, even when @option{-Wno-return-type} is specified. The only
3826 exceptions are @samp{main} and functions defined in system headers.
3828 This warning is enabled by @option{-Wall}.
3830 @item -Wshift-count-negative
3831 @opindex Wshift-count-negative
3832 @opindex Wno-shift-count-negative
3833 Warn if shift count is negative. This warning is enabled by default.
3835 @item -Wshift-count-overflow
3836 @opindex Wshift-count-overflow
3837 @opindex Wno-shift-count-overflow
3838 Warn if shift count >= width of type. This warning is enabled by default.
3843 Warn whenever a @code{switch} statement has an index of enumerated type
3844 and lacks a @code{case} for one or more of the named codes of that
3845 enumeration. (The presence of a @code{default} label prevents this
3846 warning.) @code{case} labels outside the enumeration range also
3847 provoke warnings when this option is used (even if there is a
3848 @code{default} label).
3849 This warning is enabled by @option{-Wall}.
3851 @item -Wswitch-default
3852 @opindex Wswitch-default
3853 @opindex Wno-switch-default
3854 Warn whenever a @code{switch} statement does not have a @code{default}
3858 @opindex Wswitch-enum
3859 @opindex Wno-switch-enum
3860 Warn whenever a @code{switch} statement has an index of enumerated type
3861 and lacks a @code{case} for one or more of the named codes of that
3862 enumeration. @code{case} labels outside the enumeration range also
3863 provoke warnings when this option is used. The only difference
3864 between @option{-Wswitch} and this option is that this option gives a
3865 warning about an omitted enumeration code even if there is a
3866 @code{default} label.
3869 @opindex Wswitch-bool
3870 @opindex Wno-switch-bool
3871 Warn whenever a @code{switch} statement has an index of boolean type.
3872 It is possible to suppress this warning by casting the controlling
3873 expression to a type other than @code{bool}. For example:
3876 switch ((int) (a == 4))
3882 This warning is enabled by default for C and C++ programs.
3884 @item -Wsync-nand @r{(C and C++ only)}
3886 @opindex Wno-sync-nand
3887 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3888 built-in functions are used. These functions changed semantics in GCC 4.4.
3892 @opindex Wno-trigraphs
3893 Warn if any trigraphs are encountered that might change the meaning of
3894 the program (trigraphs within comments are not warned about).
3895 This warning is enabled by @option{-Wall}.
3897 @item -Wunused-but-set-parameter
3898 @opindex Wunused-but-set-parameter
3899 @opindex Wno-unused-but-set-parameter
3900 Warn whenever a function parameter is assigned to, but otherwise unused
3901 (aside from its declaration).
3903 To suppress this warning use the @samp{unused} attribute
3904 (@pxref{Variable Attributes}).
3906 This warning is also enabled by @option{-Wunused} together with
3909 @item -Wunused-but-set-variable
3910 @opindex Wunused-but-set-variable
3911 @opindex Wno-unused-but-set-variable
3912 Warn whenever a local variable is assigned to, but otherwise unused
3913 (aside from its declaration).
3914 This warning is enabled by @option{-Wall}.
3916 To suppress this warning use the @samp{unused} attribute
3917 (@pxref{Variable Attributes}).
3919 This warning is also enabled by @option{-Wunused}, which is enabled
3922 @item -Wunused-function
3923 @opindex Wunused-function
3924 @opindex Wno-unused-function
3925 Warn whenever a static function is declared but not defined or a
3926 non-inline static function is unused.
3927 This warning is enabled by @option{-Wall}.
3929 @item -Wunused-label
3930 @opindex Wunused-label
3931 @opindex Wno-unused-label
3932 Warn whenever a label is declared but not used.
3933 This warning is enabled by @option{-Wall}.
3935 To suppress this warning use the @samp{unused} attribute
3936 (@pxref{Variable Attributes}).
3938 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
3939 @opindex Wunused-local-typedefs
3940 Warn when a typedef locally defined in a function is not used.
3941 This warning is enabled by @option{-Wall}.
3943 @item -Wunused-parameter
3944 @opindex Wunused-parameter
3945 @opindex Wno-unused-parameter
3946 Warn whenever a function parameter is unused aside from its declaration.
3948 To suppress this warning use the @samp{unused} attribute
3949 (@pxref{Variable Attributes}).
3951 @item -Wno-unused-result
3952 @opindex Wunused-result
3953 @opindex Wno-unused-result
3954 Do not warn if a caller of a function marked with attribute
3955 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
3956 its return value. The default is @option{-Wunused-result}.
3958 @item -Wunused-variable
3959 @opindex Wunused-variable
3960 @opindex Wno-unused-variable
3961 Warn whenever a local variable or non-constant static variable is unused
3962 aside from its declaration.
3963 This warning is enabled by @option{-Wall}.
3965 To suppress this warning use the @samp{unused} attribute
3966 (@pxref{Variable Attributes}).
3968 @item -Wunused-value
3969 @opindex Wunused-value
3970 @opindex Wno-unused-value
3971 Warn whenever a statement computes a result that is explicitly not
3972 used. To suppress this warning cast the unused expression to
3973 @samp{void}. This includes an expression-statement or the left-hand
3974 side of a comma expression that contains no side effects. For example,
3975 an expression such as @samp{x[i,j]} causes a warning, while
3976 @samp{x[(void)i,j]} does not.
3978 This warning is enabled by @option{-Wall}.
3983 All the above @option{-Wunused} options combined.
3985 In order to get a warning about an unused function parameter, you must
3986 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
3987 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
3989 @item -Wuninitialized
3990 @opindex Wuninitialized
3991 @opindex Wno-uninitialized
3992 Warn if an automatic variable is used without first being initialized
3993 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3994 warn if a non-static reference or non-static @samp{const} member
3995 appears in a class without constructors.
3997 If you want to warn about code that uses the uninitialized value of the
3998 variable in its own initializer, use the @option{-Winit-self} option.
4000 These warnings occur for individual uninitialized or clobbered
4001 elements of structure, union or array variables as well as for
4002 variables that are uninitialized or clobbered as a whole. They do
4003 not occur for variables or elements declared @code{volatile}. Because
4004 these warnings depend on optimization, the exact variables or elements
4005 for which there are warnings depends on the precise optimization
4006 options and version of GCC used.
4008 Note that there may be no warning about a variable that is used only
4009 to compute a value that itself is never used, because such
4010 computations may be deleted by data flow analysis before the warnings
4013 @item -Wmaybe-uninitialized
4014 @opindex Wmaybe-uninitialized
4015 @opindex Wno-maybe-uninitialized
4016 For an automatic variable, if there exists a path from the function
4017 entry to a use of the variable that is initialized, but there exist
4018 some other paths for which the variable is not initialized, the compiler
4019 emits a warning if it cannot prove the uninitialized paths are not
4020 executed at run time. These warnings are made optional because GCC is
4021 not smart enough to see all the reasons why the code might be correct
4022 in spite of appearing to have an error. Here is one example of how
4043 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4044 always initialized, but GCC doesn't know this. To suppress the
4045 warning, you need to provide a default case with assert(0) or
4048 @cindex @code{longjmp} warnings
4049 This option also warns when a non-volatile automatic variable might be
4050 changed by a call to @code{longjmp}. These warnings as well are possible
4051 only in optimizing compilation.
4053 The compiler sees only the calls to @code{setjmp}. It cannot know
4054 where @code{longjmp} will be called; in fact, a signal handler could
4055 call it at any point in the code. As a result, you may get a warning
4056 even when there is in fact no problem because @code{longjmp} cannot
4057 in fact be called at the place that would cause a problem.
4059 Some spurious warnings can be avoided if you declare all the functions
4060 you use that never return as @code{noreturn}. @xref{Function
4063 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4065 @item -Wunknown-pragmas
4066 @opindex Wunknown-pragmas
4067 @opindex Wno-unknown-pragmas
4068 @cindex warning for unknown pragmas
4069 @cindex unknown pragmas, warning
4070 @cindex pragmas, warning of unknown
4071 Warn when a @code{#pragma} directive is encountered that is not understood by
4072 GCC@. If this command-line option is used, warnings are even issued
4073 for unknown pragmas in system header files. This is not the case if
4074 the warnings are only enabled by the @option{-Wall} command-line option.
4077 @opindex Wno-pragmas
4079 Do not warn about misuses of pragmas, such as incorrect parameters,
4080 invalid syntax, or conflicts between pragmas. See also
4081 @option{-Wunknown-pragmas}.
4083 @item -Wstrict-aliasing
4084 @opindex Wstrict-aliasing
4085 @opindex Wno-strict-aliasing
4086 This option is only active when @option{-fstrict-aliasing} is active.
4087 It warns about code that might break the strict aliasing rules that the
4088 compiler is using for optimization. The warning does not catch all
4089 cases, but does attempt to catch the more common pitfalls. It is
4090 included in @option{-Wall}.
4091 It is equivalent to @option{-Wstrict-aliasing=3}
4093 @item -Wstrict-aliasing=n
4094 @opindex Wstrict-aliasing=n
4095 This option is only active when @option{-fstrict-aliasing} is active.
4096 It warns about code that might break the strict aliasing rules that the
4097 compiler is using for optimization.
4098 Higher levels correspond to higher accuracy (fewer false positives).
4099 Higher levels also correspond to more effort, similar to the way @option{-O}
4101 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4103 Level 1: Most aggressive, quick, least accurate.
4104 Possibly useful when higher levels
4105 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4106 false negatives. However, it has many false positives.
4107 Warns for all pointer conversions between possibly incompatible types,
4108 even if never dereferenced. Runs in the front end only.
4110 Level 2: Aggressive, quick, not too precise.
4111 May still have many false positives (not as many as level 1 though),
4112 and few false negatives (but possibly more than level 1).
4113 Unlike level 1, it only warns when an address is taken. Warns about
4114 incomplete types. Runs in the front end only.
4116 Level 3 (default for @option{-Wstrict-aliasing}):
4117 Should have very few false positives and few false
4118 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4119 Takes care of the common pun+dereference pattern in the front end:
4120 @code{*(int*)&some_float}.
4121 If optimization is enabled, it also runs in the back end, where it deals
4122 with multiple statement cases using flow-sensitive points-to information.
4123 Only warns when the converted pointer is dereferenced.
4124 Does not warn about incomplete types.
4126 @item -Wstrict-overflow
4127 @itemx -Wstrict-overflow=@var{n}
4128 @opindex Wstrict-overflow
4129 @opindex Wno-strict-overflow
4130 This option is only active when @option{-fstrict-overflow} is active.
4131 It warns about cases where the compiler optimizes based on the
4132 assumption that signed overflow does not occur. Note that it does not
4133 warn about all cases where the code might overflow: it only warns
4134 about cases where the compiler implements some optimization. Thus
4135 this warning depends on the optimization level.
4137 An optimization that assumes that signed overflow does not occur is
4138 perfectly safe if the values of the variables involved are such that
4139 overflow never does, in fact, occur. Therefore this warning can
4140 easily give a false positive: a warning about code that is not
4141 actually a problem. To help focus on important issues, several
4142 warning levels are defined. No warnings are issued for the use of
4143 undefined signed overflow when estimating how many iterations a loop
4144 requires, in particular when determining whether a loop will be
4148 @item -Wstrict-overflow=1
4149 Warn about cases that are both questionable and easy to avoid. For
4150 example, with @option{-fstrict-overflow}, the compiler simplifies
4151 @code{x + 1 > x} to @code{1}. This level of
4152 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4153 are not, and must be explicitly requested.
4155 @item -Wstrict-overflow=2
4156 Also warn about other cases where a comparison is simplified to a
4157 constant. For example: @code{abs (x) >= 0}. This can only be
4158 simplified when @option{-fstrict-overflow} is in effect, because
4159 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4160 zero. @option{-Wstrict-overflow} (with no level) is the same as
4161 @option{-Wstrict-overflow=2}.
4163 @item -Wstrict-overflow=3
4164 Also warn about other cases where a comparison is simplified. For
4165 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4167 @item -Wstrict-overflow=4
4168 Also warn about other simplifications not covered by the above cases.
4169 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4171 @item -Wstrict-overflow=5
4172 Also warn about cases where the compiler reduces the magnitude of a
4173 constant involved in a comparison. For example: @code{x + 2 > y} is
4174 simplified to @code{x + 1 >= y}. This is reported only at the
4175 highest warning level because this simplification applies to many
4176 comparisons, so this warning level gives a very large number of
4180 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
4181 @opindex Wsuggest-attribute=
4182 @opindex Wno-suggest-attribute=
4183 Warn for cases where adding an attribute may be beneficial. The
4184 attributes currently supported are listed below.
4187 @item -Wsuggest-attribute=pure
4188 @itemx -Wsuggest-attribute=const
4189 @itemx -Wsuggest-attribute=noreturn
4190 @opindex Wsuggest-attribute=pure
4191 @opindex Wno-suggest-attribute=pure
4192 @opindex Wsuggest-attribute=const
4193 @opindex Wno-suggest-attribute=const
4194 @opindex Wsuggest-attribute=noreturn
4195 @opindex Wno-suggest-attribute=noreturn
4197 Warn about functions that might be candidates for attributes
4198 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
4199 functions visible in other compilation units or (in the case of @code{pure} and
4200 @code{const}) if it cannot prove that the function returns normally. A function
4201 returns normally if it doesn't contain an infinite loop or return abnormally
4202 by throwing, calling @code{abort()} or trapping. This analysis requires option
4203 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
4204 higher. Higher optimization levels improve the accuracy of the analysis.
4206 @item -Wsuggest-attribute=format
4207 @itemx -Wmissing-format-attribute
4208 @opindex Wsuggest-attribute=format
4209 @opindex Wmissing-format-attribute
4210 @opindex Wno-suggest-attribute=format
4211 @opindex Wno-missing-format-attribute
4215 Warn about function pointers that might be candidates for @code{format}
4216 attributes. Note these are only possible candidates, not absolute ones.
4217 GCC guesses that function pointers with @code{format} attributes that
4218 are used in assignment, initialization, parameter passing or return
4219 statements should have a corresponding @code{format} attribute in the
4220 resulting type. I.e.@: the left-hand side of the assignment or
4221 initialization, the type of the parameter variable, or the return type
4222 of the containing function respectively should also have a @code{format}
4223 attribute to avoid the warning.
4225 GCC also warns about function definitions that might be
4226 candidates for @code{format} attributes. Again, these are only
4227 possible candidates. GCC guesses that @code{format} attributes
4228 might be appropriate for any function that calls a function like
4229 @code{vprintf} or @code{vscanf}, but this might not always be the
4230 case, and some functions for which @code{format} attributes are
4231 appropriate may not be detected.
4234 @item -Wsuggest-final-types
4235 @opindex Wno-suggest-final-types
4236 @opindex Wsuggest-final-types
4237 Warn about types with virtual methods where code quality would be improved
4238 if the type was declared with C++11 final specifier, or, if possible,
4239 declared in anonymous namespace. This allows GCC to devritualize more aggressively
4240 the polymorphic calls. This warning is more effective with link time optimization,
4241 where the information about the class hiearchy graph is more complete.
4243 @item -Wsuggest-final-methods
4244 @opindex Wno-suggest-final-methods
4245 @opindex Wsuggest-final-methods
4246 Warn about virtual methods where code quality would be improved if the method
4247 was declared with C++11 final specifier, or, if possible, its type was declared
4248 in the anonymous namespace or with final specifier. This warning is more
4249 effective with link time optimization, where the information about the class
4250 hiearchy graph is more complete. It is recommended to first consider suggestins
4251 of @option{-Wsuggest-final-types} and then rebuild with new annotations.
4253 @item -Warray-bounds
4254 @opindex Wno-array-bounds
4255 @opindex Warray-bounds
4256 This option is only active when @option{-ftree-vrp} is active
4257 (default for @option{-O2} and above). It warns about subscripts to arrays
4258 that are always out of bounds. This warning is enabled by @option{-Wall}.
4260 @item -Wbool-compare
4261 @opindex Wno-bool-compare
4262 @opindex Wbool-compare
4263 Warn about boolean expression compared with an integer value different from
4264 @code{true}/@code{false}. For instance, the following comparison is
4269 if ((n > 1) == 2) @{ @dots{} @}
4271 This warning is enabled by @option{-Wall}.
4273 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
4274 @opindex Wno-discarded-qualifiers
4275 @opindex Wdiscarded-qualifiers
4276 Do not warn if type qualifiers on pointers are being discarded.
4277 Typically, the compiler will warn if a @code{const char *} variable is
4278 passed to a function that takes @code{char *} parameter. This option
4279 can be used to suppress such a warning.
4281 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
4282 @opindex Wno-incompatible-pointer-types
4283 @opindex Wincompatible-pointer-types
4284 Do not warn when there is a conversion between pointers that have incompatible
4285 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
4286 which warns for pointer argument passing or assignment with different signedness
4288 @item -Wno-int-conversion @r{(C and Objective-C only)}
4289 @opindex Wno-int-conversion
4290 @opindex Wint-conversion
4291 Do not warn about incompatible integer to pointer and pointer to integer
4292 conversions. This warning is about implicit conversions; for explicit
4293 conversions the warnings @option{-Wno-int-to-pointer-cast} and
4294 @option{-Wno-pointer-to-int-cast} may be used.
4296 @item -Wno-div-by-zero
4297 @opindex Wno-div-by-zero
4298 @opindex Wdiv-by-zero
4299 Do not warn about compile-time integer division by zero. Floating-point
4300 division by zero is not warned about, as it can be a legitimate way of
4301 obtaining infinities and NaNs.
4303 @item -Wsystem-headers
4304 @opindex Wsystem-headers
4305 @opindex Wno-system-headers
4306 @cindex warnings from system headers
4307 @cindex system headers, warnings from
4308 Print warning messages for constructs found in system header files.
4309 Warnings from system headers are normally suppressed, on the assumption
4310 that they usually do not indicate real problems and would only make the
4311 compiler output harder to read. Using this command-line option tells
4312 GCC to emit warnings from system headers as if they occurred in user
4313 code. However, note that using @option{-Wall} in conjunction with this
4314 option does @emph{not} warn about unknown pragmas in system
4315 headers---for that, @option{-Wunknown-pragmas} must also be used.
4318 @opindex Wtrampolines
4319 @opindex Wno-trampolines
4320 Warn about trampolines generated for pointers to nested functions.
4321 A trampoline is a small piece of data or code that is created at run
4322 time on the stack when the address of a nested function is taken, and is
4323 used to call the nested function indirectly. For some targets, it is
4324 made up of data only and thus requires no special treatment. But, for
4325 most targets, it is made up of code and thus requires the stack to be
4326 made executable in order for the program to work properly.
4329 @opindex Wfloat-equal
4330 @opindex Wno-float-equal
4331 Warn if floating-point values are used in equality comparisons.
4333 The idea behind this is that sometimes it is convenient (for the
4334 programmer) to consider floating-point values as approximations to
4335 infinitely precise real numbers. If you are doing this, then you need
4336 to compute (by analyzing the code, or in some other way) the maximum or
4337 likely maximum error that the computation introduces, and allow for it
4338 when performing comparisons (and when producing output, but that's a
4339 different problem). In particular, instead of testing for equality, you
4340 should check to see whether the two values have ranges that overlap; and
4341 this is done with the relational operators, so equality comparisons are
4344 @item -Wtraditional @r{(C and Objective-C only)}
4345 @opindex Wtraditional
4346 @opindex Wno-traditional
4347 Warn about certain constructs that behave differently in traditional and
4348 ISO C@. Also warn about ISO C constructs that have no traditional C
4349 equivalent, and/or problematic constructs that should be avoided.
4353 Macro parameters that appear within string literals in the macro body.
4354 In traditional C macro replacement takes place within string literals,
4355 but in ISO C it does not.
4358 In traditional C, some preprocessor directives did not exist.
4359 Traditional preprocessors only considered a line to be a directive
4360 if the @samp{#} appeared in column 1 on the line. Therefore
4361 @option{-Wtraditional} warns about directives that traditional C
4362 understands but ignores because the @samp{#} does not appear as the
4363 first character on the line. It also suggests you hide directives like
4364 @samp{#pragma} not understood by traditional C by indenting them. Some
4365 traditional implementations do not recognize @samp{#elif}, so this option
4366 suggests avoiding it altogether.
4369 A function-like macro that appears without arguments.
4372 The unary plus operator.
4375 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
4376 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
4377 constants.) Note, these suffixes appear in macros defined in the system
4378 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
4379 Use of these macros in user code might normally lead to spurious
4380 warnings, however GCC's integrated preprocessor has enough context to
4381 avoid warning in these cases.
4384 A function declared external in one block and then used after the end of
4388 A @code{switch} statement has an operand of type @code{long}.
4391 A non-@code{static} function declaration follows a @code{static} one.
4392 This construct is not accepted by some traditional C compilers.
4395 The ISO type of an integer constant has a different width or
4396 signedness from its traditional type. This warning is only issued if
4397 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
4398 typically represent bit patterns, are not warned about.
4401 Usage of ISO string concatenation is detected.
4404 Initialization of automatic aggregates.
4407 Identifier conflicts with labels. Traditional C lacks a separate
4408 namespace for labels.
4411 Initialization of unions. If the initializer is zero, the warning is
4412 omitted. This is done under the assumption that the zero initializer in
4413 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
4414 initializer warnings and relies on default initialization to zero in the
4418 Conversions by prototypes between fixed/floating-point values and vice
4419 versa. The absence of these prototypes when compiling with traditional
4420 C causes serious problems. This is a subset of the possible
4421 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
4424 Use of ISO C style function definitions. This warning intentionally is
4425 @emph{not} issued for prototype declarations or variadic functions
4426 because these ISO C features appear in your code when using
4427 libiberty's traditional C compatibility macros, @code{PARAMS} and
4428 @code{VPARAMS}. This warning is also bypassed for nested functions
4429 because that feature is already a GCC extension and thus not relevant to
4430 traditional C compatibility.
4433 @item -Wtraditional-conversion @r{(C and Objective-C only)}
4434 @opindex Wtraditional-conversion
4435 @opindex Wno-traditional-conversion
4436 Warn if a prototype causes a type conversion that is different from what
4437 would happen to the same argument in the absence of a prototype. This
4438 includes conversions of fixed point to floating and vice versa, and
4439 conversions changing the width or signedness of a fixed-point argument
4440 except when the same as the default promotion.
4442 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
4443 @opindex Wdeclaration-after-statement
4444 @opindex Wno-declaration-after-statement
4445 Warn when a declaration is found after a statement in a block. This
4446 construct, known from C++, was introduced with ISO C99 and is by default
4447 allowed in GCC@. It is not supported by ISO C90 and was not supported by
4448 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
4453 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
4455 @item -Wno-endif-labels
4456 @opindex Wno-endif-labels
4457 @opindex Wendif-labels
4458 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
4463 Warn whenever a local variable or type declaration shadows another
4464 variable, parameter, type, class member (in C++), or instance variable
4465 (in Objective-C) or whenever a built-in function is shadowed. Note
4466 that in C++, the compiler warns if a local variable shadows an
4467 explicit typedef, but not if it shadows a struct/class/enum.
4469 @item -Wno-shadow-ivar @r{(Objective-C only)}
4470 @opindex Wno-shadow-ivar
4471 @opindex Wshadow-ivar
4472 Do not warn whenever a local variable shadows an instance variable in an
4475 @item -Wlarger-than=@var{len}
4476 @opindex Wlarger-than=@var{len}
4477 @opindex Wlarger-than-@var{len}
4478 Warn whenever an object of larger than @var{len} bytes is defined.
4480 @item -Wframe-larger-than=@var{len}
4481 @opindex Wframe-larger-than
4482 Warn if the size of a function frame is larger than @var{len} bytes.
4483 The computation done to determine the stack frame size is approximate
4484 and not conservative.
4485 The actual requirements may be somewhat greater than @var{len}
4486 even if you do not get a warning. In addition, any space allocated
4487 via @code{alloca}, variable-length arrays, or related constructs
4488 is not included by the compiler when determining
4489 whether or not to issue a warning.
4491 @item -Wno-free-nonheap-object
4492 @opindex Wno-free-nonheap-object
4493 @opindex Wfree-nonheap-object
4494 Do not warn when attempting to free an object that was not allocated
4497 @item -Wstack-usage=@var{len}
4498 @opindex Wstack-usage
4499 Warn if the stack usage of a function might be larger than @var{len} bytes.
4500 The computation done to determine the stack usage is conservative.
4501 Any space allocated via @code{alloca}, variable-length arrays, or related
4502 constructs is included by the compiler when determining whether or not to
4505 The message is in keeping with the output of @option{-fstack-usage}.
4509 If the stack usage is fully static but exceeds the specified amount, it's:
4512 warning: stack usage is 1120 bytes
4515 If the stack usage is (partly) dynamic but bounded, it's:
4518 warning: stack usage might be 1648 bytes
4521 If the stack usage is (partly) dynamic and not bounded, it's:
4524 warning: stack usage might be unbounded
4528 @item -Wunsafe-loop-optimizations
4529 @opindex Wunsafe-loop-optimizations
4530 @opindex Wno-unsafe-loop-optimizations
4531 Warn if the loop cannot be optimized because the compiler cannot
4532 assume anything on the bounds of the loop indices. With
4533 @option{-funsafe-loop-optimizations} warn if the compiler makes
4536 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4537 @opindex Wno-pedantic-ms-format
4538 @opindex Wpedantic-ms-format
4539 When used in combination with @option{-Wformat}
4540 and @option{-pedantic} without GNU extensions, this option
4541 disables the warnings about non-ISO @code{printf} / @code{scanf} format
4542 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
4543 which depend on the MS runtime.
4545 @item -Wpointer-arith
4546 @opindex Wpointer-arith
4547 @opindex Wno-pointer-arith
4548 Warn about anything that depends on the ``size of'' a function type or
4549 of @code{void}. GNU C assigns these types a size of 1, for
4550 convenience in calculations with @code{void *} pointers and pointers
4551 to functions. In C++, warn also when an arithmetic operation involves
4552 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
4555 @opindex Wtype-limits
4556 @opindex Wno-type-limits
4557 Warn if a comparison is always true or always false due to the limited
4558 range of the data type, but do not warn for constant expressions. For
4559 example, warn if an unsigned variable is compared against zero with
4560 @samp{<} or @samp{>=}. This warning is also enabled by
4563 @item -Wbad-function-cast @r{(C and Objective-C only)}
4564 @opindex Wbad-function-cast
4565 @opindex Wno-bad-function-cast
4566 Warn whenever a function call is cast to a non-matching type.
4567 For example, warn if @code{int malloc()} is cast to @code{anything *}.
4569 @item -Wc90-c99-compat @r{(C and Objective-C only)}
4570 @opindex Wc90-c99-compat
4571 @opindex Wno-c90-c99-compat
4572 Warn about features not present in ISO C90, but present in ISO C99.
4573 For instance, warn about use of variable length arrays, @code{long long}
4574 type, @code{bool} type, compound literals, designated initializers, and so
4575 on. This option is independent of the standards mode. Warnings are disabled
4576 in the expression that follows @code{__extension__}.
4578 @item -Wc99-c11-compat @r{(C and Objective-C only)}
4579 @opindex Wc99-c11-compat
4580 @opindex Wno-c99-c11-compat
4581 Warn about features not present in ISO C99, but present in ISO C11.
4582 For instance, warn about use of anonymous structures and unions,
4583 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
4584 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
4585 and so on. This option is independent of the standards mode. Warnings are
4586 disabled in the expression that follows @code{__extension__}.
4588 @item -Wc++-compat @r{(C and Objective-C only)}
4589 Warn about ISO C constructs that are outside of the common subset of
4590 ISO C and ISO C++, e.g.@: request for implicit conversion from
4591 @code{void *} to a pointer to non-@code{void} type.
4593 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4594 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4595 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4596 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4597 enabled by @option{-Wall}.
4601 @opindex Wno-cast-qual
4602 Warn whenever a pointer is cast so as to remove a type qualifier from
4603 the target type. For example, warn if a @code{const char *} is cast
4604 to an ordinary @code{char *}.
4606 Also warn when making a cast that introduces a type qualifier in an
4607 unsafe way. For example, casting @code{char **} to @code{const char **}
4608 is unsafe, as in this example:
4611 /* p is char ** value. */
4612 const char **q = (const char **) p;
4613 /* Assignment of readonly string to const char * is OK. */
4615 /* Now char** pointer points to read-only memory. */
4620 @opindex Wcast-align
4621 @opindex Wno-cast-align
4622 Warn whenever a pointer is cast such that the required alignment of the
4623 target is increased. For example, warn if a @code{char *} is cast to
4624 an @code{int *} on machines where integers can only be accessed at
4625 two- or four-byte boundaries.
4627 @item -Wwrite-strings
4628 @opindex Wwrite-strings
4629 @opindex Wno-write-strings
4630 When compiling C, give string constants the type @code{const
4631 char[@var{length}]} so that copying the address of one into a
4632 non-@code{const} @code{char *} pointer produces a warning. These
4633 warnings help you find at compile time code that can try to write
4634 into a string constant, but only if you have been very careful about
4635 using @code{const} in declarations and prototypes. Otherwise, it is
4636 just a nuisance. This is why we did not make @option{-Wall} request
4639 When compiling C++, warn about the deprecated conversion from string
4640 literals to @code{char *}. This warning is enabled by default for C++
4645 @opindex Wno-clobbered
4646 Warn for variables that might be changed by @samp{longjmp} or
4647 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4649 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
4650 @opindex Wconditionally-supported
4651 @opindex Wno-conditionally-supported
4652 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
4655 @opindex Wconversion
4656 @opindex Wno-conversion
4657 Warn for implicit conversions that may alter a value. This includes
4658 conversions between real and integer, like @code{abs (x)} when
4659 @code{x} is @code{double}; conversions between signed and unsigned,
4660 like @code{unsigned ui = -1}; and conversions to smaller types, like
4661 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4662 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4663 changed by the conversion like in @code{abs (2.0)}. Warnings about
4664 conversions between signed and unsigned integers can be disabled by
4665 using @option{-Wno-sign-conversion}.
4667 For C++, also warn for confusing overload resolution for user-defined
4668 conversions; and conversions that never use a type conversion
4669 operator: conversions to @code{void}, the same type, a base class or a
4670 reference to them. Warnings about conversions between signed and
4671 unsigned integers are disabled by default in C++ unless
4672 @option{-Wsign-conversion} is explicitly enabled.
4674 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4675 @opindex Wconversion-null
4676 @opindex Wno-conversion-null
4677 Do not warn for conversions between @code{NULL} and non-pointer
4678 types. @option{-Wconversion-null} is enabled by default.
4680 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
4681 @opindex Wzero-as-null-pointer-constant
4682 @opindex Wno-zero-as-null-pointer-constant
4683 Warn when a literal '0' is used as null pointer constant. This can
4684 be useful to facilitate the conversion to @code{nullptr} in C++11.
4688 @opindex Wno-date-time
4689 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
4690 are encountered as they might prevent bit-wise-identical reproducible
4693 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
4694 @opindex Wdelete-incomplete
4695 @opindex Wno-delete-incomplete
4696 Warn when deleting a pointer to incomplete type, which may cause
4697 undefined behavior at runtime. This warning is enabled by default.
4699 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
4700 @opindex Wuseless-cast
4701 @opindex Wno-useless-cast
4702 Warn when an expression is casted to its own type.
4705 @opindex Wempty-body
4706 @opindex Wno-empty-body
4707 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4708 while} statement. This warning is also enabled by @option{-Wextra}.
4710 @item -Wenum-compare
4711 @opindex Wenum-compare
4712 @opindex Wno-enum-compare
4713 Warn about a comparison between values of different enumerated types.
4714 In C++ enumeral mismatches in conditional expressions are also
4715 diagnosed and the warning is enabled by default. In C this warning is
4716 enabled by @option{-Wall}.
4718 @item -Wjump-misses-init @r{(C, Objective-C only)}
4719 @opindex Wjump-misses-init
4720 @opindex Wno-jump-misses-init
4721 Warn if a @code{goto} statement or a @code{switch} statement jumps
4722 forward across the initialization of a variable, or jumps backward to a
4723 label after the variable has been initialized. This only warns about
4724 variables that are initialized when they are declared. This warning is
4725 only supported for C and Objective-C; in C++ this sort of branch is an
4728 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4729 can be disabled with the @option{-Wno-jump-misses-init} option.
4731 @item -Wsign-compare
4732 @opindex Wsign-compare
4733 @opindex Wno-sign-compare
4734 @cindex warning for comparison of signed and unsigned values
4735 @cindex comparison of signed and unsigned values, warning
4736 @cindex signed and unsigned values, comparison warning
4737 Warn when a comparison between signed and unsigned values could produce
4738 an incorrect result when the signed value is converted to unsigned.
4739 This warning is also enabled by @option{-Wextra}; to get the other warnings
4740 of @option{-Wextra} without this warning, use @option{-Wextra -Wno-sign-compare}.
4742 @item -Wsign-conversion
4743 @opindex Wsign-conversion
4744 @opindex Wno-sign-conversion
4745 Warn for implicit conversions that may change the sign of an integer
4746 value, like assigning a signed integer expression to an unsigned
4747 integer variable. An explicit cast silences the warning. In C, this
4748 option is enabled also by @option{-Wconversion}.
4750 @item -Wfloat-conversion
4751 @opindex Wfloat-conversion
4752 @opindex Wno-float-conversion
4753 Warn for implicit conversions that reduce the precision of a real value.
4754 This includes conversions from real to integer, and from higher precision
4755 real to lower precision real values. This option is also enabled by
4756 @option{-Wconversion}.
4758 @item -Wsizeof-pointer-memaccess
4759 @opindex Wsizeof-pointer-memaccess
4760 @opindex Wno-sizeof-pointer-memaccess
4761 Warn for suspicious length parameters to certain string and memory built-in
4762 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
4763 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
4764 but a pointer, and suggests a possible fix, or about
4765 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
4768 @item -Wsizeof-array-argument
4769 @opindex Wsizeof-array-argument
4770 @opindex Wno-sizeof-array-argument
4771 Warn when the @code{sizeof} operator is applied to a parameter that is
4772 declared as an array in a function definition. This warning is enabled by
4773 default for C and C++ programs.
4775 @item -Wmemset-transposed-args
4776 @opindex Wmemset-transposed-args
4777 @opindex Wno-memset-transposed-args
4778 Warn for suspicious calls to the @code{memset} built-in function, if the
4779 second argument is not zero and the third argument is zero. This warns e.g.@
4780 about @code{memset (buf, sizeof buf, 0)} where most probably
4781 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
4782 is only emitted if the third argument is literal zero, if it is some expression
4783 that is folded to zero, or e.g. a cast of zero to some type etc., it
4784 is far less likely that user has mistakenly exchanged the arguments and
4785 no warning is emitted. This warning is enabled by @option{-Wall}.
4789 @opindex Wno-address
4790 Warn about suspicious uses of memory addresses. These include using
4791 the address of a function in a conditional expression, such as
4792 @code{void func(void); if (func)}, and comparisons against the memory
4793 address of a string literal, such as @code{if (x == "abc")}. Such
4794 uses typically indicate a programmer error: the address of a function
4795 always evaluates to true, so their use in a conditional usually
4796 indicate that the programmer forgot the parentheses in a function
4797 call; and comparisons against string literals result in unspecified
4798 behavior and are not portable in C, so they usually indicate that the
4799 programmer intended to use @code{strcmp}. This warning is enabled by
4803 @opindex Wlogical-op
4804 @opindex Wno-logical-op
4805 Warn about suspicious uses of logical operators in expressions.
4806 This includes using logical operators in contexts where a
4807 bit-wise operator is likely to be expected.
4809 @item -Wlogical-not-parentheses
4810 @opindex Wlogical-not-parentheses
4811 @opindex Wno-logical-not-parentheses
4812 Warn about logical not used on the left hand side operand of a comparison.
4813 This option does not warn if the RHS operand is of a boolean type. Its
4814 purpose is to detect suspicious code like the following:
4818 if (!a > 1) @{ @dots{} @}
4821 It is possible to suppress the warning by wrapping the LHS into
4824 if ((!a) > 1) @{ @dots{} @}
4827 This warning is enabled by @option{-Wall}.
4829 @item -Waggregate-return
4830 @opindex Waggregate-return
4831 @opindex Wno-aggregate-return
4832 Warn if any functions that return structures or unions are defined or
4833 called. (In languages where you can return an array, this also elicits
4836 @item -Wno-aggressive-loop-optimizations
4837 @opindex Wno-aggressive-loop-optimizations
4838 @opindex Waggressive-loop-optimizations
4839 Warn if in a loop with constant number of iterations the compiler detects
4840 undefined behavior in some statement during one or more of the iterations.
4842 @item -Wno-attributes
4843 @opindex Wno-attributes
4844 @opindex Wattributes
4845 Do not warn if an unexpected @code{__attribute__} is used, such as
4846 unrecognized attributes, function attributes applied to variables,
4847 etc. This does not stop errors for incorrect use of supported
4850 @item -Wno-builtin-macro-redefined
4851 @opindex Wno-builtin-macro-redefined
4852 @opindex Wbuiltin-macro-redefined
4853 Do not warn if certain built-in macros are redefined. This suppresses
4854 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4855 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4857 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4858 @opindex Wstrict-prototypes
4859 @opindex Wno-strict-prototypes
4860 Warn if a function is declared or defined without specifying the
4861 argument types. (An old-style function definition is permitted without
4862 a warning if preceded by a declaration that specifies the argument
4865 @item -Wold-style-declaration @r{(C and Objective-C only)}
4866 @opindex Wold-style-declaration
4867 @opindex Wno-old-style-declaration
4868 Warn for obsolescent usages, according to the C Standard, in a
4869 declaration. For example, warn if storage-class specifiers like
4870 @code{static} are not the first things in a declaration. This warning
4871 is also enabled by @option{-Wextra}.
4873 @item -Wold-style-definition @r{(C and Objective-C only)}
4874 @opindex Wold-style-definition
4875 @opindex Wno-old-style-definition
4876 Warn if an old-style function definition is used. A warning is given
4877 even if there is a previous prototype.
4879 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4880 @opindex Wmissing-parameter-type
4881 @opindex Wno-missing-parameter-type
4882 A function parameter is declared without a type specifier in K&R-style
4889 This warning is also enabled by @option{-Wextra}.
4891 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4892 @opindex Wmissing-prototypes
4893 @opindex Wno-missing-prototypes
4894 Warn if a global function is defined without a previous prototype
4895 declaration. This warning is issued even if the definition itself
4896 provides a prototype. Use this option to detect global functions
4897 that do not have a matching prototype declaration in a header file.
4898 This option is not valid for C++ because all function declarations
4899 provide prototypes and a non-matching declaration will declare an
4900 overload rather than conflict with an earlier declaration.
4901 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
4903 @item -Wmissing-declarations
4904 @opindex Wmissing-declarations
4905 @opindex Wno-missing-declarations
4906 Warn if a global function is defined without a previous declaration.
4907 Do so even if the definition itself provides a prototype.
4908 Use this option to detect global functions that are not declared in
4909 header files. In C, no warnings are issued for functions with previous
4910 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
4911 missing prototypes. In C++, no warnings are issued for function templates,
4912 or for inline functions, or for functions in anonymous namespaces.
4914 @item -Wmissing-field-initializers
4915 @opindex Wmissing-field-initializers
4916 @opindex Wno-missing-field-initializers
4920 Warn if a structure's initializer has some fields missing. For
4921 example, the following code causes such a warning, because
4922 @code{x.h} is implicitly zero:
4925 struct s @{ int f, g, h; @};
4926 struct s x = @{ 3, 4 @};
4929 This option does not warn about designated initializers, so the following
4930 modification does not trigger a warning:
4933 struct s @{ int f, g, h; @};
4934 struct s x = @{ .f = 3, .g = 4 @};
4937 In C++ this option does not warn either about the empty @{ @}
4938 initializer, for example:
4941 struct s @{ int f, g, h; @};
4945 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4946 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
4948 @item -Wno-multichar
4949 @opindex Wno-multichar
4951 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4952 Usually they indicate a typo in the user's code, as they have
4953 implementation-defined values, and should not be used in portable code.
4955 @item -Wnormalized@r{[}=@r{<}none@r{|}id@r{|}nfc@r{|}nfkc@r{>]}
4956 @opindex Wnormalized=
4957 @opindex Wnormalized
4958 @opindex Wno-normalized
4961 @cindex character set, input normalization
4962 In ISO C and ISO C++, two identifiers are different if they are
4963 different sequences of characters. However, sometimes when characters
4964 outside the basic ASCII character set are used, you can have two
4965 different character sequences that look the same. To avoid confusion,
4966 the ISO 10646 standard sets out some @dfn{normalization rules} which
4967 when applied ensure that two sequences that look the same are turned into
4968 the same sequence. GCC can warn you if you are using identifiers that
4969 have not been normalized; this option controls that warning.
4971 There are four levels of warning supported by GCC@. The default is
4972 @option{-Wnormalized=nfc}, which warns about any identifier that is
4973 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4974 recommended form for most uses. It is equivalent to
4975 @option{-Wnormalized}.
4977 Unfortunately, there are some characters allowed in identifiers by
4978 ISO C and ISO C++ that, when turned into NFC, are not allowed in
4979 identifiers. That is, there's no way to use these symbols in portable
4980 ISO C or C++ and have all your identifiers in NFC@.
4981 @option{-Wnormalized=id} suppresses the warning for these characters.
4982 It is hoped that future versions of the standards involved will correct
4983 this, which is why this option is not the default.
4985 You can switch the warning off for all characters by writing
4986 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
4987 only do this if you are using some other normalization scheme (like
4988 ``D''), because otherwise you can easily create bugs that are
4989 literally impossible to see.
4991 Some characters in ISO 10646 have distinct meanings but look identical
4992 in some fonts or display methodologies, especially once formatting has
4993 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4994 LETTER N'', displays just like a regular @code{n} that has been
4995 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4996 normalization scheme to convert all these into a standard form as
4997 well, and GCC warns if your code is not in NFKC if you use
4998 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4999 about every identifier that contains the letter O because it might be
5000 confused with the digit 0, and so is not the default, but may be
5001 useful as a local coding convention if the programming environment
5002 cannot be fixed to display these characters distinctly.
5004 @item -Wno-deprecated
5005 @opindex Wno-deprecated
5006 @opindex Wdeprecated
5007 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
5009 @item -Wno-deprecated-declarations
5010 @opindex Wno-deprecated-declarations
5011 @opindex Wdeprecated-declarations
5012 Do not warn about uses of functions (@pxref{Function Attributes}),
5013 variables (@pxref{Variable Attributes}), and types (@pxref{Type
5014 Attributes}) marked as deprecated by using the @code{deprecated}
5018 @opindex Wno-overflow
5020 Do not warn about compile-time overflow in constant expressions.
5025 Warn about One Definition Rule violations during link-time optimization.
5026 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
5029 @opindex Wopenm-simd
5030 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
5031 simd directive set by user. The @option{-fsimd-cost-model=unlimited} can
5032 be used to relax the cost model.
5034 @item -Woverride-init @r{(C and Objective-C only)}
5035 @opindex Woverride-init
5036 @opindex Wno-override-init
5040 Warn if an initialized field without side effects is overridden when
5041 using designated initializers (@pxref{Designated Inits, , Designated
5044 This warning is included in @option{-Wextra}. To get other
5045 @option{-Wextra} warnings without this one, use @option{-Wextra
5046 -Wno-override-init}.
5051 Warn if a structure is given the packed attribute, but the packed
5052 attribute has no effect on the layout or size of the structure.
5053 Such structures may be mis-aligned for little benefit. For
5054 instance, in this code, the variable @code{f.x} in @code{struct bar}
5055 is misaligned even though @code{struct bar} does not itself
5056 have the packed attribute:
5063 @} __attribute__((packed));
5071 @item -Wpacked-bitfield-compat
5072 @opindex Wpacked-bitfield-compat
5073 @opindex Wno-packed-bitfield-compat
5074 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
5075 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
5076 the change can lead to differences in the structure layout. GCC
5077 informs you when the offset of such a field has changed in GCC 4.4.
5078 For example there is no longer a 4-bit padding between field @code{a}
5079 and @code{b} in this structure:
5086 @} __attribute__ ((packed));
5089 This warning is enabled by default. Use
5090 @option{-Wno-packed-bitfield-compat} to disable this warning.
5095 Warn if padding is included in a structure, either to align an element
5096 of the structure or to align the whole structure. Sometimes when this
5097 happens it is possible to rearrange the fields of the structure to
5098 reduce the padding and so make the structure smaller.
5100 @item -Wredundant-decls
5101 @opindex Wredundant-decls
5102 @opindex Wno-redundant-decls
5103 Warn if anything is declared more than once in the same scope, even in
5104 cases where multiple declaration is valid and changes nothing.
5106 @item -Wnested-externs @r{(C and Objective-C only)}
5107 @opindex Wnested-externs
5108 @opindex Wno-nested-externs
5109 Warn if an @code{extern} declaration is encountered within a function.
5111 @item -Wno-inherited-variadic-ctor
5112 @opindex Winherited-variadic-ctor
5113 @opindex Wno-inherited-variadic-ctor
5114 Suppress warnings about use of C++11 inheriting constructors when the
5115 base class inherited from has a C variadic constructor; the warning is
5116 on by default because the ellipsis is not inherited.
5121 Warn if a function that is declared as inline cannot be inlined.
5122 Even with this option, the compiler does not warn about failures to
5123 inline functions declared in system headers.
5125 The compiler uses a variety of heuristics to determine whether or not
5126 to inline a function. For example, the compiler takes into account
5127 the size of the function being inlined and the amount of inlining
5128 that has already been done in the current function. Therefore,
5129 seemingly insignificant changes in the source program can cause the
5130 warnings produced by @option{-Winline} to appear or disappear.
5132 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
5133 @opindex Wno-invalid-offsetof
5134 @opindex Winvalid-offsetof
5135 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
5136 type. According to the 2014 ISO C++ standard, applying @samp{offsetof}
5137 to a non-standard-layout type is undefined. In existing C++ implementations,
5138 however, @samp{offsetof} typically gives meaningful results.
5139 This flag is for users who are aware that they are
5140 writing nonportable code and who have deliberately chosen to ignore the
5143 The restrictions on @samp{offsetof} may be relaxed in a future version
5144 of the C++ standard.
5146 @item -Wno-int-to-pointer-cast
5147 @opindex Wno-int-to-pointer-cast
5148 @opindex Wint-to-pointer-cast
5149 Suppress warnings from casts to pointer type of an integer of a
5150 different size. In C++, casting to a pointer type of smaller size is
5151 an error. @option{Wint-to-pointer-cast} is enabled by default.
5154 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
5155 @opindex Wno-pointer-to-int-cast
5156 @opindex Wpointer-to-int-cast
5157 Suppress warnings from casts from a pointer to an integer type of a
5161 @opindex Winvalid-pch
5162 @opindex Wno-invalid-pch
5163 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
5164 the search path but can't be used.
5168 @opindex Wno-long-long
5169 Warn if @samp{long long} type is used. This is enabled by either
5170 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
5171 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
5173 @item -Wvariadic-macros
5174 @opindex Wvariadic-macros
5175 @opindex Wno-variadic-macros
5176 Warn if variadic macros are used in ISO C90 mode, or if the GNU
5177 alternate syntax is used in ISO C99 mode. This is enabled by either
5178 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
5179 messages, use @option{-Wno-variadic-macros}.
5183 @opindex Wno-varargs
5184 Warn upon questionable usage of the macros used to handle variable
5185 arguments like @samp{va_start}. This is default. To inhibit the
5186 warning messages, use @option{-Wno-varargs}.
5188 @item -Wvector-operation-performance
5189 @opindex Wvector-operation-performance
5190 @opindex Wno-vector-operation-performance
5191 Warn if vector operation is not implemented via SIMD capabilities of the
5192 architecture. Mainly useful for the performance tuning.
5193 Vector operation can be implemented @code{piecewise}, which means that the
5194 scalar operation is performed on every vector element;
5195 @code{in parallel}, which means that the vector operation is implemented
5196 using scalars of wider type, which normally is more performance efficient;
5197 and @code{as a single scalar}, which means that vector fits into a
5200 @item -Wno-virtual-move-assign
5201 @opindex Wvirtual-move-assign
5202 @opindex Wno-virtual-move-assign
5203 Suppress warnings about inheriting from a virtual base with a
5204 non-trivial C++11 move assignment operator. This is dangerous because
5205 if the virtual base is reachable along more than one path, it will be
5206 moved multiple times, which can mean both objects end up in the
5207 moved-from state. If the move assignment operator is written to avoid
5208 moving from a moved-from object, this warning can be disabled.
5213 Warn if variable length array is used in the code.
5214 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
5215 the variable length array.
5217 @item -Wvolatile-register-var
5218 @opindex Wvolatile-register-var
5219 @opindex Wno-volatile-register-var
5220 Warn if a register variable is declared volatile. The volatile
5221 modifier does not inhibit all optimizations that may eliminate reads
5222 and/or writes to register variables. This warning is enabled by
5225 @item -Wdisabled-optimization
5226 @opindex Wdisabled-optimization
5227 @opindex Wno-disabled-optimization
5228 Warn if a requested optimization pass is disabled. This warning does
5229 not generally indicate that there is anything wrong with your code; it
5230 merely indicates that GCC's optimizers are unable to handle the code
5231 effectively. Often, the problem is that your code is too big or too
5232 complex; GCC refuses to optimize programs when the optimization
5233 itself is likely to take inordinate amounts of time.
5235 @item -Wpointer-sign @r{(C and Objective-C only)}
5236 @opindex Wpointer-sign
5237 @opindex Wno-pointer-sign
5238 Warn for pointer argument passing or assignment with different signedness.
5239 This option is only supported for C and Objective-C@. It is implied by
5240 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
5241 @option{-Wno-pointer-sign}.
5243 @item -Wstack-protector
5244 @opindex Wstack-protector
5245 @opindex Wno-stack-protector
5246 This option is only active when @option{-fstack-protector} is active. It
5247 warns about functions that are not protected against stack smashing.
5249 @item -Woverlength-strings
5250 @opindex Woverlength-strings
5251 @opindex Wno-overlength-strings
5252 Warn about string constants that are longer than the ``minimum
5253 maximum'' length specified in the C standard. Modern compilers
5254 generally allow string constants that are much longer than the
5255 standard's minimum limit, but very portable programs should avoid
5256 using longer strings.
5258 The limit applies @emph{after} string constant concatenation, and does
5259 not count the trailing NUL@. In C90, the limit was 509 characters; in
5260 C99, it was raised to 4095. C++98 does not specify a normative
5261 minimum maximum, so we do not diagnose overlength strings in C++@.
5263 This option is implied by @option{-Wpedantic}, and can be disabled with
5264 @option{-Wno-overlength-strings}.
5266 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
5267 @opindex Wunsuffixed-float-constants
5269 Issue a warning for any floating constant that does not have
5270 a suffix. When used together with @option{-Wsystem-headers} it
5271 warns about such constants in system header files. This can be useful
5272 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
5273 from the decimal floating-point extension to C99.
5275 @item -Wno-designated-init @r{(C and Objective-C only)}
5276 Suppress warnings when a positional initializer is used to initialize
5277 a structure that has been marked with the @code{designated_init}
5282 @node Debugging Options
5283 @section Options for Debugging Your Program or GCC
5284 @cindex options, debugging
5285 @cindex debugging information options
5287 GCC has various special options that are used for debugging
5288 either your program or GCC:
5293 Produce debugging information in the operating system's native format
5294 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
5297 On most systems that use stabs format, @option{-g} enables use of extra
5298 debugging information that only GDB can use; this extra information
5299 makes debugging work better in GDB but probably makes other debuggers
5301 refuse to read the program. If you want to control for certain whether
5302 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
5303 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
5305 GCC allows you to use @option{-g} with
5306 @option{-O}. The shortcuts taken by optimized code may occasionally
5307 produce surprising results: some variables you declared may not exist
5308 at all; flow of control may briefly move where you did not expect it;
5309 some statements may not be executed because they compute constant
5310 results or their values are already at hand; some statements may
5311 execute in different places because they have been moved out of loops.
5313 Nevertheless it proves possible to debug optimized output. This makes
5314 it reasonable to use the optimizer for programs that might have bugs.
5316 The following options are useful when GCC is generated with the
5317 capability for more than one debugging format.
5320 @opindex gsplit-dwarf
5321 Separate as much dwarf debugging information as possible into a
5322 separate output file with the extension .dwo. This option allows
5323 the build system to avoid linking files with debug information. To
5324 be useful, this option requires a debugger capable of reading .dwo
5329 Produce debugging information for use by GDB@. This means to use the
5330 most expressive format available (DWARF 2, stabs, or the native format
5331 if neither of those are supported), including GDB extensions if at all
5336 Generate dwarf .debug_pubnames and .debug_pubtypes sections.
5338 @item -ggnu-pubnames
5339 @opindex ggnu-pubnames
5340 Generate .debug_pubnames and .debug_pubtypes sections in a format
5341 suitable for conversion into a GDB@ index. This option is only useful
5342 with a linker that can produce GDB@ index version 7.
5346 Produce debugging information in stabs format (if that is supported),
5347 without GDB extensions. This is the format used by DBX on most BSD
5348 systems. On MIPS, Alpha and System V Release 4 systems this option
5349 produces stabs debugging output that is not understood by DBX or SDB@.
5350 On System V Release 4 systems this option requires the GNU assembler.
5352 @item -feliminate-unused-debug-symbols
5353 @opindex feliminate-unused-debug-symbols
5354 Produce debugging information in stabs format (if that is supported),
5355 for only symbols that are actually used.
5357 @item -femit-class-debug-always
5358 Instead of emitting debugging information for a C++ class in only one
5359 object file, emit it in all object files using the class. This option
5360 should be used only with debuggers that are unable to handle the way GCC
5361 normally emits debugging information for classes because using this
5362 option increases the size of debugging information by as much as a
5365 @item -fdebug-types-section
5366 @opindex fdebug-types-section
5367 @opindex fno-debug-types-section
5368 When using DWARF Version 4 or higher, type DIEs can be put into
5369 their own @code{.debug_types} section instead of making them part of the
5370 @code{.debug_info} section. It is more efficient to put them in a separate
5371 comdat sections since the linker can then remove duplicates.
5372 But not all DWARF consumers support @code{.debug_types} sections yet
5373 and on some objects @code{.debug_types} produces larger instead of smaller
5374 debugging information.
5378 Produce debugging information in stabs format (if that is supported),
5379 using GNU extensions understood only by the GNU debugger (GDB)@. The
5380 use of these extensions is likely to make other debuggers crash or
5381 refuse to read the program.
5385 Produce debugging information in COFF format (if that is supported).
5386 This is the format used by SDB on most System V systems prior to
5391 Produce debugging information in XCOFF format (if that is supported).
5392 This is the format used by the DBX debugger on IBM RS/6000 systems.
5396 Produce debugging information in XCOFF format (if that is supported),
5397 using GNU extensions understood only by the GNU debugger (GDB)@. The
5398 use of these extensions is likely to make other debuggers crash or
5399 refuse to read the program, and may cause assemblers other than the GNU
5400 assembler (GAS) to fail with an error.
5402 @item -gdwarf-@var{version}
5403 @opindex gdwarf-@var{version}
5404 Produce debugging information in DWARF format (if that is supported).
5405 The value of @var{version} may be either 2, 3 or 4; the default version
5406 for most targets is 4.
5408 Note that with DWARF Version 2, some ports require and always
5409 use some non-conflicting DWARF 3 extensions in the unwind tables.
5411 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
5412 for maximum benefit.
5414 @item -grecord-gcc-switches
5415 @opindex grecord-gcc-switches
5416 This switch causes the command-line options used to invoke the
5417 compiler that may affect code generation to be appended to the
5418 DW_AT_producer attribute in DWARF debugging information. The options
5419 are concatenated with spaces separating them from each other and from
5420 the compiler version. See also @option{-frecord-gcc-switches} for another
5421 way of storing compiler options into the object file. This is the default.
5423 @item -gno-record-gcc-switches
5424 @opindex gno-record-gcc-switches
5425 Disallow appending command-line options to the DW_AT_producer attribute
5426 in DWARF debugging information.
5428 @item -gstrict-dwarf
5429 @opindex gstrict-dwarf
5430 Disallow using extensions of later DWARF standard version than selected
5431 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
5432 DWARF extensions from later standard versions is allowed.
5434 @item -gno-strict-dwarf
5435 @opindex gno-strict-dwarf
5436 Allow using extensions of later DWARF standard version than selected with
5437 @option{-gdwarf-@var{version}}.
5439 @item -gz@r{[}=@var{type}@r{]}
5441 Produce compressed debug sections in DWARF format, if that is supported.
5442 If @var{type} is not given, the default type depends on the capabilities
5443 of the assembler and linker used. @var{type} may be one of
5444 @option{none} (don't compress debug sections), @option{zlib} (use zlib
5445 compression in ELF gABI format), or @option{zlib-gnu} (use zlib
5446 compression in traditional GNU format). If the linker doesn't support
5447 writing compressed debug sections, the option is rejected. Otherwise,
5448 if the assembler does not support them, @option{-gz} is silently ignored
5449 when producing object files.
5453 Produce debugging information in Alpha/VMS debug format (if that is
5454 supported). This is the format used by DEBUG on Alpha/VMS systems.
5457 @itemx -ggdb@var{level}
5458 @itemx -gstabs@var{level}
5459 @itemx -gcoff@var{level}
5460 @itemx -gxcoff@var{level}
5461 @itemx -gvms@var{level}
5462 Request debugging information and also use @var{level} to specify how
5463 much information. The default level is 2.
5465 Level 0 produces no debug information at all. Thus, @option{-g0} negates
5468 Level 1 produces minimal information, enough for making backtraces in
5469 parts of the program that you don't plan to debug. This includes
5470 descriptions of functions and external variables, and line number
5471 tables, but no information about local variables.
5473 Level 3 includes extra information, such as all the macro definitions
5474 present in the program. Some debuggers support macro expansion when
5475 you use @option{-g3}.
5477 @option{-gdwarf-2} does not accept a concatenated debug level, because
5478 GCC used to support an option @option{-gdwarf} that meant to generate
5479 debug information in version 1 of the DWARF format (which is very
5480 different from version 2), and it would have been too confusing. That
5481 debug format is long obsolete, but the option cannot be changed now.
5482 Instead use an additional @option{-g@var{level}} option to change the
5483 debug level for DWARF.
5487 Turn off generation of debug info, if leaving out this option
5488 generates it, or turn it on at level 2 otherwise. The position of this
5489 argument in the command line does not matter; it takes effect after all
5490 other options are processed, and it does so only once, no matter how
5491 many times it is given. This is mainly intended to be used with
5492 @option{-fcompare-debug}.
5494 @item -fsanitize=address
5495 @opindex fsanitize=address
5496 Enable AddressSanitizer, a fast memory error detector.
5497 Memory access instructions will be instrumented to detect
5498 out-of-bounds and use-after-free bugs.
5499 See @uref{http://code.google.com/p/address-sanitizer/} for
5500 more details. The run-time behavior can be influenced using the
5501 @env{ASAN_OPTIONS} environment variable; see
5502 @url{https://code.google.com/p/address-sanitizer/wiki/Flags#Run-time_flags} for
5503 a list of supported options.
5505 @item -fsanitize=kernel-address
5506 @opindex fsanitize=kernel-address
5507 Enable AddressSanitizer for Linux kernel.
5508 See @uref{http://code.google.com/p/address-sanitizer/wiki/AddressSanitizerForKernel} for more details.
5510 @item -fsanitize=thread
5511 @opindex fsanitize=thread
5512 Enable ThreadSanitizer, a fast data race detector.
5513 Memory access instructions will be instrumented to detect
5514 data race bugs. See @uref{http://code.google.com/p/thread-sanitizer/} for more
5515 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
5516 environment variable; see
5517 @url{https://code.google.com/p/thread-sanitizer/wiki/Flags} for a list of
5520 @item -fsanitize=leak
5521 @opindex fsanitize=leak
5522 Enable LeakSanitizer, a memory leak detector.
5523 This option only matters for linking of executables and if neither
5524 @option{-fsanitize=address} nor @option{-fsanitize=thread} is used. In that
5525 case it will link the executable against a library that overrides @code{malloc}
5526 and other allocator functions. See
5527 @uref{https://code.google.com/p/address-sanitizer/wiki/LeakSanitizer} for more
5528 details. The run-time behavior can be influenced using the
5529 @env{LSAN_OPTIONS} environment variable.
5531 @item -fsanitize=undefined
5532 @opindex fsanitize=undefined
5533 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
5534 Various computations will be instrumented to detect undefined behavior
5535 at runtime. Current suboptions are:
5539 @item -fsanitize=shift
5540 @opindex fsanitize=shift
5541 This option enables checking that the result of a shift operation is
5542 not undefined. Note that what exactly is considered undefined differs
5543 slightly between C and C++, as well as between ISO C90 and C99, etc.
5545 @item -fsanitize=integer-divide-by-zero
5546 @opindex fsanitize=integer-divide-by-zero
5547 Detect integer division by zero as well as @code{INT_MIN / -1} division.
5549 @item -fsanitize=unreachable
5550 @opindex fsanitize=unreachable
5551 With this option, the compiler will turn the @code{__builtin_unreachable}
5552 call into a diagnostics message call instead. When reaching the
5553 @code{__builtin_unreachable} call, the behavior is undefined.
5555 @item -fsanitize=vla-bound
5556 @opindex fsanitize=vla-bound
5557 This option instructs the compiler to check that the size of a variable
5558 length array is positive. This option does not have any effect in
5559 @option{-std=c++14} mode, as the standard requires the exception be thrown
5562 @item -fsanitize=null
5563 @opindex fsanitize=null
5564 This option enables pointer checking. Particularly, the application
5565 built with this option turned on will issue an error message when it
5566 tries to dereference a NULL pointer, or if a reference (possibly an
5567 rvalue reference) is bound to a NULL pointer, or if a method is invoked
5568 on an object pointed by a NULL pointer.
5570 @item -fsanitize=return
5571 @opindex fsanitize=return
5572 This option enables return statement checking. Programs
5573 built with this option turned on will issue an error message
5574 when the end of a non-void function is reached without actually
5575 returning a value. This option works in C++ only.
5577 @item -fsanitize=signed-integer-overflow
5578 @opindex fsanitize=signed-integer-overflow
5579 This option enables signed integer overflow checking. We check that
5580 the result of @code{+}, @code{*}, and both unary and binary @code{-}
5581 does not overflow in the signed arithmetics. Note, integer promotion
5582 rules must be taken into account. That is, the following is not an
5585 signed char a = SCHAR_MAX;
5589 @item -fsanitize=bounds
5590 @opindex fsanitize=bounds
5591 This option enables instrumentation of array bounds. Various out of bounds
5592 accesses are detected. Flexible array members and initializers of variables
5593 with static storage are not instrumented.
5595 @item -fsanitize=alignment
5596 @opindex fsanitize=alignment
5598 This option enables checking of alignment of pointers when they are
5599 dereferenced, or when a reference is bound to insufficiently aligned target,
5600 or when a method or constructor is invoked on insufficiently aligned object.
5602 @item -fsanitize=object-size
5603 @opindex fsanitize=object-size
5604 This option enables instrumentation of memory references using the
5605 @code{__builtin_object_size} function. Various out of bounds pointer
5606 accesses are detected.
5608 @item -fsanitize=float-divide-by-zero
5609 @opindex fsanitize=float-divide-by-zero
5610 Detect floating-point division by zero. Unlike other similar options,
5611 @option{-fsanitize=float-divide-by-zero} is not enabled by
5612 @option{-fsanitize=undefined}, since floating-point division by zero can
5613 be a legitimate way of obtaining infinities and NaNs.
5615 @item -fsanitize=float-cast-overflow
5616 @opindex fsanitize=float-cast-overflow
5617 This option enables floating-point type to integer conversion checking.
5618 We check that the result of the conversion does not overflow.
5619 This option does not work well with @code{FE_INVALID} exceptions enabled.
5621 @item -fsanitize=nonnull-attribute
5622 @opindex fsanitize=nonnull-attribute
5624 This option enables instrumentation of calls, checking whether null values
5625 are not passed to arguments marked as requiring a non-null value by the
5626 @code{nonnull} function attribute.
5628 @item -fsanitize=returns-nonnull-attribute
5629 @opindex fsanitize=returns-nonnull-attribute
5631 This option enables instrumentation of return statements in functions
5632 marked with @code{returns_nonnull} function attribute, to detect returning
5633 of null values from such functions.
5635 @item -fsanitize=bool
5636 @opindex fsanitize=bool
5638 This option enables instrumentation of loads from bool. If a value other
5639 than 0/1 is loaded, a run-time error is issued.
5641 @item -fsanitize=enum
5642 @opindex fsanitize=enum
5644 This option enables instrumentation of loads from an enum type. If
5645 a value outside the range of values for the enum type is loaded,
5646 a run-time error is issued.
5650 While @option{-ftrapv} causes traps for signed overflows to be emitted,
5651 @option{-fsanitize=undefined} gives a diagnostic message.
5652 This currently works only for the C family of languages.
5654 @item -fasan-shadow-offset=@var{number}
5655 @opindex fasan-shadow-offset
5656 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
5657 It is useful for experimenting with different shadow memory layouts in
5658 Kernel AddressSanitizer.
5660 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
5661 @opindex fsanitize-recover
5662 @opindex fno-sanitize-recover
5663 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
5664 mentioned in comma-separated list of @var{opts}. Enabling this option
5665 for a sanitizer component would cause it to attempt to continue
5666 running the program as if no error happened. This means multiple
5667 runtime errors can be reported in a single program run, and the exit
5668 code of the program may indicate success even when errors
5669 have been reported. The @option{-fno-sanitize-recover=} can be used to alter
5670 this behavior, only the first detected error will be reported
5671 and program will exit after that with non-zero exit code.
5673 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
5674 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
5675 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero} and
5676 @option{-fsanitize=kernel-address}. For these sanitizers error recovery is turned on by default.
5678 Syntax without explicit @var{opts} parameter is deprecated. It is equivalent to
5679 @option{-fsanitize-recover=undefined,float-cast-overflow,float-divide-by-zero,kernel-address}.
5680 Similarly @option{-fno-sanitize-recover} is equivalent to
5681 @option{-fno-sanitize-recover=undefined,float-cast-overflow,float-divide-by-zero,kernel-address}.
5683 @item -fsanitize-undefined-trap-on-error
5684 @opindex fsanitize-undefined-trap-on-error
5685 The @option{-fsanitize-undefined-trap-on-error} instructs the compiler to
5686 report undefined behavior using @code{__builtin_trap ()} rather than
5687 a @code{libubsan} library routine. The advantage of this is that the
5688 @code{libubsan} library is not needed and will not be linked in, so this
5689 is usable even for use in freestanding environments.
5691 @item -fdump-final-insns@r{[}=@var{file}@r{]}
5692 @opindex fdump-final-insns
5693 Dump the final internal representation (RTL) to @var{file}. If the
5694 optional argument is omitted (or if @var{file} is @code{.}), the name
5695 of the dump file is determined by appending @code{.gkd} to the
5696 compilation output file name.
5698 @item -fcompare-debug@r{[}=@var{opts}@r{]}
5699 @opindex fcompare-debug
5700 @opindex fno-compare-debug
5701 If no error occurs during compilation, run the compiler a second time,
5702 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
5703 passed to the second compilation. Dump the final internal
5704 representation in both compilations, and print an error if they differ.
5706 If the equal sign is omitted, the default @option{-gtoggle} is used.
5708 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
5709 and nonzero, implicitly enables @option{-fcompare-debug}. If
5710 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
5711 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
5714 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
5715 is equivalent to @option{-fno-compare-debug}, which disables the dumping
5716 of the final representation and the second compilation, preventing even
5717 @env{GCC_COMPARE_DEBUG} from taking effect.
5719 To verify full coverage during @option{-fcompare-debug} testing, set
5720 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
5721 which GCC rejects as an invalid option in any actual compilation
5722 (rather than preprocessing, assembly or linking). To get just a
5723 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
5724 not overridden} will do.
5726 @item -fcompare-debug-second
5727 @opindex fcompare-debug-second
5728 This option is implicitly passed to the compiler for the second
5729 compilation requested by @option{-fcompare-debug}, along with options to
5730 silence warnings, and omitting other options that would cause
5731 side-effect compiler outputs to files or to the standard output. Dump
5732 files and preserved temporary files are renamed so as to contain the
5733 @code{.gk} additional extension during the second compilation, to avoid
5734 overwriting those generated by the first.
5736 When this option is passed to the compiler driver, it causes the
5737 @emph{first} compilation to be skipped, which makes it useful for little
5738 other than debugging the compiler proper.
5740 @item -feliminate-dwarf2-dups
5741 @opindex feliminate-dwarf2-dups
5742 Compress DWARF 2 debugging information by eliminating duplicated
5743 information about each symbol. This option only makes sense when
5744 generating DWARF 2 debugging information with @option{-gdwarf-2}.
5746 @item -femit-struct-debug-baseonly
5747 @opindex femit-struct-debug-baseonly
5748 Emit debug information for struct-like types
5749 only when the base name of the compilation source file
5750 matches the base name of file in which the struct is defined.
5752 This option substantially reduces the size of debugging information,
5753 but at significant potential loss in type information to the debugger.
5754 See @option{-femit-struct-debug-reduced} for a less aggressive option.
5755 See @option{-femit-struct-debug-detailed} for more detailed control.
5757 This option works only with DWARF 2.
5759 @item -femit-struct-debug-reduced
5760 @opindex femit-struct-debug-reduced
5761 Emit debug information for struct-like types
5762 only when the base name of the compilation source file
5763 matches the base name of file in which the type is defined,
5764 unless the struct is a template or defined in a system header.
5766 This option significantly reduces the size of debugging information,
5767 with some potential loss in type information to the debugger.
5768 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
5769 See @option{-femit-struct-debug-detailed} for more detailed control.
5771 This option works only with DWARF 2.
5773 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
5774 Specify the struct-like types
5775 for which the compiler generates debug information.
5776 The intent is to reduce duplicate struct debug information
5777 between different object files within the same program.
5779 This option is a detailed version of
5780 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
5781 which serves for most needs.
5783 A specification has the syntax@*
5784 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
5786 The optional first word limits the specification to
5787 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
5788 A struct type is used directly when it is the type of a variable, member.
5789 Indirect uses arise through pointers to structs.
5790 That is, when use of an incomplete struct is valid, the use is indirect.
5792 @samp{struct one direct; struct two * indirect;}.
5794 The optional second word limits the specification to
5795 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
5796 Generic structs are a bit complicated to explain.
5797 For C++, these are non-explicit specializations of template classes,
5798 or non-template classes within the above.
5799 Other programming languages have generics,
5800 but @option{-femit-struct-debug-detailed} does not yet implement them.
5802 The third word specifies the source files for those
5803 structs for which the compiler should emit debug information.
5804 The values @samp{none} and @samp{any} have the normal meaning.
5805 The value @samp{base} means that
5806 the base of name of the file in which the type declaration appears
5807 must match the base of the name of the main compilation file.
5808 In practice, this means that when compiling @file{foo.c}, debug information
5809 is generated for types declared in that file and @file{foo.h},
5810 but not other header files.
5811 The value @samp{sys} means those types satisfying @samp{base}
5812 or declared in system or compiler headers.
5814 You may need to experiment to determine the best settings for your application.
5816 The default is @option{-femit-struct-debug-detailed=all}.
5818 This option works only with DWARF 2.
5820 @item -fno-merge-debug-strings
5821 @opindex fmerge-debug-strings
5822 @opindex fno-merge-debug-strings
5823 Direct the linker to not merge together strings in the debugging
5824 information that are identical in different object files. Merging is
5825 not supported by all assemblers or linkers. Merging decreases the size
5826 of the debug information in the output file at the cost of increasing
5827 link processing time. Merging is enabled by default.
5829 @item -fdebug-prefix-map=@var{old}=@var{new}
5830 @opindex fdebug-prefix-map
5831 When compiling files in directory @file{@var{old}}, record debugging
5832 information describing them as in @file{@var{new}} instead.
5834 @item -fno-dwarf2-cfi-asm
5835 @opindex fdwarf2-cfi-asm
5836 @opindex fno-dwarf2-cfi-asm
5837 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
5838 instead of using GAS @code{.cfi_*} directives.
5840 @cindex @command{prof}
5843 Generate extra code to write profile information suitable for the
5844 analysis program @command{prof}. You must use this option when compiling
5845 the source files you want data about, and you must also use it when
5848 @cindex @command{gprof}
5851 Generate extra code to write profile information suitable for the
5852 analysis program @command{gprof}. You must use this option when compiling
5853 the source files you want data about, and you must also use it when
5858 Makes the compiler print out each function name as it is compiled, and
5859 print some statistics about each pass when it finishes.
5862 @opindex ftime-report
5863 Makes the compiler print some statistics about the time consumed by each
5864 pass when it finishes.
5867 @opindex fmem-report
5868 Makes the compiler print some statistics about permanent memory
5869 allocation when it finishes.
5871 @item -fmem-report-wpa
5872 @opindex fmem-report-wpa
5873 Makes the compiler print some statistics about permanent memory
5874 allocation for the WPA phase only.
5876 @item -fpre-ipa-mem-report
5877 @opindex fpre-ipa-mem-report
5878 @item -fpost-ipa-mem-report
5879 @opindex fpost-ipa-mem-report
5880 Makes the compiler print some statistics about permanent memory
5881 allocation before or after interprocedural optimization.
5883 @item -fprofile-report
5884 @opindex fprofile-report
5885 Makes the compiler print some statistics about consistency of the
5886 (estimated) profile and effect of individual passes.
5889 @opindex fstack-usage
5890 Makes the compiler output stack usage information for the program, on a
5891 per-function basis. The filename for the dump is made by appending
5892 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
5893 the output file, if explicitly specified and it is not an executable,
5894 otherwise it is the basename of the source file. An entry is made up
5899 The name of the function.
5903 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
5906 The qualifier @code{static} means that the function manipulates the stack
5907 statically: a fixed number of bytes are allocated for the frame on function
5908 entry and released on function exit; no stack adjustments are otherwise made
5909 in the function. The second field is this fixed number of bytes.
5911 The qualifier @code{dynamic} means that the function manipulates the stack
5912 dynamically: in addition to the static allocation described above, stack
5913 adjustments are made in the body of the function, for example to push/pop
5914 arguments around function calls. If the qualifier @code{bounded} is also
5915 present, the amount of these adjustments is bounded at compile time and
5916 the second field is an upper bound of the total amount of stack used by
5917 the function. If it is not present, the amount of these adjustments is
5918 not bounded at compile time and the second field only represents the
5921 @item -fprofile-arcs
5922 @opindex fprofile-arcs
5923 Add code so that program flow @dfn{arcs} are instrumented. During
5924 execution the program records how many times each branch and call is
5925 executed and how many times it is taken or returns. When the compiled
5926 program exits it saves this data to a file called
5927 @file{@var{auxname}.gcda} for each source file. The data may be used for
5928 profile-directed optimizations (@option{-fbranch-probabilities}), or for
5929 test coverage analysis (@option{-ftest-coverage}). Each object file's
5930 @var{auxname} is generated from the name of the output file, if
5931 explicitly specified and it is not the final executable, otherwise it is
5932 the basename of the source file. In both cases any suffix is removed
5933 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
5934 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
5935 @xref{Cross-profiling}.
5937 @cindex @command{gcov}
5941 This option is used to compile and link code instrumented for coverage
5942 analysis. The option is a synonym for @option{-fprofile-arcs}
5943 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
5944 linking). See the documentation for those options for more details.
5949 Compile the source files with @option{-fprofile-arcs} plus optimization
5950 and code generation options. For test coverage analysis, use the
5951 additional @option{-ftest-coverage} option. You do not need to profile
5952 every source file in a program.
5955 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
5956 (the latter implies the former).
5959 Run the program on a representative workload to generate the arc profile
5960 information. This may be repeated any number of times. You can run
5961 concurrent instances of your program, and provided that the file system
5962 supports locking, the data files will be correctly updated. Also
5963 @code{fork} calls are detected and correctly handled (double counting
5967 For profile-directed optimizations, compile the source files again with
5968 the same optimization and code generation options plus
5969 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
5970 Control Optimization}).
5973 For test coverage analysis, use @command{gcov} to produce human readable
5974 information from the @file{.gcno} and @file{.gcda} files. Refer to the
5975 @command{gcov} documentation for further information.
5979 With @option{-fprofile-arcs}, for each function of your program GCC
5980 creates a program flow graph, then finds a spanning tree for the graph.
5981 Only arcs that are not on the spanning tree have to be instrumented: the
5982 compiler adds code to count the number of times that these arcs are
5983 executed. When an arc is the only exit or only entrance to a block, the
5984 instrumentation code can be added to the block; otherwise, a new basic
5985 block must be created to hold the instrumentation code.
5988 @item -ftest-coverage
5989 @opindex ftest-coverage
5990 Produce a notes file that the @command{gcov} code-coverage utility
5991 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
5992 show program coverage. Each source file's note file is called
5993 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
5994 above for a description of @var{auxname} and instructions on how to
5995 generate test coverage data. Coverage data matches the source files
5996 more closely if you do not optimize.
5998 @item -fdbg-cnt-list
5999 @opindex fdbg-cnt-list
6000 Print the name and the counter upper bound for all debug counters.
6003 @item -fdbg-cnt=@var{counter-value-list}
6005 Set the internal debug counter upper bound. @var{counter-value-list}
6006 is a comma-separated list of @var{name}:@var{value} pairs
6007 which sets the upper bound of each debug counter @var{name} to @var{value}.
6008 All debug counters have the initial upper bound of @code{UINT_MAX};
6009 thus @code{dbg_cnt()} returns true always unless the upper bound
6010 is set by this option.
6011 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
6012 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
6014 @item -fenable-@var{kind}-@var{pass}
6015 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
6019 This is a set of options that are used to explicitly disable/enable
6020 optimization passes. These options are intended for use for debugging GCC.
6021 Compiler users should use regular options for enabling/disabling
6026 @item -fdisable-ipa-@var{pass}
6027 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6028 statically invoked in the compiler multiple times, the pass name should be
6029 appended with a sequential number starting from 1.
