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 @gol
395 -floop-unroll-and-jam -floop-nest-optimize @gol
396 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
397 -flto-partition=@var{alg} -flto-report -flto-report-wpa -fmerge-all-constants @gol
398 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
399 -fmove-loop-invariants -fno-branch-count-reg @gol
400 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
401 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
402 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
403 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
404 -fomit-frame-pointer -foptimize-sibling-calls @gol
405 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
406 -fprefetch-loop-arrays -fprofile-report @gol
407 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
408 -fprofile-generate=@var{path} @gol
409 -fprofile-use -fprofile-use=@var{path} -fprofile-values -fprofile-reorder-functions @gol
410 -freciprocal-math -free -frename-registers -fschedule-fusion -freorder-blocks @gol
411 -freorder-blocks-and-partition -freorder-functions @gol
412 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
413 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
414 -fsched-spec-load -fsched-spec-load-dangerous @gol
415 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
416 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
417 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
418 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
419 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
420 -fselective-scheduling -fselective-scheduling2 @gol
421 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
422 -fsemantic-interposition @gol
423 -fshrink-wrap -fsignaling-nans -fsingle-precision-constant @gol
424 -fsplit-ivs-in-unroller -fsplit-wide-types -fssa-phiopt -fstack-protector @gol
425 -fstack-protector-all -fstack-protector-strong -fstrict-aliasing @gol
426 -fstrict-overflow -fthread-jumps -ftracer -ftree-bit-ccp @gol
427 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
428 -ftree-coalesce-inline-vars -ftree-coalesce-vars -ftree-copy-prop @gol
429 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
430 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
431 -ftree-loop-if-convert-stores -ftree-loop-im @gol
432 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
433 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
434 -ftree-loop-vectorize @gol
435 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
436 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
437 -ftree-switch-conversion -ftree-tail-merge -ftree-ter @gol
438 -ftree-vectorize -ftree-vrp @gol
439 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
440 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
441 -fuse-caller-save -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
442 -fweb -fwhole-program -fwpa -fuse-ld=@var{linker} -fuse-linker-plugin @gol
443 --param @var{name}=@var{value}
444 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
446 @item Preprocessor Options
447 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
448 @gccoptlist{-A@var{question}=@var{answer} @gol
449 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
450 -C -dD -dI -dM -dN @gol
451 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
452 -idirafter @var{dir} @gol
453 -include @var{file} -imacros @var{file} @gol
454 -iprefix @var{file} -iwithprefix @var{dir} @gol
455 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
456 -imultilib @var{dir} -isysroot @var{dir} @gol
457 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
458 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
459 -remap -trigraphs -undef -U@var{macro} @gol
460 -Wp,@var{option} -Xpreprocessor @var{option} -no-integrated-cpp}
462 @item Assembler Option
463 @xref{Assembler Options,,Passing Options to the Assembler}.
464 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
467 @xref{Link Options,,Options for Linking}.
468 @gccoptlist{@var{object-file-name} -l@var{library} @gol
469 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
470 -s -static -static-libgcc -static-libstdc++ @gol
471 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
472 -shared -shared-libgcc -symbolic @gol
473 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
474 -u @var{symbol} -z @var{keyword}}
476 @item Directory Options
477 @xref{Directory Options,,Options for Directory Search}.
478 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
479 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
480 --sysroot=@var{dir} --no-sysroot-suffix}
482 @item Machine Dependent Options
483 @xref{Submodel Options,,Hardware Models and Configurations}.
484 @c This list is ordered alphanumerically by subsection name.
485 @c Try and put the significant identifier (CPU or system) first,
486 @c so users have a clue at guessing where the ones they want will be.
488 @emph{AArch64 Options}
489 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
490 -mgeneral-regs-only @gol
491 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
493 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
494 -mtls-dialect=desc -mtls-dialect=traditional @gol
495 -mfix-cortex-a53-835769 -mno-fix-cortex-a53-835769 @gol
496 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
498 @emph{Adapteva Epiphany Options}
499 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
500 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
501 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
502 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
503 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
504 -msplit-vecmove-early -m1reg-@var{reg}}
507 @gccoptlist{-mbarrel-shifter @gol
508 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
509 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
510 -mea -mno-mpy -mmul32x16 -mmul64 @gol
511 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
512 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
513 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
514 -mepilogue-cfi -mlong-calls -mmedium-calls -msdata @gol
515 -mucb-mcount -mvolatile-cache @gol
516 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
517 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
518 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
519 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
520 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
521 -mtune=@var{cpu} -mmultcost=@var{num} -munalign-prob-threshold=@var{probability}}
524 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
525 -mabi=@var{name} @gol
526 -mapcs-stack-check -mno-apcs-stack-check @gol
527 -mapcs-float -mno-apcs-float @gol
528 -mapcs-reentrant -mno-apcs-reentrant @gol
529 -msched-prolog -mno-sched-prolog @gol
530 -mlittle-endian -mbig-endian @gol
531 -mfloat-abi=@var{name} @gol
532 -mfp16-format=@var{name}
533 -mthumb-interwork -mno-thumb-interwork @gol
534 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
535 -mstructure-size-boundary=@var{n} @gol
536 -mabort-on-noreturn @gol
537 -mlong-calls -mno-long-calls @gol
538 -msingle-pic-base -mno-single-pic-base @gol
539 -mpic-register=@var{reg} @gol
540 -mnop-fun-dllimport @gol
541 -mpoke-function-name @gol
543 -mtpcs-frame -mtpcs-leaf-frame @gol
544 -mcaller-super-interworking -mcallee-super-interworking @gol
545 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
546 -mword-relocations @gol
547 -mfix-cortex-m3-ldrd @gol
548 -munaligned-access @gol
549 -mneon-for-64bits @gol
550 -mslow-flash-data @gol
551 -masm-syntax-unified @gol
555 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
556 -mcall-prologues -mint8 -mno-interrupts -mrelax @gol
557 -mstrict-X -mtiny-stack -Waddr-space-convert}
559 @emph{Blackfin Options}
560 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
561 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
562 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
563 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
564 -mno-id-shared-library -mshared-library-id=@var{n} @gol
565 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
566 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
567 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
571 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
572 -msim -msdata=@var{sdata-type}}
575 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
576 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
577 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
578 -mstack-align -mdata-align -mconst-align @gol
579 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
580 -melf -maout -melinux -mlinux -sim -sim2 @gol
581 -mmul-bug-workaround -mno-mul-bug-workaround}
584 @gccoptlist{-mmac @gol
585 -mcr16cplus -mcr16c @gol
586 -msim -mint32 -mbit-ops
587 -mdata-model=@var{model}}
589 @emph{Darwin Options}
590 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
591 -arch_only -bind_at_load -bundle -bundle_loader @gol
592 -client_name -compatibility_version -current_version @gol
594 -dependency-file -dylib_file -dylinker_install_name @gol
595 -dynamic -dynamiclib -exported_symbols_list @gol
596 -filelist -flat_namespace -force_cpusubtype_ALL @gol
597 -force_flat_namespace -headerpad_max_install_names @gol
599 -image_base -init -install_name -keep_private_externs @gol
600 -multi_module -multiply_defined -multiply_defined_unused @gol
601 -noall_load -no_dead_strip_inits_and_terms @gol
602 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
603 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
604 -private_bundle -read_only_relocs -sectalign @gol
605 -sectobjectsymbols -whyload -seg1addr @gol
606 -sectcreate -sectobjectsymbols -sectorder @gol
607 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
608 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
609 -segprot -segs_read_only_addr -segs_read_write_addr @gol
610 -single_module -static -sub_library -sub_umbrella @gol
611 -twolevel_namespace -umbrella -undefined @gol
612 -unexported_symbols_list -weak_reference_mismatches @gol
613 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
614 -mkernel -mone-byte-bool}
616 @emph{DEC Alpha Options}
617 @gccoptlist{-mno-fp-regs -msoft-float @gol
618 -mieee -mieee-with-inexact -mieee-conformant @gol
619 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
620 -mtrap-precision=@var{mode} -mbuild-constants @gol
621 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
622 -mbwx -mmax -mfix -mcix @gol
623 -mfloat-vax -mfloat-ieee @gol
624 -mexplicit-relocs -msmall-data -mlarge-data @gol
625 -msmall-text -mlarge-text @gol
626 -mmemory-latency=@var{time}}
629 @gccoptlist{-msmall-model -mno-lsim}
632 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
633 -mhard-float -msoft-float @gol
634 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
635 -mdouble -mno-double @gol
636 -mmedia -mno-media -mmuladd -mno-muladd @gol
637 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
638 -mlinked-fp -mlong-calls -malign-labels @gol
639 -mlibrary-pic -macc-4 -macc-8 @gol
640 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
641 -moptimize-membar -mno-optimize-membar @gol
642 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
643 -mvliw-branch -mno-vliw-branch @gol
644 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
645 -mno-nested-cond-exec -mtomcat-stats @gol
649 @emph{GNU/Linux Options}
650 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
651 -tno-android-cc -tno-android-ld}
653 @emph{H8/300 Options}
654 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
657 @gccoptlist{-march=@var{architecture-type} @gol
658 -mdisable-fpregs -mdisable-indexing @gol
659 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
660 -mfixed-range=@var{register-range} @gol
661 -mjump-in-delay -mlinker-opt -mlong-calls @gol
662 -mlong-load-store -mno-disable-fpregs @gol
663 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
664 -mno-jump-in-delay -mno-long-load-store @gol
665 -mno-portable-runtime -mno-soft-float @gol
666 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
667 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
668 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
669 -munix=@var{unix-std} -nolibdld -static -threads}
671 @emph{i386 and x86-64 Options}
672 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
673 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
674 -mfpmath=@var{unit} @gol
675 -masm=@var{dialect} -mno-fancy-math-387 @gol
676 -mno-fp-ret-in-387 -msoft-float @gol
677 -mno-wide-multiply -mrtd -malign-double @gol
678 -mpreferred-stack-boundary=@var{num} @gol
679 -mincoming-stack-boundary=@var{num} @gol
680 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
681 -mrecip -mrecip=@var{opt} @gol
682 -mvzeroupper -mprefer-avx128 @gol
683 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
684 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -msha @gol
685 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma -mprefetchwt1 @gol
686 -mclflushopt -mxsavec -mxsaves @gol
687 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
688 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx -mthreads @gol
689 -mno-align-stringops -minline-all-stringops @gol
690 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
691 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy}
692 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
693 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
694 -mregparm=@var{num} -msseregparm @gol
695 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
696 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
697 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
698 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
699 -m32 -m64 -mx32 -m16 -mlarge-data-threshold=@var{num} @gol
700 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
701 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
702 -mstack-protector-guard=@var{guard}}
704 @emph{i386 and x86-64 Windows Options}
705 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
706 -mnop-fun-dllimport -mthread @gol
707 -municode -mwin32 -mwindows -fno-set-stack-executable}
710 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
711 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
712 -mconstant-gp -mauto-pic -mfused-madd @gol
713 -minline-float-divide-min-latency @gol
714 -minline-float-divide-max-throughput @gol
715 -mno-inline-float-divide @gol
716 -minline-int-divide-min-latency @gol
717 -minline-int-divide-max-throughput @gol
718 -mno-inline-int-divide @gol
719 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
720 -mno-inline-sqrt @gol
721 -mdwarf2-asm -mearly-stop-bits @gol
722 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
723 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
724 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
725 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
726 -msched-spec-ldc -msched-spec-control-ldc @gol
727 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
728 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
729 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
730 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
733 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
734 -msign-extend-enabled -muser-enabled}
736 @emph{M32R/D Options}
737 @gccoptlist{-m32r2 -m32rx -m32r @gol
739 -malign-loops -mno-align-loops @gol
740 -missue-rate=@var{number} @gol
741 -mbranch-cost=@var{number} @gol
742 -mmodel=@var{code-size-model-type} @gol
743 -msdata=@var{sdata-type} @gol
744 -mno-flush-func -mflush-func=@var{name} @gol
745 -mno-flush-trap -mflush-trap=@var{number} @gol
749 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
751 @emph{M680x0 Options}
752 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
753 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
754 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
755 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
756 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
757 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
758 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
759 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
763 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
764 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
765 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
766 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
767 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
770 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
771 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
772 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
773 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
776 @emph{MicroBlaze Options}
777 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
778 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
779 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
780 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
781 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
784 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
785 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
786 -mips64 -mips64r2 -mips64r3 -mips64r5 @gol
787 -mips16 -mno-mips16 -mflip-mips16 @gol
788 -minterlink-compressed -mno-interlink-compressed @gol
789 -minterlink-mips16 -mno-interlink-mips16 @gol
790 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
791 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
792 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
793 -mno-float -msingle-float -mdouble-float @gol
794 -modd-spreg -mno-odd-spreg @gol
795 -mabs=@var{mode} -mnan=@var{encoding} @gol
796 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
799 -mvirt -mno-virt @gol
801 -mmicromips -mno-micromips @gol
802 -mfpu=@var{fpu-type} @gol
803 -msmartmips -mno-smartmips @gol
804 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
805 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
806 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
807 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
808 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
809 -membedded-data -mno-embedded-data @gol
810 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
811 -mcode-readable=@var{setting} @gol
812 -msplit-addresses -mno-split-addresses @gol
813 -mexplicit-relocs -mno-explicit-relocs @gol
814 -mcheck-zero-division -mno-check-zero-division @gol
815 -mdivide-traps -mdivide-breaks @gol
816 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
817 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
818 -mfix-24k -mno-fix-24k @gol
819 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
820 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
821 -mfix-vr4120 -mno-fix-vr4120 @gol
822 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
823 -mflush-func=@var{func} -mno-flush-func @gol
824 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
825 -mfp-exceptions -mno-fp-exceptions @gol
826 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
827 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
830 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
831 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
832 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
833 -mno-base-addresses -msingle-exit -mno-single-exit}
835 @emph{MN10300 Options}
836 @gccoptlist{-mmult-bug -mno-mult-bug @gol
837 -mno-am33 -mam33 -mam33-2 -mam34 @gol
838 -mtune=@var{cpu-type} @gol
839 -mreturn-pointer-on-d0 @gol
840 -mno-crt0 -mrelax -mliw -msetlb}
843 @gccoptlist{-meb -mel -mno-crt0}
845 @emph{MSP430 Options}
846 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
850 @gccoptlist{-mbig-endian -mlittle-endian @gol
851 -mreduced-regs -mfull-regs @gol
852 -mcmov -mno-cmov @gol
853 -mperf-ext -mno-perf-ext @gol
854 -mv3push -mno-v3push @gol
855 -m16bit -mno-16bit @gol
856 -mgp-direct -mno-gp-direct @gol
857 -misr-vector-size=@var{num} @gol
858 -mcache-block-size=@var{num} @gol
859 -march=@var{arch} @gol
860 -mforce-fp-as-gp -mforbid-fp-as-gp @gol
861 -mex9 -mctor-dtor -mrelax}
863 @emph{Nios II Options}
864 @gccoptlist{-G @var{num} -mgpopt -mno-gpopt -mel -meb @gol
865 -mno-bypass-cache -mbypass-cache @gol
866 -mno-cache-volatile -mcache-volatile @gol
867 -mno-fast-sw-div -mfast-sw-div @gol
868 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
869 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
870 -mcustom-fpu-cfg=@var{name} @gol
871 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name}}
873 @emph{PDP-11 Options}
874 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
875 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
876 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
877 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
878 -mbranch-expensive -mbranch-cheap @gol
879 -munix-asm -mdec-asm}
881 @emph{picoChip Options}
882 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
883 -msymbol-as-address -mno-inefficient-warnings}
885 @emph{PowerPC Options}
886 See RS/6000 and PowerPC Options.
889 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=rl78 @gol
890 -m64bit-doubles -m32bit-doubles}
892 @emph{RS/6000 and PowerPC Options}
893 @gccoptlist{-mcpu=@var{cpu-type} @gol
894 -mtune=@var{cpu-type} @gol
895 -mcmodel=@var{code-model} @gol
897 -maltivec -mno-altivec @gol
898 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
899 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
900 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
901 -mfprnd -mno-fprnd @gol
902 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
903 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
904 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
905 -malign-power -malign-natural @gol
906 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
907 -msingle-float -mdouble-float -msimple-fpu @gol
908 -mstring -mno-string -mupdate -mno-update @gol
909 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
910 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
911 -mstrict-align -mno-strict-align -mrelocatable @gol
912 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
913 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
914 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
915 -mprioritize-restricted-insns=@var{priority} @gol
916 -msched-costly-dep=@var{dependence_type} @gol
917 -minsert-sched-nops=@var{scheme} @gol
918 -mcall-sysv -mcall-netbsd @gol
919 -maix-struct-return -msvr4-struct-return @gol
920 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
921 -mblock-move-inline-limit=@var{num} @gol
922 -misel -mno-isel @gol
923 -misel=yes -misel=no @gol
925 -mspe=yes -mspe=no @gol
927 -mgen-cell-microcode -mwarn-cell-microcode @gol
928 -mvrsave -mno-vrsave @gol
929 -mmulhw -mno-mulhw @gol
930 -mdlmzb -mno-dlmzb @gol
931 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
932 -mprototype -mno-prototype @gol
933 -msim -mmvme -mads -myellowknife -memb -msdata @gol
934 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
935 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
936 -mno-recip-precision @gol
937 -mveclibabi=@var{type} -mfriz -mno-friz @gol
938 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
939 -msave-toc-indirect -mno-save-toc-indirect @gol
940 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
941 -mcrypto -mno-crypto -mdirect-move -mno-direct-move @gol
942 -mquad-memory -mno-quad-memory @gol
943 -mquad-memory-atomic -mno-quad-memory-atomic @gol
944 -mcompat-align-parm -mno-compat-align-parm @gol
945 -mupper-regs-df -mno-upper-regs-df -mupper-regs-sf -mno-upper-regs-sf @gol
946 -mupper-regs -mno-upper-regs}
949 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
951 -mbig-endian-data -mlittle-endian-data @gol
954 -mas100-syntax -mno-as100-syntax@gol
956 -mmax-constant-size=@gol
959 -mno-warn-multiple-fast-interrupts@gol
960 -msave-acc-in-interrupts}
962 @emph{S/390 and zSeries Options}
963 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
964 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
965 -mlong-double-64 -mlong-double-128 @gol
966 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
967 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
968 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
969 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
970 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
971 -mhotpatch[=@var{halfwords}] -mno-hotpatch}
974 @gccoptlist{-meb -mel @gol
978 -mscore5 -mscore5u -mscore7 -mscore7d}
981 @gccoptlist{-m1 -m2 -m2e @gol
982 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
984 -m4-nofpu -m4-single-only -m4-single -m4 @gol
985 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
986 -m5-64media -m5-64media-nofpu @gol
987 -m5-32media -m5-32media-nofpu @gol
988 -m5-compact -m5-compact-nofpu @gol
989 -mb -ml -mdalign -mrelax @gol
990 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
991 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
992 -mspace -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
993 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
994 -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
995 -maccumulate-outgoing-args -minvalid-symbols @gol
996 -matomic-model=@var{atomic-model} @gol
997 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
998 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
999 -mpretend-cmove -mtas}
1001 @emph{Solaris 2 Options}
1002 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1005 @emph{SPARC Options}
1006 @gccoptlist{-mcpu=@var{cpu-type} @gol
1007 -mtune=@var{cpu-type} @gol
1008 -mcmodel=@var{code-model} @gol
1009 -mmemory-model=@var{mem-model} @gol
1010 -m32 -m64 -mapp-regs -mno-app-regs @gol
1011 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1012 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1013 -mhard-quad-float -msoft-quad-float @gol
1014 -mstack-bias -mno-stack-bias @gol
1015 -munaligned-doubles -mno-unaligned-doubles @gol
1016 -muser-mode -mno-user-mode @gol
1017 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1018 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1019 -mcbcond -mno-cbcond @gol
1020 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
1021 -mfix-at697f -mfix-ut699}
1024 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1025 -msafe-dma -munsafe-dma @gol
1027 -msmall-mem -mlarge-mem -mstdmain @gol
1028 -mfixed-range=@var{register-range} @gol
1030 -maddress-space-conversion -mno-address-space-conversion @gol
1031 -mcache-size=@var{cache-size} @gol
1032 -matomic-updates -mno-atomic-updates}
1034 @emph{System V Options}
1035 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1037 @emph{TILE-Gx Options}
1038 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1039 -mcmodel=@var{code-model}}
1041 @emph{TILEPro Options}
1042 @gccoptlist{-mcpu=@var{cpu} -m32}
1045 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1046 -mprolog-function -mno-prolog-function -mspace @gol
1047 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1048 -mapp-regs -mno-app-regs @gol
1049 -mdisable-callt -mno-disable-callt @gol
1050 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1051 -mv850e -mv850 -mv850e3v5 @gol
1062 @gccoptlist{-mg -mgnu -munix}
1065 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1066 -mpointer-size=@var{size}}
1068 @emph{VxWorks Options}
1069 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1070 -Xbind-lazy -Xbind-now}
1072 @emph{x86-64 Options}
1073 See i386 and x86-64 Options.
1075 @emph{Xstormy16 Options}
1078 @emph{Xtensa Options}
1079 @gccoptlist{-mconst16 -mno-const16 @gol
1080 -mfused-madd -mno-fused-madd @gol
1082 -mserialize-volatile -mno-serialize-volatile @gol
1083 -mtext-section-literals -mno-text-section-literals @gol
1084 -mtarget-align -mno-target-align @gol
1085 -mlongcalls -mno-longcalls}
1087 @emph{zSeries Options}
1088 See S/390 and zSeries Options.
1090 @item Code Generation Options
1091 @xref{Code Gen Options,,Options for Code Generation Conventions}.
1092 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
1093 -ffixed-@var{reg} -fexceptions @gol
1094 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
1095 -fasynchronous-unwind-tables @gol
1096 -fno-gnu-unique @gol
1097 -finhibit-size-directive -finstrument-functions @gol
1098 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
1099 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
1100 -fno-common -fno-ident @gol
1101 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
1102 -fno-jump-tables @gol
1103 -frecord-gcc-switches @gol
1104 -freg-struct-return -fshort-enums @gol
1105 -fshort-double -fshort-wchar @gol
1106 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
1107 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
1108 -fno-stack-limit -fsplit-stack @gol
1109 -fleading-underscore -ftls-model=@var{model} @gol
1110 -fstack-reuse=@var{reuse_level} @gol
1111 -ftrapv -fwrapv -fbounds-check @gol
1112 -fvisibility -fstrict-volatile-bitfields -fsync-libcalls}
1116 @node Overall Options
1117 @section Options Controlling the Kind of Output
1119 Compilation can involve up to four stages: preprocessing, compilation
1120 proper, assembly and linking, always in that order. GCC is capable of
1121 preprocessing and compiling several files either into several
1122 assembler input files, or into one assembler input file; then each
1123 assembler input file produces an object file, and linking combines all
1124 the object files (those newly compiled, and those specified as input)
1125 into an executable file.
1127 @cindex file name suffix
1128 For any given input file, the file name suffix determines what kind of
1129 compilation is done:
1133 C source code that must be preprocessed.
1136 C source code that should not be preprocessed.
1139 C++ source code that should not be preprocessed.
1142 Objective-C source code. Note that you must link with the @file{libobjc}
1143 library to make an Objective-C program work.
1146 Objective-C source code that should not be preprocessed.
1150 Objective-C++ source code. Note that you must link with the @file{libobjc}
1151 library to make an Objective-C++ program work. Note that @samp{.M} refers
1152 to a literal capital M@.
1154 @item @var{file}.mii
1155 Objective-C++ source code that should not be preprocessed.
1158 C, C++, Objective-C or Objective-C++ header file to be turned into a
1159 precompiled header (default), or C, C++ header file to be turned into an
1160 Ada spec (via the @option{-fdump-ada-spec} switch).
1163 @itemx @var{file}.cp
1164 @itemx @var{file}.cxx
1165 @itemx @var{file}.cpp
1166 @itemx @var{file}.CPP
1167 @itemx @var{file}.c++
1169 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1170 the last two letters must both be literally @samp{x}. Likewise,
1171 @samp{.C} refers to a literal capital C@.
1175 Objective-C++ source code that must be preprocessed.
1177 @item @var{file}.mii
1178 Objective-C++ source code that should not be preprocessed.
1182 @itemx @var{file}.hp
1183 @itemx @var{file}.hxx
1184 @itemx @var{file}.hpp
1185 @itemx @var{file}.HPP
1186 @itemx @var{file}.h++
1187 @itemx @var{file}.tcc
1188 C++ header file to be turned into a precompiled header or Ada spec.
1191 @itemx @var{file}.for
1192 @itemx @var{file}.ftn
1193 Fixed form Fortran source code that should not be preprocessed.
1196 @itemx @var{file}.FOR
1197 @itemx @var{file}.fpp
1198 @itemx @var{file}.FPP
1199 @itemx @var{file}.FTN
1200 Fixed form Fortran source code that must be preprocessed (with the traditional
1203 @item @var{file}.f90
1204 @itemx @var{file}.f95
1205 @itemx @var{file}.f03
1206 @itemx @var{file}.f08
1207 Free form Fortran source code that should not be preprocessed.
1209 @item @var{file}.F90
1210 @itemx @var{file}.F95
1211 @itemx @var{file}.F03
1212 @itemx @var{file}.F08
1213 Free form Fortran source code that must be preprocessed (with the
1214 traditional preprocessor).
1219 @c FIXME: Descriptions of Java file types.
1225 @item @var{file}.ads
1226 Ada source code file that contains a library unit declaration (a
1227 declaration of a package, subprogram, or generic, or a generic
1228 instantiation), or a library unit renaming declaration (a package,
1229 generic, or subprogram renaming declaration). Such files are also
1232 @item @var{file}.adb
1233 Ada source code file containing a library unit body (a subprogram or
1234 package body). Such files are also called @dfn{bodies}.
1236 @c GCC also knows about some suffixes for languages not yet included:
1247 @itemx @var{file}.sx
1248 Assembler code that must be preprocessed.
1251 An object file to be fed straight into linking.
1252 Any file name with no recognized suffix is treated this way.
1256 You can specify the input language explicitly with the @option{-x} option:
1259 @item -x @var{language}
1260 Specify explicitly the @var{language} for the following input files
1261 (rather than letting the compiler choose a default based on the file
1262 name suffix). This option applies to all following input files until
1263 the next @option{-x} option. Possible values for @var{language} are:
1265 c c-header cpp-output
1266 c++ c++-header c++-cpp-output
1267 objective-c objective-c-header objective-c-cpp-output
1268 objective-c++ objective-c++-header objective-c++-cpp-output
1269 assembler assembler-with-cpp
1271 f77 f77-cpp-input f95 f95-cpp-input
1277 Turn off any specification of a language, so that subsequent files are
1278 handled according to their file name suffixes (as they are if @option{-x}
1279 has not been used at all).
1281 @item -pass-exit-codes
1282 @opindex pass-exit-codes
1283 Normally the @command{gcc} program exits with the code of 1 if any
1284 phase of the compiler returns a non-success return code. If you specify
1285 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1286 the numerically highest error produced by any phase returning an error
1287 indication. The C, C++, and Fortran front ends return 4 if an internal
1288 compiler error is encountered.
1291 If you only want some of the stages of compilation, you can use
1292 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1293 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1294 @command{gcc} is to stop. Note that some combinations (for example,
1295 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1300 Compile or assemble the source files, but do not link. The linking
1301 stage simply is not done. The ultimate output is in the form of an
1302 object file for each source file.
1304 By default, the object file name for a source file is made by replacing
1305 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1307 Unrecognized input files, not requiring compilation or assembly, are
1312 Stop after the stage of compilation proper; do not assemble. The output
1313 is in the form of an assembler code file for each non-assembler input
1316 By default, the assembler file name for a source file is made by
1317 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1319 Input files that don't require compilation are ignored.
1323 Stop after the preprocessing stage; do not run the compiler proper. The
1324 output is in the form of preprocessed source code, which is sent to the
1327 Input files that don't require preprocessing are ignored.
1329 @cindex output file option
1332 Place output in file @var{file}. This applies to whatever
1333 sort of output is being produced, whether it be an executable file,
1334 an object file, an assembler file or preprocessed C code.
1336 If @option{-o} is not specified, the default is to put an executable
1337 file in @file{a.out}, the object file for
1338 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1339 assembler file in @file{@var{source}.s}, a precompiled header file in
1340 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1345 Print (on standard error output) the commands executed to run the stages
1346 of compilation. Also print the version number of the compiler driver
1347 program and of the preprocessor and the compiler proper.
1351 Like @option{-v} except the commands are not executed and arguments
1352 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1353 This is useful for shell scripts to capture the driver-generated command lines.
1357 Use pipes rather than temporary files for communication between the
1358 various stages of compilation. This fails to work on some systems where
1359 the assembler is unable to read from a pipe; but the GNU assembler has
1364 Print (on the standard output) a description of the command-line options
1365 understood by @command{gcc}. If the @option{-v} option is also specified
1366 then @option{--help} is also passed on to the various processes
1367 invoked by @command{gcc}, so that they can display the command-line options
1368 they accept. If the @option{-Wextra} option has also been specified
1369 (prior to the @option{--help} option), then command-line options that
1370 have no documentation associated with them are also displayed.
1373 @opindex target-help
1374 Print (on the standard output) a description of target-specific command-line
1375 options for each tool. For some targets extra target-specific
1376 information may also be printed.
1378 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1379 Print (on the standard output) a description of the command-line
1380 options understood by the compiler that fit into all specified classes
1381 and qualifiers. These are the supported classes:
1384 @item @samp{optimizers}
1385 Display all of the optimization options supported by the
1388 @item @samp{warnings}
1389 Display all of the options controlling warning messages
1390 produced by the compiler.
1393 Display target-specific options. Unlike the
1394 @option{--target-help} option however, target-specific options of the
1395 linker and assembler are not displayed. This is because those
1396 tools do not currently support the extended @option{--help=} syntax.
1399 Display the values recognized by the @option{--param}
1402 @item @var{language}
1403 Display the options supported for @var{language}, where
1404 @var{language} is the name of one of the languages supported in this
1408 Display the options that are common to all languages.
1411 These are the supported qualifiers:
1414 @item @samp{undocumented}
1415 Display only those options that are undocumented.
1418 Display options taking an argument that appears after an equal
1419 sign in the same continuous piece of text, such as:
1420 @samp{--help=target}.
1422 @item @samp{separate}
1423 Display options taking an argument that appears as a separate word
1424 following the original option, such as: @samp{-o output-file}.
1427 Thus for example to display all the undocumented target-specific
1428 switches supported by the compiler, use:
1431 --help=target,undocumented
1434 The sense of a qualifier can be inverted by prefixing it with the
1435 @samp{^} character, so for example to display all binary warning
1436 options (i.e., ones that are either on or off and that do not take an
1437 argument) that have a description, use:
1440 --help=warnings,^joined,^undocumented
1443 The argument to @option{--help=} should not consist solely of inverted
1446 Combining several classes is possible, although this usually
1447 restricts the output so much that there is nothing to display. One
1448 case where it does work, however, is when one of the classes is
1449 @var{target}. For example, to display all the target-specific
1450 optimization options, use:
1453 --help=target,optimizers
1456 The @option{--help=} option can be repeated on the command line. Each
1457 successive use displays its requested class of options, skipping
1458 those that have already been displayed.
1460 If the @option{-Q} option appears on the command line before the
1461 @option{--help=} option, then the descriptive text displayed by
1462 @option{--help=} is changed. Instead of describing the displayed
1463 options, an indication is given as to whether the option is enabled,
1464 disabled or set to a specific value (assuming that the compiler
1465 knows this at the point where the @option{--help=} option is used).
1467 Here is a truncated example from the ARM port of @command{gcc}:
1470 % gcc -Q -mabi=2 --help=target -c
1471 The following options are target specific:
1473 -mabort-on-noreturn [disabled]
1477 The output is sensitive to the effects of previous command-line
1478 options, so for example it is possible to find out which optimizations
1479 are enabled at @option{-O2} by using:
1482 -Q -O2 --help=optimizers
1485 Alternatively you can discover which binary optimizations are enabled
1486 by @option{-O3} by using:
1489 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1490 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1491 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1494 @item -no-canonical-prefixes
1495 @opindex no-canonical-prefixes
1496 Do not expand any symbolic links, resolve references to @samp{/../}
1497 or @samp{/./}, or make the path absolute when generating a relative
1502 Display the version number and copyrights of the invoked GCC@.
1506 Invoke all subcommands under a wrapper program. The name of the
1507 wrapper program and its parameters are passed as a comma separated
1511 gcc -c t.c -wrapper gdb,--args
1515 This invokes all subprograms of @command{gcc} under
1516 @samp{gdb --args}, thus the invocation of @command{cc1} is
1517 @samp{gdb --args cc1 @dots{}}.
1519 @item -fplugin=@var{name}.so
1521 Load the plugin code in file @var{name}.so, assumed to be a
1522 shared object to be dlopen'd by the compiler. The base name of
1523 the shared object file is used to identify the plugin for the
1524 purposes of argument parsing (See
1525 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1526 Each plugin should define the callback functions specified in the
1529 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1530 @opindex fplugin-arg
1531 Define an argument called @var{key} with a value of @var{value}
1532 for the plugin called @var{name}.
1534 @item -fdump-ada-spec@r{[}-slim@r{]}
1535 @opindex fdump-ada-spec
1536 For C and C++ source and include files, generate corresponding Ada specs.
1537 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1538 GNAT User's Guide}, which provides detailed documentation on this feature.
1540 @item -fada-spec-parent=@var{unit}
1541 @opindex fada-spec-parent
1542 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1543 Ada specs as child units of parent @var{unit}.
1545 @item -fdump-go-spec=@var{file}
1546 @opindex fdump-go-spec
1547 For input files in any language, generate corresponding Go
1548 declarations in @var{file}. This generates Go @code{const},
1549 @code{type}, @code{var}, and @code{func} declarations which may be a
1550 useful way to start writing a Go interface to code written in some
1553 @include @value{srcdir}/../libiberty/at-file.texi
1557 @section Compiling C++ Programs
1559 @cindex suffixes for C++ source
1560 @cindex C++ source file suffixes
1561 C++ source files conventionally use one of the suffixes @samp{.C},
1562 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1563 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1564 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1565 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1566 files with these names and compiles them as C++ programs even if you
1567 call the compiler the same way as for compiling C programs (usually
1568 with the name @command{gcc}).
1572 However, the use of @command{gcc} does not add the C++ library.
1573 @command{g++} is a program that calls GCC and automatically specifies linking
1574 against the C++ library. It treats @samp{.c},
1575 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1576 files unless @option{-x} is used. This program is also useful when
1577 precompiling a C header file with a @samp{.h} extension for use in C++
1578 compilations. On many systems, @command{g++} is also installed with
1579 the name @command{c++}.
1581 @cindex invoking @command{g++}
1582 When you compile C++ programs, you may specify many of the same
1583 command-line options that you use for compiling programs in any
1584 language; or command-line options meaningful for C and related
1585 languages; or options that are meaningful only for C++ programs.
1586 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1587 explanations of options for languages related to C@.
1588 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1589 explanations of options that are meaningful only for C++ programs.
1591 @node C Dialect Options
1592 @section Options Controlling C Dialect
1593 @cindex dialect options
1594 @cindex language dialect options
1595 @cindex options, dialect
1597 The following options control the dialect of C (or languages derived
1598 from C, such as C++, Objective-C and Objective-C++) that the compiler
1602 @cindex ANSI support
1606 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1607 equivalent to @option{-std=c++98}.
1609 This turns off certain features of GCC that are incompatible with ISO
1610 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1611 such as the @code{asm} and @code{typeof} keywords, and
1612 predefined macros such as @code{unix} and @code{vax} that identify the
1613 type of system you are using. It also enables the undesirable and
1614 rarely used ISO trigraph feature. For the C compiler,
1615 it disables recognition of C++ style @samp{//} comments as well as
1616 the @code{inline} keyword.
1618 The alternate keywords @code{__asm__}, @code{__extension__},
1619 @code{__inline__} and @code{__typeof__} continue to work despite
1620 @option{-ansi}. You would not want to use them in an ISO C program, of
1621 course, but it is useful to put them in header files that might be included
1622 in compilations done with @option{-ansi}. Alternate predefined macros
1623 such as @code{__unix__} and @code{__vax__} are also available, with or
1624 without @option{-ansi}.
1626 The @option{-ansi} option does not cause non-ISO programs to be
1627 rejected gratuitously. For that, @option{-Wpedantic} is required in
1628 addition to @option{-ansi}. @xref{Warning Options}.
1630 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1631 option is used. Some header files may notice this macro and refrain
1632 from declaring certain functions or defining certain macros that the
1633 ISO standard doesn't call for; this is to avoid interfering with any
1634 programs that might use these names for other things.
1636 Functions that are normally built in but do not have semantics
1637 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1638 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1639 built-in functions provided by GCC}, for details of the functions
1644 Determine the language standard. @xref{Standards,,Language Standards
1645 Supported by GCC}, for details of these standard versions. This option
1646 is currently only supported when compiling C or C++.
1648 The compiler can accept several base standards, such as @samp{c90} or
1649 @samp{c++98}, and GNU dialects of those standards, such as
1650 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1651 compiler accepts all programs following that standard plus those
1652 using GNU extensions that do not contradict it. For example,
1653 @option{-std=c90} turns off certain features of GCC that are
1654 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1655 keywords, but not other GNU extensions that do not have a meaning in
1656 ISO C90, such as omitting the middle term of a @code{?:}
1657 expression. On the other hand, when a GNU dialect of a standard is
1658 specified, all features supported by the compiler are enabled, even when
1659 those features change the meaning of the base standard. As a result, some
1660 strict-conforming programs may be rejected. The particular standard
1661 is used by @option{-Wpedantic} to identify which features are GNU
1662 extensions given that version of the standard. For example
1663 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1664 comments, while @option{-std=gnu99 -Wpedantic} does not.
1666 A value for this option must be provided; possible values are
1672 Support all ISO C90 programs (certain GNU extensions that conflict
1673 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1675 @item iso9899:199409
1676 ISO C90 as modified in amendment 1.
1682 ISO C99. This standard is substantially completely supported, modulo
1683 bugs and floating-point issues
1684 (mainly but not entirely relating to optional C99 features from
1685 Annexes F and G). See
1686 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1687 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1692 ISO C11, the 2011 revision of the ISO C standard. This standard is
1693 substantially completely supported, modulo bugs, floating-point issues
1694 (mainly but not entirely relating to optional C11 features from
1695 Annexes F and G) and the optional Annexes K (Bounds-checking
1696 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1700 GNU dialect of ISO C90 (including some C99 features).
1704 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1708 GNU dialect of ISO C11. This is the default for C code.
1709 The name @samp{gnu1x} is deprecated.
1713 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1714 additional defect reports. Same as @option{-ansi} for C++ code.
1718 GNU dialect of @option{-std=c++98}. This is the default for
1723 The 2011 ISO C++ standard plus amendments.
1724 The name @samp{c++0x} is deprecated.
1728 GNU dialect of @option{-std=c++11}.
1729 The name @samp{gnu++0x} is deprecated.
1733 The 2014 ISO C++ standard plus amendments.
1734 The name @samp{c++1y} is deprecated.
1738 GNU dialect of @option{-std=c++14}.
1739 The name @samp{gnu++1y} is deprecated.
1742 The next revision of the ISO C++ standard, tentatively planned for
1743 2017. Support is highly experimental, and will almost certainly
1744 change in incompatible ways in future releases.
1747 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1748 and will almost certainly change in incompatible ways in future
1752 @item -fgnu89-inline
1753 @opindex fgnu89-inline
1754 The option @option{-fgnu89-inline} tells GCC to use the traditional
1755 GNU semantics for @code{inline} functions when in C99 mode.
1756 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1757 is accepted and ignored by GCC versions 4.1.3 up to but not including
1758 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1759 C99 mode. Using this option is roughly equivalent to adding the
1760 @code{gnu_inline} function attribute to all inline functions
1761 (@pxref{Function Attributes}).
1763 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1764 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1765 specifies the default behavior). This option was first supported in
1766 GCC 4.3. This option is not supported in @option{-std=c90} or
1767 @option{-std=gnu90} mode.
1769 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1770 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1771 in effect for @code{inline} functions. @xref{Common Predefined
1772 Macros,,,cpp,The C Preprocessor}.
1774 @item -aux-info @var{filename}
1776 Output to the given filename prototyped declarations for all functions
1777 declared and/or defined in a translation unit, including those in header
1778 files. This option is silently ignored in any language other than C@.
1780 Besides declarations, the file indicates, in comments, the origin of
1781 each declaration (source file and line), whether the declaration was
1782 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1783 @samp{O} for old, respectively, in the first character after the line
1784 number and the colon), and whether it came from a declaration or a
1785 definition (@samp{C} or @samp{F}, respectively, in the following
1786 character). In the case of function definitions, a K&R-style list of
1787 arguments followed by their declarations is also provided, inside
1788 comments, after the declaration.
1790 @item -fallow-parameterless-variadic-functions
1791 @opindex fallow-parameterless-variadic-functions
1792 Accept variadic functions without named parameters.
1794 Although it is possible to define such a function, this is not very
1795 useful as it is not possible to read the arguments. This is only
1796 supported for C as this construct is allowed by C++.
1800 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1801 keyword, so that code can use these words as identifiers. You can use
1802 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1803 instead. @option{-ansi} implies @option{-fno-asm}.
1805 In C++, this switch only affects the @code{typeof} keyword, since
1806 @code{asm} and @code{inline} are standard keywords. You may want to
1807 use the @option{-fno-gnu-keywords} flag instead, which has the same
1808 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1809 switch only affects the @code{asm} and @code{typeof} keywords, since
1810 @code{inline} is a standard keyword in ISO C99.
1813 @itemx -fno-builtin-@var{function}
1814 @opindex fno-builtin
1815 @cindex built-in functions
1816 Don't recognize built-in functions that do not begin with
1817 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1818 functions provided by GCC}, for details of the functions affected,
1819 including those which are not built-in functions when @option{-ansi} or
1820 @option{-std} options for strict ISO C conformance are used because they
1821 do not have an ISO standard meaning.
1823 GCC normally generates special code to handle certain built-in functions
1824 more efficiently; for instance, calls to @code{alloca} may become single
1825 instructions which adjust the stack directly, and calls to @code{memcpy}
1826 may become inline copy loops. The resulting code is often both smaller
1827 and faster, but since the function calls no longer appear as such, you
1828 cannot set a breakpoint on those calls, nor can you change the behavior
1829 of the functions by linking with a different library. In addition,
1830 when a function is recognized as a built-in function, GCC may use
1831 information about that function to warn about problems with calls to
1832 that function, or to generate more efficient code, even if the
1833 resulting code still contains calls to that function. For example,
1834 warnings are given with @option{-Wformat} for bad calls to
1835 @code{printf} when @code{printf} is built in and @code{strlen} is
1836 known not to modify global memory.
1838 With the @option{-fno-builtin-@var{function}} option
1839 only the built-in function @var{function} is
1840 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1841 function is named that is not built-in in this version of GCC, this
1842 option is ignored. There is no corresponding
1843 @option{-fbuiltin-@var{function}} option; if you wish to enable
1844 built-in functions selectively when using @option{-fno-builtin} or
1845 @option{-ffreestanding}, you may define macros such as:
1848 #define abs(n) __builtin_abs ((n))
1849 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1854 @cindex hosted environment
1856 Assert that compilation targets a hosted environment. This implies
1857 @option{-fbuiltin}. A hosted environment is one in which the
1858 entire standard library is available, and in which @code{main} has a return
1859 type of @code{int}. Examples are nearly everything except a kernel.
1860 This is equivalent to @option{-fno-freestanding}.
1862 @item -ffreestanding
1863 @opindex ffreestanding
1864 @cindex hosted environment
1866 Assert that compilation targets a freestanding environment. This
1867 implies @option{-fno-builtin}. A freestanding environment
1868 is one in which the standard library may not exist, and program startup may
1869 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1870 This is equivalent to @option{-fno-hosted}.
1872 @xref{Standards,,Language Standards Supported by GCC}, for details of
1873 freestanding and hosted environments.
1877 @cindex OpenMP parallel
1878 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1879 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1880 compiler generates parallel code according to the OpenMP Application
1881 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
1882 implies @option{-pthread}, and thus is only supported on targets that
1883 have support for @option{-pthread}. @option{-fopenmp} implies
1884 @option{-fopenmp-simd}.
1887 @opindex fopenmp-simd
1890 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
1891 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
1896 @cindex Enable Cilk Plus
1897 Enable the usage of Cilk Plus language extension features for C/C++.
1898 When the option @option{-fcilkplus} is specified, enable the usage of
1899 the Cilk Plus Language extension features for C/C++. The present
1900 implementation follows ABI version 1.2. This is an experimental
1901 feature that is only partially complete, and whose interface may
1902 change in future versions of GCC as the official specification
1903 changes. Currently, all features but @code{_Cilk_for} have been
1908 When the option @option{-fgnu-tm} is specified, the compiler
1909 generates code for the Linux variant of Intel's current Transactional
1910 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1911 an experimental feature whose interface may change in future versions
1912 of GCC, as the official specification changes. Please note that not
1913 all architectures are supported for this feature.
1915 For more information on GCC's support for transactional memory,
1916 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1917 Transactional Memory Library}.
1919 Note that the transactional memory feature is not supported with
1920 non-call exceptions (@option{-fnon-call-exceptions}).
1922 @item -fms-extensions
1923 @opindex fms-extensions
1924 Accept some non-standard constructs used in Microsoft header files.
1926 In C++ code, this allows member names in structures to be similar
1927 to previous types declarations.
1936 Some cases of unnamed fields in structures and unions are only
1937 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1938 fields within structs/unions}, for details.
1940 Note that this option is off for all targets but i?86 and x86_64
1941 targets using ms-abi.
1942 @item -fplan9-extensions
1943 Accept some non-standard constructs used in Plan 9 code.
1945 This enables @option{-fms-extensions}, permits passing pointers to
1946 structures with anonymous fields to functions that expect pointers to
1947 elements of the type of the field, and permits referring to anonymous
1948 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1949 struct/union fields within structs/unions}, for details. This is only
1950 supported for C, not C++.
1954 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1955 options for strict ISO C conformance) implies @option{-trigraphs}.
1957 @cindex traditional C language
1958 @cindex C language, traditional
1960 @itemx -traditional-cpp
1961 @opindex traditional-cpp
1962 @opindex traditional
1963 Formerly, these options caused GCC to attempt to emulate a pre-standard
1964 C compiler. They are now only supported with the @option{-E} switch.
1965 The preprocessor continues to support a pre-standard mode. See the GNU
1966 CPP manual for details.
1968 @item -fcond-mismatch
1969 @opindex fcond-mismatch
1970 Allow conditional expressions with mismatched types in the second and
1971 third arguments. The value of such an expression is void. This option
1972 is not supported for C++.
1974 @item -flax-vector-conversions
1975 @opindex flax-vector-conversions
1976 Allow implicit conversions between vectors with differing numbers of
1977 elements and/or incompatible element types. This option should not be
1980 @item -funsigned-char
1981 @opindex funsigned-char
1982 Let the type @code{char} be unsigned, like @code{unsigned char}.
1984 Each kind of machine has a default for what @code{char} should
1985 be. It is either like @code{unsigned char} by default or like
1986 @code{signed char} by default.
1988 Ideally, a portable program should always use @code{signed char} or
1989 @code{unsigned char} when it depends on the signedness of an object.
1990 But many programs have been written to use plain @code{char} and
1991 expect it to be signed, or expect it to be unsigned, depending on the
1992 machines they were written for. This option, and its inverse, let you
1993 make such a program work with the opposite default.
1995 The type @code{char} is always a distinct type from each of
1996 @code{signed char} or @code{unsigned char}, even though its behavior
1997 is always just like one of those two.
2000 @opindex fsigned-char
2001 Let the type @code{char} be signed, like @code{signed char}.
2003 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2004 the negative form of @option{-funsigned-char}. Likewise, the option
2005 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2007 @item -fsigned-bitfields
2008 @itemx -funsigned-bitfields
2009 @itemx -fno-signed-bitfields
2010 @itemx -fno-unsigned-bitfields
2011 @opindex fsigned-bitfields
2012 @opindex funsigned-bitfields
2013 @opindex fno-signed-bitfields
2014 @opindex fno-unsigned-bitfields
2015 These options control whether a bit-field is signed or unsigned, when the
2016 declaration does not use either @code{signed} or @code{unsigned}. By
2017 default, such a bit-field is signed, because this is consistent: the
2018 basic integer types such as @code{int} are signed types.
2021 @node C++ Dialect Options
2022 @section Options Controlling C++ Dialect
2024 @cindex compiler options, C++
2025 @cindex C++ options, command-line
2026 @cindex options, C++
2027 This section describes the command-line options that are only meaningful
2028 for C++ programs. You can also use most of the GNU compiler options
2029 regardless of what language your program is in. For example, you
2030 might compile a file @code{firstClass.C} like this:
2033 g++ -g -frepo -O -c firstClass.C
2037 In this example, only @option{-frepo} is an option meant
2038 only for C++ programs; you can use the other options with any
2039 language supported by GCC@.
2041 Here is a list of options that are @emph{only} for compiling C++ programs:
2045 @item -fabi-version=@var{n}
2046 @opindex fabi-version
2047 Use version @var{n} of the C++ ABI@. The default is version 0.
2049 Version 0 refers to the version conforming most closely to
2050 the C++ ABI specification. Therefore, the ABI obtained using version 0
2051 will change in different versions of G++ as ABI bugs are fixed.
2053 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2055 Version 2 is the version of the C++ ABI that first appeared in G++
2056 3.4, and was the default through G++ 4.9.
2058 Version 3 corrects an error in mangling a constant address as a
2061 Version 4, which first appeared in G++ 4.5, implements a standard
2062 mangling for vector types.
2064 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2065 attribute const/volatile on function pointer types, decltype of a
2066 plain decl, and use of a function parameter in the declaration of
2069 Version 6, which first appeared in G++ 4.7, corrects the promotion
2070 behavior of C++11 scoped enums and the mangling of template argument
2071 packs, const/static_cast, prefix ++ and --, and a class scope function
2072 used as a template argument.
2074 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2075 builtin type and corrects the mangling of lambdas in default argument
2078 Version 8, which first appeared in G++ 4.9, corrects the substitution
2079 behavior of function types with function-cv-qualifiers.
2081 See also @option{-Wabi}.
2083 @item -fabi-compat-version=@var{n}
2084 @opindex fabi-compat-version
2085 Starting with GCC 4.5, on targets that support strong aliases, G++
2086 works around mangling changes by creating an alias with the correct
2087 mangled name when defining a symbol with an incorrect mangled name.
2088 This switch specifies which ABI version to use for the alias.
2090 With @option{-fabi-version=0} (the default), this defaults to 2. If
2091 another ABI version is explicitly selected, this defaults to 0.
2093 The compatibility version is also set by @option{-Wabi=@var{n}}.
2095 @item -fno-access-control
2096 @opindex fno-access-control
2097 Turn off all access checking. This switch is mainly useful for working
2098 around bugs in the access control code.
2102 Check that the pointer returned by @code{operator new} is non-null
2103 before attempting to modify the storage allocated. This check is
2104 normally unnecessary because the C++ standard specifies that
2105 @code{operator new} only returns @code{0} if it is declared
2106 @samp{throw()}, in which case the compiler always checks the
2107 return value even without this option. In all other cases, when
2108 @code{operator new} has a non-empty exception specification, memory
2109 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2110 @samp{new (nothrow)}.
2112 @item -fconstexpr-depth=@var{n}
2113 @opindex fconstexpr-depth
2114 Set the maximum nested evaluation depth for C++11 constexpr functions
2115 to @var{n}. A limit is needed to detect endless recursion during
2116 constant expression evaluation. The minimum specified by the standard
2119 @item -fdeduce-init-list
2120 @opindex fdeduce-init-list
2121 Enable deduction of a template type parameter as
2122 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2125 template <class T> auto forward(T t) -> decltype (realfn (t))
2132 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2136 This deduction was implemented as a possible extension to the
2137 originally proposed semantics for the C++11 standard, but was not part
2138 of the final standard, so it is disabled by default. This option is
2139 deprecated, and may be removed in a future version of G++.
2141 @item -ffriend-injection
2142 @opindex ffriend-injection
2143 Inject friend functions into the enclosing namespace, so that they are
2144 visible outside the scope of the class in which they are declared.
2145 Friend functions were documented to work this way in the old Annotated
2146 C++ Reference Manual, and versions of G++ before 4.1 always worked
2147 that way. However, in ISO C++ a friend function that is not declared
2148 in an enclosing scope can only be found using argument dependent
2149 lookup. This option causes friends to be injected as they were in
2152 This option is for compatibility, and may be removed in a future
2155 @item -fno-elide-constructors
2156 @opindex fno-elide-constructors
2157 The C++ standard allows an implementation to omit creating a temporary
2158 that is only used to initialize another object of the same type.
2159 Specifying this option disables that optimization, and forces G++ to
2160 call the copy constructor in all cases.
2162 @item -fno-enforce-eh-specs
2163 @opindex fno-enforce-eh-specs
2164 Don't generate code to check for violation of exception specifications
2165 at run time. This option violates the C++ standard, but may be useful
2166 for reducing code size in production builds, much like defining
2167 @samp{NDEBUG}. This does not give user code permission to throw
2168 exceptions in violation of the exception specifications; the compiler
2169 still optimizes based on the specifications, so throwing an
2170 unexpected exception results in undefined behavior at run time.
2172 @item -fextern-tls-init
2173 @itemx -fno-extern-tls-init
2174 @opindex fextern-tls-init
2175 @opindex fno-extern-tls-init
2176 The C++11 and OpenMP standards allow @samp{thread_local} and
2177 @samp{threadprivate} variables to have dynamic (runtime)
2178 initialization. To support this, any use of such a variable goes
2179 through a wrapper function that performs any necessary initialization.
2180 When the use and definition of the variable are in the same
2181 translation unit, this overhead can be optimized away, but when the
2182 use is in a different translation unit there is significant overhead
2183 even if the variable doesn't actually need dynamic initialization. If
2184 the programmer can be sure that no use of the variable in a
2185 non-defining TU needs to trigger dynamic initialization (either
2186 because the variable is statically initialized, or a use of the
2187 variable in the defining TU will be executed before any uses in
2188 another TU), they can avoid this overhead with the
2189 @option{-fno-extern-tls-init} option.
2191 On targets that support symbol aliases, the default is
2192 @option{-fextern-tls-init}. On targets that do not support symbol
2193 aliases, the default is @option{-fno-extern-tls-init}.
2196 @itemx -fno-for-scope
2198 @opindex fno-for-scope
2199 If @option{-ffor-scope} is specified, the scope of variables declared in
2200 a @i{for-init-statement} is limited to the @samp{for} loop itself,
2201 as specified by the C++ standard.
2202 If @option{-fno-for-scope} is specified, the scope of variables declared in
2203 a @i{for-init-statement} extends to the end of the enclosing scope,
2204 as was the case in old versions of G++, and other (traditional)
2205 implementations of C++.
2207 If neither flag is given, the default is to follow the standard,
2208 but to allow and give a warning for old-style code that would
2209 otherwise be invalid, or have different behavior.
2211 @item -fno-gnu-keywords
2212 @opindex fno-gnu-keywords
2213 Do not recognize @code{typeof} as a keyword, so that code can use this
2214 word as an identifier. You can use the keyword @code{__typeof__} instead.
2215 @option{-ansi} implies @option{-fno-gnu-keywords}.
2217 @item -fno-implicit-templates
2218 @opindex fno-implicit-templates
2219 Never emit code for non-inline templates that are instantiated
2220 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2221 @xref{Template Instantiation}, for more information.
2223 @item -fno-implicit-inline-templates
2224 @opindex fno-implicit-inline-templates
2225 Don't emit code for implicit instantiations of inline templates, either.
2226 The default is to handle inlines differently so that compiles with and
2227 without optimization need the same set of explicit instantiations.
2229 @item -fno-implement-inlines
2230 @opindex fno-implement-inlines
2231 To save space, do not emit out-of-line copies of inline functions
2232 controlled by @samp{#pragma implementation}. This causes linker
2233 errors if these functions are not inlined everywhere they are called.
2235 @item -fms-extensions
2236 @opindex fms-extensions
2237 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2238 int and getting a pointer to member function via non-standard syntax.
2240 @item -fno-nonansi-builtins
2241 @opindex fno-nonansi-builtins
2242 Disable built-in declarations of functions that are not mandated by
2243 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2244 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2247 @opindex fnothrow-opt
2248 Treat a @code{throw()} exception specification as if it were a
2249 @code{noexcept} specification to reduce or eliminate the text size
2250 overhead relative to a function with no exception specification. If
2251 the function has local variables of types with non-trivial
2252 destructors, the exception specification actually makes the
2253 function smaller because the EH cleanups for those variables can be
2254 optimized away. The semantic effect is that an exception thrown out of
2255 a function with such an exception specification results in a call
2256 to @code{terminate} rather than @code{unexpected}.
2258 @item -fno-operator-names
2259 @opindex fno-operator-names
2260 Do not treat the operator name keywords @code{and}, @code{bitand},
2261 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2262 synonyms as keywords.
2264 @item -fno-optional-diags
2265 @opindex fno-optional-diags
2266 Disable diagnostics that the standard says a compiler does not need to
2267 issue. Currently, the only such diagnostic issued by G++ is the one for
2268 a name having multiple meanings within a class.
2271 @opindex fpermissive
2272 Downgrade some diagnostics about nonconformant code from errors to
2273 warnings. Thus, using @option{-fpermissive} allows some
2274 nonconforming code to compile.
2276 @item -fno-pretty-templates
2277 @opindex fno-pretty-templates
2278 When an error message refers to a specialization of a function
2279 template, the compiler normally prints the signature of the
2280 template followed by the template arguments and any typedefs or
2281 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2282 rather than @code{void f(int)}) so that it's clear which template is
2283 involved. When an error message refers to a specialization of a class
2284 template, the compiler omits any template arguments that match
2285 the default template arguments for that template. If either of these
2286 behaviors make it harder to understand the error message rather than
2287 easier, you can use @option{-fno-pretty-templates} to disable them.
2291 Enable automatic template instantiation at link time. This option also
2292 implies @option{-fno-implicit-templates}. @xref{Template
2293 Instantiation}, for more information.
2297 Disable generation of information about every class with virtual
2298 functions for use by the C++ run-time type identification features
2299 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2300 of the language, you can save some space by using this flag. Note that
2301 exception handling uses the same information, but G++ generates it as
2302 needed. The @samp{dynamic_cast} operator can still be used for casts that
2303 do not require run-time type information, i.e.@: casts to @code{void *} or to
2304 unambiguous base classes.
2308 Emit statistics about front-end processing at the end of the compilation.
2309 This information is generally only useful to the G++ development team.
2311 @item -fstrict-enums
2312 @opindex fstrict-enums
2313 Allow the compiler to optimize using the assumption that a value of
2314 enumerated type can only be one of the values of the enumeration (as
2315 defined in the C++ standard; basically, a value that can be
2316 represented in the minimum number of bits needed to represent all the
2317 enumerators). This assumption may not be valid if the program uses a
2318 cast to convert an arbitrary integer value to the enumerated type.
2320 @item -ftemplate-backtrace-limit=@var{n}
2321 @opindex ftemplate-backtrace-limit
2322 Set the maximum number of template instantiation notes for a single
2323 warning or error to @var{n}. The default value is 10.
2325 @item -ftemplate-depth=@var{n}
2326 @opindex ftemplate-depth
2327 Set the maximum instantiation depth for template classes to @var{n}.
2328 A limit on the template instantiation depth is needed to detect
2329 endless recursions during template class instantiation. ANSI/ISO C++
2330 conforming programs must not rely on a maximum depth greater than 17
2331 (changed to 1024 in C++11). The default value is 900, as the compiler
2332 can run out of stack space before hitting 1024 in some situations.
2334 @item -fno-threadsafe-statics
2335 @opindex fno-threadsafe-statics
2336 Do not emit the extra code to use the routines specified in the C++
2337 ABI for thread-safe initialization of local statics. You can use this
2338 option to reduce code size slightly in code that doesn't need to be
2341 @item -fuse-cxa-atexit
2342 @opindex fuse-cxa-atexit
2343 Register destructors for objects with static storage duration with the
2344 @code{__cxa_atexit} function rather than the @code{atexit} function.
2345 This option is required for fully standards-compliant handling of static
2346 destructors, but only works if your C library supports
2347 @code{__cxa_atexit}.
2349 @item -fno-use-cxa-get-exception-ptr
2350 @opindex fno-use-cxa-get-exception-ptr
2351 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2352 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2353 if the runtime routine is not available.
2355 @item -fvisibility-inlines-hidden
2356 @opindex fvisibility-inlines-hidden
2357 This switch declares that the user does not attempt to compare
2358 pointers to inline functions or methods where the addresses of the two functions
2359 are taken in different shared objects.
2361 The effect of this is that GCC may, effectively, mark inline methods with
2362 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2363 appear in the export table of a DSO and do not require a PLT indirection
2364 when used within the DSO@. Enabling this option can have a dramatic effect
2365 on load and link times of a DSO as it massively reduces the size of the
2366 dynamic export table when the library makes heavy use of templates.