6031 @item -fdisable-rtl-@var{pass}
6032 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
6033 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
6034 statically invoked in the compiler multiple times, the pass name should be
6035 appended with a sequential number starting from 1. @var{range-list} is a
6036 comma-separated list of function ranges or assembler names. Each range is a number
6037 pair separated by a colon. The range is inclusive in both ends. If the range
6038 is trivial, the number pair can be simplified as a single number. If the
6039 function's call graph node's @var{uid} falls within one of the specified ranges,
6040 the @var{pass} is disabled for that function. The @var{uid} is shown in the
6041 function header of a dump file, and the pass names can be dumped by using
6042 option @option{-fdump-passes}.
6044 @item -fdisable-tree-@var{pass}
6045 @itemx -fdisable-tree-@var{pass}=@var{range-list}
6046 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
6049 @item -fenable-ipa-@var{pass}
6050 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6051 statically invoked in the compiler multiple times, the pass name should be
6052 appended with a sequential number starting from 1.
6054 @item -fenable-rtl-@var{pass}
6055 @itemx -fenable-rtl-@var{pass}=@var{range-list}
6056 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
6057 description and examples.
6059 @item -fenable-tree-@var{pass}
6060 @itemx -fenable-tree-@var{pass}=@var{range-list}
6061 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
6062 of option arguments.
6066 Here are some examples showing uses of these options.
6070 # disable ccp1 for all functions
6072 # disable complete unroll for function whose cgraph node uid is 1
6073 -fenable-tree-cunroll=1
6074 # disable gcse2 for functions at the following ranges [1,1],
6075 # [300,400], and [400,1000]
6076 # disable gcse2 for functions foo and foo2
6077 -fdisable-rtl-gcse2=foo,foo2
6078 # disable early inlining
6079 -fdisable-tree-einline
6080 # disable ipa inlining
6081 -fdisable-ipa-inline
6082 # enable tree full unroll
6083 -fenable-tree-unroll
6087 @item -d@var{letters}
6088 @itemx -fdump-rtl-@var{pass}
6089 @itemx -fdump-rtl-@var{pass}=@var{filename}
6091 @opindex fdump-rtl-@var{pass}
6092 Says to make debugging dumps during compilation at times specified by
6093 @var{letters}. This is used for debugging the RTL-based passes of the
6094 compiler. The file names for most of the dumps are made by appending
6095 a pass number and a word to the @var{dumpname}, and the files are
6096 created in the directory of the output file. In case of
6097 @option{=@var{filename}} option, the dump is output on the given file
6098 instead of the pass numbered dump files. Note that the pass number is
6099 computed statically as passes get registered into the pass manager.
6100 Thus the numbering is not related to the dynamic order of execution of
6101 passes. In particular, a pass installed by a plugin could have a
6102 number over 200 even if it executed quite early. @var{dumpname} is
6103 generated from the name of the output file, if explicitly specified
6104 and it is not an executable, otherwise it is the basename of the
6105 source file. These switches may have different effects when
6106 @option{-E} is used for preprocessing.
6108 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
6109 @option{-d} option @var{letters}. Here are the possible
6110 letters for use in @var{pass} and @var{letters}, and their meanings:
6114 @item -fdump-rtl-alignments
6115 @opindex fdump-rtl-alignments
6116 Dump after branch alignments have been computed.
6118 @item -fdump-rtl-asmcons
6119 @opindex fdump-rtl-asmcons
6120 Dump after fixing rtl statements that have unsatisfied in/out constraints.
6122 @item -fdump-rtl-auto_inc_dec
6123 @opindex fdump-rtl-auto_inc_dec
6124 Dump after auto-inc-dec discovery. This pass is only run on
6125 architectures that have auto inc or auto dec instructions.
6127 @item -fdump-rtl-barriers
6128 @opindex fdump-rtl-barriers
6129 Dump after cleaning up the barrier instructions.
6131 @item -fdump-rtl-bbpart
6132 @opindex fdump-rtl-bbpart
6133 Dump after partitioning hot and cold basic blocks.
6135 @item -fdump-rtl-bbro
6136 @opindex fdump-rtl-bbro
6137 Dump after block reordering.
6139 @item -fdump-rtl-btl1
6140 @itemx -fdump-rtl-btl2
6141 @opindex fdump-rtl-btl2
6142 @opindex fdump-rtl-btl2
6143 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
6144 after the two branch
6145 target load optimization passes.
6147 @item -fdump-rtl-bypass
6148 @opindex fdump-rtl-bypass
6149 Dump after jump bypassing and control flow optimizations.
6151 @item -fdump-rtl-combine
6152 @opindex fdump-rtl-combine
6153 Dump after the RTL instruction combination pass.
6155 @item -fdump-rtl-compgotos
6156 @opindex fdump-rtl-compgotos
6157 Dump after duplicating the computed gotos.
6159 @item -fdump-rtl-ce1
6160 @itemx -fdump-rtl-ce2
6161 @itemx -fdump-rtl-ce3
6162 @opindex fdump-rtl-ce1
6163 @opindex fdump-rtl-ce2
6164 @opindex fdump-rtl-ce3
6165 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
6166 @option{-fdump-rtl-ce3} enable dumping after the three
6167 if conversion passes.
6169 @item -fdump-rtl-cprop_hardreg
6170 @opindex fdump-rtl-cprop_hardreg
6171 Dump after hard register copy propagation.
6173 @item -fdump-rtl-csa
6174 @opindex fdump-rtl-csa
6175 Dump after combining stack adjustments.
6177 @item -fdump-rtl-cse1
6178 @itemx -fdump-rtl-cse2
6179 @opindex fdump-rtl-cse1
6180 @opindex fdump-rtl-cse2
6181 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
6182 the two common subexpression elimination passes.
6184 @item -fdump-rtl-dce
6185 @opindex fdump-rtl-dce
6186 Dump after the standalone dead code elimination passes.
6188 @item -fdump-rtl-dbr
6189 @opindex fdump-rtl-dbr
6190 Dump after delayed branch scheduling.
6192 @item -fdump-rtl-dce1
6193 @itemx -fdump-rtl-dce2
6194 @opindex fdump-rtl-dce1
6195 @opindex fdump-rtl-dce2
6196 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
6197 the two dead store elimination passes.
6200 @opindex fdump-rtl-eh
6201 Dump after finalization of EH handling code.
6203 @item -fdump-rtl-eh_ranges
6204 @opindex fdump-rtl-eh_ranges
6205 Dump after conversion of EH handling range regions.
6207 @item -fdump-rtl-expand
6208 @opindex fdump-rtl-expand
6209 Dump after RTL generation.
6211 @item -fdump-rtl-fwprop1
6212 @itemx -fdump-rtl-fwprop2
6213 @opindex fdump-rtl-fwprop1
6214 @opindex fdump-rtl-fwprop2
6215 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
6216 dumping after the two forward propagation passes.
6218 @item -fdump-rtl-gcse1
6219 @itemx -fdump-rtl-gcse2
6220 @opindex fdump-rtl-gcse1
6221 @opindex fdump-rtl-gcse2
6222 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
6223 after global common subexpression elimination.
6225 @item -fdump-rtl-init-regs
6226 @opindex fdump-rtl-init-regs
6227 Dump after the initialization of the registers.
6229 @item -fdump-rtl-initvals
6230 @opindex fdump-rtl-initvals
6231 Dump after the computation of the initial value sets.
6233 @item -fdump-rtl-into_cfglayout
6234 @opindex fdump-rtl-into_cfglayout
6235 Dump after converting to cfglayout mode.
6237 @item -fdump-rtl-ira
6238 @opindex fdump-rtl-ira
6239 Dump after iterated register allocation.
6241 @item -fdump-rtl-jump
6242 @opindex fdump-rtl-jump
6243 Dump after the second jump optimization.
6245 @item -fdump-rtl-loop2
6246 @opindex fdump-rtl-loop2
6247 @option{-fdump-rtl-loop2} enables dumping after the rtl
6248 loop optimization passes.
6250 @item -fdump-rtl-mach
6251 @opindex fdump-rtl-mach
6252 Dump after performing the machine dependent reorganization pass, if that
6255 @item -fdump-rtl-mode_sw
6256 @opindex fdump-rtl-mode_sw
6257 Dump after removing redundant mode switches.
6259 @item -fdump-rtl-rnreg
6260 @opindex fdump-rtl-rnreg
6261 Dump after register renumbering.
6263 @item -fdump-rtl-outof_cfglayout
6264 @opindex fdump-rtl-outof_cfglayout
6265 Dump after converting from cfglayout mode.
6267 @item -fdump-rtl-peephole2
6268 @opindex fdump-rtl-peephole2
6269 Dump after the peephole pass.
6271 @item -fdump-rtl-postreload
6272 @opindex fdump-rtl-postreload
6273 Dump after post-reload optimizations.
6275 @item -fdump-rtl-pro_and_epilogue
6276 @opindex fdump-rtl-pro_and_epilogue
6277 Dump after generating the function prologues and epilogues.
6279 @item -fdump-rtl-sched1
6280 @itemx -fdump-rtl-sched2
6281 @opindex fdump-rtl-sched1
6282 @opindex fdump-rtl-sched2
6283 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
6284 after the basic block scheduling passes.
6286 @item -fdump-rtl-ree
6287 @opindex fdump-rtl-ree
6288 Dump after sign/zero extension elimination.
6290 @item -fdump-rtl-seqabstr
6291 @opindex fdump-rtl-seqabstr
6292 Dump after common sequence discovery.
6294 @item -fdump-rtl-shorten
6295 @opindex fdump-rtl-shorten
6296 Dump after shortening branches.
6298 @item -fdump-rtl-sibling
6299 @opindex fdump-rtl-sibling
6300 Dump after sibling call optimizations.
6302 @item -fdump-rtl-split1
6303 @itemx -fdump-rtl-split2
6304 @itemx -fdump-rtl-split3
6305 @itemx -fdump-rtl-split4
6306 @itemx -fdump-rtl-split5
6307 @opindex fdump-rtl-split1
6308 @opindex fdump-rtl-split2
6309 @opindex fdump-rtl-split3
6310 @opindex fdump-rtl-split4
6311 @opindex fdump-rtl-split5
6312 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
6313 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
6314 @option{-fdump-rtl-split5} enable dumping after five rounds of
6315 instruction splitting.
6317 @item -fdump-rtl-sms
6318 @opindex fdump-rtl-sms
6319 Dump after modulo scheduling. This pass is only run on some
6322 @item -fdump-rtl-stack
6323 @opindex fdump-rtl-stack
6324 Dump after conversion from GCC's ``flat register file'' registers to the
6325 x87's stack-like registers. This pass is only run on x86 variants.
6327 @item -fdump-rtl-subreg1
6328 @itemx -fdump-rtl-subreg2
6329 @opindex fdump-rtl-subreg1
6330 @opindex fdump-rtl-subreg2
6331 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
6332 the two subreg expansion passes.
6334 @item -fdump-rtl-unshare
6335 @opindex fdump-rtl-unshare
6336 Dump after all rtl has been unshared.
6338 @item -fdump-rtl-vartrack
6339 @opindex fdump-rtl-vartrack
6340 Dump after variable tracking.
6342 @item -fdump-rtl-vregs
6343 @opindex fdump-rtl-vregs
6344 Dump after converting virtual registers to hard registers.
6346 @item -fdump-rtl-web
6347 @opindex fdump-rtl-web
6348 Dump after live range splitting.
6350 @item -fdump-rtl-regclass
6351 @itemx -fdump-rtl-subregs_of_mode_init
6352 @itemx -fdump-rtl-subregs_of_mode_finish
6353 @itemx -fdump-rtl-dfinit
6354 @itemx -fdump-rtl-dfinish
6355 @opindex fdump-rtl-regclass
6356 @opindex fdump-rtl-subregs_of_mode_init
6357 @opindex fdump-rtl-subregs_of_mode_finish
6358 @opindex fdump-rtl-dfinit
6359 @opindex fdump-rtl-dfinish
6360 These dumps are defined but always produce empty files.
6363 @itemx -fdump-rtl-all
6365 @opindex fdump-rtl-all
6366 Produce all the dumps listed above.
6370 Annotate the assembler output with miscellaneous debugging information.
6374 Dump all macro definitions, at the end of preprocessing, in addition to
6379 Produce a core dump whenever an error occurs.
6383 Annotate the assembler output with a comment indicating which
6384 pattern and alternative is used. The length of each instruction is
6389 Dump the RTL in the assembler output as a comment before each instruction.
6390 Also turns on @option{-dp} annotation.
6394 Just generate RTL for a function instead of compiling it. Usually used
6395 with @option{-fdump-rtl-expand}.
6399 @opindex fdump-noaddr
6400 When doing debugging dumps, suppress address output. This makes it more
6401 feasible to use diff on debugging dumps for compiler invocations with
6402 different compiler binaries and/or different
6403 text / bss / data / heap / stack / dso start locations.
6406 @opindex freport-bug
6407 Collect and dump debug information into temporary file if ICE in C/C++
6410 @item -fdump-unnumbered
6411 @opindex fdump-unnumbered
6412 When doing debugging dumps, suppress instruction numbers and address output.
6413 This makes it more feasible to use diff on debugging dumps for compiler
6414 invocations with different options, in particular with and without
6417 @item -fdump-unnumbered-links
6418 @opindex fdump-unnumbered-links
6419 When doing debugging dumps (see @option{-d} option above), suppress
6420 instruction numbers for the links to the previous and next instructions
6423 @item -fdump-translation-unit @r{(C++ only)}
6424 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
6425 @opindex fdump-translation-unit
6426 Dump a representation of the tree structure for the entire translation
6427 unit to a file. The file name is made by appending @file{.tu} to the
6428 source file name, and the file is created in the same directory as the
6429 output file. If the @samp{-@var{options}} form is used, @var{options}
6430 controls the details of the dump as described for the
6431 @option{-fdump-tree} options.
6433 @item -fdump-class-hierarchy @r{(C++ only)}
6434 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
6435 @opindex fdump-class-hierarchy
6436 Dump a representation of each class's hierarchy and virtual function
6437 table layout to a file. The file name is made by appending
6438 @file{.class} to the source file name, and the file is created in the
6439 same directory as the output file. If the @samp{-@var{options}} form
6440 is used, @var{options} controls the details of the dump as described
6441 for the @option{-fdump-tree} options.
6443 @item -fdump-ipa-@var{switch}
6445 Control the dumping at various stages of inter-procedural analysis
6446 language tree to a file. The file name is generated by appending a
6447 switch specific suffix to the source file name, and the file is created
6448 in the same directory as the output file. The following dumps are
6453 Enables all inter-procedural analysis dumps.
6456 Dumps information about call-graph optimization, unused function removal,
6457 and inlining decisions.
6460 Dump after function inlining.
6465 @opindex fdump-passes
6466 Dump the list of optimization passes that are turned on and off by
6467 the current command-line options.
6469 @item -fdump-statistics-@var{option}
6470 @opindex fdump-statistics
6471 Enable and control dumping of pass statistics in a separate file. The
6472 file name is generated by appending a suffix ending in
6473 @samp{.statistics} to the source file name, and the file is created in
6474 the same directory as the output file. If the @samp{-@var{option}}
6475 form is used, @samp{-stats} causes counters to be summed over the
6476 whole compilation unit while @samp{-details} dumps every event as
6477 the passes generate them. The default with no option is to sum
6478 counters for each function compiled.
6480 @item -fdump-tree-@var{switch}
6481 @itemx -fdump-tree-@var{switch}-@var{options}
6482 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
6484 Control the dumping at various stages of processing the intermediate
6485 language tree to a file. The file name is generated by appending a
6486 switch-specific suffix to the source file name, and the file is
6487 created in the same directory as the output file. In case of
6488 @option{=@var{filename}} option, the dump is output on the given file
6489 instead of the auto named dump files. If the @samp{-@var{options}}
6490 form is used, @var{options} is a list of @samp{-} separated options
6491 which control the details of the dump. Not all options are applicable
6492 to all dumps; those that are not meaningful are ignored. The
6493 following options are available
6497 Print the address of each node. Usually this is not meaningful as it
6498 changes according to the environment and source file. Its primary use
6499 is for tying up a dump file with a debug environment.
6501 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
6502 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
6503 use working backward from mangled names in the assembly file.
6505 When dumping front-end intermediate representations, inhibit dumping
6506 of members of a scope or body of a function merely because that scope
6507 has been reached. Only dump such items when they are directly reachable
6510 When dumping pretty-printed trees, this option inhibits dumping the
6511 bodies of control structures.
6513 When dumping RTL, print the RTL in slim (condensed) form instead of
6514 the default LISP-like representation.
6516 Print a raw representation of the tree. By default, trees are
6517 pretty-printed into a C-like representation.
6519 Enable more detailed dumps (not honored by every dump option). Also
6520 include information from the optimization passes.
6522 Enable dumping various statistics about the pass (not honored by every dump
6525 Enable showing basic block boundaries (disabled in raw dumps).
6527 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
6528 dump a representation of the control flow graph suitable for viewing with
6529 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
6530 the file is pretty-printed as a subgraph, so that GraphViz can render them
6531 all in a single plot.
6533 This option currently only works for RTL dumps, and the RTL is always
6534 dumped in slim form.
6536 Enable showing virtual operands for every statement.
6538 Enable showing line numbers for statements.
6540 Enable showing the unique ID (@code{DECL_UID}) for each variable.
6542 Enable showing the tree dump for each statement.
6544 Enable showing the EH region number holding each statement.
6546 Enable showing scalar evolution analysis details.
6548 Enable showing optimization information (only available in certain
6551 Enable showing missed optimization information (only available in certain
6554 Enable other detailed optimization information (only available in
6556 @item =@var{filename}
6557 Instead of an auto named dump file, output into the given file
6558 name. The file names @file{stdout} and @file{stderr} are treated
6559 specially and are considered already open standard streams. For
6563 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
6564 -fdump-tree-pre=stderr file.c
6567 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
6568 output on to @file{stderr}. If two conflicting dump filenames are
6569 given for the same pass, then the latter option overrides the earlier
6573 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
6574 and @option{lineno}.
6577 Turn on all optimization options, i.e., @option{optimized},
6578 @option{missed}, and @option{note}.
6581 The following tree dumps are possible:
6585 @opindex fdump-tree-original
6586 Dump before any tree based optimization, to @file{@var{file}.original}.
6589 @opindex fdump-tree-optimized
6590 Dump after all tree based optimization, to @file{@var{file}.optimized}.
6593 @opindex fdump-tree-gimple
6594 Dump each function before and after the gimplification pass to a file. The
6595 file name is made by appending @file{.gimple} to the source file name.
6598 @opindex fdump-tree-cfg
6599 Dump the control flow graph of each function to a file. The file name is
6600 made by appending @file{.cfg} to the source file name.
6603 @opindex fdump-tree-ch
6604 Dump each function after copying loop headers. The file name is made by
6605 appending @file{.ch} to the source file name.
6608 @opindex fdump-tree-ssa
6609 Dump SSA related information to a file. The file name is made by appending
6610 @file{.ssa} to the source file name.
6613 @opindex fdump-tree-alias
6614 Dump aliasing information for each function. The file name is made by
6615 appending @file{.alias} to the source file name.
6618 @opindex fdump-tree-ccp
6619 Dump each function after CCP@. The file name is made by appending
6620 @file{.ccp} to the source file name.
6623 @opindex fdump-tree-storeccp
6624 Dump each function after STORE-CCP@. The file name is made by appending
6625 @file{.storeccp} to the source file name.
6628 @opindex fdump-tree-pre
6629 Dump trees after partial redundancy elimination. The file name is made
6630 by appending @file{.pre} to the source file name.
6633 @opindex fdump-tree-fre
6634 Dump trees after full redundancy elimination. The file name is made
6635 by appending @file{.fre} to the source file name.
6638 @opindex fdump-tree-copyprop
6639 Dump trees after copy propagation. The file name is made
6640 by appending @file{.copyprop} to the source file name.
6642 @item store_copyprop
6643 @opindex fdump-tree-store_copyprop
6644 Dump trees after store copy-propagation. The file name is made
6645 by appending @file{.store_copyprop} to the source file name.
6648 @opindex fdump-tree-dce
6649 Dump each function after dead code elimination. The file name is made by
6650 appending @file{.dce} to the source file name.
6653 @opindex fdump-tree-sra
6654 Dump each function after performing scalar replacement of aggregates. The
6655 file name is made by appending @file{.sra} to the source file name.
6658 @opindex fdump-tree-sink
6659 Dump each function after performing code sinking. The file name is made
6660 by appending @file{.sink} to the source file name.
6663 @opindex fdump-tree-dom
6664 Dump each function after applying dominator tree optimizations. The file
6665 name is made by appending @file{.dom} to the source file name.
6668 @opindex fdump-tree-dse
6669 Dump each function after applying dead store elimination. The file
6670 name is made by appending @file{.dse} to the source file name.
6673 @opindex fdump-tree-phiopt
6674 Dump each function after optimizing PHI nodes into straightline code. The file
6675 name is made by appending @file{.phiopt} to the source file name.
6678 @opindex fdump-tree-forwprop
6679 Dump each function after forward propagating single use variables. The file
6680 name is made by appending @file{.forwprop} to the source file name.
6683 @opindex fdump-tree-copyrename
6684 Dump each function after applying the copy rename optimization. The file
6685 name is made by appending @file{.copyrename} to the source file name.
6688 @opindex fdump-tree-nrv
6689 Dump each function after applying the named return value optimization on
6690 generic trees. The file name is made by appending @file{.nrv} to the source
6694 @opindex fdump-tree-vect
6695 Dump each function after applying vectorization of loops. The file name is
6696 made by appending @file{.vect} to the source file name.
6699 @opindex fdump-tree-slp
6700 Dump each function after applying vectorization of basic blocks. The file name
6701 is made by appending @file{.slp} to the source file name.
6704 @opindex fdump-tree-vrp
6705 Dump each function after Value Range Propagation (VRP). The file name
6706 is made by appending @file{.vrp} to the source file name.
6709 @opindex fdump-tree-all
6710 Enable all the available tree dumps with the flags provided in this option.
6714 @itemx -fopt-info-@var{options}
6715 @itemx -fopt-info-@var{options}=@var{filename}
6717 Controls optimization dumps from various optimization passes. If the
6718 @samp{-@var{options}} form is used, @var{options} is a list of
6719 @samp{-} separated options to select the dump details and
6720 optimizations. If @var{options} is not specified, it defaults to
6721 @option{optimized} for details and @option{optall} for optimization
6722 groups. If the @var{filename} is not specified, it defaults to
6723 @file{stderr}. Note that the output @var{filename} will be overwritten
6724 in case of multiple translation units. If a combined output from
6725 multiple translation units is desired, @file{stderr} should be used
6728 The options can be divided into two groups, 1) options describing the
6729 verbosity of the dump, and 2) options describing which optimizations
6730 should be included. The options from both the groups can be freely
6731 mixed as they are non-overlapping. However, in case of any conflicts,
6732 the latter options override the earlier options on the command
6733 line. Though multiple -fopt-info options are accepted, only one of
6734 them can have @option{=filename}. If other filenames are provided then
6735 all but the first one are ignored.
6737 The dump verbosity has the following options
6741 Print information when an optimization is successfully applied. It is
6742 up to a pass to decide which information is relevant. For example, the
6743 vectorizer passes print the source location of loops which got
6744 successfully vectorized.
6746 Print information about missed optimizations. Individual passes
6747 control which information to include in the output. For example,
6750 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
6753 will print information about missed optimization opportunities from
6754 vectorization passes on stderr.
6756 Print verbose information about optimizations, such as certain
6757 transformations, more detailed messages about decisions etc.
6759 Print detailed optimization information. This includes
6760 @var{optimized}, @var{missed}, and @var{note}.
6763 The second set of options describes a group of optimizations and may
6764 include one or more of the following.
6768 Enable dumps from all interprocedural optimizations.
6770 Enable dumps from all loop optimizations.
6772 Enable dumps from all inlining optimizations.
6774 Enable dumps from all vectorization optimizations.
6776 Enable dumps from all optimizations. This is a superset of
6777 the optimization groups listed above.
6782 gcc -O3 -fopt-info-missed=missed.all
6785 outputs missed optimization report from all the passes into
6790 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
6793 will output information about missed optimizations as well as
6794 optimized locations from all the inlining passes into
6797 If the @var{filename} is provided, then the dumps from all the
6798 applicable optimizations are concatenated into the @file{filename}.
6799 Otherwise the dump is output onto @file{stderr}. If @var{options} is
6800 omitted, it defaults to @option{all-optall}, which means dump all
6801 available optimization info from all the passes. In the following
6802 example, all optimization info is output on to @file{stderr}.
6808 Note that @option{-fopt-info-vec-missed} behaves the same as
6809 @option{-fopt-info-missed-vec}.
6811 As another example, consider
6814 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
6817 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
6818 in conflict since only one output file is allowed. In this case, only
6819 the first option takes effect and the subsequent options are
6820 ignored. Thus only the @file{vec.miss} is produced which contains
6821 dumps from the vectorizer about missed opportunities.
6823 @item -frandom-seed=@var{number}
6824 @opindex frandom-seed
6825 This option provides a seed that GCC uses in place of
6826 random numbers in generating certain symbol names
6827 that have to be different in every compiled file. It is also used to
6828 place unique stamps in coverage data files and the object files that
6829 produce them. You can use the @option{-frandom-seed} option to produce
6830 reproducibly identical object files.
6832 The @var{number} should be different for every file you compile.
6834 @item -fsched-verbose=@var{n}
6835 @opindex fsched-verbose
6836 On targets that use instruction scheduling, this option controls the
6837 amount of debugging output the scheduler prints. This information is
6838 written to standard error, unless @option{-fdump-rtl-sched1} or
6839 @option{-fdump-rtl-sched2} is specified, in which case it is output
6840 to the usual dump listing file, @file{.sched1} or @file{.sched2}
6841 respectively. However for @var{n} greater than nine, the output is
6842 always printed to standard error.
6844 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
6845 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
6846 For @var{n} greater than one, it also output basic block probabilities,
6847 detailed ready list information and unit/insn info. For @var{n} greater
6848 than two, it includes RTL at abort point, control-flow and regions info.
6849 And for @var{n} over four, @option{-fsched-verbose} also includes
6853 @itemx -save-temps=cwd
6855 Store the usual ``temporary'' intermediate files permanently; place them
6856 in the current directory and name them based on the source file. Thus,
6857 compiling @file{foo.c} with @option{-c -save-temps} produces files
6858 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
6859 preprocessed @file{foo.i} output file even though the compiler now
6860 normally uses an integrated preprocessor.
6862 When used in combination with the @option{-x} command-line option,
6863 @option{-save-temps} is sensible enough to avoid over writing an
6864 input source file with the same extension as an intermediate file.
6865 The corresponding intermediate file may be obtained by renaming the
6866 source file before using @option{-save-temps}.
6868 If you invoke GCC in parallel, compiling several different source
6869 files that share a common base name in different subdirectories or the
6870 same source file compiled for multiple output destinations, it is
6871 likely that the different parallel compilers will interfere with each
6872 other, and overwrite the temporary files. For instance:
6875 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
6876 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
6879 may result in @file{foo.i} and @file{foo.o} being written to
6880 simultaneously by both compilers.
6882 @item -save-temps=obj
6883 @opindex save-temps=obj
6884 Store the usual ``temporary'' intermediate files permanently. If the
6885 @option{-o} option is used, the temporary files are based on the
6886 object file. If the @option{-o} option is not used, the
6887 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
6892 gcc -save-temps=obj -c foo.c
6893 gcc -save-temps=obj -c bar.c -o dir/xbar.o
6894 gcc -save-temps=obj foobar.c -o dir2/yfoobar
6898 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
6899 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
6900 @file{dir2/yfoobar.o}.
6902 @item -time@r{[}=@var{file}@r{]}
6904 Report the CPU time taken by each subprocess in the compilation
6905 sequence. For C source files, this is the compiler proper and assembler
6906 (plus the linker if linking is done).
6908 Without the specification of an output file, the output looks like this:
6915 The first number on each line is the ``user time'', that is time spent
6916 executing the program itself. The second number is ``system time'',
6917 time spent executing operating system routines on behalf of the program.
6918 Both numbers are in seconds.
6920 With the specification of an output file, the output is appended to the
6921 named file, and it looks like this:
6924 0.12 0.01 cc1 @var{options}
6925 0.00 0.01 as @var{options}
6928 The ``user time'' and the ``system time'' are moved before the program
6929 name, and the options passed to the program are displayed, so that one
6930 can later tell what file was being compiled, and with which options.
6932 @item -fvar-tracking
6933 @opindex fvar-tracking
6934 Run variable tracking pass. It computes where variables are stored at each
6935 position in code. Better debugging information is then generated
6936 (if the debugging information format supports this information).
6938 It is enabled by default when compiling with optimization (@option{-Os},
6939 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
6940 the debug info format supports it.
6942 @item -fvar-tracking-assignments
6943 @opindex fvar-tracking-assignments
6944 @opindex fno-var-tracking-assignments
6945 Annotate assignments to user variables early in the compilation and
6946 attempt to carry the annotations over throughout the compilation all the
6947 way to the end, in an attempt to improve debug information while
6948 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
6950 It can be enabled even if var-tracking is disabled, in which case
6951 annotations are created and maintained, but discarded at the end.
6953 @item -fvar-tracking-assignments-toggle
6954 @opindex fvar-tracking-assignments-toggle
6955 @opindex fno-var-tracking-assignments-toggle
6956 Toggle @option{-fvar-tracking-assignments}, in the same way that
6957 @option{-gtoggle} toggles @option{-g}.
6959 @item -print-file-name=@var{library}
6960 @opindex print-file-name
6961 Print the full absolute name of the library file @var{library} that
6962 would be used when linking---and don't do anything else. With this
6963 option, GCC does not compile or link anything; it just prints the
6966 @item -print-multi-directory
6967 @opindex print-multi-directory
6968 Print the directory name corresponding to the multilib selected by any
6969 other switches present in the command line. This directory is supposed
6970 to exist in @env{GCC_EXEC_PREFIX}.
6972 @item -print-multi-lib
6973 @opindex print-multi-lib
6974 Print the mapping from multilib directory names to compiler switches
6975 that enable them. The directory name is separated from the switches by
6976 @samp{;}, and each switch starts with an @samp{@@} instead of the
6977 @samp{-}, without spaces between multiple switches. This is supposed to
6978 ease shell processing.
6980 @item -print-multi-os-directory
6981 @opindex print-multi-os-directory
6982 Print the path to OS libraries for the selected
6983 multilib, relative to some @file{lib} subdirectory. If OS libraries are
6984 present in the @file{lib} subdirectory and no multilibs are used, this is
6985 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
6986 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
6987 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
6988 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
6990 @item -print-multiarch
6991 @opindex print-multiarch
6992 Print the path to OS libraries for the selected multiarch,
6993 relative to some @file{lib} subdirectory.
6995 @item -print-prog-name=@var{program}
6996 @opindex print-prog-name
6997 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
6999 @item -print-libgcc-file-name
7000 @opindex print-libgcc-file-name
7001 Same as @option{-print-file-name=libgcc.a}.
7003 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
7004 but you do want to link with @file{libgcc.a}. You can do:
7007 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
7010 @item -print-search-dirs
7011 @opindex print-search-dirs
7012 Print the name of the configured installation directory and a list of
7013 program and library directories @command{gcc} searches---and don't do anything else.
7015 This is useful when @command{gcc} prints the error message
7016 @samp{installation problem, cannot exec cpp0: No such file or directory}.
7017 To resolve this you either need to put @file{cpp0} and the other compiler
7018 components where @command{gcc} expects to find them, or you can set the environment
7019 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
7020 Don't forget the trailing @samp{/}.
7021 @xref{Environment Variables}.
7023 @item -print-sysroot
7024 @opindex print-sysroot
7025 Print the target sysroot directory that is used during
7026 compilation. This is the target sysroot specified either at configure
7027 time or using the @option{--sysroot} option, possibly with an extra
7028 suffix that depends on compilation options. If no target sysroot is
7029 specified, the option prints nothing.
7031 @item -print-sysroot-headers-suffix
7032 @opindex print-sysroot-headers-suffix
7033 Print the suffix added to the target sysroot when searching for
7034 headers, or give an error if the compiler is not configured with such
7035 a suffix---and don't do anything else.
7038 @opindex dumpmachine
7039 Print the compiler's target machine (for example,
7040 @samp{i686-pc-linux-gnu})---and don't do anything else.
7043 @opindex dumpversion
7044 Print the compiler version (for example, @samp{3.0})---and don't do
7049 Print the compiler's built-in specs---and don't do anything else. (This
7050 is used when GCC itself is being built.) @xref{Spec Files}.
7052 @item -fno-eliminate-unused-debug-types
7053 @opindex feliminate-unused-debug-types
7054 @opindex fno-eliminate-unused-debug-types
7055 Normally, when producing DWARF 2 output, GCC avoids producing debug symbol
7056 output for types that are nowhere used in the source file being compiled.
7057 Sometimes it is useful to have GCC emit debugging
7058 information for all types declared in a compilation
7059 unit, regardless of whether or not they are actually used
7060 in that compilation unit, for example
7061 if, in the debugger, you want to cast a value to a type that is
7062 not actually used in your program (but is declared). More often,
7063 however, this results in a significant amount of wasted space.
7066 @node Optimize Options
7067 @section Options That Control Optimization
7068 @cindex optimize options
7069 @cindex options, optimization
7071 These options control various sorts of optimizations.
7073 Without any optimization option, the compiler's goal is to reduce the
7074 cost of compilation and to make debugging produce the expected
7075 results. Statements are independent: if you stop the program with a
7076 breakpoint between statements, you can then assign a new value to any
7077 variable or change the program counter to any other statement in the
7078 function and get exactly the results you expect from the source
7081 Turning on optimization flags makes the compiler attempt to improve
7082 the performance and/or code size at the expense of compilation time
7083 and possibly the ability to debug the program.
7085 The compiler performs optimization based on the knowledge it has of the
7086 program. Compiling multiple files at once to a single output file mode allows
7087 the compiler to use information gained from all of the files when compiling
7090 Not all optimizations are controlled directly by a flag. Only
7091 optimizations that have a flag are listed in this section.
7093 Most optimizations are only enabled if an @option{-O} level is set on
7094 the command line. Otherwise they are disabled, even if individual
7095 optimization flags are specified.
7097 Depending on the target and how GCC was configured, a slightly different
7098 set of optimizations may be enabled at each @option{-O} level than
7099 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
7100 to find out the exact set of optimizations that are enabled at each level.
7101 @xref{Overall Options}, for examples.
7108 Optimize. Optimizing compilation takes somewhat more time, and a lot
7109 more memory for a large function.
7111 With @option{-O}, the compiler tries to reduce code size and execution
7112 time, without performing any optimizations that take a great deal of
7115 @option{-O} turns on the following optimization flags:
7118 -fbranch-count-reg @gol
7119 -fcombine-stack-adjustments @gol
7121 -fcprop-registers @gol
7124 -fdelayed-branch @gol
7126 -fforward-propagate @gol
7127 -fguess-branch-probability @gol
7128 -fif-conversion2 @gol
7129 -fif-conversion @gol
7130 -finline-functions-called-once @gol
7131 -fipa-pure-const @gol
7133 -fipa-reference @gol
7134 -fmerge-constants @gol
7135 -fmove-loop-invariants @gol
7137 -fsplit-wide-types @gol
7142 -ftree-copy-prop @gol
7143 -ftree-copyrename @gol
7145 -ftree-dominator-opts @gol
7147 -ftree-forwprop @gol
7157 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
7158 where doing so does not interfere with debugging.
7162 Optimize even more. GCC performs nearly all supported optimizations
7163 that do not involve a space-speed tradeoff.
7164 As compared to @option{-O}, this option increases both compilation time
7165 and the performance of the generated code.
7167 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7168 also turns on the following optimization flags:
7169 @gccoptlist{-fthread-jumps @gol
7170 -falign-functions -falign-jumps @gol
7171 -falign-loops -falign-labels @gol
7174 -fcse-follow-jumps -fcse-skip-blocks @gol
7175 -fdelete-null-pointer-checks @gol
7176 -fdevirtualize -fdevirtualize-speculatively @gol
7177 -fexpensive-optimizations @gol
7178 -fgcse -fgcse-lm @gol
7179 -fhoist-adjacent-loads @gol
7180 -finline-small-functions @gol
7181 -findirect-inlining @gol
7185 -fisolate-erroneous-paths-dereference @gol
7186 -foptimize-sibling-calls @gol
7187 -foptimize-strlen @gol
7188 -fpartial-inlining @gol
7190 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
7191 -frerun-cse-after-loop @gol
7192 -fsched-interblock -fsched-spec @gol
7193 -fschedule-insns -fschedule-insns2 @gol
7194 -fstrict-aliasing -fstrict-overflow @gol
7195 -ftree-builtin-call-dce @gol
7196 -ftree-switch-conversion -ftree-tail-merge @gol
7201 Please note the warning under @option{-fgcse} about
7202 invoking @option{-O2} on programs that use computed gotos.
7206 Optimize yet more. @option{-O3} turns on all optimizations specified
7207 by @option{-O2} and also turns on the @option{-finline-functions},
7208 @option{-funswitch-loops}, @option{-fpredictive-commoning},
7209 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
7210 @option{-ftree-loop-distribute-patterns},
7211 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
7212 @option{-ftree-partial-pre} and @option{-fipa-cp-clone} options.
7216 Reduce compilation time and make debugging produce the expected
7217 results. This is the default.
7221 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7222 do not typically increase code size. It also performs further
7223 optimizations designed to reduce code size.
7225 @option{-Os} disables the following optimization flags:
7226 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7227 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
7228 -fprefetch-loop-arrays}
7232 Disregard strict standards compliance. @option{-Ofast} enables all
7233 @option{-O3} optimizations. It also enables optimizations that are not
7234 valid for all standard-compliant programs.
7235 It turns on @option{-ffast-math} and the Fortran-specific
7236 @option{-fno-protect-parens} and @option{-fstack-arrays}.
7240 Optimize debugging experience. @option{-Og} enables optimizations
7241 that do not interfere with debugging. It should be the optimization
7242 level of choice for the standard edit-compile-debug cycle, offering
7243 a reasonable level of optimization while maintaining fast compilation
7244 and a good debugging experience.
7246 If you use multiple @option{-O} options, with or without level numbers,
7247 the last such option is the one that is effective.
7250 Options of the form @option{-f@var{flag}} specify machine-independent
7251 flags. Most flags have both positive and negative forms; the negative
7252 form of @option{-ffoo} is @option{-fno-foo}. In the table
7253 below, only one of the forms is listed---the one you typically
7254 use. You can figure out the other form by either removing @samp{no-}
7257 The following options control specific optimizations. They are either
7258 activated by @option{-O} options or are related to ones that are. You
7259 can use the following flags in the rare cases when ``fine-tuning'' of
7260 optimizations to be performed is desired.
7263 @item -fno-defer-pop
7264 @opindex fno-defer-pop
7265 Always pop the arguments to each function call as soon as that function
7266 returns. For machines that must pop arguments after a function call,
7267 the compiler normally lets arguments accumulate on the stack for several
7268 function calls and pops them all at once.
7270 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7272 @item -fforward-propagate
7273 @opindex fforward-propagate
7274 Perform a forward propagation pass on RTL@. The pass tries to combine two
7275 instructions and checks if the result can be simplified. If loop unrolling
7276 is active, two passes are performed and the second is scheduled after
7279 This option is enabled by default at optimization levels @option{-O},
7280 @option{-O2}, @option{-O3}, @option{-Os}.
7282 @item -ffp-contract=@var{style}
7283 @opindex ffp-contract
7284 @option{-ffp-contract=off} disables floating-point expression contraction.
7285 @option{-ffp-contract=fast} enables floating-point expression contraction
7286 such as forming of fused multiply-add operations if the target has
7287 native support for them.
7288 @option{-ffp-contract=on} enables floating-point expression contraction
7289 if allowed by the language standard. This is currently not implemented
7290 and treated equal to @option{-ffp-contract=off}.
7292 The default is @option{-ffp-contract=fast}.
7294 @item -fomit-frame-pointer
7295 @opindex fomit-frame-pointer
7296 Don't keep the frame pointer in a register for functions that
7297 don't need one. This avoids the instructions to save, set up and
7298 restore frame pointers; it also makes an extra register available
7299 in many functions. @strong{It also makes debugging impossible on
7302 On some machines, such as the VAX, this flag has no effect, because
7303 the standard calling sequence automatically handles the frame pointer
7304 and nothing is saved by pretending it doesn't exist. The
7305 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
7306 whether a target machine supports this flag. @xref{Registers,,Register
7307 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
7309 Starting with GCC version 4.6, the default setting (when not optimizing for
7310 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets has been changed to
7311 @option{-fomit-frame-pointer}. The default can be reverted to
7312 @option{-fno-omit-frame-pointer} by configuring GCC with the
7313 @option{--enable-frame-pointer} configure option.
7315 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7317 @item -foptimize-sibling-calls
7318 @opindex foptimize-sibling-calls
7319 Optimize sibling and tail recursive calls.
7321 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7323 @item -foptimize-strlen
7324 @opindex foptimize-strlen
7325 Optimize various standard C string functions (e.g. @code{strlen},
7326 @code{strchr} or @code{strcpy}) and
7327 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7329 Enabled at levels @option{-O2}, @option{-O3}.
7333 Do not expand any functions inline apart from those marked with
7334 the @code{always_inline} attribute. This is the default when not
7337 Single functions can be exempted from inlining by marking them
7338 with the @code{noinline} attribute.
7340 @item -finline-small-functions
7341 @opindex finline-small-functions
7342 Integrate functions into their callers when their body is smaller than expected
7343 function call code (so overall size of program gets smaller). The compiler
7344 heuristically decides which functions are simple enough to be worth integrating
7345 in this way. This inlining applies to all functions, even those not declared
7348 Enabled at level @option{-O2}.
7350 @item -findirect-inlining
7351 @opindex findirect-inlining
7352 Inline also indirect calls that are discovered to be known at compile
7353 time thanks to previous inlining. This option has any effect only
7354 when inlining itself is turned on by the @option{-finline-functions}
7355 or @option{-finline-small-functions} options.
7357 Enabled at level @option{-O2}.
7359 @item -finline-functions
7360 @opindex finline-functions
7361 Consider all functions for inlining, even if they are not declared inline.
7362 The compiler heuristically decides which functions are worth integrating
7365 If all calls to a given function are integrated, and the function is
7366 declared @code{static}, then the function is normally not output as
7367 assembler code in its own right.
7369 Enabled at level @option{-O3}.
7371 @item -finline-functions-called-once
7372 @opindex finline-functions-called-once
7373 Consider all @code{static} functions called once for inlining into their
7374 caller even if they are not marked @code{inline}. If a call to a given
7375 function is integrated, then the function is not output as assembler code
7378 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7380 @item -fearly-inlining
7381 @opindex fearly-inlining
7382 Inline functions marked by @code{always_inline} and functions whose body seems
7383 smaller than the function call overhead early before doing
7384 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7385 makes profiling significantly cheaper and usually inlining faster on programs
7386 having large chains of nested wrapper functions.
7392 Perform interprocedural scalar replacement of aggregates, removal of
7393 unused parameters and replacement of parameters passed by reference
7394 by parameters passed by value.
7396 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7398 @item -finline-limit=@var{n}
7399 @opindex finline-limit
7400 By default, GCC limits the size of functions that can be inlined. This flag
7401 allows coarse control of this limit. @var{n} is the size of functions that
7402 can be inlined in number of pseudo instructions.
7404 Inlining is actually controlled by a number of parameters, which may be
7405 specified individually by using @option{--param @var{name}=@var{value}}.
7406 The @option{-finline-limit=@var{n}} option sets some of these parameters
7410 @item max-inline-insns-single
7411 is set to @var{n}/2.
7412 @item max-inline-insns-auto
7413 is set to @var{n}/2.
7416 See below for a documentation of the individual
7417 parameters controlling inlining and for the defaults of these parameters.
7419 @emph{Note:} there may be no value to @option{-finline-limit} that results
7420 in default behavior.
7422 @emph{Note:} pseudo instruction represents, in this particular context, an
7423 abstract measurement of function's size. In no way does it represent a count
7424 of assembly instructions and as such its exact meaning might change from one
7425 release to an another.
7427 @item -fno-keep-inline-dllexport
7428 @opindex -fno-keep-inline-dllexport
7429 This is a more fine-grained version of @option{-fkeep-inline-functions},
7430 which applies only to functions that are declared using the @code{dllexport}
7431 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
7434 @item -fkeep-inline-functions
7435 @opindex fkeep-inline-functions
7436 In C, emit @code{static} functions that are declared @code{inline}
7437 into the object file, even if the function has been inlined into all
7438 of its callers. This switch does not affect functions using the
7439 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7440 inline functions into the object file.
7442 @item -fkeep-static-consts
7443 @opindex fkeep-static-consts
7444 Emit variables declared @code{static const} when optimization isn't turned
7445 on, even if the variables aren't referenced.
7447 GCC enables this option by default. If you want to force the compiler to
7448 check if a variable is referenced, regardless of whether or not
7449 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7451 @item -fmerge-constants
7452 @opindex fmerge-constants
7453 Attempt to merge identical constants (string constants and floating-point
7454 constants) across compilation units.
7456 This option is the default for optimized compilation if the assembler and
7457 linker support it. Use @option{-fno-merge-constants} to inhibit this
7460 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7462 @item -fmerge-all-constants
7463 @opindex fmerge-all-constants
7464 Attempt to merge identical constants and identical variables.
7466 This option implies @option{-fmerge-constants}. In addition to
7467 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7468 arrays or initialized constant variables with integral or floating-point
7469 types. Languages like C or C++ require each variable, including multiple
7470 instances of the same variable in recursive calls, to have distinct locations,
7471 so using this option results in non-conforming
7474 @item -fmodulo-sched
7475 @opindex fmodulo-sched
7476 Perform swing modulo scheduling immediately before the first scheduling
7477 pass. This pass looks at innermost loops and reorders their
7478 instructions by overlapping different iterations.
7480 @item -fmodulo-sched-allow-regmoves
7481 @opindex fmodulo-sched-allow-regmoves
7482 Perform more aggressive SMS-based modulo scheduling with register moves
7483 allowed. By setting this flag certain anti-dependences edges are
7484 deleted, which triggers the generation of reg-moves based on the
7485 life-range analysis. This option is effective only with
7486 @option{-fmodulo-sched} enabled.
7488 @item -fno-branch-count-reg
7489 @opindex fno-branch-count-reg
7490 Do not use ``decrement and branch'' instructions on a count register,
7491 but instead generate a sequence of instructions that decrement a
7492 register, compare it against zero, then branch based upon the result.
7493 This option is only meaningful on architectures that support such
7494 instructions, which include x86, PowerPC, IA-64 and S/390.
7496 Enabled by default at @option{-O1} and higher.
7498 The default is @option{-fbranch-count-reg}.
7500 @item -fno-function-cse
7501 @opindex fno-function-cse
7502 Do not put function addresses in registers; make each instruction that
7503 calls a constant function contain the function's address explicitly.
7505 This option results in less efficient code, but some strange hacks
7506 that alter the assembler output may be confused by the optimizations
7507 performed when this option is not used.
7509 The default is @option{-ffunction-cse}
7511 @item -fno-zero-initialized-in-bss
7512 @opindex fno-zero-initialized-in-bss
7513 If the target supports a BSS section, GCC by default puts variables that
7514 are initialized to zero into BSS@. This can save space in the resulting
7517 This option turns off this behavior because some programs explicitly
7518 rely on variables going to the data section---e.g., so that the
7519 resulting executable can find the beginning of that section and/or make
7520 assumptions based on that.
7522 The default is @option{-fzero-initialized-in-bss}.
7524 @item -fthread-jumps
7525 @opindex fthread-jumps
7526 Perform optimizations that check to see if a jump branches to a
7527 location where another comparison subsumed by the first is found. If
7528 so, the first branch is redirected to either the destination of the
7529 second branch or a point immediately following it, depending on whether
7530 the condition is known to be true or false.
7532 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7534 @item -fsplit-wide-types
7535 @opindex fsplit-wide-types
7536 When using a type that occupies multiple registers, such as @code{long
7537 long} on a 32-bit system, split the registers apart and allocate them
7538 independently. This normally generates better code for those types,
7539 but may make debugging more difficult.
7541 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7544 @item -fcse-follow-jumps
7545 @opindex fcse-follow-jumps
7546 In common subexpression elimination (CSE), scan through jump instructions
7547 when the target of the jump is not reached by any other path. For
7548 example, when CSE encounters an @code{if} statement with an
7549 @code{else} clause, CSE follows the jump when the condition
7552 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7554 @item -fcse-skip-blocks
7555 @opindex fcse-skip-blocks
7556 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
7557 follow jumps that conditionally skip over blocks. When CSE
7558 encounters a simple @code{if} statement with no else clause,
7559 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
7560 body of the @code{if}.
7562 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7564 @item -frerun-cse-after-loop
7565 @opindex frerun-cse-after-loop
7566 Re-run common subexpression elimination after loop optimizations are
7569 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7573 Perform a global common subexpression elimination pass.
7574 This pass also performs global constant and copy propagation.
7576 @emph{Note:} When compiling a program using computed gotos, a GCC
7577 extension, you may get better run-time performance if you disable
7578 the global common subexpression elimination pass by adding
7579 @option{-fno-gcse} to the command line.
7581 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7585 When @option{-fgcse-lm} is enabled, global common subexpression elimination
7586 attempts to move loads that are only killed by stores into themselves. This
7587 allows a loop containing a load/store sequence to be changed to a load outside
7588 the loop, and a copy/store within the loop.
7590 Enabled by default when @option{-fgcse} is enabled.
7594 When @option{-fgcse-sm} is enabled, a store motion pass is run after
7595 global common subexpression elimination. This pass attempts to move
7596 stores out of loops. When used in conjunction with @option{-fgcse-lm},
7597 loops containing a load/store sequence can be changed to a load before
7598 the loop and a store after the loop.
7600 Not enabled at any optimization level.
7604 When @option{-fgcse-las} is enabled, the global common subexpression
7605 elimination pass eliminates redundant loads that come after stores to the
7606 same memory location (both partial and full redundancies).
7608 Not enabled at any optimization level.
7610 @item -fgcse-after-reload
7611 @opindex fgcse-after-reload
7612 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
7613 pass is performed after reload. The purpose of this pass is to clean up
7616 @item -faggressive-loop-optimizations
7617 @opindex faggressive-loop-optimizations
7618 This option tells the loop optimizer to use language constraints to
7619 derive bounds for the number of iterations of a loop. This assumes that
7620 loop code does not invoke undefined behavior by for example causing signed
7621 integer overflows or out-of-bound array accesses. The bounds for the
7622 number of iterations of a loop are used to guide loop unrolling and peeling
7623 and loop exit test optimizations.
7624 This option is enabled by default.
7626 @item -funsafe-loop-optimizations
7627 @opindex funsafe-loop-optimizations
7628 This option tells the loop optimizer to assume that loop indices do not
7629 overflow, and that loops with nontrivial exit condition are not
7630 infinite. This enables a wider range of loop optimizations even if
7631 the loop optimizer itself cannot prove that these assumptions are valid.
7632 If you use @option{-Wunsafe-loop-optimizations}, the compiler warns you
7633 if it finds this kind of loop.
7635 @item -fcrossjumping
7636 @opindex fcrossjumping
7637 Perform cross-jumping transformation.
7638 This transformation unifies equivalent code and saves code size. The
7639 resulting code may or may not perform better than without cross-jumping.
7641 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7643 @item -fauto-inc-dec
7644 @opindex fauto-inc-dec
7645 Combine increments or decrements of addresses with memory accesses.
7646 This pass is always skipped on architectures that do not have
7647 instructions to support this. Enabled by default at @option{-O} and
7648 higher on architectures that support this.
7652 Perform dead code elimination (DCE) on RTL@.
7653 Enabled by default at @option{-O} and higher.
7657 Perform dead store elimination (DSE) on RTL@.
7658 Enabled by default at @option{-O} and higher.
7660 @item -fif-conversion
7661 @opindex fif-conversion
7662 Attempt to transform conditional jumps into branch-less equivalents. This
7663 includes use of conditional moves, min, max, set flags and abs instructions, and
7664 some tricks doable by standard arithmetics. The use of conditional execution
7665 on chips where it is available is controlled by @code{if-conversion2}.
7667 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7669 @item -fif-conversion2
7670 @opindex fif-conversion2
7671 Use conditional execution (where available) to transform conditional jumps into
7672 branch-less equivalents.
7674 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7676 @item -fdeclone-ctor-dtor
7677 @opindex fdeclone-ctor-dtor
7678 The C++ ABI requires multiple entry points for constructors and
7679 destructors: one for a base subobject, one for a complete object, and
7680 one for a virtual destructor that calls operator delete afterwards.
7681 For a hierarchy with virtual bases, the base and complete variants are
7682 clones, which means two copies of the function. With this option, the
7683 base and complete variants are changed to be thunks that call a common
7686 Enabled by @option{-Os}.
7688 @item -fdelete-null-pointer-checks
7689 @opindex fdelete-null-pointer-checks
7690 Assume that programs cannot safely dereference null pointers, and that
7691 no code or data element resides there. This enables simple constant
7692 folding optimizations at all optimization levels. In addition, other
7693 optimization passes in GCC use this flag to control global dataflow
7694 analyses that eliminate useless checks for null pointers; these assume
7695 that if a pointer is checked after it has already been dereferenced,
7698 Note however that in some environments this assumption is not true.
7699 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
7700 for programs that depend on that behavior.
7702 Some targets, especially embedded ones, disable this option at all levels.
7703 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
7704 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
7705 are enabled independently at different optimization levels.
7707 @item -fdevirtualize
7708 @opindex fdevirtualize
7709 Attempt to convert calls to virtual functions to direct calls. This
7710 is done both within a procedure and interprocedurally as part of
7711 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
7712 propagation (@option{-fipa-cp}).
7713 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7715 @item -fdevirtualize-speculatively
7716 @opindex fdevirtualize-speculatively
7717 Attempt to convert calls to virtual functions to speculative direct calls.
7718 Based on the analysis of the type inheritance graph, determine for a given call
7719 the set of likely targets. If the set is small, preferably of size 1, change
7720 the call into an conditional deciding on direct and indirect call. The
7721 speculative calls enable more optimizations, such as inlining. When they seem
7722 useless after further optimization, they are converted back into original form.
7724 @item -fexpensive-optimizations
7725 @opindex fexpensive-optimizations
7726 Perform a number of minor optimizations that are relatively expensive.
7728 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7732 Attempt to remove redundant extension instructions. This is especially
7733 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
7734 registers after writing to their lower 32-bit half.
7736 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
7737 @option{-O3}, @option{-Os}.
7739 @item -flive-range-shrinkage
7740 @opindex flive-range-shrinkage
7741 Attempt to decrease register pressure through register live range
7742 shrinkage. This is helpful for fast processors with small or moderate
7745 @item -fira-algorithm=@var{algorithm}
7746 Use the specified coloring algorithm for the integrated register
7747 allocator. The @var{algorithm} argument can be @samp{priority}, which
7748 specifies Chow's priority coloring, or @samp{CB}, which specifies
7749 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
7750 for all architectures, but for those targets that do support it, it is
7751 the default because it generates better code.
7753 @item -fira-region=@var{region}
7754 Use specified regions for the integrated register allocator. The
7755 @var{region} argument should be one of the following:
7760 Use all loops as register allocation regions.
7761 This can give the best results for machines with a small and/or
7762 irregular register set.
7765 Use all loops except for loops with small register pressure
7766 as the regions. This value usually gives
7767 the best results in most cases and for most architectures,
7768 and is enabled by default when compiling with optimization for speed
7769 (@option{-O}, @option{-O2}, @dots{}).
7772 Use all functions as a single region.
7773 This typically results in the smallest code size, and is enabled by default for
7774 @option{-Os} or @option{-O0}.
7778 @item -fira-hoist-pressure
7779 @opindex fira-hoist-pressure
7780 Use IRA to evaluate register pressure in the code hoisting pass for
7781 decisions to hoist expressions. This option usually results in smaller
7782 code, but it can slow the compiler down.
7784 This option is enabled at level @option{-Os} for all targets.
7786 @item -fira-loop-pressure
7787 @opindex fira-loop-pressure
7788 Use IRA to evaluate register pressure in loops for decisions to move
7789 loop invariants. This option usually results in generation
7790 of faster and smaller code on machines with large register files (>= 32
7791 registers), but it can slow the compiler down.
7793 This option is enabled at level @option{-O3} for some targets.
7795 @item -fno-ira-share-save-slots
7796 @opindex fno-ira-share-save-slots
7797 Disable sharing of stack slots used for saving call-used hard
7798 registers living through a call. Each hard register gets a
7799 separate stack slot, and as a result function stack frames are
7802 @item -fno-ira-share-spill-slots
7803 @opindex fno-ira-share-spill-slots
7804 Disable sharing of stack slots allocated for pseudo-registers. Each
7805 pseudo-register that does not get a hard register gets a separate
7806 stack slot, and as a result function stack frames are larger.
7808 @item -fira-verbose=@var{n}
7809 @opindex fira-verbose
7810 Control the verbosity of the dump file for the integrated register allocator.
7811 The default value is 5. If the value @var{n} is greater or equal to 10,
7812 the dump output is sent to stderr using the same format as @var{n} minus 10.
7814 @item -fdelayed-branch
7815 @opindex fdelayed-branch
7816 If supported for the target machine, attempt to reorder instructions
7817 to exploit instruction slots available after delayed branch
7820 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7822 @item -fschedule-insns
7823 @opindex fschedule-insns
7824 If supported for the target machine, attempt to reorder instructions to
7825 eliminate execution stalls due to required data being unavailable. This
7826 helps machines that have slow floating point or memory load instructions
7827 by allowing other instructions to be issued until the result of the load
7828 or floating-point instruction is required.
7830 Enabled at levels @option{-O2}, @option{-O3}.
7832 @item -fschedule-insns2
7833 @opindex fschedule-insns2
7834 Similar to @option{-fschedule-insns}, but requests an additional pass of
7835 instruction scheduling after register allocation has been done. This is
7836 especially useful on machines with a relatively small number of
7837 registers and where memory load instructions take more than one cycle.
7839 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7841 @item -fno-sched-interblock
7842 @opindex fno-sched-interblock
7843 Don't schedule instructions across basic blocks. This is normally
7844 enabled by default when scheduling before register allocation, i.e.@:
7845 with @option{-fschedule-insns} or at @option{-O2} or higher.
7847 @item -fno-sched-spec
7848 @opindex fno-sched-spec
7849 Don't allow speculative motion of non-load instructions. This is normally
7850 enabled by default when scheduling before register allocation, i.e.@:
7851 with @option{-fschedule-insns} or at @option{-O2} or higher.
7853 @item -fsched-pressure
7854 @opindex fsched-pressure
7855 Enable register pressure sensitive insn scheduling before register
7856 allocation. This only makes sense when scheduling before register
7857 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
7858 @option{-O2} or higher. Usage of this option can improve the
7859 generated code and decrease its size by preventing register pressure
7860 increase above the number of available hard registers and subsequent
7861 spills in register allocation.
7863 @item -fsched-spec-load
7864 @opindex fsched-spec-load
7865 Allow speculative motion of some load instructions. This only makes
7866 sense when scheduling before register allocation, i.e.@: with
7867 @option{-fschedule-insns} or at @option{-O2} or higher.
7869 @item -fsched-spec-load-dangerous
7870 @opindex fsched-spec-load-dangerous
7871 Allow speculative motion of more load instructions. This only makes
7872 sense when scheduling before register allocation, i.e.@: with
7873 @option{-fschedule-insns} or at @option{-O2} or higher.
7875 @item -fsched-stalled-insns
7876 @itemx -fsched-stalled-insns=@var{n}
7877 @opindex fsched-stalled-insns
7878 Define how many insns (if any) can be moved prematurely from the queue
7879 of stalled insns into the ready list during the second scheduling pass.
7880 @option{-fno-sched-stalled-insns} means that no insns are moved
7881 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
7882 on how many queued insns can be moved prematurely.
7883 @option{-fsched-stalled-insns} without a value is equivalent to
7884 @option{-fsched-stalled-insns=1}.
7886 @item -fsched-stalled-insns-dep
7887 @itemx -fsched-stalled-insns-dep=@var{n}
7888 @opindex fsched-stalled-insns-dep
7889 Define how many insn groups (cycles) are examined for a dependency
7890 on a stalled insn that is a candidate for premature removal from the queue
7891 of stalled insns. This has an effect only during the second scheduling pass,
7892 and only if @option{-fsched-stalled-insns} is used.
7893 @option{-fno-sched-stalled-insns-dep} is equivalent to
7894 @option{-fsched-stalled-insns-dep=0}.
7895 @option{-fsched-stalled-insns-dep} without a value is equivalent to
7896 @option{-fsched-stalled-insns-dep=1}.
7898 @item -fsched2-use-superblocks
7899 @opindex fsched2-use-superblocks
7900 When scheduling after register allocation, use superblock scheduling.
7901 This allows motion across basic block boundaries,
7902 resulting in faster schedules. This option is experimental, as not all machine
7903 descriptions used by GCC model the CPU closely enough to avoid unreliable
7904 results from the algorithm.
7906 This only makes sense when scheduling after register allocation, i.e.@: with
7907 @option{-fschedule-insns2} or at @option{-O2} or higher.
7909 @item -fsched-group-heuristic
7910 @opindex fsched-group-heuristic
7911 Enable the group heuristic in the scheduler. This heuristic favors
7912 the instruction that belongs to a schedule group. This is enabled
7913 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7914 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7916 @item -fsched-critical-path-heuristic
7917 @opindex fsched-critical-path-heuristic
7918 Enable the critical-path heuristic in the scheduler. This heuristic favors
7919 instructions on the critical path. This is enabled by default when
7920 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7921 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7923 @item -fsched-spec-insn-heuristic
7924 @opindex fsched-spec-insn-heuristic
7925 Enable the speculative instruction heuristic in the scheduler. This
7926 heuristic favors speculative instructions with greater dependency weakness.
7927 This is enabled by default when scheduling is enabled, i.e.@:
7928 with @option{-fschedule-insns} or @option{-fschedule-insns2}
7929 or at @option{-O2} or higher.
7931 @item -fsched-rank-heuristic
7932 @opindex fsched-rank-heuristic
7933 Enable the rank heuristic in the scheduler. This heuristic favors
7934 the instruction belonging to a basic block with greater size or frequency.
7935 This is enabled by default when scheduling is enabled, i.e.@:
7936 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7937 at @option{-O2} or higher.
7939 @item -fsched-last-insn-heuristic
7940 @opindex fsched-last-insn-heuristic
7941 Enable the last-instruction heuristic in the scheduler. This heuristic
7942 favors the instruction that is less dependent on the last instruction
7943 scheduled. This is enabled by default when scheduling is enabled,
7944 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7945 at @option{-O2} or higher.
7947 @item -fsched-dep-count-heuristic
7948 @opindex fsched-dep-count-heuristic
7949 Enable the dependent-count heuristic in the scheduler. This heuristic
7950 favors the instruction that has more instructions depending on it.
7951 This is enabled by default when scheduling is enabled, i.e.@:
7952 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7953 at @option{-O2} or higher.
7955 @item -freschedule-modulo-scheduled-loops
7956 @opindex freschedule-modulo-scheduled-loops
7957 Modulo scheduling is performed before traditional scheduling. If a loop
7958 is modulo scheduled, later scheduling passes may change its schedule.
7959 Use this option to control that behavior.
7961 @item -fselective-scheduling
7962 @opindex fselective-scheduling
7963 Schedule instructions using selective scheduling algorithm. Selective
7964 scheduling runs instead of the first scheduler pass.
7966 @item -fselective-scheduling2
7967 @opindex fselective-scheduling2
7968 Schedule instructions using selective scheduling algorithm. Selective
7969 scheduling runs instead of the second scheduler pass.
7971 @item -fsel-sched-pipelining
7972 @opindex fsel-sched-pipelining
7973 Enable software pipelining of innermost loops during selective scheduling.
7974 This option has no effect unless one of @option{-fselective-scheduling} or
7975 @option{-fselective-scheduling2} is turned on.
7977 @item -fsel-sched-pipelining-outer-loops
7978 @opindex fsel-sched-pipelining-outer-loops
7979 When pipelining loops during selective scheduling, also pipeline outer loops.
7980 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
7982 @item -fsemantic-interposition
7983 @opindex fsemantic-interposition
7984 Some object formats, like ELF, allow interposing of symbols by dynamic linker.
7985 This means that for symbols exported from the DSO compiler can not perform
7986 inter-procedural propagation, inlining and other optimizations in anticipation
7987 that the function or variable in question may change. While this feature is
7988 useful, for example, to rewrite memory allocation functions by a debugging
7989 implementation, it is expensive in the terms of code quality.
7990 With @option{-fno-semantic-inteposition} compiler assumest that if interposition
7991 happens for functions the overwritting function will have
7992 precisely same semantics (and side effects). Similarly if interposition happens
7993 for variables, the constructor of the variable will be the same. The flag
7994 has no effect for functions explicitly declared inline, where
7995 interposition changing semantic is never allowed and for symbols explicitly
7999 @opindex fshrink-wrap
8000 Emit function prologues only before parts of the function that need it,
8001 rather than at the top of the function. This flag is enabled by default at
8002 @option{-O} and higher.
8004 @item -fcaller-saves
8005 @opindex fcaller-saves
8006 Enable allocation of values to registers that are clobbered by
8007 function calls, by emitting extra instructions to save and restore the
8008 registers around such calls. Such allocation is done only when it
8009 seems to result in better code.
8011 This option is always enabled by default on certain machines, usually
8012 those which have no call-preserved registers to use instead.
8014 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8016 @item -fcombine-stack-adjustments
8017 @opindex fcombine-stack-adjustments
8018 Tracks stack adjustments (pushes and pops) and stack memory references
8019 and then tries to find ways to combine them.
8021 Enabled by default at @option{-O1} and higher.
8023 @item -fuse-caller-save
8024 Use caller save registers for allocation if those registers are not used by
8025 any called function. In that case it is not necessary to save and restore
8026 them around calls. This is only possible if called functions are part of
8027 same compilation unit as current function and they are compiled before it.
8029 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8031 @item -fconserve-stack
8032 @opindex fconserve-stack
8033 Attempt to minimize stack usage. The compiler attempts to use less
8034 stack space, even if that makes the program slower. This option
8035 implies setting the @option{large-stack-frame} parameter to 100
8036 and the @option{large-stack-frame-growth} parameter to 400.
8038 @item -ftree-reassoc
8039 @opindex ftree-reassoc
8040 Perform reassociation on trees. This flag is enabled by default
8041 at @option{-O} and higher.
8045 Perform partial redundancy elimination (PRE) on trees. This flag is
8046 enabled by default at @option{-O2} and @option{-O3}.
8048 @item -ftree-partial-pre
8049 @opindex ftree-partial-pre
8050 Make partial redundancy elimination (PRE) more aggressive. This flag is
8051 enabled by default at @option{-O3}.
8053 @item -ftree-forwprop
8054 @opindex ftree-forwprop
8055 Perform forward propagation on trees. This flag is enabled by default
8056 at @option{-O} and higher.
8060 Perform full redundancy elimination (FRE) on trees. The difference
8061 between FRE and PRE is that FRE only considers expressions
8062 that are computed on all paths leading to the redundant computation.
8063 This analysis is faster than PRE, though it exposes fewer redundancies.
8064 This flag is enabled by default at @option{-O} and higher.
8066 @item -ftree-phiprop
8067 @opindex ftree-phiprop
8068 Perform hoisting of loads from conditional pointers on trees. This
8069 pass is enabled by default at @option{-O} and higher.
8071 @item -fhoist-adjacent-loads
8072 @opindex hoist-adjacent-loads
8073 Speculatively hoist loads from both branches of an if-then-else if the
8074 loads are from adjacent locations in the same structure and the target
8075 architecture has a conditional move instruction. This flag is enabled
8076 by default at @option{-O2} and higher.
8078 @item -ftree-copy-prop
8079 @opindex ftree-copy-prop
8080 Perform copy propagation on trees. This pass eliminates unnecessary
8081 copy operations. This flag is enabled by default at @option{-O} and
8084 @item -fipa-pure-const
8085 @opindex fipa-pure-const
8086 Discover which functions are pure or constant.
8087 Enabled by default at @option{-O} and higher.
8089 @item -fipa-reference
8090 @opindex fipa-reference
8091 Discover which static variables do not escape the
8093 Enabled by default at @option{-O} and higher.
8097 Perform interprocedural pointer analysis and interprocedural modification
8098 and reference analysis. This option can cause excessive memory and
8099 compile-time usage on large compilation units. It is not enabled by
8100 default at any optimization level.
8103 @opindex fipa-profile
8104 Perform interprocedural profile propagation. The functions called only from
8105 cold functions are marked as cold. Also functions executed once (such as
8106 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8107 functions and loop less parts of functions executed once are then optimized for
8109 Enabled by default at @option{-O} and higher.
8113 Perform interprocedural constant propagation.
8114 This optimization analyzes the program to determine when values passed
8115 to functions are constants and then optimizes accordingly.
8116 This optimization can substantially increase performance
8117 if the application has constants passed to functions.
8118 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8120 @item -fipa-cp-clone
8121 @opindex fipa-cp-clone
8122 Perform function cloning to make interprocedural constant propagation stronger.
8123 When enabled, interprocedural constant propagation performs function cloning
8124 when externally visible function can be called with constant arguments.
8125 Because this optimization can create multiple copies of functions,
8126 it may significantly increase code size
8127 (see @option{--param ipcp-unit-growth=@var{value}}).
8128 This flag is enabled by default at @option{-O3}.
8132 Perform Identical Code Folding for functions and read-only variables.
8133 The optimization reduces code size and may disturb unwind stacks by replacing
8134 a function by equivalent one with a different name. The optimization works
8135 more effectively with link time optimization enabled.
8137 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8138 works on different levels and thus the optimizations are not same - there are
8139 equivalences that are found only by GCC and equivalences found only by Gold.
8141 This flag is enabled by default at @option{-O2} and @option{-Os}.
8143 @item -fisolate-erroneous-paths-dereference
8144 Detect paths which trigger erroneous or undefined behaviour due to
8145 dereferencing a NULL pointer. Isolate those paths from the main control
8146 flow and turn the statement with erroneous or undefined behaviour into a trap.
8148 @item -fisolate-erroneous-paths-attribute
8149 Detect paths which trigger erroneous or undefined behaviour due a NULL value
8150 being used in a way which is forbidden by a @code{returns_nonnull} or @code{nonnull}
8151 attribute. Isolate those paths from the main control flow and turn the
8152 statement with erroneous or undefined behaviour into a trap. This is not
8153 currently enabled, but may be enabled by @code{-O2} in the future.
8157 Perform forward store motion on trees. This flag is
8158 enabled by default at @option{-O} and higher.
8160 @item -ftree-bit-ccp
8161 @opindex ftree-bit-ccp
8162 Perform sparse conditional bit constant propagation on trees and propagate
8163 pointer alignment information.
8164 This pass only operates on local scalar variables and is enabled by default
8165 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8169 Perform sparse conditional constant propagation (CCP) on trees. This
8170 pass only operates on local scalar variables and is enabled by default
8171 at @option{-O} and higher.
8174 @opindex fssa-phiopt
8175 Perform pattern matching on SSA PHI nodes to optimize conditional
8176 code. This pass is enabled by default at @option{-O} and higher.
8178 @item -ftree-switch-conversion
8179 Perform conversion of simple initializations in a switch to
8180 initializations from a scalar array. This flag is enabled by default
8181 at @option{-O2} and higher.
8183 @item -ftree-tail-merge
8184 Look for identical code sequences. When found, replace one with a jump to the
8185 other. This optimization is known as tail merging or cross jumping. This flag
8186 is enabled by default at @option{-O2} and higher. The compilation time
8188 be limited using @option{max-tail-merge-comparisons} parameter and
8189 @option{max-tail-merge-iterations} parameter.
8193 Perform dead code elimination (DCE) on trees. This flag is enabled by
8194 default at @option{-O} and higher.
8196 @item -ftree-builtin-call-dce
8197 @opindex ftree-builtin-call-dce
8198 Perform conditional dead code elimination (DCE) for calls to built-in functions
8199 that may set @code{errno} but are otherwise side-effect free. This flag is
8200 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8203 @item -ftree-dominator-opts
8204 @opindex ftree-dominator-opts
8205 Perform a variety of simple scalar cleanups (constant/copy
8206 propagation, redundancy elimination, range propagation and expression
8207 simplification) based on a dominator tree traversal. This also
8208 performs jump threading (to reduce jumps to jumps). This flag is
8209 enabled by default at @option{-O} and higher.
8213 Perform dead store elimination (DSE) on trees. A dead store is a store into
8214 a memory location that is later overwritten by another store without
8215 any intervening loads. In this case the earlier store can be deleted. This
8216 flag is enabled by default at @option{-O} and higher.
8220 Perform loop header copying on trees. This is beneficial since it increases
8221 effectiveness of code motion optimizations. It also saves one jump. This flag
8222 is enabled by default at @option{-O} and higher. It is not enabled
8223 for @option{-Os}, since it usually increases code size.
8225 @item -ftree-loop-optimize
8226 @opindex ftree-loop-optimize
8227 Perform loop optimizations on trees. This flag is enabled by default
8228 at @option{-O} and higher.
8230 @item -ftree-loop-linear
8231 @opindex ftree-loop-linear
8232 Perform loop interchange transformations on tree. Same as
8233 @option{-floop-interchange}. To use this code transformation, GCC has
8234 to be configured with @option{--with-ppl} and @option{--with-cloog} to
8235 enable the Graphite loop transformation infrastructure.
8237 @item -floop-interchange
8238 @opindex floop-interchange
8239 Perform loop interchange transformations on loops. Interchanging two
8240 nested loops switches the inner and outer loops. For example, given a
8245 A(J, I) = A(J, I) * C
8249 loop interchange transforms the loop as if it were written:
8253 A(J, I) = A(J, I) * C
8257 which can be beneficial when @code{N} is larger than the caches,
8258 because in Fortran, the elements of an array are stored in memory
8259 contiguously by column, and the original loop iterates over rows,
8260 potentially creating at each access a cache miss. This optimization
8261 applies to all the languages supported by GCC and is not limited to
8262 Fortran. To use this code transformation, GCC has to be configured
8263 with @option{--with-ppl} and @option{--with-cloog} to enable the
8264 Graphite loop transformation infrastructure.
8266 @item -floop-strip-mine
8267 @opindex floop-strip-mine
8268 Perform loop strip mining transformations on loops. Strip mining
8269 splits a loop into two nested loops. The outer loop has strides
8270 equal to the strip size and the inner loop has strides of the
8271 original loop within a strip. The strip length can be changed
8272 using the @option{loop-block-tile-size} parameter. For example,
8279 loop strip mining transforms the loop as if it were written:
8282 DO I = II, min (II + 50, N)
8287 This optimization applies to all the languages supported by GCC and is
8288 not limited to Fortran. To use this code transformation, GCC has to
8289 be configured with @option{--with-ppl} and @option{--with-cloog} to
8290 enable the Graphite loop transformation infrastructure.
8293 @opindex floop-block
8294 Perform loop blocking transformations on loops. Blocking strip mines
8295 each loop in the loop nest such that the memory accesses of the
8296 element loops fit inside caches. The strip length can be changed
8297 using the @option{loop-block-tile-size} parameter. For example, given
8302 A(J, I) = B(I) + C(J)
8306 loop blocking transforms the loop as if it were written:
8310 DO I = II, min (II + 50, N)
8311 DO J = JJ, min (JJ + 50, M)
8312 A(J, I) = B(I) + C(J)
8318 which can be beneficial when @code{M} is larger than the caches,
8319 because the innermost loop iterates over a smaller amount of data
8320 which can be kept in the caches. This optimization applies to all the
8321 languages supported by GCC and is not limited to Fortran. To use this
8322 code transformation, GCC has to be configured with @option{--with-ppl}
8323 and @option{--with-cloog} to enable the Graphite loop transformation
8326 @item -fgraphite-identity
8327 @opindex fgraphite-identity
8328 Enable the identity transformation for graphite. For every SCoP we generate
8329 the polyhedral representation and transform it back to gimple. Using
8330 @option{-fgraphite-identity} we can check the costs or benefits of the
8331 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8332 are also performed by the code generator CLooG, like index splitting and
8333 dead code elimination in loops.
8335 @item -floop-nest-optimize
8336 @opindex floop-nest-optimize
8337 Enable the ISL based loop nest optimizer. This is a generic loop nest
8338 optimizer based on the Pluto optimization algorithms. It calculates a loop
8339 structure optimized for data-locality and parallelism. This option
8342 @item -floop-parallelize-all
8343 @opindex floop-parallelize-all
8344 Use the Graphite data dependence analysis to identify loops that can
8345 be parallelized. Parallelize all the loops that can be analyzed to
8346 not contain loop carried dependences without checking that it is
8347 profitable to parallelize the loops.
8349 @item -fcheck-data-deps
8350 @opindex fcheck-data-deps
8351 Compare the results of several data dependence analyzers. This option
8352 is used for debugging the data dependence analyzers.
8354 @item -ftree-loop-if-convert
8355 Attempt to transform conditional jumps in the innermost loops to
8356 branch-less equivalents. The intent is to remove control-flow from
8357 the innermost loops in order to improve the ability of the
8358 vectorization pass to handle these loops. This is enabled by default
8359 if vectorization is enabled.
8361 @item -ftree-loop-if-convert-stores
8362 Attempt to also if-convert conditional jumps containing memory writes.
8363 This transformation can be unsafe for multi-threaded programs as it
8364 transforms conditional memory writes into unconditional memory writes.
8367 for (i = 0; i < N; i++)
8373 for (i = 0; i < N; i++)
8374 A[i] = cond ? expr : A[i];
8376 potentially producing data races.
8378 @item -ftree-loop-distribution
8379 Perform loop distribution. This flag can improve cache performance on
8380 big loop bodies and allow further loop optimizations, like
8381 parallelization or vectorization, to take place. For example, the loop
8398 @item -ftree-loop-distribute-patterns
8399 Perform loop distribution of patterns that can be code generated with
8400 calls to a library. This flag is enabled by default at @option{-O3}.
8402 This pass distributes the initialization loops and generates a call to
8403 memset zero. For example, the loop
8419 and the initialization loop is transformed into a call to memset zero.
8421 @item -ftree-loop-im
8422 @opindex ftree-loop-im
8423 Perform loop invariant motion on trees. This pass moves only invariants that
8424 are hard to handle at RTL level (function calls, operations that expand to
8425 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8426 operands of conditions that are invariant out of the loop, so that we can use
8427 just trivial invariantness analysis in loop unswitching. The pass also includes
8430 @item -ftree-loop-ivcanon
8431 @opindex ftree-loop-ivcanon
8432 Create a canonical counter for number of iterations in loops for which
8433 determining number of iterations requires complicated analysis. Later
8434 optimizations then may determine the number easily. Useful especially
8435 in connection with unrolling.
8439 Perform induction variable optimizations (strength reduction, induction
8440 variable merging and induction variable elimination) on trees.
8442 @item -ftree-parallelize-loops=n
8443 @opindex ftree-parallelize-loops
8444 Parallelize loops, i.e., split their iteration space to run in n threads.
8445 This is only possible for loops whose iterations are independent
8446 and can be arbitrarily reordered. The optimization is only
8447 profitable on multiprocessor machines, for loops that are CPU-intensive,
8448 rather than constrained e.g.@: by memory bandwidth. This option
8449 implies @option{-pthread}, and thus is only supported on targets
8450 that have support for @option{-pthread}.
8454 Perform function-local points-to analysis on trees. This flag is
8455 enabled by default at @option{-O} and higher.
8459 Perform scalar replacement of aggregates. This pass replaces structure
8460 references with scalars to prevent committing structures to memory too
8461 early. This flag is enabled by default at @option{-O} and higher.
8463 @item -ftree-copyrename
8464 @opindex ftree-copyrename
8465 Perform copy renaming on trees. This pass attempts to rename compiler
8466 temporaries to other variables at copy locations, usually resulting in
8467 variable names which more closely resemble the original variables. This flag
8468 is enabled by default at @option{-O} and higher.
8470 @item -ftree-coalesce-inlined-vars
8471 @opindex ftree-coalesce-inlined-vars
8472 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8473 combine small user-defined variables too, but only if they were inlined
8474 from other functions. It is a more limited form of
8475 @option{-ftree-coalesce-vars}. This may harm debug information of such
8476 inlined variables, but it will keep variables of the inlined-into
8477 function apart from each other, such that they are more likely to
8478 contain the expected values in a debugging session. This was the
8479 default in GCC versions older than 4.7.
8481 @item -ftree-coalesce-vars
8482 @opindex ftree-coalesce-vars
8483 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8484 combine small user-defined variables too, instead of just compiler
8485 temporaries. This may severely limit the ability to debug an optimized
8486 program compiled with @option{-fno-var-tracking-assignments}. In the
8487 negated form, this flag prevents SSA coalescing of user variables,
8488 including inlined ones. This option is enabled by default.
8492 Perform temporary expression replacement during the SSA->normal phase. Single
8493 use/single def temporaries are replaced at their use location with their
8494 defining expression. This results in non-GIMPLE code, but gives the expanders
8495 much more complex trees to work on resulting in better RTL generation. This is
8496 enabled by default at @option{-O} and higher.
8500 Perform straight-line strength reduction on trees. This recognizes related
8501 expressions involving multiplications and replaces them by less expensive
8502 calculations when possible. This is enabled by default at @option{-O} and
8505 @item -ftree-vectorize
8506 @opindex ftree-vectorize
8507 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8508 and @option{-ftree-slp-vectorize} if not explicitly specified.
8510 @item -ftree-loop-vectorize
8511 @opindex ftree-loop-vectorize
8512 Perform loop vectorization on trees. This flag is enabled by default at
8513 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8515 @item -ftree-slp-vectorize
8516 @opindex ftree-slp-vectorize
8517 Perform basic block vectorization on trees. This flag is enabled by default at
8518 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8520 @item -fvect-cost-model=@var{model}
8521 @opindex fvect-cost-model
8522 Alter the cost model used for vectorization. The @var{model} argument
8523 should be one of @code{unlimited}, @code{dynamic} or @code{cheap}.
8524 With the @code{unlimited} model the vectorized code-path is assumed
8525 to be profitable while with the @code{dynamic} model a runtime check
8526 will guard the vectorized code-path to enable it only for iteration
8527 counts that will likely execute faster than when executing the original
8528 scalar loop. The @code{cheap} model will disable vectorization of
8529 loops where doing so would be cost prohibitive for example due to
8530 required runtime checks for data dependence or alignment but otherwise
8531 is equal to the @code{dynamic} model.
8532 The default cost model depends on other optimization flags and is
8533 either @code{dynamic} or @code{cheap}.
8535 @item -fsimd-cost-model=@var{model}
8536 @opindex fsimd-cost-model
8537 Alter the cost model used for vectorization of loops marked with the OpenMP
8538 or Cilk Plus simd directive. The @var{model} argument should be one of
8539 @code{unlimited}, @code{dynamic}, @code{cheap}. All values of @var{model}
8540 have the same meaning as described in @option{-fvect-cost-model} and by
8541 default a cost model defined with @option{-fvect-cost-model} is used.
8545 Perform Value Range Propagation on trees. This is similar to the
8546 constant propagation pass, but instead of values, ranges of values are
8547 propagated. This allows the optimizers to remove unnecessary range
8548 checks like array bound checks and null pointer checks. This is
8549 enabled by default at @option{-O2} and higher. Null pointer check
8550 elimination is only done if @option{-fdelete-null-pointer-checks} is
8555 Perform tail duplication to enlarge superblock size. This transformation
8556 simplifies the control flow of the function allowing other optimizations to do
8559 @item -funroll-loops
8560 @opindex funroll-loops
8561 Unroll loops whose number of iterations can be determined at compile
8562 time or upon entry to the loop. @option{-funroll-loops} implies
8563 @option{-frerun-cse-after-loop}. This option makes code larger,
8564 and may or may not make it run faster.
8566 @item -funroll-all-loops
8567 @opindex funroll-all-loops
8568 Unroll all loops, even if their number of iterations is uncertain when
8569 the loop is entered. This usually makes programs run more slowly.
8570 @option{-funroll-all-loops} implies the same options as
8571 @option{-funroll-loops},
8573 @item -fsplit-ivs-in-unroller
8574 @opindex fsplit-ivs-in-unroller
8575 Enables expression of values of induction variables in later iterations
8576 of the unrolled loop using the value in the first iteration. This breaks
8577 long dependency chains, thus improving efficiency of the scheduling passes.
8579 A combination of @option{-fweb} and CSE is often sufficient to obtain the
8580 same effect. However, that is not reliable in cases where the loop body
8581 is more complicated than a single basic block. It also does not work at all
8582 on some architectures due to restrictions in the CSE pass.
8584 This optimization is enabled by default.
8586 @item -fvariable-expansion-in-unroller
8587 @opindex fvariable-expansion-in-unroller
8588 With this option, the compiler creates multiple copies of some
8589 local variables when unrolling a loop, which can result in superior code.
8591 @item -fpartial-inlining
8592 @opindex fpartial-inlining
8593 Inline parts of functions. This option has any effect only
8594 when inlining itself is turned on by the @option{-finline-functions}
8595 or @option{-finline-small-functions} options.
8597 Enabled at level @option{-O2}.
8599 @item -fpredictive-commoning
8600 @opindex fpredictive-commoning
8601 Perform predictive commoning optimization, i.e., reusing computations
8602 (especially memory loads and stores) performed in previous
8603 iterations of loops.
8605 This option is enabled at level @option{-O3}.
8607 @item -fprefetch-loop-arrays
8608 @opindex fprefetch-loop-arrays
8609 If supported by the target machine, generate instructions to prefetch
8610 memory to improve the performance of loops that access large arrays.
8612 This option may generate better or worse code; results are highly
8613 dependent on the structure of loops within the source code.
8615 Disabled at level @option{-Os}.
8618 @itemx -fno-peephole2
8619 @opindex fno-peephole
8620 @opindex fno-peephole2
8621 Disable any machine-specific peephole optimizations. The difference
8622 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
8623 are implemented in the compiler; some targets use one, some use the
8624 other, a few use both.
8626 @option{-fpeephole} is enabled by default.
8627 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8629 @item -fno-guess-branch-probability
8630 @opindex fno-guess-branch-probability
8631 Do not guess branch probabilities using heuristics.
8633 GCC uses heuristics to guess branch probabilities if they are
8634 not provided by profiling feedback (@option{-fprofile-arcs}). These
8635 heuristics are based on the control flow graph. If some branch probabilities
8636 are specified by @samp{__builtin_expect}, then the heuristics are
8637 used to guess branch probabilities for the rest of the control flow graph,
8638 taking the @samp{__builtin_expect} info into account. The interactions
8639 between the heuristics and @samp{__builtin_expect} can be complex, and in
8640 some cases, it may be useful to disable the heuristics so that the effects
8641 of @samp{__builtin_expect} are easier to understand.
8643 The default is @option{-fguess-branch-probability} at levels
8644 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8646 @item -freorder-blocks
8647 @opindex freorder-blocks
8648 Reorder basic blocks in the compiled function in order to reduce number of
8649 taken branches and improve code locality.
8651 Enabled at levels @option{-O2}, @option{-O3}.
8653 @item -freorder-blocks-and-partition
8654 @opindex freorder-blocks-and-partition
8655 In addition to reordering basic blocks in the compiled function, in order
8656 to reduce number of taken branches, partitions hot and cold basic blocks
8657 into separate sections of the assembly and .o files, to improve
8658 paging and cache locality performance.
8660 This optimization is automatically turned off in the presence of
8661 exception handling, for linkonce sections, for functions with a user-defined
8662 section attribute and on any architecture that does not support named
8665 Enabled for x86 at levels @option{-O2}, @option{-O3}.
8667 @item -freorder-functions
8668 @opindex freorder-functions
8669 Reorder functions in the object file in order to
8670 improve code locality. This is implemented by using special
8671 subsections @code{.text.hot} for most frequently executed functions and
8672 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
8673 the linker so object file format must support named sections and linker must
8674 place them in a reasonable way.
8676 Also profile feedback must be available to make this option effective. See
8677 @option{-fprofile-arcs} for details.
8679 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8681 @item -fstrict-aliasing
8682 @opindex fstrict-aliasing
8683 Allow the compiler to assume the strictest aliasing rules applicable to
8684 the language being compiled. For C (and C++), this activates
8685 optimizations based on the type of expressions. In particular, an
8686 object of one type is assumed never to reside at the same address as an
8687 object of a different type, unless the types are almost the same. For
8688 example, an @code{unsigned int} can alias an @code{int}, but not a
8689 @code{void*} or a @code{double}. A character type may alias any other
8692 @anchor{Type-punning}Pay special attention to code like this:
8705 The practice of reading from a different union member than the one most
8706 recently written to (called ``type-punning'') is common. Even with
8707 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
8708 is accessed through the union type. So, the code above works as
8709 expected. @xref{Structures unions enumerations and bit-fields
8710 implementation}. However, this code might not:
8721 Similarly, access by taking the address, casting the resulting pointer
8722 and dereferencing the result has undefined behavior, even if the cast
8723 uses a union type, e.g.:
8727 return ((union a_union *) &d)->i;
8731 The @option{-fstrict-aliasing} option is enabled at levels
8732 @option{-O2}, @option{-O3}, @option{-Os}.
8734 @item -fstrict-overflow
8735 @opindex fstrict-overflow
8736 Allow the compiler to assume strict signed overflow rules, depending
8737 on the language being compiled. For C (and C++) this means that
8738 overflow when doing arithmetic with signed numbers is undefined, which
8739 means that the compiler may assume that it does not happen. This
8740 permits various optimizations. For example, the compiler assumes
8741 that an expression like @code{i + 10 > i} is always true for
8742 signed @code{i}. This assumption is only valid if signed overflow is
8743 undefined, as the expression is false if @code{i + 10} overflows when
8744 using twos complement arithmetic. When this option is in effect any
8745 attempt to determine whether an operation on signed numbers
8746 overflows must be written carefully to not actually involve overflow.
8748 This option also allows the compiler to assume strict pointer
8749 semantics: given a pointer to an object, if adding an offset to that
8750 pointer does not produce a pointer to the same object, the addition is
8751 undefined. This permits the compiler to conclude that @code{p + u >
8752 p} is always true for a pointer @code{p} and unsigned integer
8753 @code{u}. This assumption is only valid because pointer wraparound is
8754 undefined, as the expression is false if @code{p + u} overflows using
8755 twos complement arithmetic.
8757 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
8758 that integer signed overflow is fully defined: it wraps. When
8759 @option{-fwrapv} is used, there is no difference between
8760 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
8761 integers. With @option{-fwrapv} certain types of overflow are
8762 permitted. For example, if the compiler gets an overflow when doing
8763 arithmetic on constants, the overflowed value can still be used with
8764 @option{-fwrapv}, but not otherwise.
8766 The @option{-fstrict-overflow} option is enabled at levels
8767 @option{-O2}, @option{-O3}, @option{-Os}.
8769 @item -falign-functions
8770 @itemx -falign-functions=@var{n}
8771 @opindex falign-functions
8772 Align the start of functions to the next power-of-two greater than
8773 @var{n}, skipping up to @var{n} bytes. For instance,
8774 @option{-falign-functions=32} aligns functions to the next 32-byte
8775 boundary, but @option{-falign-functions=24} aligns to the next
8776 32-byte boundary only if this can be done by skipping 23 bytes or less.
8778 @option{-fno-align-functions} and @option{-falign-functions=1} are
8779 equivalent and mean that functions are not aligned.
8781 Some assemblers only support this flag when @var{n} is a power of two;
8782 in that case, it is rounded up.
8784 If @var{n} is not specified or is zero, use a machine-dependent default.
8786 Enabled at levels @option{-O2}, @option{-O3}.
8788 @item -falign-labels
8789 @itemx -falign-labels=@var{n}
8790 @opindex falign-labels
8791 Align all branch targets to a power-of-two boundary, skipping up to
8792 @var{n} bytes like @option{-falign-functions}. This option can easily
8793 make code slower, because it must insert dummy operations for when the
8794 branch target is reached in the usual flow of the code.
8796 @option{-fno-align-labels} and @option{-falign-labels=1} are
8797 equivalent and mean that labels are not aligned.
8799 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
8800 are greater than this value, then their values are used instead.
8802 If @var{n} is not specified or is zero, use a machine-dependent default
8803 which is very likely to be @samp{1}, meaning no alignment.
8805 Enabled at levels @option{-O2}, @option{-O3}.
8808 @itemx -falign-loops=@var{n}
8809 @opindex falign-loops
8810 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
8811 like @option{-falign-functions}. If the loops are
8812 executed many times, this makes up for any execution of the dummy
8815 @option{-fno-align-loops} and @option{-falign-loops=1} are
8816 equivalent and mean that loops are not aligned.
8818 If @var{n} is not specified or is zero, use a machine-dependent default.
8820 Enabled at levels @option{-O2}, @option{-O3}.
8823 @itemx -falign-jumps=@var{n}
8824 @opindex falign-jumps
8825 Align branch targets to a power-of-two boundary, for branch targets
8826 where the targets can only be reached by jumping, skipping up to @var{n}
8827 bytes like @option{-falign-functions}. In this case, no dummy operations
8830 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
8831 equivalent and mean that loops are not aligned.
8833 If @var{n} is not specified or is zero, use a machine-dependent default.
8835 Enabled at levels @option{-O2}, @option{-O3}.
8837 @item -funit-at-a-time
8838 @opindex funit-at-a-time
8839 This option is left for compatibility reasons. @option{-funit-at-a-time}
8840 has no effect, while @option{-fno-unit-at-a-time} implies
8841 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
8845 @item -fno-toplevel-reorder
8846 @opindex fno-toplevel-reorder
8847 Do not reorder top-level functions, variables, and @code{asm}
8848 statements. Output them in the same order that they appear in the
8849 input file. When this option is used, unreferenced static variables
8850 are not removed. This option is intended to support existing code
8851 that relies on a particular ordering. For new code, it is better to
8852 use attributes when possible.
8854 Enabled at level @option{-O0}. When disabled explicitly, it also implies
8855 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
8860 Constructs webs as commonly used for register allocation purposes and assign
8861 each web individual pseudo register. This allows the register allocation pass
8862 to operate on pseudos directly, but also strengthens several other optimization
8863 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
8864 however, make debugging impossible, since variables no longer stay in a
8867 Enabled by default with @option{-funroll-loops}.
8869 @item -fwhole-program
8870 @opindex fwhole-program
8871 Assume that the current compilation unit represents the whole program being
8872 compiled. All public functions and variables with the exception of @code{main}
8873 and those merged by attribute @code{externally_visible} become static functions
8874 and in effect are optimized more aggressively by interprocedural optimizers.
8876 This option should not be used in combination with @code{-flto}.
8877 Instead relying on a linker plugin should provide safer and more precise
8880 @item -flto[=@var{n}]
8882 This option runs the standard link-time optimizer. When invoked
8883 with source code, it generates GIMPLE (one of GCC's internal
8884 representations) and writes it to special ELF sections in the object
8885 file. When the object files are linked together, all the function
8886 bodies are read from these ELF sections and instantiated as if they
8887 had been part of the same translation unit.
8889 To use the link-time optimizer, @option{-flto} and optimization
8890 options should be specified at compile time and during the final link.
8894 gcc -c -O2 -flto foo.c
8895 gcc -c -O2 -flto bar.c
8896 gcc -o myprog -flto -O2 foo.o bar.o
8899 The first two invocations to GCC save a bytecode representation
8900 of GIMPLE into special ELF sections inside @file{foo.o} and
8901 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
8902 @file{foo.o} and @file{bar.o}, merges the two files into a single
8903 internal image, and compiles the result as usual. Since both
8904 @file{foo.o} and @file{bar.o} are merged into a single image, this
8905 causes all the interprocedural analyses and optimizations in GCC to
8906 work across the two files as if they were a single one. This means,
8907 for example, that the inliner is able to inline functions in
8908 @file{bar.o} into functions in @file{foo.o} and vice-versa.
8910 Another (simpler) way to enable link-time optimization is:
8913 gcc -o myprog -flto -O2 foo.c bar.c
8916 The above generates bytecode for @file{foo.c} and @file{bar.c},
8917 merges them together into a single GIMPLE representation and optimizes
8918 them as usual to produce @file{myprog}.
8920 The only important thing to keep in mind is that to enable link-time
8921 optimizations you need to use the GCC driver to perform the link-step.
8922 GCC then automatically performs link-time optimization if any of the
8923 objects involved were compiled with the @option{-flto}. You generally
8924 should specify the optimization options to be used for link-time
8925 optimization though GCC will try to be clever at guessing an
8926 optimization level to use from the options used at compile-time
8927 if you fail to specify one at link-time. You can always override
8928 the automatic decision to do link-time optimization at link-time
8929 by passing @option{-fno-lto} to the link command.
8931 To make whole program optimization effective, it is necessary to make
8932 certain whole program assumptions. The compiler needs to know
8933 what functions and variables can be accessed by libraries and runtime
8934 outside of the link-time optimized unit. When supported by the linker,
8935 the linker plugin (see @option{-fuse-linker-plugin}) passes information
8936 to the compiler about used and externally visible symbols. When
8937 the linker plugin is not available, @option{-fwhole-program} should be
8938 used to allow the compiler to make these assumptions, which leads
8939 to more aggressive optimization decisions.
8941 When @option{-fuse-linker-plugin} is not enabled then, when a file is
8942 compiled with @option{-flto}, the generated object file is larger than
8943 a regular object file because it contains GIMPLE bytecodes and the usual
8944 final code (see @option{-ffat-lto-objects}. This means that
8945 object files with LTO information can be linked as normal object
8946 files; if @option{-fno-lto} is passed to the linker, no
8947 interprocedural optimizations are applied. Note that when
8948 @option{-fno-fat-lto-objects} is enabled the compile-stage is faster
8949 but you cannot perform a regular, non-LTO link on them.
8951 Additionally, the optimization flags used to compile individual files
8952 are not necessarily related to those used at link time. For instance,
8955 gcc -c -O0 -ffat-lto-objects -flto foo.c
8956 gcc -c -O0 -ffat-lto-objects -flto bar.c
8957 gcc -o myprog -O3 foo.o bar.o
8960 This produces individual object files with unoptimized assembler
8961 code, but the resulting binary @file{myprog} is optimized at
8962 @option{-O3}. If, instead, the final binary is generated with
8963 @option{-fno-lto}, then @file{myprog} is not optimized.
8965 When producing the final binary, GCC only
8966 applies link-time optimizations to those files that contain bytecode.
8967 Therefore, you can mix and match object files and libraries with
8968 GIMPLE bytecodes and final object code. GCC automatically selects
8969 which files to optimize in LTO mode and which files to link without
8972 There are some code generation flags preserved by GCC when
8973 generating bytecodes, as they need to be used during the final link
8974 stage. Generally options specified at link-time override those
8975 specified at compile-time.
8977 If you do not specify an optimization level option @option{-O} at
8978 link-time then GCC will compute one based on the optimization levels
8979 used when compiling the object files. The highest optimization
8980 level will win here.
8982 Currently, the following options and their setting are take from
8983 the first object file that explicitely specified it:
8984 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
8985 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
8986 and all the @option{-m} target flags.
8988 Certain ABI changing flags are required to match in all compilation-units
8989 and trying to override this at link-time with a conflicting value
8990 is ignored. This includes options such as @option{-freg-struct-return}
8991 and @option{-fpcc-struct-return}.
8993 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
8994 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
8995 are passed through to the link stage and merged conservatively for
8996 conflicting translation units. Specifically
8997 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
8998 precedence and for example @option{-ffp-contract=off} takes precedence
8999 over @option{-ffp-contract=fast}. You can override them at linke-time.
9001 It is recommended that you compile all the files participating in the
9002 same link with the same options and also specify those options at
9005 If LTO encounters objects with C linkage declared with incompatible
9006 types in separate translation units to be linked together (undefined
9007 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
9008 issued. The behavior is still undefined at run time. Similar
9009 diagnostics may be raised for other languages.
9011 Another feature of LTO is that it is possible to apply interprocedural
9012 optimizations on files written in different languages:
9017 gfortran -c -flto baz.f90
9018 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
9021 Notice that the final link is done with @command{g++} to get the C++
9022 runtime libraries and @option{-lgfortran} is added to get the Fortran
9023 runtime libraries. In general, when mixing languages in LTO mode, you
9024 should use the same link command options as when mixing languages in a
9025 regular (non-LTO) compilation.
9027 If object files containing GIMPLE bytecode are stored in a library archive, say
9028 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
9029 are using a linker with plugin support. To create static libraries suitable
9030 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
9031 and @code{ranlib}; to show the symbols of object files with GIMPLE bytecode, use
9032 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
9033 and @command{nm} have been compiled with plugin support. At link time, use the the
9034 flag @option{-fuse-linker-plugin} to ensure that the library participates in
9035 the LTO optimization process:
9038 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
9041 With the linker plugin enabled, the linker extracts the needed
9042 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
9043 to make them part of the aggregated GIMPLE image to be optimized.
9045 If you are not using a linker with plugin support and/or do not
9046 enable the linker plugin, then the objects inside @file{libfoo.a}
9047 are extracted and linked as usual, but they do not participate
9048 in the LTO optimization process. In order to make a static library suitable
9049 for both LTO optimization and usual linkage, compile its object files with
9050 @option{-flto} @code{-ffat-lto-objects}.
9052 Link-time optimizations do not require the presence of the whole program to
9053 operate. If the program does not require any symbols to be exported, it is
9054 possible to combine @option{-flto} and @option{-fwhole-program} to allow
9055 the interprocedural optimizers to use more aggressive assumptions which may
9056 lead to improved optimization opportunities.
9057 Use of @option{-fwhole-program} is not needed when linker plugin is
9058 active (see @option{-fuse-linker-plugin}).
9060 The current implementation of LTO makes no
9061 attempt to generate bytecode that is portable between different
9062 types of hosts. The bytecode files are versioned and there is a
9063 strict version check, so bytecode files generated in one version of
9064 GCC will not work with an older or newer version of GCC.
9066 Link-time optimization does not work well with generation of debugging
9067 information. Combining @option{-flto} with
9068 @option{-g} is currently experimental and expected to produce unexpected
9071 If you specify the optional @var{n}, the optimization and code
9072 generation done at link time is executed in parallel using @var{n}
9073 parallel jobs by utilizing an installed @command{make} program. The
9074 environment variable @env{MAKE} may be used to override the program
9075 used. The default value for @var{n} is 1.
9077 You can also specify @option{-flto=jobserver} to use GNU make's
9078 job server mode to determine the number of parallel jobs. This
9079 is useful when the Makefile calling GCC is already executing in parallel.
9080 You must prepend a @samp{+} to the command recipe in the parent Makefile
9081 for this to work. This option likely only works if @env{MAKE} is
9084 @item -flto-partition=@var{alg}
9085 @opindex flto-partition
9086 Specify the partitioning algorithm used by the link-time optimizer.
9087 The value is either @code{1to1} to specify a partitioning mirroring
9088 the original source files or @code{balanced} to specify partitioning
9089 into equally sized chunks (whenever possible) or @code{max} to create
9090 new partition for every symbol where possible. Specifying @code{none}
9091 as an algorithm disables partitioning and streaming completely.
9092 The default value is @code{balanced}. While @code{1to1} can be used
9093 as an workaround for various code ordering issues, the @code{max}
9094 partitioning is intended for internal testing only.
9095 The value @code{one} specifies that exactly one partition should be
9096 used while the value @code{none} bypasses partitioning and executes
9097 the link-time optimization step directly from the WPA phase.
9099 @item -flto-odr-type-merging
9100 @opindex flto-odr-type-merging
9101 Enable streaming of mangled types names of C++ types and their unification
9102 at linktime. This increases size of LTO object files, but enable
9103 diagnostics about One Definition Rule violations.
9105 @item -flto-compression-level=@var{n}
9106 This option specifies the level of compression used for intermediate
9107 language written to LTO object files, and is only meaningful in
9108 conjunction with LTO mode (@option{-flto}). Valid
9109 values are 0 (no compression) to 9 (maximum compression). Values
9110 outside this range are clamped to either 0 or 9. If the option is not
9111 given, a default balanced compression setting is used.
9114 Prints a report with internal details on the workings of the link-time
9115 optimizer. The contents of this report vary from version to version.
9116 It is meant to be useful to GCC developers when processing object
9117 files in LTO mode (via @option{-flto}).
9119 Disabled by default.
9121 @item -flto-report-wpa
9122 Like @option{-flto-report}, but only print for the WPA phase of Link
9125 @item -fuse-linker-plugin
9126 Enables the use of a linker plugin during link-time optimization. This
9127 option relies on plugin support in the linker, which is available in gold
9128 or in GNU ld 2.21 or newer.
9130 This option enables the extraction of object files with GIMPLE bytecode out
9131 of library archives. This improves the quality of optimization by exposing
9132 more code to the link-time optimizer. This information specifies what
9133 symbols can be accessed externally (by non-LTO object or during dynamic
9134 linking). Resulting code quality improvements on binaries (and shared
9135 libraries that use hidden visibility) are similar to @code{-fwhole-program}.
9136 See @option{-flto} for a description of the effect of this flag and how to
9139 This option is enabled by default when LTO support in GCC is enabled
9140 and GCC was configured for use with
9141 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9143 @item -ffat-lto-objects
9144 @opindex ffat-lto-objects
9145 Fat LTO objects are object files that contain both the intermediate language
9146 and the object code. This makes them usable for both LTO linking and normal
9147 linking. This option is effective only when compiling with @option{-flto}
9148 and is ignored at link time.
9150 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9151 requires the complete toolchain to be aware of LTO. It requires a linker with
9152 linker plugin support for basic functionality. Additionally,
9153 @command{nm}, @command{ar} and @command{ranlib}
9154 need to support linker plugins to allow a full-featured build environment
9155 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9156 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9157 to these tools. With non fat LTO makefiles need to be modified to use them.
9159 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9162 @item -fcompare-elim
9163 @opindex fcompare-elim
9164 After register allocation and post-register allocation instruction splitting,
9165 identify arithmetic instructions that compute processor flags similar to a
9166 comparison operation based on that arithmetic. If possible, eliminate the
9167 explicit comparison operation.
9169 This pass only applies to certain targets that cannot explicitly represent
9170 the comparison operation before register allocation is complete.
9172 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9175 @opindex fuse-ld=bfd
9176 Use the @command{bfd} linker instead of the default linker.
9179 @opindex fuse-ld=gold
9180 Use the @command{gold} linker instead of the default linker.
9182 @item -fcprop-registers
9183 @opindex fcprop-registers
9184 After register allocation and post-register allocation instruction splitting,
9185 perform a copy-propagation pass to try to reduce scheduling dependencies
9186 and occasionally eliminate the copy.
9188 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9190 @item -fprofile-correction
9191 @opindex fprofile-correction
9192 Profiles collected using an instrumented binary for multi-threaded programs may
9193 be inconsistent due to missed counter updates. When this option is specified,
9194 GCC uses heuristics to correct or smooth out such inconsistencies. By
9195 default, GCC emits an error message when an inconsistent profile is detected.
9197 @item -fprofile-dir=@var{path}
9198 @opindex fprofile-dir
9200 Set the directory to search for the profile data files in to @var{path}.
9201 This option affects only the profile data generated by
9202 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
9203 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
9204 and its related options. Both absolute and relative paths can be used.
9205 By default, GCC uses the current directory as @var{path}, thus the
9206 profile data file appears in the same directory as the object file.
9208 @item -fprofile-generate
9209 @itemx -fprofile-generate=@var{path}
9210 @opindex fprofile-generate
9212 Enable options usually used for instrumenting application to produce
9213 profile useful for later recompilation with profile feedback based
9214 optimization. You must use @option{-fprofile-generate} both when
9215 compiling and when linking your program.
9217 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
9219 If @var{path} is specified, GCC looks at the @var{path} to find
9220 the profile feedback data files. See @option{-fprofile-dir}.
9223 @itemx -fprofile-use=@var{path}
9224 @opindex fprofile-use
9225 Enable profile feedback directed optimizations, and optimizations
9226 generally profitable only with profile feedback available.
9228 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
9229 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}, @code{-ftree-vectorize},
9230 @code{ftree-loop-distribute-patterns}
9232 By default, GCC emits an error message if the feedback profiles do not
9233 match the source code. This error can be turned into a warning by using
9234 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9237 If @var{path} is specified, GCC looks at the @var{path} to find
9238 the profile feedback data files. See @option{-fprofile-dir}.
9240 @item -fauto-profile
9241 @itemx -fauto-profile=@var{path}
9242 @opindex fauto-profile
9243 Enable sampling based feedback directed optimizations, and optimizations
9244 generally profitable only with profile feedback available.
9246 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
9247 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}, @code{-ftree-vectorize},
9248 @code{-finline-functions}, @code{-fipa-cp}, @code{-fipa-cp-clone},
9249 @code{-fpredictive-commoning}, @code{-funswitch-loops},
9250 @code{-fgcse-after-reload}, @code{-ftree-loop-distribute-patterns},
9252 If @var{path} is specified, GCC looks at the @var{path} to find
9253 the profile feedback data files.
9255 In order to collect AutoFDO profile, you need to have:
9257 1. A linux system with linux perf support
9258 2. (optional) An Intel processor with last branch record (LBR) support. This is
9259 to guarantee accurate instruction level profile, which is important for
9260 AutoFDO performance.
9262 To collect the profile, first use linux perf to collect raw profile
9263 (see @uref{https://perf.wiki.kernel.org/}).
9266 @code{perf record -e br_inst_retired:near_taken -b -o perf.data -- your_program}
9268 Then use create_gcov tool, which takes raw profile and unstripped binary to
9269 generate AutoFDO profile that can be used by GCC.
9270 (see @uref{https://github.com/google/autofdo}).
9273 @code{create_gcov --binary=your_program.unstripped --profile=perf.data --gcov=profile.afdo}
9276 The following options control compiler behavior regarding floating-point
9277 arithmetic. These options trade off between speed and
9278 correctness. All must be specifically enabled.
9282 @opindex ffloat-store
9283 Do not store floating-point variables in registers, and inhibit other
9284 options that might change whether a floating-point value is taken from a
9287 @cindex floating-point precision
9288 This option prevents undesirable excess precision on machines such as
9289 the 68000 where the floating registers (of the 68881) keep more
9290 precision than a @code{double} is supposed to have. Similarly for the
9291 x86 architecture. For most programs, the excess precision does only
9292 good, but a few programs rely on the precise definition of IEEE floating
9293 point. Use @option{-ffloat-store} for such programs, after modifying
9294 them to store all pertinent intermediate computations into variables.
9296 @item -fexcess-precision=@var{style}
9297 @opindex fexcess-precision
9298 This option allows further control over excess precision on machines
9299 where floating-point registers have more precision than the IEEE
9300 @code{float} and @code{double} types and the processor does not
9301 support operations rounding to those types. By default,
9302 @option{-fexcess-precision=fast} is in effect; this means that
9303 operations are carried out in the precision of the registers and that
9304 it is unpredictable when rounding to the types specified in the source
9305 code takes place. When compiling C, if
9306 @option{-fexcess-precision=standard} is specified then excess
9307 precision follows the rules specified in ISO C99; in particular,
9308 both casts and assignments cause values to be rounded to their
9309 semantic types (whereas @option{-ffloat-store} only affects
9310 assignments). This option is enabled by default for C if a strict
9311 conformance option such as @option{-std=c99} is used.
9314 @option{-fexcess-precision=standard} is not implemented for languages
9315 other than C, and has no effect if
9316 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
9317 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
9318 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9319 semantics apply without excess precision, and in the latter, rounding
9324 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9325 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9326 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
9328 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9330 This option is not turned on by any @option{-O} option besides
9331 @option{-Ofast} since it can result in incorrect output for programs
9332 that depend on an exact implementation of IEEE or ISO rules/specifications
9333 for math functions. It may, however, yield faster code for programs
9334 that do not require the guarantees of these specifications.
9336 @item -fno-math-errno
9337 @opindex fno-math-errno
9338 Do not set @code{errno} after calling math functions that are executed
9339 with a single instruction, e.g., @code{sqrt}. A program that relies on
9340 IEEE exceptions for math error handling may want to use this flag
9341 for speed while maintaining IEEE arithmetic compatibility.
9343 This option is not turned on by any @option{-O} option since
9344 it can result in incorrect output for programs that depend on
9345 an exact implementation of IEEE or ISO rules/specifications for
9346 math functions. It may, however, yield faster code for programs
9347 that do not require the guarantees of these specifications.
9349 The default is @option{-fmath-errno}.
9351 On Darwin systems, the math library never sets @code{errno}. There is
9352 therefore no reason for the compiler to consider the possibility that
9353 it might, and @option{-fno-math-errno} is the default.
9355 @item -funsafe-math-optimizations
9356 @opindex funsafe-math-optimizations
9358 Allow optimizations for floating-point arithmetic that (a) assume
9359 that arguments and results are valid and (b) may violate IEEE or
9360 ANSI standards. When used at link-time, it may include libraries
9361 or startup files that change the default FPU control word or other
9362 similar optimizations.
9364 This option is not turned on by any @option{-O} option since
9365 it can result in incorrect output for programs that depend on
9366 an exact implementation of IEEE or ISO rules/specifications for
9367 math functions. It may, however, yield faster code for programs
9368 that do not require the guarantees of these specifications.
9369 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9370 @option{-fassociative-math} and @option{-freciprocal-math}.
9372 The default is @option{-fno-unsafe-math-optimizations}.
9374 @item -fassociative-math
9375 @opindex fassociative-math
9377 Allow re-association of operands in series of floating-point operations.
9378 This violates the ISO C and C++ language standard by possibly changing
9379 computation result. NOTE: re-ordering may change the sign of zero as
9380 well as ignore NaNs and inhibit or create underflow or overflow (and
9381 thus cannot be used on code that relies on rounding behavior like
9382 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9383 and thus may not be used when ordered comparisons are required.
9384 This option requires that both @option{-fno-signed-zeros} and
9385 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9386 much sense with @option{-frounding-math}. For Fortran the option
9387 is automatically enabled when both @option{-fno-signed-zeros} and
9388 @option{-fno-trapping-math} are in effect.
9390 The default is @option{-fno-associative-math}.
9392 @item -freciprocal-math
9393 @opindex freciprocal-math
9395 Allow the reciprocal of a value to be used instead of dividing by
9396 the value if this enables optimizations. For example @code{x / y}
9397 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9398 is subject to common subexpression elimination. Note that this loses
9399 precision and increases the number of flops operating on the value.
9401 The default is @option{-fno-reciprocal-math}.
9403 @item -ffinite-math-only
9404 @opindex ffinite-math-only
9405 Allow optimizations for floating-point arithmetic that assume
9406 that arguments and results are not NaNs or +-Infs.
9408 This option is not turned on by any @option{-O} option since
9409 it can result in incorrect output for programs that depend on
9410 an exact implementation of IEEE or ISO rules/specifications for
9411 math functions. It may, however, yield faster code for programs
9412 that do not require the guarantees of these specifications.
9414 The default is @option{-fno-finite-math-only}.
9416 @item -fno-signed-zeros
9417 @opindex fno-signed-zeros
9418 Allow optimizations for floating-point arithmetic that ignore the
9419 signedness of zero. IEEE arithmetic specifies the behavior of
9420 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9421 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9422 This option implies that the sign of a zero result isn't significant.
9424 The default is @option{-fsigned-zeros}.
9426 @item -fno-trapping-math
9427 @opindex fno-trapping-math
9428 Compile code assuming that floating-point operations cannot generate
9429 user-visible traps. These traps include division by zero, overflow,
9430 underflow, inexact result and invalid operation. This option requires
9431 that @option{-fno-signaling-nans} be in effect. Setting this option may
9432 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9434 This option should never be turned on by any @option{-O} option since
9435 it can result in incorrect output for programs that depend on
9436 an exact implementation of IEEE or ISO rules/specifications for
9439 The default is @option{-ftrapping-math}.
9441 @item -frounding-math
9442 @opindex frounding-math
9443 Disable transformations and optimizations that assume default floating-point
9444 rounding behavior. This is round-to-zero for all floating point
9445 to integer conversions, and round-to-nearest for all other arithmetic
9446 truncations. This option should be specified for programs that change
9447 the FP rounding mode dynamically, or that may be executed with a
9448 non-default rounding mode. This option disables constant folding of
9449 floating-point expressions at compile time (which may be affected by
9450 rounding mode) and arithmetic transformations that are unsafe in the
9451 presence of sign-dependent rounding modes.
9453 The default is @option{-fno-rounding-math}.
9455 This option is experimental and does not currently guarantee to
9456 disable all GCC optimizations that are affected by rounding mode.
9457 Future versions of GCC may provide finer control of this setting
9458 using C99's @code{FENV_ACCESS} pragma. This command-line option
9459 will be used to specify the default state for @code{FENV_ACCESS}.
9461 @item -fsignaling-nans
9462 @opindex fsignaling-nans
9463 Compile code assuming that IEEE signaling NaNs may generate user-visible
9464 traps during floating-point operations. Setting this option disables
9465 optimizations that may change the number of exceptions visible with
9466 signaling NaNs. This option implies @option{-ftrapping-math}.
9468 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9471 The default is @option{-fno-signaling-nans}.
9473 This option is experimental and does not currently guarantee to
9474 disable all GCC optimizations that affect signaling NaN behavior.
9476 @item -fsingle-precision-constant
9477 @opindex fsingle-precision-constant
9478 Treat floating-point constants as single precision instead of
9479 implicitly converting them to double-precision constants.
9481 @item -fcx-limited-range
9482 @opindex fcx-limited-range
9483 When enabled, this option states that a range reduction step is not
9484 needed when performing complex division. Also, there is no checking
9485 whether the result of a complex multiplication or division is @code{NaN
9486 + I*NaN}, with an attempt to rescue the situation in that case. The
9487 default is @option{-fno-cx-limited-range}, but is enabled by
9488 @option{-ffast-math}.
9490 This option controls the default setting of the ISO C99
9491 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9494 @item -fcx-fortran-rules
9495 @opindex fcx-fortran-rules
9496 Complex multiplication and division follow Fortran rules. Range
9497 reduction is done as part of complex division, but there is no checking
9498 whether the result of a complex multiplication or division is @code{NaN
9499 + I*NaN}, with an attempt to rescue the situation in that case.
9501 The default is @option{-fno-cx-fortran-rules}.
9505 The following options control optimizations that may improve
9506 performance, but are not enabled by any @option{-O} options. This
9507 section includes experimental options that may produce broken code.
9510 @item -fbranch-probabilities
9511 @opindex fbranch-probabilities
9512 After running a program compiled with @option{-fprofile-arcs}
9513 (@pxref{Debugging Options,, Options for Debugging Your Program or
9514 @command{gcc}}), you can compile it a second time using
9515 @option{-fbranch-probabilities}, to improve optimizations based on
9516 the number of times each branch was taken. When a program
9517 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9518 counts to a file called @file{@var{sourcename}.gcda} for each source
9519 file. The information in this data file is very dependent on the
9520 structure of the generated code, so you must use the same source code
9521 and the same optimization options for both compilations.
9523 With @option{-fbranch-probabilities}, GCC puts a
9524 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9525 These can be used to improve optimization. Currently, they are only
9526 used in one place: in @file{reorg.c}, instead of guessing which path a
9527 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9528 exactly determine which path is taken more often.
9530 @item -fprofile-values
9531 @opindex fprofile-values
9532 If combined with @option{-fprofile-arcs}, it adds code so that some
9533 data about values of expressions in the program is gathered.
9535 With @option{-fbranch-probabilities}, it reads back the data gathered
9536 from profiling values of expressions for usage in optimizations.
9538 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9540 @item -fprofile-reorder-functions
9541 @opindex fprofile-reorder-functions
9542 Function reordering based on profile instrumentation collects
9543 first time of execution of a function and orders these functions
9546 Enabled with @option{-fprofile-use}.
9550 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9551 to add code to gather information about values of expressions.
9553 With @option{-fbranch-probabilities}, it reads back the data gathered
9554 and actually performs the optimizations based on them.
9555 Currently the optimizations include specialization of division operations
9556 using the knowledge about the value of the denominator.
9558 @item -frename-registers
9559 @opindex frename-registers
9560 Attempt to avoid false dependencies in scheduled code by making use
9561 of registers left over after register allocation. This optimization
9562 most benefits processors with lots of registers. Depending on the
9563 debug information format adopted by the target, however, it can
9564 make debugging impossible, since variables no longer stay in
9565 a ``home register''.
9567 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
9571 Perform tail duplication to enlarge superblock size. This transformation
9572 simplifies the control flow of the function allowing other optimizations to do
9575 Enabled with @option{-fprofile-use}.
9577 @item -funroll-loops
9578 @opindex funroll-loops
9579 Unroll loops whose number of iterations can be determined at compile time or
9580 upon entry to the loop. @option{-funroll-loops} implies
9581 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
9582 It also turns on complete loop peeling (i.e.@: complete removal of loops with
9583 a small constant number of iterations). This option makes code larger, and may
9584 or may not make it run faster.
9586 Enabled with @option{-fprofile-use}.
9588 @item -funroll-all-loops
9589 @opindex funroll-all-loops
9590 Unroll all loops, even if their number of iterations is uncertain when
9591 the loop is entered. This usually makes programs run more slowly.
9592 @option{-funroll-all-loops} implies the same options as
9593 @option{-funroll-loops}.
9596 @opindex fpeel-loops
9597 Peels loops for which there is enough information that they do not
9598 roll much (from profile feedback). It also turns on complete loop peeling
9599 (i.e.@: complete removal of loops with small constant number of iterations).
9601 Enabled with @option{-fprofile-use}.
9603 @item -fmove-loop-invariants
9604 @opindex fmove-loop-invariants
9605 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
9606 at level @option{-O1}
9608 @item -funswitch-loops
9609 @opindex funswitch-loops
9610 Move branches with loop invariant conditions out of the loop, with duplicates
9611 of the loop on both branches (modified according to result of the condition).
9613 @item -ffunction-sections
9614 @itemx -fdata-sections
9615 @opindex ffunction-sections
9616 @opindex fdata-sections
9617 Place each function or data item into its own section in the output
9618 file if the target supports arbitrary sections. The name of the
9619 function or the name of the data item determines the section's name
9622 Use these options on systems where the linker can perform optimizations
9623 to improve locality of reference in the instruction space. Most systems
9624 using the ELF object format and SPARC processors running Solaris 2 have
9625 linkers with such optimizations. AIX may have these optimizations in
9628 Only use these options when there are significant benefits from doing
9629 so. When you specify these options, the assembler and linker
9630 create larger object and executable files and are also slower.
9631 You cannot use @code{gprof} on all systems if you
9632 specify this option, and you may have problems with debugging if
9633 you specify both this option and @option{-g}.
9635 @item -fbranch-target-load-optimize
9636 @opindex fbranch-target-load-optimize
9637 Perform branch target register load optimization before prologue / epilogue
9639 The use of target registers can typically be exposed only during reload,
9640 thus hoisting loads out of loops and doing inter-block scheduling needs
9641 a separate optimization pass.
9643 @item -fbranch-target-load-optimize2
9644 @opindex fbranch-target-load-optimize2
9645 Perform branch target register load optimization after prologue / epilogue
9648 @item -fbtr-bb-exclusive
9649 @opindex fbtr-bb-exclusive
9650 When performing branch target register load optimization, don't reuse
9651 branch target registers within any basic block.
9653 @item -fstack-protector
9654 @opindex fstack-protector
9655 Emit extra code to check for buffer overflows, such as stack smashing
9656 attacks. This is done by adding a guard variable to functions with
9657 vulnerable objects. This includes functions that call @code{alloca}, and
9658 functions with buffers larger than 8 bytes. The guards are initialized
9659 when a function is entered and then checked when the function exits.
9660 If a guard check fails, an error message is printed and the program exits.
9662 @item -fstack-protector-all
9663 @opindex fstack-protector-all
9664 Like @option{-fstack-protector} except that all functions are protected.
9666 @item -fstack-protector-strong
9667 @opindex fstack-protector-strong
9668 Like @option{-fstack-protector} but includes additional functions to
9669 be protected --- those that have local array definitions, or have
9670 references to local frame addresses.
9672 @item -fsection-anchors
9673 @opindex fsection-anchors
9674 Try to reduce the number of symbolic address calculations by using
9675 shared ``anchor'' symbols to address nearby objects. This transformation
9676 can help to reduce the number of GOT entries and GOT accesses on some
9679 For example, the implementation of the following function @code{foo}:
9683 int foo (void) @{ return a + b + c; @}
9687 usually calculates the addresses of all three variables, but if you
9688 compile it with @option{-fsection-anchors}, it accesses the variables
9689 from a common anchor point instead. The effect is similar to the
9690 following pseudocode (which isn't valid C):
9695 register int *xr = &x;
9696 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
9700 Not all targets support this option.
9702 @item --param @var{name}=@var{value}
9704 In some places, GCC uses various constants to control the amount of
9705 optimization that is done. For example, GCC does not inline functions
9706 that contain more than a certain number of instructions. You can
9707 control some of these constants on the command line using the
9708 @option{--param} option.
9710 The names of specific parameters, and the meaning of the values, are
9711 tied to the internals of the compiler, and are subject to change
9712 without notice in future releases.
9714 In each case, the @var{value} is an integer. The allowable choices for
9718 @item predictable-branch-outcome
9719 When branch is predicted to be taken with probability lower than this threshold
9720 (in percent), then it is considered well predictable. The default is 10.
9722 @item max-crossjump-edges
9723 The maximum number of incoming edges to consider for cross-jumping.
9724 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
9725 the number of edges incoming to each block. Increasing values mean
9726 more aggressive optimization, making the compilation time increase with
9727 probably small improvement in executable size.
9729 @item min-crossjump-insns
9730 The minimum number of instructions that must be matched at the end
9731 of two blocks before cross-jumping is performed on them. This
9732 value is ignored in the case where all instructions in the block being
9733 cross-jumped from are matched. The default value is 5.
9735 @item max-grow-copy-bb-insns
9736 The maximum code size expansion factor when copying basic blocks
9737 instead of jumping. The expansion is relative to a jump instruction.
9738 The default value is 8.
9740 @item max-goto-duplication-insns
9741 The maximum number of instructions to duplicate to a block that jumps
9742 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
9743 passes, GCC factors computed gotos early in the compilation process,
9744 and unfactors them as late as possible. Only computed jumps at the
9745 end of a basic blocks with no more than max-goto-duplication-insns are
9746 unfactored. The default value is 8.
9748 @item max-delay-slot-insn-search
9749 The maximum number of instructions to consider when looking for an
9750 instruction to fill a delay slot. If more than this arbitrary number of
9751 instructions are searched, the time savings from filling the delay slot
9752 are minimal, so stop searching. Increasing values mean more
9753 aggressive optimization, making the compilation time increase with probably
9754 small improvement in execution time.
9756 @item max-delay-slot-live-search
9757 When trying to fill delay slots, the maximum number of instructions to
9758 consider when searching for a block with valid live register
9759 information. Increasing this arbitrarily chosen value means more
9760 aggressive optimization, increasing the compilation time. This parameter
9761 should be removed when the delay slot code is rewritten to maintain the
9764 @item max-gcse-memory
9765 The approximate maximum amount of memory that can be allocated in
9766 order to perform the global common subexpression elimination
9767 optimization. If more memory than specified is required, the
9768 optimization is not done.
9770 @item max-gcse-insertion-ratio
9771 If the ratio of expression insertions to deletions is larger than this value
9772 for any expression, then RTL PRE inserts or removes the expression and thus
9773 leaves partially redundant computations in the instruction stream. The default value is 20.
9775 @item max-pending-list-length
9776 The maximum number of pending dependencies scheduling allows
9777 before flushing the current state and starting over. Large functions
9778 with few branches or calls can create excessively large lists which
9779 needlessly consume memory and resources.
9781 @item max-modulo-backtrack-attempts
9782 The maximum number of backtrack attempts the scheduler should make
9783 when modulo scheduling a loop. Larger values can exponentially increase
9786 @item max-inline-insns-single
9787 Several parameters control the tree inliner used in GCC@.
9788 This number sets the maximum number of instructions (counted in GCC's
9789 internal representation) in a single function that the tree inliner
9790 considers for inlining. This only affects functions declared
9791 inline and methods implemented in a class declaration (C++).
9792 The default value is 400.
9794 @item max-inline-insns-auto
9795 When you use @option{-finline-functions} (included in @option{-O3}),
9796 a lot of functions that would otherwise not be considered for inlining
9797 by the compiler are investigated. To those functions, a different
9798 (more restrictive) limit compared to functions declared inline can
9800 The default value is 40.
9802 @item inline-min-speedup
9803 When estimated performance improvement of caller + callee runtime exceeds this
9804 threshold (in precent), the function can be inlined regardless the limit on
9805 @option{--param max-inline-insns-single} and @option{--param
9806 max-inline-insns-auto}.
9808 @item large-function-insns
9809 The limit specifying really large functions. For functions larger than this
9810 limit after inlining, inlining is constrained by
9811 @option{--param large-function-growth}. This parameter is useful primarily
9812 to avoid extreme compilation time caused by non-linear algorithms used by the
9814 The default value is 2700.
9816 @item large-function-growth
9817 Specifies maximal growth of large function caused by inlining in percents.
9818 The default value is 100 which limits large function growth to 2.0 times
9821 @item large-unit-insns
9822 The limit specifying large translation unit. Growth caused by inlining of
9823 units larger than this limit is limited by @option{--param inline-unit-growth}.
9824 For small units this might be too tight.
9825 For example, consider a unit consisting of function A
9826 that is inline and B that just calls A three times. If B is small relative to
9827 A, the growth of unit is 300\% and yet such inlining is very sane. For very
9828 large units consisting of small inlineable functions, however, the overall unit
9829 growth limit is needed to avoid exponential explosion of code size. Thus for
9830 smaller units, the size is increased to @option{--param large-unit-insns}
9831 before applying @option{--param inline-unit-growth}. The default is 10000.
9833 @item inline-unit-growth
9834 Specifies maximal overall growth of the compilation unit caused by inlining.
9835 The default value is 30 which limits unit growth to 1.3 times the original
9836 size. Cold functions (either marked cold via an attribute or by profile
9837 feedback) are not accounted into the unit size.
9839 @item ipcp-unit-growth
9840 Specifies maximal overall growth of the compilation unit caused by
9841 interprocedural constant propagation. The default value is 10 which limits
9842 unit growth to 1.1 times the original size.
9844 @item large-stack-frame
9845 The limit specifying large stack frames. While inlining the algorithm is trying
9846 to not grow past this limit too much. The default value is 256 bytes.
9848 @item large-stack-frame-growth
9849 Specifies maximal growth of large stack frames caused by inlining in percents.
9850 The default value is 1000 which limits large stack frame growth to 11 times
9853 @item max-inline-insns-recursive
9854 @itemx max-inline-insns-recursive-auto
9855 Specifies the maximum number of instructions an out-of-line copy of a
9856 self-recursive inline
9857 function can grow into by performing recursive inlining.
9859 For functions declared inline, @option{--param max-inline-insns-recursive} is
9860 taken into account. For functions not declared inline, recursive inlining
9861 happens only when @option{-finline-functions} (included in @option{-O3}) is
9862 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
9863 default value is 450.
9865 @item max-inline-recursive-depth
9866 @itemx max-inline-recursive-depth-auto
9867 Specifies the maximum recursion depth used for recursive inlining.
9869 For functions declared inline, @option{--param max-inline-recursive-depth} is
9870 taken into account. For functions not declared inline, recursive inlining
9871 happens only when @option{-finline-functions} (included in @option{-O3}) is
9872 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
9875 @item min-inline-recursive-probability
9876 Recursive inlining is profitable only for function having deep recursion
9877 in average and can hurt for function having little recursion depth by
9878 increasing the prologue size or complexity of function body to other
9881 When profile feedback is available (see @option{-fprofile-generate}) the actual
9882 recursion depth can be guessed from probability that function recurses via a
9883 given call expression. This parameter limits inlining only to call expressions
9884 whose probability exceeds the given threshold (in percents).
9885 The default value is 10.
9887 @item early-inlining-insns
9888 Specify growth that the early inliner can make. In effect it increases
9889 the amount of inlining for code having a large abstraction penalty.
9890 The default value is 10.
9892 @item max-early-inliner-iterations
9893 @itemx max-early-inliner-iterations
9894 Limit of iterations of the early inliner. This basically bounds
9895 the number of nested indirect calls the early inliner can resolve.
9896 Deeper chains are still handled by late inlining.
9898 @item comdat-sharing-probability
9899 @itemx comdat-sharing-probability
9900 Probability (in percent) that C++ inline function with comdat visibility
9901 are shared across multiple compilation units. The default value is 20.
9903 @item profile-func-internal-id
9904 @itemx profile-func-internal-id
9905 A parameter to control whether to use function internal id in profile
9906 database lookup. If the value is 0, the compiler will use id that
9907 is based on function assembler name and filename, which makes old profile
9908 data more tolerant to source changes such as function reordering etc.
9909 The default value is 0.
9911 @item min-vect-loop-bound
9912 The minimum number of iterations under which loops are not vectorized
9913 when @option{-ftree-vectorize} is used. The number of iterations after
9914 vectorization needs to be greater than the value specified by this option
9915 to allow vectorization. The default value is 0.
9917 @item gcse-cost-distance-ratio
9918 Scaling factor in calculation of maximum distance an expression
9919 can be moved by GCSE optimizations. This is currently supported only in the
9920 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
9921 is with simple expressions, i.e., the expressions that have cost
9922 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
9923 hoisting of simple expressions. The default value is 10.
9925 @item gcse-unrestricted-cost
9926 Cost, roughly measured as the cost of a single typical machine
9927 instruction, at which GCSE optimizations do not constrain
9928 the distance an expression can travel. This is currently
9929 supported only in the code hoisting pass. The lesser the cost,
9930 the more aggressive code hoisting is. Specifying 0
9931 allows all expressions to travel unrestricted distances.
9932 The default value is 3.
9934 @item max-hoist-depth
9935 The depth of search in the dominator tree for expressions to hoist.
9936 This is used to avoid quadratic behavior in hoisting algorithm.
9937 The value of 0 does not limit on the search, but may slow down compilation
9938 of huge functions. The default value is 30.
9940 @item max-tail-merge-comparisons
9941 The maximum amount of similar bbs to compare a bb with. This is used to
9942 avoid quadratic behavior in tree tail merging. The default value is 10.
9944 @item max-tail-merge-iterations
9945 The maximum amount of iterations of the pass over the function. This is used to
9946 limit compilation time in tree tail merging. The default value is 2.
9948 @item max-unrolled-insns
9949 The maximum number of instructions that a loop may have to be unrolled.
9950 If a loop is unrolled, this parameter also determines how many times
9951 the loop code is unrolled.
9953 @item max-average-unrolled-insns
9954 The maximum number of instructions biased by probabilities of their execution
9955 that a loop may have to be unrolled. If a loop is unrolled,
9956 this parameter also determines how many times the loop code is unrolled.
9958 @item max-unroll-times
9959 The maximum number of unrollings of a single loop.
9961 @item max-peeled-insns
9962 The maximum number of instructions that a loop may have to be peeled.
9963 If a loop is peeled, this parameter also determines how many times
9964 the loop code is peeled.
9966 @item max-peel-times
9967 The maximum number of peelings of a single loop.
9969 @item max-peel-branches
9970 The maximum number of branches on the hot path through the peeled sequence.
9972 @item max-completely-peeled-insns
9973 The maximum number of insns of a completely peeled loop.
9975 @item max-completely-peel-times
9976 The maximum number of iterations of a loop to be suitable for complete peeling.
9978 @item max-completely-peel-loop-nest-depth
9979 The maximum depth of a loop nest suitable for complete peeling.
9981 @item max-unswitch-insns
9982 The maximum number of insns of an unswitched loop.
9984 @item max-unswitch-level
9985 The maximum number of branches unswitched in a single loop.
9988 The minimum cost of an expensive expression in the loop invariant motion.
9990 @item iv-consider-all-candidates-bound
9991 Bound on number of candidates for induction variables, below which
9992 all candidates are considered for each use in induction variable
9993 optimizations. If there are more candidates than this,
9994 only the most relevant ones are considered to avoid quadratic time complexity.
9996 @item iv-max-considered-uses
9997 The induction variable optimizations give up on loops that contain more
9998 induction variable uses.
10000 @item iv-always-prune-cand-set-bound
10001 If the number of candidates in the set is smaller than this value,
10002 always try to remove unnecessary ivs from the set
10003 when adding a new one.
10005 @item scev-max-expr-size
10006 Bound on size of expressions used in the scalar evolutions analyzer.
10007 Large expressions slow the analyzer.
10009 @item scev-max-expr-complexity
10010 Bound on the complexity of the expressions in the scalar evolutions analyzer.
10011 Complex expressions slow the analyzer.
10013 @item omega-max-vars
10014 The maximum number of variables in an Omega constraint system.
10015 The default value is 128.
10017 @item omega-max-geqs
10018 The maximum number of inequalities in an Omega constraint system.
10019 The default value is 256.
10021 @item omega-max-eqs
10022 The maximum number of equalities in an Omega constraint system.
10023 The default value is 128.
10025 @item omega-max-wild-cards
10026 The maximum number of wildcard variables that the Omega solver is
10027 able to insert. The default value is 18.
10029 @item omega-hash-table-size
10030 The size of the hash table in the Omega solver. The default value is
10033 @item omega-max-keys
10034 The maximal number of keys used by the Omega solver. The default
10037 @item omega-eliminate-redundant-constraints
10038 When set to 1, use expensive methods to eliminate all redundant
10039 constraints. The default value is 0.
10041 @item vect-max-version-for-alignment-checks
10042 The maximum number of run-time checks that can be performed when
10043 doing loop versioning for alignment in the vectorizer.
10045 @item vect-max-version-for-alias-checks
10046 The maximum number of run-time checks that can be performed when
10047 doing loop versioning for alias in the vectorizer.
10049 @item vect-max-peeling-for-alignment
10050 The maximum number of loop peels to enhance access alignment
10051 for vectorizer. Value -1 means 'no limit'.
10053 @item max-iterations-to-track
10054 The maximum number of iterations of a loop the brute-force algorithm
10055 for analysis of the number of iterations of the loop tries to evaluate.
10057 @item hot-bb-count-ws-permille
10058 A basic block profile count is considered hot if it contributes to
10059 the given permillage (i.e. 0...1000) of the entire profiled execution.
10061 @item hot-bb-frequency-fraction
10062 Select fraction of the entry block frequency of executions of basic block in
10063 function given basic block needs to have to be considered hot.
10065 @item max-predicted-iterations
10066 The maximum number of loop iterations we predict statically. This is useful
10067 in cases where a function contains a single loop with known bound and
10068 another loop with unknown bound.
10069 The known number of iterations is predicted correctly, while
10070 the unknown number of iterations average to roughly 10. This means that the
10071 loop without bounds appears artificially cold relative to the other one.
10073 @item builtin-expect-probability
10074 Control the probability of the expression having the specified value. This
10075 parameter takes a percentage (i.e. 0 ... 100) as input.
10076 The default probability of 90 is obtained empirically.
10078 @item align-threshold
10080 Select fraction of the maximal frequency of executions of a basic block in
10081 a function to align the basic block.
10083 @item align-loop-iterations
10085 A loop expected to iterate at least the selected number of iterations is
10088 @item tracer-dynamic-coverage
10089 @itemx tracer-dynamic-coverage-feedback
10091 This value is used to limit superblock formation once the given percentage of
10092 executed instructions is covered. This limits unnecessary code size
10095 The @option{tracer-dynamic-coverage-feedback} is used only when profile
10096 feedback is available. The real profiles (as opposed to statically estimated
10097 ones) are much less balanced allowing the threshold to be larger value.
10099 @item tracer-max-code-growth
10100 Stop tail duplication once code growth has reached given percentage. This is
10101 a rather artificial limit, as most of the duplicates are eliminated later in
10102 cross jumping, so it may be set to much higher values than is the desired code
10105 @item tracer-min-branch-ratio
10107 Stop reverse growth when the reverse probability of best edge is less than this
10108 threshold (in percent).
10110 @item tracer-min-branch-ratio
10111 @itemx tracer-min-branch-ratio-feedback
10113 Stop forward growth if the best edge has probability lower than this
10116 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
10117 compilation for profile feedback and one for compilation without. The value
10118 for compilation with profile feedback needs to be more conservative (higher) in
10119 order to make tracer effective.
10121 @item max-cse-path-length
10123 The maximum number of basic blocks on path that CSE considers.
10126 @item max-cse-insns
10127 The maximum number of instructions CSE processes before flushing.
10128 The default is 1000.
10130 @item ggc-min-expand
10132 GCC uses a garbage collector to manage its own memory allocation. This
10133 parameter specifies the minimum percentage by which the garbage
10134 collector's heap should be allowed to expand between collections.
10135 Tuning this may improve compilation speed; it has no effect on code
10138 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10139 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10140 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10141 GCC is not able to calculate RAM on a particular platform, the lower
10142 bound of 30% is used. Setting this parameter and
10143 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10144 every opportunity. This is extremely slow, but can be useful for
10147 @item ggc-min-heapsize
10149 Minimum size of the garbage collector's heap before it begins bothering
10150 to collect garbage. The first collection occurs after the heap expands
10151 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10152 tuning this may improve compilation speed, and has no effect on code
10155 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10156 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10157 with a lower bound of 4096 (four megabytes) and an upper bound of
10158 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10159 particular platform, the lower bound is used. Setting this parameter
10160 very large effectively disables garbage collection. Setting this
10161 parameter and @option{ggc-min-expand} to zero causes a full collection
10162 to occur at every opportunity.
10164 @item max-reload-search-insns
10165 The maximum number of instruction reload should look backward for equivalent
10166 register. Increasing values mean more aggressive optimization, making the
10167 compilation time increase with probably slightly better performance.
10168 The default value is 100.
10170 @item max-cselib-memory-locations
10171 The maximum number of memory locations cselib should take into account.
10172 Increasing values mean more aggressive optimization, making the compilation time
10173 increase with probably slightly better performance. The default value is 500.
10175 @item reorder-blocks-duplicate
10176 @itemx reorder-blocks-duplicate-feedback
10178 Used by the basic block reordering pass to decide whether to use unconditional
10179 branch or duplicate the code on its destination. Code is duplicated when its
10180 estimated size is smaller than this value multiplied by the estimated size of
10181 unconditional jump in the hot spots of the program.
10183 The @option{reorder-block-duplicate-feedback} is used only when profile
10184 feedback is available. It may be set to higher values than
10185 @option{reorder-block-duplicate} since information about the hot spots is more
10188 @item max-sched-ready-insns
10189 The maximum number of instructions ready to be issued the scheduler should
10190 consider at any given time during the first scheduling pass. Increasing
10191 values mean more thorough searches, making the compilation time increase
10192 with probably little benefit. The default value is 100.
10194 @item max-sched-region-blocks
10195 The maximum number of blocks in a region to be considered for
10196 interblock scheduling. The default value is 10.
10198 @item max-pipeline-region-blocks
10199 The maximum number of blocks in a region to be considered for
10200 pipelining in the selective scheduler. The default value is 15.
10202 @item max-sched-region-insns
10203 The maximum number of insns in a region to be considered for
10204 interblock scheduling. The default value is 100.
10206 @item max-pipeline-region-insns
10207 The maximum number of insns in a region to be considered for
10208 pipelining in the selective scheduler. The default value is 200.
10210 @item min-spec-prob
10211 The minimum probability (in percents) of reaching a source block
10212 for interblock speculative scheduling. The default value is 40.
10214 @item max-sched-extend-regions-iters
10215 The maximum number of iterations through CFG to extend regions.
10216 A value of 0 (the default) disables region extensions.
10218 @item max-sched-insn-conflict-delay
10219 The maximum conflict delay for an insn to be considered for speculative motion.
10220 The default value is 3.
10222 @item sched-spec-prob-cutoff
10223 The minimal probability of speculation success (in percents), so that
10224 speculative insns are scheduled.
10225 The default value is 40.
10227 @item sched-spec-state-edge-prob-cutoff
10228 The minimum probability an edge must have for the scheduler to save its
10230 The default value is 10.
10232 @item sched-mem-true-dep-cost
10233 Minimal distance (in CPU cycles) between store and load targeting same
10234 memory locations. The default value is 1.
10236 @item selsched-max-lookahead
10237 The maximum size of the lookahead window of selective scheduling. It is a
10238 depth of search for available instructions.
10239 The default value is 50.
10241 @item selsched-max-sched-times
10242 The maximum number of times that an instruction is scheduled during
10243 selective scheduling. This is the limit on the number of iterations
10244 through which the instruction may be pipelined. The default value is 2.
10246 @item selsched-max-insns-to-rename
10247 The maximum number of best instructions in the ready list that are considered
10248 for renaming in the selective scheduler. The default value is 2.
10251 The minimum value of stage count that swing modulo scheduler
10252 generates. The default value is 2.
10254 @item max-last-value-rtl
10255 The maximum size measured as number of RTLs that can be recorded in an expression
10256 in combiner for a pseudo register as last known value of that register. The default
10259 @item max-combine-insns
10260 The maximum number of instructions the RTL combiner tries to combine.
10261 The default value is 2 at @option{-Og} and 4 otherwise.
10263 @item integer-share-limit
10264 Small integer constants can use a shared data structure, reducing the
10265 compiler's memory usage and increasing its speed. This sets the maximum
10266 value of a shared integer constant. The default value is 256.
10268 @item ssp-buffer-size
10269 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10270 protection when @option{-fstack-protection} is used.
10272 @item min-size-for-stack-sharing
10273 The minimum size of variables taking part in stack slot sharing when not
10274 optimizing. The default value is 32.
10276 @item max-jump-thread-duplication-stmts
10277 Maximum number of statements allowed in a block that needs to be
10278 duplicated when threading jumps.
10280 @item max-fields-for-field-sensitive
10281 Maximum number of fields in a structure treated in
10282 a field sensitive manner during pointer analysis. The default is zero
10283 for @option{-O0} and @option{-O1},
10284 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10286 @item prefetch-latency
10287 Estimate on average number of instructions that are executed before
10288 prefetch finishes. The distance prefetched ahead is proportional
10289 to this constant. Increasing this number may also lead to less
10290 streams being prefetched (see @option{simultaneous-prefetches}).
10292 @item simultaneous-prefetches
10293 Maximum number of prefetches that can run at the same time.
10295 @item l1-cache-line-size
10296 The size of cache line in L1 cache, in bytes.
10298 @item l1-cache-size
10299 The size of L1 cache, in kilobytes.
10301 @item l2-cache-size
10302 The size of L2 cache, in kilobytes.
10304 @item min-insn-to-prefetch-ratio
10305 The minimum ratio between the number of instructions and the
10306 number of prefetches to enable prefetching in a loop.
10308 @item prefetch-min-insn-to-mem-ratio
10309 The minimum ratio between the number of instructions and the
10310 number of memory references to enable prefetching in a loop.
10312 @item use-canonical-types
10313 Whether the compiler should use the ``canonical'' type system. By
10314 default, this should always be 1, which uses a more efficient internal
10315 mechanism for comparing types in C++ and Objective-C++. However, if
10316 bugs in the canonical type system are causing compilation failures,
10317 set this value to 0 to disable canonical types.
10319 @item switch-conversion-max-branch-ratio
10320 Switch initialization conversion refuses to create arrays that are
10321 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10322 branches in the switch.
10324 @item max-partial-antic-length
10325 Maximum length of the partial antic set computed during the tree
10326 partial redundancy elimination optimization (@option{-ftree-pre}) when
10327 optimizing at @option{-O3} and above. For some sorts of source code
10328 the enhanced partial redundancy elimination optimization can run away,
10329 consuming all of the memory available on the host machine. This
10330 parameter sets a limit on the length of the sets that are computed,
10331 which prevents the runaway behavior. Setting a value of 0 for
10332 this parameter allows an unlimited set length.
10334 @item sccvn-max-scc-size
10335 Maximum size of a strongly connected component (SCC) during SCCVN
10336 processing. If this limit is hit, SCCVN processing for the whole
10337 function is not done and optimizations depending on it are
10338 disabled. The default maximum SCC size is 10000.
10340 @item sccvn-max-alias-queries-per-access
10341 Maximum number of alias-oracle queries we perform when looking for
10342 redundancies for loads and stores. If this limit is hit the search
10343 is aborted and the load or store is not considered redundant. The
10344 number of queries is algorithmically limited to the number of
10345 stores on all paths from the load to the function entry.
10346 The default maxmimum number of queries is 1000.
10348 @item ira-max-loops-num
10349 IRA uses regional register allocation by default. If a function
10350 contains more loops than the number given by this parameter, only at most
10351 the given number of the most frequently-executed loops form regions
10352 for regional register allocation. The default value of the
10355 @item ira-max-conflict-table-size
10356 Although IRA uses a sophisticated algorithm to compress the conflict
10357 table, the table can still require excessive amounts of memory for
10358 huge functions. If the conflict table for a function could be more
10359 than the size in MB given by this parameter, the register allocator
10360 instead uses a faster, simpler, and lower-quality
10361 algorithm that does not require building a pseudo-register conflict table.
10362 The default value of the parameter is 2000.
10364 @item ira-loop-reserved-regs
10365 IRA can be used to evaluate more accurate register pressure in loops
10366 for decisions to move loop invariants (see @option{-O3}). The number
10367 of available registers reserved for some other purposes is given
10368 by this parameter. The default value of the parameter is 2, which is
10369 the minimal number of registers needed by typical instructions.
10370 This value is the best found from numerous experiments.
10372 @item loop-invariant-max-bbs-in-loop
10373 Loop invariant motion can be very expensive, both in compilation time and
10374 in amount of needed compile-time memory, with very large loops. Loops
10375 with more basic blocks than this parameter won't have loop invariant
10376 motion optimization performed on them. The default value of the
10377 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10379 @item loop-max-datarefs-for-datadeps
10380 Building data dapendencies is expensive for very large loops. This
10381 parameter limits the number of data references in loops that are
10382 considered for data dependence analysis. These large loops are no
10383 handled by the optimizations using loop data dependencies.
10384 The default value is 1000.
10386 @item max-vartrack-size
10387 Sets a maximum number of hash table slots to use during variable
10388 tracking dataflow analysis of any function. If this limit is exceeded
10389 with variable tracking at assignments enabled, analysis for that
10390 function is retried without it, after removing all debug insns from
10391 the function. If the limit is exceeded even without debug insns, var
10392 tracking analysis is completely disabled for the function. Setting
10393 the parameter to zero makes it unlimited.
10395 @item max-vartrack-expr-depth
10396 Sets a maximum number of recursion levels when attempting to map
10397 variable names or debug temporaries to value expressions. This trades
10398 compilation time for more complete debug information. If this is set too
10399 low, value expressions that are available and could be represented in
10400 debug information may end up not being used; setting this higher may
10401 enable the compiler to find more complex debug expressions, but compile
10402 time and memory use may grow. The default is 12.
10404 @item min-nondebug-insn-uid
10405 Use uids starting at this parameter for nondebug insns. The range below
10406 the parameter is reserved exclusively for debug insns created by
10407 @option{-fvar-tracking-assignments}, but debug insns may get
10408 (non-overlapping) uids above it if the reserved range is exhausted.
10410 @item ipa-sra-ptr-growth-factor
10411 IPA-SRA replaces a pointer to an aggregate with one or more new
10412 parameters only when their cumulative size is less or equal to
10413 @option{ipa-sra-ptr-growth-factor} times the size of the original
10416 @item sra-max-scalarization-size-Ospeed
10417 @item sra-max-scalarization-size-Osize
10418 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10419 replace scalar parts of aggregates with uses of independent scalar
10420 variables. These parameters control the maximum size, in storage units,
10421 of aggregate which will be considered for replacement when compiling for
10423 (@option{sra-max-scalarization-size-Ospeed}) or size
10424 (@option{sra-max-scalarization-size-Osize}) respectively.
10426 @item tm-max-aggregate-size
10427 When making copies of thread-local variables in a transaction, this
10428 parameter specifies the size in bytes after which variables are
10429 saved with the logging functions as opposed to save/restore code
10430 sequence pairs. This option only applies when using
10433 @item graphite-max-nb-scop-params
10434 To avoid exponential effects in the Graphite loop transforms, the
10435 number of parameters in a Static Control Part (SCoP) is bounded. The
10436 default value is 10 parameters. A variable whose value is unknown at
10437 compilation time and defined outside a SCoP is a parameter of the SCoP.
10439 @item graphite-max-bbs-per-function
10440 To avoid exponential effects in the detection of SCoPs, the size of
10441 the functions analyzed by Graphite is bounded. The default value is
10444 @item loop-block-tile-size
10445 Loop blocking or strip mining transforms, enabled with
10446 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10447 loop in the loop nest by a given number of iterations. The strip
10448 length can be changed using the @option{loop-block-tile-size}
10449 parameter. The default value is 51 iterations.
10451 @item ipa-cp-value-list-size
10452 IPA-CP attempts to track all possible values and types passed to a function's
10453 parameter in order to propagate them and perform devirtualization.
10454 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10455 stores per one formal parameter of a function.
10457 @item ipa-cp-eval-threshold
10458 IPA-CP calculates its own score of cloning profitability heuristics
10459 and performs those cloning opportunities with scores that exceed
10460 @option{ipa-cp-eval-threshold}.
10462 @item ipa-max-agg-items
10463 IPA-CP is also capable to propagate a number of scalar values passed
10464 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10465 number of such values per one parameter.
10467 @item ipa-cp-loop-hint-bonus
10468 When IPA-CP determines that a cloning candidate would make the number
10469 of iterations of a loop known, it adds a bonus of
10470 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10473 @item ipa-cp-array-index-hint-bonus
10474 When IPA-CP determines that a cloning candidate would make the index of
10475 an array access known, it adds a bonus of
10476 @option{ipa-cp-array-index-hint-bonus} to the profitability
10477 score of the candidate.
10479 @item ipa-max-aa-steps
10480 During its analysis of function bodies, IPA-CP employs alias analysis
10481 in order to track values pointed to by function parameters. In order
10482 not spend too much time analyzing huge functions, it will give up and
10483 consider all memory clobbered after examining
10484 @option{ipa-max-aa-steps} statements modifying memory.
10486 @item lto-partitions
10487 Specify desired number of partitions produced during WHOPR compilation.
10488 The number of partitions should exceed the number of CPUs used for compilation.
10489 The default value is 32.
10491 @item lto-minpartition
10492 Size of minimal partition for WHOPR (in estimated instructions).
10493 This prevents expenses of splitting very small programs into too many
10496 @item cxx-max-namespaces-for-diagnostic-help
10497 The maximum number of namespaces to consult for suggestions when C++
10498 name lookup fails for an identifier. The default is 1000.
10500 @item sink-frequency-threshold
10501 The maximum relative execution frequency (in percents) of the target block
10502 relative to a statement's original block to allow statement sinking of a
10503 statement. Larger numbers result in more aggressive statement sinking.
10504 The default value is 75. A small positive adjustment is applied for
10505 statements with memory operands as those are even more profitable so sink.
10507 @item max-stores-to-sink
10508 The maximum number of conditional stores paires that can be sunk. Set to 0
10509 if either vectorization (@option{-ftree-vectorize}) or if-conversion
10510 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
10512 @item allow-store-data-races
10513 Allow optimizers to introduce new data races on stores.
10514 Set to 1 to allow, otherwise to 0. This option is enabled by default
10515 at optimization level @option{-Ofast}.
10517 @item case-values-threshold
10518 The smallest number of different values for which it is best to use a
10519 jump-table instead of a tree of conditional branches. If the value is
10520 0, use the default for the machine. The default is 0.
10522 @item tree-reassoc-width
10523 Set the maximum number of instructions executed in parallel in
10524 reassociated tree. This parameter overrides target dependent
10525 heuristics used by default if has non zero value.
10527 @item sched-pressure-algorithm
10528 Choose between the two available implementations of
10529 @option{-fsched-pressure}. Algorithm 1 is the original implementation
10530 and is the more likely to prevent instructions from being reordered.