2368 The behavior of this switch is not quite the same as marking the
2369 methods as hidden directly, because it does not affect static variables
2370 local to the function or cause the compiler to deduce that
2371 the function is defined in only one shared object.
2373 You may mark a method as having a visibility explicitly to negate the
2374 effect of the switch for that method. For example, if you do want to
2375 compare pointers to a particular inline method, you might mark it as
2376 having default visibility. Marking the enclosing class with explicit
2377 visibility has no effect.
2379 Explicitly instantiated inline methods are unaffected by this option
2380 as their linkage might otherwise cross a shared library boundary.
2381 @xref{Template Instantiation}.
2383 @item -fvisibility-ms-compat
2384 @opindex fvisibility-ms-compat
2385 This flag attempts to use visibility settings to make GCC's C++
2386 linkage model compatible with that of Microsoft Visual Studio.
2388 The flag makes these changes to GCC's linkage model:
2392 It sets the default visibility to @code{hidden}, like
2393 @option{-fvisibility=hidden}.
2396 Types, but not their members, are not hidden by default.
2399 The One Definition Rule is relaxed for types without explicit
2400 visibility specifications that are defined in more than one
2401 shared object: those declarations are permitted if they are
2402 permitted when this option is not used.
2405 In new code it is better to use @option{-fvisibility=hidden} and
2406 export those classes that are intended to be externally visible.
2407 Unfortunately it is possible for code to rely, perhaps accidentally,
2408 on the Visual Studio behavior.
2410 Among the consequences of these changes are that static data members
2411 of the same type with the same name but defined in different shared
2412 objects are different, so changing one does not change the other;
2413 and that pointers to function members defined in different shared
2414 objects may not compare equal. When this flag is given, it is a
2415 violation of the ODR to define types with the same name differently.
2417 @item -fvtable-verify=@var{std|preinit|none}
2418 @opindex fvtable-verify
2419 Turn on (or off, if using @option{-fvtable-verify=none}) the security
2420 feature that verifies at runtime, for every virtual call that is made, that
2421 the vtable pointer through which the call is made is valid for the type of
2422 the object, and has not been corrupted or overwritten. If an invalid vtable
2423 pointer is detected (at runtime), an error is reported and execution of the
2424 program is immediately halted.
2426 This option causes runtime data structures to be built, at program start up,
2427 for verifying the vtable pointers. The options @code{std} and @code{preinit}
2428 control the timing of when these data structures are built. In both cases the
2429 data structures are built before execution reaches 'main'. The
2430 @option{-fvtable-verify=std} causes these data structure to be built after the
2431 shared libraries have been loaded and initialized.
2432 @option{-fvtable-verify=preinit} causes them to be built before the shared
2433 libraries have been loaded and initialized.
2435 If this option appears multiple times in the compiler line, with different
2436 values specified, 'none' will take highest priority over both 'std' and
2437 'preinit'; 'preinit' will take priority over 'std'.
2440 @opindex (fvtv-debug)
2441 Causes debug versions of the runtime functions for the vtable verification
2442 feature to be called. This assumes the @option{-fvtable-verify=std} or
2443 @option{-fvtable-verify=preinit} has been used. This flag will also cause the
2444 compiler to keep track of which vtable pointers it found for each class, and
2445 record that information in the file ``vtv_set_ptr_data.log'', in the dump
2446 file directory on the user's machine.
2448 Note: This feature APPENDS data to the log file. If you want a fresh log
2449 file, be sure to delete any existing one.
2452 @opindex fvtv-counts
2453 This is a debugging flag. When used in conjunction with
2454 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
2455 causes the compiler to keep track of the total number of virtual calls
2456 it encountered and the number of verifications it inserted. It also
2457 counts the number of calls to certain runtime library functions
2458 that it inserts. This information, for each compilation unit, is written
2459 to a file named ``vtv_count_data.log'', in the dump_file directory on
2460 the user's machine. It also counts the size of the vtable pointer sets
2461 for each class, and writes this information to ``vtv_class_set_sizes.log''
2462 in the same directory.
2464 Note: This feature APPENDS data to the log files. To get a fresh log
2465 files, be sure to delete any existing ones.
2469 Do not use weak symbol support, even if it is provided by the linker.
2470 By default, G++ uses weak symbols if they are available. This
2471 option exists only for testing, and should not be used by end-users;
2472 it results in inferior code and has no benefits. This option may
2473 be removed in a future release of G++.
2477 Do not search for header files in the standard directories specific to
2478 C++, but do still search the other standard directories. (This option
2479 is used when building the C++ library.)
2482 In addition, these optimization, warning, and code generation options
2483 have meanings only for C++ programs:
2486 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2489 When an explicit @option{-fabi-version=@var{n}} option is used, causes
2490 G++ to warn when it generates code that is probably not compatible with the
2491 vendor-neutral C++ ABI@. Since G++ now defaults to
2492 @option{-fabi-version=0}, @option{-Wabi} has no effect unless either
2493 an older ABI version is selected (with @option{-fabi-version=@var{n}})
2494 or an older compatibility version is selected (with
2495 @option{-Wabi=@var{n}} or @option{-fabi-compat-version=@var{n}}).
2497 Although an effort has been made to warn about
2498 all such cases, there are probably some cases that are not warned about,
2499 even though G++ is generating incompatible code. There may also be
2500 cases where warnings are emitted even though the code that is generated
2503 You should rewrite your code to avoid these warnings if you are
2504 concerned about the fact that code generated by G++ may not be binary
2505 compatible with code generated by other compilers.
2507 @option{-Wabi} can also be used with an explicit version number to
2508 warn about compatibility with a particular @option{-fabi-version}
2509 level, e.g. @option{-Wabi=2} to warn about changes relative to
2510 @option{-fabi-version=2}. Specifying a version number also sets
2511 @option{-fabi-compat-version=@var{n}}.
2513 The known incompatibilities in @option{-fabi-version=2} (which was the
2514 default from GCC 3.4 to 4.9) include:
2519 A template with a non-type template parameter of reference type was
2520 mangled incorrectly:
2523 template <int &> struct S @{@};
2527 This was fixed in @option{-fabi-version=3}.
2530 SIMD vector types declared using @code{__attribute ((vector_size))} were
2531 mangled in a non-standard way that does not allow for overloading of
2532 functions taking vectors of different sizes.
2534 The mangling was changed in @option{-fabi-version=4}.
2537 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2538 qualifiers, and @code{decltype} of a plain declaration was folded away.
2540 These mangling issues were fixed in @option{-fabi-version=5}.
2543 Scoped enumerators passed as arguments to a variadic function are
2544 promoted like unscoped enumerators, causing @samp{va_arg} to complain.
2545 On most targets this does not actually affect the parameter passing
2546 ABI, as there is no way to pass an argument smaller than @samp{int}.
2548 Also, the ABI changed the mangling of template argument packs,
2549 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2550 a class scope function used as a template argument.
2552 These issues were corrected in @option{-fabi-version=6}.
2555 Lambdas in default argument scope were mangled incorrectly, and the
2556 ABI changed the mangling of nullptr_t.
2558 These issues were corrected in @option{-fabi-version=7}.
2561 When mangling a function type with function-cv-qualifiers, the
2562 un-qualified function type was incorrectly treated as a substitution
2565 This was fixed in @option{-fabi-version=8}.
2568 It also warns about psABI-related changes. The known psABI changes at this
2574 For SysV/x86-64, unions with @code{long double} members are
2575 passed in memory as specified in psABI. For example:
2585 @code{union U} is always passed in memory.
2589 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2590 @opindex Wctor-dtor-privacy
2591 @opindex Wno-ctor-dtor-privacy
2592 Warn when a class seems unusable because all the constructors or
2593 destructors in that class are private, and it has neither friends nor
2594 public static member functions. Also warn if there are no non-private
2595 methods, and there's at least one private member function that isn't
2596 a constructor or destructor.
2598 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2599 @opindex Wdelete-non-virtual-dtor
2600 @opindex Wno-delete-non-virtual-dtor
2601 Warn when @samp{delete} is used to destroy an instance of a class that
2602 has virtual functions and non-virtual destructor. It is unsafe to delete
2603 an instance of a derived class through a pointer to a base class if the
2604 base class does not have a virtual destructor. This warning is enabled
2607 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2608 @opindex Wliteral-suffix
2609 @opindex Wno-literal-suffix
2610 Warn when a string or character literal is followed by a ud-suffix which does
2611 not begin with an underscore. As a conforming extension, GCC treats such
2612 suffixes as separate preprocessing tokens in order to maintain backwards
2613 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2617 #define __STDC_FORMAT_MACROS
2618 #include <inttypes.h>
2623 printf("My int64: %"PRId64"\n", i64);
2627 In this case, @code{PRId64} is treated as a separate preprocessing token.
2629 This warning is enabled by default.
2631 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2633 @opindex Wno-narrowing
2634 Warn when a narrowing conversion prohibited by C++11 occurs within
2638 int i = @{ 2.2 @}; // error: narrowing from double to int
2641 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2643 With @option{-std=c++11}, @option{-Wno-narrowing} suppresses for
2644 non-constants the diagnostic required by the standard. Note that this
2645 does not affect the meaning of well-formed code; narrowing conversions
2646 are still considered ill-formed in SFINAE context.
2648 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2650 @opindex Wno-noexcept
2651 Warn when a noexcept-expression evaluates to false because of a call
2652 to a function that does not have a non-throwing exception
2653 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2654 the compiler to never throw an exception.
2656 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2657 @opindex Wnon-virtual-dtor
2658 @opindex Wno-non-virtual-dtor
2659 Warn when a class has virtual functions and an accessible non-virtual
2660 destructor itself or in an accessible polymorphic base class, in which
2661 case it is possible but unsafe to delete an instance of a derived
2662 class through a pointer to the class itself or base class. This
2663 warning is automatically enabled if @option{-Weffc++} is specified.
2665 @item -Wreorder @r{(C++ and Objective-C++ only)}
2667 @opindex Wno-reorder
2668 @cindex reordering, warning
2669 @cindex warning for reordering of member initializers
2670 Warn when the order of member initializers given in the code does not
2671 match the order in which they must be executed. For instance:
2677 A(): j (0), i (1) @{ @}
2682 The compiler rearranges the member initializers for @samp{i}
2683 and @samp{j} to match the declaration order of the members, emitting
2684 a warning to that effect. This warning is enabled by @option{-Wall}.
2686 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2687 @opindex fext-numeric-literals
2688 @opindex fno-ext-numeric-literals
2689 Accept imaginary, fixed-point, or machine-defined
2690 literal number suffixes as GNU extensions.
2691 When this option is turned off these suffixes are treated
2692 as C++11 user-defined literal numeric suffixes.
2693 This is on by default for all pre-C++11 dialects and all GNU dialects:
2694 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2695 @option{-std=gnu++14}.
2696 This option is off by default
2697 for ISO C++11 onwards (@option{-std=c++11}, ...).
2700 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2703 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2706 Warn about violations of the following style guidelines from Scott Meyers'
2707 @cite{Effective C++} series of books:
2711 Define a copy constructor and an assignment operator for classes
2712 with dynamically-allocated memory.
2715 Prefer initialization to assignment in constructors.
2718 Have @code{operator=} return a reference to @code{*this}.
2721 Don't try to return a reference when you must return an object.
2724 Distinguish between prefix and postfix forms of increment and
2725 decrement operators.
2728 Never overload @code{&&}, @code{||}, or @code{,}.
2732 This option also enables @option{-Wnon-virtual-dtor}, which is also
2733 one of the effective C++ recommendations. However, the check is
2734 extended to warn about the lack of virtual destructor in accessible
2735 non-polymorphic bases classes too.
2737 When selecting this option, be aware that the standard library
2738 headers do not obey all of these guidelines; use @samp{grep -v}
2739 to filter out those warnings.
2741 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2742 @opindex Wstrict-null-sentinel
2743 @opindex Wno-strict-null-sentinel
2744 Warn about the use of an uncasted @code{NULL} as sentinel. When
2745 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2746 to @code{__null}. Although it is a null pointer constant rather than a
2747 null pointer, it is guaranteed to be of the same size as a pointer.
2748 But this use is not portable across different compilers.
2750 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2751 @opindex Wno-non-template-friend
2752 @opindex Wnon-template-friend
2753 Disable warnings when non-templatized friend functions are declared
2754 within a template. Since the advent of explicit template specification
2755 support in G++, if the name of the friend is an unqualified-id (i.e.,
2756 @samp{friend foo(int)}), the C++ language specification demands that the
2757 friend declare or define an ordinary, nontemplate function. (Section
2758 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2759 could be interpreted as a particular specialization of a templatized
2760 function. Because this non-conforming behavior is no longer the default
2761 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2762 check existing code for potential trouble spots and is on by default.
2763 This new compiler behavior can be turned off with
2764 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2765 but disables the helpful warning.
2767 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2768 @opindex Wold-style-cast
2769 @opindex Wno-old-style-cast
2770 Warn if an old-style (C-style) cast to a non-void type is used within
2771 a C++ program. The new-style casts (@samp{dynamic_cast},
2772 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2773 less vulnerable to unintended effects and much easier to search for.
2775 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2776 @opindex Woverloaded-virtual
2777 @opindex Wno-overloaded-virtual
2778 @cindex overloaded virtual function, warning
2779 @cindex warning for overloaded virtual function
2780 Warn when a function declaration hides virtual functions from a
2781 base class. For example, in:
2788 struct B: public A @{
2793 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2804 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2805 @opindex Wno-pmf-conversions
2806 @opindex Wpmf-conversions
2807 Disable the diagnostic for converting a bound pointer to member function
2810 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2811 @opindex Wsign-promo
2812 @opindex Wno-sign-promo
2813 Warn when overload resolution chooses a promotion from unsigned or
2814 enumerated type to a signed type, over a conversion to an unsigned type of
2815 the same size. Previous versions of G++ tried to preserve
2816 unsignedness, but the standard mandates the current behavior.
2819 @node Objective-C and Objective-C++ Dialect Options
2820 @section Options Controlling Objective-C and Objective-C++ Dialects
2822 @cindex compiler options, Objective-C and Objective-C++
2823 @cindex Objective-C and Objective-C++ options, command-line
2824 @cindex options, Objective-C and Objective-C++
2825 (NOTE: This manual does not describe the Objective-C and Objective-C++
2826 languages themselves. @xref{Standards,,Language Standards
2827 Supported by GCC}, for references.)
2829 This section describes the command-line options that are only meaningful
2830 for Objective-C and Objective-C++ programs. You can also use most of
2831 the language-independent GNU compiler options.
2832 For example, you might compile a file @code{some_class.m} like this:
2835 gcc -g -fgnu-runtime -O -c some_class.m
2839 In this example, @option{-fgnu-runtime} is an option meant only for
2840 Objective-C and Objective-C++ programs; you can use the other options with
2841 any language supported by GCC@.
2843 Note that since Objective-C is an extension of the C language, Objective-C
2844 compilations may also use options specific to the C front-end (e.g.,
2845 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2846 C++-specific options (e.g., @option{-Wabi}).
2848 Here is a list of options that are @emph{only} for compiling Objective-C
2849 and Objective-C++ programs:
2852 @item -fconstant-string-class=@var{class-name}
2853 @opindex fconstant-string-class
2854 Use @var{class-name} as the name of the class to instantiate for each
2855 literal string specified with the syntax @code{@@"@dots{}"}. The default
2856 class name is @code{NXConstantString} if the GNU runtime is being used, and
2857 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2858 @option{-fconstant-cfstrings} option, if also present, overrides the
2859 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2860 to be laid out as constant CoreFoundation strings.
2863 @opindex fgnu-runtime
2864 Generate object code compatible with the standard GNU Objective-C
2865 runtime. This is the default for most types of systems.
2867 @item -fnext-runtime
2868 @opindex fnext-runtime
2869 Generate output compatible with the NeXT runtime. This is the default
2870 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2871 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2874 @item -fno-nil-receivers
2875 @opindex fno-nil-receivers
2876 Assume that all Objective-C message dispatches (@code{[receiver
2877 message:arg]}) in this translation unit ensure that the receiver is
2878 not @code{nil}. This allows for more efficient entry points in the
2879 runtime to be used. This option is only available in conjunction with
2880 the NeXT runtime and ABI version 0 or 1.
2882 @item -fobjc-abi-version=@var{n}
2883 @opindex fobjc-abi-version
2884 Use version @var{n} of the Objective-C ABI for the selected runtime.
2885 This option is currently supported only for the NeXT runtime. In that
2886 case, Version 0 is the traditional (32-bit) ABI without support for
2887 properties and other Objective-C 2.0 additions. Version 1 is the
2888 traditional (32-bit) ABI with support for properties and other
2889 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2890 nothing is specified, the default is Version 0 on 32-bit target
2891 machines, and Version 2 on 64-bit target machines.
2893 @item -fobjc-call-cxx-cdtors
2894 @opindex fobjc-call-cxx-cdtors
2895 For each Objective-C class, check if any of its instance variables is a
2896 C++ object with a non-trivial default constructor. If so, synthesize a
2897 special @code{- (id) .cxx_construct} instance method which runs
2898 non-trivial default constructors on any such instance variables, in order,
2899 and then return @code{self}. Similarly, check if any instance variable
2900 is a C++ object with a non-trivial destructor, and if so, synthesize a
2901 special @code{- (void) .cxx_destruct} method which runs
2902 all such default destructors, in reverse order.
2904 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2905 methods thusly generated only operate on instance variables
2906 declared in the current Objective-C class, and not those inherited
2907 from superclasses. It is the responsibility of the Objective-C
2908 runtime to invoke all such methods in an object's inheritance
2909 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
2910 by the runtime immediately after a new object instance is allocated;
2911 the @code{- (void) .cxx_destruct} methods are invoked immediately
2912 before the runtime deallocates an object instance.
2914 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2915 support for invoking the @code{- (id) .cxx_construct} and
2916 @code{- (void) .cxx_destruct} methods.
2918 @item -fobjc-direct-dispatch
2919 @opindex fobjc-direct-dispatch
2920 Allow fast jumps to the message dispatcher. On Darwin this is
2921 accomplished via the comm page.
2923 @item -fobjc-exceptions
2924 @opindex fobjc-exceptions
2925 Enable syntactic support for structured exception handling in
2926 Objective-C, similar to what is offered by C++ and Java. This option
2927 is required to use the Objective-C keywords @code{@@try},
2928 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2929 @code{@@synchronized}. This option is available with both the GNU
2930 runtime and the NeXT runtime (but not available in conjunction with
2931 the NeXT runtime on Mac OS X 10.2 and earlier).
2935 Enable garbage collection (GC) in Objective-C and Objective-C++
2936 programs. This option is only available with the NeXT runtime; the
2937 GNU runtime has a different garbage collection implementation that
2938 does not require special compiler flags.
2940 @item -fobjc-nilcheck
2941 @opindex fobjc-nilcheck
2942 For the NeXT runtime with version 2 of the ABI, check for a nil
2943 receiver in method invocations before doing the actual method call.
2944 This is the default and can be disabled using
2945 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2946 checked for nil in this way no matter what this flag is set to.
2947 Currently this flag does nothing when the GNU runtime, or an older
2948 version of the NeXT runtime ABI, is used.
2950 @item -fobjc-std=objc1
2952 Conform to the language syntax of Objective-C 1.0, the language
2953 recognized by GCC 4.0. This only affects the Objective-C additions to
2954 the C/C++ language; it does not affect conformance to C/C++ standards,
2955 which is controlled by the separate C/C++ dialect option flags. When
2956 this option is used with the Objective-C or Objective-C++ compiler,
2957 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2958 This is useful if you need to make sure that your Objective-C code can
2959 be compiled with older versions of GCC@.
2961 @item -freplace-objc-classes
2962 @opindex freplace-objc-classes
2963 Emit a special marker instructing @command{ld(1)} not to statically link in
2964 the resulting object file, and allow @command{dyld(1)} to load it in at
2965 run time instead. This is used in conjunction with the Fix-and-Continue
2966 debugging mode, where the object file in question may be recompiled and
2967 dynamically reloaded in the course of program execution, without the need
2968 to restart the program itself. Currently, Fix-and-Continue functionality
2969 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2974 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2975 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2976 compile time) with static class references that get initialized at load time,
2977 which improves run-time performance. Specifying the @option{-fzero-link} flag
2978 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2979 to be retained. This is useful in Zero-Link debugging mode, since it allows
2980 for individual class implementations to be modified during program execution.
2981 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2982 regardless of command-line options.
2984 @item -fno-local-ivars
2985 @opindex fno-local-ivars
2986 @opindex flocal-ivars
2987 By default instance variables in Objective-C can be accessed as if
2988 they were local variables from within the methods of the class they're
2989 declared in. This can lead to shadowing between instance variables
2990 and other variables declared either locally inside a class method or
2991 globally with the same name. Specifying the @option{-fno-local-ivars}
2992 flag disables this behavior thus avoiding variable shadowing issues.
2994 @item -fivar-visibility=@var{public|protected|private|package}
2995 @opindex fivar-visibility
2996 Set the default instance variable visibility to the specified option
2997 so that instance variables declared outside the scope of any access
2998 modifier directives default to the specified visibility.
3002 Dump interface declarations for all classes seen in the source file to a
3003 file named @file{@var{sourcename}.decl}.
3005 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3006 @opindex Wassign-intercept
3007 @opindex Wno-assign-intercept
3008 Warn whenever an Objective-C assignment is being intercepted by the
3011 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3012 @opindex Wno-protocol
3014 If a class is declared to implement a protocol, a warning is issued for
3015 every method in the protocol that is not implemented by the class. The
3016 default behavior is to issue a warning for every method not explicitly
3017 implemented in the class, even if a method implementation is inherited
3018 from the superclass. If you use the @option{-Wno-protocol} option, then
3019 methods inherited from the superclass are considered to be implemented,
3020 and no warning is issued for them.
3022 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3024 @opindex Wno-selector
3025 Warn if multiple methods of different types for the same selector are
3026 found during compilation. The check is performed on the list of methods
3027 in the final stage of compilation. Additionally, a check is performed
3028 for each selector appearing in a @code{@@selector(@dots{})}
3029 expression, and a corresponding method for that selector has been found
3030 during compilation. Because these checks scan the method table only at
3031 the end of compilation, these warnings are not produced if the final
3032 stage of compilation is not reached, for example because an error is
3033 found during compilation, or because the @option{-fsyntax-only} option is
3036 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3037 @opindex Wstrict-selector-match
3038 @opindex Wno-strict-selector-match
3039 Warn if multiple methods with differing argument and/or return types are
3040 found for a given selector when attempting to send a message using this
3041 selector to a receiver of type @code{id} or @code{Class}. When this flag
3042 is off (which is the default behavior), the compiler omits such warnings
3043 if any differences found are confined to types that share the same size
3046 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3047 @opindex Wundeclared-selector
3048 @opindex Wno-undeclared-selector
3049 Warn if a @code{@@selector(@dots{})} expression referring to an
3050 undeclared selector is found. A selector is considered undeclared if no
3051 method with that name has been declared before the
3052 @code{@@selector(@dots{})} expression, either explicitly in an
3053 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3054 an @code{@@implementation} section. This option always performs its
3055 checks as soon as a @code{@@selector(@dots{})} expression is found,
3056 while @option{-Wselector} only performs its checks in the final stage of
3057 compilation. This also enforces the coding style convention
3058 that methods and selectors must be declared before being used.
3060 @item -print-objc-runtime-info
3061 @opindex print-objc-runtime-info
3062 Generate C header describing the largest structure that is passed by
3067 @node Language Independent Options
3068 @section Options to Control Diagnostic Messages Formatting
3069 @cindex options to control diagnostics formatting
3070 @cindex diagnostic messages
3071 @cindex message formatting
3073 Traditionally, diagnostic messages have been formatted irrespective of
3074 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3075 options described below
3076 to control the formatting algorithm for diagnostic messages,
3077 e.g.@: how many characters per line, how often source location
3078 information should be reported. Note that some language front ends may not
3079 honor these options.
3082 @item -fmessage-length=@var{n}
3083 @opindex fmessage-length
3084 Try to format error messages so that they fit on lines of about
3085 @var{n} characters. If @var{n} is zero, then no line-wrapping will be
3086 done; each error message will appear on a single line. This is the
3087 default for all front ends.
3089 @item -fdiagnostics-show-location=once
3090 @opindex fdiagnostics-show-location
3091 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3092 reporter to emit source location information @emph{once}; that is, in
3093 case the message is too long to fit on a single physical line and has to
3094 be wrapped, the source location won't be emitted (as prefix) again,
3095 over and over, in subsequent continuation lines. This is the default
3098 @item -fdiagnostics-show-location=every-line
3099 Only meaningful in line-wrapping mode. Instructs the diagnostic
3100 messages reporter to emit the same source location information (as
3101 prefix) for physical lines that result from the process of breaking
3102 a message which is too long to fit on a single line.
3104 @item -fdiagnostics-color[=@var{WHEN}]
3105 @itemx -fno-diagnostics-color
3106 @opindex fdiagnostics-color
3107 @cindex highlight, color, colour
3108 @vindex GCC_COLORS @r{environment variable}
3109 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3110 or @samp{auto}. The default depends on how the compiler has been configured,
3111 it can be any of the above @var{WHEN} options or also @samp{never}
3112 if @env{GCC_COLORS} environment variable isn't present in the environment,
3113 and @samp{auto} otherwise.
3114 @samp{auto} means to use color only when the standard error is a terminal.
3115 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3116 aliases for @option{-fdiagnostics-color=always} and
3117 @option{-fdiagnostics-color=never}, respectively.
3119 The colors are defined by the environment variable @env{GCC_COLORS}.
3120 Its value is a colon-separated list of capabilities and Select Graphic
3121 Rendition (SGR) substrings. SGR commands are interpreted by the
3122 terminal or terminal emulator. (See the section in the documentation
3123 of your text terminal for permitted values and their meanings as
3124 character attributes.) These substring values are integers in decimal
3125 representation and can be concatenated with semicolons.
3126 Common values to concatenate include
3128 @samp{4} for underline,
3130 @samp{7} for inverse,
3131 @samp{39} for default foreground color,
3132 @samp{30} to @samp{37} for foreground colors,
3133 @samp{90} to @samp{97} for 16-color mode foreground colors,
3134 @samp{38;5;0} to @samp{38;5;255}
3135 for 88-color and 256-color modes foreground colors,
3136 @samp{49} for default background color,
3137 @samp{40} to @samp{47} for background colors,
3138 @samp{100} to @samp{107} for 16-color mode background colors,
3139 and @samp{48;5;0} to @samp{48;5;255}
3140 for 88-color and 256-color modes background colors.
3142 The default @env{GCC_COLORS} is
3143 @samp{error=01;31:warning=01;35:note=01;36:caret=01;32:locus=01:quote=01}
3144 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3145 @samp{01;36} is bold cyan, @samp{01;32} is bold green and
3146 @samp{01} is bold. Setting @env{GCC_COLORS} to the empty
3147 string disables colors.
3148 Supported capabilities are as follows.
3152 @vindex error GCC_COLORS @r{capability}
3153 SGR substring for error: markers.
3156 @vindex warning GCC_COLORS @r{capability}
3157 SGR substring for warning: markers.
3160 @vindex note GCC_COLORS @r{capability}
3161 SGR substring for note: markers.
3164 @vindex caret GCC_COLORS @r{capability}
3165 SGR substring for caret line.
3168 @vindex locus GCC_COLORS @r{capability}
3169 SGR substring for location information, @samp{file:line} or
3170 @samp{file:line:column} etc.
3173 @vindex quote GCC_COLORS @r{capability}
3174 SGR substring for information printed within quotes.
3177 @item -fno-diagnostics-show-option
3178 @opindex fno-diagnostics-show-option
3179 @opindex fdiagnostics-show-option
3180 By default, each diagnostic emitted includes text indicating the
3181 command-line option that directly controls the diagnostic (if such an
3182 option is known to the diagnostic machinery). Specifying the
3183 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3185 @item -fno-diagnostics-show-caret
3186 @opindex fno-diagnostics-show-caret
3187 @opindex fdiagnostics-show-caret
3188 By default, each diagnostic emitted includes the original source line
3189 and a caret '^' indicating the column. This option suppresses this
3194 @node Warning Options
3195 @section Options to Request or Suppress Warnings
3196 @cindex options to control warnings
3197 @cindex warning messages
3198 @cindex messages, warning
3199 @cindex suppressing warnings
3201 Warnings are diagnostic messages that report constructions that
3202 are not inherently erroneous but that are risky or suggest there
3203 may have been an error.
3205 The following language-independent options do not enable specific
3206 warnings but control the kinds of diagnostics produced by GCC@.
3209 @cindex syntax checking
3211 @opindex fsyntax-only
3212 Check the code for syntax errors, but don't do anything beyond that.
3214 @item -fmax-errors=@var{n}
3215 @opindex fmax-errors
3216 Limits the maximum number of error messages to @var{n}, at which point
3217 GCC bails out rather than attempting to continue processing the source
3218 code. If @var{n} is 0 (the default), there is no limit on the number
3219 of error messages produced. If @option{-Wfatal-errors} is also
3220 specified, then @option{-Wfatal-errors} takes precedence over this
3225 Inhibit all warning messages.
3230 Make all warnings into errors.
3235 Make the specified warning into an error. The specifier for a warning
3236 is appended; for example @option{-Werror=switch} turns the warnings
3237 controlled by @option{-Wswitch} into errors. This switch takes a
3238 negative form, to be used to negate @option{-Werror} for specific
3239 warnings; for example @option{-Wno-error=switch} makes
3240 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3243 The warning message for each controllable warning includes the
3244 option that controls the warning. That option can then be used with
3245 @option{-Werror=} and @option{-Wno-error=} as described above.
3246 (Printing of the option in the warning message can be disabled using the
3247 @option{-fno-diagnostics-show-option} flag.)
3249 Note that specifying @option{-Werror=}@var{foo} automatically implies
3250 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3253 @item -Wfatal-errors
3254 @opindex Wfatal-errors
3255 @opindex Wno-fatal-errors
3256 This option causes the compiler to abort compilation on the first error
3257 occurred rather than trying to keep going and printing further error
3262 You can request many specific warnings with options beginning with
3263 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3264 implicit declarations. Each of these specific warning options also
3265 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3266 example, @option{-Wno-implicit}. This manual lists only one of the
3267 two forms, whichever is not the default. For further
3268 language-specific options also refer to @ref{C++ Dialect Options} and
3269 @ref{Objective-C and Objective-C++ Dialect Options}.
3271 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3272 options, such as @option{-Wunused}, which may turn on further options,
3273 such as @option{-Wunused-value}. The combined effect of positive and
3274 negative forms is that more specific options have priority over less
3275 specific ones, independently of their position in the command-line. For
3276 options of the same specificity, the last one takes effect. Options
3277 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3278 as if they appeared at the end of the command-line.
3280 When an unrecognized warning option is requested (e.g.,
3281 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3282 that the option is not recognized. However, if the @option{-Wno-} form
3283 is used, the behavior is slightly different: no diagnostic is
3284 produced for @option{-Wno-unknown-warning} unless other diagnostics
3285 are being produced. This allows the use of new @option{-Wno-} options
3286 with old compilers, but if something goes wrong, the compiler
3287 warns that an unrecognized option is present.
3294 Issue all the warnings demanded by strict ISO C and ISO C++;
3295 reject all programs that use forbidden extensions, and some other
3296 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3297 version of the ISO C standard specified by any @option{-std} option used.
3299 Valid ISO C and ISO C++ programs should compile properly with or without
3300 this option (though a rare few require @option{-ansi} or a
3301 @option{-std} option specifying the required version of ISO C)@. However,
3302 without this option, certain GNU extensions and traditional C and C++
3303 features are supported as well. With this option, they are rejected.
3305 @option{-Wpedantic} does not cause warning messages for use of the
3306 alternate keywords whose names begin and end with @samp{__}. Pedantic
3307 warnings are also disabled in the expression that follows
3308 @code{__extension__}. However, only system header files should use
3309 these escape routes; application programs should avoid them.
3310 @xref{Alternate Keywords}.
3312 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3313 C conformance. They soon find that it does not do quite what they want:
3314 it finds some non-ISO practices, but not all---only those for which
3315 ISO C @emph{requires} a diagnostic, and some others for which
3316 diagnostics have been added.
3318 A feature to report any failure to conform to ISO C might be useful in
3319 some instances, but would require considerable additional work and would
3320 be quite different from @option{-Wpedantic}. We don't have plans to
3321 support such a feature in the near future.
3323 Where the standard specified with @option{-std} represents a GNU
3324 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3325 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3326 extended dialect is based. Warnings from @option{-Wpedantic} are given
3327 where they are required by the base standard. (It does not make sense
3328 for such warnings to be given only for features not in the specified GNU
3329 C dialect, since by definition the GNU dialects of C include all
3330 features the compiler supports with the given option, and there would be
3331 nothing to warn about.)
3333 @item -pedantic-errors
3334 @opindex pedantic-errors
3335 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3336 requires a diagnostic, in some cases where there is undefined behavior
3337 at compile-time and in some other cases that do not prevent compilation
3338 of programs that are valid according to the standard. This is not
3339 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3340 by this option and not enabled by the latter and vice versa.
3345 This enables all the warnings about constructions that some users
3346 consider questionable, and that are easy to avoid (or modify to
3347 prevent the warning), even in conjunction with macros. This also
3348 enables some language-specific warnings described in @ref{C++ Dialect
3349 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3351 @option{-Wall} turns on the following warning flags:
3353 @gccoptlist{-Waddress @gol
3354 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
3356 -Wchar-subscripts @gol
3357 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3358 -Wimplicit-int @r{(C and Objective-C only)} @gol
3359 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3362 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3363 -Wmaybe-uninitialized @gol
3364 -Wmissing-braces @r{(only for C/ObjC)} @gol
3371 -Wsequence-point @gol
3372 -Wsign-compare @r{(only in C++)} @gol
3373 -Wstrict-aliasing @gol
3374 -Wstrict-overflow=1 @gol
3377 -Wuninitialized @gol
3378 -Wunknown-pragmas @gol
3379 -Wunused-function @gol
3382 -Wunused-variable @gol
3383 -Wvolatile-register-var @gol
3386 Note that some warning flags are not implied by @option{-Wall}. Some of
3387 them warn about constructions that users generally do not consider
3388 questionable, but which occasionally you might wish to check for;
3389 others warn about constructions that are necessary or hard to avoid in
3390 some cases, and there is no simple way to modify the code to suppress
3391 the warning. Some of them are enabled by @option{-Wextra} but many of
3392 them must be enabled individually.
3398 This enables some extra warning flags that are not enabled by
3399 @option{-Wall}. (This option used to be called @option{-W}. The older
3400 name is still supported, but the newer name is more descriptive.)
3402 @gccoptlist{-Wclobbered @gol
3404 -Wignored-qualifiers @gol
3405 -Wmissing-field-initializers @gol
3406 -Wmissing-parameter-type @r{(C only)} @gol
3407 -Wold-style-declaration @r{(C only)} @gol
3408 -Woverride-init @gol
3411 -Wuninitialized @gol
3412 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3413 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3416 The option @option{-Wextra} also prints warning messages for the
3422 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3423 @samp{>}, or @samp{>=}.
3426 (C++ only) An enumerator and a non-enumerator both appear in a
3427 conditional expression.
3430 (C++ only) Ambiguous virtual bases.
3433 (C++ only) Subscripting an array that has been declared @samp{register}.
3436 (C++ only) Taking the address of a variable that has been declared
3440 (C++ only) A base class is not initialized in a derived class's copy
3445 @item -Wchar-subscripts
3446 @opindex Wchar-subscripts
3447 @opindex Wno-char-subscripts
3448 Warn if an array subscript has type @code{char}. This is a common cause
3449 of error, as programmers often forget that this type is signed on some
3451 This warning is enabled by @option{-Wall}.
3455 @opindex Wno-comment
3456 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3457 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3458 This warning is enabled by @option{-Wall}.
3460 @item -Wno-coverage-mismatch
3461 @opindex Wno-coverage-mismatch
3462 Warn if feedback profiles do not match when using the
3463 @option{-fprofile-use} option.
3464 If a source file is changed between compiling with @option{-fprofile-gen} and
3465 with @option{-fprofile-use}, the files with the profile feedback can fail
3466 to match the source file and GCC cannot use the profile feedback
3467 information. By default, this warning is enabled and is treated as an
3468 error. @option{-Wno-coverage-mismatch} can be used to disable the
3469 warning or @option{-Wno-error=coverage-mismatch} can be used to
3470 disable the error. Disabling the error for this warning can result in
3471 poorly optimized code and is useful only in the
3472 case of very minor changes such as bug fixes to an existing code-base.
3473 Completely disabling the warning is not recommended.
3476 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3478 Suppress warning messages emitted by @code{#warning} directives.
3480 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3481 @opindex Wdouble-promotion
3482 @opindex Wno-double-promotion
3483 Give a warning when a value of type @code{float} is implicitly
3484 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3485 floating-point unit implement @code{float} in hardware, but emulate
3486 @code{double} in software. On such a machine, doing computations
3487 using @code{double} values is much more expensive because of the
3488 overhead required for software emulation.
3490 It is easy to accidentally do computations with @code{double} because
3491 floating-point literals are implicitly of type @code{double}. For
3495 float area(float radius)
3497 return 3.14159 * radius * radius;
3501 the compiler performs the entire computation with @code{double}
3502 because the floating-point literal is a @code{double}.
3505 @itemx -Wformat=@var{n}
3508 @opindex ffreestanding
3509 @opindex fno-builtin
3511 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3512 the arguments supplied have types appropriate to the format string
3513 specified, and that the conversions specified in the format string make
3514 sense. This includes standard functions, and others specified by format
3515 attributes (@pxref{Function Attributes}), in the @code{printf},
3516 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3517 not in the C standard) families (or other target-specific families).
3518 Which functions are checked without format attributes having been
3519 specified depends on the standard version selected, and such checks of
3520 functions without the attribute specified are disabled by
3521 @option{-ffreestanding} or @option{-fno-builtin}.
3523 The formats are checked against the format features supported by GNU
3524 libc version 2.2. These include all ISO C90 and C99 features, as well
3525 as features from the Single Unix Specification and some BSD and GNU
3526 extensions. Other library implementations may not support all these
3527 features; GCC does not support warning about features that go beyond a
3528 particular library's limitations. However, if @option{-Wpedantic} is used
3529 with @option{-Wformat}, warnings are given about format features not
3530 in the selected standard version (but not for @code{strfmon} formats,
3531 since those are not in any version of the C standard). @xref{C Dialect
3532 Options,,Options Controlling C Dialect}.
3539 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3540 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3541 @option{-Wformat} also checks for null format arguments for several
3542 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3543 aspects of this level of format checking can be disabled by the
3544 options: @option{-Wno-format-contains-nul},
3545 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3546 @option{-Wformat} is enabled by @option{-Wall}.
3548 @item -Wno-format-contains-nul
3549 @opindex Wno-format-contains-nul
3550 @opindex Wformat-contains-nul
3551 If @option{-Wformat} is specified, do not warn about format strings that
3554 @item -Wno-format-extra-args
3555 @opindex Wno-format-extra-args
3556 @opindex Wformat-extra-args
3557 If @option{-Wformat} is specified, do not warn about excess arguments to a
3558 @code{printf} or @code{scanf} format function. The C standard specifies
3559 that such arguments are ignored.
3561 Where the unused arguments lie between used arguments that are
3562 specified with @samp{$} operand number specifications, normally
3563 warnings are still given, since the implementation could not know what
3564 type to pass to @code{va_arg} to skip the unused arguments. However,
3565 in the case of @code{scanf} formats, this option suppresses the
3566 warning if the unused arguments are all pointers, since the Single
3567 Unix Specification says that such unused arguments are allowed.
3569 @item -Wno-format-zero-length
3570 @opindex Wno-format-zero-length
3571 @opindex Wformat-zero-length
3572 If @option{-Wformat} is specified, do not warn about zero-length formats.
3573 The C standard specifies that zero-length formats are allowed.
3578 Enable @option{-Wformat} plus additional format checks. Currently
3579 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
3580 -Wformat-signedness -Wformat-y2k}.
3582 @item -Wformat-nonliteral
3583 @opindex Wformat-nonliteral
3584 @opindex Wno-format-nonliteral
3585 If @option{-Wformat} is specified, also warn if the format string is not a
3586 string literal and so cannot be checked, unless the format function
3587 takes its format arguments as a @code{va_list}.
3589 @item -Wformat-security
3590 @opindex Wformat-security
3591 @opindex Wno-format-security
3592 If @option{-Wformat} is specified, also warn about uses of format
3593 functions that represent possible security problems. At present, this
3594 warns about calls to @code{printf} and @code{scanf} functions where the
3595 format string is not a string literal and there are no format arguments,
3596 as in @code{printf (foo);}. This may be a security hole if the format
3597 string came from untrusted input and contains @samp{%n}. (This is
3598 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3599 in future warnings may be added to @option{-Wformat-security} that are not
3600 included in @option{-Wformat-nonliteral}.)
3602 @item -Wformat-signedness
3603 @opindex Wformat-signedness
3604 @opindex Wno-format-signedness
3605 If @option{-Wformat} is specified, also warn if the format string
3606 requires an unsigned argument and the argument is signed and vice versa.
3609 @opindex Wformat-y2k
3610 @opindex Wno-format-y2k
3611 If @option{-Wformat} is specified, also warn about @code{strftime}
3612 formats that may yield only a two-digit year.
3617 @opindex Wno-nonnull
3618 Warn about passing a null pointer for arguments marked as
3619 requiring a non-null value by the @code{nonnull} function attribute.
3621 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3622 can be disabled with the @option{-Wno-nonnull} option.
3624 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3626 @opindex Wno-init-self
3627 Warn about uninitialized variables that are initialized with themselves.
3628 Note this option can only be used with the @option{-Wuninitialized} option.
3630 For example, GCC warns about @code{i} being uninitialized in the
3631 following snippet only when @option{-Winit-self} has been specified:
3642 This warning is enabled by @option{-Wall} in C++.
3644 @item -Wimplicit-int @r{(C and Objective-C only)}
3645 @opindex Wimplicit-int
3646 @opindex Wno-implicit-int
3647 Warn when a declaration does not specify a type.
3648 This warning is enabled by @option{-Wall}.
3650 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3651 @opindex Wimplicit-function-declaration
3652 @opindex Wno-implicit-function-declaration
3653 Give a warning whenever a function is used before being declared. In
3654 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3655 enabled by default and it is made into an error by
3656 @option{-pedantic-errors}. This warning is also enabled by
3659 @item -Wimplicit @r{(C and Objective-C only)}
3661 @opindex Wno-implicit
3662 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3663 This warning is enabled by @option{-Wall}.
3665 @item -Wignored-qualifiers @r{(C and C++ only)}
3666 @opindex Wignored-qualifiers
3667 @opindex Wno-ignored-qualifiers
3668 Warn if the return type of a function has a type qualifier
3669 such as @code{const}. For ISO C such a type qualifier has no effect,
3670 since the value returned by a function is not an lvalue.
3671 For C++, the warning is only emitted for scalar types or @code{void}.
3672 ISO C prohibits qualified @code{void} return types on function
3673 definitions, so such return types always receive a warning
3674 even without this option.
3676 This warning is also enabled by @option{-Wextra}.
3681 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3682 a function with external linkage, returning int, taking either zero
3683 arguments, two, or three arguments of appropriate types. This warning
3684 is enabled by default in C++ and is enabled by either @option{-Wall}
3685 or @option{-Wpedantic}.
3687 @item -Wmissing-braces
3688 @opindex Wmissing-braces
3689 @opindex Wno-missing-braces
3690 Warn if an aggregate or union initializer is not fully bracketed. In
3691 the following example, the initializer for @samp{a} is not fully
3692 bracketed, but that for @samp{b} is fully bracketed. This warning is
3693 enabled by @option{-Wall} in C.
3696 int a[2][2] = @{ 0, 1, 2, 3 @};
3697 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3700 This warning is enabled by @option{-Wall}.
3702 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3703 @opindex Wmissing-include-dirs
3704 @opindex Wno-missing-include-dirs
3705 Warn if a user-supplied include directory does not exist.
3708 @opindex Wparentheses
3709 @opindex Wno-parentheses
3710 Warn if parentheses are omitted in certain contexts, such
3711 as when there is an assignment in a context where a truth value
3712 is expected, or when operators are nested whose precedence people
3713 often get confused about.
3715 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3716 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3717 interpretation from that of ordinary mathematical notation.
3719 Also warn about constructions where there may be confusion to which
3720 @code{if} statement an @code{else} branch belongs. Here is an example of
3735 In C/C++, every @code{else} branch belongs to the innermost possible
3736 @code{if} statement, which in this example is @code{if (b)}. This is
3737 often not what the programmer expected, as illustrated in the above
3738 example by indentation the programmer chose. When there is the
3739 potential for this confusion, GCC issues a warning when this flag
3740 is specified. To eliminate the warning, add explicit braces around
3741 the innermost @code{if} statement so there is no way the @code{else}
3742 can belong to the enclosing @code{if}. The resulting code
3759 Also warn for dangerous uses of the GNU extension to
3760 @code{?:} with omitted middle operand. When the condition
3761 in the @code{?}: operator is a boolean expression, the omitted value is
3762 always 1. Often programmers expect it to be a value computed
3763 inside the conditional expression instead.
3765 This warning is enabled by @option{-Wall}.
3767 @item -Wsequence-point
3768 @opindex Wsequence-point
3769 @opindex Wno-sequence-point
3770 Warn about code that may have undefined semantics because of violations
3771 of sequence point rules in the C and C++ standards.
3773 The C and C++ standards define the order in which expressions in a C/C++
3774 program are evaluated in terms of @dfn{sequence points}, which represent
3775 a partial ordering between the execution of parts of the program: those
3776 executed before the sequence point, and those executed after it. These
3777 occur after the evaluation of a full expression (one which is not part
3778 of a larger expression), after the evaluation of the first operand of a
3779 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3780 function is called (but after the evaluation of its arguments and the
3781 expression denoting the called function), and in certain other places.
3782 Other than as expressed by the sequence point rules, the order of
3783 evaluation of subexpressions of an expression is not specified. All
3784 these rules describe only a partial order rather than a total order,
3785 since, for example, if two functions are called within one expression
3786 with no sequence point between them, the order in which the functions
3787 are called is not specified. However, the standards committee have
3788 ruled that function calls do not overlap.
3790 It is not specified when between sequence points modifications to the
3791 values of objects take effect. Programs whose behavior depends on this
3792 have undefined behavior; the C and C++ standards specify that ``Between
3793 the previous and next sequence point an object shall have its stored
3794 value modified at most once by the evaluation of an expression.
3795 Furthermore, the prior value shall be read only to determine the value
3796 to be stored.''. If a program breaks these rules, the results on any
3797 particular implementation are entirely unpredictable.
3799 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3800 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3801 diagnosed by this option, and it may give an occasional false positive
3802 result, but in general it has been found fairly effective at detecting
3803 this sort of problem in programs.
3805 The standard is worded confusingly, therefore there is some debate
3806 over the precise meaning of the sequence point rules in subtle cases.
3807 Links to discussions of the problem, including proposed formal
3808 definitions, may be found on the GCC readings page, at
3809 @uref{http://gcc.gnu.org/@/readings.html}.
3811 This warning is enabled by @option{-Wall} for C and C++.
3813 @item -Wno-return-local-addr
3814 @opindex Wno-return-local-addr
3815 @opindex Wreturn-local-addr
3816 Do not warn about returning a pointer (or in C++, a reference) to a
3817 variable that goes out of scope after the function returns.
3820 @opindex Wreturn-type
3821 @opindex Wno-return-type
3822 Warn whenever a function is defined with a return type that defaults
3823 to @code{int}. Also warn about any @code{return} statement with no
3824 return value in a function whose return type is not @code{void}
3825 (falling off the end of the function body is considered returning
3826 without a value), and about a @code{return} statement with an
3827 expression in a function whose return type is @code{void}.
3829 For C++, a function without return type always produces a diagnostic
3830 message, even when @option{-Wno-return-type} is specified. The only
3831 exceptions are @samp{main} and functions defined in system headers.
3833 This warning is enabled by @option{-Wall}.
3835 @item -Wshift-count-negative
3836 @opindex Wshift-count-negative
3837 @opindex Wno-shift-count-negative
3838 Warn if shift count is negative. This warning is enabled by default.
3840 @item -Wshift-count-overflow
3841 @opindex Wshift-count-overflow
3842 @opindex Wno-shift-count-overflow
3843 Warn if shift count >= width of type. This warning is enabled by default.
3848 Warn whenever a @code{switch} statement has an index of enumerated type
3849 and lacks a @code{case} for one or more of the named codes of that
3850 enumeration. (The presence of a @code{default} label prevents this
3851 warning.) @code{case} labels outside the enumeration range also
3852 provoke warnings when this option is used (even if there is a
3853 @code{default} label).
3854 This warning is enabled by @option{-Wall}.
3856 @item -Wswitch-default
3857 @opindex Wswitch-default
3858 @opindex Wno-switch-default
3859 Warn whenever a @code{switch} statement does not have a @code{default}
3863 @opindex Wswitch-enum
3864 @opindex Wno-switch-enum
3865 Warn whenever a @code{switch} statement has an index of enumerated type
3866 and lacks a @code{case} for one or more of the named codes of that
3867 enumeration. @code{case} labels outside the enumeration range also
3868 provoke warnings when this option is used. The only difference
3869 between @option{-Wswitch} and this option is that this option gives a
3870 warning about an omitted enumeration code even if there is a
3871 @code{default} label.
3874 @opindex Wswitch-bool
3875 @opindex Wno-switch-bool
3876 Warn whenever a @code{switch} statement has an index of boolean type.
3877 It is possible to suppress this warning by casting the controlling
3878 expression to a type other than @code{bool}. For example:
3881 switch ((int) (a == 4))
3887 This warning is enabled by default for C and C++ programs.
3889 @item -Wsync-nand @r{(C and C++ only)}
3891 @opindex Wno-sync-nand
3892 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3893 built-in functions are used. These functions changed semantics in GCC 4.4.
3897 @opindex Wno-trigraphs
3898 Warn if any trigraphs are encountered that might change the meaning of
3899 the program (trigraphs within comments are not warned about).
3900 This warning is enabled by @option{-Wall}.
3902 @item -Wunused-but-set-parameter
3903 @opindex Wunused-but-set-parameter
3904 @opindex Wno-unused-but-set-parameter
3905 Warn whenever a function parameter is assigned to, but otherwise unused
3906 (aside from its declaration).
3908 To suppress this warning use the @samp{unused} attribute
3909 (@pxref{Variable Attributes}).
3911 This warning is also enabled by @option{-Wunused} together with
3914 @item -Wunused-but-set-variable
3915 @opindex Wunused-but-set-variable
3916 @opindex Wno-unused-but-set-variable
3917 Warn whenever a local variable is assigned to, but otherwise unused
3918 (aside from its declaration).
3919 This warning is enabled by @option{-Wall}.
3921 To suppress this warning use the @samp{unused} attribute
3922 (@pxref{Variable Attributes}).
3924 This warning is also enabled by @option{-Wunused}, which is enabled
3927 @item -Wunused-function
3928 @opindex Wunused-function
3929 @opindex Wno-unused-function
3930 Warn whenever a static function is declared but not defined or a
3931 non-inline static function is unused.
3932 This warning is enabled by @option{-Wall}.
3934 @item -Wunused-label
3935 @opindex Wunused-label
3936 @opindex Wno-unused-label
3937 Warn whenever a label is declared but not used.
3938 This warning is enabled by @option{-Wall}.
3940 To suppress this warning use the @samp{unused} attribute
3941 (@pxref{Variable Attributes}).
3943 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
3944 @opindex Wunused-local-typedefs
3945 Warn when a typedef locally defined in a function is not used.
3946 This warning is enabled by @option{-Wall}.
3948 @item -Wunused-parameter
3949 @opindex Wunused-parameter
3950 @opindex Wno-unused-parameter
3951 Warn whenever a function parameter is unused aside from its declaration.
3953 To suppress this warning use the @samp{unused} attribute
3954 (@pxref{Variable Attributes}).
3956 @item -Wno-unused-result
3957 @opindex Wunused-result
3958 @opindex Wno-unused-result
3959 Do not warn if a caller of a function marked with attribute
3960 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
3961 its return value. The default is @option{-Wunused-result}.
3963 @item -Wunused-variable
3964 @opindex Wunused-variable
3965 @opindex Wno-unused-variable
3966 Warn whenever a local variable or non-constant static variable is unused
3967 aside from its declaration.
3968 This warning is enabled by @option{-Wall}.
3970 To suppress this warning use the @samp{unused} attribute
3971 (@pxref{Variable Attributes}).
3973 @item -Wunused-value
3974 @opindex Wunused-value
3975 @opindex Wno-unused-value
3976 Warn whenever a statement computes a result that is explicitly not
3977 used. To suppress this warning cast the unused expression to
3978 @samp{void}. This includes an expression-statement or the left-hand
3979 side of a comma expression that contains no side effects. For example,
3980 an expression such as @samp{x[i,j]} causes a warning, while
3981 @samp{x[(void)i,j]} does not.
3983 This warning is enabled by @option{-Wall}.
3988 All the above @option{-Wunused} options combined.
3990 In order to get a warning about an unused function parameter, you must
3991 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
3992 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
3994 @item -Wuninitialized
3995 @opindex Wuninitialized
3996 @opindex Wno-uninitialized
3997 Warn if an automatic variable is used without first being initialized
3998 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3999 warn if a non-static reference or non-static @samp{const} member
4000 appears in a class without constructors.
4002 If you want to warn about code that uses the uninitialized value of the
4003 variable in its own initializer, use the @option{-Winit-self} option.
4005 These warnings occur for individual uninitialized or clobbered
4006 elements of structure, union or array variables as well as for
4007 variables that are uninitialized or clobbered as a whole. They do
4008 not occur for variables or elements declared @code{volatile}. Because
4009 these warnings depend on optimization, the exact variables or elements
4010 for which there are warnings depends on the precise optimization
4011 options and version of GCC used.
4013 Note that there may be no warning about a variable that is used only
4014 to compute a value that itself is never used, because such
4015 computations may be deleted by data flow analysis before the warnings
4018 @item -Wmaybe-uninitialized
4019 @opindex Wmaybe-uninitialized
4020 @opindex Wno-maybe-uninitialized
4021 For an automatic variable, if there exists a path from the function
4022 entry to a use of the variable that is initialized, but there exist
4023 some other paths for which the variable is not initialized, the compiler
4024 emits a warning if it cannot prove the uninitialized paths are not
4025 executed at run time. These warnings are made optional because GCC is
4026 not smart enough to see all the reasons why the code might be correct
4027 in spite of appearing to have an error. Here is one example of how
4048 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4049 always initialized, but GCC doesn't know this. To suppress the
4050 warning, you need to provide a default case with assert(0) or
4053 @cindex @code{longjmp} warnings
4054 This option also warns when a non-volatile automatic variable might be
4055 changed by a call to @code{longjmp}. These warnings as well are possible
4056 only in optimizing compilation.
4058 The compiler sees only the calls to @code{setjmp}. It cannot know
4059 where @code{longjmp} will be called; in fact, a signal handler could
4060 call it at any point in the code. As a result, you may get a warning
4061 even when there is in fact no problem because @code{longjmp} cannot
4062 in fact be called at the place that would cause a problem.
4064 Some spurious warnings can be avoided if you declare all the functions
4065 you use that never return as @code{noreturn}. @xref{Function
4068 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4070 @item -Wunknown-pragmas
4071 @opindex Wunknown-pragmas
4072 @opindex Wno-unknown-pragmas
4073 @cindex warning for unknown pragmas
4074 @cindex unknown pragmas, warning
4075 @cindex pragmas, warning of unknown
4076 Warn when a @code{#pragma} directive is encountered that is not understood by
4077 GCC@. If this command-line option is used, warnings are even issued
4078 for unknown pragmas in system header files. This is not the case if
4079 the warnings are only enabled by the @option{-Wall} command-line option.
4082 @opindex Wno-pragmas
4084 Do not warn about misuses of pragmas, such as incorrect parameters,
4085 invalid syntax, or conflicts between pragmas. See also
4086 @option{-Wunknown-pragmas}.
4088 @item -Wstrict-aliasing
4089 @opindex Wstrict-aliasing
4090 @opindex Wno-strict-aliasing
4091 This option is only active when @option{-fstrict-aliasing} is active.
4092 It warns about code that might break the strict aliasing rules that the
4093 compiler is using for optimization. The warning does not catch all
4094 cases, but does attempt to catch the more common pitfalls. It is
4095 included in @option{-Wall}.
4096 It is equivalent to @option{-Wstrict-aliasing=3}
4098 @item -Wstrict-aliasing=n
4099 @opindex Wstrict-aliasing=n
4100 This option is only active when @option{-fstrict-aliasing} is active.
4101 It warns about code that might break the strict aliasing rules that the
4102 compiler is using for optimization.
4103 Higher levels correspond to higher accuracy (fewer false positives).
4104 Higher levels also correspond to more effort, similar to the way @option{-O}
4106 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4108 Level 1: Most aggressive, quick, least accurate.
4109 Possibly useful when higher levels
4110 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4111 false negatives. However, it has many false positives.
4112 Warns for all pointer conversions between possibly incompatible types,
4113 even if never dereferenced. Runs in the front end only.
4115 Level 2: Aggressive, quick, not too precise.
4116 May still have many false positives (not as many as level 1 though),
4117 and few false negatives (but possibly more than level 1).
4118 Unlike level 1, it only warns when an address is taken. Warns about
4119 incomplete types. Runs in the front end only.
4121 Level 3 (default for @option{-Wstrict-aliasing}):
4122 Should have very few false positives and few false
4123 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4124 Takes care of the common pun+dereference pattern in the front end:
4125 @code{*(int*)&some_float}.
4126 If optimization is enabled, it also runs in the back end, where it deals
4127 with multiple statement cases using flow-sensitive points-to information.
4128 Only warns when the converted pointer is dereferenced.
4129 Does not warn about incomplete types.
4131 @item -Wstrict-overflow
4132 @itemx -Wstrict-overflow=@var{n}
4133 @opindex Wstrict-overflow
4134 @opindex Wno-strict-overflow
4135 This option is only active when @option{-fstrict-overflow} is active.
4136 It warns about cases where the compiler optimizes based on the
4137 assumption that signed overflow does not occur. Note that it does not
4138 warn about all cases where the code might overflow: it only warns
4139 about cases where the compiler implements some optimization. Thus
4140 this warning depends on the optimization level.
4142 An optimization that assumes that signed overflow does not occur is
4143 perfectly safe if the values of the variables involved are such that
4144 overflow never does, in fact, occur. Therefore this warning can
4145 easily give a false positive: a warning about code that is not
4146 actually a problem. To help focus on important issues, several
4147 warning levels are defined. No warnings are issued for the use of
4148 undefined signed overflow when estimating how many iterations a loop
4149 requires, in particular when determining whether a loop will be
4153 @item -Wstrict-overflow=1
4154 Warn about cases that are both questionable and easy to avoid. For
4155 example, with @option{-fstrict-overflow}, the compiler simplifies
4156 @code{x + 1 > x} to @code{1}. This level of
4157 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4158 are not, and must be explicitly requested.
4160 @item -Wstrict-overflow=2
4161 Also warn about other cases where a comparison is simplified to a
4162 constant. For example: @code{abs (x) >= 0}. This can only be
4163 simplified when @option{-fstrict-overflow} is in effect, because
4164 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4165 zero. @option{-Wstrict-overflow} (with no level) is the same as
4166 @option{-Wstrict-overflow=2}.
4168 @item -Wstrict-overflow=3
4169 Also warn about other cases where a comparison is simplified. For
4170 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4172 @item -Wstrict-overflow=4
4173 Also warn about other simplifications not covered by the above cases.
4174 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4176 @item -Wstrict-overflow=5
4177 Also warn about cases where the compiler reduces the magnitude of a
4178 constant involved in a comparison. For example: @code{x + 2 > y} is
4179 simplified to @code{x + 1 >= y}. This is reported only at the
4180 highest warning level because this simplification applies to many
4181 comparisons, so this warning level gives a very large number of
4185 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
4186 @opindex Wsuggest-attribute=
4187 @opindex Wno-suggest-attribute=
4188 Warn for cases where adding an attribute may be beneficial. The
4189 attributes currently supported are listed below.
4192 @item -Wsuggest-attribute=pure
4193 @itemx -Wsuggest-attribute=const
4194 @itemx -Wsuggest-attribute=noreturn
4195 @opindex Wsuggest-attribute=pure
4196 @opindex Wno-suggest-attribute=pure
4197 @opindex Wsuggest-attribute=const
4198 @opindex Wno-suggest-attribute=const
4199 @opindex Wsuggest-attribute=noreturn
4200 @opindex Wno-suggest-attribute=noreturn
4202 Warn about functions that might be candidates for attributes
4203 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
4204 functions visible in other compilation units or (in the case of @code{pure} and
4205 @code{const}) if it cannot prove that the function returns normally. A function
4206 returns normally if it doesn't contain an infinite loop or return abnormally
4207 by throwing, calling @code{abort()} or trapping. This analysis requires option
4208 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
4209 higher. Higher optimization levels improve the accuracy of the analysis.