10531 Algorithm 2 was designed to be a compromise between the relatively
10532 conservative approach taken by algorithm 1 and the rather aggressive
10533 approach taken by the default scheduler. It relies more heavily on
10534 having a regular register file and accurate register pressure classes.
10535 See @file{haifa-sched.c} in the GCC sources for more details.
10537 The default choice depends on the target.
10539 @item max-slsr-cand-scan
10540 Set the maximum number of existing candidates that will be considered when
10541 seeking a basis for a new straight-line strength reduction candidate.
10544 Enable buffer overflow detection for global objects. This kind
10545 of protection is enabled by default if you are using
10546 @option{-fsanitize=address} option.
10547 To disable global objects protection use @option{--param asan-globals=0}.
10550 Enable buffer overflow detection for stack objects. This kind of
10551 protection is enabled by default when using@option{-fsanitize=address}.
10552 To disable stack protection use @option{--param asan-stack=0} option.
10554 @item asan-instrument-reads
10555 Enable buffer overflow detection for memory reads. This kind of
10556 protection is enabled by default when using @option{-fsanitize=address}.
10557 To disable memory reads protection use
10558 @option{--param asan-instrument-reads=0}.
10560 @item asan-instrument-writes
10561 Enable buffer overflow detection for memory writes. This kind of
10562 protection is enabled by default when using @option{-fsanitize=address}.
10563 To disable memory writes protection use
10564 @option{--param asan-instrument-writes=0} option.
10566 @item asan-memintrin
10567 Enable detection for built-in functions. This kind of protection
10568 is enabled by default when using @option{-fsanitize=address}.
10569 To disable built-in functions protection use
10570 @option{--param asan-memintrin=0}.
10572 @item asan-use-after-return
10573 Enable detection of use-after-return. This kind of protection
10574 is enabled by default when using @option{-fsanitize=address} option.
10575 To disable use-after-return detection use
10576 @option{--param asan-use-after-return=0}.
10578 @item asan-instrumentation-with-call-threshold
10579 If number of memory accesses in function being instrumented
10580 is greater or equal to this number, use callbacks instead of inline checks.
10581 E.g. to disable inline code use
10582 @option{--param asan-instrumentation-with-call-threshold=0}.
10584 @item chkp-max-ctor-size
10585 Static constructors generated by Pointer Bounds Checker may become very
10586 large and significantly increase compile time at optimization level
10587 @option{-O1} and higher. This parameter is a maximum nubmer of statements
10588 in a single generated constructor. Default value is 5000.
10593 @node Preprocessor Options
10594 @section Options Controlling the Preprocessor
10595 @cindex preprocessor options
10596 @cindex options, preprocessor
10598 These options control the C preprocessor, which is run on each C source
10599 file before actual compilation.
10601 If you use the @option{-E} option, nothing is done except preprocessing.
10602 Some of these options make sense only together with @option{-E} because
10603 they cause the preprocessor output to be unsuitable for actual
10607 @item -Wp,@var{option}
10609 You can use @option{-Wp,@var{option}} to bypass the compiler driver
10610 and pass @var{option} directly through to the preprocessor. If
10611 @var{option} contains commas, it is split into multiple options at the
10612 commas. However, many options are modified, translated or interpreted
10613 by the compiler driver before being passed to the preprocessor, and
10614 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
10615 interface is undocumented and subject to change, so whenever possible
10616 you should avoid using @option{-Wp} and let the driver handle the
10619 @item -Xpreprocessor @var{option}
10620 @opindex Xpreprocessor
10621 Pass @var{option} as an option to the preprocessor. You can use this to
10622 supply system-specific preprocessor options that GCC does not
10625 If you want to pass an option that takes an argument, you must use
10626 @option{-Xpreprocessor} twice, once for the option and once for the argument.
10628 @item -no-integrated-cpp
10629 @opindex no-integrated-cpp
10630 Perform preprocessing as a separate pass before compilation.
10631 By default, GCC performs preprocessing as an integrated part of
10632 input tokenization and parsing.
10633 If this option is provided, the appropriate language front end
10634 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
10635 and Objective-C, respectively) is instead invoked twice,
10636 once for preprocessing only and once for actual compilation
10637 of the preprocessed input.
10638 This option may be useful in conjunction with the @option{-B} or
10639 @option{-wrapper} options to specify an alternate preprocessor or
10640 perform additional processing of the program source between
10641 normal preprocessing and compilation.
10644 @include cppopts.texi
10646 @node Assembler Options
10647 @section Passing Options to the Assembler
10649 @c prevent bad page break with this line
10650 You can pass options to the assembler.
10653 @item -Wa,@var{option}
10655 Pass @var{option} as an option to the assembler. If @var{option}
10656 contains commas, it is split into multiple options at the commas.
10658 @item -Xassembler @var{option}
10659 @opindex Xassembler
10660 Pass @var{option} as an option to the assembler. You can use this to
10661 supply system-specific assembler options that GCC does not
10664 If you want to pass an option that takes an argument, you must use
10665 @option{-Xassembler} twice, once for the option and once for the argument.
10670 @section Options for Linking
10671 @cindex link options
10672 @cindex options, linking
10674 These options come into play when the compiler links object files into
10675 an executable output file. They are meaningless if the compiler is
10676 not doing a link step.
10680 @item @var{object-file-name}
10681 A file name that does not end in a special recognized suffix is
10682 considered to name an object file or library. (Object files are
10683 distinguished from libraries by the linker according to the file
10684 contents.) If linking is done, these object files are used as input
10693 If any of these options is used, then the linker is not run, and
10694 object file names should not be used as arguments. @xref{Overall
10698 @item -l@var{library}
10699 @itemx -l @var{library}
10701 Search the library named @var{library} when linking. (The second
10702 alternative with the library as a separate argument is only for
10703 POSIX compliance and is not recommended.)
10705 It makes a difference where in the command you write this option; the
10706 linker searches and processes libraries and object files in the order they
10707 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
10708 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
10709 to functions in @samp{z}, those functions may not be loaded.
10711 The linker searches a standard list of directories for the library,
10712 which is actually a file named @file{lib@var{library}.a}. The linker
10713 then uses this file as if it had been specified precisely by name.
10715 The directories searched include several standard system directories
10716 plus any that you specify with @option{-L}.
10718 Normally the files found this way are library files---archive files
10719 whose members are object files. The linker handles an archive file by
10720 scanning through it for members which define symbols that have so far
10721 been referenced but not defined. But if the file that is found is an
10722 ordinary object file, it is linked in the usual fashion. The only
10723 difference between using an @option{-l} option and specifying a file name
10724 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
10725 and searches several directories.
10729 You need this special case of the @option{-l} option in order to
10730 link an Objective-C or Objective-C++ program.
10732 @item -nostartfiles
10733 @opindex nostartfiles
10734 Do not use the standard system startup files when linking.
10735 The standard system libraries are used normally, unless @option{-nostdlib}
10736 or @option{-nodefaultlibs} is used.
10738 @item -nodefaultlibs
10739 @opindex nodefaultlibs
10740 Do not use the standard system libraries when linking.
10741 Only the libraries you specify are passed to the linker, and options
10742 specifying linkage of the system libraries, such as @code{-static-libgcc}
10743 or @code{-shared-libgcc}, are ignored.
10744 The standard startup files are used normally, unless @option{-nostartfiles}
10747 The compiler may generate calls to @code{memcmp},
10748 @code{memset}, @code{memcpy} and @code{memmove}.
10749 These entries are usually resolved by entries in
10750 libc. These entry points should be supplied through some other
10751 mechanism when this option is specified.
10755 Do not use the standard system startup files or libraries when linking.
10756 No startup files and only the libraries you specify are passed to
10757 the linker, and options specifying linkage of the system libraries, such as
10758 @code{-static-libgcc} or @code{-shared-libgcc}, are ignored.
10760 The compiler may generate calls to @code{memcmp}, @code{memset},
10761 @code{memcpy} and @code{memmove}.
10762 These entries are usually resolved by entries in
10763 libc. These entry points should be supplied through some other
10764 mechanism when this option is specified.
10766 @cindex @option{-lgcc}, use with @option{-nostdlib}
10767 @cindex @option{-nostdlib} and unresolved references
10768 @cindex unresolved references and @option{-nostdlib}
10769 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
10770 @cindex @option{-nodefaultlibs} and unresolved references
10771 @cindex unresolved references and @option{-nodefaultlibs}
10772 One of the standard libraries bypassed by @option{-nostdlib} and
10773 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
10774 which GCC uses to overcome shortcomings of particular machines, or special
10775 needs for some languages.
10776 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
10777 Collection (GCC) Internals},
10778 for more discussion of @file{libgcc.a}.)
10779 In most cases, you need @file{libgcc.a} even when you want to avoid
10780 other standard libraries. In other words, when you specify @option{-nostdlib}
10781 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
10782 This ensures that you have no unresolved references to internal GCC
10783 library subroutines.
10784 (An example of such an internal subroutine is @samp{__main}, used to ensure C++
10785 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
10786 GNU Compiler Collection (GCC) Internals}.)
10790 Produce a position independent executable on targets that support it.
10791 For predictable results, you must also specify the same set of options
10792 used for compilation (@option{-fpie}, @option{-fPIE},
10793 or model suboptions) when you specify this linker option.
10797 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
10798 that support it. This instructs the linker to add all symbols, not
10799 only used ones, to the dynamic symbol table. This option is needed
10800 for some uses of @code{dlopen} or to allow obtaining backtraces
10801 from within a program.
10805 Remove all symbol table and relocation information from the executable.
10809 On systems that support dynamic linking, this prevents linking with the shared
10810 libraries. On other systems, this option has no effect.
10814 Produce a shared object which can then be linked with other objects to
10815 form an executable. Not all systems support this option. For predictable
10816 results, you must also specify the same set of options used for compilation
10817 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
10818 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
10819 needs to build supplementary stub code for constructors to work. On
10820 multi-libbed systems, @samp{gcc -shared} must select the correct support
10821 libraries to link against. Failing to supply the correct flags may lead
10822 to subtle defects. Supplying them in cases where they are not necessary
10825 @item -shared-libgcc
10826 @itemx -static-libgcc
10827 @opindex shared-libgcc
10828 @opindex static-libgcc
10829 On systems that provide @file{libgcc} as a shared library, these options
10830 force the use of either the shared or static version, respectively.
10831 If no shared version of @file{libgcc} was built when the compiler was
10832 configured, these options have no effect.
10834 There are several situations in which an application should use the
10835 shared @file{libgcc} instead of the static version. The most common
10836 of these is when the application wishes to throw and catch exceptions
10837 across different shared libraries. In that case, each of the libraries
10838 as well as the application itself should use the shared @file{libgcc}.
10840 Therefore, the G++ and GCJ drivers automatically add
10841 @option{-shared-libgcc} whenever you build a shared library or a main
10842 executable, because C++ and Java programs typically use exceptions, so
10843 this is the right thing to do.
10845 If, instead, you use the GCC driver to create shared libraries, you may
10846 find that they are not always linked with the shared @file{libgcc}.
10847 If GCC finds, at its configuration time, that you have a non-GNU linker
10848 or a GNU linker that does not support option @option{--eh-frame-hdr},
10849 it links the shared version of @file{libgcc} into shared libraries
10850 by default. Otherwise, it takes advantage of the linker and optimizes
10851 away the linking with the shared version of @file{libgcc}, linking with
10852 the static version of libgcc by default. This allows exceptions to
10853 propagate through such shared libraries, without incurring relocation
10854 costs at library load time.
10856 However, if a library or main executable is supposed to throw or catch
10857 exceptions, you must link it using the G++ or GCJ driver, as appropriate
10858 for the languages used in the program, or using the option
10859 @option{-shared-libgcc}, such that it is linked with the shared
10862 @item -static-libasan
10863 @opindex static-libasan
10864 When the @option{-fsanitize=address} option is used to link a program,
10865 the GCC driver automatically links against @option{libasan}. If
10866 @file{libasan} is available as a shared library, and the @option{-static}
10867 option is not used, then this links against the shared version of
10868 @file{libasan}. The @option{-static-libasan} option directs the GCC
10869 driver to link @file{libasan} statically, without necessarily linking
10870 other libraries statically.
10872 @item -static-libtsan
10873 @opindex static-libtsan
10874 When the @option{-fsanitize=thread} option is used to link a program,
10875 the GCC driver automatically links against @option{libtsan}. If
10876 @file{libtsan} is available as a shared library, and the @option{-static}
10877 option is not used, then this links against the shared version of
10878 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
10879 driver to link @file{libtsan} statically, without necessarily linking
10880 other libraries statically.
10882 @item -static-liblsan
10883 @opindex static-liblsan
10884 When the @option{-fsanitize=leak} option is used to link a program,
10885 the GCC driver automatically links against @option{liblsan}. If
10886 @file{liblsan} is available as a shared library, and the @option{-static}
10887 option is not used, then this links against the shared version of
10888 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
10889 driver to link @file{liblsan} statically, without necessarily linking
10890 other libraries statically.
10892 @item -static-libubsan
10893 @opindex static-libubsan
10894 When the @option{-fsanitize=undefined} option is used to link a program,
10895 the GCC driver automatically links against @option{libubsan}. If
10896 @file{libubsan} is available as a shared library, and the @option{-static}
10897 option is not used, then this links against the shared version of
10898 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
10899 driver to link @file{libubsan} statically, without necessarily linking
10900 other libraries statically.
10902 @item -static-libstdc++
10903 @opindex static-libstdc++
10904 When the @command{g++} program is used to link a C++ program, it
10905 normally automatically links against @option{libstdc++}. If
10906 @file{libstdc++} is available as a shared library, and the
10907 @option{-static} option is not used, then this links against the
10908 shared version of @file{libstdc++}. That is normally fine. However, it
10909 is sometimes useful to freeze the version of @file{libstdc++} used by
10910 the program without going all the way to a fully static link. The
10911 @option{-static-libstdc++} option directs the @command{g++} driver to
10912 link @file{libstdc++} statically, without necessarily linking other
10913 libraries statically.
10917 Bind references to global symbols when building a shared object. Warn
10918 about any unresolved references (unless overridden by the link editor
10919 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
10922 @item -T @var{script}
10924 @cindex linker script
10925 Use @var{script} as the linker script. This option is supported by most
10926 systems using the GNU linker. On some targets, such as bare-board
10927 targets without an operating system, the @option{-T} option may be required
10928 when linking to avoid references to undefined symbols.
10930 @item -Xlinker @var{option}
10932 Pass @var{option} as an option to the linker. You can use this to
10933 supply system-specific linker options that GCC does not recognize.
10935 If you want to pass an option that takes a separate argument, you must use
10936 @option{-Xlinker} twice, once for the option and once for the argument.
10937 For example, to pass @option{-assert definitions}, you must write
10938 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
10939 @option{-Xlinker "-assert definitions"}, because this passes the entire
10940 string as a single argument, which is not what the linker expects.
10942 When using the GNU linker, it is usually more convenient to pass
10943 arguments to linker options using the @option{@var{option}=@var{value}}
10944 syntax than as separate arguments. For example, you can specify
10945 @option{-Xlinker -Map=output.map} rather than
10946 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
10947 this syntax for command-line options.
10949 @item -Wl,@var{option}
10951 Pass @var{option} as an option to the linker. If @var{option} contains
10952 commas, it is split into multiple options at the commas. You can use this
10953 syntax to pass an argument to the option.
10954 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
10955 linker. When using the GNU linker, you can also get the same effect with
10956 @option{-Wl,-Map=output.map}.
10958 @item -u @var{symbol}
10960 Pretend the symbol @var{symbol} is undefined, to force linking of
10961 library modules to define it. You can use @option{-u} multiple times with
10962 different symbols to force loading of additional library modules.
10964 @item -z @var{keyword}
10966 @option{-z} is passed directly on to the linker along with the keyword
10967 @var{keyword}. See the section in the documentation of your linker for
10968 permitted values and their meanings.
10971 @node Directory Options
10972 @section Options for Directory Search
10973 @cindex directory options
10974 @cindex options, directory search
10975 @cindex search path
10977 These options specify directories to search for header files, for
10978 libraries and for parts of the compiler:
10983 Add the directory @var{dir} to the head of the list of directories to be
10984 searched for header files. This can be used to override a system header
10985 file, substituting your own version, since these directories are
10986 searched before the system header file directories. However, you should
10987 not use this option to add directories that contain vendor-supplied
10988 system header files (use @option{-isystem} for that). If you use more than
10989 one @option{-I} option, the directories are scanned in left-to-right
10990 order; the standard system directories come after.
10992 If a standard system include directory, or a directory specified with
10993 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
10994 option is ignored. The directory is still searched but as a
10995 system directory at its normal position in the system include chain.
10996 This is to ensure that GCC's procedure to fix buggy system headers and
10997 the ordering for the @code{include_next} directive are not inadvertently changed.
10998 If you really need to change the search order for system directories,
10999 use the @option{-nostdinc} and/or @option{-isystem} options.
11001 @item -iplugindir=@var{dir}
11002 @opindex iplugindir=
11003 Set the directory to search for plugins that are passed
11004 by @option{-fplugin=@var{name}} instead of
11005 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
11006 to be used by the user, but only passed by the driver.
11008 @item -iquote@var{dir}
11010 Add the directory @var{dir} to the head of the list of directories to
11011 be searched for header files only for the case of @samp{#include
11012 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
11013 otherwise just like @option{-I}.
11017 Add directory @var{dir} to the list of directories to be searched
11020 @item -B@var{prefix}
11022 This option specifies where to find the executables, libraries,
11023 include files, and data files of the compiler itself.
11025 The compiler driver program runs one or more of the subprograms
11026 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
11027 @var{prefix} as a prefix for each program it tries to run, both with and
11028 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
11030 For each subprogram to be run, the compiler driver first tries the
11031 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
11032 is not specified, the driver tries two standard prefixes,
11033 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
11034 those results in a file name that is found, the unmodified program
11035 name is searched for using the directories specified in your
11036 @env{PATH} environment variable.
11038 The compiler checks to see if the path provided by the @option{-B}
11039 refers to a directory, and if necessary it adds a directory
11040 separator character at the end of the path.
11042 @option{-B} prefixes that effectively specify directory names also apply
11043 to libraries in the linker, because the compiler translates these
11044 options into @option{-L} options for the linker. They also apply to
11045 include files in the preprocessor, because the compiler translates these
11046 options into @option{-isystem} options for the preprocessor. In this case,
11047 the compiler appends @samp{include} to the prefix.
11049 The runtime support file @file{libgcc.a} can also be searched for using
11050 the @option{-B} prefix, if needed. If it is not found there, the two
11051 standard prefixes above are tried, and that is all. The file is left
11052 out of the link if it is not found by those means.
11054 Another way to specify a prefix much like the @option{-B} prefix is to use
11055 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
11058 As a special kludge, if the path provided by @option{-B} is
11059 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
11060 9, then it is replaced by @file{[dir/]include}. This is to help
11061 with boot-strapping the compiler.
11063 @item -specs=@var{file}
11065 Process @var{file} after the compiler reads in the standard @file{specs}
11066 file, in order to override the defaults which the @command{gcc} driver
11067 program uses when determining what switches to pass to @command{cc1},
11068 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
11069 @option{-specs=@var{file}} can be specified on the command line, and they
11070 are processed in order, from left to right.
11072 @item --sysroot=@var{dir}
11074 Use @var{dir} as the logical root directory for headers and libraries.
11075 For example, if the compiler normally searches for headers in
11076 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
11077 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
11079 If you use both this option and the @option{-isysroot} option, then
11080 the @option{--sysroot} option applies to libraries, but the
11081 @option{-isysroot} option applies to header files.
11083 The GNU linker (beginning with version 2.16) has the necessary support
11084 for this option. If your linker does not support this option, the
11085 header file aspect of @option{--sysroot} still works, but the
11086 library aspect does not.
11088 @item --no-sysroot-suffix
11089 @opindex no-sysroot-suffix
11090 For some targets, a suffix is added to the root directory specified
11091 with @option{--sysroot}, depending on the other options used, so that
11092 headers may for example be found in
11093 @file{@var{dir}/@var{suffix}/usr/include} instead of
11094 @file{@var{dir}/usr/include}. This option disables the addition of
11099 This option has been deprecated. Please use @option{-iquote} instead for
11100 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
11101 Any directories you specify with @option{-I} options before the @option{-I-}
11102 option are searched only for the case of @samp{#include "@var{file}"};
11103 they are not searched for @samp{#include <@var{file}>}.
11105 If additional directories are specified with @option{-I} options after
11106 the @option{-I-}, these directories are searched for all @samp{#include}
11107 directives. (Ordinarily @emph{all} @option{-I} directories are used
11110 In addition, the @option{-I-} option inhibits the use of the current
11111 directory (where the current input file came from) as the first search
11112 directory for @samp{#include "@var{file}"}. There is no way to
11113 override this effect of @option{-I-}. With @option{-I.} you can specify
11114 searching the directory that is current when the compiler is
11115 invoked. That is not exactly the same as what the preprocessor does
11116 by default, but it is often satisfactory.
11118 @option{-I-} does not inhibit the use of the standard system directories
11119 for header files. Thus, @option{-I-} and @option{-nostdinc} are
11126 @section Specifying subprocesses and the switches to pass to them
11129 @command{gcc} is a driver program. It performs its job by invoking a
11130 sequence of other programs to do the work of compiling, assembling and
11131 linking. GCC interprets its command-line parameters and uses these to
11132 deduce which programs it should invoke, and which command-line options
11133 it ought to place on their command lines. This behavior is controlled
11134 by @dfn{spec strings}. In most cases there is one spec string for each
11135 program that GCC can invoke, but a few programs have multiple spec
11136 strings to control their behavior. The spec strings built into GCC can
11137 be overridden by using the @option{-specs=} command-line switch to specify
11140 @dfn{Spec files} are plaintext files that are used to construct spec
11141 strings. They consist of a sequence of directives separated by blank
11142 lines. The type of directive is determined by the first non-whitespace
11143 character on the line, which can be one of the following:
11146 @item %@var{command}
11147 Issues a @var{command} to the spec file processor. The commands that can
11151 @item %include <@var{file}>
11152 @cindex @code{%include}
11153 Search for @var{file} and insert its text at the current point in the
11156 @item %include_noerr <@var{file}>
11157 @cindex @code{%include_noerr}
11158 Just like @samp{%include}, but do not generate an error message if the include
11159 file cannot be found.
11161 @item %rename @var{old_name} @var{new_name}
11162 @cindex @code{%rename}
11163 Rename the spec string @var{old_name} to @var{new_name}.
11167 @item *[@var{spec_name}]:
11168 This tells the compiler to create, override or delete the named spec
11169 string. All lines after this directive up to the next directive or
11170 blank line are considered to be the text for the spec string. If this
11171 results in an empty string then the spec is deleted. (Or, if the
11172 spec did not exist, then nothing happens.) Otherwise, if the spec
11173 does not currently exist a new spec is created. If the spec does
11174 exist then its contents are overridden by the text of this
11175 directive, unless the first character of that text is the @samp{+}
11176 character, in which case the text is appended to the spec.
11178 @item [@var{suffix}]:
11179 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
11180 and up to the next directive or blank line are considered to make up the
11181 spec string for the indicated suffix. When the compiler encounters an
11182 input file with the named suffix, it processes the spec string in
11183 order to work out how to compile that file. For example:
11187 z-compile -input %i
11190 This says that any input file whose name ends in @samp{.ZZ} should be
11191 passed to the program @samp{z-compile}, which should be invoked with the
11192 command-line switch @option{-input} and with the result of performing the
11193 @samp{%i} substitution. (See below.)
11195 As an alternative to providing a spec string, the text following a
11196 suffix directive can be one of the following:
11199 @item @@@var{language}
11200 This says that the suffix is an alias for a known @var{language}. This is
11201 similar to using the @option{-x} command-line switch to GCC to specify a
11202 language explicitly. For example:
11209 Says that .ZZ files are, in fact, C++ source files.
11212 This causes an error messages saying:
11215 @var{name} compiler not installed on this system.
11219 GCC already has an extensive list of suffixes built into it.
11220 This directive adds an entry to the end of the list of suffixes, but
11221 since the list is searched from the end backwards, it is effectively
11222 possible to override earlier entries using this technique.
11226 GCC has the following spec strings built into it. Spec files can
11227 override these strings or create their own. Note that individual
11228 targets can also add their own spec strings to this list.
11231 asm Options to pass to the assembler
11232 asm_final Options to pass to the assembler post-processor
11233 cpp Options to pass to the C preprocessor
11234 cc1 Options to pass to the C compiler
11235 cc1plus Options to pass to the C++ compiler
11236 endfile Object files to include at the end of the link
11237 link Options to pass to the linker
11238 lib Libraries to include on the command line to the linker
11239 libgcc Decides which GCC support library to pass to the linker
11240 linker Sets the name of the linker
11241 predefines Defines to be passed to the C preprocessor
11242 signed_char Defines to pass to CPP to say whether @code{char} is signed
11244 startfile Object files to include at the start of the link
11247 Here is a small example of a spec file:
11250 %rename lib old_lib
11253 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
11256 This example renames the spec called @samp{lib} to @samp{old_lib} and
11257 then overrides the previous definition of @samp{lib} with a new one.
11258 The new definition adds in some extra command-line options before
11259 including the text of the old definition.
11261 @dfn{Spec strings} are a list of command-line options to be passed to their
11262 corresponding program. In addition, the spec strings can contain
11263 @samp{%}-prefixed sequences to substitute variable text or to
11264 conditionally insert text into the command line. Using these constructs
11265 it is possible to generate quite complex command lines.
11267 Here is a table of all defined @samp{%}-sequences for spec
11268 strings. Note that spaces are not generated automatically around the
11269 results of expanding these sequences. Therefore you can concatenate them
11270 together or combine them with constant text in a single argument.
11274 Substitute one @samp{%} into the program name or argument.
11277 Substitute the name of the input file being processed.
11280 Substitute the basename of the input file being processed.
11281 This is the substring up to (and not including) the last period
11282 and not including the directory.
11285 This is the same as @samp{%b}, but include the file suffix (text after
11289 Marks the argument containing or following the @samp{%d} as a
11290 temporary file name, so that that file is deleted if GCC exits
11291 successfully. Unlike @samp{%g}, this contributes no text to the
11294 @item %g@var{suffix}
11295 Substitute a file name that has suffix @var{suffix} and is chosen
11296 once per compilation, and mark the argument in the same way as
11297 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
11298 name is now chosen in a way that is hard to predict even when previously
11299 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
11300 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
11301 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
11302 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
11303 was simply substituted with a file name chosen once per compilation,
11304 without regard to any appended suffix (which was therefore treated
11305 just like ordinary text), making such attacks more likely to succeed.
11307 @item %u@var{suffix}
11308 Like @samp{%g}, but generates a new temporary file name
11309 each time it appears instead of once per compilation.
11311 @item %U@var{suffix}
11312 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
11313 new one if there is no such last file name. In the absence of any
11314 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
11315 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
11316 involves the generation of two distinct file names, one
11317 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
11318 simply substituted with a file name chosen for the previous @samp{%u},
11319 without regard to any appended suffix.
11321 @item %j@var{suffix}
11322 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
11323 writable, and if @option{-save-temps} is not used;
11324 otherwise, substitute the name
11325 of a temporary file, just like @samp{%u}. This temporary file is not
11326 meant for communication between processes, but rather as a junk
11327 disposal mechanism.
11329 @item %|@var{suffix}
11330 @itemx %m@var{suffix}
11331 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
11332 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
11333 all. These are the two most common ways to instruct a program that it
11334 should read from standard input or write to standard output. If you
11335 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
11336 construct: see for example @file{f/lang-specs.h}.
11338 @item %.@var{SUFFIX}
11339 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
11340 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
11341 terminated by the next space or %.
11344 Marks the argument containing or following the @samp{%w} as the
11345 designated output file of this compilation. This puts the argument
11346 into the sequence of arguments that @samp{%o} substitutes.
11349 Substitutes the names of all the output files, with spaces
11350 automatically placed around them. You should write spaces
11351 around the @samp{%o} as well or the results are undefined.
11352 @samp{%o} is for use in the specs for running the linker.
11353 Input files whose names have no recognized suffix are not compiled
11354 at all, but they are included among the output files, so they are
11358 Substitutes the suffix for object files. Note that this is
11359 handled specially when it immediately follows @samp{%g, %u, or %U},
11360 because of the need for those to form complete file names. The
11361 handling is such that @samp{%O} is treated exactly as if it had already
11362 been substituted, except that @samp{%g, %u, and %U} do not currently
11363 support additional @var{suffix} characters following @samp{%O} as they do
11364 following, for example, @samp{.o}.
11367 Substitutes the standard macro predefinitions for the
11368 current target machine. Use this when running @code{cpp}.
11371 Like @samp{%p}, but puts @samp{__} before and after the name of each
11372 predefined macro, except for macros that start with @samp{__} or with
11373 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
11377 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
11378 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
11379 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
11380 and @option{-imultilib} as necessary.
11383 Current argument is the name of a library or startup file of some sort.
11384 Search for that file in a standard list of directories and substitute
11385 the full name found. The current working directory is included in the
11386 list of directories scanned.
11389 Current argument is the name of a linker script. Search for that file
11390 in the current list of directories to scan for libraries. If the file
11391 is located insert a @option{--script} option into the command line
11392 followed by the full path name found. If the file is not found then
11393 generate an error message. Note: the current working directory is not
11397 Print @var{str} as an error message. @var{str} is terminated by a newline.
11398 Use this when inconsistent options are detected.
11400 @item %(@var{name})
11401 Substitute the contents of spec string @var{name} at this point.
11403 @item %x@{@var{option}@}
11404 Accumulate an option for @samp{%X}.
11407 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
11411 Output the accumulated assembler options specified by @option{-Wa}.
11414 Output the accumulated preprocessor options specified by @option{-Wp}.
11417 Process the @code{asm} spec. This is used to compute the
11418 switches to be passed to the assembler.
11421 Process the @code{asm_final} spec. This is a spec string for
11422 passing switches to an assembler post-processor, if such a program is
11426 Process the @code{link} spec. This is the spec for computing the
11427 command line passed to the linker. Typically it makes use of the
11428 @samp{%L %G %S %D and %E} sequences.
11431 Dump out a @option{-L} option for each directory that GCC believes might
11432 contain startup files. If the target supports multilibs then the
11433 current multilib directory is prepended to each of these paths.
11436 Process the @code{lib} spec. This is a spec string for deciding which
11437 libraries are included on the command line to the linker.
11440 Process the @code{libgcc} spec. This is a spec string for deciding
11441 which GCC support library is included on the command line to the linker.
11444 Process the @code{startfile} spec. This is a spec for deciding which
11445 object files are the first ones passed to the linker. Typically
11446 this might be a file named @file{crt0.o}.
11449 Process the @code{endfile} spec. This is a spec string that specifies
11450 the last object files that are passed to the linker.
11453 Process the @code{cpp} spec. This is used to construct the arguments
11454 to be passed to the C preprocessor.
11457 Process the @code{cc1} spec. This is used to construct the options to be
11458 passed to the actual C compiler (@samp{cc1}).
11461 Process the @code{cc1plus} spec. This is used to construct the options to be
11462 passed to the actual C++ compiler (@samp{cc1plus}).
11465 Substitute the variable part of a matched option. See below.
11466 Note that each comma in the substituted string is replaced by
11470 Remove all occurrences of @code{-S} from the command line. Note---this
11471 command is position dependent. @samp{%} commands in the spec string
11472 before this one see @code{-S}, @samp{%} commands in the spec string
11473 after this one do not.
11475 @item %:@var{function}(@var{args})
11476 Call the named function @var{function}, passing it @var{args}.
11477 @var{args} is first processed as a nested spec string, then split
11478 into an argument vector in the usual fashion. The function returns
11479 a string which is processed as if it had appeared literally as part
11480 of the current spec.
11482 The following built-in spec functions are provided:
11485 @item @code{getenv}
11486 The @code{getenv} spec function takes two arguments: an environment
11487 variable name and a string. If the environment variable is not
11488 defined, a fatal error is issued. Otherwise, the return value is the
11489 value of the environment variable concatenated with the string. For
11490 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
11493 %:getenv(TOPDIR /include)
11496 expands to @file{/path/to/top/include}.
11498 @item @code{if-exists}
11499 The @code{if-exists} spec function takes one argument, an absolute
11500 pathname to a file. If the file exists, @code{if-exists} returns the
11501 pathname. Here is a small example of its usage:
11505 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
11508 @item @code{if-exists-else}
11509 The @code{if-exists-else} spec function is similar to the @code{if-exists}
11510 spec function, except that it takes two arguments. The first argument is
11511 an absolute pathname to a file. If the file exists, @code{if-exists-else}
11512 returns the pathname. If it does not exist, it returns the second argument.
11513 This way, @code{if-exists-else} can be used to select one file or another,
11514 based on the existence of the first. Here is a small example of its usage:
11518 crt0%O%s %:if-exists(crti%O%s) \
11519 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
11522 @item @code{replace-outfile}
11523 The @code{replace-outfile} spec function takes two arguments. It looks for the
11524 first argument in the outfiles array and replaces it with the second argument. Here
11525 is a small example of its usage:
11528 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
11531 @item @code{remove-outfile}
11532 The @code{remove-outfile} spec function takes one argument. It looks for the
11533 first argument in the outfiles array and removes it. Here is a small example
11537 %:remove-outfile(-lm)
11540 @item @code{pass-through-libs}
11541 The @code{pass-through-libs} spec function takes any number of arguments. It
11542 finds any @option{-l} options and any non-options ending in @file{.a} (which it
11543 assumes are the names of linker input library archive files) and returns a
11544 result containing all the found arguments each prepended by
11545 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
11546 intended to be passed to the LTO linker plugin.
11549 %:pass-through-libs(%G %L %G)
11552 @item @code{print-asm-header}
11553 The @code{print-asm-header} function takes no arguments and simply
11554 prints a banner like:
11560 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
11563 It is used to separate compiler options from assembler options
11564 in the @option{--target-help} output.
11567 @item %@{@code{S}@}
11568 Substitutes the @code{-S} switch, if that switch is given to GCC@.
11569 If that switch is not specified, this substitutes nothing. Note that
11570 the leading dash is omitted when specifying this option, and it is
11571 automatically inserted if the substitution is performed. Thus the spec
11572 string @samp{%@{foo@}} matches the command-line option @option{-foo}
11573 and outputs the command-line option @option{-foo}.
11575 @item %W@{@code{S}@}
11576 Like %@{@code{S}@} but mark last argument supplied within as a file to be
11577 deleted on failure.
11579 @item %@{@code{S}*@}
11580 Substitutes all the switches specified to GCC whose names start
11581 with @code{-S}, but which also take an argument. This is used for
11582 switches like @option{-o}, @option{-D}, @option{-I}, etc.
11583 GCC considers @option{-o foo} as being
11584 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
11585 text, including the space. Thus two arguments are generated.
11587 @item %@{@code{S}*&@code{T}*@}
11588 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
11589 (the order of @code{S} and @code{T} in the spec is not significant).
11590 There can be any number of ampersand-separated variables; for each the
11591 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
11593 @item %@{@code{S}:@code{X}@}
11594 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
11596 @item %@{!@code{S}:@code{X}@}
11597 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
11599 @item %@{@code{S}*:@code{X}@}
11600 Substitutes @code{X} if one or more switches whose names start with
11601 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
11602 once, no matter how many such switches appeared. However, if @code{%*}
11603 appears somewhere in @code{X}, then @code{X} is substituted once
11604 for each matching switch, with the @code{%*} replaced by the part of
11605 that switch matching the @code{*}.
11607 If @code{%*} appears as the last part of a spec sequence then a space
11608 will be added after the end of the last substitution. If there is more
11609 text in the sequence however then a space will not be generated. This
11610 allows the @code{%*} substitution to be used as part of a larger
11611 string. For example, a spec string like this:
11614 %@{mcu=*:--script=%*/memory.ld@}
11617 when matching an option like @code{-mcu=newchip} will produce:
11620 --script=newchip/memory.ld
11623 @item %@{.@code{S}:@code{X}@}
11624 Substitutes @code{X}, if processing a file with suffix @code{S}.
11626 @item %@{!.@code{S}:@code{X}@}
11627 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
11629 @item %@{,@code{S}:@code{X}@}
11630 Substitutes @code{X}, if processing a file for language @code{S}.
11632 @item %@{!,@code{S}:@code{X}@}
11633 Substitutes @code{X}, if not processing a file for language @code{S}.
11635 @item %@{@code{S}|@code{P}:@code{X}@}
11636 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
11637 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
11638 @code{*} sequences as well, although they have a stronger binding than
11639 the @samp{|}. If @code{%*} appears in @code{X}, all of the
11640 alternatives must be starred, and only the first matching alternative
11643 For example, a spec string like this:
11646 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
11650 outputs the following command-line options from the following input
11651 command-line options:
11656 -d fred.c -foo -baz -boggle
11657 -d jim.d -bar -baz -boggle
11660 @item %@{S:X; T:Y; :D@}
11662 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
11663 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
11664 be as many clauses as you need. This may be combined with @code{.},
11665 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
11670 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
11671 construct may contain other nested @samp{%} constructs or spaces, or
11672 even newlines. They are processed as usual, as described above.
11673 Trailing white space in @code{X} is ignored. White space may also
11674 appear anywhere on the left side of the colon in these constructs,
11675 except between @code{.} or @code{*} and the corresponding word.
11677 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
11678 handled specifically in these constructs. If another value of
11679 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
11680 @option{-W} switch is found later in the command line, the earlier
11681 switch value is ignored, except with @{@code{S}*@} where @code{S} is
11682 just one letter, which passes all matching options.
11684 The character @samp{|} at the beginning of the predicate text is used to
11685 indicate that a command should be piped to the following command, but
11686 only if @option{-pipe} is specified.
11688 It is built into GCC which switches take arguments and which do not.
11689 (You might think it would be useful to generalize this to allow each
11690 compiler's spec to say which switches take arguments. But this cannot
11691 be done in a consistent fashion. GCC cannot even decide which input
11692 files have been specified without knowing which switches take arguments,
11693 and it must know which input files to compile in order to tell which
11696 GCC also knows implicitly that arguments starting in @option{-l} are to be
11697 treated as compiler output files, and passed to the linker in their
11698 proper position among the other output files.
11700 @c man begin OPTIONS
11702 @node Target Options
11703 @section Specifying Target Machine and Compiler Version
11704 @cindex target options
11705 @cindex cross compiling
11706 @cindex specifying machine version
11707 @cindex specifying compiler version and target machine
11708 @cindex compiler version, specifying
11709 @cindex target machine, specifying
11711 The usual way to run GCC is to run the executable called @command{gcc}, or
11712 @command{@var{machine}-gcc} when cross-compiling, or
11713 @command{@var{machine}-gcc-@var{version}} to run a version other than the
11714 one that was installed last.
11716 @node Submodel Options
11717 @section Hardware Models and Configurations
11718 @cindex submodel options
11719 @cindex specifying hardware config
11720 @cindex hardware models and configurations, specifying
11721 @cindex machine dependent options
11723 Each target machine types can have its own
11724 special options, starting with @samp{-m}, to choose among various
11725 hardware models or configurations---for example, 68010 vs 68020,
11726 floating coprocessor or none. A single installed version of the
11727 compiler can compile for any model or configuration, according to the
11730 Some configurations of the compiler also support additional special
11731 options, usually for compatibility with other compilers on the same
11734 @c This list is ordered alphanumerically by subsection name.
11735 @c It should be the same order and spelling as these options are listed
11736 @c in Machine Dependent Options
11739 * AArch64 Options::
11740 * Adapteva Epiphany Options::
11744 * Blackfin Options::
11749 * DEC Alpha Options::
11752 * GNU/Linux Options::
11755 * i386 and x86-64 Options::
11756 * i386 and x86-64 Windows Options::
11764 * MicroBlaze Options::
11767 * MN10300 Options::
11771 * Nios II Options::
11773 * picoChip Options::
11774 * PowerPC Options::
11776 * RS/6000 and PowerPC Options::
11778 * S/390 and zSeries Options::
11781 * Solaris 2 Options::
11784 * System V Options::
11785 * TILE-Gx Options::
11786 * TILEPro Options::
11790 * VxWorks Options::
11792 * Xstormy16 Options::
11794 * zSeries Options::
11797 @node AArch64 Options
11798 @subsection AArch64 Options
11799 @cindex AArch64 Options
11801 These options are defined for AArch64 implementations:
11805 @item -mabi=@var{name}
11807 Generate code for the specified data model. Permissible values
11808 are @samp{ilp32} for SysV-like data model where int, long int and pointer
11809 are 32-bit, and @samp{lp64} for SysV-like data model where int is 32-bit,
11810 but long int and pointer are 64-bit.
11812 The default depends on the specific target configuration. Note that
11813 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
11814 entire program with the same ABI, and link with a compatible set of libraries.
11817 @opindex mbig-endian
11818 Generate big-endian code. This is the default when GCC is configured for an
11819 @samp{aarch64_be-*-*} target.
11821 @item -mgeneral-regs-only
11822 @opindex mgeneral-regs-only
11823 Generate code which uses only the general registers.
11825 @item -mlittle-endian
11826 @opindex mlittle-endian
11827 Generate little-endian code. This is the default when GCC is configured for an
11828 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
11830 @item -mcmodel=tiny
11831 @opindex mcmodel=tiny
11832 Generate code for the tiny code model. The program and its statically defined
11833 symbols must be within 1GB of each other. Pointers are 64 bits. Programs can
11834 be statically or dynamically linked. This model is not fully implemented and
11835 mostly treated as @samp{small}.
11837 @item -mcmodel=small
11838 @opindex mcmodel=small
11839 Generate code for the small code model. The program and its statically defined
11840 symbols must be within 4GB of each other. Pointers are 64 bits. Programs can
11841 be statically or dynamically linked. This is the default code model.
11843 @item -mcmodel=large
11844 @opindex mcmodel=large
11845 Generate code for the large code model. This makes no assumptions about
11846 addresses and sizes of sections. Pointers are 64 bits. Programs can be
11847 statically linked only.
11849 @item -mstrict-align
11850 @opindex mstrict-align
11851 Do not assume that unaligned memory references will be handled by the system.
11853 @item -momit-leaf-frame-pointer
11854 @itemx -mno-omit-leaf-frame-pointer
11855 @opindex momit-leaf-frame-pointer
11856 @opindex mno-omit-leaf-frame-pointer
11857 Omit or keep the frame pointer in leaf functions. The former behaviour is the
11860 @item -mtls-dialect=desc
11861 @opindex mtls-dialect=desc
11862 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
11863 of TLS variables. This is the default.
11865 @item -mtls-dialect=traditional
11866 @opindex mtls-dialect=traditional
11867 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
11870 @item -mfix-cortex-a53-835769
11871 @itemx -mno-fix-cortex-a53-835769
11872 @opindex -mfix-cortex-a53-835769
11873 @opindex -mno-fix-cortex-a53-835769
11874 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
11875 This will involve inserting a NOP instruction between memory instructions and
11876 64-bit integer multiply-accumulate instructions.
11878 @item -march=@var{name}
11880 Specify the name of the target architecture, optionally suffixed by one or
11881 more feature modifiers. This option has the form
11882 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
11883 only permissible value for @var{arch} is @samp{armv8-a}. The permissible
11884 values for @var{feature} are documented in the sub-section below.
11886 Where conflicting feature modifiers are specified, the right-most feature is
11889 GCC uses this name to determine what kind of instructions it can emit when
11890 generating assembly code.
11892 Where @option{-march} is specified without either of @option{-mtune}
11893 or @option{-mcpu} also being specified, the code will be tuned to perform
11894 well across a range of target processors implementing the target
11897 @item -mtune=@var{name}
11899 Specify the name of the target processor for which GCC should tune the
11900 performance of the code. Permissible values for this option are:
11901 @samp{generic}, @samp{cortex-a53}, @samp{cortex-a57}, @samp{thunderx}.
11903 Additionally, this option can specify that GCC should tune the performance
11904 of the code for a big.LITTLE system. The only permissible value is
11905 @samp{cortex-a57.cortex-a53}.
11907 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
11908 are specified, the code will be tuned to perform well across a range
11909 of target processors.
11911 This option cannot be suffixed by feature modifiers.
11913 @item -mcpu=@var{name}
11915 Specify the name of the target processor, optionally suffixed by one or more
11916 feature modifiers. This option has the form
11917 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
11918 permissible values for @var{cpu} are the same as those available for
11921 The permissible values for @var{feature} are documented in the sub-section
11924 Where conflicting feature modifiers are specified, the right-most feature is
11927 GCC uses this name to determine what kind of instructions it can emit when
11928 generating assembly code (as if by @option{-march}) and to determine
11929 the target processor for which to tune for performance (as if
11930 by @option{-mtune}). Where this option is used in conjunction
11931 with @option{-march} or @option{-mtune}, those options take precedence
11932 over the appropriate part of this option.
11935 @subsubsection @option{-march} and @option{-mcpu} feature modifiers
11936 @cindex @option{-march} feature modifiers
11937 @cindex @option{-mcpu} feature modifiers
11938 Feature modifiers used with @option{-march} and @option{-mcpu} can be one
11943 Enable CRC extension.
11945 Enable Crypto extension. This implies Advanced SIMD is enabled.
11947 Enable floating-point instructions.
11949 Enable Advanced SIMD instructions. This implies floating-point instructions
11950 are enabled. This is the default for all current possible values for options
11951 @option{-march} and @option{-mcpu=}.
11954 @node Adapteva Epiphany Options
11955 @subsection Adapteva Epiphany Options
11957 These @samp{-m} options are defined for Adapteva Epiphany:
11960 @item -mhalf-reg-file
11961 @opindex mhalf-reg-file
11962 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
11963 That allows code to run on hardware variants that lack these registers.
11965 @item -mprefer-short-insn-regs
11966 @opindex mprefer-short-insn-regs
11967 Preferrentially allocate registers that allow short instruction generation.
11968 This can result in increased instruction count, so this may either reduce or
11969 increase overall code size.
11971 @item -mbranch-cost=@var{num}
11972 @opindex mbranch-cost
11973 Set the cost of branches to roughly @var{num} ``simple'' instructions.
11974 This cost is only a heuristic and is not guaranteed to produce
11975 consistent results across releases.
11979 Enable the generation of conditional moves.
11981 @item -mnops=@var{num}
11983 Emit @var{num} NOPs before every other generated instruction.
11985 @item -mno-soft-cmpsf
11986 @opindex mno-soft-cmpsf
11987 For single-precision floating-point comparisons, emit an @code{fsub} instruction
11988 and test the flags. This is faster than a software comparison, but can
11989 get incorrect results in the presence of NaNs, or when two different small
11990 numbers are compared such that their difference is calculated as zero.
11991 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
11992 software comparisons.
11994 @item -mstack-offset=@var{num}
11995 @opindex mstack-offset
11996 Set the offset between the top of the stack and the stack pointer.
11997 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
11998 can be used by leaf functions without stack allocation.
11999 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
12000 Note also that this option changes the ABI; compiling a program with a
12001 different stack offset than the libraries have been compiled with
12002 generally does not work.
12003 This option can be useful if you want to evaluate if a different stack
12004 offset would give you better code, but to actually use a different stack
12005 offset to build working programs, it is recommended to configure the
12006 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
12008 @item -mno-round-nearest
12009 @opindex mno-round-nearest
12010 Make the scheduler assume that the rounding mode has been set to
12011 truncating. The default is @option{-mround-nearest}.
12014 @opindex mlong-calls
12015 If not otherwise specified by an attribute, assume all calls might be beyond
12016 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
12017 function address into a register before performing a (otherwise direct) call.
12018 This is the default.
12020 @item -mshort-calls
12021 @opindex short-calls
12022 If not otherwise specified by an attribute, assume all direct calls are
12023 in the range of the @code{b} / @code{bl} instructions, so use these instructions
12024 for direct calls. The default is @option{-mlong-calls}.
12028 Assume addresses can be loaded as 16-bit unsigned values. This does not
12029 apply to function addresses for which @option{-mlong-calls} semantics
12032 @item -mfp-mode=@var{mode}
12034 Set the prevailing mode of the floating-point unit.
12035 This determines the floating-point mode that is provided and expected
12036 at function call and return time. Making this mode match the mode you
12037 predominantly need at function start can make your programs smaller and
12038 faster by avoiding unnecessary mode switches.
12040 @var{mode} can be set to one the following values:
12044 Any mode at function entry is valid, and retained or restored when
12045 the function returns, and when it calls other functions.
12046 This mode is useful for compiling libraries or other compilation units
12047 you might want to incorporate into different programs with different
12048 prevailing FPU modes, and the convenience of being able to use a single
12049 object file outweighs the size and speed overhead for any extra
12050 mode switching that might be needed, compared with what would be needed
12051 with a more specific choice of prevailing FPU mode.
12054 This is the mode used for floating-point calculations with
12055 truncating (i.e.@: round towards zero) rounding mode. That includes
12056 conversion from floating point to integer.
12058 @item round-nearest
12059 This is the mode used for floating-point calculations with
12060 round-to-nearest-or-even rounding mode.
12063 This is the mode used to perform integer calculations in the FPU, e.g.@:
12064 integer multiply, or integer multiply-and-accumulate.
12067 The default is @option{-mfp-mode=caller}
12069 @item -mnosplit-lohi
12070 @itemx -mno-postinc
12071 @itemx -mno-postmodify
12072 @opindex mnosplit-lohi
12073 @opindex mno-postinc
12074 @opindex mno-postmodify
12075 Code generation tweaks that disable, respectively, splitting of 32-bit
12076 loads, generation of post-increment addresses, and generation of
12077 post-modify addresses. The defaults are @option{msplit-lohi},
12078 @option{-mpost-inc}, and @option{-mpost-modify}.
12080 @item -mnovect-double
12081 @opindex mno-vect-double
12082 Change the preferred SIMD mode to SImode. The default is
12083 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
12085 @item -max-vect-align=@var{num}
12086 @opindex max-vect-align
12087 The maximum alignment for SIMD vector mode types.
12088 @var{num} may be 4 or 8. The default is 8.
12089 Note that this is an ABI change, even though many library function
12090 interfaces are unaffected if they don't use SIMD vector modes
12091 in places that affect size and/or alignment of relevant types.
12093 @item -msplit-vecmove-early
12094 @opindex msplit-vecmove-early
12095 Split vector moves into single word moves before reload. In theory this
12096 can give better register allocation, but so far the reverse seems to be
12097 generally the case.
12099 @item -m1reg-@var{reg}
12101 Specify a register to hold the constant @minus{}1, which makes loading small negative
12102 constants and certain bitmasks faster.
12103 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
12104 which specify use of that register as a fixed register,
12105 and @samp{none}, which means that no register is used for this
12106 purpose. The default is @option{-m1reg-none}.
12111 @subsection ARC Options
12112 @cindex ARC options
12114 The following options control the architecture variant for which code
12117 @c architecture variants
12120 @item -mbarrel-shifter
12121 @opindex mbarrel-shifter
12122 Generate instructions supported by barrel shifter. This is the default
12123 unless @samp{-mcpu=ARC601} is in effect.
12125 @item -mcpu=@var{cpu}
12127 Set architecture type, register usage, and instruction scheduling
12128 parameters for @var{cpu}. There are also shortcut alias options
12129 available for backward compatibility and convenience. Supported
12130 values for @var{cpu} are
12136 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
12140 Compile for ARC601. Alias: @option{-mARC601}.
12145 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
12146 This is the default when configured with @samp{--with-cpu=arc700}@.
12151 @itemx -mdpfp-compact
12152 @opindex mdpfp-compact
12153 FPX: Generate Double Precision FPX instructions, tuned for the compact
12157 @opindex mdpfp-fast
12158 FPX: Generate Double Precision FPX instructions, tuned for the fast
12161 @item -mno-dpfp-lrsr
12162 @opindex mno-dpfp-lrsr
12163 Disable LR and SR instructions from using FPX extension aux registers.
12167 Generate Extended arithmetic instructions. Currently only
12168 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
12169 supported. This is always enabled for @samp{-mcpu=ARC700}.
12173 Do not generate mpy instructions for ARC700.
12177 Generate 32x16 bit multiply and mac instructions.
12181 Generate mul64 and mulu64 instructions. Only valid for @samp{-mcpu=ARC600}.
12185 Generate norm instruction. This is the default if @samp{-mcpu=ARC700}
12190 @itemx -mspfp-compact
12191 @opindex mspfp-compact
12192 FPX: Generate Single Precision FPX instructions, tuned for the compact
12196 @opindex mspfp-fast
12197 FPX: Generate Single Precision FPX instructions, tuned for the fast
12202 Enable generation of ARC SIMD instructions via target-specific
12203 builtins. Only valid for @samp{-mcpu=ARC700}.
12206 @opindex msoft-float
12207 This option ignored; it is provided for compatibility purposes only.
12208 Software floating point code is emitted by default, and this default
12209 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
12210 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
12211 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
12215 Generate swap instructions.
12219 The following options are passed through to the assembler, and also
12220 define preprocessor macro symbols.
12222 @c Flags used by the assembler, but for which we define preprocessor
12223 @c macro symbols as well.
12226 @opindex mdsp-packa
12227 Passed down to the assembler to enable the DSP Pack A extensions.
12228 Also sets the preprocessor symbol @code{__Xdsp_packa}.
12232 Passed down to the assembler to enable the dual viterbi butterfly
12233 extension. Also sets the preprocessor symbol @code{__Xdvbf}.
12235 @c ARC700 4.10 extension instruction
12238 Passed down to the assembler to enable the Locked Load/Store
12239 Conditional extension. Also sets the preprocessor symbol
12244 Passed down to the assembler. Also sets the preprocessor symbol
12245 @code{__Xxmac_d16}.
12249 Passed down to the assembler. Also sets the preprocessor symbol
12252 @c ARC700 4.10 extension instruction
12255 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
12256 extension instruction. Also sets the preprocessor symbol
12259 @c ARC700 4.10 extension instruction
12262 Passed down to the assembler to enable the swap byte ordering
12263 extension instruction. Also sets the preprocessor symbol
12267 @opindex mtelephony
12268 Passed down to the assembler to enable dual and single operand
12269 instructions for telephony. Also sets the preprocessor symbol
12270 @code{__Xtelephony}.
12274 Passed down to the assembler to enable the XY Memory extension. Also
12275 sets the preprocessor symbol @code{__Xxy}.
12279 The following options control how the assembly code is annotated:
12281 @c Assembly annotation options
12285 Annotate assembler instructions with estimated addresses.
12287 @item -mannotate-align
12288 @opindex mannotate-align
12289 Explain what alignment considerations lead to the decision to make an
12290 instruction short or long.
12294 The following options are passed through to the linker:
12296 @c options passed through to the linker
12300 Passed through to the linker, to specify use of the @code{arclinux} emulation.
12301 This option is enabled by default in tool chains built for
12302 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
12303 when profiling is not requested.
12305 @item -marclinux_prof
12306 @opindex marclinux_prof
12307 Passed through to the linker, to specify use of the
12308 @code{arclinux_prof} emulation. This option is enabled by default in
12309 tool chains built for @w{@code{arc-linux-uclibc}} and
12310 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
12314 The following options control the semantics of generated code:
12316 @c semantically relevant code generation options
12318 @item -mepilogue-cfi
12319 @opindex mepilogue-cfi
12320 Enable generation of call frame information for epilogues.
12322 @item -mno-epilogue-cfi
12323 @opindex mno-epilogue-cfi
12324 Disable generation of call frame information for epilogues.
12327 @opindex mlong-calls
12328 Generate call insns as register indirect calls, thus providing access
12329 to the full 32-bit address range.
12331 @item -mmedium-calls
12332 @opindex mmedium-calls
12333 Don't use less than 25 bit addressing range for calls, which is the
12334 offset available for an unconditional branch-and-link
12335 instruction. Conditional execution of function calls is suppressed, to
12336 allow use of the 25-bit range, rather than the 21-bit range with
12337 conditional branch-and-link. This is the default for tool chains built
12338 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
12342 Do not generate sdata references. This is the default for tool chains
12343 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
12347 @opindex mucb-mcount
12348 Instrument with mcount calls as used in UCB code. I.e. do the
12349 counting in the callee, not the caller. By default ARC instrumentation
12350 counts in the caller.
12352 @item -mvolatile-cache
12353 @opindex mvolatile-cache
12354 Use ordinarily cached memory accesses for volatile references. This is the
12357 @item -mno-volatile-cache
12358 @opindex mno-volatile-cache
12359 Enable cache bypass for volatile references.
12363 The following options fine tune code generation:
12364 @c code generation tuning options
12367 @opindex malign-call
12368 Do alignment optimizations for call instructions.
12370 @item -mauto-modify-reg
12371 @opindex mauto-modify-reg
12372 Enable the use of pre/post modify with register displacement.
12374 @item -mbbit-peephole
12375 @opindex mbbit-peephole
12376 Enable bbit peephole2.
12380 This option disables a target-specific pass in @file{arc_reorg} to
12381 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
12382 generation driven by the combiner pass.
12384 @item -mcase-vector-pcrel
12385 @opindex mcase-vector-pcrel
12386 Use pc-relative switch case tables - this enables case table shortening.
12387 This is the default for @option{-Os}.
12389 @item -mcompact-casesi
12390 @opindex mcompact-casesi
12391 Enable compact casesi pattern.
12392 This is the default for @option{-Os}.
12394 @item -mno-cond-exec
12395 @opindex mno-cond-exec
12396 Disable ARCompact specific pass to generate conditional execution instructions.
12397 Due to delay slot scheduling and interactions between operand numbers,
12398 literal sizes, instruction lengths, and the support for conditional execution,
12399 the target-independent pass to generate conditional execution is often lacking,
12400 so the ARC port has kept a special pass around that tries to find more
12401 conditional execution generating opportunities after register allocation,
12402 branch shortening, and delay slot scheduling have been done. This pass
12403 generally, but not always, improves performance and code size, at the cost of
12404 extra compilation time, which is why there is an option to switch it off.
12405 If you have a problem with call instructions exceeding their allowable
12406 offset range because they are conditionalized, you should consider using
12407 @option{-mmedium-calls} instead.
12409 @item -mearly-cbranchsi
12410 @opindex mearly-cbranchsi
12411 Enable pre-reload use of the cbranchsi pattern.
12413 @item -mexpand-adddi
12414 @opindex mexpand-adddi
12415 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
12416 @code{add.f}, @code{adc} etc.
12418 @item -mindexed-loads
12419 @opindex mindexed-loads
12420 Enable the use of indexed loads. This can be problematic because some
12421 optimizers will then assume the that indexed stores exist, which is not
12426 Enable Local Register Allocation. This is still experimental for ARC,
12427 so by default the compiler uses standard reload
12428 (i.e. @samp{-mno-lra}).
12430 @item -mlra-priority-none
12431 @opindex mlra-priority-none
12432 Don't indicate any priority for target registers.
12434 @item -mlra-priority-compact
12435 @opindex mlra-priority-compact
12436 Indicate target register priority for r0..r3 / r12..r15.
12438 @item -mlra-priority-noncompact
12439 @opindex mlra-priority-noncompact
12440 Reduce target regsiter priority for r0..r3 / r12..r15.
12442 @item -mno-millicode
12443 @opindex mno-millicode
12444 When optimizing for size (using @option{-Os}), prologues and epilogues
12445 that have to save or restore a large number of registers are often
12446 shortened by using call to a special function in libgcc; this is
12447 referred to as a @emph{millicode} call. As these calls can pose
12448 performance issues, and/or cause linking issues when linking in a
12449 nonstandard way, this option is provided to turn off millicode call
12453 @opindex mmixed-code
12454 Tweak register allocation to help 16-bit instruction generation.
12455 This generally has the effect of decreasing the average instruction size
12456 while increasing the instruction count.
12460 Enable 'q' instruction alternatives.
12461 This is the default for @option{-Os}.
12465 Enable Rcq constraint handling - most short code generation depends on this.
12466 This is the default.
12470 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
12471 This is the default.
12473 @item -msize-level=@var{level}
12474 @opindex msize-level
12475 Fine-tune size optimization with regards to instruction lengths and alignment.
12476 The recognized values for @var{level} are:
12479 No size optimization. This level is deprecated and treated like @samp{1}.
12482 Short instructions are used opportunistically.
12485 In addition, alignment of loops and of code after barriers are dropped.
12488 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
12492 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
12493 the behavior when this is not set is equivalent to level @samp{1}.
12495 @item -mtune=@var{cpu}
12497 Set instruction scheduling parameters for @var{cpu}, overriding any implied
12498 by @option{-mcpu=}.
12500 Supported values for @var{cpu} are
12504 Tune for ARC600 cpu.
12507 Tune for ARC601 cpu.
12510 Tune for ARC700 cpu with standard multiplier block.
12513 Tune for ARC700 cpu with XMAC block.
12516 Tune for ARC725D cpu.
12519 Tune for ARC750D cpu.
12523 @item -mmultcost=@var{num}
12525 Cost to assume for a multiply instruction, with @samp{4} being equal to a
12526 normal instruction.
12528 @item -munalign-prob-threshold=@var{probability}
12529 @opindex munalign-prob-threshold
12530 Set probability threshold for unaligning branches.
12531 When tuning for @samp{ARC700} and optimizing for speed, branches without
12532 filled delay slot are preferably emitted unaligned and long, unless
12533 profiling indicates that the probability for the branch to be taken
12534 is below @var{probability}. @xref{Cross-profiling}.
12535 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
12539 The following options are maintained for backward compatibility, but
12540 are now deprecated and will be removed in a future release:
12542 @c Deprecated options
12550 @opindex mbig-endian
12553 Compile code for big endian targets. Use of these options is now
12554 deprecated. Users wanting big-endian code, should use the
12555 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
12556 building the tool chain, for which big-endian is the default.
12558 @item -mlittle-endian
12559 @opindex mlittle-endian
12562 Compile code for little endian targets. Use of these options is now
12563 deprecated. Users wanting little-endian code should use the
12564 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
12565 building the tool chain, for which little-endian is the default.
12567 @item -mbarrel_shifter
12568 @opindex mbarrel_shifter
12569 Replaced by @samp{-mbarrel-shifter}
12571 @item -mdpfp_compact
12572 @opindex mdpfp_compact
12573 Replaced by @samp{-mdpfp-compact}
12576 @opindex mdpfp_fast
12577 Replaced by @samp{-mdpfp-fast}
12580 @opindex mdsp_packa
12581 Replaced by @samp{-mdsp-packa}
12585 Replaced by @samp{-mea}
12589 Replaced by @samp{-mmac-24}
12593 Replaced by @samp{-mmac-d16}
12595 @item -mspfp_compact
12596 @opindex mspfp_compact
12597 Replaced by @samp{-mspfp-compact}
12600 @opindex mspfp_fast
12601 Replaced by @samp{-mspfp-fast}
12603 @item -mtune=@var{cpu}
12605 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
12606 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
12607 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
12609 @item -multcost=@var{num}
12611 Replaced by @samp{-mmultcost}.
12616 @subsection ARM Options
12617 @cindex ARM options
12619 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
12623 @item -mabi=@var{name}
12625 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
12626 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
12629 @opindex mapcs-frame
12630 Generate a stack frame that is compliant with the ARM Procedure Call
12631 Standard for all functions, even if this is not strictly necessary for
12632 correct execution of the code. Specifying @option{-fomit-frame-pointer}
12633 with this option causes the stack frames not to be generated for
12634 leaf functions. The default is @option{-mno-apcs-frame}.
12638 This is a synonym for @option{-mapcs-frame}.
12641 @c not currently implemented
12642 @item -mapcs-stack-check
12643 @opindex mapcs-stack-check
12644 Generate code to check the amount of stack space available upon entry to
12645 every function (that actually uses some stack space). If there is
12646 insufficient space available then either the function
12647 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} is
12648 called, depending upon the amount of stack space required. The runtime
12649 system is required to provide these functions. The default is
12650 @option{-mno-apcs-stack-check}, since this produces smaller code.
12652 @c not currently implemented
12654 @opindex mapcs-float
12655 Pass floating-point arguments using the floating-point registers. This is
12656 one of the variants of the APCS@. This option is recommended if the
12657 target hardware has a floating-point unit or if a lot of floating-point
12658 arithmetic is going to be performed by the code. The default is
12659 @option{-mno-apcs-float}, since the size of integer-only code is
12660 slightly increased if @option{-mapcs-float} is used.
12662 @c not currently implemented
12663 @item -mapcs-reentrant
12664 @opindex mapcs-reentrant
12665 Generate reentrant, position-independent code. The default is
12666 @option{-mno-apcs-reentrant}.
12669 @item -mthumb-interwork
12670 @opindex mthumb-interwork
12671 Generate code that supports calling between the ARM and Thumb
12672 instruction sets. Without this option, on pre-v5 architectures, the
12673 two instruction sets cannot be reliably used inside one program. The
12674 default is @option{-mno-thumb-interwork}, since slightly larger code
12675 is generated when @option{-mthumb-interwork} is specified. In AAPCS
12676 configurations this option is meaningless.
12678 @item -mno-sched-prolog
12679 @opindex mno-sched-prolog
12680 Prevent the reordering of instructions in the function prologue, or the
12681 merging of those instruction with the instructions in the function's
12682 body. This means that all functions start with a recognizable set
12683 of instructions (or in fact one of a choice from a small set of
12684 different function prologues), and this information can be used to
12685 locate the start of functions inside an executable piece of code. The
12686 default is @option{-msched-prolog}.
12688 @item -mfloat-abi=@var{name}
12689 @opindex mfloat-abi
12690 Specifies which floating-point ABI to use. Permissible values
12691 are: @samp{soft}, @samp{softfp} and @samp{hard}.
12693 Specifying @samp{soft} causes GCC to generate output containing
12694 library calls for floating-point operations.
12695 @samp{softfp} allows the generation of code using hardware floating-point
12696 instructions, but still uses the soft-float calling conventions.
12697 @samp{hard} allows generation of floating-point instructions
12698 and uses FPU-specific calling conventions.
12700 The default depends on the specific target configuration. Note that
12701 the hard-float and soft-float ABIs are not link-compatible; you must
12702 compile your entire program with the same ABI, and link with a
12703 compatible set of libraries.
12705 @item -mlittle-endian
12706 @opindex mlittle-endian
12707 Generate code for a processor running in little-endian mode. This is
12708 the default for all standard configurations.
12711 @opindex mbig-endian
12712 Generate code for a processor running in big-endian mode; the default is
12713 to compile code for a little-endian processor.
12715 @item -march=@var{name}
12717 This specifies the name of the target ARM architecture. GCC uses this
12718 name to determine what kind of instructions it can emit when generating
12719 assembly code. This option can be used in conjunction with or instead
12720 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
12721 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
12722 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
12723 @samp{armv6}, @samp{armv6j},
12724 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
12725 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m}, @samp{armv7e-m},
12726 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc},
12727 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
12729 @option{-march=armv7ve} is the armv7-a architecture with virtualization
12732 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
12733 architecture together with the optional CRC32 extensions.
12735 @option{-march=native} causes the compiler to auto-detect the architecture
12736 of the build computer. At present, this feature is only supported on
12737 GNU/Linux, and not all architectures are recognized. If the auto-detect
12738 is unsuccessful the option has no effect.
12740 @item -mtune=@var{name}
12742 This option specifies the name of the target ARM processor for
12743 which GCC should tune the performance of the code.
12744 For some ARM implementations better performance can be obtained by using
12746 Permissible names are: @samp{arm2}, @samp{arm250},
12747 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
12748 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
12749 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
12750 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
12752 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
12753 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
12754 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
12755 @samp{strongarm1110},
12756 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
12757 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
12758 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
12759 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
12760 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
12761 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
12762 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
12763 @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8}, @samp{cortex-a9},
12764 @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a53}, @samp{cortex-a57},
12766 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-m7},
12771 @samp{cortex-m0plus},
12772 @samp{cortex-m1.small-multiply},
12773 @samp{cortex-m0.small-multiply},
12774 @samp{cortex-m0plus.small-multiply},
12775 @samp{marvell-pj4},
12776 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
12777 @samp{fa526}, @samp{fa626},
12778 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}.
12780 Additionally, this option can specify that GCC should tune the performance
12781 of the code for a big.LITTLE system. Permissible names are:
12782 @samp{cortex-a15.cortex-a7}, @samp{cortex-a57.cortex-a53}.
12784 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
12785 performance for a blend of processors within architecture @var{arch}.
12786 The aim is to generate code that run well on the current most popular
12787 processors, balancing between optimizations that benefit some CPUs in the
12788 range, and avoiding performance pitfalls of other CPUs. The effects of
12789 this option may change in future GCC versions as CPU models come and go.
12791 @option{-mtune=native} causes the compiler to auto-detect the CPU
12792 of the build computer. At present, this feature is only supported on
12793 GNU/Linux, and not all architectures are recognized. If the auto-detect is
12794 unsuccessful the option has no effect.
12796 @item -mcpu=@var{name}
12798 This specifies the name of the target ARM processor. GCC uses this name
12799 to derive the name of the target ARM architecture (as if specified
12800 by @option{-march}) and the ARM processor type for which to tune for
12801 performance (as if specified by @option{-mtune}). Where this option
12802 is used in conjunction with @option{-march} or @option{-mtune},
12803 those options take precedence over the appropriate part of this option.
12805 Permissible names for this option are the same as those for
12808 @option{-mcpu=generic-@var{arch}} is also permissible, and is
12809 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
12810 See @option{-mtune} for more information.
12812 @option{-mcpu=native} causes the compiler to auto-detect the CPU
12813 of the build computer. At present, this feature is only supported on
12814 GNU/Linux, and not all architectures are recognized. If the auto-detect
12815 is unsuccessful the option has no effect.
12817 @item -mfpu=@var{name}
12819 This specifies what floating-point hardware (or hardware emulation) is
12820 available on the target. Permissible names are: @samp{vfp}, @samp{vfpv3},
12821 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
12822 @samp{vfpv3xd-fp16}, @samp{neon}, @samp{neon-fp16}, @samp{vfpv4},
12823 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
12824 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
12825 @samp{fp-armv8}, @samp{neon-fp-armv8}, and @samp{crypto-neon-fp-armv8}.
12827 If @option{-msoft-float} is specified this specifies the format of
12828 floating-point values.
12830 If the selected floating-point hardware includes the NEON extension
12831 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
12832 operations are not generated by GCC's auto-vectorization pass unless
12833 @option{-funsafe-math-optimizations} is also specified. This is
12834 because NEON hardware does not fully implement the IEEE 754 standard for
12835 floating-point arithmetic (in particular denormal values are treated as
12836 zero), so the use of NEON instructions may lead to a loss of precision.
12838 @item -mfp16-format=@var{name}
12839 @opindex mfp16-format
12840 Specify the format of the @code{__fp16} half-precision floating-point type.
12841 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
12842 the default is @samp{none}, in which case the @code{__fp16} type is not
12843 defined. @xref{Half-Precision}, for more information.
12845 @item -mstructure-size-boundary=@var{n}
12846 @opindex mstructure-size-boundary
12847 The sizes of all structures and unions are rounded up to a multiple
12848 of the number of bits set by this option. Permissible values are 8, 32
12849 and 64. The default value varies for different toolchains. For the COFF
12850 targeted toolchain the default value is 8. A value of 64 is only allowed
12851 if the underlying ABI supports it.
12853 Specifying a larger number can produce faster, more efficient code, but
12854 can also increase the size of the program. Different values are potentially
12855 incompatible. Code compiled with one value cannot necessarily expect to
12856 work with code or libraries compiled with another value, if they exchange
12857 information using structures or unions.
12859 @item -mabort-on-noreturn
12860 @opindex mabort-on-noreturn
12861 Generate a call to the function @code{abort} at the end of a
12862 @code{noreturn} function. It is executed if the function tries to
12866 @itemx -mno-long-calls
12867 @opindex mlong-calls
12868 @opindex mno-long-calls
12869 Tells the compiler to perform function calls by first loading the
12870 address of the function into a register and then performing a subroutine
12871 call on this register. This switch is needed if the target function
12872 lies outside of the 64-megabyte addressing range of the offset-based
12873 version of subroutine call instruction.
12875 Even if this switch is enabled, not all function calls are turned
12876 into long calls. The heuristic is that static functions, functions
12877 that have the @samp{short-call} attribute, functions that are inside
12878 the scope of a @samp{#pragma no_long_calls} directive, and functions whose
12879 definitions have already been compiled within the current compilation
12880 unit are not turned into long calls. The exceptions to this rule are
12881 that weak function definitions, functions with the @samp{long-call}
12882 attribute or the @samp{section} attribute, and functions that are within
12883 the scope of a @samp{#pragma long_calls} directive are always
12884 turned into long calls.
12886 This feature is not enabled by default. Specifying
12887 @option{-mno-long-calls} restores the default behavior, as does
12888 placing the function calls within the scope of a @samp{#pragma
12889 long_calls_off} directive. Note these switches have no effect on how
12890 the compiler generates code to handle function calls via function
12893 @item -msingle-pic-base
12894 @opindex msingle-pic-base
12895 Treat the register used for PIC addressing as read-only, rather than
12896 loading it in the prologue for each function. The runtime system is
12897 responsible for initializing this register with an appropriate value
12898 before execution begins.
12900 @item -mpic-register=@var{reg}
12901 @opindex mpic-register
12902 Specify the register to be used for PIC addressing.
12903 For standard PIC base case, the default will be any suitable register
12904 determined by compiler. For single PIC base case, the default is
12905 @samp{R9} if target is EABI based or stack-checking is enabled,
12906 otherwise the default is @samp{R10}.
12908 @item -mpic-data-is-text-relative
12909 @opindex mpic-data-is-text-relative
12910 Assume that each data segments are relative to text segment at load time.
12911 Therefore, it permits addressing data using PC-relative operations.
12912 This option is on by default for targets other than VxWorks RTP.
12914 @item -mpoke-function-name
12915 @opindex mpoke-function-name
12916 Write the name of each function into the text section, directly
12917 preceding the function prologue. The generated code is similar to this:
12921 .ascii "arm_poke_function_name", 0
12924 .word 0xff000000 + (t1 - t0)
12925 arm_poke_function_name
12927 stmfd sp!, @{fp, ip, lr, pc@}
12931 When performing a stack backtrace, code can inspect the value of
12932 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
12933 location @code{pc - 12} and the top 8 bits are set, then we know that
12934 there is a function name embedded immediately preceding this location
12935 and has length @code{((pc[-3]) & 0xff000000)}.
12942 Select between generating code that executes in ARM and Thumb
12943 states. The default for most configurations is to generate code
12944 that executes in ARM state, but the default can be changed by
12945 configuring GCC with the @option{--with-mode=}@var{state}
12949 @opindex mtpcs-frame
12950 Generate a stack frame that is compliant with the Thumb Procedure Call
12951 Standard for all non-leaf functions. (A leaf function is one that does
12952 not call any other functions.) The default is @option{-mno-tpcs-frame}.
12954 @item -mtpcs-leaf-frame
12955 @opindex mtpcs-leaf-frame
12956 Generate a stack frame that is compliant with the Thumb Procedure Call
12957 Standard for all leaf functions. (A leaf function is one that does
12958 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
12960 @item -mcallee-super-interworking
12961 @opindex mcallee-super-interworking
12962 Gives all externally visible functions in the file being compiled an ARM
12963 instruction set header which switches to Thumb mode before executing the
12964 rest of the function. This allows these functions to be called from
12965 non-interworking code. This option is not valid in AAPCS configurations
12966 because interworking is enabled by default.
12968 @item -mcaller-super-interworking
12969 @opindex mcaller-super-interworking
12970 Allows calls via function pointers (including virtual functions) to
12971 execute correctly regardless of whether the target code has been
12972 compiled for interworking or not. There is a small overhead in the cost
12973 of executing a function pointer if this option is enabled. This option
12974 is not valid in AAPCS configurations because interworking is enabled
12977 @item -mtp=@var{name}
12979 Specify the access model for the thread local storage pointer. The valid
12980 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
12981 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
12982 (supported in the arm6k architecture), and @option{auto}, which uses the
12983 best available method for the selected processor. The default setting is
12986 @item -mtls-dialect=@var{dialect}
12987 @opindex mtls-dialect
12988 Specify the dialect to use for accessing thread local storage. Two
12989 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
12990 @samp{gnu} dialect selects the original GNU scheme for supporting
12991 local and global dynamic TLS models. The @samp{gnu2} dialect
12992 selects the GNU descriptor scheme, which provides better performance
12993 for shared libraries. The GNU descriptor scheme is compatible with
12994 the original scheme, but does require new assembler, linker and
12995 library support. Initial and local exec TLS models are unaffected by
12996 this option and always use the original scheme.
12998 @item -mword-relocations
12999 @opindex mword-relocations
13000 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
13001 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
13002 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
13005 @item -mfix-cortex-m3-ldrd
13006 @opindex mfix-cortex-m3-ldrd
13007 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
13008 with overlapping destination and base registers are used. This option avoids
13009 generating these instructions. This option is enabled by default when
13010 @option{-mcpu=cortex-m3} is specified.
13012 @item -munaligned-access
13013 @itemx -mno-unaligned-access
13014 @opindex munaligned-access
13015 @opindex mno-unaligned-access
13016 Enables (or disables) reading and writing of 16- and 32- bit values
13017 from addresses that are not 16- or 32- bit aligned. By default
13018 unaligned access is disabled for all pre-ARMv6 and all ARMv6-M
13019 architectures, and enabled for all other architectures. If unaligned
13020 access is not enabled then words in packed data structures will be
13021 accessed a byte at a time.
13023 The ARM attribute @code{Tag_CPU_unaligned_access} will be set in the
13024 generated object file to either true or false, depending upon the
13025 setting of this option. If unaligned access is enabled then the
13026 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} will also be
13029 @item -mneon-for-64bits
13030 @opindex mneon-for-64bits
13031 Enables using Neon to handle scalar 64-bits operations. This is
13032 disabled by default since the cost of moving data from core registers
13035 @item -mslow-flash-data
13036 @opindex mslow-flash-data
13037 Assume loading data from flash is slower than fetching instruction.
13038 Therefore literal load is minimized for better performance.
13039 This option is only supported when compiling for ARMv7 M-profile and
13042 @item -masm-syntax-unified
13043 @opindex masm-syntax-unified
13044 Assume inline assembler is using unified asm syntax. The default is
13045 currently off which implies divided syntax. Currently this option is
13046 available only for Thumb1 and has no effect on ARM state and Thumb2.
13047 However, this may change in future releases of GCC. Divided syntax
13048 should be considered deprecated.
13050 @item -mrestrict-it
13051 @opindex mrestrict-it
13052 Restricts generation of IT blocks to conform to the rules of ARMv8.
13053 IT blocks can only contain a single 16-bit instruction from a select
13054 set of instructions. This option is on by default for ARMv8 Thumb mode.
13058 @subsection AVR Options
13059 @cindex AVR Options
13061 These options are defined for AVR implementations:
13064 @item -mmcu=@var{mcu}
13066 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
13068 The default for this option is@tie{}@code{avr2}.
13070 GCC supports the following AVR devices and ISAs:
13072 @include avr-mmcu.texi
13074 @item -maccumulate-args
13075 @opindex maccumulate-args
13076 Accumulate outgoing function arguments and acquire/release the needed
13077 stack space for outgoing function arguments once in function
13078 prologue/epilogue. Without this option, outgoing arguments are pushed
13079 before calling a function and popped afterwards.
13081 Popping the arguments after the function call can be expensive on
13082 AVR so that accumulating the stack space might lead to smaller
13083 executables because arguments need not to be removed from the
13084 stack after such a function call.
13086 This option can lead to reduced code size for functions that perform
13087 several calls to functions that get their arguments on the stack like
13088 calls to printf-like functions.
13090 @item -mbranch-cost=@var{cost}
13091 @opindex mbranch-cost
13092 Set the branch costs for conditional branch instructions to
13093 @var{cost}. Reasonable values for @var{cost} are small, non-negative
13094 integers. The default branch cost is 0.
13096 @item -mcall-prologues
13097 @opindex mcall-prologues
13098 Functions prologues/epilogues are expanded as calls to appropriate
13099 subroutines. Code size is smaller.
13103 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
13104 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
13105 and @code{long long} is 4 bytes. Please note that this option does not
13106 conform to the C standards, but it results in smaller code
13109 @item -mno-interrupts
13110 @opindex mno-interrupts
13111 Generated code is not compatible with hardware interrupts.
13112 Code size is smaller.
13116 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
13117 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
13118 Setting @code{-mrelax} just adds the @code{--relax} option to the
13119 linker command line when the linker is called.
13121 Jump relaxing is performed by the linker because jump offsets are not
13122 known before code is located. Therefore, the assembler code generated by the
13123 compiler is the same, but the instructions in the executable may
13124 differ from instructions in the assembler code.
13126 Relaxing must be turned on if linker stubs are needed, see the
13127 section on @code{EIND} and linker stubs below.
13131 Treat the stack pointer register as an 8-bit register,
13132 i.e.@: assume the high byte of the stack pointer is zero.
13133 In general, you don't need to set this option by hand.
13135 This option is used internally by the compiler to select and
13136 build multilibs for architectures @code{avr2} and @code{avr25}.
13137 These architectures mix devices with and without @code{SPH}.
13138 For any setting other than @code{-mmcu=avr2} or @code{-mmcu=avr25}
13139 the compiler driver will add or remove this option from the compiler
13140 proper's command line, because the compiler then knows if the device
13141 or architecture has an 8-bit stack pointer and thus no @code{SPH}
13146 Use address register @code{X} in a way proposed by the hardware. This means
13147 that @code{X} is only used in indirect, post-increment or
13148 pre-decrement addressing.
13150 Without this option, the @code{X} register may be used in the same way
13151 as @code{Y} or @code{Z} which then is emulated by additional
13153 For example, loading a value with @code{X+const} addressing with a
13154 small non-negative @code{const < 64} to a register @var{Rn} is
13158 adiw r26, const ; X += const
13159 ld @var{Rn}, X ; @var{Rn} = *X
13160 sbiw r26, const ; X -= const
13164 @opindex mtiny-stack
13165 Only change the lower 8@tie{}bits of the stack pointer.
13167 @item -Waddr-space-convert
13168 @opindex Waddr-space-convert
13169 Warn about conversions between address spaces in the case where the
13170 resulting address space is not contained in the incoming address space.
13173 @subsubsection @code{EIND} and Devices with more than 128 Ki Bytes of Flash
13174 @cindex @code{EIND}
13175 Pointers in the implementation are 16@tie{}bits wide.
13176 The address of a function or label is represented as word address so
13177 that indirect jumps and calls can target any code address in the
13178 range of 64@tie{}Ki words.
13180 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
13181 bytes of program memory space, there is a special function register called
13182 @code{EIND} that serves as most significant part of the target address
13183 when @code{EICALL} or @code{EIJMP} instructions are used.
13185 Indirect jumps and calls on these devices are handled as follows by
13186 the compiler and are subject to some limitations:
13191 The compiler never sets @code{EIND}.
13194 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
13195 instructions or might read @code{EIND} directly in order to emulate an
13196 indirect call/jump by means of a @code{RET} instruction.
13199 The compiler assumes that @code{EIND} never changes during the startup
13200 code or during the application. In particular, @code{EIND} is not
13201 saved/restored in function or interrupt service routine
13205 For indirect calls to functions and computed goto, the linker
13206 generates @emph{stubs}. Stubs are jump pads sometimes also called
13207 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
13208 The stub contains a direct jump to the desired address.
13211 Linker relaxation must be turned on so that the linker will generate
13212 the stubs correctly an all situaltion. See the compiler option
13213 @code{-mrelax} and the linler option @code{--relax}.
13214 There are corner cases where the linker is supposed to generate stubs
13215 but aborts without relaxation and without a helpful error message.
13218 The default linker script is arranged for code with @code{EIND = 0}.
13219 If code is supposed to work for a setup with @code{EIND != 0}, a custom
13220 linker script has to be used in order to place the sections whose
13221 name start with @code{.trampolines} into the segment where @code{EIND}
13225 The startup code from libgcc never sets @code{EIND}.
13226 Notice that startup code is a blend of code from libgcc and AVR-LibC.
13227 For the impact of AVR-LibC on @code{EIND}, see the
13228 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
13231 It is legitimate for user-specific startup code to set up @code{EIND}
13232 early, for example by means of initialization code located in
13233 section @code{.init3}. Such code runs prior to general startup code
13234 that initializes RAM and calls constructors, but after the bit
13235 of startup code from AVR-LibC that sets @code{EIND} to the segment
13236 where the vector table is located.
13238 #include <avr/io.h>
13241 __attribute__((section(".init3"),naked,used,no_instrument_function))
13242 init3_set_eind (void)
13244 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
13245 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
13250 The @code{__trampolines_start} symbol is defined in the linker script.
13253 Stubs are generated automatically by the linker if
13254 the following two conditions are met:
13257 @item The address of a label is taken by means of the @code{gs} modifier
13258 (short for @emph{generate stubs}) like so:
13260 LDI r24, lo8(gs(@var{func}))
13261 LDI r25, hi8(gs(@var{func}))
13263 @item The final location of that label is in a code segment
13264 @emph{outside} the segment where the stubs are located.
13268 The compiler emits such @code{gs} modifiers for code labels in the
13269 following situations:
13271 @item Taking address of a function or code label.
13272 @item Computed goto.
13273 @item If prologue-save function is used, see @option{-mcall-prologues}
13274 command-line option.
13275 @item Switch/case dispatch tables. If you do not want such dispatch
13276 tables you can specify the @option{-fno-jump-tables} command-line option.
13277 @item C and C++ constructors/destructors called during startup/shutdown.
13278 @item If the tools hit a @code{gs()} modifier explained above.
13282 Jumping to non-symbolic addresses like so is @emph{not} supported:
13287 /* Call function at word address 0x2 */
13288 return ((int(*)(void)) 0x2)();
13292 Instead, a stub has to be set up, i.e.@: the function has to be called
13293 through a symbol (@code{func_4} in the example):
13298 extern int func_4 (void);
13300 /* Call function at byte address 0x4 */
13305 and the application be linked with @code{-Wl,--defsym,func_4=0x4}.
13306 Alternatively, @code{func_4} can be defined in the linker script.
13309 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
13310 @cindex @code{RAMPD}
13311 @cindex @code{RAMPX}
13312 @cindex @code{RAMPY}
13313 @cindex @code{RAMPZ}
13314 Some AVR devices support memories larger than the 64@tie{}KiB range
13315 that can be accessed with 16-bit pointers. To access memory locations
13316 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
13317 register is used as high part of the address:
13318 The @code{X}, @code{Y}, @code{Z} address register is concatenated
13319 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
13320 register, respectively, to get a wide address. Similarly,
13321 @code{RAMPD} is used together with direct addressing.
13325 The startup code initializes the @code{RAMP} special function
13326 registers with zero.
13329 If a @ref{AVR Named Address Spaces,named address space} other than
13330 generic or @code{__flash} is used, then @code{RAMPZ} is set
13331 as needed before the operation.
13334 If the device supports RAM larger than 64@tie{}KiB and the compiler
13335 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
13336 is reset to zero after the operation.
13339 If the device comes with a specific @code{RAMP} register, the ISR
13340 prologue/epilogue saves/restores that SFR and initializes it with
13341 zero in case the ISR code might (implicitly) use it.
13344 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
13345 If you use inline assembler to read from locations outside the
13346 16-bit address range and change one of the @code{RAMP} registers,
13347 you must reset it to zero after the access.
13351 @subsubsection AVR Built-in Macros
13353 GCC defines several built-in macros so that the user code can test
13354 for the presence or absence of features. Almost any of the following
13355 built-in macros are deduced from device capabilities and thus
13356 triggered by the @code{-mmcu=} command-line option.
13358 For even more AVR-specific built-in macros see
13359 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
13364 Build-in macro that resolves to a decimal number that identifies the
13365 architecture and depends on the @code{-mmcu=@var{mcu}} option.
13366 Possible values are:
13368 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
13369 @code{4}, @code{5}, @code{51}, @code{6}, @code{102}, @code{104},
13370 @code{105}, @code{106}, @code{107}
13372 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3},
13373 @code{avr31}, @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51},
13374 @code{avr6}, @code{avrxmega2}, @code{avrxmega4}, @code{avrxmega5},
13375 @code{avrxmega6}, @code{avrxmega7}, respectively.
13376 If @var{mcu} specifies a device, this built-in macro is set
13377 accordingly. For example, with @code{-mmcu=atmega8} the macro will be
13378 defined to @code{4}.
13380 @item __AVR_@var{Device}__
13381 Setting @code{-mmcu=@var{device}} defines this built-in macro which reflects
13382 the device's name. For example, @code{-mmcu=atmega8} defines the
13383 built-in macro @code{__AVR_ATmega8__}, @code{-mmcu=attiny261a} defines
13384 @code{__AVR_ATtiny261A__}, etc.
13386 The built-in macros' names follow
13387 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
13388 the device name as from the AVR user manual. The difference between
13389 @var{Device} in the built-in macro and @var{device} in
13390 @code{-mmcu=@var{device}} is that the latter is always lowercase.
13392 If @var{device} is not a device but only a core architecture like
13393 @code{avr51}, this macro will not be defined.
13395 @item __AVR_DEVICE_NAME__
13396 Setting @code{-mmcu=@var{device}} defines this built-in macro to
13397 the device's name. For example, with @code{-mmcu=atmega8} the macro
13398 will be defined to @code{atmega8}.
13400 If @var{device} is not a device but only a core architecture like
13401 @code{avr51}, this macro will not be defined.
13403 @item __AVR_XMEGA__
13404 The device / architecture belongs to the XMEGA family of devices.
13406 @item __AVR_HAVE_ELPM__
13407 The device has the the @code{ELPM} instruction.
13409 @item __AVR_HAVE_ELPMX__
13410 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
13411 R@var{n},Z+} instructions.
13413 @item __AVR_HAVE_MOVW__
13414 The device has the @code{MOVW} instruction to perform 16-bit
13415 register-register moves.
13417 @item __AVR_HAVE_LPMX__
13418 The device has the @code{LPM R@var{n},Z} and
13419 @code{LPM R@var{n},Z+} instructions.
13421 @item __AVR_HAVE_MUL__
13422 The device has a hardware multiplier.
13424 @item __AVR_HAVE_JMP_CALL__
13425 The device has the @code{JMP} and @code{CALL} instructions.
13426 This is the case for devices with at least 16@tie{}KiB of program
13429 @item __AVR_HAVE_EIJMP_EICALL__
13430 @itemx __AVR_3_BYTE_PC__
13431 The device has the @code{EIJMP} and @code{EICALL} instructions.
13432 This is the case for devices with more than 128@tie{}KiB of program memory.
13433 This also means that the program counter
13434 (PC) is 3@tie{}bytes wide.
13436 @item __AVR_2_BYTE_PC__
13437 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
13438 with up to 128@tie{}KiB of program memory.
13440 @item __AVR_HAVE_8BIT_SP__
13441 @itemx __AVR_HAVE_16BIT_SP__
13442 The stack pointer (SP) register is treated as 8-bit respectively
13443 16-bit register by the compiler.
13444 The definition of these macros is affected by @code{-mtiny-stack}.
13446 @item __AVR_HAVE_SPH__
13448 The device has the SPH (high part of stack pointer) special function
13449 register or has an 8-bit stack pointer, respectively.
13450 The definition of these macros is affected by @code{-mmcu=} and
13451 in the cases of @code{-mmcu=avr2} and @code{-mmcu=avr25} also
13454 @item __AVR_HAVE_RAMPD__
13455 @itemx __AVR_HAVE_RAMPX__
13456 @itemx __AVR_HAVE_RAMPY__
13457 @itemx __AVR_HAVE_RAMPZ__
13458 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
13459 @code{RAMPZ} special function register, respectively.
13461 @item __NO_INTERRUPTS__
13462 This macro reflects the @code{-mno-interrupts} command line option.
13464 @item __AVR_ERRATA_SKIP__
13465 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
13466 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
13467 instructions because of a hardware erratum. Skip instructions are
13468 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
13469 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
13472 @item __AVR_ISA_RMW__
13473 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
13475 @item __AVR_SFR_OFFSET__=@var{offset}
13476 Instructions that can address I/O special function registers directly
13477 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
13478 address as if addressed by an instruction to access RAM like @code{LD}
13479 or @code{STS}. This offset depends on the device architecture and has
13480 to be subtracted from the RAM address in order to get the
13481 respective I/O@tie{}address.
13483 @item __WITH_AVRLIBC__
13484 The compiler is configured to be used together with AVR-Libc.
13485 See the @code{--with-avrlibc} configure option.
13489 @node Blackfin Options
13490 @subsection Blackfin Options
13491 @cindex Blackfin Options
13494 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
13496 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
13497 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
13498 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
13499 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
13500 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
13501 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
13502 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
13503 @samp{bf561}, @samp{bf592}.
13505 The optional @var{sirevision} specifies the silicon revision of the target
13506 Blackfin processor. Any workarounds available for the targeted silicon revision
13507 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
13508 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
13509 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
13510 hexadecimal digits representing the major and minor numbers in the silicon
13511 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
13512 is not defined. If @var{sirevision} is @samp{any}, the
13513 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
13514 If this optional @var{sirevision} is not used, GCC assumes the latest known
13515 silicon revision of the targeted Blackfin processor.
13517 GCC defines a preprocessor macro for the specified @var{cpu}.
13518 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
13519 provided by libgloss to be linked in if @option{-msim} is not given.
13521 Without this option, @samp{bf532} is used as the processor by default.
13523 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
13524 only the preprocessor macro is defined.
13528 Specifies that the program will be run on the simulator. This causes
13529 the simulator BSP provided by libgloss to be linked in. This option
13530 has effect only for @samp{bfin-elf} toolchain.
13531 Certain other options, such as @option{-mid-shared-library} and
13532 @option{-mfdpic}, imply @option{-msim}.
13534 @item -momit-leaf-frame-pointer
13535 @opindex momit-leaf-frame-pointer
13536 Don't keep the frame pointer in a register for leaf functions. This
13537 avoids the instructions to save, set up and restore frame pointers and
13538 makes an extra register available in leaf functions. The option
13539 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
13540 which might make debugging harder.
13542 @item -mspecld-anomaly
13543 @opindex mspecld-anomaly
13544 When enabled, the compiler ensures that the generated code does not
13545 contain speculative loads after jump instructions. If this option is used,
13546 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
13548 @item -mno-specld-anomaly
13549 @opindex mno-specld-anomaly
13550 Don't generate extra code to prevent speculative loads from occurring.
13552 @item -mcsync-anomaly
13553 @opindex mcsync-anomaly
13554 When enabled, the compiler ensures that the generated code does not
13555 contain CSYNC or SSYNC instructions too soon after conditional branches.
13556 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
13558 @item -mno-csync-anomaly
13559 @opindex mno-csync-anomaly
13560 Don't generate extra code to prevent CSYNC or SSYNC instructions from
13561 occurring too soon after a conditional branch.
13565 When enabled, the compiler is free to take advantage of the knowledge that
13566 the entire program fits into the low 64k of memory.
13569 @opindex mno-low-64k
13570 Assume that the program is arbitrarily large. This is the default.
13572 @item -mstack-check-l1
13573 @opindex mstack-check-l1
13574 Do stack checking using information placed into L1 scratchpad memory by the
13577 @item -mid-shared-library
13578 @opindex mid-shared-library
13579 Generate code that supports shared libraries via the library ID method.
13580 This allows for execute in place and shared libraries in an environment
13581 without virtual memory management. This option implies @option{-fPIC}.
13582 With a @samp{bfin-elf} target, this option implies @option{-msim}.
13584 @item -mno-id-shared-library
13585 @opindex mno-id-shared-library
13586 Generate code that doesn't assume ID-based shared libraries are being used.
13587 This is the default.
13589 @item -mleaf-id-shared-library
13590 @opindex mleaf-id-shared-library
13591 Generate code that supports shared libraries via the library ID method,
13592 but assumes that this library or executable won't link against any other
13593 ID shared libraries. That allows the compiler to use faster code for jumps
13596 @item -mno-leaf-id-shared-library
13597 @opindex mno-leaf-id-shared-library
13598 Do not assume that the code being compiled won't link against any ID shared
13599 libraries. Slower code is generated for jump and call insns.
13601 @item -mshared-library-id=n
13602 @opindex mshared-library-id
13603 Specifies the identification number of the ID-based shared library being
13604 compiled. Specifying a value of 0 generates more compact code; specifying
13605 other values forces the allocation of that number to the current
13606 library but is no more space- or time-efficient than omitting this option.
13610 Generate code that allows the data segment to be located in a different
13611 area of memory from the text segment. This allows for execute in place in
13612 an environment without virtual memory management by eliminating relocations
13613 against the text section.
13615 @item -mno-sep-data
13616 @opindex mno-sep-data
13617 Generate code that assumes that the data segment follows the text segment.
13618 This is the default.
13621 @itemx -mno-long-calls
13622 @opindex mlong-calls
13623 @opindex mno-long-calls
13624 Tells the compiler to perform function calls by first loading the
13625 address of the function into a register and then performing a subroutine
13626 call on this register. This switch is needed if the target function
13627 lies outside of the 24-bit addressing range of the offset-based
13628 version of subroutine call instruction.
13630 This feature is not enabled by default. Specifying
13631 @option{-mno-long-calls} restores the default behavior. Note these
13632 switches have no effect on how the compiler generates code to handle
13633 function calls via function pointers.
13637 Link with the fast floating-point library. This library relaxes some of
13638 the IEEE floating-point standard's rules for checking inputs against
13639 Not-a-Number (NAN), in the interest of performance.
13642 @opindex minline-plt
13643 Enable inlining of PLT entries in function calls to functions that are
13644 not known to bind locally. It has no effect without @option{-mfdpic}.
13647 @opindex mmulticore
13648 Build a standalone application for multicore Blackfin processors.
13649 This option causes proper start files and link scripts supporting
13650 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
13651 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
13653 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
13654 selects the one-application-per-core programming model. Without
13655 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
13656 programming model is used. In this model, the main function of Core B
13657 should be named as @code{coreb_main}.
13659 If this option is not used, the single-core application programming
13664 Build a standalone application for Core A of BF561 when using
13665 the one-application-per-core programming model. Proper start files
13666 and link scripts are used to support Core A, and the macro
13667 @code{__BFIN_COREA} is defined.
13668 This option can only be used in conjunction with @option{-mmulticore}.
13672 Build a standalone application for Core B of BF561 when using
13673 the one-application-per-core programming model. Proper start files
13674 and link scripts are used to support Core B, and the macro
13675 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
13676 should be used instead of @code{main}.
13677 This option can only be used in conjunction with @option{-mmulticore}.
13681 Build a standalone application for SDRAM. Proper start files and
13682 link scripts are used to put the application into SDRAM, and the macro
13683 @code{__BFIN_SDRAM} is defined.
13684 The loader should initialize SDRAM before loading the application.
13688 Assume that ICPLBs are enabled at run time. This has an effect on certain
13689 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
13690 are enabled; for standalone applications the default is off.
13694 @subsection C6X Options
13695 @cindex C6X Options
13698 @item -march=@var{name}
13700 This specifies the name of the target architecture. GCC uses this
13701 name to determine what kind of instructions it can emit when generating
13702 assembly code. Permissible names are: @samp{c62x},
13703 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
13706 @opindex mbig-endian
13707 Generate code for a big-endian target.
13709 @item -mlittle-endian
13710 @opindex mlittle-endian
13711 Generate code for a little-endian target. This is the default.
13715 Choose startup files and linker script suitable for the simulator.
13717 @item -msdata=default
13718 @opindex msdata=default
13719 Put small global and static data in the @samp{.neardata} section,
13720 which is pointed to by register @code{B14}. Put small uninitialized
13721 global and static data in the @samp{.bss} section, which is adjacent
13722 to the @samp{.neardata} section. Put small read-only data into the
13723 @samp{.rodata} section. The corresponding sections used for large
13724 pieces of data are @samp{.fardata}, @samp{.far} and @samp{.const}.
13727 @opindex msdata=all
13728 Put all data, not just small objects, into the sections reserved for
13729 small data, and use addressing relative to the @code{B14} register to
13733 @opindex msdata=none
13734 Make no use of the sections reserved for small data, and use absolute
13735 addresses to access all data. Put all initialized global and static
13736 data in the @samp{.fardata} section, and all uninitialized data in the
13737 @samp{.far} section. Put all constant data into the @samp{.const}
13742 @subsection CRIS Options
13743 @cindex CRIS Options
13745 These options are defined specifically for the CRIS ports.
13748 @item -march=@var{architecture-type}
13749 @itemx -mcpu=@var{architecture-type}
13752 Generate code for the specified architecture. The choices for
13753 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
13754 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
13755 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
13758 @item -mtune=@var{architecture-type}
13760 Tune to @var{architecture-type} everything applicable about the generated
13761 code, except for the ABI and the set of available instructions. The
13762 choices for @var{architecture-type} are the same as for
13763 @option{-march=@var{architecture-type}}.
13765 @item -mmax-stack-frame=@var{n}
13766 @opindex mmax-stack-frame
13767 Warn when the stack frame of a function exceeds @var{n} bytes.
13773 The options @option{-metrax4} and @option{-metrax100} are synonyms for
13774 @option{-march=v3} and @option{-march=v8} respectively.
13776 @item -mmul-bug-workaround
13777 @itemx -mno-mul-bug-workaround
13778 @opindex mmul-bug-workaround
13779 @opindex mno-mul-bug-workaround
13780 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
13781 models where it applies. This option is active by default.
13785 Enable CRIS-specific verbose debug-related information in the assembly
13786 code. This option also has the effect of turning off the @samp{#NO_APP}
13787 formatted-code indicator to the assembler at the beginning of the
13792 Do not use condition-code results from previous instruction; always emit
13793 compare and test instructions before use of condition codes.
13795 @item -mno-side-effects
13796 @opindex mno-side-effects
13797 Do not emit instructions with side effects in addressing modes other than
13800 @item -mstack-align
13801 @itemx -mno-stack-align
13802 @itemx -mdata-align
13803 @itemx -mno-data-align
13804 @itemx -mconst-align
13805 @itemx -mno-const-align
13806 @opindex mstack-align
13807 @opindex mno-stack-align
13808 @opindex mdata-align
13809 @opindex mno-data-align
13810 @opindex mconst-align
13811 @opindex mno-const-align
13812 These options (@samp{no-} options) arrange (eliminate arrangements) for the
13813 stack frame, individual data and constants to be aligned for the maximum
13814 single data access size for the chosen CPU model. The default is to
13815 arrange for 32-bit alignment. ABI details such as structure layout are
13816 not affected by these options.
13824 Similar to the stack- data- and const-align options above, these options
13825 arrange for stack frame, writable data and constants to all be 32-bit,
13826 16-bit or 8-bit aligned. The default is 32-bit alignment.
13828 @item -mno-prologue-epilogue
13829 @itemx -mprologue-epilogue
13830 @opindex mno-prologue-epilogue
13831 @opindex mprologue-epilogue
13832 With @option{-mno-prologue-epilogue}, the normal function prologue and
13833 epilogue which set up the stack frame are omitted and no return
13834 instructions or return sequences are generated in the code. Use this
13835 option only together with visual inspection of the compiled code: no
13836 warnings or errors are generated when call-saved registers must be saved,
13837 or storage for local variables needs to be allocated.
13841 @opindex mno-gotplt
13843 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
13844 instruction sequences that load addresses for functions from the PLT part
13845 of the GOT rather than (traditional on other architectures) calls to the
13846 PLT@. The default is @option{-mgotplt}.
13850 Legacy no-op option only recognized with the cris-axis-elf and
13851 cris-axis-linux-gnu targets.
13855 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
13859 This option, recognized for the cris-axis-elf, arranges
13860 to link with input-output functions from a simulator library. Code,
13861 initialized data and zero-initialized data are allocated consecutively.
13865 Like @option{-sim}, but pass linker options to locate initialized data at
13866 0x40000000 and zero-initialized data at 0x80000000.
13870 @subsection CR16 Options
13871 @cindex CR16 Options
13873 These options are defined specifically for the CR16 ports.
13879 Enable the use of multiply-accumulate instructions. Disabled by default.
13883 @opindex mcr16cplus
13885 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
13890 Links the library libsim.a which is in compatible with simulator. Applicable
13891 to ELF compiler only.
13895 Choose integer type as 32-bit wide.
13899 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
13901 @item -mdata-model=@var{model}
13902 @opindex mdata-model
13903 Choose a data model. The choices for @var{model} are @samp{near},
13904 @samp{far} or @samp{medium}. @samp{medium} is default.
13905 However, @samp{far} is not valid with @option{-mcr16c}, as the
13906 CR16C architecture does not support the far data model.
13909 @node Darwin Options
13910 @subsection Darwin Options
13911 @cindex Darwin options
13913 These options are defined for all architectures running the Darwin operating
13916 FSF GCC on Darwin does not create ``fat'' object files; it creates
13917 an object file for the single architecture that GCC was built to
13918 target. Apple's GCC on Darwin does create ``fat'' files if multiple
13919 @option{-arch} options are used; it does so by running the compiler or
13920 linker multiple times and joining the results together with
13923 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
13924 @samp{i686}) is determined by the flags that specify the ISA
13925 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
13926 @option{-force_cpusubtype_ALL} option can be used to override this.
13928 The Darwin tools vary in their behavior when presented with an ISA
13929 mismatch. The assembler, @file{as}, only permits instructions to
13930 be used that are valid for the subtype of the file it is generating,
13931 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
13932 The linker for shared libraries, @file{/usr/bin/libtool}, fails
13933 and prints an error if asked to create a shared library with a less
13934 restrictive subtype than its input files (for instance, trying to put
13935 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
13936 for executables, @command{ld}, quietly gives the executable the most
13937 restrictive subtype of any of its input files.
13942 Add the framework directory @var{dir} to the head of the list of
13943 directories to be searched for header files. These directories are
13944 interleaved with those specified by @option{-I} options and are
13945 scanned in a left-to-right order.
13947 A framework directory is a directory with frameworks in it. A
13948 framework is a directory with a @file{Headers} and/or
13949 @file{PrivateHeaders} directory contained directly in it that ends
13950 in @file{.framework}. The name of a framework is the name of this
13951 directory excluding the @file{.framework}. Headers associated with
13952 the framework are found in one of those two directories, with
13953 @file{Headers} being searched first. A subframework is a framework
13954 directory that is in a framework's @file{Frameworks} directory.
13955 Includes of subframework headers can only appear in a header of a
13956 framework that contains the subframework, or in a sibling subframework
13957 header. Two subframeworks are siblings if they occur in the same
13958 framework. A subframework should not have the same name as a
13959 framework; a warning is issued if this is violated. Currently a
13960 subframework cannot have subframeworks; in the future, the mechanism
13961 may be extended to support this. The standard frameworks can be found
13962 in @file{/System/Library/Frameworks} and
13963 @file{/Library/Frameworks}. An example include looks like
13964 @code{#include <Framework/header.h>}, where @file{Framework} denotes
13965 the name of the framework and @file{header.h} is found in the
13966 @file{PrivateHeaders} or @file{Headers} directory.
13968 @item -iframework@var{dir}
13969 @opindex iframework
13970 Like @option{-F} except the directory is a treated as a system
13971 directory. The main difference between this @option{-iframework} and
13972 @option{-F} is that with @option{-iframework} the compiler does not
13973 warn about constructs contained within header files found via
13974 @var{dir}. This option is valid only for the C family of languages.
13978 Emit debugging information for symbols that are used. For stabs
13979 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
13980 This is by default ON@.
13984 Emit debugging information for all symbols and types.
13986 @item -mmacosx-version-min=@var{version}
13987 The earliest version of MacOS X that this executable will run on
13988 is @var{version}. Typical values of @var{version} include @code{10.1},
13989 @code{10.2}, and @code{10.3.9}.
13991 If the compiler was built to use the system's headers by default,
13992 then the default for this option is the system version on which the
13993 compiler is running, otherwise the default is to make choices that
13994 are compatible with as many systems and code bases as possible.
13998 Enable kernel development mode. The @option{-mkernel} option sets
13999 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
14000 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
14001 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
14002 applicable. This mode also sets @option{-mno-altivec},
14003 @option{-msoft-float}, @option{-fno-builtin} and
14004 @option{-mlong-branch} for PowerPC targets.
14006 @item -mone-byte-bool
14007 @opindex mone-byte-bool
14008 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
14009 By default @samp{sizeof(bool)} is @samp{4} when compiling for
14010 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
14011 option has no effect on x86.
14013 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
14014 to generate code that is not binary compatible with code generated
14015 without that switch. Using this switch may require recompiling all
14016 other modules in a program, including system libraries. Use this
14017 switch to conform to a non-default data model.
14019 @item -mfix-and-continue
14020 @itemx -ffix-and-continue
14021 @itemx -findirect-data
14022 @opindex mfix-and-continue
14023 @opindex ffix-and-continue
14024 @opindex findirect-data
14025 Generate code suitable for fast turnaround development, such as to
14026 allow GDB to dynamically load @code{.o} files into already-running
14027 programs. @option{-findirect-data} and @option{-ffix-and-continue}
14028 are provided for backwards compatibility.
14032 Loads all members of static archive libraries.
14033 See man ld(1) for more information.
14035 @item -arch_errors_fatal
14036 @opindex arch_errors_fatal
14037 Cause the errors having to do with files that have the wrong architecture
14040 @item -bind_at_load
14041 @opindex bind_at_load
14042 Causes the output file to be marked such that the dynamic linker will
14043 bind all undefined references when the file is loaded or launched.
14047 Produce a Mach-o bundle format file.
14048 See man ld(1) for more information.
14050 @item -bundle_loader @var{executable}
14051 @opindex bundle_loader
14052 This option specifies the @var{executable} that will load the build
14053 output file being linked. See man ld(1) for more information.
14056 @opindex dynamiclib
14057 When passed this option, GCC produces a dynamic library instead of
14058 an executable when linking, using the Darwin @file{libtool} command.
14060 @item -force_cpusubtype_ALL
14061 @opindex force_cpusubtype_ALL
14062 This causes GCC's output file to have the @var{ALL} subtype, instead of
14063 one controlled by the @option{-mcpu} or @option{-march} option.
14065 @item -allowable_client @var{client_name}
14066 @itemx -client_name
14067 @itemx -compatibility_version
14068 @itemx -current_version
14070 @itemx -dependency-file
14072 @itemx -dylinker_install_name
14074 @itemx -exported_symbols_list
14077 @itemx -flat_namespace
14078 @itemx -force_flat_namespace
14079 @itemx -headerpad_max_install_names
14082 @itemx -install_name
14083 @itemx -keep_private_externs
14084 @itemx -multi_module
14085 @itemx -multiply_defined
14086 @itemx -multiply_defined_unused
14089 @itemx -no_dead_strip_inits_and_terms
14090 @itemx -nofixprebinding
14091 @itemx -nomultidefs
14093 @itemx -noseglinkedit
14094 @itemx -pagezero_size
14096 @itemx -prebind_all_twolevel_modules
14097 @itemx -private_bundle
14099 @itemx -read_only_relocs
14101 @itemx -sectobjectsymbols
14105 @itemx -sectobjectsymbols
14108 @itemx -segs_read_only_addr
14110 @itemx -segs_read_write_addr
14111 @itemx -seg_addr_table
14112 @itemx -seg_addr_table_filename
14113 @itemx -seglinkedit
14115 @itemx -segs_read_only_addr
14116 @itemx -segs_read_write_addr
14117 @itemx -single_module
14119 @itemx -sub_library
14121 @itemx -sub_umbrella
14122 @itemx -twolevel_namespace
14125 @itemx -unexported_symbols_list
14126 @itemx -weak_reference_mismatches
14127 @itemx -whatsloaded
14128 @opindex allowable_client
14129 @opindex client_name
14130 @opindex compatibility_version
14131 @opindex current_version
14132 @opindex dead_strip
14133 @opindex dependency-file
14134 @opindex dylib_file
14135 @opindex dylinker_install_name
14137 @opindex exported_symbols_list
14139 @opindex flat_namespace
14140 @opindex force_flat_namespace
14141 @opindex headerpad_max_install_names
14142 @opindex image_base
14144 @opindex install_name
14145 @opindex keep_private_externs
14146 @opindex multi_module
14147 @opindex multiply_defined
14148 @opindex multiply_defined_unused
14149 @opindex noall_load
14150 @opindex no_dead_strip_inits_and_terms
14151 @opindex nofixprebinding
14152 @opindex nomultidefs
14154 @opindex noseglinkedit
14155 @opindex pagezero_size
14157 @opindex prebind_all_twolevel_modules
14158 @opindex private_bundle
14159 @opindex read_only_relocs
14161 @opindex sectobjectsymbols
14164 @opindex sectcreate
14165 @opindex sectobjectsymbols
14168 @opindex segs_read_only_addr
14169 @opindex segs_read_write_addr
14170 @opindex seg_addr_table
14171 @opindex seg_addr_table_filename
14172 @opindex seglinkedit
14174 @opindex segs_read_only_addr
14175 @opindex segs_read_write_addr
14176 @opindex single_module
14178 @opindex sub_library
14179 @opindex sub_umbrella
14180 @opindex twolevel_namespace
14183 @opindex unexported_symbols_list
14184 @opindex weak_reference_mismatches
14185 @opindex whatsloaded
14186 These options are passed to the Darwin linker. The Darwin linker man page
14187 describes them in detail.
14190 @node DEC Alpha Options
14191 @subsection DEC Alpha Options
14193 These @samp{-m} options are defined for the DEC Alpha implementations:
14196 @item -mno-soft-float
14197 @itemx -msoft-float
14198 @opindex mno-soft-float
14199 @opindex msoft-float
14200 Use (do not use) the hardware floating-point instructions for
14201 floating-point operations. When @option{-msoft-float} is specified,
14202 functions in @file{libgcc.a} are used to perform floating-point
14203 operations. Unless they are replaced by routines that emulate the
14204 floating-point operations, or compiled in such a way as to call such
14205 emulations routines, these routines issue floating-point
14206 operations. If you are compiling for an Alpha without floating-point
14207 operations, you must ensure that the library is built so as not to call
14210 Note that Alpha implementations without floating-point operations are
14211 required to have floating-point registers.
14214 @itemx -mno-fp-regs
14216 @opindex mno-fp-regs
14217 Generate code that uses (does not use) the floating-point register set.
14218 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
14219 register set is not used, floating-point operands are passed in integer
14220 registers as if they were integers and floating-point results are passed
14221 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
14222 so any function with a floating-point argument or return value called by code
14223 compiled with @option{-mno-fp-regs} must also be compiled with that
14226 A typical use of this option is building a kernel that does not use,
14227 and hence need not save and restore, any floating-point registers.
14231 The Alpha architecture implements floating-point hardware optimized for
14232 maximum performance. It is mostly compliant with the IEEE floating-point
14233 standard. However, for full compliance, software assistance is
14234 required. This option generates code fully IEEE-compliant code
14235 @emph{except} that the @var{inexact-flag} is not maintained (see below).
14236 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
14237 defined during compilation. The resulting code is less efficient but is
14238 able to correctly support denormalized numbers and exceptional IEEE
14239 values such as not-a-number and plus/minus infinity. Other Alpha
14240 compilers call this option @option{-ieee_with_no_inexact}.
14242 @item -mieee-with-inexact
14243 @opindex mieee-with-inexact
14244 This is like @option{-mieee} except the generated code also maintains
14245 the IEEE @var{inexact-flag}. Turning on this option causes the
14246 generated code to implement fully-compliant IEEE math. In addition to
14247 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
14248 macro. On some Alpha implementations the resulting code may execute
14249 significantly slower than the code generated by default. Since there is
14250 very little code that depends on the @var{inexact-flag}, you should
14251 normally not specify this option. Other Alpha compilers call this
14252 option @option{-ieee_with_inexact}.
14254 @item -mfp-trap-mode=@var{trap-mode}
14255 @opindex mfp-trap-mode
14256 This option controls what floating-point related traps are enabled.
14257 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
14258 The trap mode can be set to one of four values:
14262 This is the default (normal) setting. The only traps that are enabled
14263 are the ones that cannot be disabled in software (e.g., division by zero
14267 In addition to the traps enabled by @samp{n}, underflow traps are enabled
14271 Like @samp{u}, but the instructions are marked to be safe for software
14272 completion (see Alpha architecture manual for details).
14275 Like @samp{su}, but inexact traps are enabled as well.
14278 @item -mfp-rounding-mode=@var{rounding-mode}
14279 @opindex mfp-rounding-mode
14280 Selects the IEEE rounding mode. Other Alpha compilers call this option
14281 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
14286 Normal IEEE rounding mode. Floating-point numbers are rounded towards
14287 the nearest machine number or towards the even machine number in case
14291 Round towards minus infinity.
14294 Chopped rounding mode. Floating-point numbers are rounded towards zero.
14297 Dynamic rounding mode. A field in the floating-point control register
14298 (@var{fpcr}, see Alpha architecture reference manual) controls the
14299 rounding mode in effect. The C library initializes this register for
14300 rounding towards plus infinity. Thus, unless your program modifies the
14301 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
14304 @item -mtrap-precision=@var{trap-precision}
14305 @opindex mtrap-precision
14306 In the Alpha architecture, floating-point traps are imprecise. This
14307 means without software assistance it is impossible to recover from a
14308 floating trap and program execution normally needs to be terminated.
14309 GCC can generate code that can assist operating system trap handlers
14310 in determining the exact location that caused a floating-point trap.
14311 Depending on the requirements of an application, different levels of
14312 precisions can be selected:
14316 Program precision. This option is the default and means a trap handler
14317 can only identify which program caused a floating-point exception.
14320 Function precision. The trap handler can determine the function that
14321 caused a floating-point exception.
14324 Instruction precision. The trap handler can determine the exact
14325 instruction that caused a floating-point exception.
14328 Other Alpha compilers provide the equivalent options called
14329 @option{-scope_safe} and @option{-resumption_safe}.
14331 @item -mieee-conformant
14332 @opindex mieee-conformant
14333 This option marks the generated code as IEEE conformant. You must not
14334 use this option unless you also specify @option{-mtrap-precision=i} and either
14335 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
14336 is to emit the line @samp{.eflag 48} in the function prologue of the
14337 generated assembly file.
14339 @item -mbuild-constants
14340 @opindex mbuild-constants
14341 Normally GCC examines a 32- or 64-bit integer constant to
14342 see if it can construct it from smaller constants in two or three
14343 instructions. If it cannot, it outputs the constant as a literal and
14344 generates code to load it from the data segment at run time.
14346 Use this option to require GCC to construct @emph{all} integer constants
14347 using code, even if it takes more instructions (the maximum is six).
14349 You typically use this option to build a shared library dynamic
14350 loader. Itself a shared library, it must relocate itself in memory
14351 before it can find the variables and constants in its own data segment.
14369 Indicate whether GCC should generate code to use the optional BWX,
14370 CIX, FIX and MAX instruction sets. The default is to use the instruction
14371 sets supported by the CPU type specified via @option{-mcpu=} option or that
14372 of the CPU on which GCC was built if none is specified.
14375 @itemx -mfloat-ieee
14376 @opindex mfloat-vax
14377 @opindex mfloat-ieee
14378 Generate code that uses (does not use) VAX F and G floating-point
14379 arithmetic instead of IEEE single and double precision.
14381 @item -mexplicit-relocs
14382 @itemx -mno-explicit-relocs
14383 @opindex mexplicit-relocs
14384 @opindex mno-explicit-relocs
14385 Older Alpha assemblers provided no way to generate symbol relocations
14386 except via assembler macros. Use of these macros does not allow
14387 optimal instruction scheduling. GNU binutils as of version 2.12
14388 supports a new syntax that allows the compiler to explicitly mark
14389 which relocations should apply to which instructions. This option
14390 is mostly useful for debugging, as GCC detects the capabilities of
14391 the assembler when it is built and sets the default accordingly.
14394 @itemx -mlarge-data
14395 @opindex msmall-data
14396 @opindex mlarge-data
14397 When @option{-mexplicit-relocs} is in effect, static data is
14398 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
14399 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
14400 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
14401 16-bit relocations off of the @code{$gp} register. This limits the
14402 size of the small data area to 64KB, but allows the variables to be
14403 directly accessed via a single instruction.
14405 The default is @option{-mlarge-data}. With this option the data area
14406 is limited to just below 2GB@. Programs that require more than 2GB of
14407 data must use @code{malloc} or @code{mmap} to allocate the data in the
14408 heap instead of in the program's data segment.
14410 When generating code for shared libraries, @option{-fpic} implies
14411 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
14414 @itemx -mlarge-text
14415 @opindex msmall-text
14416 @opindex mlarge-text
14417 When @option{-msmall-text} is used, the compiler assumes that the
14418 code of the entire program (or shared library) fits in 4MB, and is
14419 thus reachable with a branch instruction. When @option{-msmall-data}
14420 is used, the compiler can assume that all local symbols share the
14421 same @code{$gp} value, and thus reduce the number of instructions
14422 required for a function call from 4 to 1.
14424 The default is @option{-mlarge-text}.
14426 @item -mcpu=@var{cpu_type}
14428 Set the instruction set and instruction scheduling parameters for
14429 machine type @var{cpu_type}. You can specify either the @samp{EV}
14430 style name or the corresponding chip number. GCC supports scheduling
14431 parameters for the EV4, EV5 and EV6 family of processors and
14432 chooses the default values for the instruction set from the processor
14433 you specify. If you do not specify a processor type, GCC defaults
14434 to the processor on which the compiler was built.
14436 Supported values for @var{cpu_type} are
14442 Schedules as an EV4 and has no instruction set extensions.
14446 Schedules as an EV5 and has no instruction set extensions.
14450 Schedules as an EV5 and supports the BWX extension.
14455 Schedules as an EV5 and supports the BWX and MAX extensions.
14459 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
14463 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
14466 Native toolchains also support the value @samp{native},
14467 which selects the best architecture option for the host processor.
14468 @option{-mcpu=native} has no effect if GCC does not recognize
14471 @item -mtune=@var{cpu_type}
14473 Set only the instruction scheduling parameters for machine type
14474 @var{cpu_type}. The instruction set is not changed.
14476 Native toolchains also support the value @samp{native},
14477 which selects the best architecture option for the host processor.
14478 @option{-mtune=native} has no effect if GCC does not recognize
14481 @item -mmemory-latency=@var{time}
14482 @opindex mmemory-latency
14483 Sets the latency the scheduler should assume for typical memory
14484 references as seen by the application. This number is highly
14485 dependent on the memory access patterns used by the application
14486 and the size of the external cache on the machine.
14488 Valid options for @var{time} are
14492 A decimal number representing clock cycles.
14498 The compiler contains estimates of the number of clock cycles for
14499 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
14500 (also called Dcache, Scache, and Bcache), as well as to main memory.
14501 Note that L3 is only valid for EV5.
14507 @subsection FR30 Options
14508 @cindex FR30 Options
14510 These options are defined specifically for the FR30 port.
14514 @item -msmall-model
14515 @opindex msmall-model
14516 Use the small address space model. This can produce smaller code, but
14517 it does assume that all symbolic values and addresses fit into a
14522 Assume that runtime support has been provided and so there is no need
14523 to include the simulator library (@file{libsim.a}) on the linker
14529 @subsection FRV Options
14530 @cindex FRV Options
14536 Only use the first 32 general-purpose registers.
14541 Use all 64 general-purpose registers.
14546 Use only the first 32 floating-point registers.
14551 Use all 64 floating-point registers.
14554 @opindex mhard-float
14556 Use hardware instructions for floating-point operations.
14559 @opindex msoft-float
14561 Use library routines for floating-point operations.
14566 Dynamically allocate condition code registers.
14571 Do not try to dynamically allocate condition code registers, only
14572 use @code{icc0} and @code{fcc0}.
14577 Change ABI to use double word insns.
14582 Do not use double word instructions.
14587 Use floating-point double instructions.
14590 @opindex mno-double
14592 Do not use floating-point double instructions.
14597 Use media instructions.
14602 Do not use media instructions.
14607 Use multiply and add/subtract instructions.
14610 @opindex mno-muladd
14612 Do not use multiply and add/subtract instructions.
14617 Select the FDPIC ABI, which uses function descriptors to represent
14618 pointers to functions. Without any PIC/PIE-related options, it
14619 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
14620 assumes GOT entries and small data are within a 12-bit range from the
14621 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
14622 are computed with 32 bits.
14623 With a @samp{bfin-elf} target, this option implies @option{-msim}.
14626 @opindex minline-plt
14628 Enable inlining of PLT entries in function calls to functions that are
14629 not known to bind locally. It has no effect without @option{-mfdpic}.
14630 It's enabled by default if optimizing for speed and compiling for
14631 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
14632 optimization option such as @option{-O3} or above is present in the
14638 Assume a large TLS segment when generating thread-local code.
14643 Do not assume a large TLS segment when generating thread-local code.
14648 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
14649 that is known to be in read-only sections. It's enabled by default,
14650 except for @option{-fpic} or @option{-fpie}: even though it may help
14651 make the global offset table smaller, it trades 1 instruction for 4.
14652 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
14653 one of which may be shared by multiple symbols, and it avoids the need
14654 for a GOT entry for the referenced symbol, so it's more likely to be a
14655 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
14657 @item -multilib-library-pic
14658 @opindex multilib-library-pic
14660 Link with the (library, not FD) pic libraries. It's implied by
14661 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
14662 @option{-fpic} without @option{-mfdpic}. You should never have to use
14666 @opindex mlinked-fp
14668 Follow the EABI requirement of always creating a frame pointer whenever
14669 a stack frame is allocated. This option is enabled by default and can
14670 be disabled with @option{-mno-linked-fp}.
14673 @opindex mlong-calls
14675 Use indirect addressing to call functions outside the current
14676 compilation unit. This allows the functions to be placed anywhere
14677 within the 32-bit address space.
14679 @item -malign-labels
14680 @opindex malign-labels
14682 Try to align labels to an 8-byte boundary by inserting NOPs into the
14683 previous packet. This option only has an effect when VLIW packing
14684 is enabled. It doesn't create new packets; it merely adds NOPs to
14687 @item -mlibrary-pic
14688 @opindex mlibrary-pic
14690 Generate position-independent EABI code.
14695 Use only the first four media accumulator registers.
14700 Use all eight media accumulator registers.
14705 Pack VLIW instructions.
14710 Do not pack VLIW instructions.
14713 @opindex mno-eflags
14715 Do not mark ABI switches in e_flags.
14718 @opindex mcond-move
14720 Enable the use of conditional-move instructions (default).
14722 This switch is mainly for debugging the compiler and will likely be removed
14723 in a future version.
14725 @item -mno-cond-move
14726 @opindex mno-cond-move
14728 Disable the use of conditional-move instructions.
14730 This switch is mainly for debugging the compiler and will likely be removed
14731 in a future version.
14736 Enable the use of conditional set instructions (default).
14738 This switch is mainly for debugging the compiler and will likely be removed
14739 in a future version.
14744 Disable the use of conditional set instructions.
14746 This switch is mainly for debugging the compiler and will likely be removed
14747 in a future version.
14750 @opindex mcond-exec
14752 Enable the use of conditional execution (default).
14754 This switch is mainly for debugging the compiler and will likely be removed
14755 in a future version.
14757 @item -mno-cond-exec
14758 @opindex mno-cond-exec
14760 Disable the use of conditional execution.
14762 This switch is mainly for debugging the compiler and will likely be removed
14763 in a future version.
14765 @item -mvliw-branch
14766 @opindex mvliw-branch
14768 Run a pass to pack branches into VLIW instructions (default).
14770 This switch is mainly for debugging the compiler and will likely be removed
14771 in a future version.
14773 @item -mno-vliw-branch
14774 @opindex mno-vliw-branch
14776 Do not run a pass to pack branches into VLIW instructions.
14778 This switch is mainly for debugging the compiler and will likely be removed
14779 in a future version.
14781 @item -mmulti-cond-exec
14782 @opindex mmulti-cond-exec
14784 Enable optimization of @code{&&} and @code{||} in conditional execution
14787 This switch is mainly for debugging the compiler and will likely be removed
14788 in a future version.
14790 @item -mno-multi-cond-exec
14791 @opindex mno-multi-cond-exec
14793 Disable optimization of @code{&&} and @code{||} in conditional execution.
14795 This switch is mainly for debugging the compiler and will likely be removed
14796 in a future version.
14798 @item -mnested-cond-exec
14799 @opindex mnested-cond-exec
14801 Enable nested conditional execution optimizations (default).
14803 This switch is mainly for debugging the compiler and will likely be removed
14804 in a future version.
14806 @item -mno-nested-cond-exec
14807 @opindex mno-nested-cond-exec
14809 Disable nested conditional execution optimizations.
14811 This switch is mainly for debugging the compiler and will likely be removed
14812 in a future version.
14814 @item -moptimize-membar
14815 @opindex moptimize-membar
14817 This switch removes redundant @code{membar} instructions from the
14818 compiler-generated code. It is enabled by default.
14820 @item -mno-optimize-membar
14821 @opindex mno-optimize-membar
14823 This switch disables the automatic removal of redundant @code{membar}
14824 instructions from the generated code.
14826 @item -mtomcat-stats
14827 @opindex mtomcat-stats
14829 Cause gas to print out tomcat statistics.
14831 @item -mcpu=@var{cpu}
14834 Select the processor type for which to generate code. Possible values are
14835 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
14836 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
14840 @node GNU/Linux Options
14841 @subsection GNU/Linux Options
14843 These @samp{-m} options are defined for GNU/Linux targets:
14848 Use the GNU C library. This is the default except
14849 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
14853 Use uClibc C library. This is the default on
14854 @samp{*-*-linux-*uclibc*} targets.
14858 Use Bionic C library. This is the default on
14859 @samp{*-*-linux-*android*} targets.
14863 Compile code compatible with Android platform. This is the default on
14864 @samp{*-*-linux-*android*} targets.
14866 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
14867 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
14868 this option makes the GCC driver pass Android-specific options to the linker.
14869 Finally, this option causes the preprocessor macro @code{__ANDROID__}
14872 @item -tno-android-cc
14873 @opindex tno-android-cc
14874 Disable compilation effects of @option{-mandroid}, i.e., do not enable
14875 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
14876 @option{-fno-rtti} by default.
14878 @item -tno-android-ld
14879 @opindex tno-android-ld
14880 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
14881 linking options to the linker.
14885 @node H8/300 Options
14886 @subsection H8/300 Options
14888 These @samp{-m} options are defined for the H8/300 implementations:
14893 Shorten some address references at link time, when possible; uses the
14894 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
14895 ld, Using ld}, for a fuller description.
14899 Generate code for the H8/300H@.
14903 Generate code for the H8S@.
14907 Generate code for the H8S and H8/300H in the normal mode. This switch
14908 must be used either with @option{-mh} or @option{-ms}.
14912 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
14916 Extended registers are stored on stack before execution of function
14917 with monitor attribute. Default option is @option{-mexr}.
14918 This option is valid only for H8S targets.
14922 Extended registers are not stored on stack before execution of function
14923 with monitor attribute. Default option is @option{-mno-exr}.
14924 This option is valid only for H8S targets.
14928 Make @code{int} data 32 bits by default.
14931 @opindex malign-300
14932 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
14933 The default for the H8/300H and H8S is to align longs and floats on
14935 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
14936 This option has no effect on the H8/300.
14940 @subsection HPPA Options
14941 @cindex HPPA Options
14943 These @samp{-m} options are defined for the HPPA family of computers:
14946 @item -march=@var{architecture-type}
14948 Generate code for the specified architecture. The choices for
14949 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
14950 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
14951 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
14952 architecture option for your machine. Code compiled for lower numbered
14953 architectures runs on higher numbered architectures, but not the
14956 @item -mpa-risc-1-0
14957 @itemx -mpa-risc-1-1
14958 @itemx -mpa-risc-2-0
14959 @opindex mpa-risc-1-0
14960 @opindex mpa-risc-1-1
14961 @opindex mpa-risc-2-0
14962 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
14964 @item -mjump-in-delay
14965 @opindex mjump-in-delay
14966 This option is ignored and provided for compatibility purposes only.
14968 @item -mdisable-fpregs
14969 @opindex mdisable-fpregs
14970 Prevent floating-point registers from being used in any manner. This is
14971 necessary for compiling kernels that perform lazy context switching of
14972 floating-point registers. If you use this option and attempt to perform
14973 floating-point operations, the compiler aborts.
14975 @item -mdisable-indexing
14976 @opindex mdisable-indexing
14977 Prevent the compiler from using indexing address modes. This avoids some
14978 rather obscure problems when compiling MIG generated code under MACH@.
14980 @item -mno-space-regs
14981 @opindex mno-space-regs
14982 Generate code that assumes the target has no space registers. This allows
14983 GCC to generate faster indirect calls and use unscaled index address modes.
14985 Such code is suitable for level 0 PA systems and kernels.
14987 @item -mfast-indirect-calls
14988 @opindex mfast-indirect-calls
14989 Generate code that assumes calls never cross space boundaries. This
14990 allows GCC to emit code that performs faster indirect calls.
14992 This option does not work in the presence of shared libraries or nested
14995 @item -mfixed-range=@var{register-range}
14996 @opindex mfixed-range
14997 Generate code treating the given register range as fixed registers.
14998 A fixed register is one that the register allocator cannot use. This is
14999 useful when compiling kernel code. A register range is specified as
15000 two registers separated by a dash. Multiple register ranges can be
15001 specified separated by a comma.
15003 @item -mlong-load-store
15004 @opindex mlong-load-store
15005 Generate 3-instruction load and store sequences as sometimes required by
15006 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
15009 @item -mportable-runtime
15010 @opindex mportable-runtime
15011 Use the portable calling conventions proposed by HP for ELF systems.
15015 Enable the use of assembler directives only GAS understands.
15017 @item -mschedule=@var{cpu-type}
15019 Schedule code according to the constraints for the machine type
15020 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
15021 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
15022 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
15023 proper scheduling option for your machine. The default scheduling is
15027 @opindex mlinker-opt
15028 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
15029 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
15030 linkers in which they give bogus error messages when linking some programs.
15033 @opindex msoft-float
15034 Generate output containing library calls for floating point.
15035 @strong{Warning:} the requisite libraries are not available for all HPPA
15036 targets. Normally the facilities of the machine's usual C compiler are
15037 used, but this cannot be done directly in cross-compilation. You must make
15038 your own arrangements to provide suitable library functions for
15041 @option{-msoft-float} changes the calling convention in the output file;
15042 therefore, it is only useful if you compile @emph{all} of a program with
15043 this option. In particular, you need to compile @file{libgcc.a}, the
15044 library that comes with GCC, with @option{-msoft-float} in order for
15049 Generate the predefine, @code{_SIO}, for server IO@. The default is
15050 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
15051 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
15052 options are available under HP-UX and HI-UX@.
15056 Use options specific to GNU @command{ld}.
15057 This passes @option{-shared} to @command{ld} when
15058 building a shared library. It is the default when GCC is configured,
15059 explicitly or implicitly, with the GNU linker. This option does not
15060 affect which @command{ld} is called; it only changes what parameters
15061 are passed to that @command{ld}.
15062 The @command{ld} that is called is determined by the
15063 @option{--with-ld} configure option, GCC's program search path, and
15064 finally by the user's @env{PATH}. The linker used by GCC can be printed
15065 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
15066 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15070 Use options specific to HP @command{ld}.
15071 This passes @option{-b} to @command{ld} when building
15072 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
15073 links. It is the default when GCC is configured, explicitly or
15074 implicitly, with the HP linker. This option does not affect
15075 which @command{ld} is called; it only changes what parameters are passed to that
15077 The @command{ld} that is called is determined by the @option{--with-ld}
15078 configure option, GCC's program search path, and finally by the user's
15079 @env{PATH}. The linker used by GCC can be printed using @samp{which
15080 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
15081 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15084 @opindex mno-long-calls
15085 Generate code that uses long call sequences. This ensures that a call
15086 is always able to reach linker generated stubs. The default is to generate
15087 long calls only when the distance from the call site to the beginning
15088 of the function or translation unit, as the case may be, exceeds a
15089 predefined limit set by the branch type being used. The limits for
15090 normal calls are 7,600,000 and 240,000 bytes, respectively for the
15091 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
15094 Distances are measured from the beginning of functions when using the
15095 @option{-ffunction-sections} option, or when using the @option{-mgas}
15096 and @option{-mno-portable-runtime} options together under HP-UX with
15099 It is normally not desirable to use this option as it degrades
15100 performance. However, it may be useful in large applications,
15101 particularly when partial linking is used to build the application.
15103 The types of long calls used depends on the capabilities of the
15104 assembler and linker, and the type of code being generated. The
15105 impact on systems that support long absolute calls, and long pic
15106 symbol-difference or pc-relative calls should be relatively small.
15107 However, an indirect call is used on 32-bit ELF systems in pic code
15108 and it is quite long.
15110 @item -munix=@var{unix-std}
15112 Generate compiler predefines and select a startfile for the specified
15113 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
15114 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
15115 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
15116 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
15117 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
15120 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
15121 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
15122 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
15123 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
15124 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
15125 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
15127 It is @emph{important} to note that this option changes the interfaces
15128 for various library routines. It also affects the operational behavior
15129 of the C library. Thus, @emph{extreme} care is needed in using this
15132 Library code that is intended to operate with more than one UNIX
15133 standard must test, set and restore the variable @var{__xpg4_extended_mask}
15134 as appropriate. Most GNU software doesn't provide this capability.
15138 Suppress the generation of link options to search libdld.sl when the
15139 @option{-static} option is specified on HP-UX 10 and later.
15143 The HP-UX implementation of setlocale in libc has a dependency on
15144 libdld.sl. There isn't an archive version of libdld.sl. Thus,
15145 when the @option{-static} option is specified, special link options
15146 are needed to resolve this dependency.
15148 On HP-UX 10 and later, the GCC driver adds the necessary options to
15149 link with libdld.sl when the @option{-static} option is specified.
15150 This causes the resulting binary to be dynamic. On the 64-bit port,
15151 the linkers generate dynamic binaries by default in any case. The
15152 @option{-nolibdld} option can be used to prevent the GCC driver from
15153 adding these link options.
15157 Add support for multithreading with the @dfn{dce thread} library
15158 under HP-UX@. This option sets flags for both the preprocessor and
15162 @node i386 and x86-64 Options
15163 @subsection Intel 386 and AMD x86-64 Options
15164 @cindex i386 Options
15165 @cindex x86-64 Options
15166 @cindex Intel 386 Options
15167 @cindex AMD x86-64 Options
15169 These @samp{-m} options are defined for the i386 and x86-64 family of
15174 @item -march=@var{cpu-type}
15176 Generate instructions for the machine type @var{cpu-type}. In contrast to
15177 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
15178 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
15179 to generate code that may not run at all on processors other than the one
15180 indicated. Specifying @option{-march=@var{cpu-type}} implies
15181 @option{-mtune=@var{cpu-type}}.
15183 The choices for @var{cpu-type} are:
15187 This selects the CPU to generate code for at compilation time by determining
15188 the processor type of the compiling machine. Using @option{-march=native}
15189 enables all instruction subsets supported by the local machine (hence
15190 the result might not run on different machines). Using @option{-mtune=native}
15191 produces code optimized for the local machine under the constraints
15192 of the selected instruction set.
15195 Original Intel i386 CPU@.
15198 Intel i486 CPU@. (No scheduling is implemented for this chip.)
15202 Intel Pentium CPU with no MMX support.
15205 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
15208 Intel Pentium Pro CPU@.
15211 When used with @option{-march}, the Pentium Pro
15212 instruction set is used, so the code runs on all i686 family chips.
15213 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
15216 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
15221 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
15225 Intel Pentium M; low-power version of Intel Pentium III CPU
15226 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
15230 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
15233 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
15237 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
15238 SSE2 and SSE3 instruction set support.
15241 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
15242 instruction set support.
15245 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
15246 SSE4.1, SSE4.2 and POPCNT instruction set support.
15249 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
15250 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
15253 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
15254 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
15257 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
15258 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
15259 instruction set support.
15262 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
15263 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
15264 BMI, BMI2 and F16C instruction set support.
15267 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
15268 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
15269 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
15272 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
15273 instruction set support.
15276 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
15277 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
15280 AMD K6 CPU with MMX instruction set support.
15284 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
15287 @itemx athlon-tbird
15288 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
15294 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
15295 instruction set support.
15301 Processors based on the AMD K8 core with x86-64 instruction set support,
15302 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
15303 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
15304 instruction set extensions.)
15307 @itemx opteron-sse3
15308 @itemx athlon64-sse3
15309 Improved versions of AMD K8 cores with SSE3 instruction set support.
15313 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
15314 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
15315 instruction set extensions.)
15318 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
15319 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
15320 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
15322 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
15323 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
15324 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
15327 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
15328 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
15329 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
15330 64-bit instruction set extensions.
15332 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
15333 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
15334 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
15335 SSE4.2, ABM and 64-bit instruction set extensions.
15338 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
15339 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
15340 instruction set extensions.)
15343 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
15344 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
15345 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
15348 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
15352 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
15353 instruction set support.
15356 VIA C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
15357 implemented for this chip.)
15360 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
15362 implemented for this chip.)
15365 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
15368 @item -mtune=@var{cpu-type}
15370 Tune to @var{cpu-type} everything applicable about the generated code, except
15371 for the ABI and the set of available instructions.
15372 While picking a specific @var{cpu-type} schedules things appropriately
15373 for that particular chip, the compiler does not generate any code that
15374 cannot run on the default machine type unless you use a
15375 @option{-march=@var{cpu-type}} option.
15376 For example, if GCC is configured for i686-pc-linux-gnu
15377 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
15378 but still runs on i686 machines.
15380 The choices for @var{cpu-type} are the same as for @option{-march}.
15381 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
15385 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
15386 If you know the CPU on which your code will run, then you should use
15387 the corresponding @option{-mtune} or @option{-march} option instead of
15388 @option{-mtune=generic}. But, if you do not know exactly what CPU users
15389 of your application will have, then you should use this option.
15391 As new processors are deployed in the marketplace, the behavior of this
15392 option will change. Therefore, if you upgrade to a newer version of
15393 GCC, code generation controlled by this option will change to reflect
15395 that are most common at the time that version of GCC is released.
15397 There is no @option{-march=generic} option because @option{-march}
15398 indicates the instruction set the compiler can use, and there is no
15399 generic instruction set applicable to all processors. In contrast,
15400 @option{-mtune} indicates the processor (or, in this case, collection of
15401 processors) for which the code is optimized.
15404 Produce code optimized for the most current Intel processors, which are
15405 Haswell and Silvermont for this version of GCC. If you know the CPU
15406 on which your code will run, then you should use the corresponding
15407 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
15408 But, if you want your application performs better on both Haswell and
15409 Silvermont, then you should use this option.
15411 As new Intel processors are deployed in the marketplace, the behavior of
15412 this option will change. Therefore, if you upgrade to a newer version of
15413 GCC, code generation controlled by this option will change to reflect
15414 the most current Intel processors at the time that version of GCC is
15417 There is no @option{-march=intel} option because @option{-march} indicates
15418 the instruction set the compiler can use, and there is no common
15419 instruction set applicable to all processors. In contrast,
15420 @option{-mtune} indicates the processor (or, in this case, collection of
15421 processors) for which the code is optimized.
15424 @item -mcpu=@var{cpu-type}
15426 A deprecated synonym for @option{-mtune}.
15428 @item -mfpmath=@var{unit}
15430 Generate floating-point arithmetic for selected unit @var{unit}. The choices
15431 for @var{unit} are:
15435 Use the standard 387 floating-point coprocessor present on the majority of chips and
15436 emulated otherwise. Code compiled with this option runs almost everywhere.
15437 The temporary results are computed in 80-bit precision instead of the precision
15438 specified by the type, resulting in slightly different results compared to most
15439 of other chips. See @option{-ffloat-store} for more detailed description.
15441 This is the default choice for i386 compiler.
15444 Use scalar floating-point instructions present in the SSE instruction set.
15445 This instruction set is supported by Pentium III and newer chips,
15446 and in the AMD line
15447 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
15448 instruction set supports only single-precision arithmetic, thus the double and
15449 extended-precision arithmetic are still done using 387. A later version, present
15450 only in Pentium 4 and AMD x86-64 chips, supports double-precision
15453 For the i386 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
15454 or @option{-msse2} switches to enable SSE extensions and make this option
15455 effective. For the x86-64 compiler, these extensions are enabled by default.
15457 The resulting code should be considerably faster in the majority of cases and avoid
15458 the numerical instability problems of 387 code, but may break some existing
15459 code that expects temporaries to be 80 bits.
15461 This is the default choice for the x86-64 compiler.
15466 Attempt to utilize both instruction sets at once. This effectively doubles the
15467 amount of available registers, and on chips with separate execution units for
15468 387 and SSE the execution resources too. Use this option with care, as it is
15469 still experimental, because the GCC register allocator does not model separate
15470 functional units well, resulting in unstable performance.
15473 @item -masm=@var{dialect}
15474 @opindex masm=@var{dialect}
15475 Output assembly instructions using selected @var{dialect}. Supported
15476 choices are @samp{intel} or @samp{att} (the default). Darwin does
15477 not support @samp{intel}.
15480 @itemx -mno-ieee-fp
15482 @opindex mno-ieee-fp
15483 Control whether or not the compiler uses IEEE floating-point
15484 comparisons. These correctly handle the case where the result of a
15485 comparison is unordered.
15488 @opindex msoft-float
15489 Generate output containing library calls for floating point.
15491 @strong{Warning:} the requisite libraries are not part of GCC@.
15492 Normally the facilities of the machine's usual C compiler are used, but
15493 this can't be done directly in cross-compilation. You must make your
15494 own arrangements to provide suitable library functions for
15497 On machines where a function returns floating-point results in the 80387
15498 register stack, some floating-point opcodes may be emitted even if
15499 @option{-msoft-float} is used.
15501 @item -mno-fp-ret-in-387
15502 @opindex mno-fp-ret-in-387
15503 Do not use the FPU registers for return values of functions.
15505 The usual calling convention has functions return values of types
15506 @code{float} and @code{double} in an FPU register, even if there
15507 is no FPU@. The idea is that the operating system should emulate
15510 The option @option{-mno-fp-ret-in-387} causes such values to be returned
15511 in ordinary CPU registers instead.
15513 @item -mno-fancy-math-387
15514 @opindex mno-fancy-math-387
15515 Some 387 emulators do not support the @code{sin}, @code{cos} and
15516 @code{sqrt} instructions for the 387. Specify this option to avoid
15517 generating those instructions. This option is the default on FreeBSD,
15518 OpenBSD and NetBSD@. This option is overridden when @option{-march}
15519 indicates that the target CPU always has an FPU and so the
15520 instruction does not need emulation. These
15521 instructions are not generated unless you also use the
15522 @option{-funsafe-math-optimizations} switch.
15524 @item -malign-double
15525 @itemx -mno-align-double
15526 @opindex malign-double
15527 @opindex mno-align-double
15528 Control whether GCC aligns @code{double}, @code{long double}, and
15529 @code{long long} variables on a two-word boundary or a one-word
15530 boundary. Aligning @code{double} variables on a two-word boundary
15531 produces code that runs somewhat faster on a Pentium at the
15532 expense of more memory.
15534 On x86-64, @option{-malign-double} is enabled by default.
15536 @strong{Warning:} if you use the @option{-malign-double} switch,
15537 structures containing the above types are aligned differently than
15538 the published application binary interface specifications for the 386
15539 and are not binary compatible with structures in code compiled
15540 without that switch.
15542 @item -m96bit-long-double
15543 @itemx -m128bit-long-double
15544 @opindex m96bit-long-double
15545 @opindex m128bit-long-double
15546 These switches control the size of @code{long double} type. The i386
15547 application binary interface specifies the size to be 96 bits,
15548 so @option{-m96bit-long-double} is the default in 32-bit mode.
15550 Modern architectures (Pentium and newer) prefer @code{long double}
15551 to be aligned to an 8- or 16-byte boundary. In arrays or structures
15552 conforming to the ABI, this is not possible. So specifying
15553 @option{-m128bit-long-double} aligns @code{long double}
15554 to a 16-byte boundary by padding the @code{long double} with an additional
15557 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
15558 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
15560 Notice that neither of these options enable any extra precision over the x87
15561 standard of 80 bits for a @code{long double}.
15563 @strong{Warning:} if you override the default value for your target ABI, this
15564 changes the size of
15565 structures and arrays containing @code{long double} variables,
15566 as well as modifying the function calling convention for functions taking
15567 @code{long double}. Hence they are not binary-compatible
15568 with code compiled without that switch.
15570 @item -mlong-double-64
15571 @itemx -mlong-double-80
15572 @itemx -mlong-double-128
15573 @opindex mlong-double-64
15574 @opindex mlong-double-80
15575 @opindex mlong-double-128
15576 These switches control the size of @code{long double} type. A size
15577 of 64 bits makes the @code{long double} type equivalent to the @code{double}
15578 type. This is the default for 32-bit Bionic C library. A size
15579 of 128 bits makes the @code{long double} type equivalent to the
15580 @code{__float128} type. This is the default for 64-bit Bionic C library.
15582 @strong{Warning:} if you override the default value for your target ABI, this
15583 changes the size of
15584 structures and arrays containing @code{long double} variables,
15585 as well as modifying the function calling convention for functions taking
15586 @code{long double}. Hence they are not binary-compatible
15587 with code compiled without that switch.
15589 @item -mlarge-data-threshold=@var{threshold}
15590 @opindex mlarge-data-threshold
15591 When @option{-mcmodel=medium} is specified, data objects larger than
15592 @var{threshold} are placed in the large data section. This value must be the
15593 same across all objects linked into the binary, and defaults to 65535.
15597 Use a different function-calling convention, in which functions that
15598 take a fixed number of arguments return with the @code{ret @var{num}}
15599 instruction, which pops their arguments while returning. This saves one
15600 instruction in the caller since there is no need to pop the arguments
15603 You can specify that an individual function is called with this calling
15604 sequence with the function attribute @samp{stdcall}. You can also
15605 override the @option{-mrtd} option by using the function attribute
15606 @samp{cdecl}. @xref{Function Attributes}.
15608 @strong{Warning:} this calling convention is incompatible with the one
15609 normally used on Unix, so you cannot use it if you need to call
15610 libraries compiled with the Unix compiler.
15612 Also, you must provide function prototypes for all functions that
15613 take variable numbers of arguments (including @code{printf});
15614 otherwise incorrect code is generated for calls to those
15617 In addition, seriously incorrect code results if you call a
15618 function with too many arguments. (Normally, extra arguments are
15619 harmlessly ignored.)
15621 @item -mregparm=@var{num}
15623 Control how many registers are used to pass integer arguments. By
15624 default, no registers are used to pass arguments, and at most 3
15625 registers can be used. You can control this behavior for a specific
15626 function by using the function attribute @samp{regparm}.
15627 @xref{Function Attributes}.
15629 @strong{Warning:} if you use this switch, and
15630 @var{num} is nonzero, then you must build all modules with the same
15631 value, including any libraries. This includes the system libraries and
15635 @opindex msseregparm
15636 Use SSE register passing conventions for float and double arguments
15637 and return values. You can control this behavior for a specific
15638 function by using the function attribute @samp{sseregparm}.
15639 @xref{Function Attributes}.
15641 @strong{Warning:} if you use this switch then you must build all
15642 modules with the same value, including any libraries. This includes
15643 the system libraries and startup modules.
15645 @item -mvect8-ret-in-mem
15646 @opindex mvect8-ret-in-mem
15647 Return 8-byte vectors in memory instead of MMX registers. This is the
15648 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
15649 Studio compilers until version 12. Later compiler versions (starting
15650 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
15651 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
15652 you need to remain compatible with existing code produced by those
15653 previous compiler versions or older versions of GCC@.
15662 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
15663 is specified, the significands of results of floating-point operations are
15664 rounded to 24 bits (single precision); @option{-mpc64} rounds the
15665 significands of results of floating-point operations to 53 bits (double
15666 precision) and @option{-mpc80} rounds the significands of results of
15667 floating-point operations to 64 bits (extended double precision), which is
15668 the default. When this option is used, floating-point operations in higher
15669 precisions are not available to the programmer without setting the FPU
15670 control word explicitly.
15672 Setting the rounding of floating-point operations to less than the default
15673 80 bits can speed some programs by 2% or more. Note that some mathematical
15674 libraries assume that extended-precision (80-bit) floating-point operations
15675 are enabled by default; routines in such libraries could suffer significant
15676 loss of accuracy, typically through so-called ``catastrophic cancellation'',
15677 when this option is used to set the precision to less than extended precision.
15679 @item -mstackrealign
15680 @opindex mstackrealign
15681 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
15682 option generates an alternate prologue and epilogue that realigns the
15683 run-time stack if necessary. This supports mixing legacy codes that keep
15684 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
15685 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
15686 applicable to individual functions.
15688 @item -mpreferred-stack-boundary=@var{num}
15689 @opindex mpreferred-stack-boundary
15690 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
15691 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
15692 the default is 4 (16 bytes or 128 bits).
15694 @strong{Warning:} When generating code for the x86-64 architecture with
15695 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
15696 used to keep the stack boundary aligned to 8 byte boundary. Since
15697 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
15698 intended to be used in controlled environment where stack space is
15699 important limitation. This option will lead to wrong code when functions
15700 compiled with 16 byte stack alignment (such as functions from a standard
15701 library) are called with misaligned stack. In this case, SSE
15702 instructions may lead to misaligned memory access traps. In addition,
15703 variable arguments will be handled incorrectly for 16 byte aligned
15704 objects (including x87 long double and __int128), leading to wrong
15705 results. You must build all modules with
15706 @option{-mpreferred-stack-boundary=3}, including any libraries. This
15707 includes the system libraries and startup modules.
15709 @item -mincoming-stack-boundary=@var{num}
15710 @opindex mincoming-stack-boundary
15711 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
15712 boundary. If @option{-mincoming-stack-boundary} is not specified,
15713 the one specified by @option{-mpreferred-stack-boundary} is used.
15715 On Pentium and Pentium Pro, @code{double} and @code{long double} values
15716 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
15717 suffer significant run time performance penalties. On Pentium III, the
15718 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
15719 properly if it is not 16-byte aligned.
15721 To ensure proper alignment of this values on the stack, the stack boundary
15722 must be as aligned as that required by any value stored on the stack.
15723 Further, every function must be generated such that it keeps the stack
15724 aligned. Thus calling a function compiled with a higher preferred
15725 stack boundary from a function compiled with a lower preferred stack
15726 boundary most likely misaligns the stack. It is recommended that
15727 libraries that use callbacks always use the default setting.
15729 This extra alignment does consume extra stack space, and generally
15730 increases code size. Code that is sensitive to stack space usage, such
15731 as embedded systems and operating system kernels, may want to reduce the
15732 preferred alignment to @option{-mpreferred-stack-boundary=2}.
15756 @itemx -mno-avx512f
15759 @itemx -mno-avx512pf
15761 @itemx -mno-avx512er
15763 @itemx -mno-avx512cd
15771 @itemx -mno-clflsuhopt
15774 @itemx -mno-fsgsbase
15781 @itemx -mprefetchwt1
15782 @itemx -mno-prefetchwt1
15822 These switches enable or disable the use of instructions in the MMX, SSE,
15823 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
15824 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
15825 BMI, BMI2, FXSR, XSAVE, XSAVEOPT, LZCNT, RTM, MPX or 3DNow!@:
15826 extended instruction sets.
15827 These extensions are also available as built-in functions: see
15828 @ref{X86 Built-in Functions}, for details of the functions enabled and
15829 disabled by these switches.
15831 To generate SSE/SSE2 instructions automatically from floating-point
15832 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
15834 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
15835 generates new AVX instructions or AVX equivalence for all SSEx instructions
15838 These options enable GCC to use these extended instructions in
15839 generated code, even without @option{-mfpmath=sse}. Applications that
15840 perform run-time CPU detection must compile separate files for each
15841 supported architecture, using the appropriate flags. In particular,
15842 the file containing the CPU detection code should be compiled without
15845 @item -mdump-tune-features
15846 @opindex mdump-tune-features
15847 This option instructs GCC to dump the names of the x86 performance
15848 tuning features and default settings. The names can be used in
15849 @option{-mtune-ctrl=@var{feature-list}}.