4211 @item -Wsuggest-attribute=format
4212 @itemx -Wmissing-format-attribute
4213 @opindex Wsuggest-attribute=format
4214 @opindex Wmissing-format-attribute
4215 @opindex Wno-suggest-attribute=format
4216 @opindex Wno-missing-format-attribute
4220 Warn about function pointers that might be candidates for @code{format}
4221 attributes. Note these are only possible candidates, not absolute ones.
4222 GCC guesses that function pointers with @code{format} attributes that
4223 are used in assignment, initialization, parameter passing or return
4224 statements should have a corresponding @code{format} attribute in the
4225 resulting type. I.e.@: the left-hand side of the assignment or
4226 initialization, the type of the parameter variable, or the return type
4227 of the containing function respectively should also have a @code{format}
4228 attribute to avoid the warning.
4230 GCC also warns about function definitions that might be
4231 candidates for @code{format} attributes. Again, these are only
4232 possible candidates. GCC guesses that @code{format} attributes
4233 might be appropriate for any function that calls a function like
4234 @code{vprintf} or @code{vscanf}, but this might not always be the
4235 case, and some functions for which @code{format} attributes are
4236 appropriate may not be detected.
4239 @item -Wsuggest-final-types
4240 @opindex Wno-suggest-final-types
4241 @opindex Wsuggest-final-types
4242 Warn about types with virtual methods where code quality would be improved
4243 if the type was declared with C++11 final specifier, or, if possible,
4244 declared in anonymous namespace. This allows GCC to devritualize more aggressively
4245 the polymorphic calls. This warning is more effective with link time optimization,
4246 where the information about the class hiearchy graph is more complete.
4248 @item -Wsuggest-final-methods
4249 @opindex Wno-suggest-final-methods
4250 @opindex Wsuggest-final-methods
4251 Warn about virtual methods where code quality would be improved if the method
4252 was declared with C++11 final specifier, or, if possible, its type was declared
4253 in the anonymous namespace or with final specifier. This warning is more
4254 effective with link time optimization, where the information about the class
4255 hiearchy graph is more complete. It is recommended to first consider suggestins
4256 of @option{-Wsuggest-final-types} and then rebuild with new annotations.
4258 @item -Warray-bounds
4259 @opindex Wno-array-bounds
4260 @opindex Warray-bounds
4261 This option is only active when @option{-ftree-vrp} is active
4262 (default for @option{-O2} and above). It warns about subscripts to arrays
4263 that are always out of bounds. This warning is enabled by @option{-Wall}.
4265 @item -Wbool-compare
4266 @opindex Wno-bool-compare
4267 @opindex Wbool-compare
4268 Warn about boolean expression compared with an integer value different from
4269 @code{true}/@code{false}. For instance, the following comparison is
4274 if ((n > 1) == 2) @{ @dots{} @}
4276 This warning is enabled by @option{-Wall}.
4278 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
4279 @opindex Wno-discarded-qualifiers
4280 @opindex Wdiscarded-qualifiers
4281 Do not warn if type qualifiers on pointers are being discarded.
4282 Typically, the compiler will warn if a @code{const char *} variable is
4283 passed to a function that takes @code{char *} parameter. This option
4284 can be used to suppress such a warning.
4286 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
4287 @opindex Wno-incompatible-pointer-types
4288 @opindex Wincompatible-pointer-types
4289 Do not warn when there is a conversion between pointers that have incompatible
4290 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
4291 which warns for pointer argument passing or assignment with different signedness
4293 @item -Wno-int-conversion @r{(C and Objective-C only)}
4294 @opindex Wno-int-conversion
4295 @opindex Wint-conversion
4296 Do not warn about incompatible integer to pointer and pointer to integer
4297 conversions. This warning is about implicit conversions; for explicit
4298 conversions the warnings @option{-Wno-int-to-pointer-cast} and
4299 @option{-Wno-pointer-to-int-cast} may be used.
4301 @item -Wno-div-by-zero
4302 @opindex Wno-div-by-zero
4303 @opindex Wdiv-by-zero
4304 Do not warn about compile-time integer division by zero. Floating-point
4305 division by zero is not warned about, as it can be a legitimate way of
4306 obtaining infinities and NaNs.
4308 @item -Wsystem-headers
4309 @opindex Wsystem-headers
4310 @opindex Wno-system-headers
4311 @cindex warnings from system headers
4312 @cindex system headers, warnings from
4313 Print warning messages for constructs found in system header files.
4314 Warnings from system headers are normally suppressed, on the assumption
4315 that they usually do not indicate real problems and would only make the
4316 compiler output harder to read. Using this command-line option tells
4317 GCC to emit warnings from system headers as if they occurred in user
4318 code. However, note that using @option{-Wall} in conjunction with this
4319 option does @emph{not} warn about unknown pragmas in system
4320 headers---for that, @option{-Wunknown-pragmas} must also be used.
4323 @opindex Wtrampolines
4324 @opindex Wno-trampolines
4325 Warn about trampolines generated for pointers to nested functions.
4326 A trampoline is a small piece of data or code that is created at run
4327 time on the stack when the address of a nested function is taken, and is
4328 used to call the nested function indirectly. For some targets, it is
4329 made up of data only and thus requires no special treatment. But, for
4330 most targets, it is made up of code and thus requires the stack to be
4331 made executable in order for the program to work properly.
4334 @opindex Wfloat-equal
4335 @opindex Wno-float-equal
4336 Warn if floating-point values are used in equality comparisons.
4338 The idea behind this is that sometimes it is convenient (for the
4339 programmer) to consider floating-point values as approximations to
4340 infinitely precise real numbers. If you are doing this, then you need
4341 to compute (by analyzing the code, or in some other way) the maximum or
4342 likely maximum error that the computation introduces, and allow for it
4343 when performing comparisons (and when producing output, but that's a
4344 different problem). In particular, instead of testing for equality, you
4345 should check to see whether the two values have ranges that overlap; and
4346 this is done with the relational operators, so equality comparisons are
4349 @item -Wtraditional @r{(C and Objective-C only)}
4350 @opindex Wtraditional
4351 @opindex Wno-traditional
4352 Warn about certain constructs that behave differently in traditional and
4353 ISO C@. Also warn about ISO C constructs that have no traditional C
4354 equivalent, and/or problematic constructs that should be avoided.
4358 Macro parameters that appear within string literals in the macro body.
4359 In traditional C macro replacement takes place within string literals,
4360 but in ISO C it does not.
4363 In traditional C, some preprocessor directives did not exist.
4364 Traditional preprocessors only considered a line to be a directive
4365 if the @samp{#} appeared in column 1 on the line. Therefore
4366 @option{-Wtraditional} warns about directives that traditional C
4367 understands but ignores because the @samp{#} does not appear as the
4368 first character on the line. It also suggests you hide directives like
4369 @samp{#pragma} not understood by traditional C by indenting them. Some
4370 traditional implementations do not recognize @samp{#elif}, so this option
4371 suggests avoiding it altogether.
4374 A function-like macro that appears without arguments.
4377 The unary plus operator.
4380 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
4381 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
4382 constants.) Note, these suffixes appear in macros defined in the system
4383 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
4384 Use of these macros in user code might normally lead to spurious
4385 warnings, however GCC's integrated preprocessor has enough context to
4386 avoid warning in these cases.
4389 A function declared external in one block and then used after the end of
4393 A @code{switch} statement has an operand of type @code{long}.
4396 A non-@code{static} function declaration follows a @code{static} one.
4397 This construct is not accepted by some traditional C compilers.
4400 The ISO type of an integer constant has a different width or
4401 signedness from its traditional type. This warning is only issued if
4402 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
4403 typically represent bit patterns, are not warned about.
4406 Usage of ISO string concatenation is detected.
4409 Initialization of automatic aggregates.
4412 Identifier conflicts with labels. Traditional C lacks a separate
4413 namespace for labels.
4416 Initialization of unions. If the initializer is zero, the warning is
4417 omitted. This is done under the assumption that the zero initializer in
4418 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
4419 initializer warnings and relies on default initialization to zero in the
4423 Conversions by prototypes between fixed/floating-point values and vice
4424 versa. The absence of these prototypes when compiling with traditional
4425 C causes serious problems. This is a subset of the possible
4426 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
4429 Use of ISO C style function definitions. This warning intentionally is
4430 @emph{not} issued for prototype declarations or variadic functions
4431 because these ISO C features appear in your code when using
4432 libiberty's traditional C compatibility macros, @code{PARAMS} and
4433 @code{VPARAMS}. This warning is also bypassed for nested functions
4434 because that feature is already a GCC extension and thus not relevant to
4435 traditional C compatibility.
4438 @item -Wtraditional-conversion @r{(C and Objective-C only)}
4439 @opindex Wtraditional-conversion
4440 @opindex Wno-traditional-conversion
4441 Warn if a prototype causes a type conversion that is different from what
4442 would happen to the same argument in the absence of a prototype. This
4443 includes conversions of fixed point to floating and vice versa, and
4444 conversions changing the width or signedness of a fixed-point argument
4445 except when the same as the default promotion.
4447 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
4448 @opindex Wdeclaration-after-statement
4449 @opindex Wno-declaration-after-statement
4450 Warn when a declaration is found after a statement in a block. This
4451 construct, known from C++, was introduced with ISO C99 and is by default
4452 allowed in GCC@. It is not supported by ISO C90 and was not supported by
4453 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
4458 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
4460 @item -Wno-endif-labels
4461 @opindex Wno-endif-labels
4462 @opindex Wendif-labels
4463 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
4468 Warn whenever a local variable or type declaration shadows another
4469 variable, parameter, type, class member (in C++), or instance variable
4470 (in Objective-C) or whenever a built-in function is shadowed. Note
4471 that in C++, the compiler warns if a local variable shadows an
4472 explicit typedef, but not if it shadows a struct/class/enum.
4474 @item -Wno-shadow-ivar @r{(Objective-C only)}
4475 @opindex Wno-shadow-ivar
4476 @opindex Wshadow-ivar
4477 Do not warn whenever a local variable shadows an instance variable in an
4480 @item -Wlarger-than=@var{len}
4481 @opindex Wlarger-than=@var{len}
4482 @opindex Wlarger-than-@var{len}
4483 Warn whenever an object of larger than @var{len} bytes is defined.
4485 @item -Wframe-larger-than=@var{len}
4486 @opindex Wframe-larger-than
4487 Warn if the size of a function frame is larger than @var{len} bytes.
4488 The computation done to determine the stack frame size is approximate
4489 and not conservative.
4490 The actual requirements may be somewhat greater than @var{len}
4491 even if you do not get a warning. In addition, any space allocated
4492 via @code{alloca}, variable-length arrays, or related constructs
4493 is not included by the compiler when determining
4494 whether or not to issue a warning.
4496 @item -Wno-free-nonheap-object
4497 @opindex Wno-free-nonheap-object
4498 @opindex Wfree-nonheap-object
4499 Do not warn when attempting to free an object that was not allocated
4502 @item -Wstack-usage=@var{len}
4503 @opindex Wstack-usage
4504 Warn if the stack usage of a function might be larger than @var{len} bytes.
4505 The computation done to determine the stack usage is conservative.
4506 Any space allocated via @code{alloca}, variable-length arrays, or related
4507 constructs is included by the compiler when determining whether or not to
4510 The message is in keeping with the output of @option{-fstack-usage}.
4514 If the stack usage is fully static but exceeds the specified amount, it's:
4517 warning: stack usage is 1120 bytes
4520 If the stack usage is (partly) dynamic but bounded, it's:
4523 warning: stack usage might be 1648 bytes
4526 If the stack usage is (partly) dynamic and not bounded, it's:
4529 warning: stack usage might be unbounded
4533 @item -Wunsafe-loop-optimizations
4534 @opindex Wunsafe-loop-optimizations
4535 @opindex Wno-unsafe-loop-optimizations
4536 Warn if the loop cannot be optimized because the compiler cannot
4537 assume anything on the bounds of the loop indices. With
4538 @option{-funsafe-loop-optimizations} warn if the compiler makes
4541 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4542 @opindex Wno-pedantic-ms-format
4543 @opindex Wpedantic-ms-format
4544 When used in combination with @option{-Wformat}
4545 and @option{-pedantic} without GNU extensions, this option
4546 disables the warnings about non-ISO @code{printf} / @code{scanf} format
4547 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
4548 which depend on the MS runtime.
4550 @item -Wpointer-arith
4551 @opindex Wpointer-arith
4552 @opindex Wno-pointer-arith
4553 Warn about anything that depends on the ``size of'' a function type or
4554 of @code{void}. GNU C assigns these types a size of 1, for
4555 convenience in calculations with @code{void *} pointers and pointers
4556 to functions. In C++, warn also when an arithmetic operation involves
4557 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
4560 @opindex Wtype-limits
4561 @opindex Wno-type-limits
4562 Warn if a comparison is always true or always false due to the limited
4563 range of the data type, but do not warn for constant expressions. For
4564 example, warn if an unsigned variable is compared against zero with
4565 @samp{<} or @samp{>=}. This warning is also enabled by
4568 @item -Wbad-function-cast @r{(C and Objective-C only)}
4569 @opindex Wbad-function-cast
4570 @opindex Wno-bad-function-cast
4571 Warn whenever a function call is cast to a non-matching type.
4572 For example, warn if @code{int malloc()} is cast to @code{anything *}.
4574 @item -Wc90-c99-compat @r{(C and Objective-C only)}
4575 @opindex Wc90-c99-compat
4576 @opindex Wno-c90-c99-compat
4577 Warn about features not present in ISO C90, but present in ISO C99.
4578 For instance, warn about use of variable length arrays, @code{long long}
4579 type, @code{bool} type, compound literals, designated initializers, and so
4580 on. This option is independent of the standards mode. Warnings are disabled
4581 in the expression that follows @code{__extension__}.
4583 @item -Wc99-c11-compat @r{(C and Objective-C only)}
4584 @opindex Wc99-c11-compat
4585 @opindex Wno-c99-c11-compat
4586 Warn about features not present in ISO C99, but present in ISO C11.
4587 For instance, warn about use of anonymous structures and unions,
4588 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
4589 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
4590 and so on. This option is independent of the standards mode. Warnings are
4591 disabled in the expression that follows @code{__extension__}.
4593 @item -Wc++-compat @r{(C and Objective-C only)}
4594 Warn about ISO C constructs that are outside of the common subset of
4595 ISO C and ISO C++, e.g.@: request for implicit conversion from
4596 @code{void *} to a pointer to non-@code{void} type.
4598 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4599 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4600 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4601 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4602 enabled by @option{-Wall}.
4606 @opindex Wno-cast-qual
4607 Warn whenever a pointer is cast so as to remove a type qualifier from
4608 the target type. For example, warn if a @code{const char *} is cast
4609 to an ordinary @code{char *}.
4611 Also warn when making a cast that introduces a type qualifier in an
4612 unsafe way. For example, casting @code{char **} to @code{const char **}
4613 is unsafe, as in this example:
4616 /* p is char ** value. */
4617 const char **q = (const char **) p;
4618 /* Assignment of readonly string to const char * is OK. */
4620 /* Now char** pointer points to read-only memory. */
4625 @opindex Wcast-align
4626 @opindex Wno-cast-align
4627 Warn whenever a pointer is cast such that the required alignment of the
4628 target is increased. For example, warn if a @code{char *} is cast to
4629 an @code{int *} on machines where integers can only be accessed at
4630 two- or four-byte boundaries.
4632 @item -Wwrite-strings
4633 @opindex Wwrite-strings
4634 @opindex Wno-write-strings
4635 When compiling C, give string constants the type @code{const
4636 char[@var{length}]} so that copying the address of one into a
4637 non-@code{const} @code{char *} pointer produces a warning. These
4638 warnings help you find at compile time code that can try to write
4639 into a string constant, but only if you have been very careful about
4640 using @code{const} in declarations and prototypes. Otherwise, it is
4641 just a nuisance. This is why we did not make @option{-Wall} request
4644 When compiling C++, warn about the deprecated conversion from string
4645 literals to @code{char *}. This warning is enabled by default for C++
4650 @opindex Wno-clobbered
4651 Warn for variables that might be changed by @samp{longjmp} or
4652 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4654 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
4655 @opindex Wconditionally-supported
4656 @opindex Wno-conditionally-supported
4657 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
4660 @opindex Wconversion
4661 @opindex Wno-conversion
4662 Warn for implicit conversions that may alter a value. This includes
4663 conversions between real and integer, like @code{abs (x)} when
4664 @code{x} is @code{double}; conversions between signed and unsigned,
4665 like @code{unsigned ui = -1}; and conversions to smaller types, like
4666 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4667 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4668 changed by the conversion like in @code{abs (2.0)}. Warnings about
4669 conversions between signed and unsigned integers can be disabled by
4670 using @option{-Wno-sign-conversion}.
4672 For C++, also warn for confusing overload resolution for user-defined
4673 conversions; and conversions that never use a type conversion
4674 operator: conversions to @code{void}, the same type, a base class or a
4675 reference to them. Warnings about conversions between signed and
4676 unsigned integers are disabled by default in C++ unless
4677 @option{-Wsign-conversion} is explicitly enabled.
4679 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4680 @opindex Wconversion-null
4681 @opindex Wno-conversion-null
4682 Do not warn for conversions between @code{NULL} and non-pointer
4683 types. @option{-Wconversion-null} is enabled by default.
4685 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
4686 @opindex Wzero-as-null-pointer-constant
4687 @opindex Wno-zero-as-null-pointer-constant
4688 Warn when a literal '0' is used as null pointer constant. This can
4689 be useful to facilitate the conversion to @code{nullptr} in C++11.
4693 @opindex Wno-date-time
4694 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
4695 are encountered as they might prevent bit-wise-identical reproducible
4698 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
4699 @opindex Wdelete-incomplete
4700 @opindex Wno-delete-incomplete
4701 Warn when deleting a pointer to incomplete type, which may cause
4702 undefined behavior at runtime. This warning is enabled by default.
4704 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
4705 @opindex Wuseless-cast
4706 @opindex Wno-useless-cast
4707 Warn when an expression is casted to its own type.
4710 @opindex Wempty-body
4711 @opindex Wno-empty-body
4712 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4713 while} statement. This warning is also enabled by @option{-Wextra}.
4715 @item -Wenum-compare
4716 @opindex Wenum-compare
4717 @opindex Wno-enum-compare
4718 Warn about a comparison between values of different enumerated types.
4719 In C++ enumeral mismatches in conditional expressions are also
4720 diagnosed and the warning is enabled by default. In C this warning is
4721 enabled by @option{-Wall}.
4723 @item -Wjump-misses-init @r{(C, Objective-C only)}
4724 @opindex Wjump-misses-init
4725 @opindex Wno-jump-misses-init
4726 Warn if a @code{goto} statement or a @code{switch} statement jumps
4727 forward across the initialization of a variable, or jumps backward to a
4728 label after the variable has been initialized. This only warns about
4729 variables that are initialized when they are declared. This warning is
4730 only supported for C and Objective-C; in C++ this sort of branch is an
4733 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4734 can be disabled with the @option{-Wno-jump-misses-init} option.
4736 @item -Wsign-compare
4737 @opindex Wsign-compare
4738 @opindex Wno-sign-compare
4739 @cindex warning for comparison of signed and unsigned values
4740 @cindex comparison of signed and unsigned values, warning
4741 @cindex signed and unsigned values, comparison warning
4742 Warn when a comparison between signed and unsigned values could produce
4743 an incorrect result when the signed value is converted to unsigned.
4744 This warning is also enabled by @option{-Wextra}; to get the other warnings
4745 of @option{-Wextra} without this warning, use @option{-Wextra -Wno-sign-compare}.
4747 @item -Wsign-conversion
4748 @opindex Wsign-conversion
4749 @opindex Wno-sign-conversion
4750 Warn for implicit conversions that may change the sign of an integer
4751 value, like assigning a signed integer expression to an unsigned
4752 integer variable. An explicit cast silences the warning. In C, this
4753 option is enabled also by @option{-Wconversion}.
4755 @item -Wfloat-conversion
4756 @opindex Wfloat-conversion
4757 @opindex Wno-float-conversion
4758 Warn for implicit conversions that reduce the precision of a real value.
4759 This includes conversions from real to integer, and from higher precision
4760 real to lower precision real values. This option is also enabled by
4761 @option{-Wconversion}.
4763 @item -Wsizeof-pointer-memaccess
4764 @opindex Wsizeof-pointer-memaccess
4765 @opindex Wno-sizeof-pointer-memaccess
4766 Warn for suspicious length parameters to certain string and memory built-in
4767 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
4768 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
4769 but a pointer, and suggests a possible fix, or about
4770 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
4773 @item -Wsizeof-array-argument
4774 @opindex Wsizeof-array-argument
4775 @opindex Wno-sizeof-array-argument
4776 Warn when the @code{sizeof} operator is applied to a parameter that is
4777 declared as an array in a function definition. This warning is enabled by
4778 default for C and C++ programs.
4780 @item -Wmemset-transposed-args
4781 @opindex Wmemset-transposed-args
4782 @opindex Wno-memset-transposed-args
4783 Warn for suspicious calls to the @code{memset} built-in function, if the
4784 second argument is not zero and the third argument is zero. This warns e.g.@
4785 about @code{memset (buf, sizeof buf, 0)} where most probably
4786 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
4787 is only emitted if the third argument is literal zero, if it is some expression
4788 that is folded to zero, or e.g. a cast of zero to some type etc., it
4789 is far less likely that user has mistakenly exchanged the arguments and
4790 no warning is emitted. This warning is enabled by @option{-Wall}.
4794 @opindex Wno-address
4795 Warn about suspicious uses of memory addresses. These include using
4796 the address of a function in a conditional expression, such as
4797 @code{void func(void); if (func)}, and comparisons against the memory
4798 address of a string literal, such as @code{if (x == "abc")}. Such
4799 uses typically indicate a programmer error: the address of a function
4800 always evaluates to true, so their use in a conditional usually
4801 indicate that the programmer forgot the parentheses in a function
4802 call; and comparisons against string literals result in unspecified
4803 behavior and are not portable in C, so they usually indicate that the
4804 programmer intended to use @code{strcmp}. This warning is enabled by
4808 @opindex Wlogical-op
4809 @opindex Wno-logical-op
4810 Warn about suspicious uses of logical operators in expressions.
4811 This includes using logical operators in contexts where a
4812 bit-wise operator is likely to be expected.
4814 @item -Wlogical-not-parentheses
4815 @opindex Wlogical-not-parentheses
4816 @opindex Wno-logical-not-parentheses
4817 Warn about logical not used on the left hand side operand of a comparison.
4818 This option does not warn if the RHS operand is of a boolean type. Its
4819 purpose is to detect suspicious code like the following:
4823 if (!a > 1) @{ @dots{} @}
4826 It is possible to suppress the warning by wrapping the LHS into
4829 if ((!a) > 1) @{ @dots{} @}
4832 This warning is enabled by @option{-Wall}.
4834 @item -Waggregate-return
4835 @opindex Waggregate-return
4836 @opindex Wno-aggregate-return
4837 Warn if any functions that return structures or unions are defined or
4838 called. (In languages where you can return an array, this also elicits
4841 @item -Wno-aggressive-loop-optimizations
4842 @opindex Wno-aggressive-loop-optimizations
4843 @opindex Waggressive-loop-optimizations
4844 Warn if in a loop with constant number of iterations the compiler detects
4845 undefined behavior in some statement during one or more of the iterations.
4847 @item -Wno-attributes
4848 @opindex Wno-attributes
4849 @opindex Wattributes
4850 Do not warn if an unexpected @code{__attribute__} is used, such as
4851 unrecognized attributes, function attributes applied to variables,
4852 etc. This does not stop errors for incorrect use of supported
4855 @item -Wno-builtin-macro-redefined
4856 @opindex Wno-builtin-macro-redefined
4857 @opindex Wbuiltin-macro-redefined
4858 Do not warn if certain built-in macros are redefined. This suppresses
4859 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4860 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4862 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4863 @opindex Wstrict-prototypes
4864 @opindex Wno-strict-prototypes
4865 Warn if a function is declared or defined without specifying the
4866 argument types. (An old-style function definition is permitted without
4867 a warning if preceded by a declaration that specifies the argument
4870 @item -Wold-style-declaration @r{(C and Objective-C only)}
4871 @opindex Wold-style-declaration
4872 @opindex Wno-old-style-declaration
4873 Warn for obsolescent usages, according to the C Standard, in a
4874 declaration. For example, warn if storage-class specifiers like
4875 @code{static} are not the first things in a declaration. This warning
4876 is also enabled by @option{-Wextra}.
4878 @item -Wold-style-definition @r{(C and Objective-C only)}
4879 @opindex Wold-style-definition
4880 @opindex Wno-old-style-definition
4881 Warn if an old-style function definition is used. A warning is given
4882 even if there is a previous prototype.
4884 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4885 @opindex Wmissing-parameter-type
4886 @opindex Wno-missing-parameter-type
4887 A function parameter is declared without a type specifier in K&R-style
4894 This warning is also enabled by @option{-Wextra}.
4896 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4897 @opindex Wmissing-prototypes
4898 @opindex Wno-missing-prototypes
4899 Warn if a global function is defined without a previous prototype
4900 declaration. This warning is issued even if the definition itself
4901 provides a prototype. Use this option to detect global functions
4902 that do not have a matching prototype declaration in a header file.
4903 This option is not valid for C++ because all function declarations
4904 provide prototypes and a non-matching declaration will declare an
4905 overload rather than conflict with an earlier declaration.
4906 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
4908 @item -Wmissing-declarations
4909 @opindex Wmissing-declarations
4910 @opindex Wno-missing-declarations
4911 Warn if a global function is defined without a previous declaration.
4912 Do so even if the definition itself provides a prototype.
4913 Use this option to detect global functions that are not declared in
4914 header files. In C, no warnings are issued for functions with previous
4915 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
4916 missing prototypes. In C++, no warnings are issued for function templates,
4917 or for inline functions, or for functions in anonymous namespaces.
4919 @item -Wmissing-field-initializers
4920 @opindex Wmissing-field-initializers
4921 @opindex Wno-missing-field-initializers
4925 Warn if a structure's initializer has some fields missing. For
4926 example, the following code causes such a warning, because
4927 @code{x.h} is implicitly zero:
4930 struct s @{ int f, g, h; @};
4931 struct s x = @{ 3, 4 @};
4934 This option does not warn about designated initializers, so the following
4935 modification does not trigger a warning:
4938 struct s @{ int f, g, h; @};
4939 struct s x = @{ .f = 3, .g = 4 @};
4942 In C++ this option does not warn either about the empty @{ @}
4943 initializer, for example:
4946 struct s @{ int f, g, h; @};
4950 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4951 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
4953 @item -Wno-multichar
4954 @opindex Wno-multichar
4956 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4957 Usually they indicate a typo in the user's code, as they have
4958 implementation-defined values, and should not be used in portable code.
4960 @item -Wnormalized@r{[}=@r{<}none@r{|}id@r{|}nfc@r{|}nfkc@r{>]}
4961 @opindex Wnormalized=
4962 @opindex Wnormalized
4963 @opindex Wno-normalized
4966 @cindex character set, input normalization
4967 In ISO C and ISO C++, two identifiers are different if they are
4968 different sequences of characters. However, sometimes when characters
4969 outside the basic ASCII character set are used, you can have two
4970 different character sequences that look the same. To avoid confusion,
4971 the ISO 10646 standard sets out some @dfn{normalization rules} which
4972 when applied ensure that two sequences that look the same are turned into
4973 the same sequence. GCC can warn you if you are using identifiers that
4974 have not been normalized; this option controls that warning.
4976 There are four levels of warning supported by GCC@. The default is
4977 @option{-Wnormalized=nfc}, which warns about any identifier that is
4978 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4979 recommended form for most uses. It is equivalent to
4980 @option{-Wnormalized}.
4982 Unfortunately, there are some characters allowed in identifiers by
4983 ISO C and ISO C++ that, when turned into NFC, are not allowed in
4984 identifiers. That is, there's no way to use these symbols in portable
4985 ISO C or C++ and have all your identifiers in NFC@.
4986 @option{-Wnormalized=id} suppresses the warning for these characters.
4987 It is hoped that future versions of the standards involved will correct
4988 this, which is why this option is not the default.
4990 You can switch the warning off for all characters by writing
4991 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
4992 only do this if you are using some other normalization scheme (like
4993 ``D''), because otherwise you can easily create bugs that are
4994 literally impossible to see.
4996 Some characters in ISO 10646 have distinct meanings but look identical
4997 in some fonts or display methodologies, especially once formatting has
4998 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4999 LETTER N'', displays just like a regular @code{n} that has been
5000 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
5001 normalization scheme to convert all these into a standard form as
5002 well, and GCC warns if your code is not in NFKC if you use
5003 @option{-Wnormalized=nfkc}. This warning is comparable to warning
5004 about every identifier that contains the letter O because it might be
5005 confused with the digit 0, and so is not the default, but may be
5006 useful as a local coding convention if the programming environment
5007 cannot be fixed to display these characters distinctly.
5009 @item -Wno-deprecated
5010 @opindex Wno-deprecated
5011 @opindex Wdeprecated
5012 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
5014 @item -Wno-deprecated-declarations
5015 @opindex Wno-deprecated-declarations
5016 @opindex Wdeprecated-declarations
5017 Do not warn about uses of functions (@pxref{Function Attributes}),
5018 variables (@pxref{Variable Attributes}), and types (@pxref{Type
5019 Attributes}) marked as deprecated by using the @code{deprecated}
5023 @opindex Wno-overflow
5025 Do not warn about compile-time overflow in constant expressions.
5030 Warn about One Definition Rule violations during link-time optimization.
5031 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
5034 @opindex Wopenm-simd
5035 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
5036 simd directive set by user. The @option{-fsimd-cost-model=unlimited} can
5037 be used to relax the cost model.
5039 @item -Woverride-init @r{(C and Objective-C only)}
5040 @opindex Woverride-init
5041 @opindex Wno-override-init
5045 Warn if an initialized field without side effects is overridden when
5046 using designated initializers (@pxref{Designated Inits, , Designated
5049 This warning is included in @option{-Wextra}. To get other
5050 @option{-Wextra} warnings without this one, use @option{-Wextra
5051 -Wno-override-init}.
5056 Warn if a structure is given the packed attribute, but the packed
5057 attribute has no effect on the layout or size of the structure.
5058 Such structures may be mis-aligned for little benefit. For
5059 instance, in this code, the variable @code{f.x} in @code{struct bar}
5060 is misaligned even though @code{struct bar} does not itself
5061 have the packed attribute:
5068 @} __attribute__((packed));
5076 @item -Wpacked-bitfield-compat
5077 @opindex Wpacked-bitfield-compat
5078 @opindex Wno-packed-bitfield-compat
5079 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
5080 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
5081 the change can lead to differences in the structure layout. GCC
5082 informs you when the offset of such a field has changed in GCC 4.4.
5083 For example there is no longer a 4-bit padding between field @code{a}
5084 and @code{b} in this structure:
5091 @} __attribute__ ((packed));
5094 This warning is enabled by default. Use
5095 @option{-Wno-packed-bitfield-compat} to disable this warning.
5100 Warn if padding is included in a structure, either to align an element
5101 of the structure or to align the whole structure. Sometimes when this
5102 happens it is possible to rearrange the fields of the structure to
5103 reduce the padding and so make the structure smaller.
5105 @item -Wredundant-decls
5106 @opindex Wredundant-decls
5107 @opindex Wno-redundant-decls
5108 Warn if anything is declared more than once in the same scope, even in
5109 cases where multiple declaration is valid and changes nothing.
5111 @item -Wnested-externs @r{(C and Objective-C only)}
5112 @opindex Wnested-externs
5113 @opindex Wno-nested-externs
5114 Warn if an @code{extern} declaration is encountered within a function.
5116 @item -Wno-inherited-variadic-ctor
5117 @opindex Winherited-variadic-ctor
5118 @opindex Wno-inherited-variadic-ctor
5119 Suppress warnings about use of C++11 inheriting constructors when the
5120 base class inherited from has a C variadic constructor; the warning is
5121 on by default because the ellipsis is not inherited.
5126 Warn if a function that is declared as inline cannot be inlined.
5127 Even with this option, the compiler does not warn about failures to
5128 inline functions declared in system headers.
5130 The compiler uses a variety of heuristics to determine whether or not
5131 to inline a function. For example, the compiler takes into account
5132 the size of the function being inlined and the amount of inlining
5133 that has already been done in the current function. Therefore,
5134 seemingly insignificant changes in the source program can cause the
5135 warnings produced by @option{-Winline} to appear or disappear.
5137 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
5138 @opindex Wno-invalid-offsetof
5139 @opindex Winvalid-offsetof
5140 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
5141 type. According to the 2014 ISO C++ standard, applying @samp{offsetof}
5142 to a non-standard-layout type is undefined. In existing C++ implementations,
5143 however, @samp{offsetof} typically gives meaningful results.
5144 This flag is for users who are aware that they are
5145 writing nonportable code and who have deliberately chosen to ignore the
5148 The restrictions on @samp{offsetof} may be relaxed in a future version
5149 of the C++ standard.
5151 @item -Wno-int-to-pointer-cast
5152 @opindex Wno-int-to-pointer-cast
5153 @opindex Wint-to-pointer-cast
5154 Suppress warnings from casts to pointer type of an integer of a
5155 different size. In C++, casting to a pointer type of smaller size is
5156 an error. @option{Wint-to-pointer-cast} is enabled by default.
5159 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
5160 @opindex Wno-pointer-to-int-cast
5161 @opindex Wpointer-to-int-cast
5162 Suppress warnings from casts from a pointer to an integer type of a
5166 @opindex Winvalid-pch
5167 @opindex Wno-invalid-pch
5168 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
5169 the search path but can't be used.
5173 @opindex Wno-long-long
5174 Warn if @samp{long long} type is used. This is enabled by either
5175 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
5176 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
5178 @item -Wvariadic-macros
5179 @opindex Wvariadic-macros
5180 @opindex Wno-variadic-macros
5181 Warn if variadic macros are used in ISO C90 mode, or if the GNU
5182 alternate syntax is used in ISO C99 mode. This is enabled by either
5183 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
5184 messages, use @option{-Wno-variadic-macros}.
5188 @opindex Wno-varargs
5189 Warn upon questionable usage of the macros used to handle variable
5190 arguments like @samp{va_start}. This is default. To inhibit the
5191 warning messages, use @option{-Wno-varargs}.
5193 @item -Wvector-operation-performance
5194 @opindex Wvector-operation-performance
5195 @opindex Wno-vector-operation-performance
5196 Warn if vector operation is not implemented via SIMD capabilities of the
5197 architecture. Mainly useful for the performance tuning.
5198 Vector operation can be implemented @code{piecewise}, which means that the
5199 scalar operation is performed on every vector element;
5200 @code{in parallel}, which means that the vector operation is implemented
5201 using scalars of wider type, which normally is more performance efficient;
5202 and @code{as a single scalar}, which means that vector fits into a
5205 @item -Wno-virtual-move-assign
5206 @opindex Wvirtual-move-assign
5207 @opindex Wno-virtual-move-assign
5208 Suppress warnings about inheriting from a virtual base with a
5209 non-trivial C++11 move assignment operator. This is dangerous because
5210 if the virtual base is reachable along more than one path, it will be
5211 moved multiple times, which can mean both objects end up in the
5212 moved-from state. If the move assignment operator is written to avoid
5213 moving from a moved-from object, this warning can be disabled.
5218 Warn if variable length array is used in the code.
5219 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
5220 the variable length array.
5222 @item -Wvolatile-register-var
5223 @opindex Wvolatile-register-var
5224 @opindex Wno-volatile-register-var
5225 Warn if a register variable is declared volatile. The volatile
5226 modifier does not inhibit all optimizations that may eliminate reads
5227 and/or writes to register variables. This warning is enabled by
5230 @item -Wdisabled-optimization
5231 @opindex Wdisabled-optimization
5232 @opindex Wno-disabled-optimization
5233 Warn if a requested optimization pass is disabled. This warning does
5234 not generally indicate that there is anything wrong with your code; it
5235 merely indicates that GCC's optimizers are unable to handle the code
5236 effectively. Often, the problem is that your code is too big or too
5237 complex; GCC refuses to optimize programs when the optimization
5238 itself is likely to take inordinate amounts of time.
5240 @item -Wpointer-sign @r{(C and Objective-C only)}
5241 @opindex Wpointer-sign
5242 @opindex Wno-pointer-sign
5243 Warn for pointer argument passing or assignment with different signedness.
5244 This option is only supported for C and Objective-C@. It is implied by
5245 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
5246 @option{-Wno-pointer-sign}.
5248 @item -Wstack-protector
5249 @opindex Wstack-protector
5250 @opindex Wno-stack-protector
5251 This option is only active when @option{-fstack-protector} is active. It
5252 warns about functions that are not protected against stack smashing.
5254 @item -Woverlength-strings
5255 @opindex Woverlength-strings
5256 @opindex Wno-overlength-strings
5257 Warn about string constants that are longer than the ``minimum
5258 maximum'' length specified in the C standard. Modern compilers
5259 generally allow string constants that are much longer than the
5260 standard's minimum limit, but very portable programs should avoid
5261 using longer strings.
5263 The limit applies @emph{after} string constant concatenation, and does
5264 not count the trailing NUL@. In C90, the limit was 509 characters; in
5265 C99, it was raised to 4095. C++98 does not specify a normative
5266 minimum maximum, so we do not diagnose overlength strings in C++@.
5268 This option is implied by @option{-Wpedantic}, and can be disabled with
5269 @option{-Wno-overlength-strings}.
5271 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
5272 @opindex Wunsuffixed-float-constants
5274 Issue a warning for any floating constant that does not have
5275 a suffix. When used together with @option{-Wsystem-headers} it
5276 warns about such constants in system header files. This can be useful
5277 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
5278 from the decimal floating-point extension to C99.
5280 @item -Wno-designated-init @r{(C and Objective-C only)}
5281 Suppress warnings when a positional initializer is used to initialize
5282 a structure that has been marked with the @code{designated_init}
5287 @node Debugging Options
5288 @section Options for Debugging Your Program or GCC
5289 @cindex options, debugging
5290 @cindex debugging information options
5292 GCC has various special options that are used for debugging
5293 either your program or GCC:
5298 Produce debugging information in the operating system's native format
5299 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
5302 On most systems that use stabs format, @option{-g} enables use of extra
5303 debugging information that only GDB can use; this extra information
5304 makes debugging work better in GDB but probably makes other debuggers
5306 refuse to read the program. If you want to control for certain whether
5307 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
5308 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
5310 GCC allows you to use @option{-g} with
5311 @option{-O}. The shortcuts taken by optimized code may occasionally
5312 produce surprising results: some variables you declared may not exist
5313 at all; flow of control may briefly move where you did not expect it;
5314 some statements may not be executed because they compute constant
5315 results or their values are already at hand; some statements may
5316 execute in different places because they have been moved out of loops.
5318 Nevertheless it proves possible to debug optimized output. This makes
5319 it reasonable to use the optimizer for programs that might have bugs.
5321 The following options are useful when GCC is generated with the
5322 capability for more than one debugging format.
5325 @opindex gsplit-dwarf
5326 Separate as much dwarf debugging information as possible into a
5327 separate output file with the extension .dwo. This option allows
5328 the build system to avoid linking files with debug information. To
5329 be useful, this option requires a debugger capable of reading .dwo
5334 Produce debugging information for use by GDB@. This means to use the
5335 most expressive format available (DWARF 2, stabs, or the native format
5336 if neither of those are supported), including GDB extensions if at all
5341 Generate dwarf .debug_pubnames and .debug_pubtypes sections.
5343 @item -ggnu-pubnames
5344 @opindex ggnu-pubnames
5345 Generate .debug_pubnames and .debug_pubtypes sections in a format
5346 suitable for conversion into a GDB@ index. This option is only useful
5347 with a linker that can produce GDB@ index version 7.
5351 Produce debugging information in stabs format (if that is supported),
5352 without GDB extensions. This is the format used by DBX on most BSD
5353 systems. On MIPS, Alpha and System V Release 4 systems this option
5354 produces stabs debugging output that is not understood by DBX or SDB@.
5355 On System V Release 4 systems this option requires the GNU assembler.
5357 @item -feliminate-unused-debug-symbols
5358 @opindex feliminate-unused-debug-symbols
5359 Produce debugging information in stabs format (if that is supported),
5360 for only symbols that are actually used.
5362 @item -femit-class-debug-always
5363 Instead of emitting debugging information for a C++ class in only one
5364 object file, emit it in all object files using the class. This option
5365 should be used only with debuggers that are unable to handle the way GCC
5366 normally emits debugging information for classes because using this
5367 option increases the size of debugging information by as much as a
5370 @item -fdebug-types-section
5371 @opindex fdebug-types-section
5372 @opindex fno-debug-types-section
5373 When using DWARF Version 4 or higher, type DIEs can be put into
5374 their own @code{.debug_types} section instead of making them part of the
5375 @code{.debug_info} section. It is more efficient to put them in a separate
5376 comdat sections since the linker can then remove duplicates.
5377 But not all DWARF consumers support @code{.debug_types} sections yet
5378 and on some objects @code{.debug_types} produces larger instead of smaller
5379 debugging information.
5383 Produce debugging information in stabs format (if that is supported),
5384 using GNU extensions understood only by the GNU debugger (GDB)@. The
5385 use of these extensions is likely to make other debuggers crash or
5386 refuse to read the program.
5390 Produce debugging information in COFF format (if that is supported).
5391 This is the format used by SDB on most System V systems prior to
5396 Produce debugging information in XCOFF format (if that is supported).
5397 This is the format used by the DBX debugger on IBM RS/6000 systems.
5401 Produce debugging information in XCOFF format (if that is supported),
5402 using GNU extensions understood only by the GNU debugger (GDB)@. The
5403 use of these extensions is likely to make other debuggers crash or
5404 refuse to read the program, and may cause assemblers other than the GNU
5405 assembler (GAS) to fail with an error.
5407 @item -gdwarf-@var{version}
5408 @opindex gdwarf-@var{version}
5409 Produce debugging information in DWARF format (if that is supported).
5410 The value of @var{version} may be either 2, 3 or 4; the default version
5411 for most targets is 4.
5413 Note that with DWARF Version 2, some ports require and always
5414 use some non-conflicting DWARF 3 extensions in the unwind tables.
5416 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
5417 for maximum benefit.
5419 @item -grecord-gcc-switches
5420 @opindex grecord-gcc-switches
5421 This switch causes the command-line options used to invoke the
5422 compiler that may affect code generation to be appended to the
5423 DW_AT_producer attribute in DWARF debugging information. The options
5424 are concatenated with spaces separating them from each other and from
5425 the compiler version. See also @option{-frecord-gcc-switches} for another
5426 way of storing compiler options into the object file. This is the default.
5428 @item -gno-record-gcc-switches
5429 @opindex gno-record-gcc-switches
5430 Disallow appending command-line options to the DW_AT_producer attribute
5431 in DWARF debugging information.
5433 @item -gstrict-dwarf
5434 @opindex gstrict-dwarf
5435 Disallow using extensions of later DWARF standard version than selected
5436 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
5437 DWARF extensions from later standard versions is allowed.
5439 @item -gno-strict-dwarf
5440 @opindex gno-strict-dwarf
5441 Allow using extensions of later DWARF standard version than selected with
5442 @option{-gdwarf-@var{version}}.
5444 @item -gz@r{[}=@var{type}@r{]}
5446 Produce compressed debug sections in DWARF format, if that is supported.
5447 If @var{type} is not given, the default type depends on the capabilities
5448 of the assembler and linker used. @var{type} may be one of
5449 @option{none} (don't compress debug sections), @option{zlib} (use zlib
5450 compression in ELF gABI format), or @option{zlib-gnu} (use zlib
5451 compression in traditional GNU format). If the linker doesn't support
5452 writing compressed debug sections, the option is rejected. Otherwise,
5453 if the assembler does not support them, @option{-gz} is silently ignored
5454 when producing object files.
5458 Produce debugging information in Alpha/VMS debug format (if that is
5459 supported). This is the format used by DEBUG on Alpha/VMS systems.
5462 @itemx -ggdb@var{level}
5463 @itemx -gstabs@var{level}
5464 @itemx -gcoff@var{level}
5465 @itemx -gxcoff@var{level}
5466 @itemx -gvms@var{level}
5467 Request debugging information and also use @var{level} to specify how
5468 much information. The default level is 2.
5470 Level 0 produces no debug information at all. Thus, @option{-g0} negates
5473 Level 1 produces minimal information, enough for making backtraces in
5474 parts of the program that you don't plan to debug. This includes
5475 descriptions of functions and external variables, and line number
5476 tables, but no information about local variables.
5478 Level 3 includes extra information, such as all the macro definitions
5479 present in the program. Some debuggers support macro expansion when
5480 you use @option{-g3}.
5482 @option{-gdwarf-2} does not accept a concatenated debug level, because
5483 GCC used to support an option @option{-gdwarf} that meant to generate
5484 debug information in version 1 of the DWARF format (which is very
5485 different from version 2), and it would have been too confusing. That
5486 debug format is long obsolete, but the option cannot be changed now.
5487 Instead use an additional @option{-g@var{level}} option to change the
5488 debug level for DWARF.
5492 Turn off generation of debug info, if leaving out this option
5493 generates it, or turn it on at level 2 otherwise. The position of this
5494 argument in the command line does not matter; it takes effect after all
5495 other options are processed, and it does so only once, no matter how
5496 many times it is given. This is mainly intended to be used with
5497 @option{-fcompare-debug}.
5499 @item -fsanitize=address
5500 @opindex fsanitize=address
5501 Enable AddressSanitizer, a fast memory error detector.
5502 Memory access instructions will be instrumented to detect
5503 out-of-bounds and use-after-free bugs.
5504 See @uref{http://code.google.com/p/address-sanitizer/} for
5505 more details. The run-time behavior can be influenced using the
5506 @env{ASAN_OPTIONS} environment variable; see
5507 @url{https://code.google.com/p/address-sanitizer/wiki/Flags#Run-time_flags} for
5508 a list of supported options.
5510 @item -fsanitize=kernel-address
5511 @opindex fsanitize=kernel-address
5512 Enable AddressSanitizer for Linux kernel.
5513 See @uref{http://code.google.com/p/address-sanitizer/wiki/AddressSanitizerForKernel} for more details.
5515 @item -fsanitize=thread
5516 @opindex fsanitize=thread
5517 Enable ThreadSanitizer, a fast data race detector.
5518 Memory access instructions will be instrumented to detect
5519 data race bugs. See @uref{http://code.google.com/p/thread-sanitizer/} for more
5520 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
5521 environment variable; see
5522 @url{https://code.google.com/p/thread-sanitizer/wiki/Flags} for a list of
5525 @item -fsanitize=leak
5526 @opindex fsanitize=leak
5527 Enable LeakSanitizer, a memory leak detector.
5528 This option only matters for linking of executables and if neither
5529 @option{-fsanitize=address} nor @option{-fsanitize=thread} is used. In that
5530 case it will link the executable against a library that overrides @code{malloc}
5531 and other allocator functions. See
5532 @uref{https://code.google.com/p/address-sanitizer/wiki/LeakSanitizer} for more
5533 details. The run-time behavior can be influenced using the
5534 @env{LSAN_OPTIONS} environment variable.
5536 @item -fsanitize=undefined
5537 @opindex fsanitize=undefined
5538 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
5539 Various computations will be instrumented to detect undefined behavior
5540 at runtime. Current suboptions are:
5544 @item -fsanitize=shift
5545 @opindex fsanitize=shift
5546 This option enables checking that the result of a shift operation is
5547 not undefined. Note that what exactly is considered undefined differs
5548 slightly between C and C++, as well as between ISO C90 and C99, etc.
5550 @item -fsanitize=integer-divide-by-zero
5551 @opindex fsanitize=integer-divide-by-zero
5552 Detect integer division by zero as well as @code{INT_MIN / -1} division.
5554 @item -fsanitize=unreachable
5555 @opindex fsanitize=unreachable
5556 With this option, the compiler will turn the @code{__builtin_unreachable}
5557 call into a diagnostics message call instead. When reaching the
5558 @code{__builtin_unreachable} call, the behavior is undefined.
5560 @item -fsanitize=vla-bound
5561 @opindex fsanitize=vla-bound
5562 This option instructs the compiler to check that the size of a variable
5563 length array is positive. This option does not have any effect in
5564 @option{-std=c++14} mode, as the standard requires the exception be thrown
5567 @item -fsanitize=null
5568 @opindex fsanitize=null
5569 This option enables pointer checking. Particularly, the application
5570 built with this option turned on will issue an error message when it
5571 tries to dereference a NULL pointer, or if a reference (possibly an
5572 rvalue reference) is bound to a NULL pointer, or if a method is invoked
5573 on an object pointed by a NULL pointer.
5575 @item -fsanitize=return
5576 @opindex fsanitize=return
5577 This option enables return statement checking. Programs
5578 built with this option turned on will issue an error message
5579 when the end of a non-void function is reached without actually
5580 returning a value. This option works in C++ only.
5582 @item -fsanitize=signed-integer-overflow
5583 @opindex fsanitize=signed-integer-overflow
5584 This option enables signed integer overflow checking. We check that
5585 the result of @code{+}, @code{*}, and both unary and binary @code{-}
5586 does not overflow in the signed arithmetics. Note, integer promotion
5587 rules must be taken into account. That is, the following is not an
5590 signed char a = SCHAR_MAX;
5594 @item -fsanitize=bounds
5595 @opindex fsanitize=bounds
5596 This option enables instrumentation of array bounds. Various out of bounds
5597 accesses are detected. Flexible array members and initializers of variables
5598 with static storage are not instrumented.
5600 @item -fsanitize=alignment
5601 @opindex fsanitize=alignment
5603 This option enables checking of alignment of pointers when they are
5604 dereferenced, or when a reference is bound to insufficiently aligned target,
5605 or when a method or constructor is invoked on insufficiently aligned object.
5607 @item -fsanitize=object-size
5608 @opindex fsanitize=object-size
5609 This option enables instrumentation of memory references using the
5610 @code{__builtin_object_size} function. Various out of bounds pointer
5611 accesses are detected.
5613 @item -fsanitize=float-divide-by-zero
5614 @opindex fsanitize=float-divide-by-zero
5615 Detect floating-point division by zero. Unlike other similar options,
5616 @option{-fsanitize=float-divide-by-zero} is not enabled by
5617 @option{-fsanitize=undefined}, since floating-point division by zero can
5618 be a legitimate way of obtaining infinities and NaNs.
5620 @item -fsanitize=float-cast-overflow
5621 @opindex fsanitize=float-cast-overflow
5622 This option enables floating-point type to integer conversion checking.
5623 We check that the result of the conversion does not overflow.
5624 This option does not work well with @code{FE_INVALID} exceptions enabled.
5626 @item -fsanitize=nonnull-attribute
5627 @opindex fsanitize=nonnull-attribute
5629 This option enables instrumentation of calls, checking whether null values
5630 are not passed to arguments marked as requiring a non-null value by the
5631 @code{nonnull} function attribute.
5633 @item -fsanitize=returns-nonnull-attribute
5634 @opindex fsanitize=returns-nonnull-attribute
5636 This option enables instrumentation of return statements in functions
5637 marked with @code{returns_nonnull} function attribute, to detect returning
5638 of null values from such functions.
5640 @item -fsanitize=bool
5641 @opindex fsanitize=bool
5643 This option enables instrumentation of loads from bool. If a value other
5644 than 0/1 is loaded, a run-time error is issued.
5646 @item -fsanitize=enum
5647 @opindex fsanitize=enum
5649 This option enables instrumentation of loads from an enum type. If
5650 a value outside the range of values for the enum type is loaded,
5651 a run-time error is issued.
5655 While @option{-ftrapv} causes traps for signed overflows to be emitted,
5656 @option{-fsanitize=undefined} gives a diagnostic message.
5657 This currently works only for the C family of languages.
5659 @item -fasan-shadow-offset=@var{number}
5660 @opindex fasan-shadow-offset
5661 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
5662 It is useful for experimenting with different shadow memory layouts in
5663 Kernel AddressSanitizer.
5665 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
5666 @opindex fsanitize-recover
5667 @opindex fno-sanitize-recover
5668 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
5669 mentioned in comma-separated list of @var{opts}. Enabling this option
5670 for a sanitizer component would cause it to attempt to continue
5671 running the program as if no error happened. This means multiple
5672 runtime errors can be reported in a single program run, and the exit
5673 code of the program may indicate success even when errors
5674 have been reported. The @option{-fno-sanitize-recover=} can be used to alter
5675 this behavior, only the first detected error will be reported
5676 and program will exit after that with non-zero exit code.
5678 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
5679 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
5680 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero} and
5681 @option{-fsanitize=kernel-address}. For these sanitizers error recovery is turned on by default.
5683 Syntax without explicit @var{opts} parameter is deprecated. It is equivalent to
5684 @option{-fsanitize-recover=undefined,float-cast-overflow,float-divide-by-zero,kernel-address}.
5685 Similarly @option{-fno-sanitize-recover} is equivalent to
5686 @option{-fno-sanitize-recover=undefined,float-cast-overflow,float-divide-by-zero,kernel-address}.
5688 @item -fsanitize-undefined-trap-on-error
5689 @opindex fsanitize-undefined-trap-on-error
5690 The @option{-fsanitize-undefined-trap-on-error} instructs the compiler to
5691 report undefined behavior using @code{__builtin_trap ()} rather than
5692 a @code{libubsan} library routine. The advantage of this is that the
5693 @code{libubsan} library is not needed and will not be linked in, so this
5694 is usable even for use in freestanding environments.
5696 @item -fdump-final-insns@r{[}=@var{file}@r{]}
5697 @opindex fdump-final-insns
5698 Dump the final internal representation (RTL) to @var{file}. If the
5699 optional argument is omitted (or if @var{file} is @code{.}), the name
5700 of the dump file is determined by appending @code{.gkd} to the
5701 compilation output file name.
5703 @item -fcompare-debug@r{[}=@var{opts}@r{]}
5704 @opindex fcompare-debug
5705 @opindex fno-compare-debug
5706 If no error occurs during compilation, run the compiler a second time,
5707 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
5708 passed to the second compilation. Dump the final internal
5709 representation in both compilations, and print an error if they differ.
5711 If the equal sign is omitted, the default @option{-gtoggle} is used.
5713 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
5714 and nonzero, implicitly enables @option{-fcompare-debug}. If
5715 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
5716 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
5719 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
5720 is equivalent to @option{-fno-compare-debug}, which disables the dumping
5721 of the final representation and the second compilation, preventing even
5722 @env{GCC_COMPARE_DEBUG} from taking effect.
5724 To verify full coverage during @option{-fcompare-debug} testing, set
5725 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
5726 which GCC rejects as an invalid option in any actual compilation
5727 (rather than preprocessing, assembly or linking). To get just a
5728 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
5729 not overridden} will do.
5731 @item -fcompare-debug-second
5732 @opindex fcompare-debug-second
5733 This option is implicitly passed to the compiler for the second
5734 compilation requested by @option{-fcompare-debug}, along with options to
5735 silence warnings, and omitting other options that would cause
5736 side-effect compiler outputs to files or to the standard output. Dump
5737 files and preserved temporary files are renamed so as to contain the
5738 @code{.gk} additional extension during the second compilation, to avoid
5739 overwriting those generated by the first.
5741 When this option is passed to the compiler driver, it causes the
5742 @emph{first} compilation to be skipped, which makes it useful for little
5743 other than debugging the compiler proper.
5745 @item -feliminate-dwarf2-dups
5746 @opindex feliminate-dwarf2-dups
5747 Compress DWARF 2 debugging information by eliminating duplicated
5748 information about each symbol. This option only makes sense when
5749 generating DWARF 2 debugging information with @option{-gdwarf-2}.
5751 @item -femit-struct-debug-baseonly
5752 @opindex femit-struct-debug-baseonly
5753 Emit debug information for struct-like types
5754 only when the base name of the compilation source file
5755 matches the base name of file in which the struct is defined.
5757 This option substantially reduces the size of debugging information,
5758 but at significant potential loss in type information to the debugger.
5759 See @option{-femit-struct-debug-reduced} for a less aggressive option.
5760 See @option{-femit-struct-debug-detailed} for more detailed control.
5762 This option works only with DWARF 2.
5764 @item -femit-struct-debug-reduced
5765 @opindex femit-struct-debug-reduced
5766 Emit debug information for struct-like types
5767 only when the base name of the compilation source file
5768 matches the base name of file in which the type is defined,
5769 unless the struct is a template or defined in a system header.
5771 This option significantly reduces the size of debugging information,
5772 with some potential loss in type information to the debugger.
5773 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
5774 See @option{-femit-struct-debug-detailed} for more detailed control.
5776 This option works only with DWARF 2.
5778 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
5779 Specify the struct-like types
5780 for which the compiler generates debug information.
5781 The intent is to reduce duplicate struct debug information
5782 between different object files within the same program.
5784 This option is a detailed version of
5785 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
5786 which serves for most needs.
5788 A specification has the syntax@*
5789 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
5791 The optional first word limits the specification to
5792 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
5793 A struct type is used directly when it is the type of a variable, member.
5794 Indirect uses arise through pointers to structs.
5795 That is, when use of an incomplete struct is valid, the use is indirect.
5797 @samp{struct one direct; struct two * indirect;}.
5799 The optional second word limits the specification to
5800 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
5801 Generic structs are a bit complicated to explain.
5802 For C++, these are non-explicit specializations of template classes,
5803 or non-template classes within the above.
5804 Other programming languages have generics,
5805 but @option{-femit-struct-debug-detailed} does not yet implement them.
5807 The third word specifies the source files for those
5808 structs for which the compiler should emit debug information.
5809 The values @samp{none} and @samp{any} have the normal meaning.
5810 The value @samp{base} means that
5811 the base of name of the file in which the type declaration appears
5812 must match the base of the name of the main compilation file.
5813 In practice, this means that when compiling @file{foo.c}, debug information
5814 is generated for types declared in that file and @file{foo.h},
5815 but not other header files.
5816 The value @samp{sys} means those types satisfying @samp{base}
5817 or declared in system or compiler headers.
5819 You may need to experiment to determine the best settings for your application.
5821 The default is @option{-femit-struct-debug-detailed=all}.
5823 This option works only with DWARF 2.
5825 @item -fno-merge-debug-strings
5826 @opindex fmerge-debug-strings
5827 @opindex fno-merge-debug-strings
5828 Direct the linker to not merge together strings in the debugging
5829 information that are identical in different object files. Merging is
5830 not supported by all assemblers or linkers. Merging decreases the size
5831 of the debug information in the output file at the cost of increasing
5832 link processing time. Merging is enabled by default.
5834 @item -fdebug-prefix-map=@var{old}=@var{new}
5835 @opindex fdebug-prefix-map
5836 When compiling files in directory @file{@var{old}}, record debugging
5837 information describing them as in @file{@var{new}} instead.
5839 @item -fno-dwarf2-cfi-asm
5840 @opindex fdwarf2-cfi-asm
5841 @opindex fno-dwarf2-cfi-asm
5842 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
5843 instead of using GAS @code{.cfi_*} directives.
5845 @cindex @command{prof}
5848 Generate extra code to write profile information suitable for the
5849 analysis program @command{prof}. You must use this option when compiling
5850 the source files you want data about, and you must also use it when
5853 @cindex @command{gprof}
5856 Generate extra code to write profile information suitable for the
5857 analysis program @command{gprof}. You must use this option when compiling
5858 the source files you want data about, and you must also use it when
5863 Makes the compiler print out each function name as it is compiled, and
5864 print some statistics about each pass when it finishes.
5867 @opindex ftime-report
5868 Makes the compiler print some statistics about the time consumed by each
5869 pass when it finishes.
5872 @opindex fmem-report
5873 Makes the compiler print some statistics about permanent memory
5874 allocation when it finishes.
5876 @item -fmem-report-wpa
5877 @opindex fmem-report-wpa
5878 Makes the compiler print some statistics about permanent memory
5879 allocation for the WPA phase only.
5881 @item -fpre-ipa-mem-report
5882 @opindex fpre-ipa-mem-report
5883 @item -fpost-ipa-mem-report
5884 @opindex fpost-ipa-mem-report
5885 Makes the compiler print some statistics about permanent memory
5886 allocation before or after interprocedural optimization.
5888 @item -fprofile-report
5889 @opindex fprofile-report
5890 Makes the compiler print some statistics about consistency of the
5891 (estimated) profile and effect of individual passes.
5894 @opindex fstack-usage
5895 Makes the compiler output stack usage information for the program, on a
5896 per-function basis. The filename for the dump is made by appending
5897 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
5898 the output file, if explicitly specified and it is not an executable,
5899 otherwise it is the basename of the source file. An entry is made up
5904 The name of the function.
5908 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
5911 The qualifier @code{static} means that the function manipulates the stack
5912 statically: a fixed number of bytes are allocated for the frame on function
5913 entry and released on function exit; no stack adjustments are otherwise made
5914 in the function. The second field is this fixed number of bytes.