15851 @item -mtune-ctrl=@var{feature-list}
15852 @opindex mtune-ctrl=@var{feature-list}
15853 This option is used to do fine grain control of x86 code generation features.
15854 @var{feature-list} is a comma separated list of @var{feature} names. See also
15855 @option{-mdump-tune-features}. When specified, the @var{feature} will be turned
15856 on if it is not preceded with @code{^}, otherwise, it will be turned off.
15857 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
15858 developers. Using it may lead to code paths not covered by testing and can
15859 potentially result in compiler ICEs or runtime errors.
15862 @opindex mno-default
15863 This option instructs GCC to turn off all tunable features. See also
15864 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
15868 This option instructs GCC to emit a @code{cld} instruction in the prologue
15869 of functions that use string instructions. String instructions depend on
15870 the DF flag to select between autoincrement or autodecrement mode. While the
15871 ABI specifies the DF flag to be cleared on function entry, some operating
15872 systems violate this specification by not clearing the DF flag in their
15873 exception dispatchers. The exception handler can be invoked with the DF flag
15874 set, which leads to wrong direction mode when string instructions are used.
15875 This option can be enabled by default on 32-bit x86 targets by configuring
15876 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
15877 instructions can be suppressed with the @option{-mno-cld} compiler option
15881 @opindex mvzeroupper
15882 This option instructs GCC to emit a @code{vzeroupper} instruction
15883 before a transfer of control flow out of the function to minimize
15884 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
15887 @item -mprefer-avx128
15888 @opindex mprefer-avx128
15889 This option instructs GCC to use 128-bit AVX instructions instead of
15890 256-bit AVX instructions in the auto-vectorizer.
15894 This option enables GCC to generate @code{CMPXCHG16B} instructions.
15895 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
15896 (or oword) data types.
15897 This is useful for high-resolution counters that can be updated
15898 by multiple processors (or cores). This instruction is generated as part of
15899 atomic built-in functions: see @ref{__sync Builtins} or
15900 @ref{__atomic Builtins} for details.
15904 This option enables generation of @code{SAHF} instructions in 64-bit code.
15905 Early Intel Pentium 4 CPUs with Intel 64 support,
15906 prior to the introduction of Pentium 4 G1 step in December 2005,
15907 lacked the @code{LAHF} and @code{SAHF} instructions
15908 which were supported by AMD64.
15909 These are load and store instructions, respectively, for certain status flags.
15910 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
15911 @code{drem}, and @code{remainder} built-in functions;
15912 see @ref{Other Builtins} for details.
15916 This option enables use of the @code{movbe} instruction to implement
15917 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
15921 This option enables built-in functions @code{__builtin_ia32_crc32qi},
15922 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
15923 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
15927 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
15928 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
15929 with an additional Newton-Raphson step
15930 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
15931 (and their vectorized
15932 variants) for single-precision floating-point arguments. These instructions
15933 are generated only when @option{-funsafe-math-optimizations} is enabled
15934 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
15935 Note that while the throughput of the sequence is higher than the throughput
15936 of the non-reciprocal instruction, the precision of the sequence can be
15937 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
15939 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
15940 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
15941 combination), and doesn't need @option{-mrecip}.
15943 Also note that GCC emits the above sequence with additional Newton-Raphson step
15944 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
15945 already with @option{-ffast-math} (or the above option combination), and
15946 doesn't need @option{-mrecip}.
15948 @item -mrecip=@var{opt}
15949 @opindex mrecip=opt
15950 This option controls which reciprocal estimate instructions
15951 may be used. @var{opt} is a comma-separated list of options, which may
15952 be preceded by a @samp{!} to invert the option:
15956 Enable all estimate instructions.
15959 Enable the default instructions, equivalent to @option{-mrecip}.
15962 Disable all estimate instructions, equivalent to @option{-mno-recip}.
15965 Enable the approximation for scalar division.
15968 Enable the approximation for vectorized division.
15971 Enable the approximation for scalar square root.
15974 Enable the approximation for vectorized square root.
15977 So, for example, @option{-mrecip=all,!sqrt} enables
15978 all of the reciprocal approximations, except for square root.
15980 @item -mveclibabi=@var{type}
15981 @opindex mveclibabi
15982 Specifies the ABI type to use for vectorizing intrinsics using an
15983 external library. Supported values for @var{type} are @samp{svml}
15984 for the Intel short
15985 vector math library and @samp{acml} for the AMD math core library.
15986 To use this option, both @option{-ftree-vectorize} and
15987 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
15988 ABI-compatible library must be specified at link time.
15990 GCC currently emits calls to @code{vmldExp2},
15991 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
15992 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
15993 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
15994 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
15995 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
15996 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
15997 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
15998 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
15999 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
16000 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
16001 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
16002 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
16003 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
16004 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
16005 when @option{-mveclibabi=acml} is used.
16007 @item -mabi=@var{name}
16009 Generate code for the specified calling convention. Permissible values
16010 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
16011 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
16012 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
16013 You can control this behavior for a specific function by
16014 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
16015 @xref{Function Attributes}.
16017 @item -mtls-dialect=@var{type}
16018 @opindex mtls-dialect
16019 Generate code to access thread-local storage using the @samp{gnu} or
16020 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
16021 @samp{gnu2} is more efficient, but it may add compile- and run-time
16022 requirements that cannot be satisfied on all systems.
16025 @itemx -mno-push-args
16026 @opindex mpush-args
16027 @opindex mno-push-args
16028 Use PUSH operations to store outgoing parameters. This method is shorter
16029 and usually equally fast as method using SUB/MOV operations and is enabled
16030 by default. In some cases disabling it may improve performance because of
16031 improved scheduling and reduced dependencies.
16033 @item -maccumulate-outgoing-args
16034 @opindex maccumulate-outgoing-args
16035 If enabled, the maximum amount of space required for outgoing arguments is
16036 computed in the function prologue. This is faster on most modern CPUs
16037 because of reduced dependencies, improved scheduling and reduced stack usage
16038 when the preferred stack boundary is not equal to 2. The drawback is a notable
16039 increase in code size. This switch implies @option{-mno-push-args}.
16043 Support thread-safe exception handling on MinGW. Programs that rely
16044 on thread-safe exception handling must compile and link all code with the
16045 @option{-mthreads} option. When compiling, @option{-mthreads} defines
16046 @code{-D_MT}; when linking, it links in a special thread helper library
16047 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
16049 @item -mno-align-stringops
16050 @opindex mno-align-stringops
16051 Do not align the destination of inlined string operations. This switch reduces
16052 code size and improves performance in case the destination is already aligned,
16053 but GCC doesn't know about it.
16055 @item -minline-all-stringops
16056 @opindex minline-all-stringops
16057 By default GCC inlines string operations only when the destination is
16058 known to be aligned to least a 4-byte boundary.
16059 This enables more inlining and increases code
16060 size, but may improve performance of code that depends on fast
16061 @code{memcpy}, @code{strlen},
16062 and @code{memset} for short lengths.
16064 @item -minline-stringops-dynamically
16065 @opindex minline-stringops-dynamically
16066 For string operations of unknown size, use run-time checks with
16067 inline code for small blocks and a library call for large blocks.
16069 @item -mstringop-strategy=@var{alg}
16070 @opindex mstringop-strategy=@var{alg}
16071 Override the internal decision heuristic for the particular algorithm to use
16072 for inlining string operations. The allowed values for @var{alg} are:
16078 Expand using i386 @code{rep} prefix of the specified size.
16082 @itemx unrolled_loop
16083 Expand into an inline loop.
16086 Always use a library call.
16089 @item -mmemcpy-strategy=@var{strategy}
16090 @opindex mmemcpy-strategy=@var{strategy}
16091 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
16092 should be inlined and what inline algorithm to use when the expected size
16093 of the copy operation is known. @var{strategy}
16094 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
16095 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
16096 the max byte size with which inline algorithm @var{alg} is allowed. For the last
16097 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
16098 in the list must be specified in increasing order. The minimal byte size for
16099 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
16102 @item -mmemset-strategy=@var{strategy}
16103 @opindex mmemset-strategy=@var{strategy}
16104 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
16105 @code{__builtin_memset} expansion.
16107 @item -momit-leaf-frame-pointer
16108 @opindex momit-leaf-frame-pointer
16109 Don't keep the frame pointer in a register for leaf functions. This
16110 avoids the instructions to save, set up, and restore frame pointers and
16111 makes an extra register available in leaf functions. The option
16112 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
16113 which might make debugging harder.
16115 @item -mtls-direct-seg-refs
16116 @itemx -mno-tls-direct-seg-refs
16117 @opindex mtls-direct-seg-refs
16118 Controls whether TLS variables may be accessed with offsets from the
16119 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
16120 or whether the thread base pointer must be added. Whether or not this
16121 is valid depends on the operating system, and whether it maps the
16122 segment to cover the entire TLS area.
16124 For systems that use the GNU C Library, the default is on.
16127 @itemx -mno-sse2avx
16129 Specify that the assembler should encode SSE instructions with VEX
16130 prefix. The option @option{-mavx} turns this on by default.
16135 If profiling is active (@option{-pg}), put the profiling
16136 counter call before the prologue.
16137 Note: On x86 architectures the attribute @code{ms_hook_prologue}
16138 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
16140 @item -mrecord-mcount
16141 @itemx -mno-record-mcount
16142 @opindex mrecord-mcount
16143 If profiling is active (@option{-pg}), generate a __mcount_loc section
16144 that contains pointers to each profiling call. This is useful for
16145 automatically patching and out calls.
16148 @itemx -mno-nop-mcount
16149 @opindex mnop-mcount
16150 If profiling is active (@option{-pg}), generate the calls to
16151 the profiling functions as nops. This is useful when they
16152 should be patched in later dynamically. This is likely only
16153 useful together with @option{-mrecord-mcount}.
16156 @itemx -mno-8bit-idiv
16158 On some processors, like Intel Atom, 8-bit unsigned integer divide is
16159 much faster than 32-bit/64-bit integer divide. This option generates a
16160 run-time check. If both dividend and divisor are within range of 0
16161 to 255, 8-bit unsigned integer divide is used instead of
16162 32-bit/64-bit integer divide.
16164 @item -mavx256-split-unaligned-load
16165 @itemx -mavx256-split-unaligned-store
16166 @opindex avx256-split-unaligned-load
16167 @opindex avx256-split-unaligned-store
16168 Split 32-byte AVX unaligned load and store.
16170 @item -mstack-protector-guard=@var{guard}
16171 @opindex mstack-protector-guard=@var{guard}
16172 Generate stack protection code using canary at @var{guard}. Supported
16173 locations are @samp{global} for global canary or @samp{tls} for per-thread
16174 canary in the TLS block (the default). This option has effect only when
16175 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
16179 These @samp{-m} switches are supported in addition to the above
16180 on x86-64 processors in 64-bit environments.
16191 Generate code for a 16-bit, 32-bit or 64-bit environment.
16192 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
16194 generates code that runs on any i386 system.
16196 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
16197 types to 64 bits, and generates code for the x86-64 architecture.
16198 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
16199 and @option{-mdynamic-no-pic} options.
16201 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
16203 generates code for the x86-64 architecture.
16205 The @option{-m16} option is the same as @option{-m32}, except for that
16206 it outputs the @code{.code16gcc} assembly directive at the beginning of
16207 the assembly output so that the binary can run in 16-bit mode.
16209 @item -mno-red-zone
16210 @opindex mno-red-zone
16211 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
16212 by the x86-64 ABI; it is a 128-byte area beyond the location of the
16213 stack pointer that is not modified by signal or interrupt handlers
16214 and therefore can be used for temporary data without adjusting the stack
16215 pointer. The flag @option{-mno-red-zone} disables this red zone.
16217 @item -mcmodel=small
16218 @opindex mcmodel=small
16219 Generate code for the small code model: the program and its symbols must
16220 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
16221 Programs can be statically or dynamically linked. This is the default
16224 @item -mcmodel=kernel
16225 @opindex mcmodel=kernel
16226 Generate code for the kernel code model. The kernel runs in the
16227 negative 2 GB of the address space.
16228 This model has to be used for Linux kernel code.
16230 @item -mcmodel=medium
16231 @opindex mcmodel=medium
16232 Generate code for the medium model: the program is linked in the lower 2
16233 GB of the address space. Small symbols are also placed there. Symbols
16234 with sizes larger than @option{-mlarge-data-threshold} are put into
16235 large data or BSS sections and can be located above 2GB. Programs can
16236 be statically or dynamically linked.
16238 @item -mcmodel=large
16239 @opindex mcmodel=large
16240 Generate code for the large model. This model makes no assumptions
16241 about addresses and sizes of sections.
16243 @item -maddress-mode=long
16244 @opindex maddress-mode=long
16245 Generate code for long address mode. This is only supported for 64-bit
16246 and x32 environments. It is the default address mode for 64-bit
16249 @item -maddress-mode=short
16250 @opindex maddress-mode=short
16251 Generate code for short address mode. This is only supported for 32-bit
16252 and x32 environments. It is the default address mode for 32-bit and
16256 @node i386 and x86-64 Windows Options
16257 @subsection i386 and x86-64 Windows Options
16258 @cindex i386 and x86-64 Windows Options
16260 These additional options are available for Microsoft Windows targets:
16266 specifies that a console application is to be generated, by
16267 instructing the linker to set the PE header subsystem type
16268 required for console applications.
16269 This option is available for Cygwin and MinGW targets and is
16270 enabled by default on those targets.
16274 This option is available for Cygwin and MinGW targets. It
16275 specifies that a DLL---a dynamic link library---is to be
16276 generated, enabling the selection of the required runtime
16277 startup object and entry point.
16279 @item -mnop-fun-dllimport
16280 @opindex mnop-fun-dllimport
16281 This option is available for Cygwin and MinGW targets. It
16282 specifies that the @code{dllimport} attribute should be ignored.
16286 This option is available for MinGW targets. It specifies
16287 that MinGW-specific thread support is to be used.
16291 This option is available for MinGW-w64 targets. It causes
16292 the @code{UNICODE} preprocessor macro to be predefined, and
16293 chooses Unicode-capable runtime startup code.
16297 This option is available for Cygwin and MinGW targets. It
16298 specifies that the typical Microsoft Windows predefined macros are to
16299 be set in the pre-processor, but does not influence the choice
16300 of runtime library/startup code.
16304 This option is available for Cygwin and MinGW targets. It
16305 specifies that a GUI application is to be generated by
16306 instructing the linker to set the PE header subsystem type
16309 @item -fno-set-stack-executable
16310 @opindex fno-set-stack-executable
16311 This option is available for MinGW targets. It specifies that
16312 the executable flag for the stack used by nested functions isn't
16313 set. This is necessary for binaries running in kernel mode of
16314 Microsoft Windows, as there the User32 API, which is used to set executable
16315 privileges, isn't available.
16317 @item -fwritable-relocated-rdata
16318 @opindex fno-writable-relocated-rdata
16319 This option is available for MinGW and Cygwin targets. It specifies
16320 that relocated-data in read-only section is put into .data
16321 section. This is a necessary for older runtimes not supporting
16322 modification of .rdata sections for pseudo-relocation.
16324 @item -mpe-aligned-commons
16325 @opindex mpe-aligned-commons
16326 This option is available for Cygwin and MinGW targets. It
16327 specifies that the GNU extension to the PE file format that
16328 permits the correct alignment of COMMON variables should be
16329 used when generating code. It is enabled by default if
16330 GCC detects that the target assembler found during configuration
16331 supports the feature.
16334 See also under @ref{i386 and x86-64 Options} for standard options.
16336 @node IA-64 Options
16337 @subsection IA-64 Options
16338 @cindex IA-64 Options
16340 These are the @samp{-m} options defined for the Intel IA-64 architecture.
16344 @opindex mbig-endian
16345 Generate code for a big-endian target. This is the default for HP-UX@.
16347 @item -mlittle-endian
16348 @opindex mlittle-endian
16349 Generate code for a little-endian target. This is the default for AIX5
16355 @opindex mno-gnu-as
16356 Generate (or don't) code for the GNU assembler. This is the default.
16357 @c Also, this is the default if the configure option @option{--with-gnu-as}
16363 @opindex mno-gnu-ld
16364 Generate (or don't) code for the GNU linker. This is the default.
16365 @c Also, this is the default if the configure option @option{--with-gnu-ld}
16370 Generate code that does not use a global pointer register. The result
16371 is not position independent code, and violates the IA-64 ABI@.
16373 @item -mvolatile-asm-stop
16374 @itemx -mno-volatile-asm-stop
16375 @opindex mvolatile-asm-stop
16376 @opindex mno-volatile-asm-stop
16377 Generate (or don't) a stop bit immediately before and after volatile asm
16380 @item -mregister-names
16381 @itemx -mno-register-names
16382 @opindex mregister-names
16383 @opindex mno-register-names
16384 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
16385 the stacked registers. This may make assembler output more readable.
16391 Disable (or enable) optimizations that use the small data section. This may
16392 be useful for working around optimizer bugs.
16394 @item -mconstant-gp
16395 @opindex mconstant-gp
16396 Generate code that uses a single constant global pointer value. This is
16397 useful when compiling kernel code.
16401 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
16402 This is useful when compiling firmware code.
16404 @item -minline-float-divide-min-latency
16405 @opindex minline-float-divide-min-latency
16406 Generate code for inline divides of floating-point values
16407 using the minimum latency algorithm.
16409 @item -minline-float-divide-max-throughput
16410 @opindex minline-float-divide-max-throughput
16411 Generate code for inline divides of floating-point values
16412 using the maximum throughput algorithm.
16414 @item -mno-inline-float-divide
16415 @opindex mno-inline-float-divide
16416 Do not generate inline code for divides of floating-point values.
16418 @item -minline-int-divide-min-latency
16419 @opindex minline-int-divide-min-latency
16420 Generate code for inline divides of integer values
16421 using the minimum latency algorithm.
16423 @item -minline-int-divide-max-throughput
16424 @opindex minline-int-divide-max-throughput
16425 Generate code for inline divides of integer values
16426 using the maximum throughput algorithm.
16428 @item -mno-inline-int-divide
16429 @opindex mno-inline-int-divide
16430 Do not generate inline code for divides of integer values.
16432 @item -minline-sqrt-min-latency
16433 @opindex minline-sqrt-min-latency
16434 Generate code for inline square roots
16435 using the minimum latency algorithm.
16437 @item -minline-sqrt-max-throughput
16438 @opindex minline-sqrt-max-throughput
16439 Generate code for inline square roots
16440 using the maximum throughput algorithm.
16442 @item -mno-inline-sqrt
16443 @opindex mno-inline-sqrt
16444 Do not generate inline code for @code{sqrt}.
16447 @itemx -mno-fused-madd
16448 @opindex mfused-madd
16449 @opindex mno-fused-madd
16450 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
16451 instructions. The default is to use these instructions.
16453 @item -mno-dwarf2-asm
16454 @itemx -mdwarf2-asm
16455 @opindex mno-dwarf2-asm
16456 @opindex mdwarf2-asm
16457 Don't (or do) generate assembler code for the DWARF 2 line number debugging
16458 info. This may be useful when not using the GNU assembler.
16460 @item -mearly-stop-bits
16461 @itemx -mno-early-stop-bits
16462 @opindex mearly-stop-bits
16463 @opindex mno-early-stop-bits
16464 Allow stop bits to be placed earlier than immediately preceding the
16465 instruction that triggered the stop bit. This can improve instruction
16466 scheduling, but does not always do so.
16468 @item -mfixed-range=@var{register-range}
16469 @opindex mfixed-range
16470 Generate code treating the given register range as fixed registers.
16471 A fixed register is one that the register allocator cannot use. This is
16472 useful when compiling kernel code. A register range is specified as
16473 two registers separated by a dash. Multiple register ranges can be
16474 specified separated by a comma.
16476 @item -mtls-size=@var{tls-size}
16478 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
16481 @item -mtune=@var{cpu-type}
16483 Tune the instruction scheduling for a particular CPU, Valid values are
16484 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
16485 and @samp{mckinley}.
16491 Generate code for a 32-bit or 64-bit environment.
16492 The 32-bit environment sets int, long and pointer to 32 bits.
16493 The 64-bit environment sets int to 32 bits and long and pointer
16494 to 64 bits. These are HP-UX specific flags.
16496 @item -mno-sched-br-data-spec
16497 @itemx -msched-br-data-spec
16498 @opindex mno-sched-br-data-spec
16499 @opindex msched-br-data-spec
16500 (Dis/En)able data speculative scheduling before reload.
16501 This results in generation of @code{ld.a} instructions and
16502 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16503 The default is 'disable'.
16505 @item -msched-ar-data-spec
16506 @itemx -mno-sched-ar-data-spec
16507 @opindex msched-ar-data-spec
16508 @opindex mno-sched-ar-data-spec
16509 (En/Dis)able data speculative scheduling after reload.
16510 This results in generation of @code{ld.a} instructions and
16511 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16512 The default is 'enable'.
16514 @item -mno-sched-control-spec
16515 @itemx -msched-control-spec
16516 @opindex mno-sched-control-spec
16517 @opindex msched-control-spec
16518 (Dis/En)able control speculative scheduling. This feature is
16519 available only during region scheduling (i.e.@: before reload).
16520 This results in generation of the @code{ld.s} instructions and
16521 the corresponding check instructions @code{chk.s}.
16522 The default is 'disable'.
16524 @item -msched-br-in-data-spec
16525 @itemx -mno-sched-br-in-data-spec
16526 @opindex msched-br-in-data-spec
16527 @opindex mno-sched-br-in-data-spec
16528 (En/Dis)able speculative scheduling of the instructions that
16529 are dependent on the data speculative loads before reload.
16530 This is effective only with @option{-msched-br-data-spec} enabled.
16531 The default is 'enable'.
16533 @item -msched-ar-in-data-spec
16534 @itemx -mno-sched-ar-in-data-spec
16535 @opindex msched-ar-in-data-spec
16536 @opindex mno-sched-ar-in-data-spec
16537 (En/Dis)able speculative scheduling of the instructions that
16538 are dependent on the data speculative loads after reload.
16539 This is effective only with @option{-msched-ar-data-spec} enabled.
16540 The default is 'enable'.
16542 @item -msched-in-control-spec
16543 @itemx -mno-sched-in-control-spec
16544 @opindex msched-in-control-spec
16545 @opindex mno-sched-in-control-spec
16546 (En/Dis)able speculative scheduling of the instructions that
16547 are dependent on the control speculative loads.
16548 This is effective only with @option{-msched-control-spec} enabled.
16549 The default is 'enable'.
16551 @item -mno-sched-prefer-non-data-spec-insns
16552 @itemx -msched-prefer-non-data-spec-insns
16553 @opindex mno-sched-prefer-non-data-spec-insns
16554 @opindex msched-prefer-non-data-spec-insns
16555 If enabled, data-speculative instructions are chosen for schedule
16556 only if there are no other choices at the moment. This makes
16557 the use of the data speculation much more conservative.
16558 The default is 'disable'.
16560 @item -mno-sched-prefer-non-control-spec-insns
16561 @itemx -msched-prefer-non-control-spec-insns
16562 @opindex mno-sched-prefer-non-control-spec-insns
16563 @opindex msched-prefer-non-control-spec-insns
16564 If enabled, control-speculative instructions are chosen for schedule
16565 only if there are no other choices at the moment. This makes
16566 the use of the control speculation much more conservative.
16567 The default is 'disable'.
16569 @item -mno-sched-count-spec-in-critical-path
16570 @itemx -msched-count-spec-in-critical-path
16571 @opindex mno-sched-count-spec-in-critical-path
16572 @opindex msched-count-spec-in-critical-path
16573 If enabled, speculative dependencies are considered during
16574 computation of the instructions priorities. This makes the use of the
16575 speculation a bit more conservative.
16576 The default is 'disable'.
16578 @item -msched-spec-ldc
16579 @opindex msched-spec-ldc
16580 Use a simple data speculation check. This option is on by default.
16582 @item -msched-control-spec-ldc
16583 @opindex msched-spec-ldc
16584 Use a simple check for control speculation. This option is on by default.
16586 @item -msched-stop-bits-after-every-cycle
16587 @opindex msched-stop-bits-after-every-cycle
16588 Place a stop bit after every cycle when scheduling. This option is on
16591 @item -msched-fp-mem-deps-zero-cost
16592 @opindex msched-fp-mem-deps-zero-cost
16593 Assume that floating-point stores and loads are not likely to cause a conflict
16594 when placed into the same instruction group. This option is disabled by
16597 @item -msel-sched-dont-check-control-spec
16598 @opindex msel-sched-dont-check-control-spec
16599 Generate checks for control speculation in selective scheduling.
16600 This flag is disabled by default.
16602 @item -msched-max-memory-insns=@var{max-insns}
16603 @opindex msched-max-memory-insns
16604 Limit on the number of memory insns per instruction group, giving lower
16605 priority to subsequent memory insns attempting to schedule in the same
16606 instruction group. Frequently useful to prevent cache bank conflicts.
16607 The default value is 1.
16609 @item -msched-max-memory-insns-hard-limit
16610 @opindex msched-max-memory-insns-hard-limit
16611 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
16612 disallowing more than that number in an instruction group.
16613 Otherwise, the limit is ``soft'', meaning that non-memory operations
16614 are preferred when the limit is reached, but memory operations may still
16620 @subsection LM32 Options
16621 @cindex LM32 options
16623 These @option{-m} options are defined for the LatticeMico32 architecture:
16626 @item -mbarrel-shift-enabled
16627 @opindex mbarrel-shift-enabled
16628 Enable barrel-shift instructions.
16630 @item -mdivide-enabled
16631 @opindex mdivide-enabled
16632 Enable divide and modulus instructions.
16634 @item -mmultiply-enabled
16635 @opindex multiply-enabled
16636 Enable multiply instructions.
16638 @item -msign-extend-enabled
16639 @opindex msign-extend-enabled
16640 Enable sign extend instructions.
16642 @item -muser-enabled
16643 @opindex muser-enabled
16644 Enable user-defined instructions.
16649 @subsection M32C Options
16650 @cindex M32C options
16653 @item -mcpu=@var{name}
16655 Select the CPU for which code is generated. @var{name} may be one of
16656 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
16657 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
16658 the M32C/80 series.
16662 Specifies that the program will be run on the simulator. This causes
16663 an alternate runtime library to be linked in which supports, for
16664 example, file I/O@. You must not use this option when generating
16665 programs that will run on real hardware; you must provide your own
16666 runtime library for whatever I/O functions are needed.
16668 @item -memregs=@var{number}
16670 Specifies the number of memory-based pseudo-registers GCC uses
16671 during code generation. These pseudo-registers are used like real
16672 registers, so there is a tradeoff between GCC's ability to fit the
16673 code into available registers, and the performance penalty of using
16674 memory instead of registers. Note that all modules in a program must
16675 be compiled with the same value for this option. Because of that, you
16676 must not use this option with GCC's default runtime libraries.
16680 @node M32R/D Options
16681 @subsection M32R/D Options
16682 @cindex M32R/D options
16684 These @option{-m} options are defined for Renesas M32R/D architectures:
16689 Generate code for the M32R/2@.
16693 Generate code for the M32R/X@.
16697 Generate code for the M32R@. This is the default.
16699 @item -mmodel=small
16700 @opindex mmodel=small
16701 Assume all objects live in the lower 16MB of memory (so that their addresses
16702 can be loaded with the @code{ld24} instruction), and assume all subroutines
16703 are reachable with the @code{bl} instruction.
16704 This is the default.
16706 The addressability of a particular object can be set with the
16707 @code{model} attribute.
16709 @item -mmodel=medium
16710 @opindex mmodel=medium
16711 Assume objects may be anywhere in the 32-bit address space (the compiler
16712 generates @code{seth/add3} instructions to load their addresses), and
16713 assume all subroutines are reachable with the @code{bl} instruction.
16715 @item -mmodel=large
16716 @opindex mmodel=large
16717 Assume objects may be anywhere in the 32-bit address space (the compiler
16718 generates @code{seth/add3} instructions to load their addresses), and
16719 assume subroutines may not be reachable with the @code{bl} instruction
16720 (the compiler generates the much slower @code{seth/add3/jl}
16721 instruction sequence).
16724 @opindex msdata=none
16725 Disable use of the small data area. Variables are put into
16726 one of @samp{.data}, @samp{.bss}, or @samp{.rodata} (unless the
16727 @code{section} attribute has been specified).
16728 This is the default.
16730 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
16731 Objects may be explicitly put in the small data area with the
16732 @code{section} attribute using one of these sections.
16734 @item -msdata=sdata
16735 @opindex msdata=sdata
16736 Put small global and static data in the small data area, but do not
16737 generate special code to reference them.
16740 @opindex msdata=use
16741 Put small global and static data in the small data area, and generate
16742 special instructions to reference them.
16746 @cindex smaller data references
16747 Put global and static objects less than or equal to @var{num} bytes
16748 into the small data or BSS sections instead of the normal data or BSS
16749 sections. The default value of @var{num} is 8.
16750 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
16751 for this option to have any effect.
16753 All modules should be compiled with the same @option{-G @var{num}} value.
16754 Compiling with different values of @var{num} may or may not work; if it
16755 doesn't the linker gives an error message---incorrect code is not
16760 Makes the M32R-specific code in the compiler display some statistics
16761 that might help in debugging programs.
16763 @item -malign-loops
16764 @opindex malign-loops
16765 Align all loops to a 32-byte boundary.
16767 @item -mno-align-loops
16768 @opindex mno-align-loops
16769 Do not enforce a 32-byte alignment for loops. This is the default.
16771 @item -missue-rate=@var{number}
16772 @opindex missue-rate=@var{number}
16773 Issue @var{number} instructions per cycle. @var{number} can only be 1
16776 @item -mbranch-cost=@var{number}
16777 @opindex mbranch-cost=@var{number}
16778 @var{number} can only be 1 or 2. If it is 1 then branches are
16779 preferred over conditional code, if it is 2, then the opposite applies.
16781 @item -mflush-trap=@var{number}
16782 @opindex mflush-trap=@var{number}
16783 Specifies the trap number to use to flush the cache. The default is
16784 12. Valid numbers are between 0 and 15 inclusive.
16786 @item -mno-flush-trap
16787 @opindex mno-flush-trap
16788 Specifies that the cache cannot be flushed by using a trap.
16790 @item -mflush-func=@var{name}
16791 @opindex mflush-func=@var{name}
16792 Specifies the name of the operating system function to call to flush
16793 the cache. The default is @emph{_flush_cache}, but a function call
16794 is only used if a trap is not available.
16796 @item -mno-flush-func
16797 @opindex mno-flush-func
16798 Indicates that there is no OS function for flushing the cache.
16802 @node M680x0 Options
16803 @subsection M680x0 Options
16804 @cindex M680x0 options
16806 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
16807 The default settings depend on which architecture was selected when
16808 the compiler was configured; the defaults for the most common choices
16812 @item -march=@var{arch}
16814 Generate code for a specific M680x0 or ColdFire instruction set
16815 architecture. Permissible values of @var{arch} for M680x0
16816 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
16817 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
16818 architectures are selected according to Freescale's ISA classification
16819 and the permissible values are: @samp{isaa}, @samp{isaaplus},
16820 @samp{isab} and @samp{isac}.
16822 GCC defines a macro @samp{__mcf@var{arch}__} whenever it is generating
16823 code for a ColdFire target. The @var{arch} in this macro is one of the
16824 @option{-march} arguments given above.
16826 When used together, @option{-march} and @option{-mtune} select code
16827 that runs on a family of similar processors but that is optimized
16828 for a particular microarchitecture.
16830 @item -mcpu=@var{cpu}
16832 Generate code for a specific M680x0 or ColdFire processor.
16833 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
16834 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
16835 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
16836 below, which also classifies the CPUs into families:
16838 @multitable @columnfractions 0.20 0.80
16839 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
16840 @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}
16841 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
16842 @item @samp{5206e} @tab @samp{5206e}
16843 @item @samp{5208} @tab @samp{5207} @samp{5208}
16844 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
16845 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
16846 @item @samp{5216} @tab @samp{5214} @samp{5216}
16847 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
16848 @item @samp{5225} @tab @samp{5224} @samp{5225}
16849 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
16850 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
16851 @item @samp{5249} @tab @samp{5249}
16852 @item @samp{5250} @tab @samp{5250}
16853 @item @samp{5271} @tab @samp{5270} @samp{5271}
16854 @item @samp{5272} @tab @samp{5272}
16855 @item @samp{5275} @tab @samp{5274} @samp{5275}
16856 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
16857 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
16858 @item @samp{5307} @tab @samp{5307}
16859 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
16860 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
16861 @item @samp{5407} @tab @samp{5407}
16862 @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}
16865 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
16866 @var{arch} is compatible with @var{cpu}. Other combinations of
16867 @option{-mcpu} and @option{-march} are rejected.
16869 GCC defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
16870 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
16871 where the value of @var{family} is given by the table above.
16873 @item -mtune=@var{tune}
16875 Tune the code for a particular microarchitecture within the
16876 constraints set by @option{-march} and @option{-mcpu}.
16877 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
16878 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
16879 and @samp{cpu32}. The ColdFire microarchitectures
16880 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
16882 You can also use @option{-mtune=68020-40} for code that needs
16883 to run relatively well on 68020, 68030 and 68040 targets.
16884 @option{-mtune=68020-60} is similar but includes 68060 targets
16885 as well. These two options select the same tuning decisions as
16886 @option{-m68020-40} and @option{-m68020-60} respectively.
16888 GCC defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
16889 when tuning for 680x0 architecture @var{arch}. It also defines
16890 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
16891 option is used. If GCC is tuning for a range of architectures,
16892 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
16893 it defines the macros for every architecture in the range.
16895 GCC also defines the macro @samp{__m@var{uarch}__} when tuning for
16896 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
16897 of the arguments given above.
16903 Generate output for a 68000. This is the default
16904 when the compiler is configured for 68000-based systems.
16905 It is equivalent to @option{-march=68000}.
16907 Use this option for microcontrollers with a 68000 or EC000 core,
16908 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
16912 Generate output for a 68010. This is the default
16913 when the compiler is configured for 68010-based systems.
16914 It is equivalent to @option{-march=68010}.
16920 Generate output for a 68020. This is the default
16921 when the compiler is configured for 68020-based systems.
16922 It is equivalent to @option{-march=68020}.
16926 Generate output for a 68030. This is the default when the compiler is
16927 configured for 68030-based systems. It is equivalent to
16928 @option{-march=68030}.
16932 Generate output for a 68040. This is the default when the compiler is
16933 configured for 68040-based systems. It is equivalent to
16934 @option{-march=68040}.
16936 This option inhibits the use of 68881/68882 instructions that have to be
16937 emulated by software on the 68040. Use this option if your 68040 does not
16938 have code to emulate those instructions.
16942 Generate output for a 68060. This is the default when the compiler is
16943 configured for 68060-based systems. It is equivalent to
16944 @option{-march=68060}.
16946 This option inhibits the use of 68020 and 68881/68882 instructions that
16947 have to be emulated by software on the 68060. Use this option if your 68060
16948 does not have code to emulate those instructions.
16952 Generate output for a CPU32. This is the default
16953 when the compiler is configured for CPU32-based systems.
16954 It is equivalent to @option{-march=cpu32}.
16956 Use this option for microcontrollers with a
16957 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
16958 68336, 68340, 68341, 68349 and 68360.
16962 Generate output for a 520X ColdFire CPU@. This is the default
16963 when the compiler is configured for 520X-based systems.
16964 It is equivalent to @option{-mcpu=5206}, and is now deprecated
16965 in favor of that option.
16967 Use this option for microcontroller with a 5200 core, including
16968 the MCF5202, MCF5203, MCF5204 and MCF5206.
16972 Generate output for a 5206e ColdFire CPU@. The option is now
16973 deprecated in favor of the equivalent @option{-mcpu=5206e}.
16977 Generate output for a member of the ColdFire 528X family.
16978 The option is now deprecated in favor of the equivalent
16979 @option{-mcpu=528x}.
16983 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
16984 in favor of the equivalent @option{-mcpu=5307}.
16988 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
16989 in favor of the equivalent @option{-mcpu=5407}.
16993 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
16994 This includes use of hardware floating-point instructions.
16995 The option is equivalent to @option{-mcpu=547x}, and is now
16996 deprecated in favor of that option.
17000 Generate output for a 68040, without using any of the new instructions.
17001 This results in code that can run relatively efficiently on either a
17002 68020/68881 or a 68030 or a 68040. The generated code does use the
17003 68881 instructions that are emulated on the 68040.
17005 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
17009 Generate output for a 68060, without using any of the new instructions.
17010 This results in code that can run relatively efficiently on either a
17011 68020/68881 or a 68030 or a 68040. The generated code does use the
17012 68881 instructions that are emulated on the 68060.
17014 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
17018 @opindex mhard-float
17020 Generate floating-point instructions. This is the default for 68020
17021 and above, and for ColdFire devices that have an FPU@. It defines the
17022 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
17023 on ColdFire targets.
17026 @opindex msoft-float
17027 Do not generate floating-point instructions; use library calls instead.
17028 This is the default for 68000, 68010, and 68832 targets. It is also
17029 the default for ColdFire devices that have no FPU.
17035 Generate (do not generate) ColdFire hardware divide and remainder
17036 instructions. If @option{-march} is used without @option{-mcpu},
17037 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
17038 architectures. Otherwise, the default is taken from the target CPU
17039 (either the default CPU, or the one specified by @option{-mcpu}). For
17040 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
17041 @option{-mcpu=5206e}.
17043 GCC defines the macro @samp{__mcfhwdiv__} when this option is enabled.
17047 Consider type @code{int} to be 16 bits wide, like @code{short int}.
17048 Additionally, parameters passed on the stack are also aligned to a
17049 16-bit boundary even on targets whose API mandates promotion to 32-bit.
17053 Do not consider type @code{int} to be 16 bits wide. This is the default.
17056 @itemx -mno-bitfield
17057 @opindex mnobitfield
17058 @opindex mno-bitfield
17059 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
17060 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
17064 Do use the bit-field instructions. The @option{-m68020} option implies
17065 @option{-mbitfield}. This is the default if you use a configuration
17066 designed for a 68020.
17070 Use a different function-calling convention, in which functions
17071 that take a fixed number of arguments return with the @code{rtd}
17072 instruction, which pops their arguments while returning. This
17073 saves one instruction in the caller since there is no need to pop
17074 the arguments there.
17076 This calling convention is incompatible with the one normally
17077 used on Unix, so you cannot use it if you need to call libraries
17078 compiled with the Unix compiler.
17080 Also, you must provide function prototypes for all functions that
17081 take variable numbers of arguments (including @code{printf});
17082 otherwise incorrect code is generated for calls to those
17085 In addition, seriously incorrect code results if you call a
17086 function with too many arguments. (Normally, extra arguments are
17087 harmlessly ignored.)
17089 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
17090 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
17094 Do not use the calling conventions selected by @option{-mrtd}.
17095 This is the default.
17098 @itemx -mno-align-int
17099 @opindex malign-int
17100 @opindex mno-align-int
17101 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
17102 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
17103 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
17104 Aligning variables on 32-bit boundaries produces code that runs somewhat
17105 faster on processors with 32-bit busses at the expense of more memory.
17107 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
17108 aligns structures containing the above types differently than
17109 most published application binary interface specifications for the m68k.
17113 Use the pc-relative addressing mode of the 68000 directly, instead of
17114 using a global offset table. At present, this option implies @option{-fpic},
17115 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
17116 not presently supported with @option{-mpcrel}, though this could be supported for
17117 68020 and higher processors.
17119 @item -mno-strict-align
17120 @itemx -mstrict-align
17121 @opindex mno-strict-align
17122 @opindex mstrict-align
17123 Do not (do) assume that unaligned memory references are handled by
17127 Generate code that allows the data segment to be located in a different
17128 area of memory from the text segment. This allows for execute-in-place in
17129 an environment without virtual memory management. This option implies
17132 @item -mno-sep-data
17133 Generate code that assumes that the data segment follows the text segment.
17134 This is the default.
17136 @item -mid-shared-library
17137 Generate code that supports shared libraries via the library ID method.
17138 This allows for execute-in-place and shared libraries in an environment
17139 without virtual memory management. This option implies @option{-fPIC}.
17141 @item -mno-id-shared-library
17142 Generate code that doesn't assume ID-based shared libraries are being used.
17143 This is the default.
17145 @item -mshared-library-id=n
17146 Specifies the identification number of the ID-based shared library being
17147 compiled. Specifying a value of 0 generates more compact code; specifying
17148 other values forces the allocation of that number to the current
17149 library, but is no more space- or time-efficient than omitting this option.
17155 When generating position-independent code for ColdFire, generate code
17156 that works if the GOT has more than 8192 entries. This code is
17157 larger and slower than code generated without this option. On M680x0
17158 processors, this option is not needed; @option{-fPIC} suffices.
17160 GCC normally uses a single instruction to load values from the GOT@.
17161 While this is relatively efficient, it only works if the GOT
17162 is smaller than about 64k. Anything larger causes the linker
17163 to report an error such as:
17165 @cindex relocation truncated to fit (ColdFire)
17167 relocation truncated to fit: R_68K_GOT16O foobar
17170 If this happens, you should recompile your code with @option{-mxgot}.
17171 It should then work with very large GOTs. However, code generated with
17172 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
17173 the value of a global symbol.
17175 Note that some linkers, including newer versions of the GNU linker,
17176 can create multiple GOTs and sort GOT entries. If you have such a linker,
17177 you should only need to use @option{-mxgot} when compiling a single
17178 object file that accesses more than 8192 GOT entries. Very few do.
17180 These options have no effect unless GCC is generating
17181 position-independent code.
17185 @node MCore Options
17186 @subsection MCore Options
17187 @cindex MCore options
17189 These are the @samp{-m} options defined for the Motorola M*Core
17195 @itemx -mno-hardlit
17197 @opindex mno-hardlit
17198 Inline constants into the code stream if it can be done in two
17199 instructions or less.
17205 Use the divide instruction. (Enabled by default).
17207 @item -mrelax-immediate
17208 @itemx -mno-relax-immediate
17209 @opindex mrelax-immediate
17210 @opindex mno-relax-immediate
17211 Allow arbitrary-sized immediates in bit operations.
17213 @item -mwide-bitfields
17214 @itemx -mno-wide-bitfields
17215 @opindex mwide-bitfields
17216 @opindex mno-wide-bitfields
17217 Always treat bit-fields as @code{int}-sized.
17219 @item -m4byte-functions
17220 @itemx -mno-4byte-functions
17221 @opindex m4byte-functions
17222 @opindex mno-4byte-functions
17223 Force all functions to be aligned to a 4-byte boundary.
17225 @item -mcallgraph-data
17226 @itemx -mno-callgraph-data
17227 @opindex mcallgraph-data
17228 @opindex mno-callgraph-data
17229 Emit callgraph information.
17232 @itemx -mno-slow-bytes
17233 @opindex mslow-bytes
17234 @opindex mno-slow-bytes
17235 Prefer word access when reading byte quantities.
17237 @item -mlittle-endian
17238 @itemx -mbig-endian
17239 @opindex mlittle-endian
17240 @opindex mbig-endian
17241 Generate code for a little-endian target.
17247 Generate code for the 210 processor.
17251 Assume that runtime support has been provided and so omit the
17252 simulator library (@file{libsim.a)} from the linker command line.
17254 @item -mstack-increment=@var{size}
17255 @opindex mstack-increment
17256 Set the maximum amount for a single stack increment operation. Large
17257 values can increase the speed of programs that contain functions
17258 that need a large amount of stack space, but they can also trigger a
17259 segmentation fault if the stack is extended too much. The default
17265 @subsection MeP Options
17266 @cindex MeP options
17272 Enables the @code{abs} instruction, which is the absolute difference
17273 between two registers.
17277 Enables all the optional instructions---average, multiply, divide, bit
17278 operations, leading zero, absolute difference, min/max, clip, and
17284 Enables the @code{ave} instruction, which computes the average of two
17287 @item -mbased=@var{n}
17289 Variables of size @var{n} bytes or smaller are placed in the
17290 @code{.based} section by default. Based variables use the @code{$tp}
17291 register as a base register, and there is a 128-byte limit to the
17292 @code{.based} section.
17296 Enables the bit operation instructions---bit test (@code{btstm}), set
17297 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
17298 test-and-set (@code{tas}).
17300 @item -mc=@var{name}
17302 Selects which section constant data is placed in. @var{name} may
17303 be @code{tiny}, @code{near}, or @code{far}.
17307 Enables the @code{clip} instruction. Note that @code{-mclip} is not
17308 useful unless you also provide @code{-mminmax}.
17310 @item -mconfig=@var{name}
17312 Selects one of the built-in core configurations. Each MeP chip has
17313 one or more modules in it; each module has a core CPU and a variety of
17314 coprocessors, optional instructions, and peripherals. The
17315 @code{MeP-Integrator} tool, not part of GCC, provides these
17316 configurations through this option; using this option is the same as
17317 using all the corresponding command-line options. The default
17318 configuration is @code{default}.
17322 Enables the coprocessor instructions. By default, this is a 32-bit
17323 coprocessor. Note that the coprocessor is normally enabled via the
17324 @code{-mconfig=} option.
17328 Enables the 32-bit coprocessor's instructions.
17332 Enables the 64-bit coprocessor's instructions.
17336 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
17340 Causes constant variables to be placed in the @code{.near} section.
17344 Enables the @code{div} and @code{divu} instructions.
17348 Generate big-endian code.
17352 Generate little-endian code.
17354 @item -mio-volatile
17355 @opindex mio-volatile
17356 Tells the compiler that any variable marked with the @code{io}
17357 attribute is to be considered volatile.
17361 Causes variables to be assigned to the @code{.far} section by default.
17365 Enables the @code{leadz} (leading zero) instruction.
17369 Causes variables to be assigned to the @code{.near} section by default.
17373 Enables the @code{min} and @code{max} instructions.
17377 Enables the multiplication and multiply-accumulate instructions.
17381 Disables all the optional instructions enabled by @code{-mall-opts}.
17385 Enables the @code{repeat} and @code{erepeat} instructions, used for
17386 low-overhead looping.
17390 Causes all variables to default to the @code{.tiny} section. Note
17391 that there is a 65536-byte limit to this section. Accesses to these
17392 variables use the @code{%gp} base register.
17396 Enables the saturation instructions. Note that the compiler does not
17397 currently generate these itself, but this option is included for
17398 compatibility with other tools, like @code{as}.
17402 Link the SDRAM-based runtime instead of the default ROM-based runtime.
17406 Link the simulator run-time libraries.
17410 Link the simulator runtime libraries, excluding built-in support
17411 for reset and exception vectors and tables.
17415 Causes all functions to default to the @code{.far} section. Without
17416 this option, functions default to the @code{.near} section.
17418 @item -mtiny=@var{n}
17420 Variables that are @var{n} bytes or smaller are allocated to the
17421 @code{.tiny} section. These variables use the @code{$gp} base
17422 register. The default for this option is 4, but note that there's a
17423 65536-byte limit to the @code{.tiny} section.
17427 @node MicroBlaze Options
17428 @subsection MicroBlaze Options
17429 @cindex MicroBlaze Options
17434 @opindex msoft-float
17435 Use software emulation for floating point (default).
17438 @opindex mhard-float
17439 Use hardware floating-point instructions.
17443 Do not optimize block moves, use @code{memcpy}.
17445 @item -mno-clearbss
17446 @opindex mno-clearbss
17447 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
17449 @item -mcpu=@var{cpu-type}
17451 Use features of, and schedule code for, the given CPU.
17452 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
17453 where @var{X} is a major version, @var{YY} is the minor version, and
17454 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
17455 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
17457 @item -mxl-soft-mul
17458 @opindex mxl-soft-mul
17459 Use software multiply emulation (default).
17461 @item -mxl-soft-div
17462 @opindex mxl-soft-div
17463 Use software emulation for divides (default).
17465 @item -mxl-barrel-shift
17466 @opindex mxl-barrel-shift
17467 Use the hardware barrel shifter.
17469 @item -mxl-pattern-compare
17470 @opindex mxl-pattern-compare
17471 Use pattern compare instructions.
17473 @item -msmall-divides
17474 @opindex msmall-divides
17475 Use table lookup optimization for small signed integer divisions.
17477 @item -mxl-stack-check
17478 @opindex mxl-stack-check
17479 This option is deprecated. Use @option{-fstack-check} instead.
17482 @opindex mxl-gp-opt
17483 Use GP-relative @code{.sdata}/@code{.sbss} sections.
17485 @item -mxl-multiply-high
17486 @opindex mxl-multiply-high
17487 Use multiply high instructions for high part of 32x32 multiply.
17489 @item -mxl-float-convert
17490 @opindex mxl-float-convert
17491 Use hardware floating-point conversion instructions.
17493 @item -mxl-float-sqrt
17494 @opindex mxl-float-sqrt
17495 Use hardware floating-point square root instruction.
17498 @opindex mbig-endian
17499 Generate code for a big-endian target.
17501 @item -mlittle-endian
17502 @opindex mlittle-endian
17503 Generate code for a little-endian target.
17506 @opindex mxl-reorder
17507 Use reorder instructions (swap and byte reversed load/store).
17509 @item -mxl-mode-@var{app-model}
17510 Select application model @var{app-model}. Valid models are
17513 normal executable (default), uses startup code @file{crt0.o}.
17516 for use with Xilinx Microprocessor Debugger (XMD) based
17517 software intrusive debug agent called xmdstub. This uses startup file
17518 @file{crt1.o} and sets the start address of the program to 0x800.
17521 for applications that are loaded using a bootloader.
17522 This model uses startup file @file{crt2.o} which does not contain a processor
17523 reset vector handler. This is suitable for transferring control on a
17524 processor reset to the bootloader rather than the application.
17527 for applications that do not require any of the
17528 MicroBlaze vectors. This option may be useful for applications running
17529 within a monitoring application. This model uses @file{crt3.o} as a startup file.
17532 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
17533 @option{-mxl-mode-@var{app-model}}.
17538 @subsection MIPS Options
17539 @cindex MIPS options
17545 Generate big-endian code.
17549 Generate little-endian code. This is the default for @samp{mips*el-*-*}
17552 @item -march=@var{arch}
17554 Generate code that runs on @var{arch}, which can be the name of a
17555 generic MIPS ISA, or the name of a particular processor.
17557 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
17558 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
17559 @samp{mips64}, @samp{mips64r2}, @samp{mips64r3} and @samp{mips64r5}.
17560 The processor names are:
17561 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
17562 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
17563 @samp{5kc}, @samp{5kf},
17565 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
17566 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
17567 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
17568 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
17569 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
17570 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
17572 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
17573 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
17576 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
17577 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
17578 @samp{rm7000}, @samp{rm9000},
17579 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
17582 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
17583 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
17584 @samp{xlr} and @samp{xlp}.
17585 The special value @samp{from-abi} selects the
17586 most compatible architecture for the selected ABI (that is,
17587 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
17589 The native Linux/GNU toolchain also supports the value @samp{native},
17590 which selects the best architecture option for the host processor.
17591 @option{-march=native} has no effect if GCC does not recognize
17594 In processor names, a final @samp{000} can be abbreviated as @samp{k}
17595 (for example, @option{-march=r2k}). Prefixes are optional, and
17596 @samp{vr} may be written @samp{r}.
17598 Names of the form @samp{@var{n}f2_1} refer to processors with
17599 FPUs clocked at half the rate of the core, names of the form
17600 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
17601 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
17602 processors with FPUs clocked a ratio of 3:2 with respect to the core.
17603 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
17604 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
17605 accepted as synonyms for @samp{@var{n}f1_1}.
17607 GCC defines two macros based on the value of this option. The first
17608 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
17609 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
17610 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
17611 For example, @option{-march=r2000} sets @samp{_MIPS_ARCH}
17612 to @samp{"r2000"} and defines the macro @samp{_MIPS_ARCH_R2000}.
17614 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
17615 above. In other words, it has the full prefix and does not
17616 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
17617 the macro names the resolved architecture (either @samp{"mips1"} or
17618 @samp{"mips3"}). It names the default architecture when no
17619 @option{-march} option is given.
17621 @item -mtune=@var{arch}
17623 Optimize for @var{arch}. Among other things, this option controls
17624 the way instructions are scheduled, and the perceived cost of arithmetic
17625 operations. The list of @var{arch} values is the same as for
17628 When this option is not used, GCC optimizes for the processor
17629 specified by @option{-march}. By using @option{-march} and
17630 @option{-mtune} together, it is possible to generate code that
17631 runs on a family of processors, but optimize the code for one
17632 particular member of that family.
17634 @option{-mtune} defines the macros @samp{_MIPS_TUNE} and
17635 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
17636 @option{-march} ones described above.
17640 Equivalent to @option{-march=mips1}.
17644 Equivalent to @option{-march=mips2}.
17648 Equivalent to @option{-march=mips3}.
17652 Equivalent to @option{-march=mips4}.
17656 Equivalent to @option{-march=mips32}.
17660 Equivalent to @option{-march=mips32r3}.
17664 Equivalent to @option{-march=mips32r5}.
17668 Equivalent to @option{-march=mips64}.
17672 Equivalent to @option{-march=mips64r2}.
17676 Equivalent to @option{-march=mips64r3}.
17680 Equivalent to @option{-march=mips64r5}.
17685 @opindex mno-mips16
17686 Generate (do not generate) MIPS16 code. If GCC is targeting a
17687 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
17689 MIPS16 code generation can also be controlled on a per-function basis
17690 by means of @code{mips16} and @code{nomips16} attributes.
17691 @xref{Function Attributes}, for more information.
17693 @item -mflip-mips16
17694 @opindex mflip-mips16
17695 Generate MIPS16 code on alternating functions. This option is provided
17696 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
17697 not intended for ordinary use in compiling user code.
17699 @item -minterlink-compressed
17700 @item -mno-interlink-compressed
17701 @opindex minterlink-compressed
17702 @opindex mno-interlink-compressed
17703 Require (do not require) that code using the standard (uncompressed) MIPS ISA
17704 be link-compatible with MIPS16 and microMIPS code, and vice versa.
17706 For example, code using the standard ISA encoding cannot jump directly
17707 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
17708 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
17709 knows that the target of the jump is not compressed.
17711 @item -minterlink-mips16
17712 @itemx -mno-interlink-mips16
17713 @opindex minterlink-mips16
17714 @opindex mno-interlink-mips16
17715 Aliases of @option{-minterlink-compressed} and
17716 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
17717 and are retained for backwards compatibility.
17729 Generate code for the given ABI@.
17731 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
17732 generates 64-bit code when you select a 64-bit architecture, but you
17733 can use @option{-mgp32} to get 32-bit code instead.
17735 For information about the O64 ABI, see
17736 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
17738 GCC supports a variant of the o32 ABI in which floating-point registers
17739 are 64 rather than 32 bits wide. You can select this combination with
17740 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
17741 and @code{mfhc1} instructions and is therefore only supported for
17742 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
17744 The register assignments for arguments and return values remain the
17745 same, but each scalar value is passed in a single 64-bit register
17746 rather than a pair of 32-bit registers. For example, scalar
17747 floating-point values are returned in @samp{$f0} only, not a
17748 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
17749 remains the same in that the even-numbered double-precision registers
17752 Two additional variants of the o32 ABI are supported to enable
17753 a transition from 32-bit to 64-bit registers. These are FPXX
17754 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
17755 The FPXX extension mandates that all code must execute correctly
17756 when run using 32-bit or 64-bit registers. The code can be interlinked
17757 with either FP32 or FP64, but not both.
17758 The FP64A extension is similar to the FP64 extension but forbids the
17759 use of odd-numbered single-precision registers. This can be used
17760 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
17761 processors and allows both FP32 and FP64A code to interlink and
17762 run in the same process without changing FPU modes.
17765 @itemx -mno-abicalls
17767 @opindex mno-abicalls
17768 Generate (do not generate) code that is suitable for SVR4-style
17769 dynamic objects. @option{-mabicalls} is the default for SVR4-based
17774 Generate (do not generate) code that is fully position-independent,
17775 and that can therefore be linked into shared libraries. This option
17776 only affects @option{-mabicalls}.
17778 All @option{-mabicalls} code has traditionally been position-independent,
17779 regardless of options like @option{-fPIC} and @option{-fpic}. However,
17780 as an extension, the GNU toolchain allows executables to use absolute
17781 accesses for locally-binding symbols. It can also use shorter GP
17782 initialization sequences and generate direct calls to locally-defined
17783 functions. This mode is selected by @option{-mno-shared}.
17785 @option{-mno-shared} depends on binutils 2.16 or higher and generates
17786 objects that can only be linked by the GNU linker. However, the option
17787 does not affect the ABI of the final executable; it only affects the ABI
17788 of relocatable objects. Using @option{-mno-shared} generally makes
17789 executables both smaller and quicker.
17791 @option{-mshared} is the default.
17797 Assume (do not assume) that the static and dynamic linkers
17798 support PLTs and copy relocations. This option only affects
17799 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
17800 has no effect without @option{-msym32}.
17802 You can make @option{-mplt} the default by configuring
17803 GCC with @option{--with-mips-plt}. The default is
17804 @option{-mno-plt} otherwise.
17810 Lift (do not lift) the usual restrictions on the size of the global
17813 GCC normally uses a single instruction to load values from the GOT@.
17814 While this is relatively efficient, it only works if the GOT
17815 is smaller than about 64k. Anything larger causes the linker
17816 to report an error such as:
17818 @cindex relocation truncated to fit (MIPS)
17820 relocation truncated to fit: R_MIPS_GOT16 foobar
17823 If this happens, you should recompile your code with @option{-mxgot}.
17824 This works with very large GOTs, although the code is also
17825 less efficient, since it takes three instructions to fetch the
17826 value of a global symbol.
17828 Note that some linkers can create multiple GOTs. If you have such a
17829 linker, you should only need to use @option{-mxgot} when a single object
17830 file accesses more than 64k's worth of GOT entries. Very few do.
17832 These options have no effect unless GCC is generating position
17837 Assume that general-purpose registers are 32 bits wide.
17841 Assume that general-purpose registers are 64 bits wide.
17845 Assume that floating-point registers are 32 bits wide.
17849 Assume that floating-point registers are 64 bits wide.
17853 Do not assume the width of floating-point registers.
17856 @opindex mhard-float
17857 Use floating-point coprocessor instructions.
17860 @opindex msoft-float
17861 Do not use floating-point coprocessor instructions. Implement
17862 floating-point calculations using library calls instead.
17866 Equivalent to @option{-msoft-float}, but additionally asserts that the
17867 program being compiled does not perform any floating-point operations.
17868 This option is presently supported only by some bare-metal MIPS
17869 configurations, where it may select a special set of libraries
17870 that lack all floating-point support (including, for example, the
17871 floating-point @code{printf} formats).
17872 If code compiled with @code{-mno-float} accidentally contains
17873 floating-point operations, it is likely to suffer a link-time
17874 or run-time failure.
17876 @item -msingle-float
17877 @opindex msingle-float
17878 Assume that the floating-point coprocessor only supports single-precision
17881 @item -mdouble-float
17882 @opindex mdouble-float
17883 Assume that the floating-point coprocessor supports double-precision
17884 operations. This is the default.
17887 @itemx -mno-odd-spreg
17888 @opindex modd-spreg
17889 @opindex mno-odd-spreg
17890 Enable the use of odd-numbered single-precision floating-point registers
17891 for the o32 ABI. This is the default for processors that are known to
17892 support these registers. When using the o32 FPXX ABI, @code{-mno-odd-spreg}
17896 @itemx -mabs=legacy
17898 @opindex mabs=legacy
17899 These options control the treatment of the special not-a-number (NaN)
17900 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
17901 @code{neg.@i{fmt}} machine instructions.
17903 By default or when the @option{-mabs=legacy} is used the legacy
17904 treatment is selected. In this case these instructions are considered
17905 arithmetic and avoided where correct operation is required and the
17906 input operand might be a NaN. A longer sequence of instructions that
17907 manipulate the sign bit of floating-point datum manually is used
17908 instead unless the @option{-ffinite-math-only} option has also been
17911 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
17912 this case these instructions are considered non-arithmetic and therefore
17913 operating correctly in all cases, including in particular where the
17914 input operand is a NaN. These instructions are therefore always used
17915 for the respective operations.
17918 @itemx -mnan=legacy
17920 @opindex mnan=legacy
17921 These options control the encoding of the special not-a-number (NaN)
17922 IEEE 754 floating-point data.
17924 The @option{-mnan=legacy} option selects the legacy encoding. In this
17925 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
17926 significand field being 0, whereas signalling NaNs (sNaNs) are denoted
17927 by the first bit of their trailing significand field being 1.
17929 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
17930 this case qNaNs are denoted by the first bit of their trailing
17931 significand field being 1, whereas sNaNs are denoted by the first bit of
17932 their trailing significand field being 0.
17934 The default is @option{-mnan=legacy} unless GCC has been configured with
17935 @option{--with-nan=2008}.
17941 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
17942 implement atomic memory built-in functions. When neither option is
17943 specified, GCC uses the instructions if the target architecture
17946 @option{-mllsc} is useful if the runtime environment can emulate the
17947 instructions and @option{-mno-llsc} can be useful when compiling for
17948 nonstandard ISAs. You can make either option the default by
17949 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
17950 respectively. @option{--with-llsc} is the default for some
17951 configurations; see the installation documentation for details.
17957 Use (do not use) revision 1 of the MIPS DSP ASE@.
17958 @xref{MIPS DSP Built-in Functions}. This option defines the
17959 preprocessor macro @samp{__mips_dsp}. It also defines
17960 @samp{__mips_dsp_rev} to 1.
17966 Use (do not use) revision 2 of the MIPS DSP ASE@.
17967 @xref{MIPS DSP Built-in Functions}. This option defines the
17968 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
17969 It also defines @samp{__mips_dsp_rev} to 2.
17972 @itemx -mno-smartmips
17973 @opindex msmartmips
17974 @opindex mno-smartmips
17975 Use (do not use) the MIPS SmartMIPS ASE.
17977 @item -mpaired-single
17978 @itemx -mno-paired-single
17979 @opindex mpaired-single
17980 @opindex mno-paired-single
17981 Use (do not use) paired-single floating-point instructions.
17982 @xref{MIPS Paired-Single Support}. This option requires
17983 hardware floating-point support to be enabled.
17989 Use (do not use) MIPS Digital Media Extension instructions.
17990 This option can only be used when generating 64-bit code and requires
17991 hardware floating-point support to be enabled.
17996 @opindex mno-mips3d
17997 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
17998 The option @option{-mips3d} implies @option{-mpaired-single}.
18001 @itemx -mno-micromips
18002 @opindex mmicromips
18003 @opindex mno-mmicromips
18004 Generate (do not generate) microMIPS code.
18006 MicroMIPS code generation can also be controlled on a per-function basis
18007 by means of @code{micromips} and @code{nomicromips} attributes.
18008 @xref{Function Attributes}, for more information.
18014 Use (do not use) MT Multithreading instructions.
18020 Use (do not use) the MIPS MCU ASE instructions.
18026 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
18032 Use (do not use) the MIPS Virtualization Application Specific instructions.
18038 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
18042 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
18043 an explanation of the default and the way that the pointer size is
18048 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
18050 The default size of @code{int}s, @code{long}s and pointers depends on
18051 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
18052 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
18053 32-bit @code{long}s. Pointers are the same size as @code{long}s,
18054 or the same size as integer registers, whichever is smaller.
18060 Assume (do not assume) that all symbols have 32-bit values, regardless
18061 of the selected ABI@. This option is useful in combination with
18062 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
18063 to generate shorter and faster references to symbolic addresses.
18067 Put definitions of externally-visible data in a small data section
18068 if that data is no bigger than @var{num} bytes. GCC can then generate
18069 more efficient accesses to the data; see @option{-mgpopt} for details.
18071 The default @option{-G} option depends on the configuration.
18073 @item -mlocal-sdata
18074 @itemx -mno-local-sdata
18075 @opindex mlocal-sdata
18076 @opindex mno-local-sdata
18077 Extend (do not extend) the @option{-G} behavior to local data too,
18078 such as to static variables in C@. @option{-mlocal-sdata} is the
18079 default for all configurations.
18081 If the linker complains that an application is using too much small data,
18082 you might want to try rebuilding the less performance-critical parts with
18083 @option{-mno-local-sdata}. You might also want to build large
18084 libraries with @option{-mno-local-sdata}, so that the libraries leave
18085 more room for the main program.
18087 @item -mextern-sdata
18088 @itemx -mno-extern-sdata
18089 @opindex mextern-sdata
18090 @opindex mno-extern-sdata
18091 Assume (do not assume) that externally-defined data is in
18092 a small data section if the size of that data is within the @option{-G} limit.
18093 @option{-mextern-sdata} is the default for all configurations.
18095 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
18096 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
18097 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
18098 is placed in a small data section. If @var{Var} is defined by another
18099 module, you must either compile that module with a high-enough
18100 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
18101 definition. If @var{Var} is common, you must link the application
18102 with a high-enough @option{-G} setting.
18104 The easiest way of satisfying these restrictions is to compile
18105 and link every module with the same @option{-G} option. However,
18106 you may wish to build a library that supports several different
18107 small data limits. You can do this by compiling the library with
18108 the highest supported @option{-G} setting and additionally using
18109 @option{-mno-extern-sdata} to stop the library from making assumptions
18110 about externally-defined data.
18116 Use (do not use) GP-relative accesses for symbols that are known to be
18117 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
18118 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
18121 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
18122 might not hold the value of @code{_gp}. For example, if the code is
18123 part of a library that might be used in a boot monitor, programs that
18124 call boot monitor routines pass an unknown value in @code{$gp}.
18125 (In such situations, the boot monitor itself is usually compiled
18126 with @option{-G0}.)
18128 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
18129 @option{-mno-extern-sdata}.
18131 @item -membedded-data
18132 @itemx -mno-embedded-data
18133 @opindex membedded-data
18134 @opindex mno-embedded-data
18135 Allocate variables to the read-only data section first if possible, then
18136 next in the small data section if possible, otherwise in data. This gives
18137 slightly slower code than the default, but reduces the amount of RAM required
18138 when executing, and thus may be preferred for some embedded systems.
18140 @item -muninit-const-in-rodata
18141 @itemx -mno-uninit-const-in-rodata
18142 @opindex muninit-const-in-rodata
18143 @opindex mno-uninit-const-in-rodata
18144 Put uninitialized @code{const} variables in the read-only data section.
18145 This option is only meaningful in conjunction with @option{-membedded-data}.
18147 @item -mcode-readable=@var{setting}
18148 @opindex mcode-readable
18149 Specify whether GCC may generate code that reads from executable sections.
18150 There are three possible settings:
18153 @item -mcode-readable=yes
18154 Instructions may freely access executable sections. This is the
18157 @item -mcode-readable=pcrel
18158 MIPS16 PC-relative load instructions can access executable sections,
18159 but other instructions must not do so. This option is useful on 4KSc
18160 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
18161 It is also useful on processors that can be configured to have a dual
18162 instruction/data SRAM interface and that, like the M4K, automatically
18163 redirect PC-relative loads to the instruction RAM.
18165 @item -mcode-readable=no
18166 Instructions must not access executable sections. This option can be
18167 useful on targets that are configured to have a dual instruction/data
18168 SRAM interface but that (unlike the M4K) do not automatically redirect
18169 PC-relative loads to the instruction RAM.
18172 @item -msplit-addresses
18173 @itemx -mno-split-addresses
18174 @opindex msplit-addresses
18175 @opindex mno-split-addresses
18176 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
18177 relocation operators. This option has been superseded by
18178 @option{-mexplicit-relocs} but is retained for backwards compatibility.
18180 @item -mexplicit-relocs
18181 @itemx -mno-explicit-relocs
18182 @opindex mexplicit-relocs
18183 @opindex mno-explicit-relocs
18184 Use (do not use) assembler relocation operators when dealing with symbolic
18185 addresses. The alternative, selected by @option{-mno-explicit-relocs},
18186 is to use assembler macros instead.
18188 @option{-mexplicit-relocs} is the default if GCC was configured
18189 to use an assembler that supports relocation operators.
18191 @item -mcheck-zero-division
18192 @itemx -mno-check-zero-division
18193 @opindex mcheck-zero-division
18194 @opindex mno-check-zero-division
18195 Trap (do not trap) on integer division by zero.
18197 The default is @option{-mcheck-zero-division}.
18199 @item -mdivide-traps
18200 @itemx -mdivide-breaks
18201 @opindex mdivide-traps
18202 @opindex mdivide-breaks
18203 MIPS systems check for division by zero by generating either a
18204 conditional trap or a break instruction. Using traps results in
18205 smaller code, but is only supported on MIPS II and later. Also, some
18206 versions of the Linux kernel have a bug that prevents trap from
18207 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
18208 allow conditional traps on architectures that support them and
18209 @option{-mdivide-breaks} to force the use of breaks.
18211 The default is usually @option{-mdivide-traps}, but this can be
18212 overridden at configure time using @option{--with-divide=breaks}.
18213 Divide-by-zero checks can be completely disabled using
18214 @option{-mno-check-zero-division}.
18219 @opindex mno-memcpy
18220 Force (do not force) the use of @code{memcpy()} for non-trivial block
18221 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
18222 most constant-sized copies.
18225 @itemx -mno-long-calls
18226 @opindex mlong-calls
18227 @opindex mno-long-calls
18228 Disable (do not disable) use of the @code{jal} instruction. Calling
18229 functions using @code{jal} is more efficient but requires the caller
18230 and callee to be in the same 256 megabyte segment.
18232 This option has no effect on abicalls code. The default is
18233 @option{-mno-long-calls}.
18239 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
18240 instructions, as provided by the R4650 ISA@.
18246 Enable (disable) use of the @code{madd} and @code{msub} integer
18247 instructions. The default is @option{-mimadd} on architectures
18248 that support @code{madd} and @code{msub} except for the 74k
18249 architecture where it was found to generate slower code.
18252 @itemx -mno-fused-madd
18253 @opindex mfused-madd
18254 @opindex mno-fused-madd
18255 Enable (disable) use of the floating-point multiply-accumulate
18256 instructions, when they are available. The default is
18257 @option{-mfused-madd}.
18259 On the R8000 CPU when multiply-accumulate instructions are used,
18260 the intermediate product is calculated to infinite precision
18261 and is not subject to the FCSR Flush to Zero bit. This may be
18262 undesirable in some circumstances. On other processors the result
18263 is numerically identical to the equivalent computation using
18264 separate multiply, add, subtract and negate instructions.
18268 Tell the MIPS assembler to not run its preprocessor over user
18269 assembler files (with a @samp{.s} suffix) when assembling them.
18274 @opindex mno-fix-24k
18275 Work around the 24K E48 (lost data on stores during refill) errata.
18276 The workarounds are implemented by the assembler rather than by GCC@.
18279 @itemx -mno-fix-r4000
18280 @opindex mfix-r4000
18281 @opindex mno-fix-r4000
18282 Work around certain R4000 CPU errata:
18285 A double-word or a variable shift may give an incorrect result if executed
18286 immediately after starting an integer division.
18288 A double-word or a variable shift may give an incorrect result if executed
18289 while an integer multiplication is in progress.
18291 An integer division may give an incorrect result if started in a delay slot
18292 of a taken branch or a jump.
18296 @itemx -mno-fix-r4400
18297 @opindex mfix-r4400
18298 @opindex mno-fix-r4400
18299 Work around certain R4400 CPU errata:
18302 A double-word or a variable shift may give an incorrect result if executed
18303 immediately after starting an integer division.
18307 @itemx -mno-fix-r10000
18308 @opindex mfix-r10000
18309 @opindex mno-fix-r10000
18310 Work around certain R10000 errata:
18313 @code{ll}/@code{sc} sequences may not behave atomically on revisions
18314 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
18317 This option can only be used if the target architecture supports
18318 branch-likely instructions. @option{-mfix-r10000} is the default when
18319 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
18323 @itemx -mno-fix-rm7000
18324 @opindex mfix-rm7000
18325 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
18326 workarounds are implemented by the assembler rather than by GCC@.
18329 @itemx -mno-fix-vr4120
18330 @opindex mfix-vr4120
18331 Work around certain VR4120 errata:
18334 @code{dmultu} does not always produce the correct result.
18336 @code{div} and @code{ddiv} do not always produce the correct result if one
18337 of the operands is negative.
18339 The workarounds for the division errata rely on special functions in
18340 @file{libgcc.a}. At present, these functions are only provided by
18341 the @code{mips64vr*-elf} configurations.
18343 Other VR4120 errata require a NOP to be inserted between certain pairs of
18344 instructions. These errata are handled by the assembler, not by GCC itself.
18347 @opindex mfix-vr4130
18348 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
18349 workarounds are implemented by the assembler rather than by GCC,
18350 although GCC avoids using @code{mflo} and @code{mfhi} if the
18351 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
18352 instructions are available instead.
18355 @itemx -mno-fix-sb1
18357 Work around certain SB-1 CPU core errata.
18358 (This flag currently works around the SB-1 revision 2
18359 ``F1'' and ``F2'' floating-point errata.)
18361 @item -mr10k-cache-barrier=@var{setting}
18362 @opindex mr10k-cache-barrier
18363 Specify whether GCC should insert cache barriers to avoid the
18364 side-effects of speculation on R10K processors.
18366 In common with many processors, the R10K tries to predict the outcome
18367 of a conditional branch and speculatively executes instructions from
18368 the ``taken'' branch. It later aborts these instructions if the
18369 predicted outcome is wrong. However, on the R10K, even aborted
18370 instructions can have side effects.
18372 This problem only affects kernel stores and, depending on the system,
18373 kernel loads. As an example, a speculatively-executed store may load
18374 the target memory into cache and mark the cache line as dirty, even if
18375 the store itself is later aborted. If a DMA operation writes to the
18376 same area of memory before the ``dirty'' line is flushed, the cached
18377 data overwrites the DMA-ed data. See the R10K processor manual
18378 for a full description, including other potential problems.
18380 One workaround is to insert cache barrier instructions before every memory
18381 access that might be speculatively executed and that might have side
18382 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
18383 controls GCC's implementation of this workaround. It assumes that
18384 aborted accesses to any byte in the following regions does not have
18389 the memory occupied by the current function's stack frame;
18392 the memory occupied by an incoming stack argument;
18395 the memory occupied by an object with a link-time-constant address.
18398 It is the kernel's responsibility to ensure that speculative
18399 accesses to these regions are indeed safe.
18401 If the input program contains a function declaration such as:
18407 then the implementation of @code{foo} must allow @code{j foo} and
18408 @code{jal foo} to be executed speculatively. GCC honors this
18409 restriction for functions it compiles itself. It expects non-GCC
18410 functions (such as hand-written assembly code) to do the same.
18412 The option has three forms:
18415 @item -mr10k-cache-barrier=load-store
18416 Insert a cache barrier before a load or store that might be
18417 speculatively executed and that might have side effects even
18420 @item -mr10k-cache-barrier=store
18421 Insert a cache barrier before a store that might be speculatively
18422 executed and that might have side effects even if aborted.
18424 @item -mr10k-cache-barrier=none
18425 Disable the insertion of cache barriers. This is the default setting.
18428 @item -mflush-func=@var{func}
18429 @itemx -mno-flush-func
18430 @opindex mflush-func
18431 Specifies the function to call to flush the I and D caches, or to not
18432 call any such function. If called, the function must take the same
18433 arguments as the common @code{_flush_func()}, that is, the address of the
18434 memory range for which the cache is being flushed, the size of the
18435 memory range, and the number 3 (to flush both caches). The default
18436 depends on the target GCC was configured for, but commonly is either
18437 @samp{_flush_func} or @samp{__cpu_flush}.
18439 @item mbranch-cost=@var{num}
18440 @opindex mbranch-cost
18441 Set the cost of branches to roughly @var{num} ``simple'' instructions.
18442 This cost is only a heuristic and is not guaranteed to produce
18443 consistent results across releases. A zero cost redundantly selects
18444 the default, which is based on the @option{-mtune} setting.
18446 @item -mbranch-likely
18447 @itemx -mno-branch-likely
18448 @opindex mbranch-likely
18449 @opindex mno-branch-likely
18450 Enable or disable use of Branch Likely instructions, regardless of the
18451 default for the selected architecture. By default, Branch Likely
18452 instructions may be generated if they are supported by the selected
18453 architecture. An exception is for the MIPS32 and MIPS64 architectures
18454 and processors that implement those architectures; for those, Branch
18455 Likely instructions are not be generated by default because the MIPS32
18456 and MIPS64 architectures specifically deprecate their use.
18458 @item -mfp-exceptions
18459 @itemx -mno-fp-exceptions
18460 @opindex mfp-exceptions
18461 Specifies whether FP exceptions are enabled. This affects how
18462 FP instructions are scheduled for some processors.
18463 The default is that FP exceptions are
18466 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
18467 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
18470 @item -mvr4130-align
18471 @itemx -mno-vr4130-align
18472 @opindex mvr4130-align
18473 The VR4130 pipeline is two-way superscalar, but can only issue two
18474 instructions together if the first one is 8-byte aligned. When this
18475 option is enabled, GCC aligns pairs of instructions that it
18476 thinks should execute in parallel.
18478 This option only has an effect when optimizing for the VR4130.
18479 It normally makes code faster, but at the expense of making it bigger.
18480 It is enabled by default at optimization level @option{-O3}.
18485 Enable (disable) generation of @code{synci} instructions on
18486 architectures that support it. The @code{synci} instructions (if
18487 enabled) are generated when @code{__builtin___clear_cache()} is
18490 This option defaults to @code{-mno-synci}, but the default can be
18491 overridden by configuring with @code{--with-synci}.
18493 When compiling code for single processor systems, it is generally safe
18494 to use @code{synci}. However, on many multi-core (SMP) systems, it
18495 does not invalidate the instruction caches on all cores and may lead
18496 to undefined behavior.
18498 @item -mrelax-pic-calls
18499 @itemx -mno-relax-pic-calls
18500 @opindex mrelax-pic-calls
18501 Try to turn PIC calls that are normally dispatched via register
18502 @code{$25} into direct calls. This is only possible if the linker can
18503 resolve the destination at link-time and if the destination is within
18504 range for a direct call.