5916 The qualifier @code{dynamic} means that the function manipulates the stack
5917 dynamically: in addition to the static allocation described above, stack
5918 adjustments are made in the body of the function, for example to push/pop
5919 arguments around function calls. If the qualifier @code{bounded} is also
5920 present, the amount of these adjustments is bounded at compile time and
5921 the second field is an upper bound of the total amount of stack used by
5922 the function. If it is not present, the amount of these adjustments is
5923 not bounded at compile time and the second field only represents the
5926 @item -fprofile-arcs
5927 @opindex fprofile-arcs
5928 Add code so that program flow @dfn{arcs} are instrumented. During
5929 execution the program records how many times each branch and call is
5930 executed and how many times it is taken or returns. When the compiled
5931 program exits it saves this data to a file called
5932 @file{@var{auxname}.gcda} for each source file. The data may be used for
5933 profile-directed optimizations (@option{-fbranch-probabilities}), or for
5934 test coverage analysis (@option{-ftest-coverage}). Each object file's
5935 @var{auxname} is generated from the name of the output file, if
5936 explicitly specified and it is not the final executable, otherwise it is
5937 the basename of the source file. In both cases any suffix is removed
5938 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
5939 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
5940 @xref{Cross-profiling}.
5942 @cindex @command{gcov}
5946 This option is used to compile and link code instrumented for coverage
5947 analysis. The option is a synonym for @option{-fprofile-arcs}
5948 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
5949 linking). See the documentation for those options for more details.
5954 Compile the source files with @option{-fprofile-arcs} plus optimization
5955 and code generation options. For test coverage analysis, use the
5956 additional @option{-ftest-coverage} option. You do not need to profile
5957 every source file in a program.
5960 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
5961 (the latter implies the former).
5964 Run the program on a representative workload to generate the arc profile
5965 information. This may be repeated any number of times. You can run
5966 concurrent instances of your program, and provided that the file system
5967 supports locking, the data files will be correctly updated. Also
5968 @code{fork} calls are detected and correctly handled (double counting
5972 For profile-directed optimizations, compile the source files again with
5973 the same optimization and code generation options plus
5974 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
5975 Control Optimization}).
5978 For test coverage analysis, use @command{gcov} to produce human readable
5979 information from the @file{.gcno} and @file{.gcda} files. Refer to the
5980 @command{gcov} documentation for further information.
5984 With @option{-fprofile-arcs}, for each function of your program GCC
5985 creates a program flow graph, then finds a spanning tree for the graph.
5986 Only arcs that are not on the spanning tree have to be instrumented: the
5987 compiler adds code to count the number of times that these arcs are
5988 executed. When an arc is the only exit or only entrance to a block, the
5989 instrumentation code can be added to the block; otherwise, a new basic
5990 block must be created to hold the instrumentation code.
5993 @item -ftest-coverage
5994 @opindex ftest-coverage
5995 Produce a notes file that the @command{gcov} code-coverage utility
5996 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
5997 show program coverage. Each source file's note file is called
5998 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
5999 above for a description of @var{auxname} and instructions on how to
6000 generate test coverage data. Coverage data matches the source files
6001 more closely if you do not optimize.
6003 @item -fdbg-cnt-list
6004 @opindex fdbg-cnt-list
6005 Print the name and the counter upper bound for all debug counters.
6008 @item -fdbg-cnt=@var{counter-value-list}
6010 Set the internal debug counter upper bound. @var{counter-value-list}
6011 is a comma-separated list of @var{name}:@var{value} pairs
6012 which sets the upper bound of each debug counter @var{name} to @var{value}.
6013 All debug counters have the initial upper bound of @code{UINT_MAX};
6014 thus @code{dbg_cnt()} returns true always unless the upper bound
6015 is set by this option.
6016 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
6017 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
6019 @item -fenable-@var{kind}-@var{pass}
6020 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
6024 This is a set of options that are used to explicitly disable/enable
6025 optimization passes. These options are intended for use for debugging GCC.
6026 Compiler users should use regular options for enabling/disabling
6031 @item -fdisable-ipa-@var{pass}
6032 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6033 statically invoked in the compiler multiple times, the pass name should be
6034 appended with a sequential number starting from 1.
6036 @item -fdisable-rtl-@var{pass}
6037 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
6038 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
6039 statically invoked in the compiler multiple times, the pass name should be
6040 appended with a sequential number starting from 1. @var{range-list} is a
6041 comma-separated list of function ranges or assembler names. Each range is a number
6042 pair separated by a colon. The range is inclusive in both ends. If the range
6043 is trivial, the number pair can be simplified as a single number. If the
6044 function's call graph node's @var{uid} falls within one of the specified ranges,
6045 the @var{pass} is disabled for that function. The @var{uid} is shown in the
6046 function header of a dump file, and the pass names can be dumped by using
6047 option @option{-fdump-passes}.
6049 @item -fdisable-tree-@var{pass}
6050 @itemx -fdisable-tree-@var{pass}=@var{range-list}
6051 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
6054 @item -fenable-ipa-@var{pass}
6055 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6056 statically invoked in the compiler multiple times, the pass name should be
6057 appended with a sequential number starting from 1.
6059 @item -fenable-rtl-@var{pass}
6060 @itemx -fenable-rtl-@var{pass}=@var{range-list}
6061 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
6062 description and examples.
6064 @item -fenable-tree-@var{pass}
6065 @itemx -fenable-tree-@var{pass}=@var{range-list}
6066 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
6067 of option arguments.
6071 Here are some examples showing uses of these options.
6075 # disable ccp1 for all functions
6077 # disable complete unroll for function whose cgraph node uid is 1
6078 -fenable-tree-cunroll=1
6079 # disable gcse2 for functions at the following ranges [1,1],
6080 # [300,400], and [400,1000]
6081 # disable gcse2 for functions foo and foo2
6082 -fdisable-rtl-gcse2=foo,foo2
6083 # disable early inlining
6084 -fdisable-tree-einline
6085 # disable ipa inlining
6086 -fdisable-ipa-inline
6087 # enable tree full unroll
6088 -fenable-tree-unroll
6092 @item -d@var{letters}
6093 @itemx -fdump-rtl-@var{pass}
6094 @itemx -fdump-rtl-@var{pass}=@var{filename}
6096 @opindex fdump-rtl-@var{pass}
6097 Says to make debugging dumps during compilation at times specified by
6098 @var{letters}. This is used for debugging the RTL-based passes of the
6099 compiler. The file names for most of the dumps are made by appending
6100 a pass number and a word to the @var{dumpname}, and the files are
6101 created in the directory of the output file. In case of
6102 @option{=@var{filename}} option, the dump is output on the given file
6103 instead of the pass numbered dump files. Note that the pass number is
6104 computed statically as passes get registered into the pass manager.
6105 Thus the numbering is not related to the dynamic order of execution of
6106 passes. In particular, a pass installed by a plugin could have a
6107 number over 200 even if it executed quite early. @var{dumpname} is
6108 generated from the name of the output file, if explicitly specified
6109 and it is not an executable, otherwise it is the basename of the
6110 source file. These switches may have different effects when
6111 @option{-E} is used for preprocessing.
6113 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
6114 @option{-d} option @var{letters}. Here are the possible
6115 letters for use in @var{pass} and @var{letters}, and their meanings:
6119 @item -fdump-rtl-alignments
6120 @opindex fdump-rtl-alignments
6121 Dump after branch alignments have been computed.
6123 @item -fdump-rtl-asmcons
6124 @opindex fdump-rtl-asmcons
6125 Dump after fixing rtl statements that have unsatisfied in/out constraints.
6127 @item -fdump-rtl-auto_inc_dec
6128 @opindex fdump-rtl-auto_inc_dec
6129 Dump after auto-inc-dec discovery. This pass is only run on
6130 architectures that have auto inc or auto dec instructions.
6132 @item -fdump-rtl-barriers
6133 @opindex fdump-rtl-barriers
6134 Dump after cleaning up the barrier instructions.
6136 @item -fdump-rtl-bbpart
6137 @opindex fdump-rtl-bbpart
6138 Dump after partitioning hot and cold basic blocks.
6140 @item -fdump-rtl-bbro
6141 @opindex fdump-rtl-bbro
6142 Dump after block reordering.
6144 @item -fdump-rtl-btl1
6145 @itemx -fdump-rtl-btl2
6146 @opindex fdump-rtl-btl2
6147 @opindex fdump-rtl-btl2
6148 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
6149 after the two branch
6150 target load optimization passes.
6152 @item -fdump-rtl-bypass
6153 @opindex fdump-rtl-bypass
6154 Dump after jump bypassing and control flow optimizations.
6156 @item -fdump-rtl-combine
6157 @opindex fdump-rtl-combine
6158 Dump after the RTL instruction combination pass.
6160 @item -fdump-rtl-compgotos
6161 @opindex fdump-rtl-compgotos
6162 Dump after duplicating the computed gotos.
6164 @item -fdump-rtl-ce1
6165 @itemx -fdump-rtl-ce2
6166 @itemx -fdump-rtl-ce3
6167 @opindex fdump-rtl-ce1
6168 @opindex fdump-rtl-ce2
6169 @opindex fdump-rtl-ce3
6170 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
6171 @option{-fdump-rtl-ce3} enable dumping after the three
6172 if conversion passes.
6174 @item -fdump-rtl-cprop_hardreg
6175 @opindex fdump-rtl-cprop_hardreg
6176 Dump after hard register copy propagation.
6178 @item -fdump-rtl-csa
6179 @opindex fdump-rtl-csa
6180 Dump after combining stack adjustments.
6182 @item -fdump-rtl-cse1
6183 @itemx -fdump-rtl-cse2
6184 @opindex fdump-rtl-cse1
6185 @opindex fdump-rtl-cse2
6186 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
6187 the two common subexpression elimination passes.
6189 @item -fdump-rtl-dce
6190 @opindex fdump-rtl-dce
6191 Dump after the standalone dead code elimination passes.
6193 @item -fdump-rtl-dbr
6194 @opindex fdump-rtl-dbr
6195 Dump after delayed branch scheduling.
6197 @item -fdump-rtl-dce1
6198 @itemx -fdump-rtl-dce2
6199 @opindex fdump-rtl-dce1
6200 @opindex fdump-rtl-dce2
6201 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
6202 the two dead store elimination passes.
6205 @opindex fdump-rtl-eh
6206 Dump after finalization of EH handling code.
6208 @item -fdump-rtl-eh_ranges
6209 @opindex fdump-rtl-eh_ranges
6210 Dump after conversion of EH handling range regions.
6212 @item -fdump-rtl-expand
6213 @opindex fdump-rtl-expand
6214 Dump after RTL generation.
6216 @item -fdump-rtl-fwprop1
6217 @itemx -fdump-rtl-fwprop2
6218 @opindex fdump-rtl-fwprop1
6219 @opindex fdump-rtl-fwprop2
6220 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
6221 dumping after the two forward propagation passes.
6223 @item -fdump-rtl-gcse1
6224 @itemx -fdump-rtl-gcse2
6225 @opindex fdump-rtl-gcse1
6226 @opindex fdump-rtl-gcse2
6227 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
6228 after global common subexpression elimination.
6230 @item -fdump-rtl-init-regs
6231 @opindex fdump-rtl-init-regs
6232 Dump after the initialization of the registers.
6234 @item -fdump-rtl-initvals
6235 @opindex fdump-rtl-initvals
6236 Dump after the computation of the initial value sets.
6238 @item -fdump-rtl-into_cfglayout
6239 @opindex fdump-rtl-into_cfglayout
6240 Dump after converting to cfglayout mode.
6242 @item -fdump-rtl-ira
6243 @opindex fdump-rtl-ira
6244 Dump after iterated register allocation.
6246 @item -fdump-rtl-jump
6247 @opindex fdump-rtl-jump
6248 Dump after the second jump optimization.
6250 @item -fdump-rtl-loop2
6251 @opindex fdump-rtl-loop2
6252 @option{-fdump-rtl-loop2} enables dumping after the rtl
6253 loop optimization passes.
6255 @item -fdump-rtl-mach
6256 @opindex fdump-rtl-mach
6257 Dump after performing the machine dependent reorganization pass, if that
6260 @item -fdump-rtl-mode_sw
6261 @opindex fdump-rtl-mode_sw
6262 Dump after removing redundant mode switches.
6264 @item -fdump-rtl-rnreg
6265 @opindex fdump-rtl-rnreg
6266 Dump after register renumbering.
6268 @item -fdump-rtl-outof_cfglayout
6269 @opindex fdump-rtl-outof_cfglayout
6270 Dump after converting from cfglayout mode.
6272 @item -fdump-rtl-peephole2
6273 @opindex fdump-rtl-peephole2
6274 Dump after the peephole pass.
6276 @item -fdump-rtl-postreload
6277 @opindex fdump-rtl-postreload
6278 Dump after post-reload optimizations.
6280 @item -fdump-rtl-pro_and_epilogue
6281 @opindex fdump-rtl-pro_and_epilogue
6282 Dump after generating the function prologues and epilogues.
6284 @item -fdump-rtl-sched1
6285 @itemx -fdump-rtl-sched2
6286 @opindex fdump-rtl-sched1
6287 @opindex fdump-rtl-sched2
6288 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
6289 after the basic block scheduling passes.
6291 @item -fdump-rtl-ree
6292 @opindex fdump-rtl-ree
6293 Dump after sign/zero extension elimination.
6295 @item -fdump-rtl-seqabstr
6296 @opindex fdump-rtl-seqabstr
6297 Dump after common sequence discovery.
6299 @item -fdump-rtl-shorten
6300 @opindex fdump-rtl-shorten
6301 Dump after shortening branches.
6303 @item -fdump-rtl-sibling
6304 @opindex fdump-rtl-sibling
6305 Dump after sibling call optimizations.
6307 @item -fdump-rtl-split1
6308 @itemx -fdump-rtl-split2
6309 @itemx -fdump-rtl-split3
6310 @itemx -fdump-rtl-split4
6311 @itemx -fdump-rtl-split5
6312 @opindex fdump-rtl-split1
6313 @opindex fdump-rtl-split2
6314 @opindex fdump-rtl-split3
6315 @opindex fdump-rtl-split4
6316 @opindex fdump-rtl-split5
6317 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
6318 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
6319 @option{-fdump-rtl-split5} enable dumping after five rounds of
6320 instruction splitting.
6322 @item -fdump-rtl-sms
6323 @opindex fdump-rtl-sms
6324 Dump after modulo scheduling. This pass is only run on some
6327 @item -fdump-rtl-stack
6328 @opindex fdump-rtl-stack
6329 Dump after conversion from GCC's ``flat register file'' registers to the
6330 x87's stack-like registers. This pass is only run on x86 variants.
6332 @item -fdump-rtl-subreg1
6333 @itemx -fdump-rtl-subreg2
6334 @opindex fdump-rtl-subreg1
6335 @opindex fdump-rtl-subreg2
6336 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
6337 the two subreg expansion passes.
6339 @item -fdump-rtl-unshare
6340 @opindex fdump-rtl-unshare
6341 Dump after all rtl has been unshared.
6343 @item -fdump-rtl-vartrack
6344 @opindex fdump-rtl-vartrack
6345 Dump after variable tracking.
6347 @item -fdump-rtl-vregs
6348 @opindex fdump-rtl-vregs
6349 Dump after converting virtual registers to hard registers.
6351 @item -fdump-rtl-web
6352 @opindex fdump-rtl-web
6353 Dump after live range splitting.
6355 @item -fdump-rtl-regclass
6356 @itemx -fdump-rtl-subregs_of_mode_init
6357 @itemx -fdump-rtl-subregs_of_mode_finish
6358 @itemx -fdump-rtl-dfinit
6359 @itemx -fdump-rtl-dfinish
6360 @opindex fdump-rtl-regclass
6361 @opindex fdump-rtl-subregs_of_mode_init
6362 @opindex fdump-rtl-subregs_of_mode_finish
6363 @opindex fdump-rtl-dfinit
6364 @opindex fdump-rtl-dfinish
6365 These dumps are defined but always produce empty files.
6368 @itemx -fdump-rtl-all
6370 @opindex fdump-rtl-all
6371 Produce all the dumps listed above.
6375 Annotate the assembler output with miscellaneous debugging information.
6379 Dump all macro definitions, at the end of preprocessing, in addition to
6384 Produce a core dump whenever an error occurs.
6388 Annotate the assembler output with a comment indicating which
6389 pattern and alternative is used. The length of each instruction is
6394 Dump the RTL in the assembler output as a comment before each instruction.
6395 Also turns on @option{-dp} annotation.
6399 Just generate RTL for a function instead of compiling it. Usually used
6400 with @option{-fdump-rtl-expand}.
6404 @opindex fdump-noaddr
6405 When doing debugging dumps, suppress address output. This makes it more
6406 feasible to use diff on debugging dumps for compiler invocations with
6407 different compiler binaries and/or different
6408 text / bss / data / heap / stack / dso start locations.
6411 @opindex freport-bug
6412 Collect and dump debug information into temporary file if ICE in C/C++
6415 @item -fdump-unnumbered
6416 @opindex fdump-unnumbered
6417 When doing debugging dumps, suppress instruction numbers and address output.
6418 This makes it more feasible to use diff on debugging dumps for compiler
6419 invocations with different options, in particular with and without
6422 @item -fdump-unnumbered-links
6423 @opindex fdump-unnumbered-links
6424 When doing debugging dumps (see @option{-d} option above), suppress
6425 instruction numbers for the links to the previous and next instructions
6428 @item -fdump-translation-unit @r{(C++ only)}
6429 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
6430 @opindex fdump-translation-unit
6431 Dump a representation of the tree structure for the entire translation
6432 unit to a file. The file name is made by appending @file{.tu} to the
6433 source file name, and the file is created in the same directory as the
6434 output file. If the @samp{-@var{options}} form is used, @var{options}
6435 controls the details of the dump as described for the
6436 @option{-fdump-tree} options.
6438 @item -fdump-class-hierarchy @r{(C++ only)}
6439 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
6440 @opindex fdump-class-hierarchy
6441 Dump a representation of each class's hierarchy and virtual function
6442 table layout to a file. The file name is made by appending
6443 @file{.class} to the source file name, and the file is created in the
6444 same directory as the output file. If the @samp{-@var{options}} form
6445 is used, @var{options} controls the details of the dump as described
6446 for the @option{-fdump-tree} options.
6448 @item -fdump-ipa-@var{switch}
6450 Control the dumping at various stages of inter-procedural analysis
6451 language tree to a file. The file name is generated by appending a
6452 switch specific suffix to the source file name, and the file is created
6453 in the same directory as the output file. The following dumps are
6458 Enables all inter-procedural analysis dumps.
6461 Dumps information about call-graph optimization, unused function removal,
6462 and inlining decisions.
6465 Dump after function inlining.
6470 @opindex fdump-passes
6471 Dump the list of optimization passes that are turned on and off by
6472 the current command-line options.
6474 @item -fdump-statistics-@var{option}
6475 @opindex fdump-statistics
6476 Enable and control dumping of pass statistics in a separate file. The
6477 file name is generated by appending a suffix ending in
6478 @samp{.statistics} to the source file name, and the file is created in
6479 the same directory as the output file. If the @samp{-@var{option}}
6480 form is used, @samp{-stats} causes counters to be summed over the
6481 whole compilation unit while @samp{-details} dumps every event as
6482 the passes generate them. The default with no option is to sum
6483 counters for each function compiled.
6485 @item -fdump-tree-@var{switch}
6486 @itemx -fdump-tree-@var{switch}-@var{options}
6487 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
6489 Control the dumping at various stages of processing the intermediate
6490 language tree to a file. The file name is generated by appending a
6491 switch-specific suffix to the source file name, and the file is
6492 created in the same directory as the output file. In case of
6493 @option{=@var{filename}} option, the dump is output on the given file
6494 instead of the auto named dump files. If the @samp{-@var{options}}
6495 form is used, @var{options} is a list of @samp{-} separated options
6496 which control the details of the dump. Not all options are applicable
6497 to all dumps; those that are not meaningful are ignored. The
6498 following options are available
6502 Print the address of each node. Usually this is not meaningful as it
6503 changes according to the environment and source file. Its primary use
6504 is for tying up a dump file with a debug environment.
6506 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
6507 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
6508 use working backward from mangled names in the assembly file.
6510 When dumping front-end intermediate representations, inhibit dumping
6511 of members of a scope or body of a function merely because that scope
6512 has been reached. Only dump such items when they are directly reachable
6515 When dumping pretty-printed trees, this option inhibits dumping the
6516 bodies of control structures.
6518 When dumping RTL, print the RTL in slim (condensed) form instead of
6519 the default LISP-like representation.
6521 Print a raw representation of the tree. By default, trees are
6522 pretty-printed into a C-like representation.
6524 Enable more detailed dumps (not honored by every dump option). Also
6525 include information from the optimization passes.
6527 Enable dumping various statistics about the pass (not honored by every dump
6530 Enable showing basic block boundaries (disabled in raw dumps).
6532 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
6533 dump a representation of the control flow graph suitable for viewing with
6534 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
6535 the file is pretty-printed as a subgraph, so that GraphViz can render them
6536 all in a single plot.
6538 This option currently only works for RTL dumps, and the RTL is always
6539 dumped in slim form.
6541 Enable showing virtual operands for every statement.
6543 Enable showing line numbers for statements.
6545 Enable showing the unique ID (@code{DECL_UID}) for each variable.
6547 Enable showing the tree dump for each statement.
6549 Enable showing the EH region number holding each statement.
6551 Enable showing scalar evolution analysis details.
6553 Enable showing optimization information (only available in certain
6556 Enable showing missed optimization information (only available in certain
6559 Enable other detailed optimization information (only available in
6561 @item =@var{filename}
6562 Instead of an auto named dump file, output into the given file
6563 name. The file names @file{stdout} and @file{stderr} are treated
6564 specially and are considered already open standard streams. For
6568 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
6569 -fdump-tree-pre=stderr file.c
6572 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
6573 output on to @file{stderr}. If two conflicting dump filenames are
6574 given for the same pass, then the latter option overrides the earlier
6578 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
6579 and @option{lineno}.
6582 Turn on all optimization options, i.e., @option{optimized},
6583 @option{missed}, and @option{note}.
6586 The following tree dumps are possible:
6590 @opindex fdump-tree-original
6591 Dump before any tree based optimization, to @file{@var{file}.original}.
6594 @opindex fdump-tree-optimized
6595 Dump after all tree based optimization, to @file{@var{file}.optimized}.
6598 @opindex fdump-tree-gimple
6599 Dump each function before and after the gimplification pass to a file. The
6600 file name is made by appending @file{.gimple} to the source file name.
6603 @opindex fdump-tree-cfg
6604 Dump the control flow graph of each function to a file. The file name is
6605 made by appending @file{.cfg} to the source file name.
6608 @opindex fdump-tree-ch
6609 Dump each function after copying loop headers. The file name is made by
6610 appending @file{.ch} to the source file name.
6613 @opindex fdump-tree-ssa
6614 Dump SSA related information to a file. The file name is made by appending
6615 @file{.ssa} to the source file name.
6618 @opindex fdump-tree-alias
6619 Dump aliasing information for each function. The file name is made by
6620 appending @file{.alias} to the source file name.
6623 @opindex fdump-tree-ccp
6624 Dump each function after CCP@. The file name is made by appending
6625 @file{.ccp} to the source file name.
6628 @opindex fdump-tree-storeccp
6629 Dump each function after STORE-CCP@. The file name is made by appending
6630 @file{.storeccp} to the source file name.
6633 @opindex fdump-tree-pre
6634 Dump trees after partial redundancy elimination. The file name is made
6635 by appending @file{.pre} to the source file name.
6638 @opindex fdump-tree-fre
6639 Dump trees after full redundancy elimination. The file name is made
6640 by appending @file{.fre} to the source file name.
6643 @opindex fdump-tree-copyprop
6644 Dump trees after copy propagation. The file name is made
6645 by appending @file{.copyprop} to the source file name.
6647 @item store_copyprop
6648 @opindex fdump-tree-store_copyprop
6649 Dump trees after store copy-propagation. The file name is made
6650 by appending @file{.store_copyprop} to the source file name.
6653 @opindex fdump-tree-dce
6654 Dump each function after dead code elimination. The file name is made by
6655 appending @file{.dce} to the source file name.
6658 @opindex fdump-tree-sra
6659 Dump each function after performing scalar replacement of aggregates. The
6660 file name is made by appending @file{.sra} to the source file name.
6663 @opindex fdump-tree-sink
6664 Dump each function after performing code sinking. The file name is made
6665 by appending @file{.sink} to the source file name.
6668 @opindex fdump-tree-dom
6669 Dump each function after applying dominator tree optimizations. The file
6670 name is made by appending @file{.dom} to the source file name.
6673 @opindex fdump-tree-dse
6674 Dump each function after applying dead store elimination. The file
6675 name is made by appending @file{.dse} to the source file name.
6678 @opindex fdump-tree-phiopt
6679 Dump each function after optimizing PHI nodes into straightline code. The file
6680 name is made by appending @file{.phiopt} to the source file name.
6683 @opindex fdump-tree-forwprop
6684 Dump each function after forward propagating single use variables. The file
6685 name is made by appending @file{.forwprop} to the source file name.
6688 @opindex fdump-tree-copyrename
6689 Dump each function after applying the copy rename optimization. The file
6690 name is made by appending @file{.copyrename} to the source file name.
6693 @opindex fdump-tree-nrv
6694 Dump each function after applying the named return value optimization on
6695 generic trees. The file name is made by appending @file{.nrv} to the source
6699 @opindex fdump-tree-vect
6700 Dump each function after applying vectorization of loops. The file name is
6701 made by appending @file{.vect} to the source file name.
6704 @opindex fdump-tree-slp
6705 Dump each function after applying vectorization of basic blocks. The file name
6706 is made by appending @file{.slp} to the source file name.
6709 @opindex fdump-tree-vrp
6710 Dump each function after Value Range Propagation (VRP). The file name
6711 is made by appending @file{.vrp} to the source file name.
6714 @opindex fdump-tree-all
6715 Enable all the available tree dumps with the flags provided in this option.
6719 @itemx -fopt-info-@var{options}
6720 @itemx -fopt-info-@var{options}=@var{filename}
6722 Controls optimization dumps from various optimization passes. If the
6723 @samp{-@var{options}} form is used, @var{options} is a list of
6724 @samp{-} separated options to select the dump details and
6725 optimizations. If @var{options} is not specified, it defaults to
6726 @option{optimized} for details and @option{optall} for optimization
6727 groups. If the @var{filename} is not specified, it defaults to
6728 @file{stderr}. Note that the output @var{filename} will be overwritten
6729 in case of multiple translation units. If a combined output from
6730 multiple translation units is desired, @file{stderr} should be used
6733 The options can be divided into two groups, 1) options describing the
6734 verbosity of the dump, and 2) options describing which optimizations
6735 should be included. The options from both the groups can be freely
6736 mixed as they are non-overlapping. However, in case of any conflicts,
6737 the latter options override the earlier options on the command
6738 line. Though multiple -fopt-info options are accepted, only one of
6739 them can have @option{=filename}. If other filenames are provided then
6740 all but the first one are ignored.
6742 The dump verbosity has the following options
6746 Print information when an optimization is successfully applied. It is
6747 up to a pass to decide which information is relevant. For example, the
6748 vectorizer passes print the source location of loops which got
6749 successfully vectorized.
6751 Print information about missed optimizations. Individual passes
6752 control which information to include in the output. For example,
6755 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
6758 will print information about missed optimization opportunities from
6759 vectorization passes on stderr.
6761 Print verbose information about optimizations, such as certain
6762 transformations, more detailed messages about decisions etc.
6764 Print detailed optimization information. This includes
6765 @var{optimized}, @var{missed}, and @var{note}.
6768 The second set of options describes a group of optimizations and may
6769 include one or more of the following.
6773 Enable dumps from all interprocedural optimizations.
6775 Enable dumps from all loop optimizations.
6777 Enable dumps from all inlining optimizations.
6779 Enable dumps from all vectorization optimizations.
6781 Enable dumps from all optimizations. This is a superset of
6782 the optimization groups listed above.
6787 gcc -O3 -fopt-info-missed=missed.all
6790 outputs missed optimization report from all the passes into
6795 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
6798 will output information about missed optimizations as well as
6799 optimized locations from all the inlining passes into
6802 If the @var{filename} is provided, then the dumps from all the
6803 applicable optimizations are concatenated into the @file{filename}.
6804 Otherwise the dump is output onto @file{stderr}. If @var{options} is
6805 omitted, it defaults to @option{all-optall}, which means dump all
6806 available optimization info from all the passes. In the following
6807 example, all optimization info is output on to @file{stderr}.
6813 Note that @option{-fopt-info-vec-missed} behaves the same as
6814 @option{-fopt-info-missed-vec}.
6816 As another example, consider
6819 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
6822 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
6823 in conflict since only one output file is allowed. In this case, only
6824 the first option takes effect and the subsequent options are
6825 ignored. Thus only the @file{vec.miss} is produced which contains
6826 dumps from the vectorizer about missed opportunities.
6828 @item -frandom-seed=@var{number}
6829 @opindex frandom-seed
6830 This option provides a seed that GCC uses in place of
6831 random numbers in generating certain symbol names
6832 that have to be different in every compiled file. It is also used to
6833 place unique stamps in coverage data files and the object files that
6834 produce them. You can use the @option{-frandom-seed} option to produce
6835 reproducibly identical object files.
6837 The @var{number} should be different for every file you compile.
6839 @item -fsched-verbose=@var{n}
6840 @opindex fsched-verbose
6841 On targets that use instruction scheduling, this option controls the
6842 amount of debugging output the scheduler prints. This information is
6843 written to standard error, unless @option{-fdump-rtl-sched1} or
6844 @option{-fdump-rtl-sched2} is specified, in which case it is output
6845 to the usual dump listing file, @file{.sched1} or @file{.sched2}
6846 respectively. However for @var{n} greater than nine, the output is
6847 always printed to standard error.
6849 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
6850 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
6851 For @var{n} greater than one, it also output basic block probabilities,
6852 detailed ready list information and unit/insn info. For @var{n} greater
6853 than two, it includes RTL at abort point, control-flow and regions info.
6854 And for @var{n} over four, @option{-fsched-verbose} also includes
6858 @itemx -save-temps=cwd
6860 Store the usual ``temporary'' intermediate files permanently; place them
6861 in the current directory and name them based on the source file. Thus,
6862 compiling @file{foo.c} with @option{-c -save-temps} produces files
6863 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
6864 preprocessed @file{foo.i} output file even though the compiler now
6865 normally uses an integrated preprocessor.
6867 When used in combination with the @option{-x} command-line option,
6868 @option{-save-temps} is sensible enough to avoid over writing an
6869 input source file with the same extension as an intermediate file.
6870 The corresponding intermediate file may be obtained by renaming the
6871 source file before using @option{-save-temps}.
6873 If you invoke GCC in parallel, compiling several different source
6874 files that share a common base name in different subdirectories or the
6875 same source file compiled for multiple output destinations, it is
6876 likely that the different parallel compilers will interfere with each
6877 other, and overwrite the temporary files. For instance:
6880 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
6881 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
6884 may result in @file{foo.i} and @file{foo.o} being written to
6885 simultaneously by both compilers.
6887 @item -save-temps=obj
6888 @opindex save-temps=obj
6889 Store the usual ``temporary'' intermediate files permanently. If the
6890 @option{-o} option is used, the temporary files are based on the
6891 object file. If the @option{-o} option is not used, the
6892 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
6897 gcc -save-temps=obj -c foo.c
6898 gcc -save-temps=obj -c bar.c -o dir/xbar.o
6899 gcc -save-temps=obj foobar.c -o dir2/yfoobar
6903 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
6904 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
6905 @file{dir2/yfoobar.o}.
6907 @item -time@r{[}=@var{file}@r{]}
6909 Report the CPU time taken by each subprocess in the compilation
6910 sequence. For C source files, this is the compiler proper and assembler
6911 (plus the linker if linking is done).
6913 Without the specification of an output file, the output looks like this:
6920 The first number on each line is the ``user time'', that is time spent
6921 executing the program itself. The second number is ``system time'',
6922 time spent executing operating system routines on behalf of the program.
6923 Both numbers are in seconds.
6925 With the specification of an output file, the output is appended to the
6926 named file, and it looks like this:
6929 0.12 0.01 cc1 @var{options}
6930 0.00 0.01 as @var{options}
6933 The ``user time'' and the ``system time'' are moved before the program
6934 name, and the options passed to the program are displayed, so that one
6935 can later tell what file was being compiled, and with which options.
6937 @item -fvar-tracking
6938 @opindex fvar-tracking
6939 Run variable tracking pass. It computes where variables are stored at each
6940 position in code. Better debugging information is then generated
6941 (if the debugging information format supports this information).
6943 It is enabled by default when compiling with optimization (@option{-Os},
6944 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
6945 the debug info format supports it.
6947 @item -fvar-tracking-assignments
6948 @opindex fvar-tracking-assignments
6949 @opindex fno-var-tracking-assignments
6950 Annotate assignments to user variables early in the compilation and
6951 attempt to carry the annotations over throughout the compilation all the
6952 way to the end, in an attempt to improve debug information while
6953 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
6955 It can be enabled even if var-tracking is disabled, in which case
6956 annotations are created and maintained, but discarded at the end.
6958 @item -fvar-tracking-assignments-toggle
6959 @opindex fvar-tracking-assignments-toggle
6960 @opindex fno-var-tracking-assignments-toggle
6961 Toggle @option{-fvar-tracking-assignments}, in the same way that
6962 @option{-gtoggle} toggles @option{-g}.
6964 @item -print-file-name=@var{library}
6965 @opindex print-file-name
6966 Print the full absolute name of the library file @var{library} that
6967 would be used when linking---and don't do anything else. With this
6968 option, GCC does not compile or link anything; it just prints the
6971 @item -print-multi-directory
6972 @opindex print-multi-directory
6973 Print the directory name corresponding to the multilib selected by any
6974 other switches present in the command line. This directory is supposed
6975 to exist in @env{GCC_EXEC_PREFIX}.
6977 @item -print-multi-lib
6978 @opindex print-multi-lib
6979 Print the mapping from multilib directory names to compiler switches
6980 that enable them. The directory name is separated from the switches by
6981 @samp{;}, and each switch starts with an @samp{@@} instead of the
6982 @samp{-}, without spaces between multiple switches. This is supposed to
6983 ease shell processing.
6985 @item -print-multi-os-directory
6986 @opindex print-multi-os-directory
6987 Print the path to OS libraries for the selected
6988 multilib, relative to some @file{lib} subdirectory. If OS libraries are
6989 present in the @file{lib} subdirectory and no multilibs are used, this is
6990 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
6991 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
6992 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
6993 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
6995 @item -print-multiarch
6996 @opindex print-multiarch
6997 Print the path to OS libraries for the selected multiarch,
6998 relative to some @file{lib} subdirectory.
7000 @item -print-prog-name=@var{program}
7001 @opindex print-prog-name
7002 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
7004 @item -print-libgcc-file-name
7005 @opindex print-libgcc-file-name
7006 Same as @option{-print-file-name=libgcc.a}.
7008 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
7009 but you do want to link with @file{libgcc.a}. You can do:
7012 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
7015 @item -print-search-dirs
7016 @opindex print-search-dirs
7017 Print the name of the configured installation directory and a list of
7018 program and library directories @command{gcc} searches---and don't do anything else.
7020 This is useful when @command{gcc} prints the error message
7021 @samp{installation problem, cannot exec cpp0: No such file or directory}.
7022 To resolve this you either need to put @file{cpp0} and the other compiler
7023 components where @command{gcc} expects to find them, or you can set the environment
7024 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
7025 Don't forget the trailing @samp{/}.
7026 @xref{Environment Variables}.
7028 @item -print-sysroot
7029 @opindex print-sysroot
7030 Print the target sysroot directory that is used during
7031 compilation. This is the target sysroot specified either at configure
7032 time or using the @option{--sysroot} option, possibly with an extra
7033 suffix that depends on compilation options. If no target sysroot is
7034 specified, the option prints nothing.
7036 @item -print-sysroot-headers-suffix
7037 @opindex print-sysroot-headers-suffix
7038 Print the suffix added to the target sysroot when searching for
7039 headers, or give an error if the compiler is not configured with such
7040 a suffix---and don't do anything else.
7043 @opindex dumpmachine
7044 Print the compiler's target machine (for example,
7045 @samp{i686-pc-linux-gnu})---and don't do anything else.
7048 @opindex dumpversion
7049 Print the compiler version (for example, @samp{3.0})---and don't do
7054 Print the compiler's built-in specs---and don't do anything else. (This
7055 is used when GCC itself is being built.) @xref{Spec Files}.
7057 @item -fno-eliminate-unused-debug-types
7058 @opindex feliminate-unused-debug-types
7059 @opindex fno-eliminate-unused-debug-types
7060 Normally, when producing DWARF 2 output, GCC avoids producing debug symbol
7061 output for types that are nowhere used in the source file being compiled.
7062 Sometimes it is useful to have GCC emit debugging
7063 information for all types declared in a compilation
7064 unit, regardless of whether or not they are actually used
7065 in that compilation unit, for example
7066 if, in the debugger, you want to cast a value to a type that is
7067 not actually used in your program (but is declared). More often,
7068 however, this results in a significant amount of wasted space.
7071 @node Optimize Options
7072 @section Options That Control Optimization
7073 @cindex optimize options
7074 @cindex options, optimization
7076 These options control various sorts of optimizations.
7078 Without any optimization option, the compiler's goal is to reduce the
7079 cost of compilation and to make debugging produce the expected
7080 results. Statements are independent: if you stop the program with a
7081 breakpoint between statements, you can then assign a new value to any
7082 variable or change the program counter to any other statement in the
7083 function and get exactly the results you expect from the source
7086 Turning on optimization flags makes the compiler attempt to improve
7087 the performance and/or code size at the expense of compilation time
7088 and possibly the ability to debug the program.
7090 The compiler performs optimization based on the knowledge it has of the
7091 program. Compiling multiple files at once to a single output file mode allows
7092 the compiler to use information gained from all of the files when compiling
7095 Not all optimizations are controlled directly by a flag. Only
7096 optimizations that have a flag are listed in this section.
7098 Most optimizations are only enabled if an @option{-O} level is set on
7099 the command line. Otherwise they are disabled, even if individual
7100 optimization flags are specified.
7102 Depending on the target and how GCC was configured, a slightly different
7103 set of optimizations may be enabled at each @option{-O} level than
7104 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
7105 to find out the exact set of optimizations that are enabled at each level.
7106 @xref{Overall Options}, for examples.
7113 Optimize. Optimizing compilation takes somewhat more time, and a lot
7114 more memory for a large function.
7116 With @option{-O}, the compiler tries to reduce code size and execution
7117 time, without performing any optimizations that take a great deal of
7120 @option{-O} turns on the following optimization flags:
7123 -fbranch-count-reg @gol
7124 -fcombine-stack-adjustments @gol
7126 -fcprop-registers @gol
7129 -fdelayed-branch @gol
7131 -fforward-propagate @gol
7132 -fguess-branch-probability @gol
7133 -fif-conversion2 @gol
7134 -fif-conversion @gol
7135 -finline-functions-called-once @gol
7136 -fipa-pure-const @gol
7138 -fipa-reference @gol
7139 -fmerge-constants @gol
7140 -fmove-loop-invariants @gol
7142 -fsplit-wide-types @gol
7147 -ftree-copy-prop @gol
7148 -ftree-copyrename @gol
7150 -ftree-dominator-opts @gol
7152 -ftree-forwprop @gol
7162 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
7163 where doing so does not interfere with debugging.
7167 Optimize even more. GCC performs nearly all supported optimizations
7168 that do not involve a space-speed tradeoff.
7169 As compared to @option{-O}, this option increases both compilation time
7170 and the performance of the generated code.
7172 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7173 also turns on the following optimization flags:
7174 @gccoptlist{-fthread-jumps @gol
7175 -falign-functions -falign-jumps @gol
7176 -falign-loops -falign-labels @gol
7179 -fcse-follow-jumps -fcse-skip-blocks @gol
7180 -fdelete-null-pointer-checks @gol
7181 -fdevirtualize -fdevirtualize-speculatively @gol
7182 -fexpensive-optimizations @gol
7183 -fgcse -fgcse-lm @gol
7184 -fhoist-adjacent-loads @gol
7185 -finline-small-functions @gol
7186 -findirect-inlining @gol
7190 -fisolate-erroneous-paths-dereference @gol
7192 -foptimize-sibling-calls @gol
7193 -foptimize-strlen @gol
7194 -fpartial-inlining @gol
7196 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
7197 -frerun-cse-after-loop @gol
7198 -fsched-interblock -fsched-spec @gol
7199 -fschedule-insns -fschedule-insns2 @gol
7200 -fstrict-aliasing -fstrict-overflow @gol
7201 -ftree-builtin-call-dce @gol
7202 -ftree-switch-conversion -ftree-tail-merge @gol
7207 Please note the warning under @option{-fgcse} about
7208 invoking @option{-O2} on programs that use computed gotos.
7212 Optimize yet more. @option{-O3} turns on all optimizations specified
7213 by @option{-O2} and also turns on the @option{-finline-functions},
7214 @option{-funswitch-loops}, @option{-fpredictive-commoning},
7215 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
7216 @option{-ftree-loop-distribute-patterns},
7217 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
7218 @option{-ftree-partial-pre} and @option{-fipa-cp-clone} options.
7222 Reduce compilation time and make debugging produce the expected
7223 results. This is the default.
7227 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7228 do not typically increase code size. It also performs further
7229 optimizations designed to reduce code size.
7231 @option{-Os} disables the following optimization flags:
7232 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7233 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
7234 -fprefetch-loop-arrays}
7238 Disregard strict standards compliance. @option{-Ofast} enables all
7239 @option{-O3} optimizations. It also enables optimizations that are not
7240 valid for all standard-compliant programs.
7241 It turns on @option{-ffast-math} and the Fortran-specific
7242 @option{-fno-protect-parens} and @option{-fstack-arrays}.
7246 Optimize debugging experience. @option{-Og} enables optimizations
7247 that do not interfere with debugging. It should be the optimization
7248 level of choice for the standard edit-compile-debug cycle, offering
7249 a reasonable level of optimization while maintaining fast compilation
7250 and a good debugging experience.
7252 If you use multiple @option{-O} options, with or without level numbers,
7253 the last such option is the one that is effective.
7256 Options of the form @option{-f@var{flag}} specify machine-independent
7257 flags. Most flags have both positive and negative forms; the negative
7258 form of @option{-ffoo} is @option{-fno-foo}. In the table
7259 below, only one of the forms is listed---the one you typically
7260 use. You can figure out the other form by either removing @samp{no-}
7263 The following options control specific optimizations. They are either
7264 activated by @option{-O} options or are related to ones that are. You
7265 can use the following flags in the rare cases when ``fine-tuning'' of
7266 optimizations to be performed is desired.
7269 @item -fno-defer-pop
7270 @opindex fno-defer-pop
7271 Always pop the arguments to each function call as soon as that function
7272 returns. For machines that must pop arguments after a function call,
7273 the compiler normally lets arguments accumulate on the stack for several
7274 function calls and pops them all at once.
7276 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7278 @item -fforward-propagate
7279 @opindex fforward-propagate
7280 Perform a forward propagation pass on RTL@. The pass tries to combine two
7281 instructions and checks if the result can be simplified. If loop unrolling
7282 is active, two passes are performed and the second is scheduled after
7285 This option is enabled by default at optimization levels @option{-O},
7286 @option{-O2}, @option{-O3}, @option{-Os}.
7288 @item -ffp-contract=@var{style}
7289 @opindex ffp-contract
7290 @option{-ffp-contract=off} disables floating-point expression contraction.
7291 @option{-ffp-contract=fast} enables floating-point expression contraction
7292 such as forming of fused multiply-add operations if the target has
7293 native support for them.
7294 @option{-ffp-contract=on} enables floating-point expression contraction
7295 if allowed by the language standard. This is currently not implemented
7296 and treated equal to @option{-ffp-contract=off}.
7298 The default is @option{-ffp-contract=fast}.
7300 @item -fomit-frame-pointer
7301 @opindex fomit-frame-pointer
7302 Don't keep the frame pointer in a register for functions that
7303 don't need one. This avoids the instructions to save, set up and
7304 restore frame pointers; it also makes an extra register available
7305 in many functions. @strong{It also makes debugging impossible on
7308 On some machines, such as the VAX, this flag has no effect, because
7309 the standard calling sequence automatically handles the frame pointer
7310 and nothing is saved by pretending it doesn't exist. The
7311 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
7312 whether a target machine supports this flag. @xref{Registers,,Register
7313 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
7315 Starting with GCC version 4.6, the default setting (when not optimizing for
7316 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets has been changed to
7317 @option{-fomit-frame-pointer}. The default can be reverted to
7318 @option{-fno-omit-frame-pointer} by configuring GCC with the
7319 @option{--enable-frame-pointer} configure option.
7321 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7323 @item -foptimize-sibling-calls
7324 @opindex foptimize-sibling-calls
7325 Optimize sibling and tail recursive calls.
7327 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7329 @item -foptimize-strlen
7330 @opindex foptimize-strlen
7331 Optimize various standard C string functions (e.g. @code{strlen},
7332 @code{strchr} or @code{strcpy}) and
7333 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7335 Enabled at levels @option{-O2}, @option{-O3}.
7339 Do not expand any functions inline apart from those marked with
7340 the @code{always_inline} attribute. This is the default when not
7343 Single functions can be exempted from inlining by marking them
7344 with the @code{noinline} attribute.
7346 @item -finline-small-functions
7347 @opindex finline-small-functions
7348 Integrate functions into their callers when their body is smaller than expected
7349 function call code (so overall size of program gets smaller). The compiler
7350 heuristically decides which functions are simple enough to be worth integrating
7351 in this way. This inlining applies to all functions, even those not declared
7354 Enabled at level @option{-O2}.
7356 @item -findirect-inlining
7357 @opindex findirect-inlining
7358 Inline also indirect calls that are discovered to be known at compile
7359 time thanks to previous inlining. This option has any effect only
7360 when inlining itself is turned on by the @option{-finline-functions}
7361 or @option{-finline-small-functions} options.
7363 Enabled at level @option{-O2}.
7365 @item -finline-functions
7366 @opindex finline-functions
7367 Consider all functions for inlining, even if they are not declared inline.
7368 The compiler heuristically decides which functions are worth integrating
7371 If all calls to a given function are integrated, and the function is
7372 declared @code{static}, then the function is normally not output as
7373 assembler code in its own right.
7375 Enabled at level @option{-O3}.
7377 @item -finline-functions-called-once
7378 @opindex finline-functions-called-once
7379 Consider all @code{static} functions called once for inlining into their
7380 caller even if they are not marked @code{inline}. If a call to a given
7381 function is integrated, then the function is not output as assembler code
7384 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7386 @item -fearly-inlining
7387 @opindex fearly-inlining
7388 Inline functions marked by @code{always_inline} and functions whose body seems
7389 smaller than the function call overhead early before doing
7390 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7391 makes profiling significantly cheaper and usually inlining faster on programs
7392 having large chains of nested wrapper functions.
7398 Perform interprocedural scalar replacement of aggregates, removal of
7399 unused parameters and replacement of parameters passed by reference
7400 by parameters passed by value.
7402 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7404 @item -finline-limit=@var{n}
7405 @opindex finline-limit
7406 By default, GCC limits the size of functions that can be inlined. This flag
7407 allows coarse control of this limit. @var{n} is the size of functions that
7408 can be inlined in number of pseudo instructions.
7410 Inlining is actually controlled by a number of parameters, which may be
7411 specified individually by using @option{--param @var{name}=@var{value}}.
7412 The @option{-finline-limit=@var{n}} option sets some of these parameters
7416 @item max-inline-insns-single
7417 is set to @var{n}/2.
7418 @item max-inline-insns-auto
7419 is set to @var{n}/2.
7422 See below for a documentation of the individual
7423 parameters controlling inlining and for the defaults of these parameters.
7425 @emph{Note:} there may be no value to @option{-finline-limit} that results
7426 in default behavior.
7428 @emph{Note:} pseudo instruction represents, in this particular context, an
7429 abstract measurement of function's size. In no way does it represent a count
7430 of assembly instructions and as such its exact meaning might change from one
7431 release to an another.
7433 @item -fno-keep-inline-dllexport
7434 @opindex -fno-keep-inline-dllexport
7435 This is a more fine-grained version of @option{-fkeep-inline-functions},
7436 which applies only to functions that are declared using the @code{dllexport}
7437 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
7440 @item -fkeep-inline-functions
7441 @opindex fkeep-inline-functions
7442 In C, emit @code{static} functions that are declared @code{inline}
7443 into the object file, even if the function has been inlined into all
7444 of its callers. This switch does not affect functions using the
7445 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7446 inline functions into the object file.
7448 @item -fkeep-static-consts
7449 @opindex fkeep-static-consts
7450 Emit variables declared @code{static const} when optimization isn't turned
7451 on, even if the variables aren't referenced.
7453 GCC enables this option by default. If you want to force the compiler to
7454 check if a variable is referenced, regardless of whether or not
7455 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7457 @item -fmerge-constants
7458 @opindex fmerge-constants
7459 Attempt to merge identical constants (string constants and floating-point
7460 constants) across compilation units.
7462 This option is the default for optimized compilation if the assembler and
7463 linker support it. Use @option{-fno-merge-constants} to inhibit this
7466 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7468 @item -fmerge-all-constants
7469 @opindex fmerge-all-constants
7470 Attempt to merge identical constants and identical variables.
7472 This option implies @option{-fmerge-constants}. In addition to
7473 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7474 arrays or initialized constant variables with integral or floating-point
7475 types. Languages like C or C++ require each variable, including multiple
7476 instances of the same variable in recursive calls, to have distinct locations,
7477 so using this option results in non-conforming
7480 @item -fmodulo-sched
7481 @opindex fmodulo-sched
7482 Perform swing modulo scheduling immediately before the first scheduling
7483 pass. This pass looks at innermost loops and reorders their
7484 instructions by overlapping different iterations.
7486 @item -fmodulo-sched-allow-regmoves
7487 @opindex fmodulo-sched-allow-regmoves
7488 Perform more aggressive SMS-based modulo scheduling with register moves
7489 allowed. By setting this flag certain anti-dependences edges are
7490 deleted, which triggers the generation of reg-moves based on the
7491 life-range analysis. This option is effective only with
7492 @option{-fmodulo-sched} enabled.
7494 @item -fno-branch-count-reg
7495 @opindex fno-branch-count-reg
7496 Do not use ``decrement and branch'' instructions on a count register,
7497 but instead generate a sequence of instructions that decrement a
7498 register, compare it against zero, then branch based upon the result.
7499 This option is only meaningful on architectures that support such
7500 instructions, which include x86, PowerPC, IA-64 and S/390.
7502 Enabled by default at @option{-O1} and higher.
7504 The default is @option{-fbranch-count-reg}.
7506 @item -fno-function-cse
7507 @opindex fno-function-cse
7508 Do not put function addresses in registers; make each instruction that
7509 calls a constant function contain the function's address explicitly.
7511 This option results in less efficient code, but some strange hacks
7512 that alter the assembler output may be confused by the optimizations
7513 performed when this option is not used.
7515 The default is @option{-ffunction-cse}
7517 @item -fno-zero-initialized-in-bss
7518 @opindex fno-zero-initialized-in-bss
7519 If the target supports a BSS section, GCC by default puts variables that
7520 are initialized to zero into BSS@. This can save space in the resulting
7523 This option turns off this behavior because some programs explicitly
7524 rely on variables going to the data section---e.g., so that the
7525 resulting executable can find the beginning of that section and/or make
7526 assumptions based on that.
7528 The default is @option{-fzero-initialized-in-bss}.
7530 @item -fthread-jumps
7531 @opindex fthread-jumps
7532 Perform optimizations that check to see if a jump branches to a
7533 location where another comparison subsumed by the first is found. If
7534 so, the first branch is redirected to either the destination of the
7535 second branch or a point immediately following it, depending on whether
7536 the condition is known to be true or false.
7538 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7540 @item -fsplit-wide-types
7541 @opindex fsplit-wide-types
7542 When using a type that occupies multiple registers, such as @code{long
7543 long} on a 32-bit system, split the registers apart and allocate them
7544 independently. This normally generates better code for those types,
7545 but may make debugging more difficult.
7547 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7550 @item -fcse-follow-jumps
7551 @opindex fcse-follow-jumps
7552 In common subexpression elimination (CSE), scan through jump instructions
7553 when the target of the jump is not reached by any other path. For
7554 example, when CSE encounters an @code{if} statement with an
7555 @code{else} clause, CSE follows the jump when the condition
7558 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7560 @item -fcse-skip-blocks
7561 @opindex fcse-skip-blocks
7562 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
7563 follow jumps that conditionally skip over blocks. When CSE
7564 encounters a simple @code{if} statement with no else clause,
7565 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
7566 body of the @code{if}.
7568 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7570 @item -frerun-cse-after-loop
7571 @opindex frerun-cse-after-loop
7572 Re-run common subexpression elimination after loop optimizations are
7575 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7579 Perform a global common subexpression elimination pass.
7580 This pass also performs global constant and copy propagation.
7582 @emph{Note:} When compiling a program using computed gotos, a GCC
7583 extension, you may get better run-time performance if you disable
7584 the global common subexpression elimination pass by adding
7585 @option{-fno-gcse} to the command line.
7587 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7591 When @option{-fgcse-lm} is enabled, global common subexpression elimination
7592 attempts to move loads that are only killed by stores into themselves. This
7593 allows a loop containing a load/store sequence to be changed to a load outside
7594 the loop, and a copy/store within the loop.
7596 Enabled by default when @option{-fgcse} is enabled.
7600 When @option{-fgcse-sm} is enabled, a store motion pass is run after
7601 global common subexpression elimination. This pass attempts to move
7602 stores out of loops. When used in conjunction with @option{-fgcse-lm},
7603 loops containing a load/store sequence can be changed to a load before
7604 the loop and a store after the loop.
7606 Not enabled at any optimization level.
7610 When @option{-fgcse-las} is enabled, the global common subexpression
7611 elimination pass eliminates redundant loads that come after stores to the
7612 same memory location (both partial and full redundancies).
7614 Not enabled at any optimization level.
7616 @item -fgcse-after-reload
7617 @opindex fgcse-after-reload
7618 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
7619 pass is performed after reload. The purpose of this pass is to clean up
7622 @item -faggressive-loop-optimizations
7623 @opindex faggressive-loop-optimizations
7624 This option tells the loop optimizer to use language constraints to
7625 derive bounds for the number of iterations of a loop. This assumes that
7626 loop code does not invoke undefined behavior by for example causing signed
7627 integer overflows or out-of-bound array accesses. The bounds for the
7628 number of iterations of a loop are used to guide loop unrolling and peeling
7629 and loop exit test optimizations.
7630 This option is enabled by default.
7632 @item -funsafe-loop-optimizations
7633 @opindex funsafe-loop-optimizations
7634 This option tells the loop optimizer to assume that loop indices do not
7635 overflow, and that loops with nontrivial exit condition are not
7636 infinite. This enables a wider range of loop optimizations even if
7637 the loop optimizer itself cannot prove that these assumptions are valid.
7638 If you use @option{-Wunsafe-loop-optimizations}, the compiler warns you
7639 if it finds this kind of loop.
7641 @item -fcrossjumping
7642 @opindex fcrossjumping
7643 Perform cross-jumping transformation.
7644 This transformation unifies equivalent code and saves code size. The
7645 resulting code may or may not perform better than without cross-jumping.
7647 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7649 @item -fauto-inc-dec
7650 @opindex fauto-inc-dec
7651 Combine increments or decrements of addresses with memory accesses.
7652 This pass is always skipped on architectures that do not have
7653 instructions to support this. Enabled by default at @option{-O} and
7654 higher on architectures that support this.
7658 Perform dead code elimination (DCE) on RTL@.
7659 Enabled by default at @option{-O} and higher.
7663 Perform dead store elimination (DSE) on RTL@.
7664 Enabled by default at @option{-O} and higher.
7666 @item -fif-conversion
7667 @opindex fif-conversion
7668 Attempt to transform conditional jumps into branch-less equivalents. This
7669 includes use of conditional moves, min, max, set flags and abs instructions, and
7670 some tricks doable by standard arithmetics. The use of conditional execution
7671 on chips where it is available is controlled by @code{if-conversion2}.
7673 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7675 @item -fif-conversion2
7676 @opindex fif-conversion2
7677 Use conditional execution (where available) to transform conditional jumps into
7678 branch-less equivalents.
7680 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7682 @item -fdeclone-ctor-dtor
7683 @opindex fdeclone-ctor-dtor
7684 The C++ ABI requires multiple entry points for constructors and
7685 destructors: one for a base subobject, one for a complete object, and
7686 one for a virtual destructor that calls operator delete afterwards.
7687 For a hierarchy with virtual bases, the base and complete variants are
7688 clones, which means two copies of the function. With this option, the
7689 base and complete variants are changed to be thunks that call a common
7692 Enabled by @option{-Os}.
7694 @item -fdelete-null-pointer-checks
7695 @opindex fdelete-null-pointer-checks
7696 Assume that programs cannot safely dereference null pointers, and that
7697 no code or data element resides there. This enables simple constant
7698 folding optimizations at all optimization levels. In addition, other
7699 optimization passes in GCC use this flag to control global dataflow
7700 analyses that eliminate useless checks for null pointers; these assume
7701 that if a pointer is checked after it has already been dereferenced,
7704 Note however that in some environments this assumption is not true.
7705 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
7706 for programs that depend on that behavior.
7708 Some targets, especially embedded ones, disable this option at all levels.
7709 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
7710 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
7711 are enabled independently at different optimization levels.
7713 @item -fdevirtualize
7714 @opindex fdevirtualize
7715 Attempt to convert calls to virtual functions to direct calls. This
7716 is done both within a procedure and interprocedurally as part of
7717 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
7718 propagation (@option{-fipa-cp}).
7719 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7721 @item -fdevirtualize-speculatively
7722 @opindex fdevirtualize-speculatively
7723 Attempt to convert calls to virtual functions to speculative direct calls.
7724 Based on the analysis of the type inheritance graph, determine for a given call
7725 the set of likely targets. If the set is small, preferably of size 1, change
7726 the call into an conditional deciding on direct and indirect call. The
7727 speculative calls enable more optimizations, such as inlining. When they seem
7728 useless after further optimization, they are converted back into original form.
7730 @item -fexpensive-optimizations
7731 @opindex fexpensive-optimizations
7732 Perform a number of minor optimizations that are relatively expensive.
7734 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7738 Attempt to remove redundant extension instructions. This is especially
7739 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
7740 registers after writing to their lower 32-bit half.
7742 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
7743 @option{-O3}, @option{-Os}.
7745 @item -flive-range-shrinkage
7746 @opindex flive-range-shrinkage
7747 Attempt to decrease register pressure through register live range
7748 shrinkage. This is helpful for fast processors with small or moderate
7751 @item -fira-algorithm=@var{algorithm}
7752 Use the specified coloring algorithm for the integrated register
7753 allocator. The @var{algorithm} argument can be @samp{priority}, which
7754 specifies Chow's priority coloring, or @samp{CB}, which specifies
7755 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
7756 for all architectures, but for those targets that do support it, it is
7757 the default because it generates better code.
7759 @item -fira-region=@var{region}
7760 Use specified regions for the integrated register allocator. The
7761 @var{region} argument should be one of the following:
7766 Use all loops as register allocation regions.
7767 This can give the best results for machines with a small and/or
7768 irregular register set.
7771 Use all loops except for loops with small register pressure
7772 as the regions. This value usually gives
7773 the best results in most cases and for most architectures,
7774 and is enabled by default when compiling with optimization for speed
7775 (@option{-O}, @option{-O2}, @dots{}).
7778 Use all functions as a single region.
7779 This typically results in the smallest code size, and is enabled by default for
7780 @option{-Os} or @option{-O0}.
7784 @item -fira-hoist-pressure
7785 @opindex fira-hoist-pressure
7786 Use IRA to evaluate register pressure in the code hoisting pass for
7787 decisions to hoist expressions. This option usually results in smaller
7788 code, but it can slow the compiler down.
7790 This option is enabled at level @option{-Os} for all targets.
7792 @item -fira-loop-pressure
7793 @opindex fira-loop-pressure
7794 Use IRA to evaluate register pressure in loops for decisions to move
7795 loop invariants. This option usually results in generation
7796 of faster and smaller code on machines with large register files (>= 32
7797 registers), but it can slow the compiler down.
7799 This option is enabled at level @option{-O3} for some targets.
7801 @item -fno-ira-share-save-slots
7802 @opindex fno-ira-share-save-slots
7803 Disable sharing of stack slots used for saving call-used hard
7804 registers living through a call. Each hard register gets a
7805 separate stack slot, and as a result function stack frames are
7808 @item -fno-ira-share-spill-slots
7809 @opindex fno-ira-share-spill-slots
7810 Disable sharing of stack slots allocated for pseudo-registers. Each
7811 pseudo-register that does not get a hard register gets a separate
7812 stack slot, and as a result function stack frames are larger.
7814 @item -fira-verbose=@var{n}
7815 @opindex fira-verbose
7816 Control the verbosity of the dump file for the integrated register allocator.
7817 The default value is 5. If the value @var{n} is greater or equal to 10,
7818 the dump output is sent to stderr using the same format as @var{n} minus 10.
7821 @opindex fcaller-saves
7822 Enable CFG-sensitive rematerialization in LRA. Instead of loading
7823 values of spilled pseudos, LRA tries to rematerialize (recalculate)
7824 values if it is profitable.
7826 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7828 @item -fdelayed-branch
7829 @opindex fdelayed-branch
7830 If supported for the target machine, attempt to reorder instructions
7831 to exploit instruction slots available after delayed branch
7834 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7836 @item -fschedule-insns
7837 @opindex fschedule-insns
7838 If supported for the target machine, attempt to reorder instructions to
7839 eliminate execution stalls due to required data being unavailable. This
7840 helps machines that have slow floating point or memory load instructions
7841 by allowing other instructions to be issued until the result of the load
7842 or floating-point instruction is required.
7844 Enabled at levels @option{-O2}, @option{-O3}.
7846 @item -fschedule-insns2
7847 @opindex fschedule-insns2
7848 Similar to @option{-fschedule-insns}, but requests an additional pass of
7849 instruction scheduling after register allocation has been done. This is
7850 especially useful on machines with a relatively small number of
7851 registers and where memory load instructions take more than one cycle.
7853 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7855 @item -fno-sched-interblock
7856 @opindex fno-sched-interblock
7857 Don't schedule instructions across basic blocks. This is normally
7858 enabled by default when scheduling before register allocation, i.e.@:
7859 with @option{-fschedule-insns} or at @option{-O2} or higher.
7861 @item -fno-sched-spec
7862 @opindex fno-sched-spec
7863 Don't allow speculative motion of non-load instructions. This is normally
7864 enabled by default when scheduling before register allocation, i.e.@:
7865 with @option{-fschedule-insns} or at @option{-O2} or higher.
7867 @item -fsched-pressure
7868 @opindex fsched-pressure
7869 Enable register pressure sensitive insn scheduling before register
7870 allocation. This only makes sense when scheduling before register
7871 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
7872 @option{-O2} or higher. Usage of this option can improve the
7873 generated code and decrease its size by preventing register pressure
7874 increase above the number of available hard registers and subsequent
7875 spills in register allocation.
7877 @item -fsched-spec-load
7878 @opindex fsched-spec-load
7879 Allow speculative motion of some load instructions. This only makes
7880 sense when scheduling before register allocation, i.e.@: with
7881 @option{-fschedule-insns} or at @option{-O2} or higher.
7883 @item -fsched-spec-load-dangerous
7884 @opindex fsched-spec-load-dangerous
7885 Allow speculative motion of more load instructions. This only makes
7886 sense when scheduling before register allocation, i.e.@: with
7887 @option{-fschedule-insns} or at @option{-O2} or higher.
7889 @item -fsched-stalled-insns
7890 @itemx -fsched-stalled-insns=@var{n}
7891 @opindex fsched-stalled-insns
7892 Define how many insns (if any) can be moved prematurely from the queue
7893 of stalled insns into the ready list during the second scheduling pass.
7894 @option{-fno-sched-stalled-insns} means that no insns are moved
7895 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
7896 on how many queued insns can be moved prematurely.
7897 @option{-fsched-stalled-insns} without a value is equivalent to
7898 @option{-fsched-stalled-insns=1}.
7900 @item -fsched-stalled-insns-dep
7901 @itemx -fsched-stalled-insns-dep=@var{n}
7902 @opindex fsched-stalled-insns-dep
7903 Define how many insn groups (cycles) are examined for a dependency
7904 on a stalled insn that is a candidate for premature removal from the queue
7905 of stalled insns. This has an effect only during the second scheduling pass,
7906 and only if @option{-fsched-stalled-insns} is used.
7907 @option{-fno-sched-stalled-insns-dep} is equivalent to
7908 @option{-fsched-stalled-insns-dep=0}.
7909 @option{-fsched-stalled-insns-dep} without a value is equivalent to
7910 @option{-fsched-stalled-insns-dep=1}.
7912 @item -fsched2-use-superblocks
7913 @opindex fsched2-use-superblocks
7914 When scheduling after register allocation, use superblock scheduling.
7915 This allows motion across basic block boundaries,
7916 resulting in faster schedules. This option is experimental, as not all machine
7917 descriptions used by GCC model the CPU closely enough to avoid unreliable
7918 results from the algorithm.
7920 This only makes sense when scheduling after register allocation, i.e.@: with
7921 @option{-fschedule-insns2} or at @option{-O2} or higher.
7923 @item -fsched-group-heuristic
7924 @opindex fsched-group-heuristic
7925 Enable the group heuristic in the scheduler. This heuristic favors
7926 the instruction that belongs to a schedule group. This is enabled
7927 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7928 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7930 @item -fsched-critical-path-heuristic
7931 @opindex fsched-critical-path-heuristic
7932 Enable the critical-path heuristic in the scheduler. This heuristic favors
7933 instructions on the critical path. This is enabled by default when
7934 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7935 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7937 @item -fsched-spec-insn-heuristic
7938 @opindex fsched-spec-insn-heuristic
7939 Enable the speculative instruction heuristic in the scheduler. This
7940 heuristic favors speculative instructions with greater dependency weakness.
7941 This is enabled by default when scheduling is enabled, i.e.@:
7942 with @option{-fschedule-insns} or @option{-fschedule-insns2}
7943 or at @option{-O2} or higher.
7945 @item -fsched-rank-heuristic
7946 @opindex fsched-rank-heuristic
7947 Enable the rank heuristic in the scheduler. This heuristic favors
7948 the instruction belonging to a basic block with greater size or frequency.
7949 This is enabled by default when scheduling is enabled, i.e.@:
7950 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7951 at @option{-O2} or higher.
7953 @item -fsched-last-insn-heuristic
7954 @opindex fsched-last-insn-heuristic
7955 Enable the last-instruction heuristic in the scheduler. This heuristic
7956 favors the instruction that is less dependent on the last instruction
7957 scheduled. This is enabled by default when scheduling is enabled,
7958 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7959 at @option{-O2} or higher.
7961 @item -fsched-dep-count-heuristic
7962 @opindex fsched-dep-count-heuristic
7963 Enable the dependent-count heuristic in the scheduler. This heuristic
7964 favors the instruction that has more instructions depending on it.
7965 This is enabled by default when scheduling is enabled, i.e.@:
7966 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7967 at @option{-O2} or higher.
7969 @item -freschedule-modulo-scheduled-loops
7970 @opindex freschedule-modulo-scheduled-loops
7971 Modulo scheduling is performed before traditional scheduling. If a loop
7972 is modulo scheduled, later scheduling passes may change its schedule.
7973 Use this option to control that behavior.
7975 @item -fselective-scheduling
7976 @opindex fselective-scheduling
7977 Schedule instructions using selective scheduling algorithm. Selective
7978 scheduling runs instead of the first scheduler pass.
7980 @item -fselective-scheduling2
7981 @opindex fselective-scheduling2
7982 Schedule instructions using selective scheduling algorithm. Selective
7983 scheduling runs instead of the second scheduler pass.
7985 @item -fsel-sched-pipelining
7986 @opindex fsel-sched-pipelining
7987 Enable software pipelining of innermost loops during selective scheduling.
7988 This option has no effect unless one of @option{-fselective-scheduling} or
7989 @option{-fselective-scheduling2} is turned on.
7991 @item -fsel-sched-pipelining-outer-loops
7992 @opindex fsel-sched-pipelining-outer-loops
7993 When pipelining loops during selective scheduling, also pipeline outer loops.
7994 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
7996 @item -fsemantic-interposition
7997 @opindex fsemantic-interposition
7998 Some object formats, like ELF, allow interposing of symbols by dynamic linker.
7999 This means that for symbols exported from the DSO compiler can not perform
8000 inter-procedural propagation, inlining and other optimizations in anticipation
8001 that the function or variable in question may change. While this feature is
8002 useful, for example, to rewrite memory allocation functions by a debugging
8003 implementation, it is expensive in the terms of code quality.
8004 With @option{-fno-semantic-inteposition} compiler assumest that if interposition
8005 happens for functions the overwritting function will have
8006 precisely same semantics (and side effects). Similarly if interposition happens
8007 for variables, the constructor of the variable will be the same. The flag
8008 has no effect for functions explicitly declared inline, where
8009 interposition changing semantic is never allowed and for symbols explicitly
8013 @opindex fshrink-wrap
8014 Emit function prologues only before parts of the function that need it,
8015 rather than at the top of the function. This flag is enabled by default at
8016 @option{-O} and higher.
8018 @item -fcaller-saves
8019 @opindex fcaller-saves
8020 Enable allocation of values to registers that are clobbered by
8021 function calls, by emitting extra instructions to save and restore the
8022 registers around such calls. Such allocation is done only when it
8023 seems to result in better code.
8025 This option is always enabled by default on certain machines, usually
8026 those which have no call-preserved registers to use instead.
8028 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8030 @item -fcombine-stack-adjustments
8031 @opindex fcombine-stack-adjustments
8032 Tracks stack adjustments (pushes and pops) and stack memory references
8033 and then tries to find ways to combine them.
8035 Enabled by default at @option{-O1} and higher.
8037 @item -fuse-caller-save
8038 Use caller save registers for allocation if those registers are not used by
8039 any called function. In that case it is not necessary to save and restore
8040 them around calls. This is only possible if called functions are part of
8041 same compilation unit as current function and they are compiled before it.
8043 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8045 @item -fconserve-stack
8046 @opindex fconserve-stack
8047 Attempt to minimize stack usage. The compiler attempts to use less
8048 stack space, even if that makes the program slower. This option
8049 implies setting the @option{large-stack-frame} parameter to 100
8050 and the @option{large-stack-frame-growth} parameter to 400.
8052 @item -ftree-reassoc
8053 @opindex ftree-reassoc
8054 Perform reassociation on trees. This flag is enabled by default
8055 at @option{-O} and higher.
8059 Perform partial redundancy elimination (PRE) on trees. This flag is
8060 enabled by default at @option{-O2} and @option{-O3}.
8062 @item -ftree-partial-pre
8063 @opindex ftree-partial-pre
8064 Make partial redundancy elimination (PRE) more aggressive. This flag is
8065 enabled by default at @option{-O3}.
8067 @item -ftree-forwprop
8068 @opindex ftree-forwprop
8069 Perform forward propagation on trees. This flag is enabled by default
8070 at @option{-O} and higher.
8074 Perform full redundancy elimination (FRE) on trees. The difference
8075 between FRE and PRE is that FRE only considers expressions
8076 that are computed on all paths leading to the redundant computation.
8077 This analysis is faster than PRE, though it exposes fewer redundancies.
8078 This flag is enabled by default at @option{-O} and higher.
8080 @item -ftree-phiprop
8081 @opindex ftree-phiprop
8082 Perform hoisting of loads from conditional pointers on trees. This
8083 pass is enabled by default at @option{-O} and higher.
8085 @item -fhoist-adjacent-loads
8086 @opindex hoist-adjacent-loads
8087 Speculatively hoist loads from both branches of an if-then-else if the
8088 loads are from adjacent locations in the same structure and the target
8089 architecture has a conditional move instruction. This flag is enabled
8090 by default at @option{-O2} and higher.
8092 @item -ftree-copy-prop
8093 @opindex ftree-copy-prop
8094 Perform copy propagation on trees. This pass eliminates unnecessary
8095 copy operations. This flag is enabled by default at @option{-O} and
8098 @item -fipa-pure-const
8099 @opindex fipa-pure-const
8100 Discover which functions are pure or constant.
8101 Enabled by default at @option{-O} and higher.
8103 @item -fipa-reference
8104 @opindex fipa-reference
8105 Discover which static variables do not escape the
8107 Enabled by default at @option{-O} and higher.
8111 Perform interprocedural pointer analysis and interprocedural modification
8112 and reference analysis. This option can cause excessive memory and
8113 compile-time usage on large compilation units. It is not enabled by
8114 default at any optimization level.
8117 @opindex fipa-profile
8118 Perform interprocedural profile propagation. The functions called only from
8119 cold functions are marked as cold. Also functions executed once (such as
8120 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8121 functions and loop less parts of functions executed once are then optimized for
8123 Enabled by default at @option{-O} and higher.
8127 Perform interprocedural constant propagation.
8128 This optimization analyzes the program to determine when values passed
8129 to functions are constants and then optimizes accordingly.
8130 This optimization can substantially increase performance
8131 if the application has constants passed to functions.
8132 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8134 @item -fipa-cp-clone
8135 @opindex fipa-cp-clone
8136 Perform function cloning to make interprocedural constant propagation stronger.
8137 When enabled, interprocedural constant propagation performs function cloning
8138 when externally visible function can be called with constant arguments.
8139 Because this optimization can create multiple copies of functions,
8140 it may significantly increase code size
8141 (see @option{--param ipcp-unit-growth=@var{value}}).
8142 This flag is enabled by default at @option{-O3}.
8146 Perform Identical Code Folding for functions and read-only variables.
8147 The optimization reduces code size and may disturb unwind stacks by replacing
8148 a function by equivalent one with a different name. The optimization works
8149 more effectively with link time optimization enabled.
8151 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8152 works on different levels and thus the optimizations are not same - there are
8153 equivalences that are found only by GCC and equivalences found only by Gold.
8155 This flag is enabled by default at @option{-O2} and @option{-Os}.
8157 @item -fisolate-erroneous-paths-dereference
8158 Detect paths which trigger erroneous or undefined behaviour due to
8159 dereferencing a NULL pointer. Isolate those paths from the main control
8160 flow and turn the statement with erroneous or undefined behaviour into a trap.
8162 @item -fisolate-erroneous-paths-attribute
8163 Detect paths which trigger erroneous or undefined behaviour due a NULL value
8164 being used in a way which is forbidden by a @code{returns_nonnull} or @code{nonnull}
8165 attribute. Isolate those paths from the main control flow and turn the
8166 statement with erroneous or undefined behaviour into a trap. This is not
8167 currently enabled, but may be enabled by @code{-O2} in the future.
8171 Perform forward store motion on trees. This flag is
8172 enabled by default at @option{-O} and higher.
8174 @item -ftree-bit-ccp
8175 @opindex ftree-bit-ccp
8176 Perform sparse conditional bit constant propagation on trees and propagate
8177 pointer alignment information.
8178 This pass only operates on local scalar variables and is enabled by default
8179 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8183 Perform sparse conditional constant propagation (CCP) on trees. This
8184 pass only operates on local scalar variables and is enabled by default
8185 at @option{-O} and higher.
8188 @opindex fssa-phiopt
8189 Perform pattern matching on SSA PHI nodes to optimize conditional
8190 code. This pass is enabled by default at @option{-O} and higher.
8192 @item -ftree-switch-conversion
8193 Perform conversion of simple initializations in a switch to
8194 initializations from a scalar array. This flag is enabled by default
8195 at @option{-O2} and higher.
8197 @item -ftree-tail-merge
8198 Look for identical code sequences. When found, replace one with a jump to the
8199 other. This optimization is known as tail merging or cross jumping. This flag
8200 is enabled by default at @option{-O2} and higher. The compilation time
8202 be limited using @option{max-tail-merge-comparisons} parameter and
8203 @option{max-tail-merge-iterations} parameter.
8207 Perform dead code elimination (DCE) on trees. This flag is enabled by
8208 default at @option{-O} and higher.
8210 @item -ftree-builtin-call-dce
8211 @opindex ftree-builtin-call-dce
8212 Perform conditional dead code elimination (DCE) for calls to built-in functions
8213 that may set @code{errno} but are otherwise side-effect free. This flag is
8214 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8217 @item -ftree-dominator-opts
8218 @opindex ftree-dominator-opts
8219 Perform a variety of simple scalar cleanups (constant/copy
8220 propagation, redundancy elimination, range propagation and expression
8221 simplification) based on a dominator tree traversal. This also
8222 performs jump threading (to reduce jumps to jumps). This flag is
8223 enabled by default at @option{-O} and higher.
8227 Perform dead store elimination (DSE) on trees. A dead store is a store into
8228 a memory location that is later overwritten by another store without
8229 any intervening loads. In this case the earlier store can be deleted. This
8230 flag is enabled by default at @option{-O} and higher.
8234 Perform loop header copying on trees. This is beneficial since it increases
8235 effectiveness of code motion optimizations. It also saves one jump. This flag
8236 is enabled by default at @option{-O} and higher. It is not enabled
8237 for @option{-Os}, since it usually increases code size.
8239 @item -ftree-loop-optimize
8240 @opindex ftree-loop-optimize
8241 Perform loop optimizations on trees. This flag is enabled by default
8242 at @option{-O} and higher.
8244 @item -ftree-loop-linear
8245 @opindex ftree-loop-linear
8246 Perform loop interchange transformations on tree. Same as
8247 @option{-floop-interchange}. To use this code transformation, GCC has
8248 to be configured with @option{--with-ppl} and @option{--with-cloog} to
8249 enable the Graphite loop transformation infrastructure.
8251 @item -floop-interchange
8252 @opindex floop-interchange
8253 Perform loop interchange transformations on loops. Interchanging two
8254 nested loops switches the inner and outer loops. For example, given a
8259 A(J, I) = A(J, I) * C
8263 loop interchange transforms the loop as if it were written:
8267 A(J, I) = A(J, I) * C
8271 which can be beneficial when @code{N} is larger than the caches,
8272 because in Fortran, the elements of an array are stored in memory
8273 contiguously by column, and the original loop iterates over rows,
8274 potentially creating at each access a cache miss. This optimization
8275 applies to all the languages supported by GCC and is not limited to
8276 Fortran. To use this code transformation, GCC has to be configured
8277 with @option{--with-ppl} and @option{--with-cloog} to enable the
8278 Graphite loop transformation infrastructure.
8280 @item -floop-strip-mine
8281 @opindex floop-strip-mine
8282 Perform loop strip mining transformations on loops. Strip mining
8283 splits a loop into two nested loops. The outer loop has strides
8284 equal to the strip size and the inner loop has strides of the
8285 original loop within a strip. The strip length can be changed
8286 using the @option{loop-block-tile-size} parameter. For example,
8293 loop strip mining transforms the loop as if it were written:
8296 DO I = II, min (II + 50, N)
8301 This optimization applies to all the languages supported by GCC and is
8302 not limited to Fortran. To use this code transformation, GCC has to
8303 be configured with @option{--with-ppl} and @option{--with-cloog} to
8304 enable the Graphite loop transformation infrastructure.
8307 @opindex floop-block
8308 Perform loop blocking transformations on loops. Blocking strip mines
8309 each loop in the loop nest such that the memory accesses of the
8310 element loops fit inside caches. The strip length can be changed
8311 using the @option{loop-block-tile-size} parameter. For example, given
8316 A(J, I) = B(I) + C(J)
8320 loop blocking transforms the loop as if it were written:
8324 DO I = II, min (II + 50, N)
8325 DO J = JJ, min (JJ + 50, M)
8326 A(J, I) = B(I) + C(J)
8332 which can be beneficial when @code{M} is larger than the caches,
8333 because the innermost loop iterates over a smaller amount of data
8334 which can be kept in the caches. This optimization applies to all the
8335 languages supported by GCC and is not limited to Fortran. To use this
8336 code transformation, GCC has to be configured with @option{--with-ppl}
8337 and @option{--with-cloog} to enable the Graphite loop transformation
8340 @item -fgraphite-identity
8341 @opindex fgraphite-identity
8342 Enable the identity transformation for graphite. For every SCoP we generate
8343 the polyhedral representation and transform it back to gimple. Using
8344 @option{-fgraphite-identity} we can check the costs or benefits of the
8345 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8346 are also performed by the code generator CLooG, like index splitting and
8347 dead code elimination in loops.
8349 @item -floop-nest-optimize
8350 @opindex floop-nest-optimize
8351 Enable the ISL based loop nest optimizer. This is a generic loop nest
8352 optimizer based on the Pluto optimization algorithms. It calculates a loop
8353 structure optimized for data-locality and parallelism. This option
8356 @item -floop-unroll-and-jam
8357 @opindex floop-unroll-and-jam
8358 Enable unroll and jam for the ISL based loop nest optimizer. The unroll
8359 factor can be changed using the @option{loop-unroll-jam-size} parameter.
8360 The unrolled dimension (counting from the most inner one) can be changed
8361 using the @option{loop-unroll-jam-depth} parameter. .
8363 @item -floop-parallelize-all
8364 @opindex floop-parallelize-all
8365 Use the Graphite data dependence analysis to identify loops that can
8366 be parallelized. Parallelize all the loops that can be analyzed to
8367 not contain loop carried dependences without checking that it is
8368 profitable to parallelize the loops.
8370 @item -fcheck-data-deps
8371 @opindex fcheck-data-deps
8372 Compare the results of several data dependence analyzers. This option
8373 is used for debugging the data dependence analyzers.
8375 @item -ftree-loop-if-convert
8376 Attempt to transform conditional jumps in the innermost loops to
8377 branch-less equivalents. The intent is to remove control-flow from
8378 the innermost loops in order to improve the ability of the
8379 vectorization pass to handle these loops. This is enabled by default
8380 if vectorization is enabled.
8382 @item -ftree-loop-if-convert-stores
8383 Attempt to also if-convert conditional jumps containing memory writes.
8384 This transformation can be unsafe for multi-threaded programs as it
8385 transforms conditional memory writes into unconditional memory writes.
8388 for (i = 0; i < N; i++)
8394 for (i = 0; i < N; i++)
8395 A[i] = cond ? expr : A[i];
8397 potentially producing data races.
8399 @item -ftree-loop-distribution
8400 Perform loop distribution. This flag can improve cache performance on
8401 big loop bodies and allow further loop optimizations, like
8402 parallelization or vectorization, to take place. For example, the loop
8419 @item -ftree-loop-distribute-patterns
8420 Perform loop distribution of patterns that can be code generated with
8421 calls to a library. This flag is enabled by default at @option{-O3}.
8423 This pass distributes the initialization loops and generates a call to
8424 memset zero. For example, the loop
8440 and the initialization loop is transformed into a call to memset zero.
8442 @item -ftree-loop-im
8443 @opindex ftree-loop-im
8444 Perform loop invariant motion on trees. This pass moves only invariants that
8445 are hard to handle at RTL level (function calls, operations that expand to
8446 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8447 operands of conditions that are invariant out of the loop, so that we can use
8448 just trivial invariantness analysis in loop unswitching. The pass also includes
8451 @item -ftree-loop-ivcanon
8452 @opindex ftree-loop-ivcanon
8453 Create a canonical counter for number of iterations in loops for which
8454 determining number of iterations requires complicated analysis. Later
8455 optimizations then may determine the number easily. Useful especially
8456 in connection with unrolling.
8460 Perform induction variable optimizations (strength reduction, induction
8461 variable merging and induction variable elimination) on trees.
8463 @item -ftree-parallelize-loops=n
8464 @opindex ftree-parallelize-loops
8465 Parallelize loops, i.e., split their iteration space to run in n threads.
8466 This is only possible for loops whose iterations are independent
8467 and can be arbitrarily reordered. The optimization is only
8468 profitable on multiprocessor machines, for loops that are CPU-intensive,
8469 rather than constrained e.g.@: by memory bandwidth. This option
8470 implies @option{-pthread}, and thus is only supported on targets
8471 that have support for @option{-pthread}.
8475 Perform function-local points-to analysis on trees. This flag is
8476 enabled by default at @option{-O} and higher.
8480 Perform scalar replacement of aggregates. This pass replaces structure
8481 references with scalars to prevent committing structures to memory too
8482 early. This flag is enabled by default at @option{-O} and higher.
8484 @item -ftree-copyrename
8485 @opindex ftree-copyrename
8486 Perform copy renaming on trees. This pass attempts to rename compiler
8487 temporaries to other variables at copy locations, usually resulting in
8488 variable names which more closely resemble the original variables. This flag
8489 is enabled by default at @option{-O} and higher.
8491 @item -ftree-coalesce-inlined-vars
8492 @opindex ftree-coalesce-inlined-vars
8493 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8494 combine small user-defined variables too, but only if they were inlined
8495 from other functions. It is a more limited form of
8496 @option{-ftree-coalesce-vars}. This may harm debug information of such
8497 inlined variables, but it will keep variables of the inlined-into
8498 function apart from each other, such that they are more likely to
8499 contain the expected values in a debugging session. This was the
8500 default in GCC versions older than 4.7.
8502 @item -ftree-coalesce-vars
8503 @opindex ftree-coalesce-vars
8504 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8505 combine small user-defined variables too, instead of just compiler
8506 temporaries. This may severely limit the ability to debug an optimized
8507 program compiled with @option{-fno-var-tracking-assignments}. In the
8508 negated form, this flag prevents SSA coalescing of user variables,
8509 including inlined ones. This option is enabled by default.
8513 Perform temporary expression replacement during the SSA->normal phase. Single
8514 use/single def temporaries are replaced at their use location with their
8515 defining expression. This results in non-GIMPLE code, but gives the expanders
8516 much more complex trees to work on resulting in better RTL generation. This is
8517 enabled by default at @option{-O} and higher.
8521 Perform straight-line strength reduction on trees. This recognizes related
8522 expressions involving multiplications and replaces them by less expensive
8523 calculations when possible. This is enabled by default at @option{-O} and
8526 @item -ftree-vectorize
8527 @opindex ftree-vectorize
8528 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8529 and @option{-ftree-slp-vectorize} if not explicitly specified.
8531 @item -ftree-loop-vectorize
8532 @opindex ftree-loop-vectorize
8533 Perform loop vectorization on trees. This flag is enabled by default at
8534 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8536 @item -ftree-slp-vectorize
8537 @opindex ftree-slp-vectorize
8538 Perform basic block vectorization on trees. This flag is enabled by default at
8539 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8541 @item -fvect-cost-model=@var{model}
8542 @opindex fvect-cost-model
8543 Alter the cost model used for vectorization. The @var{model} argument
8544 should be one of @code{unlimited}, @code{dynamic} or @code{cheap}.
8545 With the @code{unlimited} model the vectorized code-path is assumed
8546 to be profitable while with the @code{dynamic} model a runtime check
8547 will guard the vectorized code-path to enable it only for iteration
8548 counts that will likely execute faster than when executing the original
8549 scalar loop. The @code{cheap} model will disable vectorization of
8550 loops where doing so would be cost prohibitive for example due to
8551 required runtime checks for data dependence or alignment but otherwise
8552 is equal to the @code{dynamic} model.
8553 The default cost model depends on other optimization flags and is
8554 either @code{dynamic} or @code{cheap}.
8556 @item -fsimd-cost-model=@var{model}
8557 @opindex fsimd-cost-model
8558 Alter the cost model used for vectorization of loops marked with the OpenMP
8559 or Cilk Plus simd directive. The @var{model} argument should be one of
8560 @code{unlimited}, @code{dynamic}, @code{cheap}. All values of @var{model}
8561 have the same meaning as described in @option{-fvect-cost-model} and by
8562 default a cost model defined with @option{-fvect-cost-model} is used.
8566 Perform Value Range Propagation on trees. This is similar to the
8567 constant propagation pass, but instead of values, ranges of values are
8568 propagated. This allows the optimizers to remove unnecessary range
8569 checks like array bound checks and null pointer checks. This is
8570 enabled by default at @option{-O2} and higher. Null pointer check
8571 elimination is only done if @option{-fdelete-null-pointer-checks} is
8576 Perform tail duplication to enlarge superblock size. This transformation
8577 simplifies the control flow of the function allowing other optimizations to do
8580 @item -funroll-loops
8581 @opindex funroll-loops
8582 Unroll loops whose number of iterations can be determined at compile
8583 time or upon entry to the loop. @option{-funroll-loops} implies
8584 @option{-frerun-cse-after-loop}. This option makes code larger,
8585 and may or may not make it run faster.
8587 @item -funroll-all-loops
8588 @opindex funroll-all-loops
8589 Unroll all loops, even if their number of iterations is uncertain when
8590 the loop is entered. This usually makes programs run more slowly.
8591 @option{-funroll-all-loops} implies the same options as
8592 @option{-funroll-loops},
8594 @item -fsplit-ivs-in-unroller
8595 @opindex fsplit-ivs-in-unroller
8596 Enables expression of values of induction variables in later iterations
8597 of the unrolled loop using the value in the first iteration. This breaks
8598 long dependency chains, thus improving efficiency of the scheduling passes.
8600 A combination of @option{-fweb} and CSE is often sufficient to obtain the
8601 same effect. However, that is not reliable in cases where the loop body
8602 is more complicated than a single basic block. It also does not work at all
8603 on some architectures due to restrictions in the CSE pass.
8605 This optimization is enabled by default.
8607 @item -fvariable-expansion-in-unroller
8608 @opindex fvariable-expansion-in-unroller
8609 With this option, the compiler creates multiple copies of some
8610 local variables when unrolling a loop, which can result in superior code.
8612 @item -fpartial-inlining
8613 @opindex fpartial-inlining
8614 Inline parts of functions. This option has any effect only
8615 when inlining itself is turned on by the @option{-finline-functions}
8616 or @option{-finline-small-functions} options.
8618 Enabled at level @option{-O2}.
8620 @item -fpredictive-commoning
8621 @opindex fpredictive-commoning
8622 Perform predictive commoning optimization, i.e., reusing computations
8623 (especially memory loads and stores) performed in previous
8624 iterations of loops.
8626 This option is enabled at level @option{-O3}.
8628 @item -fprefetch-loop-arrays
8629 @opindex fprefetch-loop-arrays
8630 If supported by the target machine, generate instructions to prefetch
8631 memory to improve the performance of loops that access large arrays.
8633 This option may generate better or worse code; results are highly
8634 dependent on the structure of loops within the source code.
8636 Disabled at level @option{-Os}.
8639 @itemx -fno-peephole2
8640 @opindex fno-peephole
8641 @opindex fno-peephole2
8642 Disable any machine-specific peephole optimizations. The difference
8643 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
8644 are implemented in the compiler; some targets use one, some use the
8645 other, a few use both.
8647 @option{-fpeephole} is enabled by default.
8648 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8650 @item -fno-guess-branch-probability
8651 @opindex fno-guess-branch-probability
8652 Do not guess branch probabilities using heuristics.
8654 GCC uses heuristics to guess branch probabilities if they are
8655 not provided by profiling feedback (@option{-fprofile-arcs}). These
8656 heuristics are based on the control flow graph. If some branch probabilities
8657 are specified by @samp{__builtin_expect}, then the heuristics are
8658 used to guess branch probabilities for the rest of the control flow graph,
8659 taking the @samp{__builtin_expect} info into account. The interactions
8660 between the heuristics and @samp{__builtin_expect} can be complex, and in
8661 some cases, it may be useful to disable the heuristics so that the effects
8662 of @samp{__builtin_expect} are easier to understand.
8664 The default is @option{-fguess-branch-probability} at levels
8665 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8667 @item -freorder-blocks
8668 @opindex freorder-blocks
8669 Reorder basic blocks in the compiled function in order to reduce number of
8670 taken branches and improve code locality.
8672 Enabled at levels @option{-O2}, @option{-O3}.
8674 @item -freorder-blocks-and-partition
8675 @opindex freorder-blocks-and-partition
8676 In addition to reordering basic blocks in the compiled function, in order
8677 to reduce number of taken branches, partitions hot and cold basic blocks
8678 into separate sections of the assembly and .o files, to improve
8679 paging and cache locality performance.
8681 This optimization is automatically turned off in the presence of
8682 exception handling, for linkonce sections, for functions with a user-defined
8683 section attribute and on any architecture that does not support named
8686 Enabled for x86 at levels @option{-O2}, @option{-O3}.
8688 @item -freorder-functions
8689 @opindex freorder-functions
8690 Reorder functions in the object file in order to
8691 improve code locality. This is implemented by using special
8692 subsections @code{.text.hot} for most frequently executed functions and
8693 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
8694 the linker so object file format must support named sections and linker must
8695 place them in a reasonable way.
8697 Also profile feedback must be available to make this option effective. See
8698 @option{-fprofile-arcs} for details.
8700 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8702 @item -fstrict-aliasing
8703 @opindex fstrict-aliasing
8704 Allow the compiler to assume the strictest aliasing rules applicable to
8705 the language being compiled. For C (and C++), this activates
8706 optimizations based on the type of expressions. In particular, an
8707 object of one type is assumed never to reside at the same address as an
8708 object of a different type, unless the types are almost the same. For
8709 example, an @code{unsigned int} can alias an @code{int}, but not a
8710 @code{void*} or a @code{double}. A character type may alias any other
8713 @anchor{Type-punning}Pay special attention to code like this:
8726 The practice of reading from a different union member than the one most
8727 recently written to (called ``type-punning'') is common. Even with
8728 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
8729 is accessed through the union type. So, the code above works as
8730 expected. @xref{Structures unions enumerations and bit-fields
8731 implementation}. However, this code might not:
8742 Similarly, access by taking the address, casting the resulting pointer
8743 and dereferencing the result has undefined behavior, even if the cast
8744 uses a union type, e.g.:
8748 return ((union a_union *) &d)->i;
8752 The @option{-fstrict-aliasing} option is enabled at levels
8753 @option{-O2}, @option{-O3}, @option{-Os}.
8755 @item -fstrict-overflow
8756 @opindex fstrict-overflow
8757 Allow the compiler to assume strict signed overflow rules, depending
8758 on the language being compiled. For C (and C++) this means that
8759 overflow when doing arithmetic with signed numbers is undefined, which
8760 means that the compiler may assume that it does not happen. This
8761 permits various optimizations. For example, the compiler assumes
8762 that an expression like @code{i + 10 > i} is always true for
8763 signed @code{i}. This assumption is only valid if signed overflow is
8764 undefined, as the expression is false if @code{i + 10} overflows when
8765 using twos complement arithmetic. When this option is in effect any
8766 attempt to determine whether an operation on signed numbers
8767 overflows must be written carefully to not actually involve overflow.
8769 This option also allows the compiler to assume strict pointer
8770 semantics: given a pointer to an object, if adding an offset to that
8771 pointer does not produce a pointer to the same object, the addition is
8772 undefined. This permits the compiler to conclude that @code{p + u >
8773 p} is always true for a pointer @code{p} and unsigned integer
8774 @code{u}. This assumption is only valid because pointer wraparound is
8775 undefined, as the expression is false if @code{p + u} overflows using
8776 twos complement arithmetic.
8778 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
8779 that integer signed overflow is fully defined: it wraps. When
8780 @option{-fwrapv} is used, there is no difference between
8781 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
8782 integers. With @option{-fwrapv} certain types of overflow are
8783 permitted. For example, if the compiler gets an overflow when doing
8784 arithmetic on constants, the overflowed value can still be used with
8785 @option{-fwrapv}, but not otherwise.
8787 The @option{-fstrict-overflow} option is enabled at levels
8788 @option{-O2}, @option{-O3}, @option{-Os}.
8790 @item -falign-functions
8791 @itemx -falign-functions=@var{n}
8792 @opindex falign-functions
8793 Align the start of functions to the next power-of-two greater than
8794 @var{n}, skipping up to @var{n} bytes. For instance,
8795 @option{-falign-functions=32} aligns functions to the next 32-byte
8796 boundary, but @option{-falign-functions=24} aligns to the next
8797 32-byte boundary only if this can be done by skipping 23 bytes or less.
8799 @option{-fno-align-functions} and @option{-falign-functions=1} are
8800 equivalent and mean that functions are not aligned.
8802 Some assemblers only support this flag when @var{n} is a power of two;
8803 in that case, it is rounded up.
8805 If @var{n} is not specified or is zero, use a machine-dependent default.
8807 Enabled at levels @option{-O2}, @option{-O3}.
8809 @item -falign-labels
8810 @itemx -falign-labels=@var{n}
8811 @opindex falign-labels
8812 Align all branch targets to a power-of-two boundary, skipping up to
8813 @var{n} bytes like @option{-falign-functions}. This option can easily
8814 make code slower, because it must insert dummy operations for when the
8815 branch target is reached in the usual flow of the code.
8817 @option{-fno-align-labels} and @option{-falign-labels=1} are
8818 equivalent and mean that labels are not aligned.
8820 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
8821 are greater than this value, then their values are used instead.
8823 If @var{n} is not specified or is zero, use a machine-dependent default
8824 which is very likely to be @samp{1}, meaning no alignment.
8826 Enabled at levels @option{-O2}, @option{-O3}.
8829 @itemx -falign-loops=@var{n}
8830 @opindex falign-loops
8831 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
8832 like @option{-falign-functions}. If the loops are
8833 executed many times, this makes up for any execution of the dummy
8836 @option{-fno-align-loops} and @option{-falign-loops=1} are
8837 equivalent and mean that loops are not aligned.
8839 If @var{n} is not specified or is zero, use a machine-dependent default.
8841 Enabled at levels @option{-O2}, @option{-O3}.
8844 @itemx -falign-jumps=@var{n}
8845 @opindex falign-jumps
8846 Align branch targets to a power-of-two boundary, for branch targets
8847 where the targets can only be reached by jumping, skipping up to @var{n}
8848 bytes like @option{-falign-functions}. In this case, no dummy operations
8851 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
8852 equivalent and mean that loops are not aligned.
8854 If @var{n} is not specified or is zero, use a machine-dependent default.
8856 Enabled at levels @option{-O2}, @option{-O3}.
8858 @item -funit-at-a-time
8859 @opindex funit-at-a-time
8860 This option is left for compatibility reasons. @option{-funit-at-a-time}
8861 has no effect, while @option{-fno-unit-at-a-time} implies
8862 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
8866 @item -fno-toplevel-reorder
8867 @opindex fno-toplevel-reorder
8868 Do not reorder top-level functions, variables, and @code{asm}
8869 statements. Output them in the same order that they appear in the
8870 input file. When this option is used, unreferenced static variables
8871 are not removed. This option is intended to support existing code
8872 that relies on a particular ordering. For new code, it is better to
8873 use attributes when possible.
8875 Enabled at level @option{-O0}. When disabled explicitly, it also implies
8876 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
8881 Constructs webs as commonly used for register allocation purposes and assign
8882 each web individual pseudo register. This allows the register allocation pass
8883 to operate on pseudos directly, but also strengthens several other optimization
8884 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
8885 however, make debugging impossible, since variables no longer stay in a
8888 Enabled by default with @option{-funroll-loops}.
8890 @item -fwhole-program
8891 @opindex fwhole-program
8892 Assume that the current compilation unit represents the whole program being
8893 compiled. All public functions and variables with the exception of @code{main}
8894 and those merged by attribute @code{externally_visible} become static functions
8895 and in effect are optimized more aggressively by interprocedural optimizers.
8897 This option should not be used in combination with @code{-flto}.
8898 Instead relying on a linker plugin should provide safer and more precise
8901 @item -flto[=@var{n}]
8903 This option runs the standard link-time optimizer. When invoked
8904 with source code, it generates GIMPLE (one of GCC's internal
8905 representations) and writes it to special ELF sections in the object
8906 file. When the object files are linked together, all the function
8907 bodies are read from these ELF sections and instantiated as if they
8908 had been part of the same translation unit.
8910 To use the link-time optimizer, @option{-flto} and optimization
8911 options should be specified at compile time and during the final link.
8915 gcc -c -O2 -flto foo.c
8916 gcc -c -O2 -flto bar.c
8917 gcc -o myprog -flto -O2 foo.o bar.o
8920 The first two invocations to GCC save a bytecode representation
8921 of GIMPLE into special ELF sections inside @file{foo.o} and
8922 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
8923 @file{foo.o} and @file{bar.o}, merges the two files into a single
8924 internal image, and compiles the result as usual. Since both
8925 @file{foo.o} and @file{bar.o} are merged into a single image, this
8926 causes all the interprocedural analyses and optimizations in GCC to
8927 work across the two files as if they were a single one. This means,
8928 for example, that the inliner is able to inline functions in
8929 @file{bar.o} into functions in @file{foo.o} and vice-versa.
8931 Another (simpler) way to enable link-time optimization is:
8934 gcc -o myprog -flto -O2 foo.c bar.c
8937 The above generates bytecode for @file{foo.c} and @file{bar.c},
8938 merges them together into a single GIMPLE representation and optimizes
8939 them as usual to produce @file{myprog}.
8941 The only important thing to keep in mind is that to enable link-time
8942 optimizations you need to use the GCC driver to perform the link-step.
8943 GCC then automatically performs link-time optimization if any of the
8944 objects involved were compiled with the @option{-flto}. You generally
8945 should specify the optimization options to be used for link-time
8946 optimization though GCC will try to be clever at guessing an
8947 optimization level to use from the options used at compile-time
8948 if you fail to specify one at link-time. You can always override
8949 the automatic decision to do link-time optimization at link-time
8950 by passing @option{-fno-lto} to the link command.
8952 To make whole program optimization effective, it is necessary to make
8953 certain whole program assumptions. The compiler needs to know
8954 what functions and variables can be accessed by libraries and runtime
8955 outside of the link-time optimized unit. When supported by the linker,
8956 the linker plugin (see @option{-fuse-linker-plugin}) passes information
8957 to the compiler about used and externally visible symbols. When
8958 the linker plugin is not available, @option{-fwhole-program} should be
8959 used to allow the compiler to make these assumptions, which leads
8960 to more aggressive optimization decisions.
8962 When @option{-fuse-linker-plugin} is not enabled then, when a file is
8963 compiled with @option{-flto}, the generated object file is larger than
8964 a regular object file because it contains GIMPLE bytecodes and the usual
8965 final code (see @option{-ffat-lto-objects}. This means that
8966 object files with LTO information can be linked as normal object
8967 files; if @option{-fno-lto} is passed to the linker, no
8968 interprocedural optimizations are applied. Note that when
8969 @option{-fno-fat-lto-objects} is enabled the compile-stage is faster
8970 but you cannot perform a regular, non-LTO link on them.
8972 Additionally, the optimization flags used to compile individual files
8973 are not necessarily related to those used at link time. For instance,
8976 gcc -c -O0 -ffat-lto-objects -flto foo.c
8977 gcc -c -O0 -ffat-lto-objects -flto bar.c
8978 gcc -o myprog -O3 foo.o bar.o
8981 This produces individual object files with unoptimized assembler
8982 code, but the resulting binary @file{myprog} is optimized at
8983 @option{-O3}. If, instead, the final binary is generated with
8984 @option{-fno-lto}, then @file{myprog} is not optimized.
8986 When producing the final binary, GCC only
8987 applies link-time optimizations to those files that contain bytecode.
8988 Therefore, you can mix and match object files and libraries with
8989 GIMPLE bytecodes and final object code. GCC automatically selects
8990 which files to optimize in LTO mode and which files to link without
8993 There are some code generation flags preserved by GCC when
8994 generating bytecodes, as they need to be used during the final link
8995 stage. Generally options specified at link-time override those
8996 specified at compile-time.
8998 If you do not specify an optimization level option @option{-O} at
8999 link-time then GCC will compute one based on the optimization levels
9000 used when compiling the object files. The highest optimization
9001 level will win here.
9003 Currently, the following options and their setting are take from
9004 the first object file that explicitely specified it:
9005 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
9006 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
9007 and all the @option{-m} target flags.
9009 Certain ABI changing flags are required to match in all compilation-units
9010 and trying to override this at link-time with a conflicting value
9011 is ignored. This includes options such as @option{-freg-struct-return}
9012 and @option{-fpcc-struct-return}.
9014 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
9015 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
9016 are passed through to the link stage and merged conservatively for
9017 conflicting translation units. Specifically
9018 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
9019 precedence and for example @option{-ffp-contract=off} takes precedence
9020 over @option{-ffp-contract=fast}. You can override them at linke-time.
9022 It is recommended that you compile all the files participating in the
9023 same link with the same options and also specify those options at
9026 If LTO encounters objects with C linkage declared with incompatible
9027 types in separate translation units to be linked together (undefined
9028 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
9029 issued. The behavior is still undefined at run time. Similar
9030 diagnostics may be raised for other languages.
9032 Another feature of LTO is that it is possible to apply interprocedural
9033 optimizations on files written in different languages:
9038 gfortran -c -flto baz.f90
9039 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
9042 Notice that the final link is done with @command{g++} to get the C++
9043 runtime libraries and @option{-lgfortran} is added to get the Fortran
9044 runtime libraries. In general, when mixing languages in LTO mode, you
9045 should use the same link command options as when mixing languages in a
9046 regular (non-LTO) compilation.
9048 If object files containing GIMPLE bytecode are stored in a library archive, say
9049 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
9050 are using a linker with plugin support. To create static libraries suitable
9051 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
9052 and @code{ranlib}; to show the symbols of object files with GIMPLE bytecode, use
9053 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
9054 and @command{nm} have been compiled with plugin support. At link time, use the the
9055 flag @option{-fuse-linker-plugin} to ensure that the library participates in
9056 the LTO optimization process:
9059 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
9062 With the linker plugin enabled, the linker extracts the needed
9063 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
9064 to make them part of the aggregated GIMPLE image to be optimized.
9066 If you are not using a linker with plugin support and/or do not
9067 enable the linker plugin, then the objects inside @file{libfoo.a}
9068 are extracted and linked as usual, but they do not participate
9069 in the LTO optimization process. In order to make a static library suitable
9070 for both LTO optimization and usual linkage, compile its object files with
9071 @option{-flto} @code{-ffat-lto-objects}.
9073 Link-time optimizations do not require the presence of the whole program to
9074 operate. If the program does not require any symbols to be exported, it is
9075 possible to combine @option{-flto} and @option{-fwhole-program} to allow
9076 the interprocedural optimizers to use more aggressive assumptions which may
9077 lead to improved optimization opportunities.
9078 Use of @option{-fwhole-program} is not needed when linker plugin is
9079 active (see @option{-fuse-linker-plugin}).
9081 The current implementation of LTO makes no
9082 attempt to generate bytecode that is portable between different
9083 types of hosts. The bytecode files are versioned and there is a
9084 strict version check, so bytecode files generated in one version of
9085 GCC will not work with an older or newer version of GCC.
9087 Link-time optimization does not work well with generation of debugging
9088 information. Combining @option{-flto} with
9089 @option{-g} is currently experimental and expected to produce unexpected
9092 If you specify the optional @var{n}, the optimization and code
9093 generation done at link time is executed in parallel using @var{n}
9094 parallel jobs by utilizing an installed @command{make} program. The
9095 environment variable @env{MAKE} may be used to override the program
9096 used. The default value for @var{n} is 1.
9098 You can also specify @option{-flto=jobserver} to use GNU make's
9099 job server mode to determine the number of parallel jobs. This
9100 is useful when the Makefile calling GCC is already executing in parallel.
9101 You must prepend a @samp{+} to the command recipe in the parent Makefile
9102 for this to work. This option likely only works if @env{MAKE} is
9105 @item -flto-partition=@var{alg}
9106 @opindex flto-partition
9107 Specify the partitioning algorithm used by the link-time optimizer.
9108 The value is either @code{1to1} to specify a partitioning mirroring
9109 the original source files or @code{balanced} to specify partitioning
9110 into equally sized chunks (whenever possible) or @code{max} to create
9111 new partition for every symbol where possible. Specifying @code{none}
9112 as an algorithm disables partitioning and streaming completely.
9113 The default value is @code{balanced}. While @code{1to1} can be used
9114 as an workaround for various code ordering issues, the @code{max}
9115 partitioning is intended for internal testing only.
9116 The value @code{one} specifies that exactly one partition should be
9117 used while the value @code{none} bypasses partitioning and executes
9118 the link-time optimization step directly from the WPA phase.
9120 @item -flto-odr-type-merging
9121 @opindex flto-odr-type-merging
9122 Enable streaming of mangled types names of C++ types and their unification
9123 at linktime. This increases size of LTO object files, but enable
9124 diagnostics about One Definition Rule violations.
9126 @item -flto-compression-level=@var{n}
9127 This option specifies the level of compression used for intermediate
9128 language written to LTO object files, and is only meaningful in
9129 conjunction with LTO mode (@option{-flto}). Valid
9130 values are 0 (no compression) to 9 (maximum compression). Values
9131 outside this range are clamped to either 0 or 9. If the option is not
9132 given, a default balanced compression setting is used.
9135 Prints a report with internal details on the workings of the link-time
9136 optimizer. The contents of this report vary from version to version.
9137 It is meant to be useful to GCC developers when processing object
9138 files in LTO mode (via @option{-flto}).
9140 Disabled by default.
9142 @item -flto-report-wpa
9143 Like @option{-flto-report}, but only print for the WPA phase of Link
9146 @item -fuse-linker-plugin
9147 Enables the use of a linker plugin during link-time optimization. This
9148 option relies on plugin support in the linker, which is available in gold
9149 or in GNU ld 2.21 or newer.
9151 This option enables the extraction of object files with GIMPLE bytecode out
9152 of library archives. This improves the quality of optimization by exposing
9153 more code to the link-time optimizer. This information specifies what
9154 symbols can be accessed externally (by non-LTO object or during dynamic
9155 linking). Resulting code quality improvements on binaries (and shared
9156 libraries that use hidden visibility) are similar to @code{-fwhole-program}.
9157 See @option{-flto} for a description of the effect of this flag and how to
9160 This option is enabled by default when LTO support in GCC is enabled
9161 and GCC was configured for use with
9162 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9164 @item -ffat-lto-objects
9165 @opindex ffat-lto-objects
9166 Fat LTO objects are object files that contain both the intermediate language
9167 and the object code. This makes them usable for both LTO linking and normal
9168 linking. This option is effective only when compiling with @option{-flto}
9169 and is ignored at link time.
9171 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9172 requires the complete toolchain to be aware of LTO. It requires a linker with
9173 linker plugin support for basic functionality. Additionally,
9174 @command{nm}, @command{ar} and @command{ranlib}
9175 need to support linker plugins to allow a full-featured build environment
9176 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9177 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9178 to these tools. With non fat LTO makefiles need to be modified to use them.
9180 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9183 @item -fcompare-elim
9184 @opindex fcompare-elim
9185 After register allocation and post-register allocation instruction splitting,
9186 identify arithmetic instructions that compute processor flags similar to a
9187 comparison operation based on that arithmetic. If possible, eliminate the
9188 explicit comparison operation.
9190 This pass only applies to certain targets that cannot explicitly represent
9191 the comparison operation before register allocation is complete.
9193 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9196 @opindex fuse-ld=bfd
9197 Use the @command{bfd} linker instead of the default linker.
9200 @opindex fuse-ld=gold
9201 Use the @command{gold} linker instead of the default linker.
9203 @item -fcprop-registers
9204 @opindex fcprop-registers
9205 After register allocation and post-register allocation instruction splitting,
9206 perform a copy-propagation pass to try to reduce scheduling dependencies
9207 and occasionally eliminate the copy.
9209 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9211 @item -fprofile-correction
9212 @opindex fprofile-correction
9213 Profiles collected using an instrumented binary for multi-threaded programs may
9214 be inconsistent due to missed counter updates. When this option is specified,
9215 GCC uses heuristics to correct or smooth out such inconsistencies. By
9216 default, GCC emits an error message when an inconsistent profile is detected.
9218 @item -fprofile-dir=@var{path}
9219 @opindex fprofile-dir
9221 Set the directory to search for the profile data files in to @var{path}.
9222 This option affects only the profile data generated by
9223 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
9224 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
9225 and its related options. Both absolute and relative paths can be used.
9226 By default, GCC uses the current directory as @var{path}, thus the
9227 profile data file appears in the same directory as the object file.
9229 @item -fprofile-generate
9230 @itemx -fprofile-generate=@var{path}
9231 @opindex fprofile-generate
9233 Enable options usually used for instrumenting application to produce
9234 profile useful for later recompilation with profile feedback based
9235 optimization. You must use @option{-fprofile-generate} both when
9236 compiling and when linking your program.
9238 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
9240 If @var{path} is specified, GCC looks at the @var{path} to find
9241 the profile feedback data files. See @option{-fprofile-dir}.
9244 @itemx -fprofile-use=@var{path}
9245 @opindex fprofile-use
9246 Enable profile feedback directed optimizations, and optimizations
9247 generally profitable only with profile feedback available.
9249 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
9250 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}, @code{-ftree-vectorize},
9251 @code{ftree-loop-distribute-patterns}
9253 By default, GCC emits an error message if the feedback profiles do not
9254 match the source code. This error can be turned into a warning by using
9255 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9258 If @var{path} is specified, GCC looks at the @var{path} to find
9259 the profile feedback data files. See @option{-fprofile-dir}.
9261 @item -fauto-profile
9262 @itemx -fauto-profile=@var{path}
9263 @opindex fauto-profile
9264 Enable sampling based feedback directed optimizations, and optimizations
9265 generally profitable only with profile feedback available.
9267 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
9268 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}, @code{-ftree-vectorize},
9269 @code{-finline-functions}, @code{-fipa-cp}, @code{-fipa-cp-clone},
9270 @code{-fpredictive-commoning}, @code{-funswitch-loops},
9271 @code{-fgcse-after-reload}, @code{-ftree-loop-distribute-patterns},
9273 If @var{path} is specified, GCC looks at the @var{path} to find
9274 the profile feedback data files.
9276 In order to collect AutoFDO profile, you need to have:
9278 1. A linux system with linux perf support
9279 2. (optional) An Intel processor with last branch record (LBR) support. This is
9280 to guarantee accurate instruction level profile, which is important for
9281 AutoFDO performance.
9283 To collect the profile, first use linux perf to collect raw profile
9284 (see @uref{https://perf.wiki.kernel.org/}).
9287 @code{perf record -e br_inst_retired:near_taken -b -o perf.data -- your_program}
9289 Then use create_gcov tool, which takes raw profile and unstripped binary to
9290 generate AutoFDO profile that can be used by GCC.
9291 (see @uref{https://github.com/google/autofdo}).
9294 @code{create_gcov --binary=your_program.unstripped --profile=perf.data --gcov=profile.afdo}
9297 The following options control compiler behavior regarding floating-point
9298 arithmetic. These options trade off between speed and
9299 correctness. All must be specifically enabled.
9303 @opindex ffloat-store
9304 Do not store floating-point variables in registers, and inhibit other
9305 options that might change whether a floating-point value is taken from a
9308 @cindex floating-point precision
9309 This option prevents undesirable excess precision on machines such as
9310 the 68000 where the floating registers (of the 68881) keep more
9311 precision than a @code{double} is supposed to have. Similarly for the
9312 x86 architecture. For most programs, the excess precision does only
9313 good, but a few programs rely on the precise definition of IEEE floating
9314 point. Use @option{-ffloat-store} for such programs, after modifying
9315 them to store all pertinent intermediate computations into variables.
9317 @item -fexcess-precision=@var{style}
9318 @opindex fexcess-precision
9319 This option allows further control over excess precision on machines
9320 where floating-point registers have more precision than the IEEE
9321 @code{float} and @code{double} types and the processor does not
9322 support operations rounding to those types. By default,
9323 @option{-fexcess-precision=fast} is in effect; this means that
9324 operations are carried out in the precision of the registers and that
9325 it is unpredictable when rounding to the types specified in the source
9326 code takes place. When compiling C, if
9327 @option{-fexcess-precision=standard} is specified then excess
9328 precision follows the rules specified in ISO C99; in particular,
9329 both casts and assignments cause values to be rounded to their
9330 semantic types (whereas @option{-ffloat-store} only affects
9331 assignments). This option is enabled by default for C if a strict
9332 conformance option such as @option{-std=c99} is used.
9335 @option{-fexcess-precision=standard} is not implemented for languages
9336 other than C, and has no effect if
9337 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
9338 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
9339 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9340 semantics apply without excess precision, and in the latter, rounding
9345 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9346 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9347 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
9349 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9351 This option is not turned on by any @option{-O} option besides
9352 @option{-Ofast} since it can result in incorrect output for programs
9353 that depend on an exact implementation of IEEE or ISO rules/specifications
9354 for math functions. It may, however, yield faster code for programs
9355 that do not require the guarantees of these specifications.
9357 @item -fno-math-errno
9358 @opindex fno-math-errno
9359 Do not set @code{errno} after calling math functions that are executed
9360 with a single instruction, e.g., @code{sqrt}. A program that relies on
9361 IEEE exceptions for math error handling may want to use this flag
9362 for speed while maintaining IEEE arithmetic compatibility.
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.
9370 The default is @option{-fmath-errno}.
9372 On Darwin systems, the math library never sets @code{errno}. There is
9373 therefore no reason for the compiler to consider the possibility that
9374 it might, and @option{-fno-math-errno} is the default.
9376 @item -funsafe-math-optimizations
9377 @opindex funsafe-math-optimizations
9379 Allow optimizations for floating-point arithmetic that (a) assume
9380 that arguments and results are valid and (b) may violate IEEE or
9381 ANSI standards. When used at link-time, it may include libraries
9382 or startup files that change the default FPU control word or other
9383 similar optimizations.
9385 This option is not turned on by any @option{-O} option since
9386 it can result in incorrect output for programs that depend on
9387 an exact implementation of IEEE or ISO rules/specifications for
9388 math functions. It may, however, yield faster code for programs
9389 that do not require the guarantees of these specifications.
9390 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9391 @option{-fassociative-math} and @option{-freciprocal-math}.
9393 The default is @option{-fno-unsafe-math-optimizations}.
9395 @item -fassociative-math
9396 @opindex fassociative-math
9398 Allow re-association of operands in series of floating-point operations.
9399 This violates the ISO C and C++ language standard by possibly changing
9400 computation result. NOTE: re-ordering may change the sign of zero as
9401 well as ignore NaNs and inhibit or create underflow or overflow (and
9402 thus cannot be used on code that relies on rounding behavior like
9403 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9404 and thus may not be used when ordered comparisons are required.
9405 This option requires that both @option{-fno-signed-zeros} and
9406 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9407 much sense with @option{-frounding-math}. For Fortran the option
9408 is automatically enabled when both @option{-fno-signed-zeros} and
9409 @option{-fno-trapping-math} are in effect.
9411 The default is @option{-fno-associative-math}.
9413 @item -freciprocal-math
9414 @opindex freciprocal-math
9416 Allow the reciprocal of a value to be used instead of dividing by
9417 the value if this enables optimizations. For example @code{x / y}
9418 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9419 is subject to common subexpression elimination. Note that this loses
9420 precision and increases the number of flops operating on the value.
9422 The default is @option{-fno-reciprocal-math}.
9424 @item -ffinite-math-only
9425 @opindex ffinite-math-only
9426 Allow optimizations for floating-point arithmetic that assume
9427 that arguments and results are not NaNs or +-Infs.
9429 This option is not turned on by any @option{-O} option since
9430 it can result in incorrect output for programs that depend on
9431 an exact implementation of IEEE or ISO rules/specifications for
9432 math functions. It may, however, yield faster code for programs
9433 that do not require the guarantees of these specifications.
9435 The default is @option{-fno-finite-math-only}.
9437 @item -fno-signed-zeros
9438 @opindex fno-signed-zeros
9439 Allow optimizations for floating-point arithmetic that ignore the
9440 signedness of zero. IEEE arithmetic specifies the behavior of
9441 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9442 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9443 This option implies that the sign of a zero result isn't significant.
9445 The default is @option{-fsigned-zeros}.
9447 @item -fno-trapping-math
9448 @opindex fno-trapping-math
9449 Compile code assuming that floating-point operations cannot generate
9450 user-visible traps. These traps include division by zero, overflow,
9451 underflow, inexact result and invalid operation. This option requires
9452 that @option{-fno-signaling-nans} be in effect. Setting this option may
9453 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9455 This option should never be turned on by any @option{-O} option since
9456 it can result in incorrect output for programs that depend on
9457 an exact implementation of IEEE or ISO rules/specifications for
9460 The default is @option{-ftrapping-math}.
9462 @item -frounding-math
9463 @opindex frounding-math
9464 Disable transformations and optimizations that assume default floating-point
9465 rounding behavior. This is round-to-zero for all floating point
9466 to integer conversions, and round-to-nearest for all other arithmetic
9467 truncations. This option should be specified for programs that change
9468 the FP rounding mode dynamically, or that may be executed with a
9469 non-default rounding mode. This option disables constant folding of
9470 floating-point expressions at compile time (which may be affected by
9471 rounding mode) and arithmetic transformations that are unsafe in the
9472 presence of sign-dependent rounding modes.
9474 The default is @option{-fno-rounding-math}.
9476 This option is experimental and does not currently guarantee to
9477 disable all GCC optimizations that are affected by rounding mode.
9478 Future versions of GCC may provide finer control of this setting
9479 using C99's @code{FENV_ACCESS} pragma. This command-line option
9480 will be used to specify the default state for @code{FENV_ACCESS}.
9482 @item -fsignaling-nans
9483 @opindex fsignaling-nans
9484 Compile code assuming that IEEE signaling NaNs may generate user-visible
9485 traps during floating-point operations. Setting this option disables
9486 optimizations that may change the number of exceptions visible with
9487 signaling NaNs. This option implies @option{-ftrapping-math}.