18506 @option{-mrelax-pic-calls} is the default if GCC was configured to use
18507 an assembler and a linker that support the @code{.reloc} assembly
18508 directive and @code{-mexplicit-relocs} is in effect. With
18509 @code{-mno-explicit-relocs}, this optimization can be performed by the
18510 assembler and the linker alone without help from the compiler.
18512 @item -mmcount-ra-address
18513 @itemx -mno-mcount-ra-address
18514 @opindex mmcount-ra-address
18515 @opindex mno-mcount-ra-address
18516 Emit (do not emit) code that allows @code{_mcount} to modify the
18517 calling function's return address. When enabled, this option extends
18518 the usual @code{_mcount} interface with a new @var{ra-address}
18519 parameter, which has type @code{intptr_t *} and is passed in register
18520 @code{$12}. @code{_mcount} can then modify the return address by
18521 doing both of the following:
18524 Returning the new address in register @code{$31}.
18526 Storing the new address in @code{*@var{ra-address}},
18527 if @var{ra-address} is nonnull.
18530 The default is @option{-mno-mcount-ra-address}.
18535 @subsection MMIX Options
18536 @cindex MMIX Options
18538 These options are defined for the MMIX:
18542 @itemx -mno-libfuncs
18544 @opindex mno-libfuncs
18545 Specify that intrinsic library functions are being compiled, passing all
18546 values in registers, no matter the size.
18549 @itemx -mno-epsilon
18551 @opindex mno-epsilon
18552 Generate floating-point comparison instructions that compare with respect
18553 to the @code{rE} epsilon register.
18555 @item -mabi=mmixware
18557 @opindex mabi=mmixware
18559 Generate code that passes function parameters and return values that (in
18560 the called function) are seen as registers @code{$0} and up, as opposed to
18561 the GNU ABI which uses global registers @code{$231} and up.
18563 @item -mzero-extend
18564 @itemx -mno-zero-extend
18565 @opindex mzero-extend
18566 @opindex mno-zero-extend
18567 When reading data from memory in sizes shorter than 64 bits, use (do not
18568 use) zero-extending load instructions by default, rather than
18569 sign-extending ones.
18572 @itemx -mno-knuthdiv
18574 @opindex mno-knuthdiv
18575 Make the result of a division yielding a remainder have the same sign as
18576 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
18577 remainder follows the sign of the dividend. Both methods are
18578 arithmetically valid, the latter being almost exclusively used.
18580 @item -mtoplevel-symbols
18581 @itemx -mno-toplevel-symbols
18582 @opindex mtoplevel-symbols
18583 @opindex mno-toplevel-symbols
18584 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
18585 code can be used with the @code{PREFIX} assembly directive.
18589 Generate an executable in the ELF format, rather than the default
18590 @samp{mmo} format used by the @command{mmix} simulator.
18592 @item -mbranch-predict
18593 @itemx -mno-branch-predict
18594 @opindex mbranch-predict
18595 @opindex mno-branch-predict
18596 Use (do not use) the probable-branch instructions, when static branch
18597 prediction indicates a probable branch.
18599 @item -mbase-addresses
18600 @itemx -mno-base-addresses
18601 @opindex mbase-addresses
18602 @opindex mno-base-addresses
18603 Generate (do not generate) code that uses @emph{base addresses}. Using a
18604 base address automatically generates a request (handled by the assembler
18605 and the linker) for a constant to be set up in a global register. The
18606 register is used for one or more base address requests within the range 0
18607 to 255 from the value held in the register. The generally leads to short
18608 and fast code, but the number of different data items that can be
18609 addressed is limited. This means that a program that uses lots of static
18610 data may require @option{-mno-base-addresses}.
18612 @item -msingle-exit
18613 @itemx -mno-single-exit
18614 @opindex msingle-exit
18615 @opindex mno-single-exit
18616 Force (do not force) generated code to have a single exit point in each
18620 @node MN10300 Options
18621 @subsection MN10300 Options
18622 @cindex MN10300 options
18624 These @option{-m} options are defined for Matsushita MN10300 architectures:
18629 Generate code to avoid bugs in the multiply instructions for the MN10300
18630 processors. This is the default.
18632 @item -mno-mult-bug
18633 @opindex mno-mult-bug
18634 Do not generate code to avoid bugs in the multiply instructions for the
18635 MN10300 processors.
18639 Generate code using features specific to the AM33 processor.
18643 Do not generate code using features specific to the AM33 processor. This
18648 Generate code using features specific to the AM33/2.0 processor.
18652 Generate code using features specific to the AM34 processor.
18654 @item -mtune=@var{cpu-type}
18656 Use the timing characteristics of the indicated CPU type when
18657 scheduling instructions. This does not change the targeted processor
18658 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
18659 @samp{am33-2} or @samp{am34}.
18661 @item -mreturn-pointer-on-d0
18662 @opindex mreturn-pointer-on-d0
18663 When generating a function that returns a pointer, return the pointer
18664 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
18665 only in @code{a0}, and attempts to call such functions without a prototype
18666 result in errors. Note that this option is on by default; use
18667 @option{-mno-return-pointer-on-d0} to disable it.
18671 Do not link in the C run-time initialization object file.
18675 Indicate to the linker that it should perform a relaxation optimization pass
18676 to shorten branches, calls and absolute memory addresses. This option only
18677 has an effect when used on the command line for the final link step.
18679 This option makes symbolic debugging impossible.
18683 Allow the compiler to generate @emph{Long Instruction Word}
18684 instructions if the target is the @samp{AM33} or later. This is the
18685 default. This option defines the preprocessor macro @samp{__LIW__}.
18689 Do not allow the compiler to generate @emph{Long Instruction Word}
18690 instructions. This option defines the preprocessor macro
18695 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
18696 instructions if the target is the @samp{AM33} or later. This is the
18697 default. This option defines the preprocessor macro @samp{__SETLB__}.
18701 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
18702 instructions. This option defines the preprocessor macro
18703 @samp{__NO_SETLB__}.
18707 @node Moxie Options
18708 @subsection Moxie Options
18709 @cindex Moxie Options
18715 Generate big-endian code. This is the default for @samp{moxie-*-*}
18720 Generate little-endian code.
18724 Do not link in the C run-time initialization object file.
18728 @node MSP430 Options
18729 @subsection MSP430 Options
18730 @cindex MSP430 Options
18732 These options are defined for the MSP430:
18738 Force assembly output to always use hex constants. Normally such
18739 constants are signed decimals, but this option is available for
18740 testsuite and/or aesthetic purposes.
18744 Select the MCU to target. This is used to create a C preprocessor
18745 symbol based upon the MCU name, converted to upper case and pre- and
18746 post- fixed with @code{__}. This in turn will be used by the
18747 @code{msp430.h} header file to select an MCU specific supplimentary
18750 The option also sets the ISA to use. If the MCU name is one that is
18751 known to only support the 430 ISA then that is selected, otherwise the
18752 430X ISA is selected. A generic MCU name of @code{msp430} can also be
18753 used to select the 430 ISA. Similarly the generic @code{msp430x} MCU
18754 name will select the 430X ISA.
18756 In addition an MCU specific linker script will be added to the linker
18757 command line. The script's name is the name of the MCU with
18758 @code{.ld} appended. Thus specifying @option{-mmcu=xxx} on the gcc
18759 command line will define the C preprocessor symbol @code{__XXX__} and
18760 cause the linker to search for a script called @file{xxx.ld}.
18762 This option is also passed on to the assembler.
18766 Specifies the ISA to use. Accepted values are @code{msp430},
18767 @code{msp430x} and @code{msp430xv2}. This option is deprecated. The
18768 @option{-mmcu=} option should be used to select the ISA.
18772 Link to the simulator runtime libraries and linker script. Overrides
18773 any scripts that would be selected by the @option{-mmcu=} option.
18777 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
18781 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
18785 This option is passed to the assembler and linker, and allows the
18786 linker to perform certain optimizations that cannot be done until
18791 Describes the type of hardware multiply supported by the target.
18792 Accepted values are @code{none} for no hardware multiply, @code{16bit}
18793 for the original 16-bit-only multiply supported by early MCUs.
18794 @code{32bit} for the 16/32-bit multiply supported by later MCUs and
18795 @code{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
18796 A value of @code{auto} can also be given. This tells GCC to deduce
18797 the hardware multiply support based upon the MCU name provided by the
18798 @option{-mmcu} option. If no @option{-mmcu} option is specified then
18799 @code{32bit} hardware multiply support is assumed. @code{auto} is the
18802 Hardware multiplies are normally performed by calling a library
18803 routine. This saves space in the generated code. When compiling at
18804 @code{-O3} or higher however the hardware multiplier is invoked
18805 inline. This makes for bigger, but faster code.
18807 The hardware multiply routines disable interrupts whilst running and
18808 restore the previous interrupt state when they finish. This makes
18809 them safe to use inside interrupt handlers as well as in normal code.
18813 Enable the use of a minimum runtime environment - no static
18814 initializers or constructors. This is intended for memory-constrained
18815 devices. The compiler will include special symbols in some objects
18816 that tell the linker and runtime which code fragments are required.
18820 @node NDS32 Options
18821 @subsection NDS32 Options
18822 @cindex NDS32 Options
18824 These options are defined for NDS32 implementations:
18829 @opindex mbig-endian
18830 Generate code in big-endian mode.
18832 @item -mlittle-endian
18833 @opindex mlittle-endian
18834 Generate code in little-endian mode.
18836 @item -mreduced-regs
18837 @opindex mreduced-regs
18838 Use reduced-set registers for register allocation.
18841 @opindex mfull-regs
18842 Use full-set registers for register allocation.
18846 Generate conditional move instructions.
18850 Do not generate conditional move instructions.
18854 Generate performance extension instructions.
18856 @item -mno-perf-ext
18857 @opindex mno-perf-ext
18858 Do not generate performance extension instructions.
18862 Generate v3 push25/pop25 instructions.
18865 @opindex mno-v3push
18866 Do not generate v3 push25/pop25 instructions.
18870 Generate 16-bit instructions.
18873 @opindex mno-16-bit
18874 Do not generate 16-bit instructions.
18877 @opindex mgp-direct
18878 Generate GP base instructions directly.
18880 @item -mno-gp-direct
18881 @opindex mno-gp-direct
18882 Do no generate GP base instructions directly.
18884 @item -misr-vector-size=@var{num}
18885 @opindex misr-vector-size
18886 Specify the size of each interrupt vector, which must be 4 or 16.
18888 @item -mcache-block-size=@var{num}
18889 @opindex mcache-block-size
18890 Specify the size of each cache block,
18891 which must be a power of 2 between 4 and 512.
18893 @item -march=@var{arch}
18895 Specify the name of the target architecture.
18897 @item -mforce-fp-as-gp
18898 @opindex mforce-fp-as-gp
18899 Prevent $fp being allocated during register allocation so that compiler
18900 is able to force performing fp-as-gp optimization.
18902 @item -mforbid-fp-as-gp
18903 @opindex mforbid-fp-as-gp
18904 Forbid using $fp to access static and global variables.
18905 This option strictly forbids fp-as-gp optimization
18906 regardless of @option{-mforce-fp-as-gp}.
18910 Use special directives to guide linker doing ex9 optimization.
18913 @opindex mctor-dtor
18914 Enable constructor/destructor feature.
18918 Guide linker to relax instructions.
18922 @node Nios II Options
18923 @subsection Nios II Options
18924 @cindex Nios II options
18925 @cindex Altera Nios II options
18927 These are the options defined for the Altera Nios II processor.
18933 @cindex smaller data references
18934 Put global and static objects less than or equal to @var{num} bytes
18935 into the small data or BSS sections instead of the normal data or BSS
18936 sections. The default value of @var{num} is 8.
18942 Generate (do not generate) GP-relative accesses for objects in the
18943 small data or BSS sections. The default is @option{-mgpopt} except
18944 when @option{-fpic} or @option{-fPIC} is specified to generate
18945 position-independent code. Note that the Nios II ABI does not permit
18946 GP-relative accesses from shared libraries.
18948 You may need to specify @option{-mno-gpopt} explicitly when building
18949 programs that include large amounts of small data, including large
18950 GOT data sections. In this case, the 16-bit offset for GP-relative
18951 addressing may not be large enough to allow access to the entire
18952 small data section.
18958 Generate little-endian (default) or big-endian (experimental) code,
18961 @item -mbypass-cache
18962 @itemx -mno-bypass-cache
18963 @opindex mno-bypass-cache
18964 @opindex mbypass-cache
18965 Force all load and store instructions to always bypass cache by
18966 using I/O variants of the instructions. The default is not to
18969 @item -mno-cache-volatile
18970 @itemx -mcache-volatile
18971 @opindex mcache-volatile
18972 @opindex mno-cache-volatile
18973 Volatile memory access bypass the cache using the I/O variants of
18974 the load and store instructions. The default is not to bypass the cache.
18976 @item -mno-fast-sw-div
18977 @itemx -mfast-sw-div
18978 @opindex mno-fast-sw-div
18979 @opindex mfast-sw-div
18980 Do not use table-based fast divide for small numbers. The default
18981 is to use the fast divide at @option{-O3} and above.
18985 @itemx -mno-hw-mulx
18989 @opindex mno-hw-mul
18991 @opindex mno-hw-mulx
18993 @opindex mno-hw-div
18995 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
18996 instructions by the compiler. The default is to emit @code{mul}
18997 and not emit @code{div} and @code{mulx}.
18999 @item -mcustom-@var{insn}=@var{N}
19000 @itemx -mno-custom-@var{insn}
19001 @opindex mcustom-@var{insn}
19002 @opindex mno-custom-@var{insn}
19003 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
19004 custom instruction with encoding @var{N} when generating code that uses
19005 @var{insn}. For example, @code{-mcustom-fadds=253} generates custom
19006 instruction 253 for single-precision floating-point add operations instead
19007 of the default behavior of using a library call.
19009 The following values of @var{insn} are supported. Except as otherwise
19010 noted, floating-point operations are expected to be implemented with
19011 normal IEEE 754 semantics and correspond directly to the C operators or the
19012 equivalent GCC built-in functions (@pxref{Other Builtins}).
19014 Single-precision floating point:
19017 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
19018 Binary arithmetic operations.
19024 Unary absolute value.
19026 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
19027 Comparison operations.
19029 @item @samp{fmins}, @samp{fmaxs}
19030 Floating-point minimum and maximum. These instructions are only
19031 generated if @option{-ffinite-math-only} is specified.
19033 @item @samp{fsqrts}
19034 Unary square root operation.
19036 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
19037 Floating-point trigonometric and exponential functions. These instructions
19038 are only generated if @option{-funsafe-math-optimizations} is also specified.
19042 Double-precision floating point:
19045 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
19046 Binary arithmetic operations.
19052 Unary absolute value.
19054 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
19055 Comparison operations.
19057 @item @samp{fmind}, @samp{fmaxd}
19058 Double-precision minimum and maximum. These instructions are only
19059 generated if @option{-ffinite-math-only} is specified.
19061 @item @samp{fsqrtd}
19062 Unary square root operation.
19064 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
19065 Double-precision trigonometric and exponential functions. These instructions
19066 are only generated if @option{-funsafe-math-optimizations} is also specified.
19072 @item @samp{fextsd}
19073 Conversion from single precision to double precision.
19075 @item @samp{ftruncds}
19076 Conversion from double precision to single precision.
19078 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
19079 Conversion from floating point to signed or unsigned integer types, with
19080 truncation towards zero.
19083 Conversion from single-precision floating point to signed integer,
19084 rounding to the nearest integer and ties away from zero.
19085 This corresponds to the @code{__builtin_lroundf} function when
19086 @option{-fno-math-errno} is used.
19088 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
19089 Conversion from signed or unsigned integer types to floating-point types.
19093 In addition, all of the following transfer instructions for internal
19094 registers X and Y must be provided to use any of the double-precision
19095 floating-point instructions. Custom instructions taking two
19096 double-precision source operands expect the first operand in the
19097 64-bit register X. The other operand (or only operand of a unary
19098 operation) is given to the custom arithmetic instruction with the
19099 least significant half in source register @var{src1} and the most
19100 significant half in @var{src2}. A custom instruction that returns a
19101 double-precision result returns the most significant 32 bits in the
19102 destination register and the other half in 32-bit register Y.
19103 GCC automatically generates the necessary code sequences to write
19104 register X and/or read register Y when double-precision floating-point
19105 instructions are used.
19110 Write @var{src1} into the least significant half of X and @var{src2} into
19111 the most significant half of X.
19114 Write @var{src1} into Y.
19116 @item @samp{frdxhi}, @samp{frdxlo}
19117 Read the most or least (respectively) significant half of X and store it in
19121 Read the value of Y and store it into @var{dest}.
19124 Note that you can gain more local control over generation of Nios II custom
19125 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
19126 and @code{target("no-custom-@var{insn}")} function attributes
19127 (@pxref{Function Attributes})
19128 or pragmas (@pxref{Function Specific Option Pragmas}).
19130 @item -mcustom-fpu-cfg=@var{name}
19131 @opindex mcustom-fpu-cfg
19133 This option enables a predefined, named set of custom instruction encodings
19134 (see @option{-mcustom-@var{insn}} above).
19135 Currently, the following sets are defined:
19137 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
19138 @gccoptlist{-mcustom-fmuls=252 @gol
19139 -mcustom-fadds=253 @gol
19140 -mcustom-fsubs=254 @gol
19141 -fsingle-precision-constant}
19143 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
19144 @gccoptlist{-mcustom-fmuls=252 @gol
19145 -mcustom-fadds=253 @gol
19146 -mcustom-fsubs=254 @gol
19147 -mcustom-fdivs=255 @gol
19148 -fsingle-precision-constant}
19150 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
19151 @gccoptlist{-mcustom-floatus=243 @gol
19152 -mcustom-fixsi=244 @gol
19153 -mcustom-floatis=245 @gol
19154 -mcustom-fcmpgts=246 @gol
19155 -mcustom-fcmples=249 @gol
19156 -mcustom-fcmpeqs=250 @gol
19157 -mcustom-fcmpnes=251 @gol
19158 -mcustom-fmuls=252 @gol
19159 -mcustom-fadds=253 @gol
19160 -mcustom-fsubs=254 @gol
19161 -mcustom-fdivs=255 @gol
19162 -fsingle-precision-constant}
19164 Custom instruction assignments given by individual
19165 @option{-mcustom-@var{insn}=} options override those given by
19166 @option{-mcustom-fpu-cfg=}, regardless of the
19167 order of the options on the command line.
19169 Note that you can gain more local control over selection of a FPU
19170 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
19171 function attribute (@pxref{Function Attributes})
19172 or pragma (@pxref{Function Specific Option Pragmas}).
19176 These additional @samp{-m} options are available for the Altera Nios II
19177 ELF (bare-metal) target:
19183 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
19184 startup and termination code, and is typically used in conjunction with
19185 @option{-msys-crt0=} to specify the location of the alternate startup code
19186 provided by the HAL BSP.
19190 Link with a limited version of the C library, @option{-lsmallc}, rather than
19193 @item -msys-crt0=@var{startfile}
19195 @var{startfile} is the file name of the startfile (crt0) to use
19196 when linking. This option is only useful in conjunction with @option{-mhal}.
19198 @item -msys-lib=@var{systemlib}
19200 @var{systemlib} is the library name of the library that provides
19201 low-level system calls required by the C library,
19202 e.g. @code{read} and @code{write}.
19203 This option is typically used to link with a library provided by a HAL BSP.
19207 @node PDP-11 Options
19208 @subsection PDP-11 Options
19209 @cindex PDP-11 Options
19211 These options are defined for the PDP-11:
19216 Use hardware FPP floating point. This is the default. (FIS floating
19217 point on the PDP-11/40 is not supported.)
19220 @opindex msoft-float
19221 Do not use hardware floating point.
19225 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
19229 Return floating-point results in memory. This is the default.
19233 Generate code for a PDP-11/40.
19237 Generate code for a PDP-11/45. This is the default.
19241 Generate code for a PDP-11/10.
19243 @item -mbcopy-builtin
19244 @opindex mbcopy-builtin
19245 Use inline @code{movmemhi} patterns for copying memory. This is the
19250 Do not use inline @code{movmemhi} patterns for copying memory.
19256 Use 16-bit @code{int}. This is the default.
19262 Use 32-bit @code{int}.
19265 @itemx -mno-float32
19267 @opindex mno-float32
19268 Use 64-bit @code{float}. This is the default.
19271 @itemx -mno-float64
19273 @opindex mno-float64
19274 Use 32-bit @code{float}.
19278 Use @code{abshi2} pattern. This is the default.
19282 Do not use @code{abshi2} pattern.
19284 @item -mbranch-expensive
19285 @opindex mbranch-expensive
19286 Pretend that branches are expensive. This is for experimenting with
19287 code generation only.
19289 @item -mbranch-cheap
19290 @opindex mbranch-cheap
19291 Do not pretend that branches are expensive. This is the default.
19295 Use Unix assembler syntax. This is the default when configured for
19296 @samp{pdp11-*-bsd}.
19300 Use DEC assembler syntax. This is the default when configured for any
19301 PDP-11 target other than @samp{pdp11-*-bsd}.
19304 @node picoChip Options
19305 @subsection picoChip Options
19306 @cindex picoChip options
19308 These @samp{-m} options are defined for picoChip implementations:
19312 @item -mae=@var{ae_type}
19314 Set the instruction set, register set, and instruction scheduling
19315 parameters for array element type @var{ae_type}. Supported values
19316 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
19318 @option{-mae=ANY} selects a completely generic AE type. Code
19319 generated with this option runs on any of the other AE types. The
19320 code is not as efficient as it would be if compiled for a specific
19321 AE type, and some types of operation (e.g., multiplication) do not
19322 work properly on all types of AE.
19324 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
19325 for compiled code, and is the default.
19327 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
19328 option may suffer from poor performance of byte (char) manipulation,
19329 since the DSP AE does not provide hardware support for byte load/stores.
19331 @item -msymbol-as-address
19332 Enable the compiler to directly use a symbol name as an address in a
19333 load/store instruction, without first loading it into a
19334 register. Typically, the use of this option generates larger
19335 programs, which run faster than when the option isn't used. However, the
19336 results vary from program to program, so it is left as a user option,
19337 rather than being permanently enabled.
19339 @item -mno-inefficient-warnings
19340 Disables warnings about the generation of inefficient code. These
19341 warnings can be generated, for example, when compiling code that
19342 performs byte-level memory operations on the MAC AE type. The MAC AE has
19343 no hardware support for byte-level memory operations, so all byte
19344 load/stores must be synthesized from word load/store operations. This is
19345 inefficient and a warning is generated to indicate
19346 that you should rewrite the code to avoid byte operations, or to target
19347 an AE type that has the necessary hardware support. This option disables
19352 @node PowerPC Options
19353 @subsection PowerPC Options
19354 @cindex PowerPC options
19356 These are listed under @xref{RS/6000 and PowerPC Options}.
19359 @subsection RL78 Options
19360 @cindex RL78 Options
19366 Links in additional target libraries to support operation within a
19373 Specifies the type of hardware multiplication support to be used. The
19374 default is @code{none}, which uses software multiplication functions.
19375 The @code{g13} option is for the hardware multiply/divide peripheral
19376 only on the RL78/G13 targets. The @code{rl78} option is for the
19377 standard hardware multiplication defined in the RL78 software manual.
19379 @item -m64bit-doubles
19380 @itemx -m32bit-doubles
19381 @opindex m64bit-doubles
19382 @opindex m32bit-doubles
19383 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
19384 or 32 bits (@option{-m32bit-doubles}) in size. The default is
19385 @option{-m32bit-doubles}.
19389 @node RS/6000 and PowerPC Options
19390 @subsection IBM RS/6000 and PowerPC Options
19391 @cindex RS/6000 and PowerPC Options
19392 @cindex IBM RS/6000 and PowerPC Options
19394 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
19396 @item -mpowerpc-gpopt
19397 @itemx -mno-powerpc-gpopt
19398 @itemx -mpowerpc-gfxopt
19399 @itemx -mno-powerpc-gfxopt
19402 @itemx -mno-powerpc64
19406 @itemx -mno-popcntb
19408 @itemx -mno-popcntd
19417 @itemx -mno-hard-dfp
19418 @opindex mpowerpc-gpopt
19419 @opindex mno-powerpc-gpopt
19420 @opindex mpowerpc-gfxopt
19421 @opindex mno-powerpc-gfxopt
19422 @opindex mpowerpc64
19423 @opindex mno-powerpc64
19427 @opindex mno-popcntb
19429 @opindex mno-popcntd
19435 @opindex mno-mfpgpr
19437 @opindex mno-hard-dfp
19438 You use these options to specify which instructions are available on the
19439 processor you are using. The default value of these options is
19440 determined when configuring GCC@. Specifying the
19441 @option{-mcpu=@var{cpu_type}} overrides the specification of these
19442 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
19443 rather than the options listed above.
19445 Specifying @option{-mpowerpc-gpopt} allows
19446 GCC to use the optional PowerPC architecture instructions in the
19447 General Purpose group, including floating-point square root. Specifying
19448 @option{-mpowerpc-gfxopt} allows GCC to
19449 use the optional PowerPC architecture instructions in the Graphics
19450 group, including floating-point select.
19452 The @option{-mmfcrf} option allows GCC to generate the move from
19453 condition register field instruction implemented on the POWER4
19454 processor and other processors that support the PowerPC V2.01
19456 The @option{-mpopcntb} option allows GCC to generate the popcount and
19457 double-precision FP reciprocal estimate instruction implemented on the
19458 POWER5 processor and other processors that support the PowerPC V2.02
19460 The @option{-mpopcntd} option allows GCC to generate the popcount
19461 instruction implemented on the POWER7 processor and other processors
19462 that support the PowerPC V2.06 architecture.
19463 The @option{-mfprnd} option allows GCC to generate the FP round to
19464 integer instructions implemented on the POWER5+ processor and other
19465 processors that support the PowerPC V2.03 architecture.
19466 The @option{-mcmpb} option allows GCC to generate the compare bytes
19467 instruction implemented on the POWER6 processor and other processors
19468 that support the PowerPC V2.05 architecture.
19469 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
19470 general-purpose register instructions implemented on the POWER6X
19471 processor and other processors that support the extended PowerPC V2.05
19473 The @option{-mhard-dfp} option allows GCC to generate the decimal
19474 floating-point instructions implemented on some POWER processors.
19476 The @option{-mpowerpc64} option allows GCC to generate the additional
19477 64-bit instructions that are found in the full PowerPC64 architecture
19478 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
19479 @option{-mno-powerpc64}.
19481 @item -mcpu=@var{cpu_type}
19483 Set architecture type, register usage, and
19484 instruction scheduling parameters for machine type @var{cpu_type}.
19485 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
19486 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
19487 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
19488 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
19489 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
19490 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
19491 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
19492 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
19493 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
19494 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8}, @samp{powerpc},
19495 @samp{powerpc64}, and @samp{rs64}.
19497 @option{-mcpu=powerpc}, and @option{-mcpu=powerpc64} specify pure 32-bit
19498 PowerPC and 64-bit PowerPC architecture machine
19499 types, with an appropriate, generic processor model assumed for
19500 scheduling purposes.
19502 The other options specify a specific processor. Code generated under
19503 those options runs best on that processor, and may not run at all on
19506 The @option{-mcpu} options automatically enable or disable the
19509 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
19510 -mpopcntb -mpopcntd -mpowerpc64 @gol
19511 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
19512 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
19513 -mcrypto -mdirect-move -mpower8-fusion -mpower8-vector @gol
19514 -mquad-memory -mquad-memory-atomic}
19516 The particular options set for any particular CPU varies between
19517 compiler versions, depending on what setting seems to produce optimal
19518 code for that CPU; it doesn't necessarily reflect the actual hardware's
19519 capabilities. If you wish to set an individual option to a particular
19520 value, you may specify it after the @option{-mcpu} option, like
19521 @option{-mcpu=970 -mno-altivec}.
19523 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
19524 not enabled or disabled by the @option{-mcpu} option at present because
19525 AIX does not have full support for these options. You may still
19526 enable or disable them individually if you're sure it'll work in your
19529 @item -mtune=@var{cpu_type}
19531 Set the instruction scheduling parameters for machine type
19532 @var{cpu_type}, but do not set the architecture type or register usage,
19533 as @option{-mcpu=@var{cpu_type}} does. The same
19534 values for @var{cpu_type} are used for @option{-mtune} as for
19535 @option{-mcpu}. If both are specified, the code generated uses the
19536 architecture and registers set by @option{-mcpu}, but the
19537 scheduling parameters set by @option{-mtune}.
19539 @item -mcmodel=small
19540 @opindex mcmodel=small
19541 Generate PowerPC64 code for the small model: The TOC is limited to
19544 @item -mcmodel=medium
19545 @opindex mcmodel=medium
19546 Generate PowerPC64 code for the medium model: The TOC and other static
19547 data may be up to a total of 4G in size.
19549 @item -mcmodel=large
19550 @opindex mcmodel=large
19551 Generate PowerPC64 code for the large model: The TOC may be up to 4G
19552 in size. Other data and code is only limited by the 64-bit address
19556 @itemx -mno-altivec
19558 @opindex mno-altivec
19559 Generate code that uses (does not use) AltiVec instructions, and also
19560 enable the use of built-in functions that allow more direct access to
19561 the AltiVec instruction set. You may also need to set
19562 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
19565 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
19566 @option{-maltivec=be}, the element order for Altivec intrinsics such
19567 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert} will
19568 match array element order corresponding to the endianness of the
19569 target. That is, element zero identifies the leftmost element in a
19570 vector register when targeting a big-endian platform, and identifies
19571 the rightmost element in a vector register when targeting a
19572 little-endian platform.
19575 @opindex maltivec=be
19576 Generate Altivec instructions using big-endian element order,
19577 regardless of whether the target is big- or little-endian. This is
19578 the default when targeting a big-endian platform.
19580 The element order is used to interpret element numbers in Altivec
19581 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19582 @code{vec_insert}. By default, these will match array element order
19583 corresponding to the endianness for the target.
19586 @opindex maltivec=le
19587 Generate Altivec instructions using little-endian element order,
19588 regardless of whether the target is big- or little-endian. This is
19589 the default when targeting a little-endian platform. This option is
19590 currently ignored when targeting a big-endian platform.
19592 The element order is used to interpret element numbers in Altivec
19593 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19594 @code{vec_insert}. By default, these will match array element order
19595 corresponding to the endianness for the target.
19600 @opindex mno-vrsave
19601 Generate VRSAVE instructions when generating AltiVec code.
19603 @item -mgen-cell-microcode
19604 @opindex mgen-cell-microcode
19605 Generate Cell microcode instructions.
19607 @item -mwarn-cell-microcode
19608 @opindex mwarn-cell-microcode
19609 Warn when a Cell microcode instruction is emitted. An example
19610 of a Cell microcode instruction is a variable shift.
19613 @opindex msecure-plt
19614 Generate code that allows @command{ld} and @command{ld.so}
19615 to build executables and shared
19616 libraries with non-executable @code{.plt} and @code{.got} sections.
19618 32-bit SYSV ABI option.
19622 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
19624 requires @code{.plt} and @code{.got}
19625 sections that are both writable and executable.
19626 This is a PowerPC 32-bit SYSV ABI option.
19632 This switch enables or disables the generation of ISEL instructions.
19634 @item -misel=@var{yes/no}
19635 This switch has been deprecated. Use @option{-misel} and
19636 @option{-mno-isel} instead.
19642 This switch enables or disables the generation of SPE simd
19648 @opindex mno-paired
19649 This switch enables or disables the generation of PAIRED simd
19652 @item -mspe=@var{yes/no}
19653 This option has been deprecated. Use @option{-mspe} and
19654 @option{-mno-spe} instead.
19660 Generate code that uses (does not use) vector/scalar (VSX)
19661 instructions, and also enable the use of built-in functions that allow
19662 more direct access to the VSX instruction set.
19667 @opindex mno-crypto
19668 Enable the use (disable) of the built-in functions that allow direct
19669 access to the cryptographic instructions that were added in version
19670 2.07 of the PowerPC ISA.
19672 @item -mdirect-move
19673 @itemx -mno-direct-move
19674 @opindex mdirect-move
19675 @opindex mno-direct-move
19676 Generate code that uses (does not use) the instructions to move data
19677 between the general purpose registers and the vector/scalar (VSX)
19678 registers that were added in version 2.07 of the PowerPC ISA.
19680 @item -mpower8-fusion
19681 @itemx -mno-power8-fusion
19682 @opindex mpower8-fusion
19683 @opindex mno-power8-fusion
19684 Generate code that keeps (does not keeps) some integer operations
19685 adjacent so that the instructions can be fused together on power8 and
19688 @item -mpower8-vector
19689 @itemx -mno-power8-vector
19690 @opindex mpower8-vector
19691 @opindex mno-power8-vector
19692 Generate code that uses (does not use) the vector and scalar
19693 instructions that were added in version 2.07 of the PowerPC ISA. Also
19694 enable the use of built-in functions that allow more direct access to
19695 the vector instructions.
19697 @item -mquad-memory
19698 @itemx -mno-quad-memory
19699 @opindex mquad-memory
19700 @opindex mno-quad-memory
19701 Generate code that uses (does not use) the non-atomic quad word memory
19702 instructions. The @option{-mquad-memory} option requires use of
19705 @item -mquad-memory-atomic
19706 @itemx -mno-quad-memory-atomic
19707 @opindex mquad-memory-atomic
19708 @opindex mno-quad-memory-atomic
19709 Generate code that uses (does not use) the atomic quad word memory
19710 instructions. The @option{-mquad-memory-atomic} option requires use of
19713 @item -mfloat-gprs=@var{yes/single/double/no}
19714 @itemx -mfloat-gprs
19715 @opindex mfloat-gprs
19716 This switch enables or disables the generation of floating-point
19717 operations on the general-purpose registers for architectures that
19720 The argument @var{yes} or @var{single} enables the use of
19721 single-precision floating-point operations.
19723 The argument @var{double} enables the use of single and
19724 double-precision floating-point operations.
19726 The argument @var{no} disables floating-point operations on the
19727 general-purpose registers.
19729 This option is currently only available on the MPC854x.
19735 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
19736 targets (including GNU/Linux). The 32-bit environment sets int, long
19737 and pointer to 32 bits and generates code that runs on any PowerPC
19738 variant. The 64-bit environment sets int to 32 bits and long and
19739 pointer to 64 bits, and generates code for PowerPC64, as for
19740 @option{-mpowerpc64}.
19743 @itemx -mno-fp-in-toc
19744 @itemx -mno-sum-in-toc
19745 @itemx -mminimal-toc
19747 @opindex mno-fp-in-toc
19748 @opindex mno-sum-in-toc
19749 @opindex mminimal-toc
19750 Modify generation of the TOC (Table Of Contents), which is created for
19751 every executable file. The @option{-mfull-toc} option is selected by
19752 default. In that case, GCC allocates at least one TOC entry for
19753 each unique non-automatic variable reference in your program. GCC
19754 also places floating-point constants in the TOC@. However, only
19755 16,384 entries are available in the TOC@.
19757 If you receive a linker error message that saying you have overflowed
19758 the available TOC space, you can reduce the amount of TOC space used
19759 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
19760 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
19761 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
19762 generate code to calculate the sum of an address and a constant at
19763 run time instead of putting that sum into the TOC@. You may specify one
19764 or both of these options. Each causes GCC to produce very slightly
19765 slower and larger code at the expense of conserving TOC space.
19767 If you still run out of space in the TOC even when you specify both of
19768 these options, specify @option{-mminimal-toc} instead. This option causes
19769 GCC to make only one TOC entry for every file. When you specify this
19770 option, GCC produces code that is slower and larger but which
19771 uses extremely little TOC space. You may wish to use this option
19772 only on files that contain less frequently-executed code.
19778 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
19779 @code{long} type, and the infrastructure needed to support them.
19780 Specifying @option{-maix64} implies @option{-mpowerpc64},
19781 while @option{-maix32} disables the 64-bit ABI and
19782 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
19785 @itemx -mno-xl-compat
19786 @opindex mxl-compat
19787 @opindex mno-xl-compat
19788 Produce code that conforms more closely to IBM XL compiler semantics
19789 when using AIX-compatible ABI@. Pass floating-point arguments to
19790 prototyped functions beyond the register save area (RSA) on the stack
19791 in addition to argument FPRs. Do not assume that most significant
19792 double in 128-bit long double value is properly rounded when comparing
19793 values and converting to double. Use XL symbol names for long double
19796 The AIX calling convention was extended but not initially documented to
19797 handle an obscure K&R C case of calling a function that takes the
19798 address of its arguments with fewer arguments than declared. IBM XL
19799 compilers access floating-point arguments that do not fit in the
19800 RSA from the stack when a subroutine is compiled without
19801 optimization. Because always storing floating-point arguments on the
19802 stack is inefficient and rarely needed, this option is not enabled by
19803 default and only is necessary when calling subroutines compiled by IBM
19804 XL compilers without optimization.
19808 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
19809 application written to use message passing with special startup code to
19810 enable the application to run. The system must have PE installed in the
19811 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
19812 must be overridden with the @option{-specs=} option to specify the
19813 appropriate directory location. The Parallel Environment does not
19814 support threads, so the @option{-mpe} option and the @option{-pthread}
19815 option are incompatible.
19817 @item -malign-natural
19818 @itemx -malign-power
19819 @opindex malign-natural
19820 @opindex malign-power
19821 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
19822 @option{-malign-natural} overrides the ABI-defined alignment of larger
19823 types, such as floating-point doubles, on their natural size-based boundary.
19824 The option @option{-malign-power} instructs GCC to follow the ABI-specified
19825 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
19827 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
19831 @itemx -mhard-float
19832 @opindex msoft-float
19833 @opindex mhard-float
19834 Generate code that does not use (uses) the floating-point register set.
19835 Software floating-point emulation is provided if you use the
19836 @option{-msoft-float} option, and pass the option to GCC when linking.
19838 @item -msingle-float
19839 @itemx -mdouble-float
19840 @opindex msingle-float
19841 @opindex mdouble-float
19842 Generate code for single- or double-precision floating-point operations.
19843 @option{-mdouble-float} implies @option{-msingle-float}.
19846 @opindex msimple-fpu
19847 Do not generate @code{sqrt} and @code{div} instructions for hardware
19848 floating-point unit.
19850 @item -mfpu=@var{name}
19852 Specify type of floating-point unit. Valid values for @var{name} are
19853 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
19854 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
19855 @samp{sp_full} (equivalent to @option{-msingle-float}),
19856 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
19859 @opindex mxilinx-fpu
19860 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
19863 @itemx -mno-multiple
19865 @opindex mno-multiple
19866 Generate code that uses (does not use) the load multiple word
19867 instructions and the store multiple word instructions. These
19868 instructions are generated by default on POWER systems, and not
19869 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
19870 PowerPC systems, since those instructions do not work when the
19871 processor is in little-endian mode. The exceptions are PPC740 and
19872 PPC750 which permit these instructions in little-endian mode.
19877 @opindex mno-string
19878 Generate code that uses (does not use) the load string instructions
19879 and the store string word instructions to save multiple registers and
19880 do small block moves. These instructions are generated by default on
19881 POWER systems, and not generated on PowerPC systems. Do not use
19882 @option{-mstring} on little-endian PowerPC systems, since those
19883 instructions do not work when the processor is in little-endian mode.
19884 The exceptions are PPC740 and PPC750 which permit these instructions
19885 in little-endian mode.
19890 @opindex mno-update
19891 Generate code that uses (does not use) the load or store instructions
19892 that update the base register to the address of the calculated memory
19893 location. These instructions are generated by default. If you use
19894 @option{-mno-update}, there is a small window between the time that the
19895 stack pointer is updated and the address of the previous frame is
19896 stored, which means code that walks the stack frame across interrupts or
19897 signals may get corrupted data.
19899 @item -mavoid-indexed-addresses
19900 @itemx -mno-avoid-indexed-addresses
19901 @opindex mavoid-indexed-addresses
19902 @opindex mno-avoid-indexed-addresses
19903 Generate code that tries to avoid (not avoid) the use of indexed load
19904 or store instructions. These instructions can incur a performance
19905 penalty on Power6 processors in certain situations, such as when
19906 stepping through large arrays that cross a 16M boundary. This option
19907 is enabled by default when targeting Power6 and disabled otherwise.
19910 @itemx -mno-fused-madd
19911 @opindex mfused-madd
19912 @opindex mno-fused-madd
19913 Generate code that uses (does not use) the floating-point multiply and
19914 accumulate instructions. These instructions are generated by default
19915 if hardware floating point is used. The machine-dependent
19916 @option{-mfused-madd} option is now mapped to the machine-independent
19917 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
19918 mapped to @option{-ffp-contract=off}.
19924 Generate code that uses (does not use) the half-word multiply and
19925 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
19926 These instructions are generated by default when targeting those
19933 Generate code that uses (does not use) the string-search @samp{dlmzb}
19934 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
19935 generated by default when targeting those processors.
19937 @item -mno-bit-align
19939 @opindex mno-bit-align
19940 @opindex mbit-align
19941 On System V.4 and embedded PowerPC systems do not (do) force structures
19942 and unions that contain bit-fields to be aligned to the base type of the
19945 For example, by default a structure containing nothing but 8
19946 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
19947 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
19948 the structure is aligned to a 1-byte boundary and is 1 byte in
19951 @item -mno-strict-align
19952 @itemx -mstrict-align
19953 @opindex mno-strict-align
19954 @opindex mstrict-align
19955 On System V.4 and embedded PowerPC systems do not (do) assume that
19956 unaligned memory references are handled by the system.
19958 @item -mrelocatable
19959 @itemx -mno-relocatable
19960 @opindex mrelocatable
19961 @opindex mno-relocatable
19962 Generate code that allows (does not allow) a static executable to be
19963 relocated to a different address at run time. A simple embedded
19964 PowerPC system loader should relocate the entire contents of
19965 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
19966 a table of 32-bit addresses generated by this option. For this to
19967 work, all objects linked together must be compiled with
19968 @option{-mrelocatable} or @option{-mrelocatable-lib}.
19969 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
19971 @item -mrelocatable-lib
19972 @itemx -mno-relocatable-lib
19973 @opindex mrelocatable-lib
19974 @opindex mno-relocatable-lib
19975 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
19976 @code{.fixup} section to allow static executables to be relocated at
19977 run time, but @option{-mrelocatable-lib} does not use the smaller stack
19978 alignment of @option{-mrelocatable}. Objects compiled with
19979 @option{-mrelocatable-lib} may be linked with objects compiled with
19980 any combination of the @option{-mrelocatable} options.
19986 On System V.4 and embedded PowerPC systems do not (do) assume that
19987 register 2 contains a pointer to a global area pointing to the addresses
19988 used in the program.
19991 @itemx -mlittle-endian
19993 @opindex mlittle-endian
19994 On System V.4 and embedded PowerPC systems compile code for the
19995 processor in little-endian mode. The @option{-mlittle-endian} option is
19996 the same as @option{-mlittle}.
19999 @itemx -mbig-endian
20001 @opindex mbig-endian
20002 On System V.4 and embedded PowerPC systems compile code for the
20003 processor in big-endian mode. The @option{-mbig-endian} option is
20004 the same as @option{-mbig}.
20006 @item -mdynamic-no-pic
20007 @opindex mdynamic-no-pic
20008 On Darwin and Mac OS X systems, compile code so that it is not
20009 relocatable, but that its external references are relocatable. The
20010 resulting code is suitable for applications, but not shared
20013 @item -msingle-pic-base
20014 @opindex msingle-pic-base
20015 Treat the register used for PIC addressing as read-only, rather than
20016 loading it in the prologue for each function. The runtime system is
20017 responsible for initializing this register with an appropriate value
20018 before execution begins.
20020 @item -mprioritize-restricted-insns=@var{priority}
20021 @opindex mprioritize-restricted-insns
20022 This option controls the priority that is assigned to
20023 dispatch-slot restricted instructions during the second scheduling
20024 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
20025 or @samp{2} to assign no, highest, or second-highest (respectively)
20026 priority to dispatch-slot restricted
20029 @item -msched-costly-dep=@var{dependence_type}
20030 @opindex msched-costly-dep
20031 This option controls which dependences are considered costly
20032 by the target during instruction scheduling. The argument
20033 @var{dependence_type} takes one of the following values:
20037 No dependence is costly.
20040 All dependences are costly.
20042 @item @samp{true_store_to_load}
20043 A true dependence from store to load is costly.
20045 @item @samp{store_to_load}
20046 Any dependence from store to load is costly.
20049 Any dependence for which the latency is greater than or equal to
20050 @var{number} is costly.
20053 @item -minsert-sched-nops=@var{scheme}
20054 @opindex minsert-sched-nops
20055 This option controls which NOP insertion scheme is used during
20056 the second scheduling pass. The argument @var{scheme} takes one of the
20064 Pad with NOPs any dispatch group that has vacant issue slots,
20065 according to the scheduler's grouping.
20067 @item @samp{regroup_exact}
20068 Insert NOPs to force costly dependent insns into
20069 separate groups. Insert exactly as many NOPs as needed to force an insn
20070 to a new group, according to the estimated processor grouping.
20073 Insert NOPs to force costly dependent insns into
20074 separate groups. Insert @var{number} NOPs to force an insn to a new group.
20078 @opindex mcall-sysv
20079 On System V.4 and embedded PowerPC systems compile code using calling
20080 conventions that adhere to the March 1995 draft of the System V
20081 Application Binary Interface, PowerPC processor supplement. This is the
20082 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
20084 @item -mcall-sysv-eabi
20086 @opindex mcall-sysv-eabi
20087 @opindex mcall-eabi
20088 Specify both @option{-mcall-sysv} and @option{-meabi} options.
20090 @item -mcall-sysv-noeabi
20091 @opindex mcall-sysv-noeabi
20092 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
20094 @item -mcall-aixdesc
20096 On System V.4 and embedded PowerPC systems compile code for the AIX
20100 @opindex mcall-linux
20101 On System V.4 and embedded PowerPC systems compile code for the
20102 Linux-based GNU system.
20104 @item -mcall-freebsd
20105 @opindex mcall-freebsd
20106 On System V.4 and embedded PowerPC systems compile code for the
20107 FreeBSD operating system.
20109 @item -mcall-netbsd
20110 @opindex mcall-netbsd
20111 On System V.4 and embedded PowerPC systems compile code for the
20112 NetBSD operating system.
20114 @item -mcall-openbsd
20115 @opindex mcall-netbsd
20116 On System V.4 and embedded PowerPC systems compile code for the
20117 OpenBSD operating system.
20119 @item -maix-struct-return
20120 @opindex maix-struct-return
20121 Return all structures in memory (as specified by the AIX ABI)@.
20123 @item -msvr4-struct-return
20124 @opindex msvr4-struct-return
20125 Return structures smaller than 8 bytes in registers (as specified by the
20128 @item -mabi=@var{abi-type}
20130 Extend the current ABI with a particular extension, or remove such extension.
20131 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
20132 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble},
20133 @var{elfv1}, @var{elfv2}@.
20137 Extend the current ABI with SPE ABI extensions. This does not change
20138 the default ABI, instead it adds the SPE ABI extensions to the current
20142 @opindex mabi=no-spe
20143 Disable Book-E SPE ABI extensions for the current ABI@.
20145 @item -mabi=ibmlongdouble
20146 @opindex mabi=ibmlongdouble
20147 Change the current ABI to use IBM extended-precision long double.
20148 This is a PowerPC 32-bit SYSV ABI option.
20150 @item -mabi=ieeelongdouble
20151 @opindex mabi=ieeelongdouble
20152 Change the current ABI to use IEEE extended-precision long double.
20153 This is a PowerPC 32-bit Linux ABI option.
20156 @opindex mabi=elfv1
20157 Change the current ABI to use the ELFv1 ABI.
20158 This is the default ABI for big-endian PowerPC 64-bit Linux.
20159 Overriding the default ABI requires special system support and is
20160 likely to fail in spectacular ways.
20163 @opindex mabi=elfv2
20164 Change the current ABI to use the ELFv2 ABI.
20165 This is the default ABI for little-endian PowerPC 64-bit Linux.
20166 Overriding the default ABI requires special system support and is
20167 likely to fail in spectacular ways.
20170 @itemx -mno-prototype
20171 @opindex mprototype
20172 @opindex mno-prototype
20173 On System V.4 and embedded PowerPC systems assume that all calls to
20174 variable argument functions are properly prototyped. Otherwise, the
20175 compiler must insert an instruction before every non-prototyped call to
20176 set or clear bit 6 of the condition code register (@var{CR}) to
20177 indicate whether floating-point values are passed in the floating-point
20178 registers in case the function takes variable arguments. With
20179 @option{-mprototype}, only calls to prototyped variable argument functions
20180 set or clear the bit.
20184 On embedded PowerPC systems, assume that the startup module is called
20185 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
20186 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
20191 On embedded PowerPC systems, assume that the startup module is called
20192 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
20197 On embedded PowerPC systems, assume that the startup module is called
20198 @file{crt0.o} and the standard C libraries are @file{libads.a} and
20201 @item -myellowknife
20202 @opindex myellowknife
20203 On embedded PowerPC systems, assume that the startup module is called
20204 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
20209 On System V.4 and embedded PowerPC systems, specify that you are
20210 compiling for a VxWorks system.
20214 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
20215 header to indicate that @samp{eabi} extended relocations are used.
20221 On System V.4 and embedded PowerPC systems do (do not) adhere to the
20222 Embedded Applications Binary Interface (EABI), which is a set of
20223 modifications to the System V.4 specifications. Selecting @option{-meabi}
20224 means that the stack is aligned to an 8-byte boundary, a function
20225 @code{__eabi} is called from @code{main} to set up the EABI
20226 environment, and the @option{-msdata} option can use both @code{r2} and
20227 @code{r13} to point to two separate small data areas. Selecting
20228 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
20229 no EABI initialization function is called from @code{main}, and the
20230 @option{-msdata} option only uses @code{r13} to point to a single
20231 small data area. The @option{-meabi} option is on by default if you
20232 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
20235 @opindex msdata=eabi
20236 On System V.4 and embedded PowerPC systems, put small initialized
20237 @code{const} global and static data in the @samp{.sdata2} section, which
20238 is pointed to by register @code{r2}. Put small initialized
20239 non-@code{const} global and static data in the @samp{.sdata} section,
20240 which is pointed to by register @code{r13}. Put small uninitialized
20241 global and static data in the @samp{.sbss} section, which is adjacent to
20242 the @samp{.sdata} section. The @option{-msdata=eabi} option is
20243 incompatible with the @option{-mrelocatable} option. The
20244 @option{-msdata=eabi} option also sets the @option{-memb} option.
20247 @opindex msdata=sysv
20248 On System V.4 and embedded PowerPC systems, put small global and static
20249 data in the @samp{.sdata} section, which is pointed to by register
20250 @code{r13}. Put small uninitialized global and static data in the
20251 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
20252 The @option{-msdata=sysv} option is incompatible with the
20253 @option{-mrelocatable} option.
20255 @item -msdata=default
20257 @opindex msdata=default
20259 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
20260 compile code the same as @option{-msdata=eabi}, otherwise compile code the
20261 same as @option{-msdata=sysv}.
20264 @opindex msdata=data
20265 On System V.4 and embedded PowerPC systems, put small global
20266 data in the @samp{.sdata} section. Put small uninitialized global
20267 data in the @samp{.sbss} section. Do not use register @code{r13}
20268 to address small data however. This is the default behavior unless
20269 other @option{-msdata} options are used.
20273 @opindex msdata=none
20275 On embedded PowerPC systems, put all initialized global and static data
20276 in the @samp{.data} section, and all uninitialized data in the
20277 @samp{.bss} section.
20279 @item -mblock-move-inline-limit=@var{num}
20280 @opindex mblock-move-inline-limit
20281 Inline all block moves (such as calls to @code{memcpy} or structure
20282 copies) less than or equal to @var{num} bytes. The minimum value for
20283 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
20284 targets. The default value is target-specific.
20288 @cindex smaller data references (PowerPC)
20289 @cindex .sdata/.sdata2 references (PowerPC)
20290 On embedded PowerPC systems, put global and static items less than or
20291 equal to @var{num} bytes into the small data or BSS sections instead of
20292 the normal data or BSS section. By default, @var{num} is 8. The
20293 @option{-G @var{num}} switch is also passed to the linker.
20294 All modules should be compiled with the same @option{-G @var{num}} value.
20297 @itemx -mno-regnames
20299 @opindex mno-regnames
20300 On System V.4 and embedded PowerPC systems do (do not) emit register
20301 names in the assembly language output using symbolic forms.
20304 @itemx -mno-longcall
20306 @opindex mno-longcall
20307 By default assume that all calls are far away so that a longer and more
20308 expensive calling sequence is required. This is required for calls
20309 farther than 32 megabytes (33,554,432 bytes) from the current location.
20310 A short call is generated if the compiler knows
20311 the call cannot be that far away. This setting can be overridden by
20312 the @code{shortcall} function attribute, or by @code{#pragma
20315 Some linkers are capable of detecting out-of-range calls and generating
20316 glue code on the fly. On these systems, long calls are unnecessary and
20317 generate slower code. As of this writing, the AIX linker can do this,
20318 as can the GNU linker for PowerPC/64. It is planned to add this feature
20319 to the GNU linker for 32-bit PowerPC systems as well.
20321 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
20322 callee, L42}, plus a @dfn{branch island} (glue code). The two target
20323 addresses represent the callee and the branch island. The
20324 Darwin/PPC linker prefers the first address and generates a @code{bl
20325 callee} if the PPC @code{bl} instruction reaches the callee directly;
20326 otherwise, the linker generates @code{bl L42} to call the branch
20327 island. The branch island is appended to the body of the
20328 calling function; it computes the full 32-bit address of the callee
20331 On Mach-O (Darwin) systems, this option directs the compiler emit to
20332 the glue for every direct call, and the Darwin linker decides whether
20333 to use or discard it.
20335 In the future, GCC may ignore all longcall specifications
20336 when the linker is known to generate glue.
20338 @item -mtls-markers
20339 @itemx -mno-tls-markers
20340 @opindex mtls-markers
20341 @opindex mno-tls-markers
20342 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
20343 specifying the function argument. The relocation allows the linker to
20344 reliably associate function call with argument setup instructions for
20345 TLS optimization, which in turn allows GCC to better schedule the
20350 Adds support for multithreading with the @dfn{pthreads} library.
20351 This option sets flags for both the preprocessor and linker.
20356 This option enables use of the reciprocal estimate and
20357 reciprocal square root estimate instructions with additional
20358 Newton-Raphson steps to increase precision instead of doing a divide or
20359 square root and divide for floating-point arguments. You should use
20360 the @option{-ffast-math} option when using @option{-mrecip} (or at
20361 least @option{-funsafe-math-optimizations},
20362 @option{-finite-math-only}, @option{-freciprocal-math} and
20363 @option{-fno-trapping-math}). Note that while the throughput of the
20364 sequence is generally higher than the throughput of the non-reciprocal
20365 instruction, the precision of the sequence can be decreased by up to 2
20366 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
20369 @item -mrecip=@var{opt}
20370 @opindex mrecip=opt
20371 This option controls which reciprocal estimate instructions
20372 may be used. @var{opt} is a comma-separated list of options, which may
20373 be preceded by a @code{!} to invert the option:
20374 @code{all}: enable all estimate instructions,
20375 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
20376 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
20377 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
20378 @code{divf}: enable the single-precision reciprocal approximation instructions;
20379 @code{divd}: enable the double-precision reciprocal approximation instructions;
20380 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
20381 @code{rsqrtf}: enable the single-precision reciprocal square root approximation instructions;
20382 @code{rsqrtd}: enable the double-precision reciprocal square root approximation instructions;
20384 So, for example, @option{-mrecip=all,!rsqrtd} enables
20385 all of the reciprocal estimate instructions, except for the
20386 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
20387 which handle the double-precision reciprocal square root calculations.
20389 @item -mrecip-precision
20390 @itemx -mno-recip-precision
20391 @opindex mrecip-precision
20392 Assume (do not assume) that the reciprocal estimate instructions
20393 provide higher-precision estimates than is mandated by the PowerPC
20394 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
20395 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
20396 The double-precision square root estimate instructions are not generated by
20397 default on low-precision machines, since they do not provide an
20398 estimate that converges after three steps.
20400 @item -mveclibabi=@var{type}
20401 @opindex mveclibabi
20402 Specifies the ABI type to use for vectorizing intrinsics using an
20403 external library. The only type supported at present is @code{mass},
20404 which specifies to use IBM's Mathematical Acceleration Subsystem
20405 (MASS) libraries for vectorizing intrinsics using external libraries.
20406 GCC currently emits calls to @code{acosd2}, @code{acosf4},
20407 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
20408 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
20409 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
20410 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
20411 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
20412 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
20413 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
20414 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
20415 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
20416 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
20417 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
20418 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
20419 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
20420 for power7. Both @option{-ftree-vectorize} and
20421 @option{-funsafe-math-optimizations} must also be enabled. The MASS
20422 libraries must be specified at link time.
20427 Generate (do not generate) the @code{friz} instruction when the
20428 @option{-funsafe-math-optimizations} option is used to optimize
20429 rounding of floating-point values to 64-bit integer and back to floating
20430 point. The @code{friz} instruction does not return the same value if
20431 the floating-point number is too large to fit in an integer.
20433 @item -mpointers-to-nested-functions
20434 @itemx -mno-pointers-to-nested-functions
20435 @opindex mpointers-to-nested-functions
20436 Generate (do not generate) code to load up the static chain register
20437 (@var{r11}) when calling through a pointer on AIX and 64-bit Linux
20438 systems where a function pointer points to a 3-word descriptor giving
20439 the function address, TOC value to be loaded in register @var{r2}, and
20440 static chain value to be loaded in register @var{r11}. The
20441 @option{-mpointers-to-nested-functions} is on by default. You cannot
20442 call through pointers to nested functions or pointers
20443 to functions compiled in other languages that use the static chain if
20444 you use the @option{-mno-pointers-to-nested-functions}.
20446 @item -msave-toc-indirect
20447 @itemx -mno-save-toc-indirect
20448 @opindex msave-toc-indirect
20449 Generate (do not generate) code to save the TOC value in the reserved
20450 stack location in the function prologue if the function calls through
20451 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
20452 saved in the prologue, it is saved just before the call through the
20453 pointer. The @option{-mno-save-toc-indirect} option is the default.
20455 @item -mcompat-align-parm
20456 @itemx -mno-compat-align-parm
20457 @opindex mcompat-align-parm
20458 Generate (do not generate) code to pass structure parameters with a
20459 maximum alignment of 64 bits, for compatibility with older versions
20462 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
20463 structure parameter on a 128-bit boundary when that structure contained
20464 a member requiring 128-bit alignment. This is corrected in more
20465 recent versions of GCC. This option may be used to generate code
20466 that is compatible with functions compiled with older versions of
20469 The @option{-mno-compat-align-parm} option is the default.
20473 @subsection RX Options
20476 These command-line options are defined for RX targets:
20479 @item -m64bit-doubles
20480 @itemx -m32bit-doubles
20481 @opindex m64bit-doubles
20482 @opindex m32bit-doubles
20483 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
20484 or 32 bits (@option{-m32bit-doubles}) in size. The default is
20485 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
20486 works on 32-bit values, which is why the default is
20487 @option{-m32bit-doubles}.
20493 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
20494 floating-point hardware. The default is enabled for the @var{RX600}
20495 series and disabled for the @var{RX200} series.
20497 Floating-point instructions are only generated for 32-bit floating-point
20498 values, however, so the FPU hardware is not used for doubles if the
20499 @option{-m64bit-doubles} option is used.
20501 @emph{Note} If the @option{-fpu} option is enabled then
20502 @option{-funsafe-math-optimizations} is also enabled automatically.
20503 This is because the RX FPU instructions are themselves unsafe.
20505 @item -mcpu=@var{name}
20507 Selects the type of RX CPU to be targeted. Currently three types are
20508 supported, the generic @var{RX600} and @var{RX200} series hardware and
20509 the specific @var{RX610} CPU. The default is @var{RX600}.
20511 The only difference between @var{RX600} and @var{RX610} is that the
20512 @var{RX610} does not support the @code{MVTIPL} instruction.
20514 The @var{RX200} series does not have a hardware floating-point unit
20515 and so @option{-nofpu} is enabled by default when this type is
20518 @item -mbig-endian-data
20519 @itemx -mlittle-endian-data
20520 @opindex mbig-endian-data
20521 @opindex mlittle-endian-data
20522 Store data (but not code) in the big-endian format. The default is
20523 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
20526 @item -msmall-data-limit=@var{N}
20527 @opindex msmall-data-limit
20528 Specifies the maximum size in bytes of global and static variables
20529 which can be placed into the small data area. Using the small data
20530 area can lead to smaller and faster code, but the size of area is
20531 limited and it is up to the programmer to ensure that the area does
20532 not overflow. Also when the small data area is used one of the RX's
20533 registers (usually @code{r13}) is reserved for use pointing to this
20534 area, so it is no longer available for use by the compiler. This
20535 could result in slower and/or larger code if variables are pushed onto
20536 the stack instead of being held in this register.
20538 Note, common variables (variables that have not been initialized) and
20539 constants are not placed into the small data area as they are assigned
20540 to other sections in the output executable.
20542 The default value is zero, which disables this feature. Note, this
20543 feature is not enabled by default with higher optimization levels
20544 (@option{-O2} etc) because of the potentially detrimental effects of
20545 reserving a register. It is up to the programmer to experiment and
20546 discover whether this feature is of benefit to their program. See the
20547 description of the @option{-mpid} option for a description of how the
20548 actual register to hold the small data area pointer is chosen.
20554 Use the simulator runtime. The default is to use the libgloss
20555 board-specific runtime.
20557 @item -mas100-syntax
20558 @itemx -mno-as100-syntax
20559 @opindex mas100-syntax
20560 @opindex mno-as100-syntax
20561 When generating assembler output use a syntax that is compatible with
20562 Renesas's AS100 assembler. This syntax can also be handled by the GAS
20563 assembler, but it has some restrictions so it is not generated by default.
20565 @item -mmax-constant-size=@var{N}
20566 @opindex mmax-constant-size
20567 Specifies the maximum size, in bytes, of a constant that can be used as
20568 an operand in a RX instruction. Although the RX instruction set does
20569 allow constants of up to 4 bytes in length to be used in instructions,
20570 a longer value equates to a longer instruction. Thus in some
20571 circumstances it can be beneficial to restrict the size of constants
20572 that are used in instructions. Constants that are too big are instead
20573 placed into a constant pool and referenced via register indirection.
20575 The value @var{N} can be between 0 and 4. A value of 0 (the default)
20576 or 4 means that constants of any size are allowed.
20580 Enable linker relaxation. Linker relaxation is a process whereby the
20581 linker attempts to reduce the size of a program by finding shorter
20582 versions of various instructions. Disabled by default.
20584 @item -mint-register=@var{N}
20585 @opindex mint-register
20586 Specify the number of registers to reserve for fast interrupt handler
20587 functions. The value @var{N} can be between 0 and 4. A value of 1
20588 means that register @code{r13} is reserved for the exclusive use
20589 of fast interrupt handlers. A value of 2 reserves @code{r13} and
20590 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
20591 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
20592 A value of 0, the default, does not reserve any registers.
20594 @item -msave-acc-in-interrupts
20595 @opindex msave-acc-in-interrupts
20596 Specifies that interrupt handler functions should preserve the
20597 accumulator register. This is only necessary if normal code might use
20598 the accumulator register, for example because it performs 64-bit
20599 multiplications. The default is to ignore the accumulator as this
20600 makes the interrupt handlers faster.
20606 Enables the generation of position independent data. When enabled any
20607 access to constant data is done via an offset from a base address
20608 held in a register. This allows the location of constant data to be
20609 determined at run time without requiring the executable to be
20610 relocated, which is a benefit to embedded applications with tight
20611 memory constraints. Data that can be modified is not affected by this
20614 Note, using this feature reserves a register, usually @code{r13}, for
20615 the constant data base address. This can result in slower and/or
20616 larger code, especially in complicated functions.
20618 The actual register chosen to hold the constant data base address
20619 depends upon whether the @option{-msmall-data-limit} and/or the
20620 @option{-mint-register} command-line options are enabled. Starting
20621 with register @code{r13} and proceeding downwards, registers are
20622 allocated first to satisfy the requirements of @option{-mint-register},
20623 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
20624 is possible for the small data area register to be @code{r8} if both
20625 @option{-mint-register=4} and @option{-mpid} are specified on the
20628 By default this feature is not enabled. The default can be restored
20629 via the @option{-mno-pid} command-line option.
20631 @item -mno-warn-multiple-fast-interrupts
20632 @itemx -mwarn-multiple-fast-interrupts
20633 @opindex mno-warn-multiple-fast-interrupts
20634 @opindex mwarn-multiple-fast-interrupts
20635 Prevents GCC from issuing a warning message if it finds more than one
20636 fast interrupt handler when it is compiling a file. The default is to
20637 issue a warning for each extra fast interrupt handler found, as the RX
20638 only supports one such interrupt.
20642 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
20643 has special significance to the RX port when used with the
20644 @code{interrupt} function attribute. This attribute indicates a
20645 function intended to process fast interrupts. GCC ensures
20646 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
20647 and/or @code{r13} and only provided that the normal use of the
20648 corresponding registers have been restricted via the
20649 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
20652 @node S/390 and zSeries Options
20653 @subsection S/390 and zSeries Options
20654 @cindex S/390 and zSeries Options
20656 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
20660 @itemx -msoft-float
20661 @opindex mhard-float
20662 @opindex msoft-float
20663 Use (do not use) the hardware floating-point instructions and registers
20664 for floating-point operations. When @option{-msoft-float} is specified,
20665 functions in @file{libgcc.a} are used to perform floating-point
20666 operations. When @option{-mhard-float} is specified, the compiler
20667 generates IEEE floating-point instructions. This is the default.
20670 @itemx -mno-hard-dfp
20672 @opindex mno-hard-dfp
20673 Use (do not use) the hardware decimal-floating-point instructions for
20674 decimal-floating-point operations. When @option{-mno-hard-dfp} is
20675 specified, functions in @file{libgcc.a} are used to perform
20676 decimal-floating-point operations. When @option{-mhard-dfp} is
20677 specified, the compiler generates decimal-floating-point hardware
20678 instructions. This is the default for @option{-march=z9-ec} or higher.
20680 @item -mlong-double-64
20681 @itemx -mlong-double-128
20682 @opindex mlong-double-64
20683 @opindex mlong-double-128
20684 These switches control the size of @code{long double} type. A size
20685 of 64 bits makes the @code{long double} type equivalent to the @code{double}
20686 type. This is the default.
20689 @itemx -mno-backchain
20690 @opindex mbackchain
20691 @opindex mno-backchain
20692 Store (do not store) the address of the caller's frame as backchain pointer
20693 into the callee's stack frame.
20694 A backchain may be needed to allow debugging using tools that do not understand
20695 DWARF 2 call frame information.
20696 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
20697 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
20698 the backchain is placed into the topmost word of the 96/160 byte register
20701 In general, code compiled with @option{-mbackchain} is call-compatible with
20702 code compiled with @option{-mmo-backchain}; however, use of the backchain
20703 for debugging purposes usually requires that the whole binary is built with
20704 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
20705 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20706 to build a linux kernel use @option{-msoft-float}.
20708 The default is to not maintain the backchain.
20710 @item -mpacked-stack
20711 @itemx -mno-packed-stack
20712 @opindex mpacked-stack
20713 @opindex mno-packed-stack
20714 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
20715 specified, the compiler uses the all fields of the 96/160 byte register save
20716 area only for their default purpose; unused fields still take up stack space.
20717 When @option{-mpacked-stack} is specified, register save slots are densely
20718 packed at the top of the register save area; unused space is reused for other
20719 purposes, allowing for more efficient use of the available stack space.
20720 However, when @option{-mbackchain} is also in effect, the topmost word of
20721 the save area is always used to store the backchain, and the return address
20722 register is always saved two words below the backchain.
20724 As long as the stack frame backchain is not used, code generated with
20725 @option{-mpacked-stack} is call-compatible with code generated with
20726 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
20727 S/390 or zSeries generated code that uses the stack frame backchain at run
20728 time, not just for debugging purposes. Such code is not call-compatible
20729 with code compiled with @option{-mpacked-stack}. Also, note that the
20730 combination of @option{-mbackchain},
20731 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20732 to build a linux kernel use @option{-msoft-float}.
20734 The default is to not use the packed stack layout.
20737 @itemx -mno-small-exec
20738 @opindex msmall-exec
20739 @opindex mno-small-exec
20740 Generate (or do not generate) code using the @code{bras} instruction
20741 to do subroutine calls.
20742 This only works reliably if the total executable size does not
20743 exceed 64k. The default is to use the @code{basr} instruction instead,
20744 which does not have this limitation.
20750 When @option{-m31} is specified, generate code compliant to the
20751 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
20752 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
20753 particular to generate 64-bit instructions. For the @samp{s390}
20754 targets, the default is @option{-m31}, while the @samp{s390x}
20755 targets default to @option{-m64}.
20761 When @option{-mzarch} is specified, generate code using the
20762 instructions available on z/Architecture.
20763 When @option{-mesa} is specified, generate code using the
20764 instructions available on ESA/390. Note that @option{-mesa} is
20765 not possible with @option{-m64}.
20766 When generating code compliant to the GNU/Linux for S/390 ABI,
20767 the default is @option{-mesa}. When generating code compliant
20768 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
20774 Generate (or do not generate) code using the @code{mvcle} instruction
20775 to perform block moves. When @option{-mno-mvcle} is specified,
20776 use a @code{mvc} loop instead. This is the default unless optimizing for
20783 Print (or do not print) additional debug information when compiling.
20784 The default is to not print debug information.
20786 @item -march=@var{cpu-type}
20788 Generate code that runs on @var{cpu-type}, which is the name of a system
20789 representing a certain processor type. Possible values for
20790 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
20791 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
20792 When generating code using the instructions available on z/Architecture,
20793 the default is @option{-march=z900}. Otherwise, the default is
20794 @option{-march=g5}.
20796 @item -mtune=@var{cpu-type}
20798 Tune to @var{cpu-type} everything applicable about the generated code,
20799 except for the ABI and the set of available instructions.
20800 The list of @var{cpu-type} values is the same as for @option{-march}.
20801 The default is the value used for @option{-march}.
20804 @itemx -mno-tpf-trace
20805 @opindex mtpf-trace
20806 @opindex mno-tpf-trace
20807 Generate code that adds (does not add) in TPF OS specific branches to trace
20808 routines in the operating system. This option is off by default, even
20809 when compiling for the TPF OS@.
20812 @itemx -mno-fused-madd
20813 @opindex mfused-madd
20814 @opindex mno-fused-madd
20815 Generate code that uses (does not use) the floating-point multiply and
20816 accumulate instructions. These instructions are generated by default if
20817 hardware floating point is used.
20819 @item -mwarn-framesize=@var{framesize}
20820 @opindex mwarn-framesize
20821 Emit a warning if the current function exceeds the given frame size. Because
20822 this is a compile-time check it doesn't need to be a real problem when the program
20823 runs. It is intended to identify functions that most probably cause
20824 a stack overflow. It is useful to be used in an environment with limited stack
20825 size e.g.@: the linux kernel.
20827 @item -mwarn-dynamicstack
20828 @opindex mwarn-dynamicstack
20829 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
20830 arrays. This is generally a bad idea with a limited stack size.
20832 @item -mstack-guard=@var{stack-guard}
20833 @itemx -mstack-size=@var{stack-size}
20834 @opindex mstack-guard
20835 @opindex mstack-size
20836 If these options are provided the S/390 back end emits additional instructions in
20837 the function prologue that trigger a trap if the stack size is @var{stack-guard}
20838 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
20839 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
20840 the frame size of the compiled function is chosen.
20841 These options are intended to be used to help debugging stack overflow problems.
20842 The additionally emitted code causes only little overhead and hence can also be
20843 used in production-like systems without greater performance degradation. The given
20844 values have to be exact powers of 2 and @var{stack-size} has to be greater than
20845 @var{stack-guard} without exceeding 64k.
20846 In order to be efficient the extra code makes the assumption that the stack starts
20847 at an address aligned to the value given by @var{stack-size}.
20848 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
20850 @item -mhotpatch[=@var{halfwords}]
20851 @itemx -mno-hotpatch
20853 If the hotpatch option is enabled, a ``hot-patching'' function
20854 prologue is generated for all functions in the compilation unit.
20855 The funtion label is prepended with the given number of two-byte
20856 Nop instructions (@var{halfwords}, maximum 1000000) or 12 Nop
20857 instructions if no argument is present. Functions with a
20858 hot-patching prologue are never inlined automatically, and a
20859 hot-patching prologue is never generated for functions
20860 that are explicitly inline.
20862 This option can be overridden for individual functions with the
20863 @code{hotpatch} attribute.
20866 @node Score Options
20867 @subsection Score Options
20868 @cindex Score Options
20870 These options are defined for Score implementations:
20875 Compile code for big-endian mode. This is the default.
20879 Compile code for little-endian mode.
20883 Disable generation of @code{bcnz} instructions.
20887 Enable generation of unaligned load and store instructions.
20891 Enable the use of multiply-accumulate instructions. Disabled by default.
20895 Specify the SCORE5 as the target architecture.
20899 Specify the SCORE5U of the target architecture.
20903 Specify the SCORE7 as the target architecture. This is the default.
20907 Specify the SCORE7D as the target architecture.
20911 @subsection SH Options
20913 These @samp{-m} options are defined for the SH implementations:
20918 Generate code for the SH1.
20922 Generate code for the SH2.
20925 Generate code for the SH2e.
20929 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
20930 that the floating-point unit is not used.
20932 @item -m2a-single-only
20933 @opindex m2a-single-only
20934 Generate code for the SH2a-FPU, in such a way that no double-precision
20935 floating-point operations are used.
20938 @opindex m2a-single
20939 Generate code for the SH2a-FPU assuming the floating-point unit is in
20940 single-precision mode by default.
20944 Generate code for the SH2a-FPU assuming the floating-point unit is in
20945 double-precision mode by default.
20949 Generate code for the SH3.
20953 Generate code for the SH3e.
20957 Generate code for the SH4 without a floating-point unit.
20959 @item -m4-single-only
20960 @opindex m4-single-only
20961 Generate code for the SH4 with a floating-point unit that only
20962 supports single-precision arithmetic.
20966 Generate code for the SH4 assuming the floating-point unit is in
20967 single-precision mode by default.
20971 Generate code for the SH4.
20975 Generate code for SH4-100.
20977 @item -m4-100-nofpu
20978 @opindex m4-100-nofpu
20979 Generate code for SH4-100 in such a way that the
20980 floating-point unit is not used.
20982 @item -m4-100-single
20983 @opindex m4-100-single
20984 Generate code for SH4-100 assuming the floating-point unit is in
20985 single-precision mode by default.
20987 @item -m4-100-single-only
20988 @opindex m4-100-single-only
20989 Generate code for SH4-100 in such a way that no double-precision
20990 floating-point operations are used.
20994 Generate code for SH4-200.
20996 @item -m4-200-nofpu
20997 @opindex m4-200-nofpu
20998 Generate code for SH4-200 without in such a way that the
20999 floating-point unit is not used.
21001 @item -m4-200-single
21002 @opindex m4-200-single
21003 Generate code for SH4-200 assuming the floating-point unit is in
21004 single-precision mode by default.
21006 @item -m4-200-single-only
21007 @opindex m4-200-single-only
21008 Generate code for SH4-200 in such a way that no double-precision
21009 floating-point operations are used.
21013 Generate code for SH4-300.
21015 @item -m4-300-nofpu
21016 @opindex m4-300-nofpu
21017 Generate code for SH4-300 without in such a way that the
21018 floating-point unit is not used.
21020 @item -m4-300-single
21021 @opindex m4-300-single
21022 Generate code for SH4-300 in such a way that no double-precision
21023 floating-point operations are used.
21025 @item -m4-300-single-only
21026 @opindex m4-300-single-only
21027 Generate code for SH4-300 in such a way that no double-precision
21028 floating-point operations are used.
21032 Generate code for SH4-340 (no MMU, no FPU).
21036 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
21041 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
21042 floating-point unit is not used.
21044 @item -m4a-single-only
21045 @opindex m4a-single-only
21046 Generate code for the SH4a, in such a way that no double-precision
21047 floating-point operations are used.
21050 @opindex m4a-single
21051 Generate code for the SH4a assuming the floating-point unit is in
21052 single-precision mode by default.
21056 Generate code for the SH4a.
21060 Same as @option{-m4a-nofpu}, except that it implicitly passes
21061 @option{-dsp} to the assembler. GCC doesn't generate any DSP
21062 instructions at the moment.
21065 @opindex m5-32media
21066 Generate 32-bit code for SHmedia.
21068 @item -m5-32media-nofpu
21069 @opindex m5-32media-nofpu
21070 Generate 32-bit code for SHmedia in such a way that the
21071 floating-point unit is not used.
21074 @opindex m5-64media
21075 Generate 64-bit code for SHmedia.
21077 @item -m5-64media-nofpu
21078 @opindex m5-64media-nofpu
21079 Generate 64-bit code for SHmedia in such a way that the
21080 floating-point unit is not used.
21083 @opindex m5-compact
21084 Generate code for SHcompact.
21086 @item -m5-compact-nofpu
21087 @opindex m5-compact-nofpu
21088 Generate code for SHcompact in such a way that the
21089 floating-point unit is not used.
21093 Compile code for the processor in big-endian mode.
21097 Compile code for the processor in little-endian mode.
21101 Align doubles at 64-bit boundaries. Note that this changes the calling
21102 conventions, and thus some functions from the standard C library do
21103 not work unless you recompile it first with @option{-mdalign}.
21107 Shorten some address references at link time, when possible; uses the
21108 linker option @option{-relax}.
21112 Use 32-bit offsets in @code{switch} tables. The default is to use
21117 Enable the use of bit manipulation instructions on SH2A.
21121 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
21122 alignment constraints.
21126 Comply with the calling conventions defined by Renesas.
21129 @opindex mno-renesas
21130 Comply with the calling conventions defined for GCC before the Renesas
21131 conventions were available. This option is the default for all
21132 targets of the SH toolchain.
21135 @opindex mnomacsave
21136 Mark the @code{MAC} register as call-clobbered, even if
21137 @option{-mrenesas} is given.