9489 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9492 The default is @option{-fno-signaling-nans}.
9494 This option is experimental and does not currently guarantee to
9495 disable all GCC optimizations that affect signaling NaN behavior.
9497 @item -fsingle-precision-constant
9498 @opindex fsingle-precision-constant
9499 Treat floating-point constants as single precision instead of
9500 implicitly converting them to double-precision constants.
9502 @item -fcx-limited-range
9503 @opindex fcx-limited-range
9504 When enabled, this option states that a range reduction step is not
9505 needed when performing complex division. Also, there is no checking
9506 whether the result of a complex multiplication or division is @code{NaN
9507 + I*NaN}, with an attempt to rescue the situation in that case. The
9508 default is @option{-fno-cx-limited-range}, but is enabled by
9509 @option{-ffast-math}.
9511 This option controls the default setting of the ISO C99
9512 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9515 @item -fcx-fortran-rules
9516 @opindex fcx-fortran-rules
9517 Complex multiplication and division follow Fortran rules. Range
9518 reduction is done as part of complex division, but there is no checking
9519 whether the result of a complex multiplication or division is @code{NaN
9520 + I*NaN}, with an attempt to rescue the situation in that case.
9522 The default is @option{-fno-cx-fortran-rules}.
9526 The following options control optimizations that may improve
9527 performance, but are not enabled by any @option{-O} options. This
9528 section includes experimental options that may produce broken code.
9531 @item -fbranch-probabilities
9532 @opindex fbranch-probabilities
9533 After running a program compiled with @option{-fprofile-arcs}
9534 (@pxref{Debugging Options,, Options for Debugging Your Program or
9535 @command{gcc}}), you can compile it a second time using
9536 @option{-fbranch-probabilities}, to improve optimizations based on
9537 the number of times each branch was taken. When a program
9538 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9539 counts to a file called @file{@var{sourcename}.gcda} for each source
9540 file. The information in this data file is very dependent on the
9541 structure of the generated code, so you must use the same source code
9542 and the same optimization options for both compilations.
9544 With @option{-fbranch-probabilities}, GCC puts a
9545 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9546 These can be used to improve optimization. Currently, they are only
9547 used in one place: in @file{reorg.c}, instead of guessing which path a
9548 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9549 exactly determine which path is taken more often.
9551 @item -fprofile-values
9552 @opindex fprofile-values
9553 If combined with @option{-fprofile-arcs}, it adds code so that some
9554 data about values of expressions in the program is gathered.
9556 With @option{-fbranch-probabilities}, it reads back the data gathered
9557 from profiling values of expressions for usage in optimizations.
9559 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9561 @item -fprofile-reorder-functions
9562 @opindex fprofile-reorder-functions
9563 Function reordering based on profile instrumentation collects
9564 first time of execution of a function and orders these functions
9567 Enabled with @option{-fprofile-use}.
9571 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9572 to add code to gather information about values of expressions.
9574 With @option{-fbranch-probabilities}, it reads back the data gathered
9575 and actually performs the optimizations based on them.
9576 Currently the optimizations include specialization of division operations
9577 using the knowledge about the value of the denominator.
9579 @item -frename-registers
9580 @opindex frename-registers
9581 Attempt to avoid false dependencies in scheduled code by making use
9582 of registers left over after register allocation. This optimization
9583 most benefits processors with lots of registers. Depending on the
9584 debug information format adopted by the target, however, it can
9585 make debugging impossible, since variables no longer stay in
9586 a ``home register''.
9588 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
9590 @item -fschedule-fusion
9591 @opindex fschedule-fusion
9592 Performs a target dependent pass over the instruction stream to schedule
9593 instructions of same type together because target machine can execute them
9594 more efficiently if they are adjacent to each other in the instruction flow.
9596 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9600 Perform tail duplication to enlarge superblock size. This transformation
9601 simplifies the control flow of the function allowing other optimizations to do
9604 Enabled with @option{-fprofile-use}.
9606 @item -funroll-loops
9607 @opindex funroll-loops
9608 Unroll loops whose number of iterations can be determined at compile time or
9609 upon entry to the loop. @option{-funroll-loops} implies
9610 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
9611 It also turns on complete loop peeling (i.e.@: complete removal of loops with
9612 a small constant number of iterations). This option makes code larger, and may
9613 or may not make it run faster.
9615 Enabled with @option{-fprofile-use}.
9617 @item -funroll-all-loops
9618 @opindex funroll-all-loops
9619 Unroll all loops, even if their number of iterations is uncertain when
9620 the loop is entered. This usually makes programs run more slowly.
9621 @option{-funroll-all-loops} implies the same options as
9622 @option{-funroll-loops}.
9625 @opindex fpeel-loops
9626 Peels loops for which there is enough information that they do not
9627 roll much (from profile feedback). It also turns on complete loop peeling
9628 (i.e.@: complete removal of loops with small constant number of iterations).
9630 Enabled with @option{-fprofile-use}.
9632 @item -fmove-loop-invariants
9633 @opindex fmove-loop-invariants
9634 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
9635 at level @option{-O1}
9637 @item -funswitch-loops
9638 @opindex funswitch-loops
9639 Move branches with loop invariant conditions out of the loop, with duplicates
9640 of the loop on both branches (modified according to result of the condition).
9642 @item -ffunction-sections
9643 @itemx -fdata-sections
9644 @opindex ffunction-sections
9645 @opindex fdata-sections
9646 Place each function or data item into its own section in the output
9647 file if the target supports arbitrary sections. The name of the
9648 function or the name of the data item determines the section's name
9651 Use these options on systems where the linker can perform optimizations
9652 to improve locality of reference in the instruction space. Most systems
9653 using the ELF object format and SPARC processors running Solaris 2 have
9654 linkers with such optimizations. AIX may have these optimizations in
9657 Only use these options when there are significant benefits from doing
9658 so. When you specify these options, the assembler and linker
9659 create larger object and executable files and are also slower.
9660 You cannot use @code{gprof} on all systems if you
9661 specify this option, and you may have problems with debugging if
9662 you specify both this option and @option{-g}.
9664 @item -fbranch-target-load-optimize
9665 @opindex fbranch-target-load-optimize
9666 Perform branch target register load optimization before prologue / epilogue
9668 The use of target registers can typically be exposed only during reload,
9669 thus hoisting loads out of loops and doing inter-block scheduling needs
9670 a separate optimization pass.
9672 @item -fbranch-target-load-optimize2
9673 @opindex fbranch-target-load-optimize2
9674 Perform branch target register load optimization after prologue / epilogue
9677 @item -fbtr-bb-exclusive
9678 @opindex fbtr-bb-exclusive
9679 When performing branch target register load optimization, don't reuse
9680 branch target registers within any basic block.
9682 @item -fstack-protector
9683 @opindex fstack-protector
9684 Emit extra code to check for buffer overflows, such as stack smashing
9685 attacks. This is done by adding a guard variable to functions with
9686 vulnerable objects. This includes functions that call @code{alloca}, and
9687 functions with buffers larger than 8 bytes. The guards are initialized
9688 when a function is entered and then checked when the function exits.
9689 If a guard check fails, an error message is printed and the program exits.
9691 @item -fstack-protector-all
9692 @opindex fstack-protector-all
9693 Like @option{-fstack-protector} except that all functions are protected.
9695 @item -fstack-protector-strong
9696 @opindex fstack-protector-strong
9697 Like @option{-fstack-protector} but includes additional functions to
9698 be protected --- those that have local array definitions, or have
9699 references to local frame addresses.
9701 @item -fsection-anchors
9702 @opindex fsection-anchors
9703 Try to reduce the number of symbolic address calculations by using
9704 shared ``anchor'' symbols to address nearby objects. This transformation
9705 can help to reduce the number of GOT entries and GOT accesses on some
9708 For example, the implementation of the following function @code{foo}:
9712 int foo (void) @{ return a + b + c; @}
9716 usually calculates the addresses of all three variables, but if you
9717 compile it with @option{-fsection-anchors}, it accesses the variables
9718 from a common anchor point instead. The effect is similar to the
9719 following pseudocode (which isn't valid C):
9724 register int *xr = &x;
9725 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
9729 Not all targets support this option.
9731 @item --param @var{name}=@var{value}
9733 In some places, GCC uses various constants to control the amount of
9734 optimization that is done. For example, GCC does not inline functions
9735 that contain more than a certain number of instructions. You can
9736 control some of these constants on the command line using the
9737 @option{--param} option.
9739 The names of specific parameters, and the meaning of the values, are
9740 tied to the internals of the compiler, and are subject to change
9741 without notice in future releases.
9743 In each case, the @var{value} is an integer. The allowable choices for
9747 @item predictable-branch-outcome
9748 When branch is predicted to be taken with probability lower than this threshold
9749 (in percent), then it is considered well predictable. The default is 10.
9751 @item max-crossjump-edges
9752 The maximum number of incoming edges to consider for cross-jumping.
9753 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
9754 the number of edges incoming to each block. Increasing values mean
9755 more aggressive optimization, making the compilation time increase with
9756 probably small improvement in executable size.
9758 @item min-crossjump-insns
9759 The minimum number of instructions that must be matched at the end
9760 of two blocks before cross-jumping is performed on them. This
9761 value is ignored in the case where all instructions in the block being
9762 cross-jumped from are matched. The default value is 5.
9764 @item max-grow-copy-bb-insns
9765 The maximum code size expansion factor when copying basic blocks
9766 instead of jumping. The expansion is relative to a jump instruction.
9767 The default value is 8.
9769 @item max-goto-duplication-insns
9770 The maximum number of instructions to duplicate to a block that jumps
9771 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
9772 passes, GCC factors computed gotos early in the compilation process,
9773 and unfactors them as late as possible. Only computed jumps at the
9774 end of a basic blocks with no more than max-goto-duplication-insns are
9775 unfactored. The default value is 8.
9777 @item max-delay-slot-insn-search
9778 The maximum number of instructions to consider when looking for an
9779 instruction to fill a delay slot. If more than this arbitrary number of
9780 instructions are searched, the time savings from filling the delay slot
9781 are minimal, so stop searching. Increasing values mean more
9782 aggressive optimization, making the compilation time increase with probably
9783 small improvement in execution time.
9785 @item max-delay-slot-live-search
9786 When trying to fill delay slots, the maximum number of instructions to
9787 consider when searching for a block with valid live register
9788 information. Increasing this arbitrarily chosen value means more
9789 aggressive optimization, increasing the compilation time. This parameter
9790 should be removed when the delay slot code is rewritten to maintain the
9793 @item max-gcse-memory
9794 The approximate maximum amount of memory that can be allocated in
9795 order to perform the global common subexpression elimination
9796 optimization. If more memory than specified is required, the
9797 optimization is not done.
9799 @item max-gcse-insertion-ratio
9800 If the ratio of expression insertions to deletions is larger than this value
9801 for any expression, then RTL PRE inserts or removes the expression and thus
9802 leaves partially redundant computations in the instruction stream. The default value is 20.
9804 @item max-pending-list-length
9805 The maximum number of pending dependencies scheduling allows
9806 before flushing the current state and starting over. Large functions
9807 with few branches or calls can create excessively large lists which
9808 needlessly consume memory and resources.
9810 @item max-modulo-backtrack-attempts
9811 The maximum number of backtrack attempts the scheduler should make
9812 when modulo scheduling a loop. Larger values can exponentially increase
9815 @item max-inline-insns-single
9816 Several parameters control the tree inliner used in GCC@.
9817 This number sets the maximum number of instructions (counted in GCC's
9818 internal representation) in a single function that the tree inliner
9819 considers for inlining. This only affects functions declared
9820 inline and methods implemented in a class declaration (C++).
9821 The default value is 400.
9823 @item max-inline-insns-auto
9824 When you use @option{-finline-functions} (included in @option{-O3}),
9825 a lot of functions that would otherwise not be considered for inlining
9826 by the compiler are investigated. To those functions, a different
9827 (more restrictive) limit compared to functions declared inline can
9829 The default value is 40.
9831 @item inline-min-speedup
9832 When estimated performance improvement of caller + callee runtime exceeds this
9833 threshold (in precent), the function can be inlined regardless the limit on
9834 @option{--param max-inline-insns-single} and @option{--param
9835 max-inline-insns-auto}.
9837 @item large-function-insns
9838 The limit specifying really large functions. For functions larger than this
9839 limit after inlining, inlining is constrained by
9840 @option{--param large-function-growth}. This parameter is useful primarily
9841 to avoid extreme compilation time caused by non-linear algorithms used by the
9843 The default value is 2700.
9845 @item large-function-growth
9846 Specifies maximal growth of large function caused by inlining in percents.
9847 The default value is 100 which limits large function growth to 2.0 times
9850 @item large-unit-insns
9851 The limit specifying large translation unit. Growth caused by inlining of
9852 units larger than this limit is limited by @option{--param inline-unit-growth}.
9853 For small units this might be too tight.
9854 For example, consider a unit consisting of function A
9855 that is inline and B that just calls A three times. If B is small relative to
9856 A, the growth of unit is 300\% and yet such inlining is very sane. For very
9857 large units consisting of small inlineable functions, however, the overall unit
9858 growth limit is needed to avoid exponential explosion of code size. Thus for
9859 smaller units, the size is increased to @option{--param large-unit-insns}
9860 before applying @option{--param inline-unit-growth}. The default is 10000.
9862 @item inline-unit-growth
9863 Specifies maximal overall growth of the compilation unit caused by inlining.
9864 The default value is 30 which limits unit growth to 1.3 times the original
9865 size. Cold functions (either marked cold via an attribute or by profile
9866 feedback) are not accounted into the unit size.
9868 @item ipcp-unit-growth
9869 Specifies maximal overall growth of the compilation unit caused by
9870 interprocedural constant propagation. The default value is 10 which limits
9871 unit growth to 1.1 times the original size.
9873 @item large-stack-frame
9874 The limit specifying large stack frames. While inlining the algorithm is trying
9875 to not grow past this limit too much. The default value is 256 bytes.
9877 @item large-stack-frame-growth
9878 Specifies maximal growth of large stack frames caused by inlining in percents.
9879 The default value is 1000 which limits large stack frame growth to 11 times
9882 @item max-inline-insns-recursive
9883 @itemx max-inline-insns-recursive-auto
9884 Specifies the maximum number of instructions an out-of-line copy of a
9885 self-recursive inline
9886 function can grow into by performing recursive inlining.
9888 For functions declared inline, @option{--param max-inline-insns-recursive} is
9889 taken into account. For functions not declared inline, recursive inlining
9890 happens only when @option{-finline-functions} (included in @option{-O3}) is
9891 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
9892 default value is 450.
9894 @item max-inline-recursive-depth
9895 @itemx max-inline-recursive-depth-auto
9896 Specifies the maximum recursion depth used for recursive inlining.
9898 For functions declared inline, @option{--param max-inline-recursive-depth} is
9899 taken into account. For functions not declared inline, recursive inlining
9900 happens only when @option{-finline-functions} (included in @option{-O3}) is
9901 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
9904 @item min-inline-recursive-probability
9905 Recursive inlining is profitable only for function having deep recursion
9906 in average and can hurt for function having little recursion depth by
9907 increasing the prologue size or complexity of function body to other
9910 When profile feedback is available (see @option{-fprofile-generate}) the actual
9911 recursion depth can be guessed from probability that function recurses via a
9912 given call expression. This parameter limits inlining only to call expressions
9913 whose probability exceeds the given threshold (in percents).
9914 The default value is 10.
9916 @item early-inlining-insns
9917 Specify growth that the early inliner can make. In effect it increases
9918 the amount of inlining for code having a large abstraction penalty.
9919 The default value is 10.
9921 @item max-early-inliner-iterations
9922 @itemx max-early-inliner-iterations
9923 Limit of iterations of the early inliner. This basically bounds
9924 the number of nested indirect calls the early inliner can resolve.
9925 Deeper chains are still handled by late inlining.
9927 @item comdat-sharing-probability
9928 @itemx comdat-sharing-probability
9929 Probability (in percent) that C++ inline function with comdat visibility
9930 are shared across multiple compilation units. The default value is 20.
9932 @item profile-func-internal-id
9933 @itemx profile-func-internal-id
9934 A parameter to control whether to use function internal id in profile
9935 database lookup. If the value is 0, the compiler will use id that
9936 is based on function assembler name and filename, which makes old profile
9937 data more tolerant to source changes such as function reordering etc.
9938 The default value is 0.
9940 @item min-vect-loop-bound
9941 The minimum number of iterations under which loops are not vectorized
9942 when @option{-ftree-vectorize} is used. The number of iterations after
9943 vectorization needs to be greater than the value specified by this option
9944 to allow vectorization. The default value is 0.
9946 @item gcse-cost-distance-ratio
9947 Scaling factor in calculation of maximum distance an expression
9948 can be moved by GCSE optimizations. This is currently supported only in the
9949 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
9950 is with simple expressions, i.e., the expressions that have cost
9951 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
9952 hoisting of simple expressions. The default value is 10.
9954 @item gcse-unrestricted-cost
9955 Cost, roughly measured as the cost of a single typical machine
9956 instruction, at which GCSE optimizations do not constrain
9957 the distance an expression can travel. This is currently
9958 supported only in the code hoisting pass. The lesser the cost,
9959 the more aggressive code hoisting is. Specifying 0
9960 allows all expressions to travel unrestricted distances.
9961 The default value is 3.
9963 @item max-hoist-depth
9964 The depth of search in the dominator tree for expressions to hoist.
9965 This is used to avoid quadratic behavior in hoisting algorithm.
9966 The value of 0 does not limit on the search, but may slow down compilation
9967 of huge functions. The default value is 30.
9969 @item max-tail-merge-comparisons
9970 The maximum amount of similar bbs to compare a bb with. This is used to
9971 avoid quadratic behavior in tree tail merging. The default value is 10.
9973 @item max-tail-merge-iterations
9974 The maximum amount of iterations of the pass over the function. This is used to
9975 limit compilation time in tree tail merging. The default value is 2.
9977 @item max-unrolled-insns
9978 The maximum number of instructions that a loop may have to be unrolled.
9979 If a loop is unrolled, this parameter also determines how many times
9980 the loop code is unrolled.
9982 @item max-average-unrolled-insns
9983 The maximum number of instructions biased by probabilities of their execution
9984 that a loop may have to be unrolled. If a loop is unrolled,
9985 this parameter also determines how many times the loop code is unrolled.
9987 @item max-unroll-times
9988 The maximum number of unrollings of a single loop.
9990 @item max-peeled-insns
9991 The maximum number of instructions that a loop may have to be peeled.
9992 If a loop is peeled, this parameter also determines how many times
9993 the loop code is peeled.
9995 @item max-peel-times
9996 The maximum number of peelings of a single loop.
9998 @item max-peel-branches
9999 The maximum number of branches on the hot path through the peeled sequence.
10001 @item max-completely-peeled-insns
10002 The maximum number of insns of a completely peeled loop.
10004 @item max-completely-peel-times
10005 The maximum number of iterations of a loop to be suitable for complete peeling.
10007 @item max-completely-peel-loop-nest-depth
10008 The maximum depth of a loop nest suitable for complete peeling.
10010 @item max-unswitch-insns
10011 The maximum number of insns of an unswitched loop.
10013 @item max-unswitch-level
10014 The maximum number of branches unswitched in a single loop.
10016 @item lim-expensive
10017 The minimum cost of an expensive expression in the loop invariant motion.
10019 @item iv-consider-all-candidates-bound
10020 Bound on number of candidates for induction variables, below which
10021 all candidates are considered for each use in induction variable
10022 optimizations. If there are more candidates than this,
10023 only the most relevant ones are considered to avoid quadratic time complexity.
10025 @item iv-max-considered-uses
10026 The induction variable optimizations give up on loops that contain more
10027 induction variable uses.
10029 @item iv-always-prune-cand-set-bound
10030 If the number of candidates in the set is smaller than this value,
10031 always try to remove unnecessary ivs from the set
10032 when adding a new one.
10034 @item scev-max-expr-size
10035 Bound on size of expressions used in the scalar evolutions analyzer.
10036 Large expressions slow the analyzer.
10038 @item scev-max-expr-complexity
10039 Bound on the complexity of the expressions in the scalar evolutions analyzer.
10040 Complex expressions slow the analyzer.
10042 @item omega-max-vars
10043 The maximum number of variables in an Omega constraint system.
10044 The default value is 128.
10046 @item omega-max-geqs
10047 The maximum number of inequalities in an Omega constraint system.
10048 The default value is 256.
10050 @item omega-max-eqs
10051 The maximum number of equalities in an Omega constraint system.
10052 The default value is 128.
10054 @item omega-max-wild-cards
10055 The maximum number of wildcard variables that the Omega solver is
10056 able to insert. The default value is 18.
10058 @item omega-hash-table-size
10059 The size of the hash table in the Omega solver. The default value is
10062 @item omega-max-keys
10063 The maximal number of keys used by the Omega solver. The default
10066 @item omega-eliminate-redundant-constraints
10067 When set to 1, use expensive methods to eliminate all redundant
10068 constraints. The default value is 0.
10070 @item vect-max-version-for-alignment-checks
10071 The maximum number of run-time checks that can be performed when
10072 doing loop versioning for alignment in the vectorizer.
10074 @item vect-max-version-for-alias-checks
10075 The maximum number of run-time checks that can be performed when
10076 doing loop versioning for alias in the vectorizer.
10078 @item vect-max-peeling-for-alignment
10079 The maximum number of loop peels to enhance access alignment
10080 for vectorizer. Value -1 means 'no limit'.
10082 @item max-iterations-to-track
10083 The maximum number of iterations of a loop the brute-force algorithm
10084 for analysis of the number of iterations of the loop tries to evaluate.
10086 @item hot-bb-count-ws-permille
10087 A basic block profile count is considered hot if it contributes to
10088 the given permillage (i.e. 0...1000) of the entire profiled execution.
10090 @item hot-bb-frequency-fraction
10091 Select fraction of the entry block frequency of executions of basic block in
10092 function given basic block needs to have to be considered hot.
10094 @item max-predicted-iterations
10095 The maximum number of loop iterations we predict statically. This is useful
10096 in cases where a function contains a single loop with known bound and
10097 another loop with unknown bound.
10098 The known number of iterations is predicted correctly, while
10099 the unknown number of iterations average to roughly 10. This means that the
10100 loop without bounds appears artificially cold relative to the other one.
10102 @item builtin-expect-probability
10103 Control the probability of the expression having the specified value. This
10104 parameter takes a percentage (i.e. 0 ... 100) as input.
10105 The default probability of 90 is obtained empirically.
10107 @item align-threshold
10109 Select fraction of the maximal frequency of executions of a basic block in
10110 a function to align the basic block.
10112 @item align-loop-iterations
10114 A loop expected to iterate at least the selected number of iterations is
10117 @item tracer-dynamic-coverage
10118 @itemx tracer-dynamic-coverage-feedback
10120 This value is used to limit superblock formation once the given percentage of
10121 executed instructions is covered. This limits unnecessary code size
10124 The @option{tracer-dynamic-coverage-feedback} is used only when profile
10125 feedback is available. The real profiles (as opposed to statically estimated
10126 ones) are much less balanced allowing the threshold to be larger value.
10128 @item tracer-max-code-growth
10129 Stop tail duplication once code growth has reached given percentage. This is
10130 a rather artificial limit, as most of the duplicates are eliminated later in
10131 cross jumping, so it may be set to much higher values than is the desired code
10134 @item tracer-min-branch-ratio
10136 Stop reverse growth when the reverse probability of best edge is less than this
10137 threshold (in percent).
10139 @item tracer-min-branch-ratio
10140 @itemx tracer-min-branch-ratio-feedback
10142 Stop forward growth if the best edge has probability lower than this
10145 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
10146 compilation for profile feedback and one for compilation without. The value
10147 for compilation with profile feedback needs to be more conservative (higher) in
10148 order to make tracer effective.
10150 @item max-cse-path-length
10152 The maximum number of basic blocks on path that CSE considers.
10155 @item max-cse-insns
10156 The maximum number of instructions CSE processes before flushing.
10157 The default is 1000.
10159 @item ggc-min-expand
10161 GCC uses a garbage collector to manage its own memory allocation. This
10162 parameter specifies the minimum percentage by which the garbage
10163 collector's heap should be allowed to expand between collections.
10164 Tuning this may improve compilation speed; it has no effect on code
10167 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10168 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10169 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10170 GCC is not able to calculate RAM on a particular platform, the lower
10171 bound of 30% is used. Setting this parameter and
10172 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10173 every opportunity. This is extremely slow, but can be useful for
10176 @item ggc-min-heapsize
10178 Minimum size of the garbage collector's heap before it begins bothering
10179 to collect garbage. The first collection occurs after the heap expands
10180 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10181 tuning this may improve compilation speed, and has no effect on code
10184 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10185 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10186 with a lower bound of 4096 (four megabytes) and an upper bound of
10187 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10188 particular platform, the lower bound is used. Setting this parameter
10189 very large effectively disables garbage collection. Setting this
10190 parameter and @option{ggc-min-expand} to zero causes a full collection
10191 to occur at every opportunity.
10193 @item max-reload-search-insns
10194 The maximum number of instruction reload should look backward for equivalent
10195 register. Increasing values mean more aggressive optimization, making the
10196 compilation time increase with probably slightly better performance.
10197 The default value is 100.
10199 @item max-cselib-memory-locations
10200 The maximum number of memory locations cselib should take into account.
10201 Increasing values mean more aggressive optimization, making the compilation time
10202 increase with probably slightly better performance. The default value is 500.
10204 @item reorder-blocks-duplicate
10205 @itemx reorder-blocks-duplicate-feedback
10207 Used by the basic block reordering pass to decide whether to use unconditional
10208 branch or duplicate the code on its destination. Code is duplicated when its
10209 estimated size is smaller than this value multiplied by the estimated size of
10210 unconditional jump in the hot spots of the program.
10212 The @option{reorder-block-duplicate-feedback} is used only when profile
10213 feedback is available. It may be set to higher values than
10214 @option{reorder-block-duplicate} since information about the hot spots is more
10217 @item max-sched-ready-insns
10218 The maximum number of instructions ready to be issued the scheduler should
10219 consider at any given time during the first scheduling pass. Increasing
10220 values mean more thorough searches, making the compilation time increase
10221 with probably little benefit. The default value is 100.
10223 @item max-sched-region-blocks
10224 The maximum number of blocks in a region to be considered for
10225 interblock scheduling. The default value is 10.
10227 @item max-pipeline-region-blocks
10228 The maximum number of blocks in a region to be considered for
10229 pipelining in the selective scheduler. The default value is 15.
10231 @item max-sched-region-insns
10232 The maximum number of insns in a region to be considered for
10233 interblock scheduling. The default value is 100.
10235 @item max-pipeline-region-insns
10236 The maximum number of insns in a region to be considered for
10237 pipelining in the selective scheduler. The default value is 200.
10239 @item min-spec-prob
10240 The minimum probability (in percents) of reaching a source block
10241 for interblock speculative scheduling. The default value is 40.
10243 @item max-sched-extend-regions-iters
10244 The maximum number of iterations through CFG to extend regions.
10245 A value of 0 (the default) disables region extensions.
10247 @item max-sched-insn-conflict-delay
10248 The maximum conflict delay for an insn to be considered for speculative motion.
10249 The default value is 3.
10251 @item sched-spec-prob-cutoff
10252 The minimal probability of speculation success (in percents), so that
10253 speculative insns are scheduled.
10254 The default value is 40.
10256 @item sched-spec-state-edge-prob-cutoff
10257 The minimum probability an edge must have for the scheduler to save its
10259 The default value is 10.
10261 @item sched-mem-true-dep-cost
10262 Minimal distance (in CPU cycles) between store and load targeting same
10263 memory locations. The default value is 1.
10265 @item selsched-max-lookahead
10266 The maximum size of the lookahead window of selective scheduling. It is a
10267 depth of search for available instructions.
10268 The default value is 50.
10270 @item selsched-max-sched-times
10271 The maximum number of times that an instruction is scheduled during
10272 selective scheduling. This is the limit on the number of iterations
10273 through which the instruction may be pipelined. The default value is 2.
10275 @item selsched-max-insns-to-rename
10276 The maximum number of best instructions in the ready list that are considered
10277 for renaming in the selective scheduler. The default value is 2.
10280 The minimum value of stage count that swing modulo scheduler
10281 generates. The default value is 2.
10283 @item max-last-value-rtl
10284 The maximum size measured as number of RTLs that can be recorded in an expression
10285 in combiner for a pseudo register as last known value of that register. The default
10288 @item max-combine-insns
10289 The maximum number of instructions the RTL combiner tries to combine.
10290 The default value is 2 at @option{-Og} and 4 otherwise.
10292 @item integer-share-limit
10293 Small integer constants can use a shared data structure, reducing the
10294 compiler's memory usage and increasing its speed. This sets the maximum
10295 value of a shared integer constant. The default value is 256.
10297 @item ssp-buffer-size
10298 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10299 protection when @option{-fstack-protection} is used.
10301 @item min-size-for-stack-sharing
10302 The minimum size of variables taking part in stack slot sharing when not
10303 optimizing. The default value is 32.
10305 @item max-jump-thread-duplication-stmts
10306 Maximum number of statements allowed in a block that needs to be
10307 duplicated when threading jumps.
10309 @item max-fields-for-field-sensitive
10310 Maximum number of fields in a structure treated in
10311 a field sensitive manner during pointer analysis. The default is zero
10312 for @option{-O0} and @option{-O1},
10313 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10315 @item prefetch-latency
10316 Estimate on average number of instructions that are executed before
10317 prefetch finishes. The distance prefetched ahead is proportional
10318 to this constant. Increasing this number may also lead to less
10319 streams being prefetched (see @option{simultaneous-prefetches}).
10321 @item simultaneous-prefetches
10322 Maximum number of prefetches that can run at the same time.
10324 @item l1-cache-line-size
10325 The size of cache line in L1 cache, in bytes.
10327 @item l1-cache-size
10328 The size of L1 cache, in kilobytes.
10330 @item l2-cache-size
10331 The size of L2 cache, in kilobytes.
10333 @item min-insn-to-prefetch-ratio
10334 The minimum ratio between the number of instructions and the
10335 number of prefetches to enable prefetching in a loop.
10337 @item prefetch-min-insn-to-mem-ratio
10338 The minimum ratio between the number of instructions and the
10339 number of memory references to enable prefetching in a loop.
10341 @item use-canonical-types
10342 Whether the compiler should use the ``canonical'' type system. By
10343 default, this should always be 1, which uses a more efficient internal
10344 mechanism for comparing types in C++ and Objective-C++. However, if
10345 bugs in the canonical type system are causing compilation failures,
10346 set this value to 0 to disable canonical types.
10348 @item switch-conversion-max-branch-ratio
10349 Switch initialization conversion refuses to create arrays that are
10350 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10351 branches in the switch.
10353 @item max-partial-antic-length
10354 Maximum length of the partial antic set computed during the tree
10355 partial redundancy elimination optimization (@option{-ftree-pre}) when
10356 optimizing at @option{-O3} and above. For some sorts of source code
10357 the enhanced partial redundancy elimination optimization can run away,
10358 consuming all of the memory available on the host machine. This
10359 parameter sets a limit on the length of the sets that are computed,
10360 which prevents the runaway behavior. Setting a value of 0 for
10361 this parameter allows an unlimited set length.
10363 @item sccvn-max-scc-size
10364 Maximum size of a strongly connected component (SCC) during SCCVN
10365 processing. If this limit is hit, SCCVN processing for the whole
10366 function is not done and optimizations depending on it are
10367 disabled. The default maximum SCC size is 10000.
10369 @item sccvn-max-alias-queries-per-access
10370 Maximum number of alias-oracle queries we perform when looking for
10371 redundancies for loads and stores. If this limit is hit the search
10372 is aborted and the load or store is not considered redundant. The
10373 number of queries is algorithmically limited to the number of
10374 stores on all paths from the load to the function entry.
10375 The default maxmimum number of queries is 1000.
10377 @item ira-max-loops-num
10378 IRA uses regional register allocation by default. If a function
10379 contains more loops than the number given by this parameter, only at most
10380 the given number of the most frequently-executed loops form regions
10381 for regional register allocation. The default value of the
10384 @item ira-max-conflict-table-size
10385 Although IRA uses a sophisticated algorithm to compress the conflict
10386 table, the table can still require excessive amounts of memory for
10387 huge functions. If the conflict table for a function could be more
10388 than the size in MB given by this parameter, the register allocator
10389 instead uses a faster, simpler, and lower-quality
10390 algorithm that does not require building a pseudo-register conflict table.
10391 The default value of the parameter is 2000.
10393 @item ira-loop-reserved-regs
10394 IRA can be used to evaluate more accurate register pressure in loops
10395 for decisions to move loop invariants (see @option{-O3}). The number
10396 of available registers reserved for some other purposes is given
10397 by this parameter. The default value of the parameter is 2, which is
10398 the minimal number of registers needed by typical instructions.
10399 This value is the best found from numerous experiments.
10401 @item loop-invariant-max-bbs-in-loop
10402 Loop invariant motion can be very expensive, both in compilation time and
10403 in amount of needed compile-time memory, with very large loops. Loops
10404 with more basic blocks than this parameter won't have loop invariant
10405 motion optimization performed on them. The default value of the
10406 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10408 @item loop-max-datarefs-for-datadeps
10409 Building data dapendencies is expensive for very large loops. This
10410 parameter limits the number of data references in loops that are
10411 considered for data dependence analysis. These large loops are no
10412 handled by the optimizations using loop data dependencies.
10413 The default value is 1000.
10415 @item max-vartrack-size
10416 Sets a maximum number of hash table slots to use during variable
10417 tracking dataflow analysis of any function. If this limit is exceeded
10418 with variable tracking at assignments enabled, analysis for that
10419 function is retried without it, after removing all debug insns from
10420 the function. If the limit is exceeded even without debug insns, var
10421 tracking analysis is completely disabled for the function. Setting
10422 the parameter to zero makes it unlimited.
10424 @item max-vartrack-expr-depth
10425 Sets a maximum number of recursion levels when attempting to map
10426 variable names or debug temporaries to value expressions. This trades
10427 compilation time for more complete debug information. If this is set too
10428 low, value expressions that are available and could be represented in
10429 debug information may end up not being used; setting this higher may
10430 enable the compiler to find more complex debug expressions, but compile
10431 time and memory use may grow. The default is 12.
10433 @item min-nondebug-insn-uid
10434 Use uids starting at this parameter for nondebug insns. The range below
10435 the parameter is reserved exclusively for debug insns created by
10436 @option{-fvar-tracking-assignments}, but debug insns may get
10437 (non-overlapping) uids above it if the reserved range is exhausted.
10439 @item ipa-sra-ptr-growth-factor
10440 IPA-SRA replaces a pointer to an aggregate with one or more new
10441 parameters only when their cumulative size is less or equal to
10442 @option{ipa-sra-ptr-growth-factor} times the size of the original
10445 @item sra-max-scalarization-size-Ospeed
10446 @item sra-max-scalarization-size-Osize
10447 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10448 replace scalar parts of aggregates with uses of independent scalar
10449 variables. These parameters control the maximum size, in storage units,
10450 of aggregate which will be considered for replacement when compiling for
10452 (@option{sra-max-scalarization-size-Ospeed}) or size
10453 (@option{sra-max-scalarization-size-Osize}) respectively.
10455 @item tm-max-aggregate-size
10456 When making copies of thread-local variables in a transaction, this
10457 parameter specifies the size in bytes after which variables are
10458 saved with the logging functions as opposed to save/restore code
10459 sequence pairs. This option only applies when using
10462 @item graphite-max-nb-scop-params
10463 To avoid exponential effects in the Graphite loop transforms, the
10464 number of parameters in a Static Control Part (SCoP) is bounded. The
10465 default value is 10 parameters. A variable whose value is unknown at
10466 compilation time and defined outside a SCoP is a parameter of the SCoP.
10468 @item graphite-max-bbs-per-function
10469 To avoid exponential effects in the detection of SCoPs, the size of
10470 the functions analyzed by Graphite is bounded. The default value is
10473 @item loop-block-tile-size
10474 Loop blocking or strip mining transforms, enabled with
10475 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10476 loop in the loop nest by a given number of iterations. The strip
10477 length can be changed using the @option{loop-block-tile-size}
10478 parameter. The default value is 51 iterations.
10480 @item loop-unroll-jam-size
10481 Specify the unroll factor for the @option{-floop-unroll-and-jam}. The
10482 default value is 4.
10484 @item loop-unroll-jam-depth
10485 Specify the dimension to be unrolled (counting from the most inner loop)
10486 for the @option{-floop-unroll-and-jam}. The default value is 2.
10488 @item ipa-cp-value-list-size
10489 IPA-CP attempts to track all possible values and types passed to a function's
10490 parameter in order to propagate them and perform devirtualization.
10491 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10492 stores per one formal parameter of a function.
10494 @item ipa-cp-eval-threshold
10495 IPA-CP calculates its own score of cloning profitability heuristics
10496 and performs those cloning opportunities with scores that exceed
10497 @option{ipa-cp-eval-threshold}.
10499 @item ipa-max-agg-items
10500 IPA-CP is also capable to propagate a number of scalar values passed
10501 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10502 number of such values per one parameter.
10504 @item ipa-cp-loop-hint-bonus
10505 When IPA-CP determines that a cloning candidate would make the number
10506 of iterations of a loop known, it adds a bonus of
10507 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10510 @item ipa-cp-array-index-hint-bonus
10511 When IPA-CP determines that a cloning candidate would make the index of
10512 an array access known, it adds a bonus of
10513 @option{ipa-cp-array-index-hint-bonus} to the profitability
10514 score of the candidate.
10516 @item ipa-max-aa-steps
10517 During its analysis of function bodies, IPA-CP employs alias analysis
10518 in order to track values pointed to by function parameters. In order
10519 not spend too much time analyzing huge functions, it will give up and
10520 consider all memory clobbered after examining
10521 @option{ipa-max-aa-steps} statements modifying memory.
10523 @item lto-partitions
10524 Specify desired number of partitions produced during WHOPR compilation.
10525 The number of partitions should exceed the number of CPUs used for compilation.
10526 The default value is 32.
10528 @item lto-minpartition
10529 Size of minimal partition for WHOPR (in estimated instructions).
10530 This prevents expenses of splitting very small programs into too many
10533 @item cxx-max-namespaces-for-diagnostic-help
10534 The maximum number of namespaces to consult for suggestions when C++
10535 name lookup fails for an identifier. The default is 1000.
10537 @item sink-frequency-threshold
10538 The maximum relative execution frequency (in percents) of the target block
10539 relative to a statement's original block to allow statement sinking of a
10540 statement. Larger numbers result in more aggressive statement sinking.
10541 The default value is 75. A small positive adjustment is applied for
10542 statements with memory operands as those are even more profitable so sink.
10544 @item max-stores-to-sink
10545 The maximum number of conditional stores paires that can be sunk. Set to 0
10546 if either vectorization (@option{-ftree-vectorize}) or if-conversion
10547 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
10549 @item allow-store-data-races
10550 Allow optimizers to introduce new data races on stores.
10551 Set to 1 to allow, otherwise to 0. This option is enabled by default
10552 at optimization level @option{-Ofast}.
10554 @item case-values-threshold
10555 The smallest number of different values for which it is best to use a
10556 jump-table instead of a tree of conditional branches. If the value is
10557 0, use the default for the machine. The default is 0.
10559 @item tree-reassoc-width
10560 Set the maximum number of instructions executed in parallel in
10561 reassociated tree. This parameter overrides target dependent
10562 heuristics used by default if has non zero value.
10564 @item sched-pressure-algorithm
10565 Choose between the two available implementations of
10566 @option{-fsched-pressure}. Algorithm 1 is the original implementation
10567 and is the more likely to prevent instructions from being reordered.
10568 Algorithm 2 was designed to be a compromise between the relatively
10569 conservative approach taken by algorithm 1 and the rather aggressive
10570 approach taken by the default scheduler. It relies more heavily on
10571 having a regular register file and accurate register pressure classes.
10572 See @file{haifa-sched.c} in the GCC sources for more details.
10574 The default choice depends on the target.
10576 @item max-slsr-cand-scan
10577 Set the maximum number of existing candidates that will be considered when
10578 seeking a basis for a new straight-line strength reduction candidate.
10581 Enable buffer overflow detection for global objects. This kind
10582 of protection is enabled by default if you are using
10583 @option{-fsanitize=address} option.
10584 To disable global objects protection use @option{--param asan-globals=0}.
10587 Enable buffer overflow detection for stack objects. This kind of
10588 protection is enabled by default when using@option{-fsanitize=address}.
10589 To disable stack protection use @option{--param asan-stack=0} option.
10591 @item asan-instrument-reads
10592 Enable buffer overflow detection for memory reads. This kind of
10593 protection is enabled by default when using @option{-fsanitize=address}.
10594 To disable memory reads protection use
10595 @option{--param asan-instrument-reads=0}.
10597 @item asan-instrument-writes
10598 Enable buffer overflow detection for memory writes. This kind of
10599 protection is enabled by default when using @option{-fsanitize=address}.
10600 To disable memory writes protection use
10601 @option{--param asan-instrument-writes=0} option.
10603 @item asan-memintrin
10604 Enable detection for built-in functions. This kind of protection
10605 is enabled by default when using @option{-fsanitize=address}.
10606 To disable built-in functions protection use
10607 @option{--param asan-memintrin=0}.
10609 @item asan-use-after-return
10610 Enable detection of use-after-return. This kind of protection
10611 is enabled by default when using @option{-fsanitize=address} option.
10612 To disable use-after-return detection use
10613 @option{--param asan-use-after-return=0}.
10615 @item asan-instrumentation-with-call-threshold
10616 If number of memory accesses in function being instrumented
10617 is greater or equal to this number, use callbacks instead of inline checks.
10618 E.g. to disable inline code use
10619 @option{--param asan-instrumentation-with-call-threshold=0}.
10621 @item chkp-max-ctor-size
10622 Static constructors generated by Pointer Bounds Checker may become very
10623 large and significantly increase compile time at optimization level
10624 @option{-O1} and higher. This parameter is a maximum nubmer of statements
10625 in a single generated constructor. Default value is 5000.
10630 @node Preprocessor Options
10631 @section Options Controlling the Preprocessor
10632 @cindex preprocessor options
10633 @cindex options, preprocessor
10635 These options control the C preprocessor, which is run on each C source
10636 file before actual compilation.
10638 If you use the @option{-E} option, nothing is done except preprocessing.
10639 Some of these options make sense only together with @option{-E} because
10640 they cause the preprocessor output to be unsuitable for actual
10644 @item -Wp,@var{option}
10646 You can use @option{-Wp,@var{option}} to bypass the compiler driver
10647 and pass @var{option} directly through to the preprocessor. If
10648 @var{option} contains commas, it is split into multiple options at the
10649 commas. However, many options are modified, translated or interpreted
10650 by the compiler driver before being passed to the preprocessor, and
10651 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
10652 interface is undocumented and subject to change, so whenever possible
10653 you should avoid using @option{-Wp} and let the driver handle the
10656 @item -Xpreprocessor @var{option}
10657 @opindex Xpreprocessor
10658 Pass @var{option} as an option to the preprocessor. You can use this to
10659 supply system-specific preprocessor options that GCC does not
10662 If you want to pass an option that takes an argument, you must use
10663 @option{-Xpreprocessor} twice, once for the option and once for the argument.
10665 @item -no-integrated-cpp
10666 @opindex no-integrated-cpp
10667 Perform preprocessing as a separate pass before compilation.
10668 By default, GCC performs preprocessing as an integrated part of
10669 input tokenization and parsing.
10670 If this option is provided, the appropriate language front end
10671 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
10672 and Objective-C, respectively) is instead invoked twice,
10673 once for preprocessing only and once for actual compilation
10674 of the preprocessed input.
10675 This option may be useful in conjunction with the @option{-B} or
10676 @option{-wrapper} options to specify an alternate preprocessor or
10677 perform additional processing of the program source between
10678 normal preprocessing and compilation.
10681 @include cppopts.texi
10683 @node Assembler Options
10684 @section Passing Options to the Assembler
10686 @c prevent bad page break with this line
10687 You can pass options to the assembler.
10690 @item -Wa,@var{option}
10692 Pass @var{option} as an option to the assembler. If @var{option}
10693 contains commas, it is split into multiple options at the commas.
10695 @item -Xassembler @var{option}
10696 @opindex Xassembler
10697 Pass @var{option} as an option to the assembler. You can use this to
10698 supply system-specific assembler options that GCC does not
10701 If you want to pass an option that takes an argument, you must use
10702 @option{-Xassembler} twice, once for the option and once for the argument.
10707 @section Options for Linking
10708 @cindex link options
10709 @cindex options, linking
10711 These options come into play when the compiler links object files into
10712 an executable output file. They are meaningless if the compiler is
10713 not doing a link step.
10717 @item @var{object-file-name}
10718 A file name that does not end in a special recognized suffix is
10719 considered to name an object file or library. (Object files are
10720 distinguished from libraries by the linker according to the file
10721 contents.) If linking is done, these object files are used as input
10730 If any of these options is used, then the linker is not run, and
10731 object file names should not be used as arguments. @xref{Overall
10735 @item -l@var{library}
10736 @itemx -l @var{library}
10738 Search the library named @var{library} when linking. (The second
10739 alternative with the library as a separate argument is only for
10740 POSIX compliance and is not recommended.)
10742 It makes a difference where in the command you write this option; the
10743 linker searches and processes libraries and object files in the order they
10744 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
10745 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
10746 to functions in @samp{z}, those functions may not be loaded.
10748 The linker searches a standard list of directories for the library,
10749 which is actually a file named @file{lib@var{library}.a}. The linker
10750 then uses this file as if it had been specified precisely by name.
10752 The directories searched include several standard system directories
10753 plus any that you specify with @option{-L}.
10755 Normally the files found this way are library files---archive files
10756 whose members are object files. The linker handles an archive file by
10757 scanning through it for members which define symbols that have so far
10758 been referenced but not defined. But if the file that is found is an
10759 ordinary object file, it is linked in the usual fashion. The only
10760 difference between using an @option{-l} option and specifying a file name
10761 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
10762 and searches several directories.
10766 You need this special case of the @option{-l} option in order to
10767 link an Objective-C or Objective-C++ program.
10769 @item -nostartfiles
10770 @opindex nostartfiles
10771 Do not use the standard system startup files when linking.
10772 The standard system libraries are used normally, unless @option{-nostdlib}
10773 or @option{-nodefaultlibs} is used.
10775 @item -nodefaultlibs
10776 @opindex nodefaultlibs
10777 Do not use the standard system libraries when linking.
10778 Only the libraries you specify are passed to the linker, and options
10779 specifying linkage of the system libraries, such as @code{-static-libgcc}
10780 or @code{-shared-libgcc}, are ignored.
10781 The standard startup files are used normally, unless @option{-nostartfiles}
10784 The compiler may generate calls to @code{memcmp},
10785 @code{memset}, @code{memcpy} and @code{memmove}.
10786 These entries are usually resolved by entries in
10787 libc. These entry points should be supplied through some other
10788 mechanism when this option is specified.
10792 Do not use the standard system startup files or libraries when linking.
10793 No startup files and only the libraries you specify are passed to
10794 the linker, and options specifying linkage of the system libraries, such as
10795 @code{-static-libgcc} or @code{-shared-libgcc}, are ignored.
10797 The compiler may generate calls to @code{memcmp}, @code{memset},
10798 @code{memcpy} and @code{memmove}.
10799 These entries are usually resolved by entries in
10800 libc. These entry points should be supplied through some other
10801 mechanism when this option is specified.
10803 @cindex @option{-lgcc}, use with @option{-nostdlib}
10804 @cindex @option{-nostdlib} and unresolved references
10805 @cindex unresolved references and @option{-nostdlib}
10806 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
10807 @cindex @option{-nodefaultlibs} and unresolved references
10808 @cindex unresolved references and @option{-nodefaultlibs}
10809 One of the standard libraries bypassed by @option{-nostdlib} and
10810 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
10811 which GCC uses to overcome shortcomings of particular machines, or special
10812 needs for some languages.
10813 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
10814 Collection (GCC) Internals},
10815 for more discussion of @file{libgcc.a}.)
10816 In most cases, you need @file{libgcc.a} even when you want to avoid
10817 other standard libraries. In other words, when you specify @option{-nostdlib}
10818 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
10819 This ensures that you have no unresolved references to internal GCC
10820 library subroutines.
10821 (An example of such an internal subroutine is @samp{__main}, used to ensure C++
10822 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
10823 GNU Compiler Collection (GCC) Internals}.)
10827 Produce a position independent executable on targets that support it.
10828 For predictable results, you must also specify the same set of options
10829 used for compilation (@option{-fpie}, @option{-fPIE},
10830 or model suboptions) when you specify this linker option.
10834 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
10835 that support it. This instructs the linker to add all symbols, not
10836 only used ones, to the dynamic symbol table. This option is needed
10837 for some uses of @code{dlopen} or to allow obtaining backtraces
10838 from within a program.
10842 Remove all symbol table and relocation information from the executable.
10846 On systems that support dynamic linking, this prevents linking with the shared
10847 libraries. On other systems, this option has no effect.
10851 Produce a shared object which can then be linked with other objects to
10852 form an executable. Not all systems support this option. For predictable
10853 results, you must also specify the same set of options used for compilation
10854 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
10855 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
10856 needs to build supplementary stub code for constructors to work. On
10857 multi-libbed systems, @samp{gcc -shared} must select the correct support
10858 libraries to link against. Failing to supply the correct flags may lead
10859 to subtle defects. Supplying them in cases where they are not necessary
10862 @item -shared-libgcc
10863 @itemx -static-libgcc
10864 @opindex shared-libgcc
10865 @opindex static-libgcc
10866 On systems that provide @file{libgcc} as a shared library, these options
10867 force the use of either the shared or static version, respectively.
10868 If no shared version of @file{libgcc} was built when the compiler was
10869 configured, these options have no effect.
10871 There are several situations in which an application should use the
10872 shared @file{libgcc} instead of the static version. The most common
10873 of these is when the application wishes to throw and catch exceptions
10874 across different shared libraries. In that case, each of the libraries
10875 as well as the application itself should use the shared @file{libgcc}.
10877 Therefore, the G++ and GCJ drivers automatically add
10878 @option{-shared-libgcc} whenever you build a shared library or a main
10879 executable, because C++ and Java programs typically use exceptions, so
10880 this is the right thing to do.
10882 If, instead, you use the GCC driver to create shared libraries, you may
10883 find that they are not always linked with the shared @file{libgcc}.
10884 If GCC finds, at its configuration time, that you have a non-GNU linker
10885 or a GNU linker that does not support option @option{--eh-frame-hdr},
10886 it links the shared version of @file{libgcc} into shared libraries
10887 by default. Otherwise, it takes advantage of the linker and optimizes
10888 away the linking with the shared version of @file{libgcc}, linking with
10889 the static version of libgcc by default. This allows exceptions to
10890 propagate through such shared libraries, without incurring relocation
10891 costs at library load time.
10893 However, if a library or main executable is supposed to throw or catch
10894 exceptions, you must link it using the G++ or GCJ driver, as appropriate
10895 for the languages used in the program, or using the option
10896 @option{-shared-libgcc}, such that it is linked with the shared
10899 @item -static-libasan
10900 @opindex static-libasan
10901 When the @option{-fsanitize=address} option is used to link a program,
10902 the GCC driver automatically links against @option{libasan}. If
10903 @file{libasan} is available as a shared library, and the @option{-static}
10904 option is not used, then this links against the shared version of
10905 @file{libasan}. The @option{-static-libasan} option directs the GCC
10906 driver to link @file{libasan} statically, without necessarily linking
10907 other libraries statically.
10909 @item -static-libtsan
10910 @opindex static-libtsan
10911 When the @option{-fsanitize=thread} option is used to link a program,
10912 the GCC driver automatically links against @option{libtsan}. If
10913 @file{libtsan} is available as a shared library, and the @option{-static}
10914 option is not used, then this links against the shared version of
10915 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
10916 driver to link @file{libtsan} statically, without necessarily linking
10917 other libraries statically.
10919 @item -static-liblsan
10920 @opindex static-liblsan
10921 When the @option{-fsanitize=leak} option is used to link a program,
10922 the GCC driver automatically links against @option{liblsan}. If
10923 @file{liblsan} is available as a shared library, and the @option{-static}
10924 option is not used, then this links against the shared version of
10925 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
10926 driver to link @file{liblsan} statically, without necessarily linking
10927 other libraries statically.
10929 @item -static-libubsan
10930 @opindex static-libubsan
10931 When the @option{-fsanitize=undefined} option is used to link a program,
10932 the GCC driver automatically links against @option{libubsan}. If
10933 @file{libubsan} is available as a shared library, and the @option{-static}
10934 option is not used, then this links against the shared version of
10935 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
10936 driver to link @file{libubsan} statically, without necessarily linking
10937 other libraries statically.
10939 @item -static-libstdc++
10940 @opindex static-libstdc++
10941 When the @command{g++} program is used to link a C++ program, it
10942 normally automatically links against @option{libstdc++}. If
10943 @file{libstdc++} is available as a shared library, and the
10944 @option{-static} option is not used, then this links against the
10945 shared version of @file{libstdc++}. That is normally fine. However, it
10946 is sometimes useful to freeze the version of @file{libstdc++} used by
10947 the program without going all the way to a fully static link. The
10948 @option{-static-libstdc++} option directs the @command{g++} driver to
10949 link @file{libstdc++} statically, without necessarily linking other
10950 libraries statically.
10954 Bind references to global symbols when building a shared object. Warn
10955 about any unresolved references (unless overridden by the link editor
10956 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
10959 @item -T @var{script}
10961 @cindex linker script
10962 Use @var{script} as the linker script. This option is supported by most
10963 systems using the GNU linker. On some targets, such as bare-board
10964 targets without an operating system, the @option{-T} option may be required
10965 when linking to avoid references to undefined symbols.
10967 @item -Xlinker @var{option}
10969 Pass @var{option} as an option to the linker. You can use this to
10970 supply system-specific linker options that GCC does not recognize.
10972 If you want to pass an option that takes a separate argument, you must use
10973 @option{-Xlinker} twice, once for the option and once for the argument.
10974 For example, to pass @option{-assert definitions}, you must write
10975 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
10976 @option{-Xlinker "-assert definitions"}, because this passes the entire
10977 string as a single argument, which is not what the linker expects.
10979 When using the GNU linker, it is usually more convenient to pass
10980 arguments to linker options using the @option{@var{option}=@var{value}}
10981 syntax than as separate arguments. For example, you can specify
10982 @option{-Xlinker -Map=output.map} rather than
10983 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
10984 this syntax for command-line options.
10986 @item -Wl,@var{option}
10988 Pass @var{option} as an option to the linker. If @var{option} contains
10989 commas, it is split into multiple options at the commas. You can use this
10990 syntax to pass an argument to the option.
10991 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
10992 linker. When using the GNU linker, you can also get the same effect with
10993 @option{-Wl,-Map=output.map}.
10995 @item -u @var{symbol}
10997 Pretend the symbol @var{symbol} is undefined, to force linking of
10998 library modules to define it. You can use @option{-u} multiple times with
10999 different symbols to force loading of additional library modules.
11001 @item -z @var{keyword}
11003 @option{-z} is passed directly on to the linker along with the keyword
11004 @var{keyword}. See the section in the documentation of your linker for
11005 permitted values and their meanings.
11008 @node Directory Options
11009 @section Options for Directory Search
11010 @cindex directory options
11011 @cindex options, directory search
11012 @cindex search path
11014 These options specify directories to search for header files, for
11015 libraries and for parts of the compiler:
11020 Add the directory @var{dir} to the head of the list of directories to be
11021 searched for header files. This can be used to override a system header
11022 file, substituting your own version, since these directories are
11023 searched before the system header file directories. However, you should
11024 not use this option to add directories that contain vendor-supplied
11025 system header files (use @option{-isystem} for that). If you use more than
11026 one @option{-I} option, the directories are scanned in left-to-right
11027 order; the standard system directories come after.
11029 If a standard system include directory, or a directory specified with
11030 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
11031 option is ignored. The directory is still searched but as a
11032 system directory at its normal position in the system include chain.
11033 This is to ensure that GCC's procedure to fix buggy system headers and
11034 the ordering for the @code{include_next} directive are not inadvertently changed.
11035 If you really need to change the search order for system directories,
11036 use the @option{-nostdinc} and/or @option{-isystem} options.
11038 @item -iplugindir=@var{dir}
11039 @opindex iplugindir=
11040 Set the directory to search for plugins that are passed
11041 by @option{-fplugin=@var{name}} instead of
11042 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
11043 to be used by the user, but only passed by the driver.
11045 @item -iquote@var{dir}
11047 Add the directory @var{dir} to the head of the list of directories to
11048 be searched for header files only for the case of @samp{#include
11049 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
11050 otherwise just like @option{-I}.
11054 Add directory @var{dir} to the list of directories to be searched
11057 @item -B@var{prefix}
11059 This option specifies where to find the executables, libraries,
11060 include files, and data files of the compiler itself.
11062 The compiler driver program runs one or more of the subprograms
11063 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
11064 @var{prefix} as a prefix for each program it tries to run, both with and
11065 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
11067 For each subprogram to be run, the compiler driver first tries the
11068 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
11069 is not specified, the driver tries two standard prefixes,
11070 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
11071 those results in a file name that is found, the unmodified program
11072 name is searched for using the directories specified in your
11073 @env{PATH} environment variable.
11075 The compiler checks to see if the path provided by the @option{-B}
11076 refers to a directory, and if necessary it adds a directory
11077 separator character at the end of the path.
11079 @option{-B} prefixes that effectively specify directory names also apply
11080 to libraries in the linker, because the compiler translates these
11081 options into @option{-L} options for the linker. They also apply to
11082 include files in the preprocessor, because the compiler translates these
11083 options into @option{-isystem} options for the preprocessor. In this case,
11084 the compiler appends @samp{include} to the prefix.
11086 The runtime support file @file{libgcc.a} can also be searched for using
11087 the @option{-B} prefix, if needed. If it is not found there, the two
11088 standard prefixes above are tried, and that is all. The file is left
11089 out of the link if it is not found by those means.
11091 Another way to specify a prefix much like the @option{-B} prefix is to use
11092 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
11095 As a special kludge, if the path provided by @option{-B} is
11096 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
11097 9, then it is replaced by @file{[dir/]include}. This is to help
11098 with boot-strapping the compiler.
11100 @item -specs=@var{file}
11102 Process @var{file} after the compiler reads in the standard @file{specs}
11103 file, in order to override the defaults which the @command{gcc} driver
11104 program uses when determining what switches to pass to @command{cc1},
11105 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
11106 @option{-specs=@var{file}} can be specified on the command line, and they
11107 are processed in order, from left to right.
11109 @item --sysroot=@var{dir}
11111 Use @var{dir} as the logical root directory for headers and libraries.
11112 For example, if the compiler normally searches for headers in
11113 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
11114 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
11116 If you use both this option and the @option{-isysroot} option, then
11117 the @option{--sysroot} option applies to libraries, but the
11118 @option{-isysroot} option applies to header files.
11120 The GNU linker (beginning with version 2.16) has the necessary support
11121 for this option. If your linker does not support this option, the
11122 header file aspect of @option{--sysroot} still works, but the
11123 library aspect does not.
11125 @item --no-sysroot-suffix
11126 @opindex no-sysroot-suffix
11127 For some targets, a suffix is added to the root directory specified
11128 with @option{--sysroot}, depending on the other options used, so that
11129 headers may for example be found in
11130 @file{@var{dir}/@var{suffix}/usr/include} instead of
11131 @file{@var{dir}/usr/include}. This option disables the addition of
11136 This option has been deprecated. Please use @option{-iquote} instead for
11137 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
11138 Any directories you specify with @option{-I} options before the @option{-I-}
11139 option are searched only for the case of @samp{#include "@var{file}"};
11140 they are not searched for @samp{#include <@var{file}>}.
11142 If additional directories are specified with @option{-I} options after
11143 the @option{-I-}, these directories are searched for all @samp{#include}
11144 directives. (Ordinarily @emph{all} @option{-I} directories are used
11147 In addition, the @option{-I-} option inhibits the use of the current
11148 directory (where the current input file came from) as the first search
11149 directory for @samp{#include "@var{file}"}. There is no way to
11150 override this effect of @option{-I-}. With @option{-I.} you can specify
11151 searching the directory that is current when the compiler is
11152 invoked. That is not exactly the same as what the preprocessor does
11153 by default, but it is often satisfactory.
11155 @option{-I-} does not inhibit the use of the standard system directories
11156 for header files. Thus, @option{-I-} and @option{-nostdinc} are
11163 @section Specifying subprocesses and the switches to pass to them
11166 @command{gcc} is a driver program. It performs its job by invoking a
11167 sequence of other programs to do the work of compiling, assembling and
11168 linking. GCC interprets its command-line parameters and uses these to
11169 deduce which programs it should invoke, and which command-line options
11170 it ought to place on their command lines. This behavior is controlled
11171 by @dfn{spec strings}. In most cases there is one spec string for each
11172 program that GCC can invoke, but a few programs have multiple spec
11173 strings to control their behavior. The spec strings built into GCC can
11174 be overridden by using the @option{-specs=} command-line switch to specify
11177 @dfn{Spec files} are plaintext files that are used to construct spec
11178 strings. They consist of a sequence of directives separated by blank
11179 lines. The type of directive is determined by the first non-whitespace
11180 character on the line, which can be one of the following:
11183 @item %@var{command}
11184 Issues a @var{command} to the spec file processor. The commands that can
11188 @item %include <@var{file}>
11189 @cindex @code{%include}
11190 Search for @var{file} and insert its text at the current point in the
11193 @item %include_noerr <@var{file}>
11194 @cindex @code{%include_noerr}
11195 Just like @samp{%include}, but do not generate an error message if the include
11196 file cannot be found.
11198 @item %rename @var{old_name} @var{new_name}
11199 @cindex @code{%rename}
11200 Rename the spec string @var{old_name} to @var{new_name}.
11204 @item *[@var{spec_name}]:
11205 This tells the compiler to create, override or delete the named spec
11206 string. All lines after this directive up to the next directive or
11207 blank line are considered to be the text for the spec string. If this
11208 results in an empty string then the spec is deleted. (Or, if the
11209 spec did not exist, then nothing happens.) Otherwise, if the spec
11210 does not currently exist a new spec is created. If the spec does
11211 exist then its contents are overridden by the text of this
11212 directive, unless the first character of that text is the @samp{+}
11213 character, in which case the text is appended to the spec.
11215 @item [@var{suffix}]:
11216 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
11217 and up to the next directive or blank line are considered to make up the
11218 spec string for the indicated suffix. When the compiler encounters an
11219 input file with the named suffix, it processes the spec string in
11220 order to work out how to compile that file. For example:
11224 z-compile -input %i
11227 This says that any input file whose name ends in @samp{.ZZ} should be
11228 passed to the program @samp{z-compile}, which should be invoked with the
11229 command-line switch @option{-input} and with the result of performing the
11230 @samp{%i} substitution. (See below.)
11232 As an alternative to providing a spec string, the text following a
11233 suffix directive can be one of the following:
11236 @item @@@var{language}
11237 This says that the suffix is an alias for a known @var{language}. This is
11238 similar to using the @option{-x} command-line switch to GCC to specify a
11239 language explicitly. For example:
11246 Says that .ZZ files are, in fact, C++ source files.
11249 This causes an error messages saying:
11252 @var{name} compiler not installed on this system.
11256 GCC already has an extensive list of suffixes built into it.
11257 This directive adds an entry to the end of the list of suffixes, but
11258 since the list is searched from the end backwards, it is effectively
11259 possible to override earlier entries using this technique.
11263 GCC has the following spec strings built into it. Spec files can
11264 override these strings or create their own. Note that individual
11265 targets can also add their own spec strings to this list.
11268 asm Options to pass to the assembler
11269 asm_final Options to pass to the assembler post-processor
11270 cpp Options to pass to the C preprocessor
11271 cc1 Options to pass to the C compiler
11272 cc1plus Options to pass to the C++ compiler
11273 endfile Object files to include at the end of the link
11274 link Options to pass to the linker
11275 lib Libraries to include on the command line to the linker
11276 libgcc Decides which GCC support library to pass to the linker
11277 linker Sets the name of the linker
11278 predefines Defines to be passed to the C preprocessor
11279 signed_char Defines to pass to CPP to say whether @code{char} is signed
11281 startfile Object files to include at the start of the link
11284 Here is a small example of a spec file:
11287 %rename lib old_lib
11290 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
11293 This example renames the spec called @samp{lib} to @samp{old_lib} and
11294 then overrides the previous definition of @samp{lib} with a new one.
11295 The new definition adds in some extra command-line options before
11296 including the text of the old definition.
11298 @dfn{Spec strings} are a list of command-line options to be passed to their
11299 corresponding program. In addition, the spec strings can contain
11300 @samp{%}-prefixed sequences to substitute variable text or to
11301 conditionally insert text into the command line. Using these constructs
11302 it is possible to generate quite complex command lines.
11304 Here is a table of all defined @samp{%}-sequences for spec
11305 strings. Note that spaces are not generated automatically around the
11306 results of expanding these sequences. Therefore you can concatenate them
11307 together or combine them with constant text in a single argument.
11311 Substitute one @samp{%} into the program name or argument.
11314 Substitute the name of the input file being processed.
11317 Substitute the basename of the input file being processed.
11318 This is the substring up to (and not including) the last period
11319 and not including the directory.
11322 This is the same as @samp{%b}, but include the file suffix (text after
11326 Marks the argument containing or following the @samp{%d} as a
11327 temporary file name, so that that file is deleted if GCC exits
11328 successfully. Unlike @samp{%g}, this contributes no text to the
11331 @item %g@var{suffix}
11332 Substitute a file name that has suffix @var{suffix} and is chosen
11333 once per compilation, and mark the argument in the same way as
11334 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
11335 name is now chosen in a way that is hard to predict even when previously
11336 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
11337 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
11338 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
11339 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
11340 was simply substituted with a file name chosen once per compilation,
11341 without regard to any appended suffix (which was therefore treated
11342 just like ordinary text), making such attacks more likely to succeed.
11344 @item %u@var{suffix}
11345 Like @samp{%g}, but generates a new temporary file name
11346 each time it appears instead of once per compilation.
11348 @item %U@var{suffix}
11349 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
11350 new one if there is no such last file name. In the absence of any
11351 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
11352 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
11353 involves the generation of two distinct file names, one
11354 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
11355 simply substituted with a file name chosen for the previous @samp{%u},
11356 without regard to any appended suffix.
11358 @item %j@var{suffix}
11359 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
11360 writable, and if @option{-save-temps} is not used;
11361 otherwise, substitute the name
11362 of a temporary file, just like @samp{%u}. This temporary file is not
11363 meant for communication between processes, but rather as a junk
11364 disposal mechanism.
11366 @item %|@var{suffix}
11367 @itemx %m@var{suffix}
11368 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
11369 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
11370 all. These are the two most common ways to instruct a program that it
11371 should read from standard input or write to standard output. If you
11372 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
11373 construct: see for example @file{f/lang-specs.h}.
11375 @item %.@var{SUFFIX}
11376 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
11377 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
11378 terminated by the next space or %.
11381 Marks the argument containing or following the @samp{%w} as the
11382 designated output file of this compilation. This puts the argument
11383 into the sequence of arguments that @samp{%o} substitutes.
11386 Substitutes the names of all the output files, with spaces
11387 automatically placed around them. You should write spaces
11388 around the @samp{%o} as well or the results are undefined.
11389 @samp{%o} is for use in the specs for running the linker.
11390 Input files whose names have no recognized suffix are not compiled
11391 at all, but they are included among the output files, so they are
11395 Substitutes the suffix for object files. Note that this is
11396 handled specially when it immediately follows @samp{%g, %u, or %U},
11397 because of the need for those to form complete file names. The
11398 handling is such that @samp{%O} is treated exactly as if it had already
11399 been substituted, except that @samp{%g, %u, and %U} do not currently
11400 support additional @var{suffix} characters following @samp{%O} as they do
11401 following, for example, @samp{.o}.
11404 Substitutes the standard macro predefinitions for the
11405 current target machine. Use this when running @code{cpp}.
11408 Like @samp{%p}, but puts @samp{__} before and after the name of each
11409 predefined macro, except for macros that start with @samp{__} or with
11410 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
11414 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
11415 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
11416 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
11417 and @option{-imultilib} as necessary.
11420 Current argument is the name of a library or startup file of some sort.
11421 Search for that file in a standard list of directories and substitute
11422 the full name found. The current working directory is included in the
11423 list of directories scanned.
11426 Current argument is the name of a linker script. Search for that file
11427 in the current list of directories to scan for libraries. If the file
11428 is located insert a @option{--script} option into the command line
11429 followed by the full path name found. If the file is not found then
11430 generate an error message. Note: the current working directory is not
11434 Print @var{str} as an error message. @var{str} is terminated by a newline.
11435 Use this when inconsistent options are detected.
11437 @item %(@var{name})
11438 Substitute the contents of spec string @var{name} at this point.
11440 @item %x@{@var{option}@}
11441 Accumulate an option for @samp{%X}.
11444 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
11448 Output the accumulated assembler options specified by @option{-Wa}.
11451 Output the accumulated preprocessor options specified by @option{-Wp}.
11454 Process the @code{asm} spec. This is used to compute the
11455 switches to be passed to the assembler.
11458 Process the @code{asm_final} spec. This is a spec string for
11459 passing switches to an assembler post-processor, if such a program is
11463 Process the @code{link} spec. This is the spec for computing the
11464 command line passed to the linker. Typically it makes use of the
11465 @samp{%L %G %S %D and %E} sequences.
11468 Dump out a @option{-L} option for each directory that GCC believes might
11469 contain startup files. If the target supports multilibs then the
11470 current multilib directory is prepended to each of these paths.
11473 Process the @code{lib} spec. This is a spec string for deciding which
11474 libraries are included on the command line to the linker.
11477 Process the @code{libgcc} spec. This is a spec string for deciding
11478 which GCC support library is included on the command line to the linker.
11481 Process the @code{startfile} spec. This is a spec for deciding which
11482 object files are the first ones passed to the linker. Typically
11483 this might be a file named @file{crt0.o}.
11486 Process the @code{endfile} spec. This is a spec string that specifies
11487 the last object files that are passed to the linker.
11490 Process the @code{cpp} spec. This is used to construct the arguments
11491 to be passed to the C preprocessor.
11494 Process the @code{cc1} spec. This is used to construct the options to be
11495 passed to the actual C compiler (@samp{cc1}).
11498 Process the @code{cc1plus} spec. This is used to construct the options to be
11499 passed to the actual C++ compiler (@samp{cc1plus}).
11502 Substitute the variable part of a matched option. See below.
11503 Note that each comma in the substituted string is replaced by
11507 Remove all occurrences of @code{-S} from the command line. Note---this
11508 command is position dependent. @samp{%} commands in the spec string
11509 before this one see @code{-S}, @samp{%} commands in the spec string
11510 after this one do not.
11512 @item %:@var{function}(@var{args})
11513 Call the named function @var{function}, passing it @var{args}.
11514 @var{args} is first processed as a nested spec string, then split
11515 into an argument vector in the usual fashion. The function returns
11516 a string which is processed as if it had appeared literally as part
11517 of the current spec.
11519 The following built-in spec functions are provided:
11522 @item @code{getenv}
11523 The @code{getenv} spec function takes two arguments: an environment
11524 variable name and a string. If the environment variable is not
11525 defined, a fatal error is issued. Otherwise, the return value is the
11526 value of the environment variable concatenated with the string. For
11527 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
11530 %:getenv(TOPDIR /include)
11533 expands to @file{/path/to/top/include}.
11535 @item @code{if-exists}
11536 The @code{if-exists} spec function takes one argument, an absolute
11537 pathname to a file. If the file exists, @code{if-exists} returns the
11538 pathname. Here is a small example of its usage:
11542 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
11545 @item @code{if-exists-else}
11546 The @code{if-exists-else} spec function is similar to the @code{if-exists}
11547 spec function, except that it takes two arguments. The first argument is
11548 an absolute pathname to a file. If the file exists, @code{if-exists-else}
11549 returns the pathname. If it does not exist, it returns the second argument.
11550 This way, @code{if-exists-else} can be used to select one file or another,
11551 based on the existence of the first. Here is a small example of its usage:
11555 crt0%O%s %:if-exists(crti%O%s) \
11556 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
11559 @item @code{replace-outfile}
11560 The @code{replace-outfile} spec function takes two arguments. It looks for the
11561 first argument in the outfiles array and replaces it with the second argument. Here
11562 is a small example of its usage:
11565 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
11568 @item @code{remove-outfile}
11569 The @code{remove-outfile} spec function takes one argument. It looks for the
11570 first argument in the outfiles array and removes it. Here is a small example
11574 %:remove-outfile(-lm)
11577 @item @code{pass-through-libs}
11578 The @code{pass-through-libs} spec function takes any number of arguments. It
11579 finds any @option{-l} options and any non-options ending in @file{.a} (which it
11580 assumes are the names of linker input library archive files) and returns a
11581 result containing all the found arguments each prepended by
11582 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
11583 intended to be passed to the LTO linker plugin.
11586 %:pass-through-libs(%G %L %G)
11589 @item @code{print-asm-header}
11590 The @code{print-asm-header} function takes no arguments and simply
11591 prints a banner like:
11597 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
11600 It is used to separate compiler options from assembler options
11601 in the @option{--target-help} output.
11604 @item %@{@code{S}@}
11605 Substitutes the @code{-S} switch, if that switch is given to GCC@.
11606 If that switch is not specified, this substitutes nothing. Note that
11607 the leading dash is omitted when specifying this option, and it is
11608 automatically inserted if the substitution is performed. Thus the spec
11609 string @samp{%@{foo@}} matches the command-line option @option{-foo}
11610 and outputs the command-line option @option{-foo}.
11612 @item %W@{@code{S}@}
11613 Like %@{@code{S}@} but mark last argument supplied within as a file to be
11614 deleted on failure.
11616 @item %@{@code{S}*@}
11617 Substitutes all the switches specified to GCC whose names start
11618 with @code{-S}, but which also take an argument. This is used for
11619 switches like @option{-o}, @option{-D}, @option{-I}, etc.
11620 GCC considers @option{-o foo} as being
11621 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
11622 text, including the space. Thus two arguments are generated.
11624 @item %@{@code{S}*&@code{T}*@}
11625 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
11626 (the order of @code{S} and @code{T} in the spec is not significant).
11627 There can be any number of ampersand-separated variables; for each the
11628 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
11630 @item %@{@code{S}:@code{X}@}
11631 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
11633 @item %@{!@code{S}:@code{X}@}
11634 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
11636 @item %@{@code{S}*:@code{X}@}
11637 Substitutes @code{X} if one or more switches whose names start with
11638 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
11639 once, no matter how many such switches appeared. However, if @code{%*}
11640 appears somewhere in @code{X}, then @code{X} is substituted once
11641 for each matching switch, with the @code{%*} replaced by the part of
11642 that switch matching the @code{*}.
11644 If @code{%*} appears as the last part of a spec sequence then a space
11645 will be added after the end of the last substitution. If there is more
11646 text in the sequence however then a space will not be generated. This
11647 allows the @code{%*} substitution to be used as part of a larger
11648 string. For example, a spec string like this:
11651 %@{mcu=*:--script=%*/memory.ld@}
11654 when matching an option like @code{-mcu=newchip} will produce:
11657 --script=newchip/memory.ld
11660 @item %@{.@code{S}:@code{X}@}
11661 Substitutes @code{X}, if processing a file with suffix @code{S}.
11663 @item %@{!.@code{S}:@code{X}@}
11664 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
11666 @item %@{,@code{S}:@code{X}@}
11667 Substitutes @code{X}, if processing a file for language @code{S}.
11669 @item %@{!,@code{S}:@code{X}@}
11670 Substitutes @code{X}, if not processing a file for language @code{S}.
11672 @item %@{@code{S}|@code{P}:@code{X}@}
11673 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
11674 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
11675 @code{*} sequences as well, although they have a stronger binding than
11676 the @samp{|}. If @code{%*} appears in @code{X}, all of the
11677 alternatives must be starred, and only the first matching alternative
11680 For example, a spec string like this:
11683 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
11687 outputs the following command-line options from the following input
11688 command-line options:
11693 -d fred.c -foo -baz -boggle
11694 -d jim.d -bar -baz -boggle
11697 @item %@{S:X; T:Y; :D@}
11699 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
11700 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
11701 be as many clauses as you need. This may be combined with @code{.},
11702 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
11707 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
11708 construct may contain other nested @samp{%} constructs or spaces, or
11709 even newlines. They are processed as usual, as described above.
11710 Trailing white space in @code{X} is ignored. White space may also
11711 appear anywhere on the left side of the colon in these constructs,
11712 except between @code{.} or @code{*} and the corresponding word.
11714 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
11715 handled specifically in these constructs. If another value of
11716 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
11717 @option{-W} switch is found later in the command line, the earlier
11718 switch value is ignored, except with @{@code{S}*@} where @code{S} is
11719 just one letter, which passes all matching options.
11721 The character @samp{|} at the beginning of the predicate text is used to
11722 indicate that a command should be piped to the following command, but
11723 only if @option{-pipe} is specified.
11725 It is built into GCC which switches take arguments and which do not.
11726 (You might think it would be useful to generalize this to allow each
11727 compiler's spec to say which switches take arguments. But this cannot
11728 be done in a consistent fashion. GCC cannot even decide which input
11729 files have been specified without knowing which switches take arguments,
11730 and it must know which input files to compile in order to tell which
11733 GCC also knows implicitly that arguments starting in @option{-l} are to be
11734 treated as compiler output files, and passed to the linker in their
11735 proper position among the other output files.
11737 @c man begin OPTIONS
11739 @node Target Options
11740 @section Specifying Target Machine and Compiler Version
11741 @cindex target options
11742 @cindex cross compiling
11743 @cindex specifying machine version
11744 @cindex specifying compiler version and target machine
11745 @cindex compiler version, specifying
11746 @cindex target machine, specifying
11748 The usual way to run GCC is to run the executable called @command{gcc}, or
11749 @command{@var{machine}-gcc} when cross-compiling, or
11750 @command{@var{machine}-gcc-@var{version}} to run a version other than the
11751 one that was installed last.
11753 @node Submodel Options
11754 @section Hardware Models and Configurations
11755 @cindex submodel options
11756 @cindex specifying hardware config
11757 @cindex hardware models and configurations, specifying
11758 @cindex machine dependent options
11760 Each target machine types can have its own
11761 special options, starting with @samp{-m}, to choose among various
11762 hardware models or configurations---for example, 68010 vs 68020,
11763 floating coprocessor or none. A single installed version of the
11764 compiler can compile for any model or configuration, according to the
11767 Some configurations of the compiler also support additional special
11768 options, usually for compatibility with other compilers on the same
11771 @c This list is ordered alphanumerically by subsection name.
11772 @c It should be the same order and spelling as these options are listed
11773 @c in Machine Dependent Options
11776 * AArch64 Options::
11777 * Adapteva Epiphany Options::
11781 * Blackfin Options::
11786 * DEC Alpha Options::
11789 * GNU/Linux Options::
11792 * i386 and x86-64 Options::
11793 * i386 and x86-64 Windows Options::
11801 * MicroBlaze Options::
11804 * MN10300 Options::
11808 * Nios II Options::
11810 * picoChip Options::
11811 * PowerPC Options::
11813 * RS/6000 and PowerPC Options::
11815 * S/390 and zSeries Options::
11818 * Solaris 2 Options::
11821 * System V Options::
11822 * TILE-Gx Options::
11823 * TILEPro Options::
11827 * VxWorks Options::
11829 * Xstormy16 Options::
11831 * zSeries Options::
11834 @node AArch64 Options
11835 @subsection AArch64 Options
11836 @cindex AArch64 Options
11838 These options are defined for AArch64 implementations:
11842 @item -mabi=@var{name}
11844 Generate code for the specified data model. Permissible values
11845 are @samp{ilp32} for SysV-like data model where int, long int and pointer
11846 are 32-bit, and @samp{lp64} for SysV-like data model where int is 32-bit,
11847 but long int and pointer are 64-bit.
11849 The default depends on the specific target configuration. Note that
11850 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
11851 entire program with the same ABI, and link with a compatible set of libraries.
11854 @opindex mbig-endian
11855 Generate big-endian code. This is the default when GCC is configured for an
11856 @samp{aarch64_be-*-*} target.
11858 @item -mgeneral-regs-only
11859 @opindex mgeneral-regs-only
11860 Generate code which uses only the general registers.
11862 @item -mlittle-endian
11863 @opindex mlittle-endian
11864 Generate little-endian code. This is the default when GCC is configured for an
11865 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
11867 @item -mcmodel=tiny
11868 @opindex mcmodel=tiny
11869 Generate code for the tiny code model. The program and its statically defined
11870 symbols must be within 1GB of each other. Pointers are 64 bits. Programs can
11871 be statically or dynamically linked. This model is not fully implemented and
11872 mostly treated as @samp{small}.
11874 @item -mcmodel=small
11875 @opindex mcmodel=small
11876 Generate code for the small code model. The program and its statically defined
11877 symbols must be within 4GB of each other. Pointers are 64 bits. Programs can
11878 be statically or dynamically linked. This is the default code model.
11880 @item -mcmodel=large
11881 @opindex mcmodel=large
11882 Generate code for the large code model. This makes no assumptions about
11883 addresses and sizes of sections. Pointers are 64 bits. Programs can be
11884 statically linked only.
11886 @item -mstrict-align
11887 @opindex mstrict-align
11888 Do not assume that unaligned memory references will be handled by the system.
11890 @item -momit-leaf-frame-pointer
11891 @itemx -mno-omit-leaf-frame-pointer
11892 @opindex momit-leaf-frame-pointer
11893 @opindex mno-omit-leaf-frame-pointer
11894 Omit or keep the frame pointer in leaf functions. The former behaviour is the
11897 @item -mtls-dialect=desc
11898 @opindex mtls-dialect=desc
11899 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
11900 of TLS variables. This is the default.
11902 @item -mtls-dialect=traditional
11903 @opindex mtls-dialect=traditional
11904 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
11907 @item -mfix-cortex-a53-835769
11908 @itemx -mno-fix-cortex-a53-835769
11909 @opindex -mfix-cortex-a53-835769
11910 @opindex -mno-fix-cortex-a53-835769
11911 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
11912 This will involve inserting a NOP instruction between memory instructions and
11913 64-bit integer multiply-accumulate instructions.
11915 @item -march=@var{name}
11917 Specify the name of the target architecture, optionally suffixed by one or
11918 more feature modifiers. This option has the form
11919 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
11920 only permissible value for @var{arch} is @samp{armv8-a}. The permissible
11921 values for @var{feature} are documented in the sub-section below.
11923 Where conflicting feature modifiers are specified, the right-most feature is
11926 GCC uses this name to determine what kind of instructions it can emit when
11927 generating assembly code.
11929 Where @option{-march} is specified without either of @option{-mtune}
11930 or @option{-mcpu} also being specified, the code will be tuned to perform
11931 well across a range of target processors implementing the target
11934 @item -mtune=@var{name}
11936 Specify the name of the target processor for which GCC should tune the
11937 performance of the code. Permissible values for this option are:
11938 @samp{generic}, @samp{cortex-a53}, @samp{cortex-a57}, @samp{thunderx}.
11940 Additionally, this option can specify that GCC should tune the performance
11941 of the code for a big.LITTLE system. The only permissible value is
11942 @samp{cortex-a57.cortex-a53}.
11944 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
11945 are specified, the code will be tuned to perform well across a range
11946 of target processors.
11948 This option cannot be suffixed by feature modifiers.
11950 @item -mcpu=@var{name}
11952 Specify the name of the target processor, optionally suffixed by one or more
11953 feature modifiers. This option has the form
11954 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
11955 permissible values for @var{cpu} are the same as those available for
11958 The permissible values for @var{feature} are documented in the sub-section
11961 Where conflicting feature modifiers are specified, the right-most feature is
11964 GCC uses this name to determine what kind of instructions it can emit when
11965 generating assembly code (as if by @option{-march}) and to determine
11966 the target processor for which to tune for performance (as if
11967 by @option{-mtune}). Where this option is used in conjunction
11968 with @option{-march} or @option{-mtune}, those options take precedence
11969 over the appropriate part of this option.
11972 @subsubsection @option{-march} and @option{-mcpu} feature modifiers
11973 @cindex @option{-march} feature modifiers
11974 @cindex @option{-mcpu} feature modifiers
11975 Feature modifiers used with @option{-march} and @option{-mcpu} can be one
11980 Enable CRC extension.
11982 Enable Crypto extension. This implies Advanced SIMD is enabled.
11984 Enable floating-point instructions.
11986 Enable Advanced SIMD instructions. This implies floating-point instructions
11987 are enabled. This is the default for all current possible values for options
11988 @option{-march} and @option{-mcpu=}.
11991 @node Adapteva Epiphany Options
11992 @subsection Adapteva Epiphany Options
11994 These @samp{-m} options are defined for Adapteva Epiphany:
11997 @item -mhalf-reg-file
11998 @opindex mhalf-reg-file
11999 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
12000 That allows code to run on hardware variants that lack these registers.
12002 @item -mprefer-short-insn-regs
12003 @opindex mprefer-short-insn-regs
12004 Preferrentially allocate registers that allow short instruction generation.
12005 This can result in increased instruction count, so this may either reduce or
12006 increase overall code size.
12008 @item -mbranch-cost=@var{num}
12009 @opindex mbranch-cost
12010 Set the cost of branches to roughly @var{num} ``simple'' instructions.
12011 This cost is only a heuristic and is not guaranteed to produce
12012 consistent results across releases.
12016 Enable the generation of conditional moves.
12018 @item -mnops=@var{num}
12020 Emit @var{num} NOPs before every other generated instruction.
12022 @item -mno-soft-cmpsf
12023 @opindex mno-soft-cmpsf
12024 For single-precision floating-point comparisons, emit an @code{fsub} instruction
12025 and test the flags. This is faster than a software comparison, but can
12026 get incorrect results in the presence of NaNs, or when two different small
12027 numbers are compared such that their difference is calculated as zero.
12028 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
12029 software comparisons.
12031 @item -mstack-offset=@var{num}
12032 @opindex mstack-offset
12033 Set the offset between the top of the stack and the stack pointer.
12034 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
12035 can be used by leaf functions without stack allocation.
12036 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
12037 Note also that this option changes the ABI; compiling a program with a
12038 different stack offset than the libraries have been compiled with
12039 generally does not work.
12040 This option can be useful if you want to evaluate if a different stack
12041 offset would give you better code, but to actually use a different stack
12042 offset to build working programs, it is recommended to configure the
12043 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
12045 @item -mno-round-nearest
12046 @opindex mno-round-nearest
12047 Make the scheduler assume that the rounding mode has been set to
12048 truncating. The default is @option{-mround-nearest}.
12051 @opindex mlong-calls
12052 If not otherwise specified by an attribute, assume all calls might be beyond
12053 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
12054 function address into a register before performing a (otherwise direct) call.
12055 This is the default.
12057 @item -mshort-calls
12058 @opindex short-calls
12059 If not otherwise specified by an attribute, assume all direct calls are
12060 in the range of the @code{b} / @code{bl} instructions, so use these instructions
12061 for direct calls. The default is @option{-mlong-calls}.
12065 Assume addresses can be loaded as 16-bit unsigned values. This does not
12066 apply to function addresses for which @option{-mlong-calls} semantics
12069 @item -mfp-mode=@var{mode}
12071 Set the prevailing mode of the floating-point unit.
12072 This determines the floating-point mode that is provided and expected
12073 at function call and return time. Making this mode match the mode you
12074 predominantly need at function start can make your programs smaller and
12075 faster by avoiding unnecessary mode switches.
12077 @var{mode} can be set to one the following values:
12081 Any mode at function entry is valid, and retained or restored when
12082 the function returns, and when it calls other functions.
12083 This mode is useful for compiling libraries or other compilation units
12084 you might want to incorporate into different programs with different
12085 prevailing FPU modes, and the convenience of being able to use a single
12086 object file outweighs the size and speed overhead for any extra
12087 mode switching that might be needed, compared with what would be needed
12088 with a more specific choice of prevailing FPU mode.
12091 This is the mode used for floating-point calculations with
12092 truncating (i.e.@: round towards zero) rounding mode. That includes
12093 conversion from floating point to integer.
12095 @item round-nearest
12096 This is the mode used for floating-point calculations with
12097 round-to-nearest-or-even rounding mode.
12100 This is the mode used to perform integer calculations in the FPU, e.g.@:
12101 integer multiply, or integer multiply-and-accumulate.
12104 The default is @option{-mfp-mode=caller}
12106 @item -mnosplit-lohi
12107 @itemx -mno-postinc
12108 @itemx -mno-postmodify
12109 @opindex mnosplit-lohi
12110 @opindex mno-postinc
12111 @opindex mno-postmodify
12112 Code generation tweaks that disable, respectively, splitting of 32-bit
12113 loads, generation of post-increment addresses, and generation of
12114 post-modify addresses. The defaults are @option{msplit-lohi},
12115 @option{-mpost-inc}, and @option{-mpost-modify}.
12117 @item -mnovect-double
12118 @opindex mno-vect-double
12119 Change the preferred SIMD mode to SImode. The default is
12120 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
12122 @item -max-vect-align=@var{num}
12123 @opindex max-vect-align
12124 The maximum alignment for SIMD vector mode types.
12125 @var{num} may be 4 or 8. The default is 8.
12126 Note that this is an ABI change, even though many library function
12127 interfaces are unaffected if they don't use SIMD vector modes
12128 in places that affect size and/or alignment of relevant types.
12130 @item -msplit-vecmove-early
12131 @opindex msplit-vecmove-early
12132 Split vector moves into single word moves before reload. In theory this
12133 can give better register allocation, but so far the reverse seems to be
12134 generally the case.
12136 @item -m1reg-@var{reg}
12138 Specify a register to hold the constant @minus{}1, which makes loading small negative
12139 constants and certain bitmasks faster.
12140 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
12141 which specify use of that register as a fixed register,
12142 and @samp{none}, which means that no register is used for this
12143 purpose. The default is @option{-m1reg-none}.
12148 @subsection ARC Options
12149 @cindex ARC options
12151 The following options control the architecture variant for which code
12154 @c architecture variants
12157 @item -mbarrel-shifter
12158 @opindex mbarrel-shifter
12159 Generate instructions supported by barrel shifter. This is the default
12160 unless @samp{-mcpu=ARC601} is in effect.
12162 @item -mcpu=@var{cpu}
12164 Set architecture type, register usage, and instruction scheduling
12165 parameters for @var{cpu}. There are also shortcut alias options
12166 available for backward compatibility and convenience. Supported
12167 values for @var{cpu} are
12173 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
12177 Compile for ARC601. Alias: @option{-mARC601}.
12182 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
12183 This is the default when configured with @samp{--with-cpu=arc700}@.
12188 @itemx -mdpfp-compact
12189 @opindex mdpfp-compact
12190 FPX: Generate Double Precision FPX instructions, tuned for the compact
12194 @opindex mdpfp-fast
12195 FPX: Generate Double Precision FPX instructions, tuned for the fast
12198 @item -mno-dpfp-lrsr
12199 @opindex mno-dpfp-lrsr
12200 Disable LR and SR instructions from using FPX extension aux registers.
12204 Generate Extended arithmetic instructions. Currently only
12205 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
12206 supported. This is always enabled for @samp{-mcpu=ARC700}.
12210 Do not generate mpy instructions for ARC700.
12214 Generate 32x16 bit multiply and mac instructions.
12218 Generate mul64 and mulu64 instructions. Only valid for @samp{-mcpu=ARC600}.
12222 Generate norm instruction. This is the default if @samp{-mcpu=ARC700}
12227 @itemx -mspfp-compact
12228 @opindex mspfp-compact
12229 FPX: Generate Single Precision FPX instructions, tuned for the compact
12233 @opindex mspfp-fast
12234 FPX: Generate Single Precision FPX instructions, tuned for the fast
12239 Enable generation of ARC SIMD instructions via target-specific
12240 builtins. Only valid for @samp{-mcpu=ARC700}.
12243 @opindex msoft-float
12244 This option ignored; it is provided for compatibility purposes only.
12245 Software floating point code is emitted by default, and this default
12246 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
12247 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
12248 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
12252 Generate swap instructions.
12256 The following options are passed through to the assembler, and also
12257 define preprocessor macro symbols.
12259 @c Flags used by the assembler, but for which we define preprocessor
12260 @c macro symbols as well.
12263 @opindex mdsp-packa
12264 Passed down to the assembler to enable the DSP Pack A extensions.
12265 Also sets the preprocessor symbol @code{__Xdsp_packa}.
12269 Passed down to the assembler to enable the dual viterbi butterfly
12270 extension. Also sets the preprocessor symbol @code{__Xdvbf}.
12272 @c ARC700 4.10 extension instruction
12275 Passed down to the assembler to enable the Locked Load/Store
12276 Conditional extension. Also sets the preprocessor symbol
12281 Passed down to the assembler. Also sets the preprocessor symbol
12282 @code{__Xxmac_d16}.
12286 Passed down to the assembler. Also sets the preprocessor symbol
12289 @c ARC700 4.10 extension instruction
12292 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
12293 extension instruction. Also sets the preprocessor symbol
12296 @c ARC700 4.10 extension instruction
12299 Passed down to the assembler to enable the swap byte ordering
12300 extension instruction. Also sets the preprocessor symbol
12304 @opindex mtelephony
12305 Passed down to the assembler to enable dual and single operand
12306 instructions for telephony. Also sets the preprocessor symbol
12307 @code{__Xtelephony}.
12311 Passed down to the assembler to enable the XY Memory extension. Also
12312 sets the preprocessor symbol @code{__Xxy}.
12316 The following options control how the assembly code is annotated:
12318 @c Assembly annotation options
12322 Annotate assembler instructions with estimated addresses.
12324 @item -mannotate-align
12325 @opindex mannotate-align
12326 Explain what alignment considerations lead to the decision to make an
12327 instruction short or long.
12331 The following options are passed through to the linker:
12333 @c options passed through to the linker
12337 Passed through to the linker, to specify use of the @code{arclinux} emulation.
12338 This option is enabled by default in tool chains built for
12339 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
12340 when profiling is not requested.
12342 @item -marclinux_prof
12343 @opindex marclinux_prof
12344 Passed through to the linker, to specify use of the
12345 @code{arclinux_prof} emulation. This option is enabled by default in
12346 tool chains built for @w{@code{arc-linux-uclibc}} and
12347 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
12351 The following options control the semantics of generated code:
12353 @c semantically relevant code generation options
12355 @item -mepilogue-cfi
12356 @opindex mepilogue-cfi
12357 Enable generation of call frame information for epilogues.
12359 @item -mno-epilogue-cfi
12360 @opindex mno-epilogue-cfi
12361 Disable generation of call frame information for epilogues.
12364 @opindex mlong-calls
12365 Generate call insns as register indirect calls, thus providing access
12366 to the full 32-bit address range.
12368 @item -mmedium-calls
12369 @opindex mmedium-calls
12370 Don't use less than 25 bit addressing range for calls, which is the
12371 offset available for an unconditional branch-and-link
12372 instruction. Conditional execution of function calls is suppressed, to
12373 allow use of the 25-bit range, rather than the 21-bit range with
12374 conditional branch-and-link. This is the default for tool chains built
12375 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
12379 Do not generate sdata references. This is the default for tool chains
12380 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
12384 @opindex mucb-mcount
12385 Instrument with mcount calls as used in UCB code. I.e. do the
12386 counting in the callee, not the caller. By default ARC instrumentation
12387 counts in the caller.
12389 @item -mvolatile-cache
12390 @opindex mvolatile-cache
12391 Use ordinarily cached memory accesses for volatile references. This is the
12394 @item -mno-volatile-cache
12395 @opindex mno-volatile-cache
12396 Enable cache bypass for volatile references.
12400 The following options fine tune code generation:
12401 @c code generation tuning options
12404 @opindex malign-call
12405 Do alignment optimizations for call instructions.
12407 @item -mauto-modify-reg
12408 @opindex mauto-modify-reg
12409 Enable the use of pre/post modify with register displacement.
12411 @item -mbbit-peephole
12412 @opindex mbbit-peephole
12413 Enable bbit peephole2.
12417 This option disables a target-specific pass in @file{arc_reorg} to
12418 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
12419 generation driven by the combiner pass.
12421 @item -mcase-vector-pcrel
12422 @opindex mcase-vector-pcrel
12423 Use pc-relative switch case tables - this enables case table shortening.
12424 This is the default for @option{-Os}.
12426 @item -mcompact-casesi
12427 @opindex mcompact-casesi
12428 Enable compact casesi pattern.
12429 This is the default for @option{-Os}.
12431 @item -mno-cond-exec
12432 @opindex mno-cond-exec
12433 Disable ARCompact specific pass to generate conditional execution instructions.
12434 Due to delay slot scheduling and interactions between operand numbers,
12435 literal sizes, instruction lengths, and the support for conditional execution,
12436 the target-independent pass to generate conditional execution is often lacking,
12437 so the ARC port has kept a special pass around that tries to find more
12438 conditional execution generating opportunities after register allocation,
12439 branch shortening, and delay slot scheduling have been done. This pass
12440 generally, but not always, improves performance and code size, at the cost of
12441 extra compilation time, which is why there is an option to switch it off.
12442 If you have a problem with call instructions exceeding their allowable
12443 offset range because they are conditionalized, you should consider using
12444 @option{-mmedium-calls} instead.
12446 @item -mearly-cbranchsi
12447 @opindex mearly-cbranchsi
12448 Enable pre-reload use of the cbranchsi pattern.
12450 @item -mexpand-adddi
12451 @opindex mexpand-adddi
12452 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
12453 @code{add.f}, @code{adc} etc.
12455 @item -mindexed-loads
12456 @opindex mindexed-loads
12457 Enable the use of indexed loads. This can be problematic because some
12458 optimizers will then assume the that indexed stores exist, which is not
12463 Enable Local Register Allocation. This is still experimental for ARC,
12464 so by default the compiler uses standard reload
12465 (i.e. @samp{-mno-lra}).
12467 @item -mlra-priority-none
12468 @opindex mlra-priority-none
12469 Don't indicate any priority for target registers.
12471 @item -mlra-priority-compact
12472 @opindex mlra-priority-compact
12473 Indicate target register priority for r0..r3 / r12..r15.
12475 @item -mlra-priority-noncompact
12476 @opindex mlra-priority-noncompact
12477 Reduce target regsiter priority for r0..r3 / r12..r15.
12479 @item -mno-millicode
12480 @opindex mno-millicode
12481 When optimizing for size (using @option{-Os}), prologues and epilogues
12482 that have to save or restore a large number of registers are often
12483 shortened by using call to a special function in libgcc; this is
12484 referred to as a @emph{millicode} call. As these calls can pose
12485 performance issues, and/or cause linking issues when linking in a
12486 nonstandard way, this option is provided to turn off millicode call
12490 @opindex mmixed-code
12491 Tweak register allocation to help 16-bit instruction generation.
12492 This generally has the effect of decreasing the average instruction size
12493 while increasing the instruction count.
12497 Enable 'q' instruction alternatives.
12498 This is the default for @option{-Os}.
12502 Enable Rcq constraint handling - most short code generation depends on this.
12503 This is the default.
12507 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
12508 This is the default.
12510 @item -msize-level=@var{level}
12511 @opindex msize-level
12512 Fine-tune size optimization with regards to instruction lengths and alignment.
12513 The recognized values for @var{level} are:
12516 No size optimization. This level is deprecated and treated like @samp{1}.
12519 Short instructions are used opportunistically.
12522 In addition, alignment of loops and of code after barriers are dropped.
12525 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
12529 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
12530 the behavior when this is not set is equivalent to level @samp{1}.
12532 @item -mtune=@var{cpu}
12534 Set instruction scheduling parameters for @var{cpu}, overriding any implied
12535 by @option{-mcpu=}.
12537 Supported values for @var{cpu} are
12541 Tune for ARC600 cpu.
12544 Tune for ARC601 cpu.
12547 Tune for ARC700 cpu with standard multiplier block.
12550 Tune for ARC700 cpu with XMAC block.
12553 Tune for ARC725D cpu.
12556 Tune for ARC750D cpu.
12560 @item -mmultcost=@var{num}
12562 Cost to assume for a multiply instruction, with @samp{4} being equal to a
12563 normal instruction.
12565 @item -munalign-prob-threshold=@var{probability}
12566 @opindex munalign-prob-threshold
12567 Set probability threshold for unaligning branches.
12568 When tuning for @samp{ARC700} and optimizing for speed, branches without
12569 filled delay slot are preferably emitted unaligned and long, unless
12570 profiling indicates that the probability for the branch to be taken
12571 is below @var{probability}. @xref{Cross-profiling}.
12572 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
12576 The following options are maintained for backward compatibility, but
12577 are now deprecated and will be removed in a future release:
12579 @c Deprecated options
12587 @opindex mbig-endian
12590 Compile code for big endian targets. Use of these options is now
12591 deprecated. Users wanting big-endian code, should use the
12592 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
12593 building the tool chain, for which big-endian is the default.
12595 @item -mlittle-endian
12596 @opindex mlittle-endian
12599 Compile code for little endian targets. Use of these options is now
12600 deprecated. Users wanting little-endian code should use the
12601 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
12602 building the tool chain, for which little-endian is the default.
12604 @item -mbarrel_shifter
12605 @opindex mbarrel_shifter
12606 Replaced by @samp{-mbarrel-shifter}
12608 @item -mdpfp_compact
12609 @opindex mdpfp_compact
12610 Replaced by @samp{-mdpfp-compact}
12613 @opindex mdpfp_fast
12614 Replaced by @samp{-mdpfp-fast}
12617 @opindex mdsp_packa
12618 Replaced by @samp{-mdsp-packa}
12622 Replaced by @samp{-mea}
12626 Replaced by @samp{-mmac-24}
12630 Replaced by @samp{-mmac-d16}
12632 @item -mspfp_compact
12633 @opindex mspfp_compact
12634 Replaced by @samp{-mspfp-compact}
12637 @opindex mspfp_fast
12638 Replaced by @samp{-mspfp-fast}
12640 @item -mtune=@var{cpu}
12642 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
12643 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
12644 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
12646 @item -multcost=@var{num}
12648 Replaced by @samp{-mmultcost}.
12653 @subsection ARM Options
12654 @cindex ARM options
12656 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
12660 @item -mabi=@var{name}
12662 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
12663 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
12666 @opindex mapcs-frame
12667 Generate a stack frame that is compliant with the ARM Procedure Call
12668 Standard for all functions, even if this is not strictly necessary for
12669 correct execution of the code. Specifying @option{-fomit-frame-pointer}
12670 with this option causes the stack frames not to be generated for
12671 leaf functions. The default is @option{-mno-apcs-frame}.
12675 This is a synonym for @option{-mapcs-frame}.
12678 @c not currently implemented
12679 @item -mapcs-stack-check
12680 @opindex mapcs-stack-check
12681 Generate code to check the amount of stack space available upon entry to
12682 every function (that actually uses some stack space). If there is
12683 insufficient space available then either the function
12684 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} is
12685 called, depending upon the amount of stack space required. The runtime
12686 system is required to provide these functions. The default is
12687 @option{-mno-apcs-stack-check}, since this produces smaller code.
12689 @c not currently implemented
12691 @opindex mapcs-float
12692 Pass floating-point arguments using the floating-point registers. This is
12693 one of the variants of the APCS@. This option is recommended if the
12694 target hardware has a floating-point unit or if a lot of floating-point
12695 arithmetic is going to be performed by the code. The default is
12696 @option{-mno-apcs-float}, since the size of integer-only code is
12697 slightly increased if @option{-mapcs-float} is used.
12699 @c not currently implemented
12700 @item -mapcs-reentrant
12701 @opindex mapcs-reentrant
12702 Generate reentrant, position-independent code. The default is
12703 @option{-mno-apcs-reentrant}.
12706 @item -mthumb-interwork
12707 @opindex mthumb-interwork
12708 Generate code that supports calling between the ARM and Thumb
12709 instruction sets. Without this option, on pre-v5 architectures, the
12710 two instruction sets cannot be reliably used inside one program. The
12711 default is @option{-mno-thumb-interwork}, since slightly larger code
12712 is generated when @option{-mthumb-interwork} is specified. In AAPCS
12713 configurations this option is meaningless.
12715 @item -mno-sched-prolog
12716 @opindex mno-sched-prolog
12717 Prevent the reordering of instructions in the function prologue, or the
12718 merging of those instruction with the instructions in the function's
12719 body. This means that all functions start with a recognizable set
12720 of instructions (or in fact one of a choice from a small set of
12721 different function prologues), and this information can be used to
12722 locate the start of functions inside an executable piece of code. The
12723 default is @option{-msched-prolog}.
12725 @item -mfloat-abi=@var{name}
12726 @opindex mfloat-abi
12727 Specifies which floating-point ABI to use. Permissible values
12728 are: @samp{soft}, @samp{softfp} and @samp{hard}.
12730 Specifying @samp{soft} causes GCC to generate output containing
12731 library calls for floating-point operations.
12732 @samp{softfp} allows the generation of code using hardware floating-point
12733 instructions, but still uses the soft-float calling conventions.
12734 @samp{hard} allows generation of floating-point instructions
12735 and uses FPU-specific calling conventions.
12737 The default depends on the specific target configuration. Note that
12738 the hard-float and soft-float ABIs are not link-compatible; you must
12739 compile your entire program with the same ABI, and link with a
12740 compatible set of libraries.
12742 @item -mlittle-endian
12743 @opindex mlittle-endian
12744 Generate code for a processor running in little-endian mode. This is
12745 the default for all standard configurations.
12748 @opindex mbig-endian
12749 Generate code for a processor running in big-endian mode; the default is
12750 to compile code for a little-endian processor.
12752 @item -march=@var{name}
12754 This specifies the name of the target ARM architecture. GCC uses this
12755 name to determine what kind of instructions it can emit when generating
12756 assembly code. This option can be used in conjunction with or instead
12757 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
12758 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
12759 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
12760 @samp{armv6}, @samp{armv6j},
12761 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
12762 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m}, @samp{armv7e-m},
12763 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc},
12764 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
12766 @option{-march=armv7ve} is the armv7-a architecture with virtualization
12769 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
12770 architecture together with the optional CRC32 extensions.
12772 @option{-march=native} causes the compiler to auto-detect the architecture
12773 of the build computer. At present, this feature is only supported on
12774 GNU/Linux, and not all architectures are recognized. If the auto-detect
12775 is unsuccessful the option has no effect.
12777 @item -mtune=@var{name}
12779 This option specifies the name of the target ARM processor for
12780 which GCC should tune the performance of the code.
12781 For some ARM implementations better performance can be obtained by using
12783 Permissible names are: @samp{arm2}, @samp{arm250},
12784 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
12785 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
12786 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
12787 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
12789 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
12790 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
12791 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
12792 @samp{strongarm1110},
12793 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
12794 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
12795 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
12796 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
12797 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
12798 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
12799 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
12800 @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8}, @samp{cortex-a9},
12801 @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a53}, @samp{cortex-a57},
12803 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-m7},
12808 @samp{cortex-m0plus},
12809 @samp{cortex-m1.small-multiply},
12810 @samp{cortex-m0.small-multiply},
12811 @samp{cortex-m0plus.small-multiply},
12812 @samp{marvell-pj4},
12813 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
12814 @samp{fa526}, @samp{fa626},
12815 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}.
12817 Additionally, this option can specify that GCC should tune the performance
12818 of the code for a big.LITTLE system. Permissible names are:
12819 @samp{cortex-a15.cortex-a7}, @samp{cortex-a57.cortex-a53}.
12821 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
12822 performance for a blend of processors within architecture @var{arch}.
12823 The aim is to generate code that run well on the current most popular
12824 processors, balancing between optimizations that benefit some CPUs in the
12825 range, and avoiding performance pitfalls of other CPUs. The effects of
12826 this option may change in future GCC versions as CPU models come and go.
12828 @option{-mtune=native} causes the compiler to auto-detect the CPU
12829 of the build computer. At present, this feature is only supported on
12830 GNU/Linux, and not all architectures are recognized. If the auto-detect is
12831 unsuccessful the option has no effect.
12833 @item -mcpu=@var{name}
12835 This specifies the name of the target ARM processor. GCC uses this name
12836 to derive the name of the target ARM architecture (as if specified
12837 by @option{-march}) and the ARM processor type for which to tune for
12838 performance (as if specified by @option{-mtune}). Where this option
12839 is used in conjunction with @option{-march} or @option{-mtune},
12840 those options take precedence over the appropriate part of this option.
12842 Permissible names for this option are the same as those for
12845 @option{-mcpu=generic-@var{arch}} is also permissible, and is
12846 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
12847 See @option{-mtune} for more information.
12849 @option{-mcpu=native} causes the compiler to auto-detect the CPU
12850 of the build computer. At present, this feature is only supported on
12851 GNU/Linux, and not all architectures are recognized. If the auto-detect
12852 is unsuccessful the option has no effect.
12854 @item -mfpu=@var{name}
12856 This specifies what floating-point hardware (or hardware emulation) is
12857 available on the target. Permissible names are: @samp{vfp}, @samp{vfpv3},
12858 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
12859 @samp{vfpv3xd-fp16}, @samp{neon}, @samp{neon-fp16}, @samp{vfpv4},
12860 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
12861 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
12862 @samp{fp-armv8}, @samp{neon-fp-armv8}, and @samp{crypto-neon-fp-armv8}.
12864 If @option{-msoft-float} is specified this specifies the format of
12865 floating-point values.
12867 If the selected floating-point hardware includes the NEON extension
12868 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
12869 operations are not generated by GCC's auto-vectorization pass unless
12870 @option{-funsafe-math-optimizations} is also specified. This is
12871 because NEON hardware does not fully implement the IEEE 754 standard for
12872 floating-point arithmetic (in particular denormal values are treated as
12873 zero), so the use of NEON instructions may lead to a loss of precision.
12875 @item -mfp16-format=@var{name}
12876 @opindex mfp16-format
12877 Specify the format of the @code{__fp16} half-precision floating-point type.
12878 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
12879 the default is @samp{none}, in which case the @code{__fp16} type is not
12880 defined. @xref{Half-Precision}, for more information.
12882 @item -mstructure-size-boundary=@var{n}
12883 @opindex mstructure-size-boundary
12884 The sizes of all structures and unions are rounded up to a multiple
12885 of the number of bits set by this option. Permissible values are 8, 32
12886 and 64. The default value varies for different toolchains. For the COFF
12887 targeted toolchain the default value is 8. A value of 64 is only allowed
12888 if the underlying ABI supports it.
12890 Specifying a larger number can produce faster, more efficient code, but
12891 can also increase the size of the program. Different values are potentially
12892 incompatible. Code compiled with one value cannot necessarily expect to
12893 work with code or libraries compiled with another value, if they exchange
12894 information using structures or unions.
12896 @item -mabort-on-noreturn
12897 @opindex mabort-on-noreturn
12898 Generate a call to the function @code{abort} at the end of a
12899 @code{noreturn} function. It is executed if the function tries to
12903 @itemx -mno-long-calls
12904 @opindex mlong-calls
12905 @opindex mno-long-calls
12906 Tells the compiler to perform function calls by first loading the
12907 address of the function into a register and then performing a subroutine
12908 call on this register. This switch is needed if the target function
12909 lies outside of the 64-megabyte addressing range of the offset-based
12910 version of subroutine call instruction.
12912 Even if this switch is enabled, not all function calls are turned
12913 into long calls. The heuristic is that static functions, functions
12914 that have the @samp{short-call} attribute, functions that are inside
12915 the scope of a @samp{#pragma no_long_calls} directive, and functions whose
12916 definitions have already been compiled within the current compilation
12917 unit are not turned into long calls. The exceptions to this rule are
12918 that weak function definitions, functions with the @samp{long-call}
12919 attribute or the @samp{section} attribute, and functions that are within
12920 the scope of a @samp{#pragma long_calls} directive are always
12921 turned into long calls.
12923 This feature is not enabled by default. Specifying
12924 @option{-mno-long-calls} restores the default behavior, as does
12925 placing the function calls within the scope of a @samp{#pragma
12926 long_calls_off} directive. Note these switches have no effect on how
12927 the compiler generates code to handle function calls via function
12930 @item -msingle-pic-base
12931 @opindex msingle-pic-base
12932 Treat the register used for PIC addressing as read-only, rather than
12933 loading it in the prologue for each function. The runtime system is
12934 responsible for initializing this register with an appropriate value
12935 before execution begins.
12937 @item -mpic-register=@var{reg}
12938 @opindex mpic-register
12939 Specify the register to be used for PIC addressing.
12940 For standard PIC base case, the default will be any suitable register
12941 determined by compiler. For single PIC base case, the default is
12942 @samp{R9} if target is EABI based or stack-checking is enabled,
12943 otherwise the default is @samp{R10}.
12945 @item -mpic-data-is-text-relative
12946 @opindex mpic-data-is-text-relative
12947 Assume that each data segments are relative to text segment at load time.
12948 Therefore, it permits addressing data using PC-relative operations.
12949 This option is on by default for targets other than VxWorks RTP.
12951 @item -mpoke-function-name
12952 @opindex mpoke-function-name
12953 Write the name of each function into the text section, directly
12954 preceding the function prologue. The generated code is similar to this:
12958 .ascii "arm_poke_function_name", 0
12961 .word 0xff000000 + (t1 - t0)
12962 arm_poke_function_name
12964 stmfd sp!, @{fp, ip, lr, pc@}
12968 When performing a stack backtrace, code can inspect the value of
12969 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
12970 location @code{pc - 12} and the top 8 bits are set, then we know that
12971 there is a function name embedded immediately preceding this location
12972 and has length @code{((pc[-3]) & 0xff000000)}.
12979 Select between generating code that executes in ARM and Thumb
12980 states. The default for most configurations is to generate code
12981 that executes in ARM state, but the default can be changed by
12982 configuring GCC with the @option{--with-mode=}@var{state}
12986 @opindex mtpcs-frame
12987 Generate a stack frame that is compliant with the Thumb Procedure Call
12988 Standard for all non-leaf functions. (A leaf function is one that does
12989 not call any other functions.) The default is @option{-mno-tpcs-frame}.
12991 @item -mtpcs-leaf-frame
12992 @opindex mtpcs-leaf-frame
12993 Generate a stack frame that is compliant with the Thumb Procedure Call
12994 Standard for all leaf functions. (A leaf function is one that does
12995 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
12997 @item -mcallee-super-interworking
12998 @opindex mcallee-super-interworking
12999 Gives all externally visible functions in the file being compiled an ARM
13000 instruction set header which switches to Thumb mode before executing the
13001 rest of the function. This allows these functions to be called from
13002 non-interworking code. This option is not valid in AAPCS configurations
13003 because interworking is enabled by default.
13005 @item -mcaller-super-interworking
13006 @opindex mcaller-super-interworking
13007 Allows calls via function pointers (including virtual functions) to
13008 execute correctly regardless of whether the target code has been
13009 compiled for interworking or not. There is a small overhead in the cost
13010 of executing a function pointer if this option is enabled. This option
13011 is not valid in AAPCS configurations because interworking is enabled
13014 @item -mtp=@var{name}
13016 Specify the access model for the thread local storage pointer. The valid
13017 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
13018 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
13019 (supported in the arm6k architecture), and @option{auto}, which uses the
13020 best available method for the selected processor. The default setting is
13023 @item -mtls-dialect=@var{dialect}
13024 @opindex mtls-dialect
13025 Specify the dialect to use for accessing thread local storage. Two
13026 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
13027 @samp{gnu} dialect selects the original GNU scheme for supporting
13028 local and global dynamic TLS models. The @samp{gnu2} dialect
13029 selects the GNU descriptor scheme, which provides better performance
13030 for shared libraries. The GNU descriptor scheme is compatible with
13031 the original scheme, but does require new assembler, linker and
13032 library support. Initial and local exec TLS models are unaffected by
13033 this option and always use the original scheme.
13035 @item -mword-relocations
13036 @opindex mword-relocations
13037 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
13038 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
13039 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
13042 @item -mfix-cortex-m3-ldrd
13043 @opindex mfix-cortex-m3-ldrd
13044 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
13045 with overlapping destination and base registers are used. This option avoids
13046 generating these instructions. This option is enabled by default when
13047 @option{-mcpu=cortex-m3} is specified.
13049 @item -munaligned-access
13050 @itemx -mno-unaligned-access
13051 @opindex munaligned-access
13052 @opindex mno-unaligned-access
13053 Enables (or disables) reading and writing of 16- and 32- bit values
13054 from addresses that are not 16- or 32- bit aligned. By default
13055 unaligned access is disabled for all pre-ARMv6 and all ARMv6-M
13056 architectures, and enabled for all other architectures. If unaligned
13057 access is not enabled then words in packed data structures will be
13058 accessed a byte at a time.
13060 The ARM attribute @code{Tag_CPU_unaligned_access} will be set in the
13061 generated object file to either true or false, depending upon the
13062 setting of this option. If unaligned access is enabled then the
13063 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} will also be
13066 @item -mneon-for-64bits
13067 @opindex mneon-for-64bits
13068 Enables using Neon to handle scalar 64-bits operations. This is
13069 disabled by default since the cost of moving data from core registers
13072 @item -mslow-flash-data
13073 @opindex mslow-flash-data
13074 Assume loading data from flash is slower than fetching instruction.
13075 Therefore literal load is minimized for better performance.
13076 This option is only supported when compiling for ARMv7 M-profile and
13079 @item -masm-syntax-unified
13080 @opindex masm-syntax-unified
13081 Assume inline assembler is using unified asm syntax. The default is
13082 currently off which implies divided syntax. Currently this option is
13083 available only for Thumb1 and has no effect on ARM state and Thumb2.
13084 However, this may change in future releases of GCC. Divided syntax
13085 should be considered deprecated.
13087 @item -mrestrict-it
13088 @opindex mrestrict-it
13089 Restricts generation of IT blocks to conform to the rules of ARMv8.
13090 IT blocks can only contain a single 16-bit instruction from a select
13091 set of instructions. This option is on by default for ARMv8 Thumb mode.
13095 @subsection AVR Options
13096 @cindex AVR Options
13098 These options are defined for AVR implementations:
13101 @item -mmcu=@var{mcu}
13103 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
13105 The default for this option is@tie{}@code{avr2}.
13107 GCC supports the following AVR devices and ISAs:
13109 @include avr-mmcu.texi
13111 @item -maccumulate-args
13112 @opindex maccumulate-args
13113 Accumulate outgoing function arguments and acquire/release the needed
13114 stack space for outgoing function arguments once in function
13115 prologue/epilogue. Without this option, outgoing arguments are pushed
13116 before calling a function and popped afterwards.
13118 Popping the arguments after the function call can be expensive on
13119 AVR so that accumulating the stack space might lead to smaller
13120 executables because arguments need not to be removed from the
13121 stack after such a function call.
13123 This option can lead to reduced code size for functions that perform
13124 several calls to functions that get their arguments on the stack like
13125 calls to printf-like functions.
13127 @item -mbranch-cost=@var{cost}
13128 @opindex mbranch-cost
13129 Set the branch costs for conditional branch instructions to
13130 @var{cost}. Reasonable values for @var{cost} are small, non-negative
13131 integers. The default branch cost is 0.
13133 @item -mcall-prologues
13134 @opindex mcall-prologues
13135 Functions prologues/epilogues are expanded as calls to appropriate
13136 subroutines. Code size is smaller.
13140 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
13141 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
13142 and @code{long long} is 4 bytes. Please note that this option does not
13143 conform to the C standards, but it results in smaller code
13146 @item -mno-interrupts
13147 @opindex mno-interrupts
13148 Generated code is not compatible with hardware interrupts.
13149 Code size is smaller.
13153 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
13154 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
13155 Setting @code{-mrelax} just adds the @code{--relax} option to the
13156 linker command line when the linker is called.
13158 Jump relaxing is performed by the linker because jump offsets are not
13159 known before code is located. Therefore, the assembler code generated by the
13160 compiler is the same, but the instructions in the executable may
13161 differ from instructions in the assembler code.
13163 Relaxing must be turned on if linker stubs are needed, see the
13164 section on @code{EIND} and linker stubs below.
13168 Treat the stack pointer register as an 8-bit register,
13169 i.e.@: assume the high byte of the stack pointer is zero.
13170 In general, you don't need to set this option by hand.
13172 This option is used internally by the compiler to select and
13173 build multilibs for architectures @code{avr2} and @code{avr25}.
13174 These architectures mix devices with and without @code{SPH}.
13175 For any setting other than @code{-mmcu=avr2} or @code{-mmcu=avr25}
13176 the compiler driver will add or remove this option from the compiler
13177 proper's command line, because the compiler then knows if the device
13178 or architecture has an 8-bit stack pointer and thus no @code{SPH}
13183 Use address register @code{X} in a way proposed by the hardware. This means
13184 that @code{X} is only used in indirect, post-increment or
13185 pre-decrement addressing.
13187 Without this option, the @code{X} register may be used in the same way
13188 as @code{Y} or @code{Z} which then is emulated by additional
13190 For example, loading a value with @code{X+const} addressing with a
13191 small non-negative @code{const < 64} to a register @var{Rn} is
13195 adiw r26, const ; X += const
13196 ld @var{Rn}, X ; @var{Rn} = *X
13197 sbiw r26, const ; X -= const
13201 @opindex mtiny-stack
13202 Only change the lower 8@tie{}bits of the stack pointer.
13204 @item -Waddr-space-convert
13205 @opindex Waddr-space-convert
13206 Warn about conversions between address spaces in the case where the
13207 resulting address space is not contained in the incoming address space.
13210 @subsubsection @code{EIND} and Devices with more than 128 Ki Bytes of Flash
13211 @cindex @code{EIND}
13212 Pointers in the implementation are 16@tie{}bits wide.
13213 The address of a function or label is represented as word address so
13214 that indirect jumps and calls can target any code address in the
13215 range of 64@tie{}Ki words.