21143 Control the IEEE compliance of floating-point comparisons, which affects the
21144 handling of cases where the result of a comparison is unordered. By default
21145 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
21146 enabled @option{-mno-ieee} is implicitly set, which results in faster
21147 floating-point greater-equal and less-equal comparisons. The implcit settings
21148 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
21150 @item -minline-ic_invalidate
21151 @opindex minline-ic_invalidate
21152 Inline code to invalidate instruction cache entries after setting up
21153 nested function trampolines.
21154 This option has no effect if @option{-musermode} is in effect and the selected
21155 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
21157 If the selected code generation option does not allow the use of the @code{icbi}
21158 instruction, and @option{-musermode} is not in effect, the inlined code
21159 manipulates the instruction cache address array directly with an associative
21160 write. This not only requires privileged mode at run time, but it also
21161 fails if the cache line had been mapped via the TLB and has become unmapped.
21165 Dump instruction size and location in the assembly code.
21168 @opindex mpadstruct
21169 This option is deprecated. It pads structures to multiple of 4 bytes,
21170 which is incompatible with the SH ABI@.
21172 @item -matomic-model=@var{model}
21173 @opindex matomic-model=@var{model}
21174 Sets the model of atomic operations and additional parameters as a comma
21175 separated list. For details on the atomic built-in functions see
21176 @ref{__atomic Builtins}. The following models and parameters are supported:
21181 Disable compiler generated atomic sequences and emit library calls for atomic
21182 operations. This is the default if the target is not @code{sh*-*-linux*}.
21185 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
21186 built-in functions. The generated atomic sequences require additional support
21187 from the interrupt/exception handling code of the system and are only suitable
21188 for SH3* and SH4* single-core systems. This option is enabled by default when
21189 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
21190 this option will also partially utilize the hardware atomic instructions
21191 @code{movli.l} and @code{movco.l} to create more efficient code, unless
21192 @samp{strict} is specified.
21195 Generate software atomic sequences that use a variable in the thread control
21196 block. This is a variation of the gUSA sequences which can also be used on
21197 SH1* and SH2* targets. The generated atomic sequences require additional
21198 support from the interrupt/exception handling code of the system and are only
21199 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
21200 parameter has to be specified as well.
21203 Generate software atomic sequences that temporarily disable interrupts by
21204 setting @code{SR.IMASK = 1111}. This model works only when the program runs
21205 in privileged mode and is only suitable for single-core systems. Additional
21206 support from the interrupt/exception handling code of the system is not
21207 required. This model is enabled by default when the target is
21208 @code{sh*-*-linux*} and SH1* or SH2*.
21211 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
21212 instructions only. This is only available on SH4A and is suitable for
21213 multi-core systems. Since the hardware instructions support only 32 bit atomic
21214 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
21215 Code compiled with this option will also be compatible with other software
21216 atomic model interrupt/exception handling systems if executed on an SH4A
21217 system. Additional support from the interrupt/exception handling code of the
21218 system is not required for this model.
21221 This parameter specifies the offset in bytes of the variable in the thread
21222 control block structure that should be used by the generated atomic sequences
21223 when the @samp{soft-tcb} model has been selected. For other models this
21224 parameter is ignored. The specified value must be an integer multiple of four
21225 and in the range 0-1020.
21228 This parameter prevents mixed usage of multiple atomic models, even though they
21229 would be compatible, and will make the compiler generate atomic sequences of the
21230 specified model only.
21236 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
21237 Notice that depending on the particular hardware and software configuration
21238 this can degrade overall performance due to the operand cache line flushes
21239 that are implied by the @code{tas.b} instruction. On multi-core SH4A
21240 processors the @code{tas.b} instruction must be used with caution since it
21241 can result in data corruption for certain cache configurations.
21244 @opindex mprefergot
21245 When generating position-independent code, emit function calls using
21246 the Global Offset Table instead of the Procedure Linkage Table.
21249 @itemx -mno-usermode
21251 @opindex mno-usermode
21252 Don't allow (allow) the compiler generating privileged mode code. Specifying
21253 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
21254 inlined code would not work in user mode. @option{-musermode} is the default
21255 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
21256 @option{-musermode} has no effect, since there is no user mode.
21258 @item -multcost=@var{number}
21259 @opindex multcost=@var{number}
21260 Set the cost to assume for a multiply insn.
21262 @item -mdiv=@var{strategy}
21263 @opindex mdiv=@var{strategy}
21264 Set the division strategy to be used for integer division operations.
21265 For SHmedia @var{strategy} can be one of:
21270 Performs the operation in floating point. This has a very high latency,
21271 but needs only a few instructions, so it might be a good choice if
21272 your code has enough easily-exploitable ILP to allow the compiler to
21273 schedule the floating-point instructions together with other instructions.
21274 Division by zero causes a floating-point exception.
21277 Uses integer operations to calculate the inverse of the divisor,
21278 and then multiplies the dividend with the inverse. This strategy allows
21279 CSE and hoisting of the inverse calculation. Division by zero calculates
21280 an unspecified result, but does not trap.
21283 A variant of @samp{inv} where, if no CSE or hoisting opportunities
21284 have been found, or if the entire operation has been hoisted to the same
21285 place, the last stages of the inverse calculation are intertwined with the
21286 final multiply to reduce the overall latency, at the expense of using a few
21287 more instructions, and thus offering fewer scheduling opportunities with
21291 Calls a library function that usually implements the @samp{inv:minlat}
21293 This gives high code density for @code{m5-*media-nofpu} compilations.
21296 Uses a different entry point of the same library function, where it
21297 assumes that a pointer to a lookup table has already been set up, which
21298 exposes the pointer load to CSE and code hoisting optimizations.
21303 Use the @samp{inv} algorithm for initial
21304 code generation, but if the code stays unoptimized, revert to the @samp{call},
21305 @samp{call2}, or @samp{fp} strategies, respectively. Note that the
21306 potentially-trapping side effect of division by zero is carried by a
21307 separate instruction, so it is possible that all the integer instructions
21308 are hoisted out, but the marker for the side effect stays where it is.
21309 A recombination to floating-point operations or a call is not possible
21314 Variants of the @samp{inv:minlat} strategy. In the case
21315 that the inverse calculation is not separated from the multiply, they speed
21316 up division where the dividend fits into 20 bits (plus sign where applicable)
21317 by inserting a test to skip a number of operations in this case; this test
21318 slows down the case of larger dividends. @samp{inv20u} assumes the case of a such
21319 a small dividend to be unlikely, and @samp{inv20l} assumes it to be likely.
21323 For targets other than SHmedia @var{strategy} can be one of:
21328 Calls a library function that uses the single-step division instruction
21329 @code{div1} to perform the operation. Division by zero calculates an
21330 unspecified result and does not trap. This is the default except for SH4,
21331 SH2A and SHcompact.
21334 Calls a library function that performs the operation in double precision
21335 floating point. Division by zero causes a floating-point exception. This is
21336 the default for SHcompact with FPU. Specifying this for targets that do not
21337 have a double precision FPU will default to @code{call-div1}.
21340 Calls a library function that uses a lookup table for small divisors and
21341 the @code{div1} instruction with case distinction for larger divisors. Division
21342 by zero calculates an unspecified result and does not trap. This is the default
21343 for SH4. Specifying this for targets that do not have dynamic shift
21344 instructions will default to @code{call-div1}.
21348 When a division strategy has not been specified the default strategy will be
21349 selected based on the current target. For SH2A the default strategy is to
21350 use the @code{divs} and @code{divu} instructions instead of library function
21353 @item -maccumulate-outgoing-args
21354 @opindex maccumulate-outgoing-args
21355 Reserve space once for outgoing arguments in the function prologue rather
21356 than around each call. Generally beneficial for performance and size. Also
21357 needed for unwinding to avoid changing the stack frame around conditional code.
21359 @item -mdivsi3_libfunc=@var{name}
21360 @opindex mdivsi3_libfunc=@var{name}
21361 Set the name of the library function used for 32-bit signed division to
21363 This only affects the name used in the @samp{call} and @samp{inv:call}
21364 division strategies, and the compiler still expects the same
21365 sets of input/output/clobbered registers as if this option were not present.
21367 @item -mfixed-range=@var{register-range}
21368 @opindex mfixed-range
21369 Generate code treating the given register range as fixed registers.
21370 A fixed register is one that the register allocator can not use. This is
21371 useful when compiling kernel code. A register range is specified as
21372 two registers separated by a dash. Multiple register ranges can be
21373 specified separated by a comma.
21375 @item -mindexed-addressing
21376 @opindex mindexed-addressing
21377 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
21378 This is only safe if the hardware and/or OS implement 32-bit wrap-around
21379 semantics for the indexed addressing mode. The architecture allows the
21380 implementation of processors with 64-bit MMU, which the OS could use to
21381 get 32-bit addressing, but since no current hardware implementation supports
21382 this or any other way to make the indexed addressing mode safe to use in
21383 the 32-bit ABI, the default is @option{-mno-indexed-addressing}.
21385 @item -mgettrcost=@var{number}
21386 @opindex mgettrcost=@var{number}
21387 Set the cost assumed for the @code{gettr} instruction to @var{number}.
21388 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
21392 Assume @code{pt*} instructions won't trap. This generally generates
21393 better-scheduled code, but is unsafe on current hardware.
21394 The current architecture
21395 definition says that @code{ptabs} and @code{ptrel} trap when the target
21397 This has the unintentional effect of making it unsafe to schedule these
21398 instructions before a branch, or hoist them out of a loop. For example,
21399 @code{__do_global_ctors}, a part of @file{libgcc}
21400 that runs constructors at program
21401 startup, calls functions in a list which is delimited by @minus{}1. With the
21402 @option{-mpt-fixed} option, the @code{ptabs} is done before testing against @minus{}1.
21403 That means that all the constructors run a bit more quickly, but when
21404 the loop comes to the end of the list, the program crashes because @code{ptabs}
21405 loads @minus{}1 into a target register.
21407 Since this option is unsafe for any
21408 hardware implementing the current architecture specification, the default
21409 is @option{-mno-pt-fixed}. Unless specified explicitly with
21410 @option{-mgettrcost}, @option{-mno-pt-fixed} also implies @option{-mgettrcost=100};
21411 this deters register allocation from using target registers for storing
21414 @item -minvalid-symbols
21415 @opindex minvalid-symbols
21416 Assume symbols might be invalid. Ordinary function symbols generated by
21417 the compiler are always valid to load with
21418 @code{movi}/@code{shori}/@code{ptabs} or
21419 @code{movi}/@code{shori}/@code{ptrel},
21420 but with assembler and/or linker tricks it is possible
21421 to generate symbols that cause @code{ptabs} or @code{ptrel} to trap.
21422 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
21423 It prevents cross-basic-block CSE, hoisting and most scheduling
21424 of symbol loads. The default is @option{-mno-invalid-symbols}.
21426 @item -mbranch-cost=@var{num}
21427 @opindex mbranch-cost=@var{num}
21428 Assume @var{num} to be the cost for a branch instruction. Higher numbers
21429 make the compiler try to generate more branch-free code if possible.
21430 If not specified the value is selected depending on the processor type that
21431 is being compiled for.
21434 @itemx -mno-zdcbranch
21435 @opindex mzdcbranch
21436 @opindex mno-zdcbranch
21437 Assume (do not assume) that zero displacement conditional branch instructions
21438 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
21439 compiler will try to prefer zero displacement branch code sequences. This is
21440 enabled by default when generating code for SH4 and SH4A. It can be explicitly
21441 disabled by specifying @option{-mno-zdcbranch}.
21444 @itemx -mno-fused-madd
21445 @opindex mfused-madd
21446 @opindex mno-fused-madd
21447 Generate code that uses (does not use) the floating-point multiply and
21448 accumulate instructions. These instructions are generated by default
21449 if hardware floating point is used. The machine-dependent
21450 @option{-mfused-madd} option is now mapped to the machine-independent
21451 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
21452 mapped to @option{-ffp-contract=off}.
21458 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
21459 and cosine approximations. The option @code{-mfsca} must be used in
21460 combination with @code{-funsafe-math-optimizations}. It is enabled by default
21461 when generating code for SH4A. Using @code{-mno-fsca} disables sine and cosine
21462 approximations even if @code{-funsafe-math-optimizations} is in effect.
21468 Allow or disallow the compiler to emit the @code{fsrra} instruction for
21469 reciprocal square root approximations. The option @code{-mfsrra} must be used
21470 in combination with @code{-funsafe-math-optimizations} and
21471 @code{-ffinite-math-only}. It is enabled by default when generating code for
21472 SH4A. Using @code{-mno-fsrra} disables reciprocal square root approximations
21473 even if @code{-funsafe-math-optimizations} and @code{-ffinite-math-only} are
21476 @item -mpretend-cmove
21477 @opindex mpretend-cmove
21478 Prefer zero-displacement conditional branches for conditional move instruction
21479 patterns. This can result in faster code on the SH4 processor.
21483 @node Solaris 2 Options
21484 @subsection Solaris 2 Options
21485 @cindex Solaris 2 options
21487 These @samp{-m} options are supported on Solaris 2:
21490 @item -mclear-hwcap
21491 @opindex mclear-hwcap
21492 @option{-mclear-hwcap} tells the compiler to remove the hardware
21493 capabilities generated by the Solaris assembler. This is only necessary
21494 when object files use ISA extensions not supported by the current
21495 machine, but check at runtime whether or not to use them.
21497 @item -mimpure-text
21498 @opindex mimpure-text
21499 @option{-mimpure-text}, used in addition to @option{-shared}, tells
21500 the compiler to not pass @option{-z text} to the linker when linking a
21501 shared object. Using this option, you can link position-dependent
21502 code into a shared object.
21504 @option{-mimpure-text} suppresses the ``relocations remain against
21505 allocatable but non-writable sections'' linker error message.
21506 However, the necessary relocations trigger copy-on-write, and the
21507 shared object is not actually shared across processes. Instead of
21508 using @option{-mimpure-text}, you should compile all source code with
21509 @option{-fpic} or @option{-fPIC}.
21513 These switches are supported in addition to the above on Solaris 2:
21518 Add support for multithreading using the POSIX threads library. This
21519 option sets flags for both the preprocessor and linker. This option does
21520 not affect the thread safety of object code produced by the compiler or
21521 that of libraries supplied with it.
21525 This is a synonym for @option{-pthreads}.
21528 @node SPARC Options
21529 @subsection SPARC Options
21530 @cindex SPARC options
21532 These @samp{-m} options are supported on the SPARC:
21535 @item -mno-app-regs
21537 @opindex mno-app-regs
21539 Specify @option{-mapp-regs} to generate output using the global registers
21540 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
21541 global register 1, each global register 2 through 4 is then treated as an
21542 allocable register that is clobbered by function calls. This is the default.
21544 To be fully SVR4 ABI-compliant at the cost of some performance loss,
21545 specify @option{-mno-app-regs}. You should compile libraries and system
21546 software with this option.
21552 With @option{-mflat}, the compiler does not generate save/restore instructions
21553 and uses a ``flat'' or single register window model. This model is compatible
21554 with the regular register window model. The local registers and the input
21555 registers (0--5) are still treated as ``call-saved'' registers and are
21556 saved on the stack as needed.
21558 With @option{-mno-flat} (the default), the compiler generates save/restore
21559 instructions (except for leaf functions). This is the normal operating mode.
21562 @itemx -mhard-float
21564 @opindex mhard-float
21565 Generate output containing floating-point instructions. This is the
21569 @itemx -msoft-float
21571 @opindex msoft-float
21572 Generate output containing library calls for floating point.
21573 @strong{Warning:} the requisite libraries are not available for all SPARC
21574 targets. Normally the facilities of the machine's usual C compiler are
21575 used, but this cannot be done directly in cross-compilation. You must make
21576 your own arrangements to provide suitable library functions for
21577 cross-compilation. The embedded targets @samp{sparc-*-aout} and
21578 @samp{sparclite-*-*} do provide software floating-point support.
21580 @option{-msoft-float} changes the calling convention in the output file;
21581 therefore, it is only useful if you compile @emph{all} of a program with
21582 this option. In particular, you need to compile @file{libgcc.a}, the
21583 library that comes with GCC, with @option{-msoft-float} in order for
21586 @item -mhard-quad-float
21587 @opindex mhard-quad-float
21588 Generate output containing quad-word (long double) floating-point
21591 @item -msoft-quad-float
21592 @opindex msoft-quad-float
21593 Generate output containing library calls for quad-word (long double)
21594 floating-point instructions. The functions called are those specified
21595 in the SPARC ABI@. This is the default.
21597 As of this writing, there are no SPARC implementations that have hardware
21598 support for the quad-word floating-point instructions. They all invoke
21599 a trap handler for one of these instructions, and then the trap handler
21600 emulates the effect of the instruction. Because of the trap handler overhead,
21601 this is much slower than calling the ABI library routines. Thus the
21602 @option{-msoft-quad-float} option is the default.
21604 @item -mno-unaligned-doubles
21605 @itemx -munaligned-doubles
21606 @opindex mno-unaligned-doubles
21607 @opindex munaligned-doubles
21608 Assume that doubles have 8-byte alignment. This is the default.
21610 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
21611 alignment only if they are contained in another type, or if they have an
21612 absolute address. Otherwise, it assumes they have 4-byte alignment.
21613 Specifying this option avoids some rare compatibility problems with code
21614 generated by other compilers. It is not the default because it results
21615 in a performance loss, especially for floating-point code.
21618 @itemx -mno-user-mode
21619 @opindex muser-mode
21620 @opindex mno-user-mode
21621 Do not generate code that can only run in supervisor mode. This is relevant
21622 only for the @code{casa} instruction emitted for the LEON3 processor. The
21623 default is @option{-mno-user-mode}.
21625 @item -mno-faster-structs
21626 @itemx -mfaster-structs
21627 @opindex mno-faster-structs
21628 @opindex mfaster-structs
21629 With @option{-mfaster-structs}, the compiler assumes that structures
21630 should have 8-byte alignment. This enables the use of pairs of
21631 @code{ldd} and @code{std} instructions for copies in structure
21632 assignment, in place of twice as many @code{ld} and @code{st} pairs.
21633 However, the use of this changed alignment directly violates the SPARC
21634 ABI@. Thus, it's intended only for use on targets where the developer
21635 acknowledges that their resulting code is not directly in line with
21636 the rules of the ABI@.
21638 @item -mcpu=@var{cpu_type}
21640 Set the instruction set, register set, and instruction scheduling parameters
21641 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
21642 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
21643 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
21644 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
21645 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21646 @samp{niagara3} and @samp{niagara4}.
21648 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
21649 which selects the best architecture option for the host processor.
21650 @option{-mcpu=native} has no effect if GCC does not recognize
21653 Default instruction scheduling parameters are used for values that select
21654 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
21655 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
21657 Here is a list of each supported architecture and their supported
21665 supersparc, hypersparc, leon, leon3
21668 f930, f934, sparclite86x
21674 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
21677 By default (unless configured otherwise), GCC generates code for the V7
21678 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
21679 additionally optimizes it for the Cypress CY7C602 chip, as used in the
21680 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
21681 SPARCStation 1, 2, IPX etc.
21683 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
21684 architecture. The only difference from V7 code is that the compiler emits
21685 the integer multiply and integer divide instructions which exist in SPARC-V8
21686 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
21687 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
21690 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
21691 the SPARC architecture. This adds the integer multiply, integer divide step
21692 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
21693 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
21694 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
21695 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
21696 MB86934 chip, which is the more recent SPARClite with FPU@.
21698 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
21699 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
21700 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
21701 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
21702 optimizes it for the TEMIC SPARClet chip.
21704 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
21705 architecture. This adds 64-bit integer and floating-point move instructions,
21706 3 additional floating-point condition code registers and conditional move
21707 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
21708 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
21709 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
21710 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
21711 @option{-mcpu=niagara}, the compiler additionally optimizes it for
21712 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
21713 additionally optimizes it for Sun UltraSPARC T2 chips. With
21714 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
21715 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
21716 additionally optimizes it for Sun UltraSPARC T4 chips.
21718 @item -mtune=@var{cpu_type}
21720 Set the instruction scheduling parameters for machine type
21721 @var{cpu_type}, but do not set the instruction set or register set that the
21722 option @option{-mcpu=@var{cpu_type}} does.
21724 The same values for @option{-mcpu=@var{cpu_type}} can be used for
21725 @option{-mtune=@var{cpu_type}}, but the only useful values are those
21726 that select a particular CPU implementation. Those are @samp{cypress},
21727 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{leon3},
21728 @samp{leon3v7}, @samp{f930}, @samp{f934}, @samp{sparclite86x}, @samp{tsc701},
21729 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21730 @samp{niagara3} and @samp{niagara4}. With native Solaris and GNU/Linux
21731 toolchains, @samp{native} can also be used.
21736 @opindex mno-v8plus
21737 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
21738 difference from the V8 ABI is that the global and out registers are
21739 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
21740 mode for all SPARC-V9 processors.
21746 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
21747 Visual Instruction Set extensions. The default is @option{-mno-vis}.
21753 With @option{-mvis2}, GCC generates code that takes advantage of
21754 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
21755 default is @option{-mvis2} when targeting a cpu that supports such
21756 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
21757 also sets @option{-mvis}.
21763 With @option{-mvis3}, GCC generates code that takes advantage of
21764 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
21765 default is @option{-mvis3} when targeting a cpu that supports such
21766 instructions, such as niagara-3 and later. Setting @option{-mvis3}
21767 also sets @option{-mvis2} and @option{-mvis}.
21772 @opindex mno-cbcond
21773 With @option{-mcbcond}, GCC generates code that takes advantage of
21774 compare-and-branch instructions, as defined in the Sparc Architecture 2011.
21775 The default is @option{-mcbcond} when targeting a cpu that supports such
21776 instructions, such as niagara-4 and later.
21782 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
21783 population count instruction. The default is @option{-mpopc}
21784 when targeting a cpu that supports such instructions, such as Niagara-2 and
21791 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
21792 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
21793 when targeting a cpu that supports such instructions, such as Niagara-3 and
21797 @opindex mfix-at697f
21798 Enable the documented workaround for the single erratum of the Atmel AT697F
21799 processor (which corresponds to erratum #13 of the AT697E processor).
21802 @opindex mfix-ut699
21803 Enable the documented workarounds for the floating-point errata and the data
21804 cache nullify errata of the UT699 processor.
21807 These @samp{-m} options are supported in addition to the above
21808 on SPARC-V9 processors in 64-bit environments:
21815 Generate code for a 32-bit or 64-bit environment.
21816 The 32-bit environment sets int, long and pointer to 32 bits.
21817 The 64-bit environment sets int to 32 bits and long and pointer
21820 @item -mcmodel=@var{which}
21822 Set the code model to one of
21826 The Medium/Low code model: 64-bit addresses, programs
21827 must be linked in the low 32 bits of memory. Programs can be statically
21828 or dynamically linked.
21831 The Medium/Middle code model: 64-bit addresses, programs
21832 must be linked in the low 44 bits of memory, the text and data segments must
21833 be less than 2GB in size and the data segment must be located within 2GB of
21837 The Medium/Anywhere code model: 64-bit addresses, programs
21838 may be linked anywhere in memory, the text and data segments must be less
21839 than 2GB in size and the data segment must be located within 2GB of the
21843 The Medium/Anywhere code model for embedded systems:
21844 64-bit addresses, the text and data segments must be less than 2GB in
21845 size, both starting anywhere in memory (determined at link time). The
21846 global register %g4 points to the base of the data segment. Programs
21847 are statically linked and PIC is not supported.
21850 @item -mmemory-model=@var{mem-model}
21851 @opindex mmemory-model
21852 Set the memory model in force on the processor to one of
21856 The default memory model for the processor and operating system.
21859 Relaxed Memory Order
21862 Partial Store Order
21868 Sequential Consistency
21871 These memory models are formally defined in Appendix D of the Sparc V9
21872 architecture manual, as set in the processor's @code{PSTATE.MM} field.
21875 @itemx -mno-stack-bias
21876 @opindex mstack-bias
21877 @opindex mno-stack-bias
21878 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
21879 frame pointer if present, are offset by @minus{}2047 which must be added back
21880 when making stack frame references. This is the default in 64-bit mode.
21881 Otherwise, assume no such offset is present.
21885 @subsection SPU Options
21886 @cindex SPU options
21888 These @samp{-m} options are supported on the SPU:
21892 @itemx -merror-reloc
21893 @opindex mwarn-reloc
21894 @opindex merror-reloc
21896 The loader for SPU does not handle dynamic relocations. By default, GCC
21897 gives an error when it generates code that requires a dynamic
21898 relocation. @option{-mno-error-reloc} disables the error,
21899 @option{-mwarn-reloc} generates a warning instead.
21902 @itemx -munsafe-dma
21904 @opindex munsafe-dma
21906 Instructions that initiate or test completion of DMA must not be
21907 reordered with respect to loads and stores of the memory that is being
21909 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
21910 memory accesses, but that can lead to inefficient code in places where the
21911 memory is known to not change. Rather than mark the memory as volatile,
21912 you can use @option{-msafe-dma} to tell the compiler to treat
21913 the DMA instructions as potentially affecting all memory.
21915 @item -mbranch-hints
21916 @opindex mbranch-hints
21918 By default, GCC generates a branch hint instruction to avoid
21919 pipeline stalls for always-taken or probably-taken branches. A hint
21920 is not generated closer than 8 instructions away from its branch.
21921 There is little reason to disable them, except for debugging purposes,
21922 or to make an object a little bit smaller.
21926 @opindex msmall-mem
21927 @opindex mlarge-mem
21929 By default, GCC generates code assuming that addresses are never larger
21930 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
21931 a full 32-bit address.
21936 By default, GCC links against startup code that assumes the SPU-style
21937 main function interface (which has an unconventional parameter list).
21938 With @option{-mstdmain}, GCC links your program against startup
21939 code that assumes a C99-style interface to @code{main}, including a
21940 local copy of @code{argv} strings.
21942 @item -mfixed-range=@var{register-range}
21943 @opindex mfixed-range
21944 Generate code treating the given register range as fixed registers.
21945 A fixed register is one that the register allocator cannot use. This is
21946 useful when compiling kernel code. A register range is specified as
21947 two registers separated by a dash. Multiple register ranges can be
21948 specified separated by a comma.
21954 Compile code assuming that pointers to the PPU address space accessed
21955 via the @code{__ea} named address space qualifier are either 32 or 64
21956 bits wide. The default is 32 bits. As this is an ABI-changing option,
21957 all object code in an executable must be compiled with the same setting.
21959 @item -maddress-space-conversion
21960 @itemx -mno-address-space-conversion
21961 @opindex maddress-space-conversion
21962 @opindex mno-address-space-conversion
21963 Allow/disallow treating the @code{__ea} address space as superset
21964 of the generic address space. This enables explicit type casts
21965 between @code{__ea} and generic pointer as well as implicit
21966 conversions of generic pointers to @code{__ea} pointers. The
21967 default is to allow address space pointer conversions.
21969 @item -mcache-size=@var{cache-size}
21970 @opindex mcache-size
21971 This option controls the version of libgcc that the compiler links to an
21972 executable and selects a software-managed cache for accessing variables
21973 in the @code{__ea} address space with a particular cache size. Possible
21974 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
21975 and @samp{128}. The default cache size is 64KB.
21977 @item -matomic-updates
21978 @itemx -mno-atomic-updates
21979 @opindex matomic-updates
21980 @opindex mno-atomic-updates
21981 This option controls the version of libgcc that the compiler links to an
21982 executable and selects whether atomic updates to the software-managed
21983 cache of PPU-side variables are used. If you use atomic updates, changes
21984 to a PPU variable from SPU code using the @code{__ea} named address space
21985 qualifier do not interfere with changes to other PPU variables residing
21986 in the same cache line from PPU code. If you do not use atomic updates,
21987 such interference may occur; however, writing back cache lines is
21988 more efficient. The default behavior is to use atomic updates.
21991 @itemx -mdual-nops=@var{n}
21992 @opindex mdual-nops
21993 By default, GCC inserts nops to increase dual issue when it expects
21994 it to increase performance. @var{n} can be a value from 0 to 10. A
21995 smaller @var{n} inserts fewer nops. 10 is the default, 0 is the
21996 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
21998 @item -mhint-max-nops=@var{n}
21999 @opindex mhint-max-nops
22000 Maximum number of nops to insert for a branch hint. A branch hint must
22001 be at least 8 instructions away from the branch it is affecting. GCC
22002 inserts up to @var{n} nops to enforce this, otherwise it does not
22003 generate the branch hint.
22005 @item -mhint-max-distance=@var{n}
22006 @opindex mhint-max-distance
22007 The encoding of the branch hint instruction limits the hint to be within
22008 256 instructions of the branch it is affecting. By default, GCC makes
22009 sure it is within 125.
22012 @opindex msafe-hints
22013 Work around a hardware bug that causes the SPU to stall indefinitely.
22014 By default, GCC inserts the @code{hbrp} instruction to make sure
22015 this stall won't happen.
22019 @node System V Options
22020 @subsection Options for System V
22022 These additional options are available on System V Release 4 for
22023 compatibility with other compilers on those systems:
22028 Create a shared object.
22029 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
22033 Identify the versions of each tool used by the compiler, in a
22034 @code{.ident} assembler directive in the output.
22038 Refrain from adding @code{.ident} directives to the output file (this is
22041 @item -YP,@var{dirs}
22043 Search the directories @var{dirs}, and no others, for libraries
22044 specified with @option{-l}.
22046 @item -Ym,@var{dir}
22048 Look in the directory @var{dir} to find the M4 preprocessor.
22049 The assembler uses this option.
22050 @c This is supposed to go with a -Yd for predefined M4 macro files, but
22051 @c the generic assembler that comes with Solaris takes just -Ym.
22054 @node TILE-Gx Options
22055 @subsection TILE-Gx Options
22056 @cindex TILE-Gx options
22058 These @samp{-m} options are supported on the TILE-Gx:
22061 @item -mcmodel=small
22062 @opindex mcmodel=small
22063 Generate code for the small model. The distance for direct calls is
22064 limited to 500M in either direction. PC-relative addresses are 32
22065 bits. Absolute addresses support the full address range.
22067 @item -mcmodel=large
22068 @opindex mcmodel=large
22069 Generate code for the large model. There is no limitation on call
22070 distance, pc-relative addresses, or absolute addresses.
22072 @item -mcpu=@var{name}
22074 Selects the type of CPU to be targeted. Currently the only supported
22075 type is @samp{tilegx}.
22081 Generate code for a 32-bit or 64-bit environment. The 32-bit
22082 environment sets int, long, and pointer to 32 bits. The 64-bit
22083 environment sets int to 32 bits and long and pointer to 64 bits.
22086 @itemx -mlittle-endian
22087 @opindex mbig-endian
22088 @opindex mlittle-endian
22089 Generate code in big/little endian mode, respectively.
22092 @node TILEPro Options
22093 @subsection TILEPro Options
22094 @cindex TILEPro options
22096 These @samp{-m} options are supported on the TILEPro:
22099 @item -mcpu=@var{name}
22101 Selects the type of CPU to be targeted. Currently the only supported
22102 type is @samp{tilepro}.
22106 Generate code for a 32-bit environment, which sets int, long, and
22107 pointer to 32 bits. This is the only supported behavior so the flag
22108 is essentially ignored.
22112 @subsection V850 Options
22113 @cindex V850 Options
22115 These @samp{-m} options are defined for V850 implementations:
22119 @itemx -mno-long-calls
22120 @opindex mlong-calls
22121 @opindex mno-long-calls
22122 Treat all calls as being far away (near). If calls are assumed to be
22123 far away, the compiler always loads the function's address into a
22124 register, and calls indirect through the pointer.
22130 Do not optimize (do optimize) basic blocks that use the same index
22131 pointer 4 or more times to copy pointer into the @code{ep} register, and
22132 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
22133 option is on by default if you optimize.
22135 @item -mno-prolog-function
22136 @itemx -mprolog-function
22137 @opindex mno-prolog-function
22138 @opindex mprolog-function
22139 Do not use (do use) external functions to save and restore registers
22140 at the prologue and epilogue of a function. The external functions
22141 are slower, but use less code space if more than one function saves
22142 the same number of registers. The @option{-mprolog-function} option
22143 is on by default if you optimize.
22147 Try to make the code as small as possible. At present, this just turns
22148 on the @option{-mep} and @option{-mprolog-function} options.
22150 @item -mtda=@var{n}
22152 Put static or global variables whose size is @var{n} bytes or less into
22153 the tiny data area that register @code{ep} points to. The tiny data
22154 area can hold up to 256 bytes in total (128 bytes for byte references).
22156 @item -msda=@var{n}
22158 Put static or global variables whose size is @var{n} bytes or less into
22159 the small data area that register @code{gp} points to. The small data
22160 area can hold up to 64 kilobytes.
22162 @item -mzda=@var{n}
22164 Put static or global variables whose size is @var{n} bytes or less into
22165 the first 32 kilobytes of memory.
22169 Specify that the target processor is the V850.
22173 Specify that the target processor is the V850E3V5. The preprocessor
22174 constant @samp{__v850e3v5__} is defined if this option is used.
22178 Specify that the target processor is the V850E3V5. This is an alias for
22179 the @option{-mv850e3v5} option.
22183 Specify that the target processor is the V850E2V3. The preprocessor
22184 constant @samp{__v850e2v3__} is defined if this option is used.
22188 Specify that the target processor is the V850E2. The preprocessor
22189 constant @samp{__v850e2__} is defined if this option is used.
22193 Specify that the target processor is the V850E1. The preprocessor
22194 constants @samp{__v850e1__} and @samp{__v850e__} are defined if
22195 this option is used.
22199 Specify that the target processor is the V850ES. This is an alias for
22200 the @option{-mv850e1} option.
22204 Specify that the target processor is the V850E@. The preprocessor
22205 constant @samp{__v850e__} is defined if this option is used.
22207 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
22208 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
22209 are defined then a default target processor is chosen and the
22210 relevant @samp{__v850*__} preprocessor constant is defined.
22212 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
22213 defined, regardless of which processor variant is the target.
22215 @item -mdisable-callt
22216 @itemx -mno-disable-callt
22217 @opindex mdisable-callt
22218 @opindex mno-disable-callt
22219 This option suppresses generation of the @code{CALLT} instruction for the
22220 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
22223 This option is enabled by default when the RH850 ABI is
22224 in use (see @option{-mrh850-abi}), and disabled by default when the
22225 GCC ABI is in use. If @code{CALLT} instructions are being generated
22226 then the C preprocessor symbol @code{__V850_CALLT__} will be defined.
22232 Pass on (or do not pass on) the @option{-mrelax} command line option
22236 @itemx -mno-long-jumps
22237 @opindex mlong-jumps
22238 @opindex mno-long-jumps
22239 Disable (or re-enable) the generation of PC-relative jump instructions.
22242 @itemx -mhard-float
22243 @opindex msoft-float
22244 @opindex mhard-float
22245 Disable (or re-enable) the generation of hardware floating point
22246 instructions. This option is only significant when the target
22247 architecture is @samp{V850E2V3} or higher. If hardware floating point
22248 instructions are being generated then the C preprocessor symbol
22249 @code{__FPU_OK__} will be defined, otherwise the symbol
22250 @code{__NO_FPU__} will be defined.
22254 Enables the use of the e3v5 LOOP instruction. The use of this
22255 instruction is not enabled by default when the e3v5 architecture is
22256 selected because its use is still experimental.
22260 @opindex mrh850-abi
22262 Enables support for the RH850 version of the V850 ABI. This is the
22263 default. With this version of the ABI the following rules apply:
22267 Integer sized structures and unions are returned via a memory pointer
22268 rather than a register.
22271 Large structures and unions (more than 8 bytes in size) are passed by
22275 Functions are aligned to 16-bit boundaries.
22278 The @option{-m8byte-align} command line option is supported.
22281 The @option{-mdisable-callt} command line option is enabled by
22282 default. The @option{-mno-disable-callt} command line option is not
22286 When this version of the ABI is enabled the C preprocessor symbol
22287 @code{__V850_RH850_ABI__} is defined.
22291 Enables support for the old GCC version of the V850 ABI. With this
22292 version of the ABI the following rules apply:
22296 Integer sized structures and unions are returned in register @code{r10}.
22299 Large structures and unions (more than 8 bytes in size) are passed by
22303 Functions are aligned to 32-bit boundaries, unless optimizing for
22307 The @option{-m8byte-align} command line option is not supported.
22310 The @option{-mdisable-callt} command line option is supported but not
22311 enabled by default.
22314 When this version of the ABI is enabled the C preprocessor symbol
22315 @code{__V850_GCC_ABI__} is defined.
22317 @item -m8byte-align
22318 @itemx -mno-8byte-align
22319 @opindex m8byte-align
22320 @opindex mno-8byte-align
22321 Enables support for @code{doubles} and @code{long long} types to be
22322 aligned on 8-byte boundaries. The default is to restrict the
22323 alignment of all objects to at most 4-bytes. When
22324 @option{-m8byte-align} is in effect the C preprocessor symbol
22325 @code{__V850_8BYTE_ALIGN__} will be defined.
22328 @opindex mbig-switch
22329 Generate code suitable for big switch tables. Use this option only if
22330 the assembler/linker complain about out of range branches within a switch
22335 This option causes r2 and r5 to be used in the code generated by
22336 the compiler. This setting is the default.
22338 @item -mno-app-regs
22339 @opindex mno-app-regs
22340 This option causes r2 and r5 to be treated as fixed registers.
22345 @subsection VAX Options
22346 @cindex VAX options
22348 These @samp{-m} options are defined for the VAX:
22353 Do not output certain jump instructions (@code{aobleq} and so on)
22354 that the Unix assembler for the VAX cannot handle across long
22359 Do output those jump instructions, on the assumption that the
22360 GNU assembler is being used.
22364 Output code for G-format floating-point numbers instead of D-format.
22368 @subsection VMS Options
22370 These @samp{-m} options are defined for the VMS implementations:
22373 @item -mvms-return-codes
22374 @opindex mvms-return-codes
22375 Return VMS condition codes from @code{main}. The default is to return POSIX-style
22376 condition (e.g.@ error) codes.
22378 @item -mdebug-main=@var{prefix}
22379 @opindex mdebug-main=@var{prefix}
22380 Flag the first routine whose name starts with @var{prefix} as the main
22381 routine for the debugger.
22385 Default to 64-bit memory allocation routines.
22387 @item -mpointer-size=@var{size}
22388 @opindex -mpointer-size=@var{size}
22389 Set the default size of pointers. Possible options for @var{size} are
22390 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
22391 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
22392 The later option disables @code{pragma pointer_size}.
22395 @node VxWorks Options
22396 @subsection VxWorks Options
22397 @cindex VxWorks Options
22399 The options in this section are defined for all VxWorks targets.
22400 Options specific to the target hardware are listed with the other
22401 options for that target.
22406 GCC can generate code for both VxWorks kernels and real time processes
22407 (RTPs). This option switches from the former to the latter. It also
22408 defines the preprocessor macro @code{__RTP__}.
22411 @opindex non-static
22412 Link an RTP executable against shared libraries rather than static
22413 libraries. The options @option{-static} and @option{-shared} can
22414 also be used for RTPs (@pxref{Link Options}); @option{-static}
22421 These options are passed down to the linker. They are defined for
22422 compatibility with Diab.
22425 @opindex Xbind-lazy
22426 Enable lazy binding of function calls. This option is equivalent to
22427 @option{-Wl,-z,now} and is defined for compatibility with Diab.
22431 Disable lazy binding of function calls. This option is the default and
22432 is defined for compatibility with Diab.
22435 @node x86-64 Options
22436 @subsection x86-64 Options
22437 @cindex x86-64 options
22439 These are listed under @xref{i386 and x86-64 Options}.
22441 @node Xstormy16 Options
22442 @subsection Xstormy16 Options
22443 @cindex Xstormy16 Options
22445 These options are defined for Xstormy16:
22450 Choose startup files and linker script suitable for the simulator.
22453 @node Xtensa Options
22454 @subsection Xtensa Options
22455 @cindex Xtensa Options
22457 These options are supported for Xtensa targets:
22461 @itemx -mno-const16
22463 @opindex mno-const16
22464 Enable or disable use of @code{CONST16} instructions for loading
22465 constant values. The @code{CONST16} instruction is currently not a
22466 standard option from Tensilica. When enabled, @code{CONST16}
22467 instructions are always used in place of the standard @code{L32R}
22468 instructions. The use of @code{CONST16} is enabled by default only if
22469 the @code{L32R} instruction is not available.
22472 @itemx -mno-fused-madd
22473 @opindex mfused-madd
22474 @opindex mno-fused-madd
22475 Enable or disable use of fused multiply/add and multiply/subtract
22476 instructions in the floating-point option. This has no effect if the
22477 floating-point option is not also enabled. Disabling fused multiply/add
22478 and multiply/subtract instructions forces the compiler to use separate
22479 instructions for the multiply and add/subtract operations. This may be
22480 desirable in some cases where strict IEEE 754-compliant results are
22481 required: the fused multiply add/subtract instructions do not round the
22482 intermediate result, thereby producing results with @emph{more} bits of
22483 precision than specified by the IEEE standard. Disabling fused multiply
22484 add/subtract instructions also ensures that the program output is not
22485 sensitive to the compiler's ability to combine multiply and add/subtract
22488 @item -mserialize-volatile
22489 @itemx -mno-serialize-volatile
22490 @opindex mserialize-volatile
22491 @opindex mno-serialize-volatile
22492 When this option is enabled, GCC inserts @code{MEMW} instructions before
22493 @code{volatile} memory references to guarantee sequential consistency.
22494 The default is @option{-mserialize-volatile}. Use
22495 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
22497 @item -mforce-no-pic
22498 @opindex mforce-no-pic
22499 For targets, like GNU/Linux, where all user-mode Xtensa code must be
22500 position-independent code (PIC), this option disables PIC for compiling
22503 @item -mtext-section-literals
22504 @itemx -mno-text-section-literals
22505 @opindex mtext-section-literals
22506 @opindex mno-text-section-literals
22507 Control the treatment of literal pools. The default is
22508 @option{-mno-text-section-literals}, which places literals in a separate
22509 section in the output file. This allows the literal pool to be placed
22510 in a data RAM/ROM, and it also allows the linker to combine literal
22511 pools from separate object files to remove redundant literals and
22512 improve code size. With @option{-mtext-section-literals}, the literals
22513 are interspersed in the text section in order to keep them as close as
22514 possible to their references. This may be necessary for large assembly
22517 @item -mtarget-align
22518 @itemx -mno-target-align
22519 @opindex mtarget-align
22520 @opindex mno-target-align
22521 When this option is enabled, GCC instructs the assembler to
22522 automatically align instructions to reduce branch penalties at the
22523 expense of some code density. The assembler attempts to widen density
22524 instructions to align branch targets and the instructions following call
22525 instructions. If there are not enough preceding safe density
22526 instructions to align a target, no widening is performed. The
22527 default is @option{-mtarget-align}. These options do not affect the
22528 treatment of auto-aligned instructions like @code{LOOP}, which the
22529 assembler always aligns, either by widening density instructions or
22530 by inserting NOP instructions.
22533 @itemx -mno-longcalls
22534 @opindex mlongcalls
22535 @opindex mno-longcalls
22536 When this option is enabled, GCC instructs the assembler to translate
22537 direct calls to indirect calls unless it can determine that the target
22538 of a direct call is in the range allowed by the call instruction. This
22539 translation typically occurs for calls to functions in other source
22540 files. Specifically, the assembler translates a direct @code{CALL}
22541 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
22542 The default is @option{-mno-longcalls}. This option should be used in
22543 programs where the call target can potentially be out of range. This
22544 option is implemented in the assembler, not the compiler, so the
22545 assembly code generated by GCC still shows direct call
22546 instructions---look at the disassembled object code to see the actual
22547 instructions. Note that the assembler uses an indirect call for
22548 every cross-file call, not just those that really are out of range.
22551 @node zSeries Options
22552 @subsection zSeries Options
22553 @cindex zSeries options
22555 These are listed under @xref{S/390 and zSeries Options}.
22557 @node Code Gen Options
22558 @section Options for Code Generation Conventions
22559 @cindex code generation conventions
22560 @cindex options, code generation
22561 @cindex run-time options
22563 These machine-independent options control the interface conventions
22564 used in code generation.
22566 Most of them have both positive and negative forms; the negative form
22567 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
22568 one of the forms is listed---the one that is not the default. You
22569 can figure out the other form by either removing @samp{no-} or adding
22573 @item -fbounds-check
22574 @opindex fbounds-check
22575 For front ends that support it, generate additional code to check that
22576 indices used to access arrays are within the declared range. This is
22577 currently only supported by the Java and Fortran front ends, where
22578 this option defaults to true and false respectively.
22580 @item -fstack-reuse=@var{reuse-level}
22581 @opindex fstack_reuse
22582 This option controls stack space reuse for user declared local/auto variables
22583 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
22584 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
22585 local variables and temporaries, @samp{named_vars} enables the reuse only for
22586 user defined local variables with names, and @samp{none} disables stack reuse
22587 completely. The default value is @samp{all}. The option is needed when the
22588 program extends the lifetime of a scoped local variable or a compiler generated
22589 temporary beyond the end point defined by the language. When a lifetime of
22590 a variable ends, and if the variable lives in memory, the optimizing compiler
22591 has the freedom to reuse its stack space with other temporaries or scoped
22592 local variables whose live range does not overlap with it. Legacy code extending
22593 local lifetime will likely to break with the stack reuse optimization.
22612 if (*p == 10) // out of scope use of local1
22623 A(int k) : i(k), j(k) @{ @}
22630 void foo(const A& ar)
22637 foo(A(10)); // temp object's lifetime ends when foo returns
22643 ap->i+= 10; // ap references out of scope temp whose space
22644 // is reused with a. What is the value of ap->i?
22649 The lifetime of a compiler generated temporary is well defined by the C++
22650 standard. When a lifetime of a temporary ends, and if the temporary lives
22651 in memory, the optimizing compiler has the freedom to reuse its stack
22652 space with other temporaries or scoped local variables whose live range
22653 does not overlap with it. However some of the legacy code relies on
22654 the behavior of older compilers in which temporaries' stack space is
22655 not reused, the aggressive stack reuse can lead to runtime errors. This
22656 option is used to control the temporary stack reuse optimization.
22660 This option generates traps for signed overflow on addition, subtraction,
22661 multiplication operations.
22665 This option instructs the compiler to assume that signed arithmetic
22666 overflow of addition, subtraction and multiplication wraps around
22667 using twos-complement representation. This flag enables some optimizations
22668 and disables others. This option is enabled by default for the Java
22669 front end, as required by the Java language specification.
22672 @opindex fexceptions
22673 Enable exception handling. Generates extra code needed to propagate
22674 exceptions. For some targets, this implies GCC generates frame
22675 unwind information for all functions, which can produce significant data
22676 size overhead, although it does not affect execution. If you do not
22677 specify this option, GCC enables it by default for languages like
22678 C++ that normally require exception handling, and disables it for
22679 languages like C that do not normally require it. However, you may need
22680 to enable this option when compiling C code that needs to interoperate
22681 properly with exception handlers written in C++. You may also wish to
22682 disable this option if you are compiling older C++ programs that don't
22683 use exception handling.
22685 @item -fnon-call-exceptions
22686 @opindex fnon-call-exceptions
22687 Generate code that allows trapping instructions to throw exceptions.
22688 Note that this requires platform-specific runtime support that does
22689 not exist everywhere. Moreover, it only allows @emph{trapping}
22690 instructions to throw exceptions, i.e.@: memory references or floating-point
22691 instructions. It does not allow exceptions to be thrown from
22692 arbitrary signal handlers such as @code{SIGALRM}.
22694 @item -fdelete-dead-exceptions
22695 @opindex fdelete-dead-exceptions
22696 Consider that instructions that may throw exceptions but don't otherwise
22697 contribute to the execution of the program can be optimized away.
22698 This option is enabled by default for the Ada front end, as permitted by
22699 the Ada language specification.
22700 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
22702 @item -funwind-tables
22703 @opindex funwind-tables
22704 Similar to @option{-fexceptions}, except that it just generates any needed
22705 static data, but does not affect the generated code in any other way.
22706 You normally do not need to enable this option; instead, a language processor
22707 that needs this handling enables it on your behalf.
22709 @item -fasynchronous-unwind-tables
22710 @opindex fasynchronous-unwind-tables
22711 Generate unwind table in DWARF 2 format, if supported by target machine. The
22712 table is exact at each instruction boundary, so it can be used for stack
22713 unwinding from asynchronous events (such as debugger or garbage collector).
22715 @item -fno-gnu-unique
22716 @opindex fno-gnu-unique
22717 On systems with recent GNU assembler and C library, the C++ compiler
22718 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
22719 of template static data members and static local variables in inline
22720 functions are unique even in the presence of @code{RTLD_LOCAL}; this
22721 is necessary to avoid problems with a library used by two different
22722 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
22723 therefore disagreeing with the other one about the binding of the
22724 symbol. But this causes @code{dlclose} to be ignored for affected
22725 DSOs; if your program relies on reinitialization of a DSO via
22726 @code{dlclose} and @code{dlopen}, you can use
22727 @option{-fno-gnu-unique}.
22729 @item -fpcc-struct-return
22730 @opindex fpcc-struct-return
22731 Return ``short'' @code{struct} and @code{union} values in memory like
22732 longer ones, rather than in registers. This convention is less
22733 efficient, but it has the advantage of allowing intercallability between
22734 GCC-compiled files and files compiled with other compilers, particularly
22735 the Portable C Compiler (pcc).
22737 The precise convention for returning structures in memory depends
22738 on the target configuration macros.
22740 Short structures and unions are those whose size and alignment match
22741 that of some integer type.
22743 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
22744 switch is not binary compatible with code compiled with the
22745 @option{-freg-struct-return} switch.
22746 Use it to conform to a non-default application binary interface.
22748 @item -freg-struct-return
22749 @opindex freg-struct-return
22750 Return @code{struct} and @code{union} values in registers when possible.
22751 This is more efficient for small structures than
22752 @option{-fpcc-struct-return}.
22754 If you specify neither @option{-fpcc-struct-return} nor
22755 @option{-freg-struct-return}, GCC defaults to whichever convention is
22756 standard for the target. If there is no standard convention, GCC
22757 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
22758 the principal compiler. In those cases, we can choose the standard, and
22759 we chose the more efficient register return alternative.
22761 @strong{Warning:} code compiled with the @option{-freg-struct-return}
22762 switch is not binary compatible with code compiled with the
22763 @option{-fpcc-struct-return} switch.
22764 Use it to conform to a non-default application binary interface.
22766 @item -fshort-enums
22767 @opindex fshort-enums
22768 Allocate to an @code{enum} type only as many bytes as it needs for the
22769 declared range of possible values. Specifically, the @code{enum} type
22770 is equivalent to the smallest integer type that has enough room.
22772 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
22773 code that is not binary compatible with code generated without that switch.
22774 Use it to conform to a non-default application binary interface.
22776 @item -fshort-double
22777 @opindex fshort-double
22778 Use the same size for @code{double} as for @code{float}.
22780 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
22781 code that is not binary compatible with code generated without that switch.
22782 Use it to conform to a non-default application binary interface.
22784 @item -fshort-wchar
22785 @opindex fshort-wchar
22786 Override the underlying type for @samp{wchar_t} to be @samp{short
22787 unsigned int} instead of the default for the target. This option is
22788 useful for building programs to run under WINE@.
22790 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
22791 code that is not binary compatible with code generated without that switch.
22792 Use it to conform to a non-default application binary interface.
22795 @opindex fno-common
22796 In C code, controls the placement of uninitialized global variables.
22797 Unix C compilers have traditionally permitted multiple definitions of
22798 such variables in different compilation units by placing the variables
22800 This is the behavior specified by @option{-fcommon}, and is the default
22801 for GCC on most targets.
22802 On the other hand, this behavior is not required by ISO C, and on some
22803 targets may carry a speed or code size penalty on variable references.
22804 The @option{-fno-common} option specifies that the compiler should place
22805 uninitialized global variables in the data section of the object file,
22806 rather than generating them as common blocks.
22807 This has the effect that if the same variable is declared
22808 (without @code{extern}) in two different compilations,
22809 you get a multiple-definition error when you link them.
22810 In this case, you must compile with @option{-fcommon} instead.
22811 Compiling with @option{-fno-common} is useful on targets for which
22812 it provides better performance, or if you wish to verify that the
22813 program will work on other systems that always treat uninitialized
22814 variable declarations this way.
22818 Ignore the @samp{#ident} directive.
22820 @item -finhibit-size-directive
22821 @opindex finhibit-size-directive
22822 Don't output a @code{.size} assembler directive, or anything else that
22823 would cause trouble if the function is split in the middle, and the
22824 two halves are placed at locations far apart in memory. This option is
22825 used when compiling @file{crtstuff.c}; you should not need to use it
22828 @item -fverbose-asm
22829 @opindex fverbose-asm
22830 Put extra commentary information in the generated assembly code to
22831 make it more readable. This option is generally only of use to those
22832 who actually need to read the generated assembly code (perhaps while
22833 debugging the compiler itself).
22835 @option{-fno-verbose-asm}, the default, causes the
22836 extra information to be omitted and is useful when comparing two assembler
22839 @item -frecord-gcc-switches
22840 @opindex frecord-gcc-switches
22841 This switch causes the command line used to invoke the
22842 compiler to be recorded into the object file that is being created.
22843 This switch is only implemented on some targets and the exact format
22844 of the recording is target and binary file format dependent, but it
22845 usually takes the form of a section containing ASCII text. This
22846 switch is related to the @option{-fverbose-asm} switch, but that
22847 switch only records information in the assembler output file as
22848 comments, so it never reaches the object file.
22849 See also @option{-grecord-gcc-switches} for another
22850 way of storing compiler options into the object file.
22854 @cindex global offset table
22856 Generate position-independent code (PIC) suitable for use in a shared
22857 library, if supported for the target machine. Such code accesses all
22858 constant addresses through a global offset table (GOT)@. The dynamic
22859 loader resolves the GOT entries when the program starts (the dynamic
22860 loader is not part of GCC; it is part of the operating system). If
22861 the GOT size for the linked executable exceeds a machine-specific
22862 maximum size, you get an error message from the linker indicating that
22863 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
22864 instead. (These maximums are 8k on the SPARC and 32k
22865 on the m68k and RS/6000. The 386 has no such limit.)
22867 Position-independent code requires special support, and therefore works
22868 only on certain machines. For the 386, GCC supports PIC for System V
22869 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
22870 position-independent.
22872 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
22877 If supported for the target machine, emit position-independent code,
22878 suitable for dynamic linking and avoiding any limit on the size of the
22879 global offset table. This option makes a difference on the m68k,
22880 PowerPC and SPARC@.
22882 Position-independent code requires special support, and therefore works
22883 only on certain machines.
22885 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
22892 These options are similar to @option{-fpic} and @option{-fPIC}, but
22893 generated position independent code can be only linked into executables.
22894 Usually these options are used when @option{-pie} GCC option is
22895 used during linking.
22897 @option{-fpie} and @option{-fPIE} both define the macros
22898 @code{__pie__} and @code{__PIE__}. The macros have the value 1
22899 for @option{-fpie} and 2 for @option{-fPIE}.
22901 @item -fno-jump-tables
22902 @opindex fno-jump-tables
22903 Do not use jump tables for switch statements even where it would be
22904 more efficient than other code generation strategies. This option is
22905 of use in conjunction with @option{-fpic} or @option{-fPIC} for
22906 building code that forms part of a dynamic linker and cannot
22907 reference the address of a jump table. On some targets, jump tables
22908 do not require a GOT and this option is not needed.
22910 @item -ffixed-@var{reg}
22912 Treat the register named @var{reg} as a fixed register; generated code
22913 should never refer to it (except perhaps as a stack pointer, frame
22914 pointer or in some other fixed role).
22916 @var{reg} must be the name of a register. The register names accepted
22917 are machine-specific and are defined in the @code{REGISTER_NAMES}
22918 macro in the machine description macro file.
22920 This flag does not have a negative form, because it specifies a
22923 @item -fcall-used-@var{reg}
22924 @opindex fcall-used
22925 Treat the register named @var{reg} as an allocable register that is
22926 clobbered by function calls. It may be allocated for temporaries or
22927 variables that do not live across a call. Functions compiled this way
22928 do not save and restore the register @var{reg}.
22930 It is an error to use this flag with the frame pointer or stack pointer.
22931 Use of this flag for other registers that have fixed pervasive roles in
22932 the machine's execution model produces disastrous results.
22934 This flag does not have a negative form, because it specifies a
22937 @item -fcall-saved-@var{reg}
22938 @opindex fcall-saved
22939 Treat the register named @var{reg} as an allocable register saved by
22940 functions. It may be allocated even for temporaries or variables that
22941 live across a call. Functions compiled this way save and restore
22942 the register @var{reg} if they use it.
22944 It is an error to use this flag with the frame pointer or stack pointer.
22945 Use of this flag for other registers that have fixed pervasive roles in
22946 the machine's execution model produces disastrous results.
22948 A different sort of disaster results from the use of this flag for
22949 a register in which function values may be returned.
22951 This flag does not have a negative form, because it specifies a
22954 @item -fpack-struct[=@var{n}]
22955 @opindex fpack-struct
22956 Without a value specified, pack all structure members together without
22957 holes. When a value is specified (which must be a small power of two), pack
22958 structure members according to this value, representing the maximum
22959 alignment (that is, objects with default alignment requirements larger than
22960 this are output potentially unaligned at the next fitting location.
22962 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
22963 code that is not binary compatible with code generated without that switch.
22964 Additionally, it makes the code suboptimal.
22965 Use it to conform to a non-default application binary interface.
22967 @item -finstrument-functions
22968 @opindex finstrument-functions
22969 Generate instrumentation calls for entry and exit to functions. Just
22970 after function entry and just before function exit, the following
22971 profiling functions are called with the address of the current
22972 function and its call site. (On some platforms,
22973 @code{__builtin_return_address} does not work beyond the current
22974 function, so the call site information may not be available to the
22975 profiling functions otherwise.)
22978 void __cyg_profile_func_enter (void *this_fn,
22980 void __cyg_profile_func_exit (void *this_fn,
22984 The first argument is the address of the start of the current function,
22985 which may be looked up exactly in the symbol table.
22987 This instrumentation is also done for functions expanded inline in other
22988 functions. The profiling calls indicate where, conceptually, the
22989 inline function is entered and exited. This means that addressable
22990 versions of such functions must be available. If all your uses of a
22991 function are expanded inline, this may mean an additional expansion of
22992 code size. If you use @samp{extern inline} in your C code, an
22993 addressable version of such functions must be provided. (This is
22994 normally the case anyway, but if you get lucky and the optimizer always
22995 expands the functions inline, you might have gotten away without
22996 providing static copies.)
22998 A function may be given the attribute @code{no_instrument_function}, in
22999 which case this instrumentation is not done. This can be used, for
23000 example, for the profiling functions listed above, high-priority
23001 interrupt routines, and any functions from which the profiling functions
23002 cannot safely be called (perhaps signal handlers, if the profiling
23003 routines generate output or allocate memory).
23005 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
23006 @opindex finstrument-functions-exclude-file-list
23008 Set the list of functions that are excluded from instrumentation (see
23009 the description of @code{-finstrument-functions}). If the file that
23010 contains a function definition matches with one of @var{file}, then
23011 that function is not instrumented. The match is done on substrings:
23012 if the @var{file} parameter is a substring of the file name, it is
23013 considered to be a match.
23018 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
23022 excludes any inline function defined in files whose pathnames
23023 contain @code{/bits/stl} or @code{include/sys}.
23025 If, for some reason, you want to include letter @code{','} in one of
23026 @var{sym}, write @code{'\,'}. For example,
23027 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
23028 (note the single quote surrounding the option).
23030 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
23031 @opindex finstrument-functions-exclude-function-list
23033 This is similar to @code{-finstrument-functions-exclude-file-list},
23034 but this option sets the list of function names to be excluded from
23035 instrumentation. The function name to be matched is its user-visible
23036 name, such as @code{vector<int> blah(const vector<int> &)}, not the
23037 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
23038 match is done on substrings: if the @var{sym} parameter is a substring
23039 of the function name, it is considered to be a match. For C99 and C++
23040 extended identifiers, the function name must be given in UTF-8, not
23041 using universal character names.
23043 @item -fstack-check
23044 @opindex fstack-check
23045 Generate code to verify that you do not go beyond the boundary of the
23046 stack. You should specify this flag if you are running in an
23047 environment with multiple threads, but you only rarely need to specify it in
23048 a single-threaded environment since stack overflow is automatically
23049 detected on nearly all systems if there is only one stack.
23051 Note that this switch does not actually cause checking to be done; the
23052 operating system or the language runtime must do that. The switch causes
23053 generation of code to ensure that they see the stack being extended.
23055 You can additionally specify a string parameter: @code{no} means no
23056 checking, @code{generic} means force the use of old-style checking,
23057 @code{specific} means use the best checking method and is equivalent
23058 to bare @option{-fstack-check}.
23060 Old-style checking is a generic mechanism that requires no specific
23061 target support in the compiler but comes with the following drawbacks:
23065 Modified allocation strategy for large objects: they are always
23066 allocated dynamically if their size exceeds a fixed threshold.
23069 Fixed limit on the size of the static frame of functions: when it is
23070 topped by a particular function, stack checking is not reliable and
23071 a warning is issued by the compiler.
23074 Inefficiency: because of both the modified allocation strategy and the
23075 generic implementation, code performance is hampered.
23078 Note that old-style stack checking is also the fallback method for
23079 @code{specific} if no target support has been added in the compiler.
23081 @item -fstack-limit-register=@var{reg}
23082 @itemx -fstack-limit-symbol=@var{sym}
23083 @itemx -fno-stack-limit
23084 @opindex fstack-limit-register
23085 @opindex fstack-limit-symbol
23086 @opindex fno-stack-limit
23087 Generate code to ensure that the stack does not grow beyond a certain value,
23088 either the value of a register or the address of a symbol. If a larger
23089 stack is required, a signal is raised at run time. For most targets,
23090 the signal is raised before the stack overruns the boundary, so
23091 it is possible to catch the signal without taking special precautions.
23093 For instance, if the stack starts at absolute address @samp{0x80000000}
23094 and grows downwards, you can use the flags
23095 @option{-fstack-limit-symbol=__stack_limit} and
23096 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
23097 of 128KB@. Note that this may only work with the GNU linker.
23099 @item -fsplit-stack
23100 @opindex fsplit-stack
23101 Generate code to automatically split the stack before it overflows.
23102 The resulting program has a discontiguous stack which can only
23103 overflow if the program is unable to allocate any more memory. This
23104 is most useful when running threaded programs, as it is no longer
23105 necessary to calculate a good stack size to use for each thread. This
23106 is currently only implemented for the i386 and x86_64 back ends running
23109 When code compiled with @option{-fsplit-stack} calls code compiled
23110 without @option{-fsplit-stack}, there may not be much stack space
23111 available for the latter code to run. If compiling all code,
23112 including library code, with @option{-fsplit-stack} is not an option,
23113 then the linker can fix up these calls so that the code compiled
23114 without @option{-fsplit-stack} always has a large stack. Support for
23115 this is implemented in the gold linker in GNU binutils release 2.21
23118 @item -fleading-underscore
23119 @opindex fleading-underscore
23120 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
23121 change the way C symbols are represented in the object file. One use
23122 is to help link with legacy assembly code.
23124 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
23125 generate code that is not binary compatible with code generated without that
23126 switch. Use it to conform to a non-default application binary interface.
23127 Not all targets provide complete support for this switch.
23129 @item -ftls-model=@var{model}
23130 @opindex ftls-model
23131 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
23132 The @var{model} argument should be one of @code{global-dynamic},
23133 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
23134 Note that the choice is subject to optimization: the compiler may use
23135 a more efficient model for symbols not visible outside of the translation
23136 unit, or if @option{-fpic} is not given on the command line.
23138 The default without @option{-fpic} is @code{initial-exec}; with
23139 @option{-fpic} the default is @code{global-dynamic}.
23141 @item -fvisibility=@var{default|internal|hidden|protected}
23142 @opindex fvisibility
23143 Set the default ELF image symbol visibility to the specified option---all
23144 symbols are marked with this unless overridden within the code.
23145 Using this feature can very substantially improve linking and
23146 load times of shared object libraries, produce more optimized
23147 code, provide near-perfect API export and prevent symbol clashes.
23148 It is @strong{strongly} recommended that you use this in any shared objects
23151 Despite the nomenclature, @code{default} always means public; i.e.,
23152 available to be linked against from outside the shared object.
23153 @code{protected} and @code{internal} are pretty useless in real-world
23154 usage so the only other commonly used option is @code{hidden}.
23155 The default if @option{-fvisibility} isn't specified is
23156 @code{default}, i.e., make every
23157 symbol public---this causes the same behavior as previous versions of
23160 A good explanation of the benefits offered by ensuring ELF
23161 symbols have the correct visibility is given by ``How To Write
23162 Shared Libraries'' by Ulrich Drepper (which can be found at
23163 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
23164 solution made possible by this option to marking things hidden when
23165 the default is public is to make the default hidden and mark things
23166 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
23167 and @code{__attribute__ ((visibility("default")))} instead of
23168 @code{__declspec(dllexport)} you get almost identical semantics with
23169 identical syntax. This is a great boon to those working with
23170 cross-platform projects.
23172 For those adding visibility support to existing code, you may find
23173 @samp{#pragma GCC visibility} of use. This works by you enclosing
23174 the declarations you wish to set visibility for with (for example)
23175 @samp{#pragma GCC visibility push(hidden)} and
23176 @samp{#pragma GCC visibility pop}.
23177 Bear in mind that symbol visibility should be viewed @strong{as
23178 part of the API interface contract} and thus all new code should
23179 always specify visibility when it is not the default; i.e., declarations
23180 only for use within the local DSO should @strong{always} be marked explicitly
23181 as hidden as so to avoid PLT indirection overheads---making this
23182 abundantly clear also aids readability and self-documentation of the code.
23183 Note that due to ISO C++ specification requirements, @code{operator new} and
23184 @code{operator delete} must always be of default visibility.
23186 Be aware that headers from outside your project, in particular system
23187 headers and headers from any other library you use, may not be
23188 expecting to be compiled with visibility other than the default. You
23189 may need to explicitly say @samp{#pragma GCC visibility push(default)}
23190 before including any such headers.
23192 @samp{extern} declarations are not affected by @option{-fvisibility}, so
23193 a lot of code can be recompiled with @option{-fvisibility=hidden} with
23194 no modifications. However, this means that calls to @code{extern}
23195 functions with no explicit visibility use the PLT, so it is more
23196 effective to use @code{__attribute ((visibility))} and/or
23197 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
23198 declarations should be treated as hidden.
23200 Note that @option{-fvisibility} does affect C++ vague linkage
23201 entities. This means that, for instance, an exception class that is
23202 be thrown between DSOs must be explicitly marked with default
23203 visibility so that the @samp{type_info} nodes are unified between
23206 An overview of these techniques, their benefits and how to use them
23207 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
23209 @item -fstrict-volatile-bitfields
23210 @opindex fstrict-volatile-bitfields
23211 This option should be used if accesses to volatile bit-fields (or other
23212 structure fields, although the compiler usually honors those types
23213 anyway) should use a single access of the width of the
23214 field's type, aligned to a natural alignment if possible. For
23215 example, targets with memory-mapped peripheral registers might require
23216 all such accesses to be 16 bits wide; with this flag you can
23217 declare all peripheral bit-fields as @code{unsigned short} (assuming short
23218 is 16 bits on these targets) to force GCC to use 16-bit accesses
23219 instead of, perhaps, a more efficient 32-bit access.
23221 If this option is disabled, the compiler uses the most efficient
23222 instruction. In the previous example, that might be a 32-bit load
23223 instruction, even though that accesses bytes that do not contain
23224 any portion of the bit-field, or memory-mapped registers unrelated to
23225 the one being updated.
23227 In some cases, such as when the @code{packed} attribute is applied to a
23228 structure field, it may not be possible to access the field with a single
23229 read or write that is correctly aligned for the target machine. In this
23230 case GCC falls back to generating multiple accesses rather than code that
23231 will fault or truncate the result at run time.
23233 Note: Due to restrictions of the C/C++11 memory model, write accesses are
23234 not allowed to touch non bit-field members. It is therefore recommended
23235 to define all bits of the field's type as bit-field members.
23237 The default value of this option is determined by the application binary
23238 interface for the target processor.
23240 @item -fsync-libcalls
23241 @opindex fsync-libcalls
23242 This option controls whether any out-of-line instance of the @code{__sync}
23243 family of functions may be used to implement the C++11 @code{__atomic}
23244 family of functions.
23246 The default value of this option is enabled, thus the only useful form
23247 of the option is @option{-fno-sync-libcalls}. This option is used in
23248 the implementation of the @file{libatomic} runtime library.
23254 @node Environment Variables
23255 @section Environment Variables Affecting GCC
23256 @cindex environment variables
23258 @c man begin ENVIRONMENT
23259 This section describes several environment variables that affect how GCC
23260 operates. Some of them work by specifying directories or prefixes to use
23261 when searching for various kinds of files. Some are used to specify other
23262 aspects of the compilation environment.
23264 Note that you can also specify places to search using options such as
23265 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
23266 take precedence over places specified using environment variables, which
23267 in turn take precedence over those specified by the configuration of GCC@.
23268 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
23269 GNU Compiler Collection (GCC) Internals}.
23274 @c @itemx LC_COLLATE
23276 @c @itemx LC_MONETARY
23277 @c @itemx LC_NUMERIC
23282 @c @findex LC_COLLATE
23283 @findex LC_MESSAGES
23284 @c @findex LC_MONETARY
23285 @c @findex LC_NUMERIC
23289 These environment variables control the way that GCC uses
23290 localization information which allows GCC to work with different
23291 national conventions. GCC inspects the locale categories
23292 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
23293 so. These locale categories can be set to any value supported by your
23294 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
23295 Kingdom encoded in UTF-8.
23297 The @env{LC_CTYPE} environment variable specifies character
23298 classification. GCC uses it to determine the character boundaries in
23299 a string; this is needed for some multibyte encodings that contain quote
23300 and escape characters that are otherwise interpreted as a string
23303 The @env{LC_MESSAGES} environment variable specifies the language to
23304 use in diagnostic messages.
23306 If the @env{LC_ALL} environment variable is set, it overrides the value
23307 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
23308 and @env{LC_MESSAGES} default to the value of the @env{LANG}
23309 environment variable. If none of these variables are set, GCC
23310 defaults to traditional C English behavior.
23314 If @env{TMPDIR} is set, it specifies the directory to use for temporary
23315 files. GCC uses temporary files to hold the output of one stage of
23316 compilation which is to be used as input to the next stage: for example,
23317 the output of the preprocessor, which is the input to the compiler
23320 @item GCC_COMPARE_DEBUG
23321 @findex GCC_COMPARE_DEBUG
23322 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
23323 @option{-fcompare-debug} to the compiler driver. See the documentation
23324 of this option for more details.
23326 @item GCC_EXEC_PREFIX
23327 @findex GCC_EXEC_PREFIX
23328 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
23329 names of the subprograms executed by the compiler. No slash is added
23330 when this prefix is combined with the name of a subprogram, but you can
23331 specify a prefix that ends with a slash if you wish.
23333 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
23334 an appropriate prefix to use based on the pathname it is invoked with.
23336 If GCC cannot find the subprogram using the specified prefix, it
23337 tries looking in the usual places for the subprogram.
23339 The default value of @env{GCC_EXEC_PREFIX} is
23340 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
23341 the installed compiler. In many cases @var{prefix} is the value
23342 of @code{prefix} when you ran the @file{configure} script.
23344 Other prefixes specified with @option{-B} take precedence over this prefix.
23346 This prefix is also used for finding files such as @file{crt0.o} that are
23349 In addition, the prefix is used in an unusual way in finding the
23350 directories to search for header files. For each of the standard
23351 directories whose name normally begins with @samp{/usr/local/lib/gcc}
23352 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
23353 replacing that beginning with the specified prefix to produce an
23354 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
23355 @file{foo/bar} just before it searches the standard directory
23356 @file{/usr/local/lib/bar}.
23357 If a standard directory begins with the configured
23358 @var{prefix} then the value of @var{prefix} is replaced by
23359 @env{GCC_EXEC_PREFIX} when looking for header files.
23361 @item COMPILER_PATH
23362 @findex COMPILER_PATH
23363 The value of @env{COMPILER_PATH} is a colon-separated list of
23364 directories, much like @env{PATH}. GCC tries the directories thus
23365 specified when searching for subprograms, if it can't find the
23366 subprograms using @env{GCC_EXEC_PREFIX}.
23369 @findex LIBRARY_PATH
23370 The value of @env{LIBRARY_PATH} is a colon-separated list of
23371 directories, much like @env{PATH}. When configured as a native compiler,
23372 GCC tries the directories thus specified when searching for special
23373 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
23374 using GCC also uses these directories when searching for ordinary
23375 libraries for the @option{-l} option (but directories specified with
23376 @option{-L} come first).
23380 @cindex locale definition
23381 This variable is used to pass locale information to the compiler. One way in
23382 which this information is used is to determine the character set to be used
23383 when character literals, string literals and comments are parsed in C and C++.
23384 When the compiler is configured to allow multibyte characters,
23385 the following values for @env{LANG} are recognized:
23389 Recognize JIS characters.
23391 Recognize SJIS characters.
23393 Recognize EUCJP characters.
23396 If @env{LANG} is not defined, or if it has some other value, then the
23397 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
23398 recognize and translate multibyte characters.
23402 Some additional environment variables affect the behavior of the
23405 @include cppenv.texi
23409 @node Precompiled Headers
23410 @section Using Precompiled Headers
23411 @cindex precompiled headers
23412 @cindex speed of compilation
23414 Often large projects have many header files that are included in every
23415 source file. The time the compiler takes to process these header files
23416 over and over again can account for nearly all of the time required to
23417 build the project. To make builds faster, GCC allows you to
23418 @dfn{precompile} a header file.
23420 To create a precompiled header file, simply compile it as you would any
23421 other file, if necessary using the @option{-x} option to make the driver
23422 treat it as a C or C++ header file. You may want to use a
23423 tool like @command{make} to keep the precompiled header up-to-date when
23424 the headers it contains change.
23426 A precompiled header file is searched for when @code{#include} is
23427 seen in the compilation. As it searches for the included file
23428 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
23429 compiler looks for a precompiled header in each directory just before it
23430 looks for the include file in that directory. The name searched for is
23431 the name specified in the @code{#include} with @samp{.gch} appended. If
23432 the precompiled header file can't be used, it is ignored.
23434 For instance, if you have @code{#include "all.h"}, and you have
23435 @file{all.h.gch} in the same directory as @file{all.h}, then the
23436 precompiled header file is used if possible, and the original
23437 header is used otherwise.
23439 Alternatively, you might decide to put the precompiled header file in a
23440 directory and use @option{-I} to ensure that directory is searched
23441 before (or instead of) the directory containing the original header.
23442 Then, if you want to check that the precompiled header file is always
23443 used, you can put a file of the same name as the original header in this
23444 directory containing an @code{#error} command.
23446 This also works with @option{-include}. So yet another way to use
23447 precompiled headers, good for projects not designed with precompiled
23448 header files in mind, is to simply take most of the header files used by
23449 a project, include them from another header file, precompile that header
23450 file, and @option{-include} the precompiled header. If the header files
23451 have guards against multiple inclusion, they are skipped because
23452 they've already been included (in the precompiled header).
23454 If you need to precompile the same header file for different
23455 languages, targets, or compiler options, you can instead make a
23456 @emph{directory} named like @file{all.h.gch}, and put each precompiled
23457 header in the directory, perhaps using @option{-o}. It doesn't matter
23458 what you call the files in the directory; every precompiled header in
23459 the directory is considered. The first precompiled header
23460 encountered in the directory that is valid for this compilation is
23461 used; they're searched in no particular order.
23463 There are many other possibilities, limited only by your imagination,
23464 good sense, and the constraints of your build system.
23466 A precompiled header file can be used only when these conditions apply:
23470 Only one precompiled header can be used in a particular compilation.
23473 A precompiled header can't be used once the first C token is seen. You
23474 can have preprocessor directives before a precompiled header; you cannot
23475 include a precompiled header from inside another header.
23478 The precompiled header file must be produced for the same language as
23479 the current compilation. You can't use a C precompiled header for a C++
23483 The precompiled header file must have been produced by the same compiler
23484 binary as the current compilation is using.
23487 Any macros defined before the precompiled header is included must
23488 either be defined in the same way as when the precompiled header was
23489 generated, or must not affect the precompiled header, which usually
23490 means that they don't appear in the precompiled header at all.
23492 The @option{-D} option is one way to define a macro before a
23493 precompiled header is included; using a @code{#define} can also do it.
23494 There are also some options that define macros implicitly, like
23495 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
23498 @item If debugging information is output when using the precompiled
23499 header, using @option{-g} or similar, the same kind of debugging information
23500 must have been output when building the precompiled header. However,
23501 a precompiled header built using @option{-g} can be used in a compilation
23502 when no debugging information is being output.
23504 @item The same @option{-m} options must generally be used when building
23505 and using the precompiled header. @xref{Submodel Options},
23506 for any cases where this rule is relaxed.
23508 @item Each of the following options must be the same when building and using
23509 the precompiled header:
23511 @gccoptlist{-fexceptions}
23514 Some other command-line options starting with @option{-f},
23515 @option{-p}, or @option{-O} must be defined in the same way as when
23516 the precompiled header was generated. At present, it's not clear
23517 which options are safe to change and which are not; the safest choice
23518 is to use exactly the same options when generating and using the
23519 precompiled header. The following are known to be safe:
23521 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
23522 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
23523 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
23528 For all of these except the last, the compiler automatically
23529 ignores the precompiled header if the conditions aren't met. If you
23530 find an option combination that doesn't work and doesn't cause the
23531 precompiled header to be ignored, please consider filing a bug report,
23534 If you do use differing options when generating and using the
23535 precompiled header, the actual behavior is a mixture of the
23536 behavior for the options. For instance, if you use @option{-g} to
23537 generate the precompiled header but not when using it, you may or may
23538 not get debugging information for routines in the precompiled header.