13217 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
13218 bytes of program memory space, there is a special function register called
13219 @code{EIND} that serves as most significant part of the target address
13220 when @code{EICALL} or @code{EIJMP} instructions are used.
13222 Indirect jumps and calls on these devices are handled as follows by
13223 the compiler and are subject to some limitations:
13228 The compiler never sets @code{EIND}.
13231 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
13232 instructions or might read @code{EIND} directly in order to emulate an
13233 indirect call/jump by means of a @code{RET} instruction.
13236 The compiler assumes that @code{EIND} never changes during the startup
13237 code or during the application. In particular, @code{EIND} is not
13238 saved/restored in function or interrupt service routine
13242 For indirect calls to functions and computed goto, the linker
13243 generates @emph{stubs}. Stubs are jump pads sometimes also called
13244 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
13245 The stub contains a direct jump to the desired address.
13248 Linker relaxation must be turned on so that the linker will generate
13249 the stubs correctly an all situaltion. See the compiler option
13250 @code{-mrelax} and the linler option @code{--relax}.
13251 There are corner cases where the linker is supposed to generate stubs
13252 but aborts without relaxation and without a helpful error message.
13255 The default linker script is arranged for code with @code{EIND = 0}.
13256 If code is supposed to work for a setup with @code{EIND != 0}, a custom
13257 linker script has to be used in order to place the sections whose
13258 name start with @code{.trampolines} into the segment where @code{EIND}
13262 The startup code from libgcc never sets @code{EIND}.
13263 Notice that startup code is a blend of code from libgcc and AVR-LibC.
13264 For the impact of AVR-LibC on @code{EIND}, see the
13265 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
13268 It is legitimate for user-specific startup code to set up @code{EIND}
13269 early, for example by means of initialization code located in
13270 section @code{.init3}. Such code runs prior to general startup code
13271 that initializes RAM and calls constructors, but after the bit
13272 of startup code from AVR-LibC that sets @code{EIND} to the segment
13273 where the vector table is located.
13275 #include <avr/io.h>
13278 __attribute__((section(".init3"),naked,used,no_instrument_function))
13279 init3_set_eind (void)
13281 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
13282 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
13287 The @code{__trampolines_start} symbol is defined in the linker script.
13290 Stubs are generated automatically by the linker if
13291 the following two conditions are met:
13294 @item The address of a label is taken by means of the @code{gs} modifier
13295 (short for @emph{generate stubs}) like so:
13297 LDI r24, lo8(gs(@var{func}))
13298 LDI r25, hi8(gs(@var{func}))
13300 @item The final location of that label is in a code segment
13301 @emph{outside} the segment where the stubs are located.
13305 The compiler emits such @code{gs} modifiers for code labels in the
13306 following situations:
13308 @item Taking address of a function or code label.
13309 @item Computed goto.
13310 @item If prologue-save function is used, see @option{-mcall-prologues}
13311 command-line option.
13312 @item Switch/case dispatch tables. If you do not want such dispatch
13313 tables you can specify the @option{-fno-jump-tables} command-line option.
13314 @item C and C++ constructors/destructors called during startup/shutdown.
13315 @item If the tools hit a @code{gs()} modifier explained above.
13319 Jumping to non-symbolic addresses like so is @emph{not} supported:
13324 /* Call function at word address 0x2 */
13325 return ((int(*)(void)) 0x2)();
13329 Instead, a stub has to be set up, i.e.@: the function has to be called
13330 through a symbol (@code{func_4} in the example):
13335 extern int func_4 (void);
13337 /* Call function at byte address 0x4 */
13342 and the application be linked with @code{-Wl,--defsym,func_4=0x4}.
13343 Alternatively, @code{func_4} can be defined in the linker script.
13346 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
13347 @cindex @code{RAMPD}
13348 @cindex @code{RAMPX}
13349 @cindex @code{RAMPY}
13350 @cindex @code{RAMPZ}
13351 Some AVR devices support memories larger than the 64@tie{}KiB range
13352 that can be accessed with 16-bit pointers. To access memory locations
13353 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
13354 register is used as high part of the address:
13355 The @code{X}, @code{Y}, @code{Z} address register is concatenated
13356 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
13357 register, respectively, to get a wide address. Similarly,
13358 @code{RAMPD} is used together with direct addressing.
13362 The startup code initializes the @code{RAMP} special function
13363 registers with zero.
13366 If a @ref{AVR Named Address Spaces,named address space} other than
13367 generic or @code{__flash} is used, then @code{RAMPZ} is set
13368 as needed before the operation.
13371 If the device supports RAM larger than 64@tie{}KiB and the compiler
13372 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
13373 is reset to zero after the operation.
13376 If the device comes with a specific @code{RAMP} register, the ISR
13377 prologue/epilogue saves/restores that SFR and initializes it with
13378 zero in case the ISR code might (implicitly) use it.
13381 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
13382 If you use inline assembler to read from locations outside the
13383 16-bit address range and change one of the @code{RAMP} registers,
13384 you must reset it to zero after the access.
13388 @subsubsection AVR Built-in Macros
13390 GCC defines several built-in macros so that the user code can test
13391 for the presence or absence of features. Almost any of the following
13392 built-in macros are deduced from device capabilities and thus
13393 triggered by the @code{-mmcu=} command-line option.
13395 For even more AVR-specific built-in macros see
13396 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
13401 Build-in macro that resolves to a decimal number that identifies the
13402 architecture and depends on the @code{-mmcu=@var{mcu}} option.
13403 Possible values are:
13405 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
13406 @code{4}, @code{5}, @code{51}, @code{6}, @code{102}, @code{104},
13407 @code{105}, @code{106}, @code{107}
13409 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3},
13410 @code{avr31}, @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51},
13411 @code{avr6}, @code{avrxmega2}, @code{avrxmega4}, @code{avrxmega5},
13412 @code{avrxmega6}, @code{avrxmega7}, respectively.
13413 If @var{mcu} specifies a device, this built-in macro is set
13414 accordingly. For example, with @code{-mmcu=atmega8} the macro will be
13415 defined to @code{4}.
13417 @item __AVR_@var{Device}__
13418 Setting @code{-mmcu=@var{device}} defines this built-in macro which reflects
13419 the device's name. For example, @code{-mmcu=atmega8} defines the
13420 built-in macro @code{__AVR_ATmega8__}, @code{-mmcu=attiny261a} defines
13421 @code{__AVR_ATtiny261A__}, etc.
13423 The built-in macros' names follow
13424 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
13425 the device name as from the AVR user manual. The difference between
13426 @var{Device} in the built-in macro and @var{device} in
13427 @code{-mmcu=@var{device}} is that the latter is always lowercase.
13429 If @var{device} is not a device but only a core architecture like
13430 @code{avr51}, this macro will not be defined.
13432 @item __AVR_DEVICE_NAME__
13433 Setting @code{-mmcu=@var{device}} defines this built-in macro to
13434 the device's name. For example, with @code{-mmcu=atmega8} the macro
13435 will be defined to @code{atmega8}.
13437 If @var{device} is not a device but only a core architecture like
13438 @code{avr51}, this macro will not be defined.
13440 @item __AVR_XMEGA__
13441 The device / architecture belongs to the XMEGA family of devices.
13443 @item __AVR_HAVE_ELPM__
13444 The device has the the @code{ELPM} instruction.
13446 @item __AVR_HAVE_ELPMX__
13447 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
13448 R@var{n},Z+} instructions.
13450 @item __AVR_HAVE_MOVW__
13451 The device has the @code{MOVW} instruction to perform 16-bit
13452 register-register moves.
13454 @item __AVR_HAVE_LPMX__
13455 The device has the @code{LPM R@var{n},Z} and
13456 @code{LPM R@var{n},Z+} instructions.
13458 @item __AVR_HAVE_MUL__
13459 The device has a hardware multiplier.
13461 @item __AVR_HAVE_JMP_CALL__
13462 The device has the @code{JMP} and @code{CALL} instructions.
13463 This is the case for devices with at least 16@tie{}KiB of program
13466 @item __AVR_HAVE_EIJMP_EICALL__
13467 @itemx __AVR_3_BYTE_PC__
13468 The device has the @code{EIJMP} and @code{EICALL} instructions.
13469 This is the case for devices with more than 128@tie{}KiB of program memory.
13470 This also means that the program counter
13471 (PC) is 3@tie{}bytes wide.
13473 @item __AVR_2_BYTE_PC__
13474 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
13475 with up to 128@tie{}KiB of program memory.
13477 @item __AVR_HAVE_8BIT_SP__
13478 @itemx __AVR_HAVE_16BIT_SP__
13479 The stack pointer (SP) register is treated as 8-bit respectively
13480 16-bit register by the compiler.
13481 The definition of these macros is affected by @code{-mtiny-stack}.
13483 @item __AVR_HAVE_SPH__
13485 The device has the SPH (high part of stack pointer) special function
13486 register or has an 8-bit stack pointer, respectively.
13487 The definition of these macros is affected by @code{-mmcu=} and
13488 in the cases of @code{-mmcu=avr2} and @code{-mmcu=avr25} also
13491 @item __AVR_HAVE_RAMPD__
13492 @itemx __AVR_HAVE_RAMPX__
13493 @itemx __AVR_HAVE_RAMPY__
13494 @itemx __AVR_HAVE_RAMPZ__
13495 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
13496 @code{RAMPZ} special function register, respectively.
13498 @item __NO_INTERRUPTS__
13499 This macro reflects the @code{-mno-interrupts} command line option.
13501 @item __AVR_ERRATA_SKIP__
13502 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
13503 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
13504 instructions because of a hardware erratum. Skip instructions are
13505 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
13506 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
13509 @item __AVR_ISA_RMW__
13510 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
13512 @item __AVR_SFR_OFFSET__=@var{offset}
13513 Instructions that can address I/O special function registers directly
13514 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
13515 address as if addressed by an instruction to access RAM like @code{LD}
13516 or @code{STS}. This offset depends on the device architecture and has
13517 to be subtracted from the RAM address in order to get the
13518 respective I/O@tie{}address.
13520 @item __WITH_AVRLIBC__
13521 The compiler is configured to be used together with AVR-Libc.
13522 See the @code{--with-avrlibc} configure option.
13526 @node Blackfin Options
13527 @subsection Blackfin Options
13528 @cindex Blackfin Options
13531 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
13533 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
13534 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
13535 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
13536 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
13537 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
13538 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
13539 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
13540 @samp{bf561}, @samp{bf592}.
13542 The optional @var{sirevision} specifies the silicon revision of the target
13543 Blackfin processor. Any workarounds available for the targeted silicon revision
13544 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
13545 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
13546 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
13547 hexadecimal digits representing the major and minor numbers in the silicon
13548 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
13549 is not defined. If @var{sirevision} is @samp{any}, the
13550 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
13551 If this optional @var{sirevision} is not used, GCC assumes the latest known
13552 silicon revision of the targeted Blackfin processor.
13554 GCC defines a preprocessor macro for the specified @var{cpu}.
13555 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
13556 provided by libgloss to be linked in if @option{-msim} is not given.
13558 Without this option, @samp{bf532} is used as the processor by default.
13560 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
13561 only the preprocessor macro is defined.
13565 Specifies that the program will be run on the simulator. This causes
13566 the simulator BSP provided by libgloss to be linked in. This option
13567 has effect only for @samp{bfin-elf} toolchain.
13568 Certain other options, such as @option{-mid-shared-library} and
13569 @option{-mfdpic}, imply @option{-msim}.
13571 @item -momit-leaf-frame-pointer
13572 @opindex momit-leaf-frame-pointer
13573 Don't keep the frame pointer in a register for leaf functions. This
13574 avoids the instructions to save, set up and restore frame pointers and
13575 makes an extra register available in leaf functions. The option
13576 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
13577 which might make debugging harder.
13579 @item -mspecld-anomaly
13580 @opindex mspecld-anomaly
13581 When enabled, the compiler ensures that the generated code does not
13582 contain speculative loads after jump instructions. If this option is used,
13583 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
13585 @item -mno-specld-anomaly
13586 @opindex mno-specld-anomaly
13587 Don't generate extra code to prevent speculative loads from occurring.
13589 @item -mcsync-anomaly
13590 @opindex mcsync-anomaly
13591 When enabled, the compiler ensures that the generated code does not
13592 contain CSYNC or SSYNC instructions too soon after conditional branches.
13593 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
13595 @item -mno-csync-anomaly
13596 @opindex mno-csync-anomaly
13597 Don't generate extra code to prevent CSYNC or SSYNC instructions from
13598 occurring too soon after a conditional branch.
13602 When enabled, the compiler is free to take advantage of the knowledge that
13603 the entire program fits into the low 64k of memory.
13606 @opindex mno-low-64k
13607 Assume that the program is arbitrarily large. This is the default.
13609 @item -mstack-check-l1
13610 @opindex mstack-check-l1
13611 Do stack checking using information placed into L1 scratchpad memory by the
13614 @item -mid-shared-library
13615 @opindex mid-shared-library
13616 Generate code that supports shared libraries via the library ID method.
13617 This allows for execute in place and shared libraries in an environment
13618 without virtual memory management. This option implies @option{-fPIC}.
13619 With a @samp{bfin-elf} target, this option implies @option{-msim}.
13621 @item -mno-id-shared-library
13622 @opindex mno-id-shared-library
13623 Generate code that doesn't assume ID-based shared libraries are being used.
13624 This is the default.
13626 @item -mleaf-id-shared-library
13627 @opindex mleaf-id-shared-library
13628 Generate code that supports shared libraries via the library ID method,
13629 but assumes that this library or executable won't link against any other
13630 ID shared libraries. That allows the compiler to use faster code for jumps
13633 @item -mno-leaf-id-shared-library
13634 @opindex mno-leaf-id-shared-library
13635 Do not assume that the code being compiled won't link against any ID shared
13636 libraries. Slower code is generated for jump and call insns.
13638 @item -mshared-library-id=n
13639 @opindex mshared-library-id
13640 Specifies the identification number of the ID-based shared library being
13641 compiled. Specifying a value of 0 generates more compact code; specifying
13642 other values forces the allocation of that number to the current
13643 library but is no more space- or time-efficient than omitting this option.
13647 Generate code that allows the data segment to be located in a different
13648 area of memory from the text segment. This allows for execute in place in
13649 an environment without virtual memory management by eliminating relocations
13650 against the text section.
13652 @item -mno-sep-data
13653 @opindex mno-sep-data
13654 Generate code that assumes that the data segment follows the text segment.
13655 This is the default.
13658 @itemx -mno-long-calls
13659 @opindex mlong-calls
13660 @opindex mno-long-calls
13661 Tells the compiler to perform function calls by first loading the
13662 address of the function into a register and then performing a subroutine
13663 call on this register. This switch is needed if the target function
13664 lies outside of the 24-bit addressing range of the offset-based
13665 version of subroutine call instruction.
13667 This feature is not enabled by default. Specifying
13668 @option{-mno-long-calls} restores the default behavior. Note these
13669 switches have no effect on how the compiler generates code to handle
13670 function calls via function pointers.
13674 Link with the fast floating-point library. This library relaxes some of
13675 the IEEE floating-point standard's rules for checking inputs against
13676 Not-a-Number (NAN), in the interest of performance.
13679 @opindex minline-plt
13680 Enable inlining of PLT entries in function calls to functions that are
13681 not known to bind locally. It has no effect without @option{-mfdpic}.
13684 @opindex mmulticore
13685 Build a standalone application for multicore Blackfin processors.
13686 This option causes proper start files and link scripts supporting
13687 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
13688 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
13690 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
13691 selects the one-application-per-core programming model. Without
13692 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
13693 programming model is used. In this model, the main function of Core B
13694 should be named as @code{coreb_main}.
13696 If this option is not used, the single-core application programming
13701 Build a standalone application for Core A of BF561 when using
13702 the one-application-per-core programming model. Proper start files
13703 and link scripts are used to support Core A, and the macro
13704 @code{__BFIN_COREA} is defined.
13705 This option can only be used in conjunction with @option{-mmulticore}.
13709 Build a standalone application for Core B of BF561 when using
13710 the one-application-per-core programming model. Proper start files
13711 and link scripts are used to support Core B, and the macro
13712 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
13713 should be used instead of @code{main}.
13714 This option can only be used in conjunction with @option{-mmulticore}.
13718 Build a standalone application for SDRAM. Proper start files and
13719 link scripts are used to put the application into SDRAM, and the macro
13720 @code{__BFIN_SDRAM} is defined.
13721 The loader should initialize SDRAM before loading the application.
13725 Assume that ICPLBs are enabled at run time. This has an effect on certain
13726 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
13727 are enabled; for standalone applications the default is off.
13731 @subsection C6X Options
13732 @cindex C6X Options
13735 @item -march=@var{name}
13737 This specifies the name of the target architecture. GCC uses this
13738 name to determine what kind of instructions it can emit when generating
13739 assembly code. Permissible names are: @samp{c62x},
13740 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
13743 @opindex mbig-endian
13744 Generate code for a big-endian target.
13746 @item -mlittle-endian
13747 @opindex mlittle-endian
13748 Generate code for a little-endian target. This is the default.
13752 Choose startup files and linker script suitable for the simulator.
13754 @item -msdata=default
13755 @opindex msdata=default
13756 Put small global and static data in the @samp{.neardata} section,
13757 which is pointed to by register @code{B14}. Put small uninitialized
13758 global and static data in the @samp{.bss} section, which is adjacent
13759 to the @samp{.neardata} section. Put small read-only data into the
13760 @samp{.rodata} section. The corresponding sections used for large
13761 pieces of data are @samp{.fardata}, @samp{.far} and @samp{.const}.
13764 @opindex msdata=all
13765 Put all data, not just small objects, into the sections reserved for
13766 small data, and use addressing relative to the @code{B14} register to
13770 @opindex msdata=none
13771 Make no use of the sections reserved for small data, and use absolute
13772 addresses to access all data. Put all initialized global and static
13773 data in the @samp{.fardata} section, and all uninitialized data in the
13774 @samp{.far} section. Put all constant data into the @samp{.const}
13779 @subsection CRIS Options
13780 @cindex CRIS Options
13782 These options are defined specifically for the CRIS ports.
13785 @item -march=@var{architecture-type}
13786 @itemx -mcpu=@var{architecture-type}
13789 Generate code for the specified architecture. The choices for
13790 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
13791 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
13792 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
13795 @item -mtune=@var{architecture-type}
13797 Tune to @var{architecture-type} everything applicable about the generated
13798 code, except for the ABI and the set of available instructions. The
13799 choices for @var{architecture-type} are the same as for
13800 @option{-march=@var{architecture-type}}.
13802 @item -mmax-stack-frame=@var{n}
13803 @opindex mmax-stack-frame
13804 Warn when the stack frame of a function exceeds @var{n} bytes.
13810 The options @option{-metrax4} and @option{-metrax100} are synonyms for
13811 @option{-march=v3} and @option{-march=v8} respectively.
13813 @item -mmul-bug-workaround
13814 @itemx -mno-mul-bug-workaround
13815 @opindex mmul-bug-workaround
13816 @opindex mno-mul-bug-workaround
13817 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
13818 models where it applies. This option is active by default.
13822 Enable CRIS-specific verbose debug-related information in the assembly
13823 code. This option also has the effect of turning off the @samp{#NO_APP}
13824 formatted-code indicator to the assembler at the beginning of the
13829 Do not use condition-code results from previous instruction; always emit
13830 compare and test instructions before use of condition codes.
13832 @item -mno-side-effects
13833 @opindex mno-side-effects
13834 Do not emit instructions with side effects in addressing modes other than
13837 @item -mstack-align
13838 @itemx -mno-stack-align
13839 @itemx -mdata-align
13840 @itemx -mno-data-align
13841 @itemx -mconst-align
13842 @itemx -mno-const-align
13843 @opindex mstack-align
13844 @opindex mno-stack-align
13845 @opindex mdata-align
13846 @opindex mno-data-align
13847 @opindex mconst-align
13848 @opindex mno-const-align
13849 These options (@samp{no-} options) arrange (eliminate arrangements) for the
13850 stack frame, individual data and constants to be aligned for the maximum
13851 single data access size for the chosen CPU model. The default is to
13852 arrange for 32-bit alignment. ABI details such as structure layout are
13853 not affected by these options.
13861 Similar to the stack- data- and const-align options above, these options
13862 arrange for stack frame, writable data and constants to all be 32-bit,
13863 16-bit or 8-bit aligned. The default is 32-bit alignment.
13865 @item -mno-prologue-epilogue
13866 @itemx -mprologue-epilogue
13867 @opindex mno-prologue-epilogue
13868 @opindex mprologue-epilogue
13869 With @option{-mno-prologue-epilogue}, the normal function prologue and
13870 epilogue which set up the stack frame are omitted and no return
13871 instructions or return sequences are generated in the code. Use this
13872 option only together with visual inspection of the compiled code: no
13873 warnings or errors are generated when call-saved registers must be saved,
13874 or storage for local variables needs to be allocated.
13878 @opindex mno-gotplt
13880 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
13881 instruction sequences that load addresses for functions from the PLT part
13882 of the GOT rather than (traditional on other architectures) calls to the
13883 PLT@. The default is @option{-mgotplt}.
13887 Legacy no-op option only recognized with the cris-axis-elf and
13888 cris-axis-linux-gnu targets.
13892 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
13896 This option, recognized for the cris-axis-elf, arranges
13897 to link with input-output functions from a simulator library. Code,
13898 initialized data and zero-initialized data are allocated consecutively.
13902 Like @option{-sim}, but pass linker options to locate initialized data at
13903 0x40000000 and zero-initialized data at 0x80000000.
13907 @subsection CR16 Options
13908 @cindex CR16 Options
13910 These options are defined specifically for the CR16 ports.
13916 Enable the use of multiply-accumulate instructions. Disabled by default.
13920 @opindex mcr16cplus
13922 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
13927 Links the library libsim.a which is in compatible with simulator. Applicable
13928 to ELF compiler only.
13932 Choose integer type as 32-bit wide.
13936 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
13938 @item -mdata-model=@var{model}
13939 @opindex mdata-model
13940 Choose a data model. The choices for @var{model} are @samp{near},
13941 @samp{far} or @samp{medium}. @samp{medium} is default.
13942 However, @samp{far} is not valid with @option{-mcr16c}, as the
13943 CR16C architecture does not support the far data model.
13946 @node Darwin Options
13947 @subsection Darwin Options
13948 @cindex Darwin options
13950 These options are defined for all architectures running the Darwin operating
13953 FSF GCC on Darwin does not create ``fat'' object files; it creates
13954 an object file for the single architecture that GCC was built to
13955 target. Apple's GCC on Darwin does create ``fat'' files if multiple
13956 @option{-arch} options are used; it does so by running the compiler or
13957 linker multiple times and joining the results together with
13960 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
13961 @samp{i686}) is determined by the flags that specify the ISA
13962 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
13963 @option{-force_cpusubtype_ALL} option can be used to override this.
13965 The Darwin tools vary in their behavior when presented with an ISA
13966 mismatch. The assembler, @file{as}, only permits instructions to
13967 be used that are valid for the subtype of the file it is generating,
13968 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
13969 The linker for shared libraries, @file{/usr/bin/libtool}, fails
13970 and prints an error if asked to create a shared library with a less
13971 restrictive subtype than its input files (for instance, trying to put
13972 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
13973 for executables, @command{ld}, quietly gives the executable the most
13974 restrictive subtype of any of its input files.
13979 Add the framework directory @var{dir} to the head of the list of
13980 directories to be searched for header files. These directories are
13981 interleaved with those specified by @option{-I} options and are
13982 scanned in a left-to-right order.
13984 A framework directory is a directory with frameworks in it. A
13985 framework is a directory with a @file{Headers} and/or
13986 @file{PrivateHeaders} directory contained directly in it that ends
13987 in @file{.framework}. The name of a framework is the name of this
13988 directory excluding the @file{.framework}. Headers associated with
13989 the framework are found in one of those two directories, with
13990 @file{Headers} being searched first. A subframework is a framework
13991 directory that is in a framework's @file{Frameworks} directory.
13992 Includes of subframework headers can only appear in a header of a
13993 framework that contains the subframework, or in a sibling subframework
13994 header. Two subframeworks are siblings if they occur in the same
13995 framework. A subframework should not have the same name as a
13996 framework; a warning is issued if this is violated. Currently a
13997 subframework cannot have subframeworks; in the future, the mechanism
13998 may be extended to support this. The standard frameworks can be found
13999 in @file{/System/Library/Frameworks} and
14000 @file{/Library/Frameworks}. An example include looks like
14001 @code{#include <Framework/header.h>}, where @file{Framework} denotes
14002 the name of the framework and @file{header.h} is found in the
14003 @file{PrivateHeaders} or @file{Headers} directory.
14005 @item -iframework@var{dir}
14006 @opindex iframework
14007 Like @option{-F} except the directory is a treated as a system
14008 directory. The main difference between this @option{-iframework} and
14009 @option{-F} is that with @option{-iframework} the compiler does not
14010 warn about constructs contained within header files found via
14011 @var{dir}. This option is valid only for the C family of languages.
14015 Emit debugging information for symbols that are used. For stabs
14016 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
14017 This is by default ON@.
14021 Emit debugging information for all symbols and types.
14023 @item -mmacosx-version-min=@var{version}
14024 The earliest version of MacOS X that this executable will run on
14025 is @var{version}. Typical values of @var{version} include @code{10.1},
14026 @code{10.2}, and @code{10.3.9}.
14028 If the compiler was built to use the system's headers by default,
14029 then the default for this option is the system version on which the
14030 compiler is running, otherwise the default is to make choices that
14031 are compatible with as many systems and code bases as possible.
14035 Enable kernel development mode. The @option{-mkernel} option sets
14036 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
14037 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
14038 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
14039 applicable. This mode also sets @option{-mno-altivec},
14040 @option{-msoft-float}, @option{-fno-builtin} and
14041 @option{-mlong-branch} for PowerPC targets.
14043 @item -mone-byte-bool
14044 @opindex mone-byte-bool
14045 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
14046 By default @samp{sizeof(bool)} is @samp{4} when compiling for
14047 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
14048 option has no effect on x86.
14050 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
14051 to generate code that is not binary compatible with code generated
14052 without that switch. Using this switch may require recompiling all
14053 other modules in a program, including system libraries. Use this
14054 switch to conform to a non-default data model.
14056 @item -mfix-and-continue
14057 @itemx -ffix-and-continue
14058 @itemx -findirect-data
14059 @opindex mfix-and-continue
14060 @opindex ffix-and-continue
14061 @opindex findirect-data
14062 Generate code suitable for fast turnaround development, such as to
14063 allow GDB to dynamically load @code{.o} files into already-running
14064 programs. @option{-findirect-data} and @option{-ffix-and-continue}
14065 are provided for backwards compatibility.
14069 Loads all members of static archive libraries.
14070 See man ld(1) for more information.
14072 @item -arch_errors_fatal
14073 @opindex arch_errors_fatal
14074 Cause the errors having to do with files that have the wrong architecture
14077 @item -bind_at_load
14078 @opindex bind_at_load
14079 Causes the output file to be marked such that the dynamic linker will
14080 bind all undefined references when the file is loaded or launched.
14084 Produce a Mach-o bundle format file.
14085 See man ld(1) for more information.
14087 @item -bundle_loader @var{executable}
14088 @opindex bundle_loader
14089 This option specifies the @var{executable} that will load the build
14090 output file being linked. See man ld(1) for more information.
14093 @opindex dynamiclib
14094 When passed this option, GCC produces a dynamic library instead of
14095 an executable when linking, using the Darwin @file{libtool} command.
14097 @item -force_cpusubtype_ALL
14098 @opindex force_cpusubtype_ALL
14099 This causes GCC's output file to have the @var{ALL} subtype, instead of
14100 one controlled by the @option{-mcpu} or @option{-march} option.
14102 @item -allowable_client @var{client_name}
14103 @itemx -client_name
14104 @itemx -compatibility_version
14105 @itemx -current_version
14107 @itemx -dependency-file
14109 @itemx -dylinker_install_name
14111 @itemx -exported_symbols_list
14114 @itemx -flat_namespace
14115 @itemx -force_flat_namespace
14116 @itemx -headerpad_max_install_names
14119 @itemx -install_name
14120 @itemx -keep_private_externs
14121 @itemx -multi_module
14122 @itemx -multiply_defined
14123 @itemx -multiply_defined_unused
14126 @itemx -no_dead_strip_inits_and_terms
14127 @itemx -nofixprebinding
14128 @itemx -nomultidefs
14130 @itemx -noseglinkedit
14131 @itemx -pagezero_size
14133 @itemx -prebind_all_twolevel_modules
14134 @itemx -private_bundle
14136 @itemx -read_only_relocs
14138 @itemx -sectobjectsymbols
14142 @itemx -sectobjectsymbols
14145 @itemx -segs_read_only_addr
14147 @itemx -segs_read_write_addr
14148 @itemx -seg_addr_table
14149 @itemx -seg_addr_table_filename
14150 @itemx -seglinkedit
14152 @itemx -segs_read_only_addr
14153 @itemx -segs_read_write_addr
14154 @itemx -single_module
14156 @itemx -sub_library
14158 @itemx -sub_umbrella
14159 @itemx -twolevel_namespace
14162 @itemx -unexported_symbols_list
14163 @itemx -weak_reference_mismatches
14164 @itemx -whatsloaded
14165 @opindex allowable_client
14166 @opindex client_name
14167 @opindex compatibility_version
14168 @opindex current_version
14169 @opindex dead_strip
14170 @opindex dependency-file
14171 @opindex dylib_file
14172 @opindex dylinker_install_name
14174 @opindex exported_symbols_list
14176 @opindex flat_namespace
14177 @opindex force_flat_namespace
14178 @opindex headerpad_max_install_names
14179 @opindex image_base
14181 @opindex install_name
14182 @opindex keep_private_externs
14183 @opindex multi_module
14184 @opindex multiply_defined
14185 @opindex multiply_defined_unused
14186 @opindex noall_load
14187 @opindex no_dead_strip_inits_and_terms
14188 @opindex nofixprebinding
14189 @opindex nomultidefs
14191 @opindex noseglinkedit
14192 @opindex pagezero_size
14194 @opindex prebind_all_twolevel_modules
14195 @opindex private_bundle
14196 @opindex read_only_relocs
14198 @opindex sectobjectsymbols
14201 @opindex sectcreate
14202 @opindex sectobjectsymbols
14205 @opindex segs_read_only_addr
14206 @opindex segs_read_write_addr
14207 @opindex seg_addr_table
14208 @opindex seg_addr_table_filename
14209 @opindex seglinkedit
14211 @opindex segs_read_only_addr
14212 @opindex segs_read_write_addr
14213 @opindex single_module
14215 @opindex sub_library
14216 @opindex sub_umbrella
14217 @opindex twolevel_namespace
14220 @opindex unexported_symbols_list
14221 @opindex weak_reference_mismatches
14222 @opindex whatsloaded
14223 These options are passed to the Darwin linker. The Darwin linker man page
14224 describes them in detail.
14227 @node DEC Alpha Options
14228 @subsection DEC Alpha Options
14230 These @samp{-m} options are defined for the DEC Alpha implementations:
14233 @item -mno-soft-float
14234 @itemx -msoft-float
14235 @opindex mno-soft-float
14236 @opindex msoft-float
14237 Use (do not use) the hardware floating-point instructions for
14238 floating-point operations. When @option{-msoft-float} is specified,
14239 functions in @file{libgcc.a} are used to perform floating-point
14240 operations. Unless they are replaced by routines that emulate the
14241 floating-point operations, or compiled in such a way as to call such
14242 emulations routines, these routines issue floating-point
14243 operations. If you are compiling for an Alpha without floating-point
14244 operations, you must ensure that the library is built so as not to call
14247 Note that Alpha implementations without floating-point operations are
14248 required to have floating-point registers.
14251 @itemx -mno-fp-regs
14253 @opindex mno-fp-regs
14254 Generate code that uses (does not use) the floating-point register set.
14255 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
14256 register set is not used, floating-point operands are passed in integer
14257 registers as if they were integers and floating-point results are passed
14258 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
14259 so any function with a floating-point argument or return value called by code
14260 compiled with @option{-mno-fp-regs} must also be compiled with that
14263 A typical use of this option is building a kernel that does not use,
14264 and hence need not save and restore, any floating-point registers.
14268 The Alpha architecture implements floating-point hardware optimized for
14269 maximum performance. It is mostly compliant with the IEEE floating-point
14270 standard. However, for full compliance, software assistance is
14271 required. This option generates code fully IEEE-compliant code
14272 @emph{except} that the @var{inexact-flag} is not maintained (see below).
14273 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
14274 defined during compilation. The resulting code is less efficient but is
14275 able to correctly support denormalized numbers and exceptional IEEE
14276 values such as not-a-number and plus/minus infinity. Other Alpha
14277 compilers call this option @option{-ieee_with_no_inexact}.
14279 @item -mieee-with-inexact
14280 @opindex mieee-with-inexact
14281 This is like @option{-mieee} except the generated code also maintains
14282 the IEEE @var{inexact-flag}. Turning on this option causes the
14283 generated code to implement fully-compliant IEEE math. In addition to
14284 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
14285 macro. On some Alpha implementations the resulting code may execute
14286 significantly slower than the code generated by default. Since there is
14287 very little code that depends on the @var{inexact-flag}, you should
14288 normally not specify this option. Other Alpha compilers call this
14289 option @option{-ieee_with_inexact}.
14291 @item -mfp-trap-mode=@var{trap-mode}
14292 @opindex mfp-trap-mode
14293 This option controls what floating-point related traps are enabled.
14294 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
14295 The trap mode can be set to one of four values:
14299 This is the default (normal) setting. The only traps that are enabled
14300 are the ones that cannot be disabled in software (e.g., division by zero
14304 In addition to the traps enabled by @samp{n}, underflow traps are enabled
14308 Like @samp{u}, but the instructions are marked to be safe for software
14309 completion (see Alpha architecture manual for details).
14312 Like @samp{su}, but inexact traps are enabled as well.
14315 @item -mfp-rounding-mode=@var{rounding-mode}
14316 @opindex mfp-rounding-mode
14317 Selects the IEEE rounding mode. Other Alpha compilers call this option
14318 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
14323 Normal IEEE rounding mode. Floating-point numbers are rounded towards
14324 the nearest machine number or towards the even machine number in case
14328 Round towards minus infinity.
14331 Chopped rounding mode. Floating-point numbers are rounded towards zero.
14334 Dynamic rounding mode. A field in the floating-point control register
14335 (@var{fpcr}, see Alpha architecture reference manual) controls the
14336 rounding mode in effect. The C library initializes this register for
14337 rounding towards plus infinity. Thus, unless your program modifies the
14338 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
14341 @item -mtrap-precision=@var{trap-precision}
14342 @opindex mtrap-precision
14343 In the Alpha architecture, floating-point traps are imprecise. This
14344 means without software assistance it is impossible to recover from a
14345 floating trap and program execution normally needs to be terminated.
14346 GCC can generate code that can assist operating system trap handlers
14347 in determining the exact location that caused a floating-point trap.
14348 Depending on the requirements of an application, different levels of
14349 precisions can be selected:
14353 Program precision. This option is the default and means a trap handler
14354 can only identify which program caused a floating-point exception.
14357 Function precision. The trap handler can determine the function that
14358 caused a floating-point exception.
14361 Instruction precision. The trap handler can determine the exact
14362 instruction that caused a floating-point exception.
14365 Other Alpha compilers provide the equivalent options called
14366 @option{-scope_safe} and @option{-resumption_safe}.
14368 @item -mieee-conformant
14369 @opindex mieee-conformant
14370 This option marks the generated code as IEEE conformant. You must not
14371 use this option unless you also specify @option{-mtrap-precision=i} and either
14372 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
14373 is to emit the line @samp{.eflag 48} in the function prologue of the
14374 generated assembly file.
14376 @item -mbuild-constants
14377 @opindex mbuild-constants
14378 Normally GCC examines a 32- or 64-bit integer constant to
14379 see if it can construct it from smaller constants in two or three
14380 instructions. If it cannot, it outputs the constant as a literal and
14381 generates code to load it from the data segment at run time.
14383 Use this option to require GCC to construct @emph{all} integer constants
14384 using code, even if it takes more instructions (the maximum is six).
14386 You typically use this option to build a shared library dynamic
14387 loader. Itself a shared library, it must relocate itself in memory
14388 before it can find the variables and constants in its own data segment.
14406 Indicate whether GCC should generate code to use the optional BWX,
14407 CIX, FIX and MAX instruction sets. The default is to use the instruction
14408 sets supported by the CPU type specified via @option{-mcpu=} option or that
14409 of the CPU on which GCC was built if none is specified.
14412 @itemx -mfloat-ieee
14413 @opindex mfloat-vax
14414 @opindex mfloat-ieee
14415 Generate code that uses (does not use) VAX F and G floating-point
14416 arithmetic instead of IEEE single and double precision.
14418 @item -mexplicit-relocs
14419 @itemx -mno-explicit-relocs
14420 @opindex mexplicit-relocs
14421 @opindex mno-explicit-relocs
14422 Older Alpha assemblers provided no way to generate symbol relocations
14423 except via assembler macros. Use of these macros does not allow
14424 optimal instruction scheduling. GNU binutils as of version 2.12
14425 supports a new syntax that allows the compiler to explicitly mark
14426 which relocations should apply to which instructions. This option
14427 is mostly useful for debugging, as GCC detects the capabilities of
14428 the assembler when it is built and sets the default accordingly.
14431 @itemx -mlarge-data
14432 @opindex msmall-data
14433 @opindex mlarge-data
14434 When @option{-mexplicit-relocs} is in effect, static data is
14435 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
14436 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
14437 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
14438 16-bit relocations off of the @code{$gp} register. This limits the
14439 size of the small data area to 64KB, but allows the variables to be
14440 directly accessed via a single instruction.
14442 The default is @option{-mlarge-data}. With this option the data area
14443 is limited to just below 2GB@. Programs that require more than 2GB of
14444 data must use @code{malloc} or @code{mmap} to allocate the data in the
14445 heap instead of in the program's data segment.
14447 When generating code for shared libraries, @option{-fpic} implies
14448 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
14451 @itemx -mlarge-text
14452 @opindex msmall-text
14453 @opindex mlarge-text
14454 When @option{-msmall-text} is used, the compiler assumes that the
14455 code of the entire program (or shared library) fits in 4MB, and is
14456 thus reachable with a branch instruction. When @option{-msmall-data}
14457 is used, the compiler can assume that all local symbols share the
14458 same @code{$gp} value, and thus reduce the number of instructions
14459 required for a function call from 4 to 1.
14461 The default is @option{-mlarge-text}.
14463 @item -mcpu=@var{cpu_type}
14465 Set the instruction set and instruction scheduling parameters for
14466 machine type @var{cpu_type}. You can specify either the @samp{EV}
14467 style name or the corresponding chip number. GCC supports scheduling
14468 parameters for the EV4, EV5 and EV6 family of processors and
14469 chooses the default values for the instruction set from the processor
14470 you specify. If you do not specify a processor type, GCC defaults
14471 to the processor on which the compiler was built.
14473 Supported values for @var{cpu_type} are
14479 Schedules as an EV4 and has no instruction set extensions.
14483 Schedules as an EV5 and has no instruction set extensions.
14487 Schedules as an EV5 and supports the BWX extension.
14492 Schedules as an EV5 and supports the BWX and MAX extensions.
14496 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
14500 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
14503 Native toolchains also support the value @samp{native},
14504 which selects the best architecture option for the host processor.
14505 @option{-mcpu=native} has no effect if GCC does not recognize
14508 @item -mtune=@var{cpu_type}
14510 Set only the instruction scheduling parameters for machine type
14511 @var{cpu_type}. The instruction set is not changed.
14513 Native toolchains also support the value @samp{native},
14514 which selects the best architecture option for the host processor.
14515 @option{-mtune=native} has no effect if GCC does not recognize
14518 @item -mmemory-latency=@var{time}
14519 @opindex mmemory-latency
14520 Sets the latency the scheduler should assume for typical memory
14521 references as seen by the application. This number is highly
14522 dependent on the memory access patterns used by the application
14523 and the size of the external cache on the machine.
14525 Valid options for @var{time} are
14529 A decimal number representing clock cycles.
14535 The compiler contains estimates of the number of clock cycles for
14536 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
14537 (also called Dcache, Scache, and Bcache), as well as to main memory.
14538 Note that L3 is only valid for EV5.
14544 @subsection FR30 Options
14545 @cindex FR30 Options
14547 These options are defined specifically for the FR30 port.
14551 @item -msmall-model
14552 @opindex msmall-model
14553 Use the small address space model. This can produce smaller code, but
14554 it does assume that all symbolic values and addresses fit into a
14559 Assume that runtime support has been provided and so there is no need
14560 to include the simulator library (@file{libsim.a}) on the linker
14566 @subsection FRV Options
14567 @cindex FRV Options
14573 Only use the first 32 general-purpose registers.
14578 Use all 64 general-purpose registers.
14583 Use only the first 32 floating-point registers.
14588 Use all 64 floating-point registers.
14591 @opindex mhard-float
14593 Use hardware instructions for floating-point operations.
14596 @opindex msoft-float
14598 Use library routines for floating-point operations.
14603 Dynamically allocate condition code registers.
14608 Do not try to dynamically allocate condition code registers, only
14609 use @code{icc0} and @code{fcc0}.
14614 Change ABI to use double word insns.
14619 Do not use double word instructions.
14624 Use floating-point double instructions.
14627 @opindex mno-double
14629 Do not use floating-point double instructions.
14634 Use media instructions.
14639 Do not use media instructions.
14644 Use multiply and add/subtract instructions.
14647 @opindex mno-muladd
14649 Do not use multiply and add/subtract instructions.
14654 Select the FDPIC ABI, which uses function descriptors to represent
14655 pointers to functions. Without any PIC/PIE-related options, it
14656 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
14657 assumes GOT entries and small data are within a 12-bit range from the
14658 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
14659 are computed with 32 bits.
14660 With a @samp{bfin-elf} target, this option implies @option{-msim}.
14663 @opindex minline-plt
14665 Enable inlining of PLT entries in function calls to functions that are
14666 not known to bind locally. It has no effect without @option{-mfdpic}.
14667 It's enabled by default if optimizing for speed and compiling for
14668 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
14669 optimization option such as @option{-O3} or above is present in the
14675 Assume a large TLS segment when generating thread-local code.
14680 Do not assume a large TLS segment when generating thread-local code.
14685 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
14686 that is known to be in read-only sections. It's enabled by default,
14687 except for @option{-fpic} or @option{-fpie}: even though it may help
14688 make the global offset table smaller, it trades 1 instruction for 4.
14689 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
14690 one of which may be shared by multiple symbols, and it avoids the need
14691 for a GOT entry for the referenced symbol, so it's more likely to be a
14692 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
14694 @item -multilib-library-pic
14695 @opindex multilib-library-pic
14697 Link with the (library, not FD) pic libraries. It's implied by
14698 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
14699 @option{-fpic} without @option{-mfdpic}. You should never have to use
14703 @opindex mlinked-fp
14705 Follow the EABI requirement of always creating a frame pointer whenever
14706 a stack frame is allocated. This option is enabled by default and can
14707 be disabled with @option{-mno-linked-fp}.
14710 @opindex mlong-calls
14712 Use indirect addressing to call functions outside the current
14713 compilation unit. This allows the functions to be placed anywhere
14714 within the 32-bit address space.
14716 @item -malign-labels
14717 @opindex malign-labels
14719 Try to align labels to an 8-byte boundary by inserting NOPs into the
14720 previous packet. This option only has an effect when VLIW packing
14721 is enabled. It doesn't create new packets; it merely adds NOPs to
14724 @item -mlibrary-pic
14725 @opindex mlibrary-pic
14727 Generate position-independent EABI code.
14732 Use only the first four media accumulator registers.
14737 Use all eight media accumulator registers.
14742 Pack VLIW instructions.
14747 Do not pack VLIW instructions.
14750 @opindex mno-eflags
14752 Do not mark ABI switches in e_flags.
14755 @opindex mcond-move
14757 Enable the use of conditional-move instructions (default).
14759 This switch is mainly for debugging the compiler and will likely be removed
14760 in a future version.
14762 @item -mno-cond-move
14763 @opindex mno-cond-move
14765 Disable the use of conditional-move instructions.
14767 This switch is mainly for debugging the compiler and will likely be removed
14768 in a future version.
14773 Enable the use of conditional set instructions (default).
14775 This switch is mainly for debugging the compiler and will likely be removed
14776 in a future version.
14781 Disable the use of conditional set instructions.
14783 This switch is mainly for debugging the compiler and will likely be removed
14784 in a future version.
14787 @opindex mcond-exec
14789 Enable the use of conditional execution (default).
14791 This switch is mainly for debugging the compiler and will likely be removed
14792 in a future version.
14794 @item -mno-cond-exec
14795 @opindex mno-cond-exec
14797 Disable the use of conditional execution.
14799 This switch is mainly for debugging the compiler and will likely be removed
14800 in a future version.
14802 @item -mvliw-branch
14803 @opindex mvliw-branch
14805 Run a pass to pack branches into VLIW instructions (default).
14807 This switch is mainly for debugging the compiler and will likely be removed
14808 in a future version.
14810 @item -mno-vliw-branch
14811 @opindex mno-vliw-branch
14813 Do not run a pass to pack branches into VLIW instructions.
14815 This switch is mainly for debugging the compiler and will likely be removed
14816 in a future version.
14818 @item -mmulti-cond-exec
14819 @opindex mmulti-cond-exec
14821 Enable optimization of @code{&&} and @code{||} in conditional execution
14824 This switch is mainly for debugging the compiler and will likely be removed
14825 in a future version.
14827 @item -mno-multi-cond-exec
14828 @opindex mno-multi-cond-exec
14830 Disable optimization of @code{&&} and @code{||} in conditional execution.
14832 This switch is mainly for debugging the compiler and will likely be removed
14833 in a future version.
14835 @item -mnested-cond-exec
14836 @opindex mnested-cond-exec
14838 Enable nested conditional execution optimizations (default).
14840 This switch is mainly for debugging the compiler and will likely be removed
14841 in a future version.
14843 @item -mno-nested-cond-exec
14844 @opindex mno-nested-cond-exec
14846 Disable nested conditional execution optimizations.
14848 This switch is mainly for debugging the compiler and will likely be removed
14849 in a future version.
14851 @item -moptimize-membar
14852 @opindex moptimize-membar
14854 This switch removes redundant @code{membar} instructions from the
14855 compiler-generated code. It is enabled by default.
14857 @item -mno-optimize-membar
14858 @opindex mno-optimize-membar
14860 This switch disables the automatic removal of redundant @code{membar}
14861 instructions from the generated code.
14863 @item -mtomcat-stats
14864 @opindex mtomcat-stats
14866 Cause gas to print out tomcat statistics.
14868 @item -mcpu=@var{cpu}
14871 Select the processor type for which to generate code. Possible values are
14872 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
14873 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
14877 @node GNU/Linux Options
14878 @subsection GNU/Linux Options
14880 These @samp{-m} options are defined for GNU/Linux targets:
14885 Use the GNU C library. This is the default except
14886 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
14890 Use uClibc C library. This is the default on
14891 @samp{*-*-linux-*uclibc*} targets.
14895 Use Bionic C library. This is the default on
14896 @samp{*-*-linux-*android*} targets.
14900 Compile code compatible with Android platform. This is the default on
14901 @samp{*-*-linux-*android*} targets.
14903 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
14904 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
14905 this option makes the GCC driver pass Android-specific options to the linker.
14906 Finally, this option causes the preprocessor macro @code{__ANDROID__}
14909 @item -tno-android-cc
14910 @opindex tno-android-cc
14911 Disable compilation effects of @option{-mandroid}, i.e., do not enable
14912 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
14913 @option{-fno-rtti} by default.
14915 @item -tno-android-ld
14916 @opindex tno-android-ld
14917 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
14918 linking options to the linker.
14922 @node H8/300 Options
14923 @subsection H8/300 Options
14925 These @samp{-m} options are defined for the H8/300 implementations:
14930 Shorten some address references at link time, when possible; uses the
14931 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
14932 ld, Using ld}, for a fuller description.
14936 Generate code for the H8/300H@.
14940 Generate code for the H8S@.
14944 Generate code for the H8S and H8/300H in the normal mode. This switch
14945 must be used either with @option{-mh} or @option{-ms}.
14949 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
14953 Extended registers are stored on stack before execution of function
14954 with monitor attribute. Default option is @option{-mexr}.
14955 This option is valid only for H8S targets.
14959 Extended registers are not stored on stack before execution of function
14960 with monitor attribute. Default option is @option{-mno-exr}.
14961 This option is valid only for H8S targets.
14965 Make @code{int} data 32 bits by default.
14968 @opindex malign-300
14969 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
14970 The default for the H8/300H and H8S is to align longs and floats on
14972 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
14973 This option has no effect on the H8/300.
14977 @subsection HPPA Options
14978 @cindex HPPA Options
14980 These @samp{-m} options are defined for the HPPA family of computers:
14983 @item -march=@var{architecture-type}
14985 Generate code for the specified architecture. The choices for
14986 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
14987 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
14988 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
14989 architecture option for your machine. Code compiled for lower numbered
14990 architectures runs on higher numbered architectures, but not the
14993 @item -mpa-risc-1-0
14994 @itemx -mpa-risc-1-1
14995 @itemx -mpa-risc-2-0
14996 @opindex mpa-risc-1-0
14997 @opindex mpa-risc-1-1
14998 @opindex mpa-risc-2-0
14999 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
15001 @item -mjump-in-delay
15002 @opindex mjump-in-delay
15003 This option is ignored and provided for compatibility purposes only.
15005 @item -mdisable-fpregs
15006 @opindex mdisable-fpregs
15007 Prevent floating-point registers from being used in any manner. This is
15008 necessary for compiling kernels that perform lazy context switching of
15009 floating-point registers. If you use this option and attempt to perform
15010 floating-point operations, the compiler aborts.
15012 @item -mdisable-indexing
15013 @opindex mdisable-indexing
15014 Prevent the compiler from using indexing address modes. This avoids some
15015 rather obscure problems when compiling MIG generated code under MACH@.
15017 @item -mno-space-regs
15018 @opindex mno-space-regs
15019 Generate code that assumes the target has no space registers. This allows
15020 GCC to generate faster indirect calls and use unscaled index address modes.
15022 Such code is suitable for level 0 PA systems and kernels.
15024 @item -mfast-indirect-calls
15025 @opindex mfast-indirect-calls
15026 Generate code that assumes calls never cross space boundaries. This
15027 allows GCC to emit code that performs faster indirect calls.
15029 This option does not work in the presence of shared libraries or nested
15032 @item -mfixed-range=@var{register-range}
15033 @opindex mfixed-range
15034 Generate code treating the given register range as fixed registers.
15035 A fixed register is one that the register allocator cannot use. This is
15036 useful when compiling kernel code. A register range is specified as
15037 two registers separated by a dash. Multiple register ranges can be
15038 specified separated by a comma.
15040 @item -mlong-load-store
15041 @opindex mlong-load-store
15042 Generate 3-instruction load and store sequences as sometimes required by
15043 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
15046 @item -mportable-runtime
15047 @opindex mportable-runtime
15048 Use the portable calling conventions proposed by HP for ELF systems.
15052 Enable the use of assembler directives only GAS understands.
15054 @item -mschedule=@var{cpu-type}
15056 Schedule code according to the constraints for the machine type
15057 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
15058 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
15059 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
15060 proper scheduling option for your machine. The default scheduling is
15064 @opindex mlinker-opt
15065 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
15066 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
15067 linkers in which they give bogus error messages when linking some programs.
15070 @opindex msoft-float
15071 Generate output containing library calls for floating point.
15072 @strong{Warning:} the requisite libraries are not available for all HPPA
15073 targets. Normally the facilities of the machine's usual C compiler are
15074 used, but this cannot be done directly in cross-compilation. You must make
15075 your own arrangements to provide suitable library functions for
15078 @option{-msoft-float} changes the calling convention in the output file;
15079 therefore, it is only useful if you compile @emph{all} of a program with
15080 this option. In particular, you need to compile @file{libgcc.a}, the
15081 library that comes with GCC, with @option{-msoft-float} in order for
15086 Generate the predefine, @code{_SIO}, for server IO@. The default is
15087 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
15088 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
15089 options are available under HP-UX and HI-UX@.
15093 Use options specific to GNU @command{ld}.
15094 This passes @option{-shared} to @command{ld} when
15095 building a shared library. It is the default when GCC is configured,
15096 explicitly or implicitly, with the GNU linker. This option does not
15097 affect which @command{ld} is called; it only changes what parameters
15098 are passed to that @command{ld}.
15099 The @command{ld} that is called is determined by the
15100 @option{--with-ld} configure option, GCC's program search path, and
15101 finally by the user's @env{PATH}. The linker used by GCC can be printed
15102 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
15103 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15107 Use options specific to HP @command{ld}.
15108 This passes @option{-b} to @command{ld} when building
15109 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
15110 links. It is the default when GCC is configured, explicitly or
15111 implicitly, with the HP linker. This option does not affect
15112 which @command{ld} is called; it only changes what parameters are passed to that
15114 The @command{ld} that is called is determined by the @option{--with-ld}
15115 configure option, GCC's program search path, and finally by the user's
15116 @env{PATH}. The linker used by GCC can be printed using @samp{which
15117 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
15118 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15121 @opindex mno-long-calls
15122 Generate code that uses long call sequences. This ensures that a call
15123 is always able to reach linker generated stubs. The default is to generate
15124 long calls only when the distance from the call site to the beginning
15125 of the function or translation unit, as the case may be, exceeds a
15126 predefined limit set by the branch type being used. The limits for
15127 normal calls are 7,600,000 and 240,000 bytes, respectively for the
15128 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
15131 Distances are measured from the beginning of functions when using the
15132 @option{-ffunction-sections} option, or when using the @option{-mgas}
15133 and @option{-mno-portable-runtime} options together under HP-UX with
15136 It is normally not desirable to use this option as it degrades
15137 performance. However, it may be useful in large applications,
15138 particularly when partial linking is used to build the application.
15140 The types of long calls used depends on the capabilities of the
15141 assembler and linker, and the type of code being generated. The
15142 impact on systems that support long absolute calls, and long pic
15143 symbol-difference or pc-relative calls should be relatively small.
15144 However, an indirect call is used on 32-bit ELF systems in pic code
15145 and it is quite long.
15147 @item -munix=@var{unix-std}
15149 Generate compiler predefines and select a startfile for the specified
15150 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
15151 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
15152 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
15153 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
15154 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
15157 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
15158 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
15159 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
15160 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
15161 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
15162 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
15164 It is @emph{important} to note that this option changes the interfaces
15165 for various library routines. It also affects the operational behavior
15166 of the C library. Thus, @emph{extreme} care is needed in using this
15169 Library code that is intended to operate with more than one UNIX
15170 standard must test, set and restore the variable @var{__xpg4_extended_mask}
15171 as appropriate. Most GNU software doesn't provide this capability.
15175 Suppress the generation of link options to search libdld.sl when the
15176 @option{-static} option is specified on HP-UX 10 and later.
15180 The HP-UX implementation of setlocale in libc has a dependency on
15181 libdld.sl. There isn't an archive version of libdld.sl. Thus,
15182 when the @option{-static} option is specified, special link options
15183 are needed to resolve this dependency.
15185 On HP-UX 10 and later, the GCC driver adds the necessary options to
15186 link with libdld.sl when the @option{-static} option is specified.
15187 This causes the resulting binary to be dynamic. On the 64-bit port,
15188 the linkers generate dynamic binaries by default in any case. The
15189 @option{-nolibdld} option can be used to prevent the GCC driver from
15190 adding these link options.
15194 Add support for multithreading with the @dfn{dce thread} library
15195 under HP-UX@. This option sets flags for both the preprocessor and
15199 @node i386 and x86-64 Options
15200 @subsection Intel 386 and AMD x86-64 Options
15201 @cindex i386 Options
15202 @cindex x86-64 Options
15203 @cindex Intel 386 Options
15204 @cindex AMD x86-64 Options
15206 These @samp{-m} options are defined for the i386 and x86-64 family of
15211 @item -march=@var{cpu-type}
15213 Generate instructions for the machine type @var{cpu-type}. In contrast to
15214 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
15215 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
15216 to generate code that may not run at all on processors other than the one
15217 indicated. Specifying @option{-march=@var{cpu-type}} implies
15218 @option{-mtune=@var{cpu-type}}.
15220 The choices for @var{cpu-type} are:
15224 This selects the CPU to generate code for at compilation time by determining
15225 the processor type of the compiling machine. Using @option{-march=native}
15226 enables all instruction subsets supported by the local machine (hence
15227 the result might not run on different machines). Using @option{-mtune=native}
15228 produces code optimized for the local machine under the constraints
15229 of the selected instruction set.
15232 Original Intel i386 CPU@.
15235 Intel i486 CPU@. (No scheduling is implemented for this chip.)
15239 Intel Pentium CPU with no MMX support.
15242 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
15245 Intel Pentium Pro CPU@.
15248 When used with @option{-march}, the Pentium Pro
15249 instruction set is used, so the code runs on all i686 family chips.
15250 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
15253 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
15258 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
15262 Intel Pentium M; low-power version of Intel Pentium III CPU
15263 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
15267 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
15270 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
15274 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
15275 SSE2 and SSE3 instruction set support.
15278 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
15279 instruction set support.
15282 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
15283 SSE4.1, SSE4.2 and POPCNT instruction set support.
15286 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
15287 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
15290 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
15291 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
15294 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
15295 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
15296 instruction set support.
15299 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
15300 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
15301 BMI, BMI2 and F16C instruction set support.
15304 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
15305 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
15306 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
15309 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
15310 instruction set support.
15313 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
15314 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
15317 AMD K6 CPU with MMX instruction set support.
15321 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
15324 @itemx athlon-tbird
15325 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
15331 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
15332 instruction set support.
15338 Processors based on the AMD K8 core with x86-64 instruction set support,
15339 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
15340 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
15341 instruction set extensions.)
15344 @itemx opteron-sse3
15345 @itemx athlon64-sse3
15346 Improved versions of AMD K8 cores with SSE3 instruction set support.
15350 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
15351 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
15352 instruction set extensions.)
15355 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
15356 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
15357 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
15359 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
15360 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
15361 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
15364 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
15365 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
15366 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
15367 64-bit instruction set extensions.
15369 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
15370 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
15371 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
15372 SSE4.2, ABM and 64-bit instruction set extensions.
15375 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
15376 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
15377 instruction set extensions.)
15380 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
15381 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
15382 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
15385 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
15389 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
15390 instruction set support.
15393 VIA C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
15394 implemented for this chip.)
15397 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
15399 implemented for this chip.)
15402 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
15405 @item -mtune=@var{cpu-type}
15407 Tune to @var{cpu-type} everything applicable about the generated code, except
15408 for the ABI and the set of available instructions.
15409 While picking a specific @var{cpu-type} schedules things appropriately
15410 for that particular chip, the compiler does not generate any code that
15411 cannot run on the default machine type unless you use a
15412 @option{-march=@var{cpu-type}} option.
15413 For example, if GCC is configured for i686-pc-linux-gnu
15414 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
15415 but still runs on i686 machines.
15417 The choices for @var{cpu-type} are the same as for @option{-march}.
15418 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
15422 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
15423 If you know the CPU on which your code will run, then you should use
15424 the corresponding @option{-mtune} or @option{-march} option instead of
15425 @option{-mtune=generic}. But, if you do not know exactly what CPU users
15426 of your application will have, then you should use this option.
15428 As new processors are deployed in the marketplace, the behavior of this
15429 option will change. Therefore, if you upgrade to a newer version of
15430 GCC, code generation controlled by this option will change to reflect
15432 that are most common at the time that version of GCC is released.
15434 There is no @option{-march=generic} option because @option{-march}
15435 indicates the instruction set the compiler can use, and there is no
15436 generic instruction set applicable to all processors. In contrast,
15437 @option{-mtune} indicates the processor (or, in this case, collection of
15438 processors) for which the code is optimized.
15441 Produce code optimized for the most current Intel processors, which are
15442 Haswell and Silvermont for this version of GCC. If you know the CPU
15443 on which your code will run, then you should use the corresponding
15444 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
15445 But, if you want your application performs better on both Haswell and
15446 Silvermont, then you should use this option.
15448 As new Intel processors are deployed in the marketplace, the behavior of
15449 this option will change. Therefore, if you upgrade to a newer version of
15450 GCC, code generation controlled by this option will change to reflect
15451 the most current Intel processors at the time that version of GCC is
15454 There is no @option{-march=intel} option because @option{-march} indicates
15455 the instruction set the compiler can use, and there is no common
15456 instruction set applicable to all processors. In contrast,
15457 @option{-mtune} indicates the processor (or, in this case, collection of
15458 processors) for which the code is optimized.
15461 @item -mcpu=@var{cpu-type}
15463 A deprecated synonym for @option{-mtune}.
15465 @item -mfpmath=@var{unit}
15467 Generate floating-point arithmetic for selected unit @var{unit}. The choices
15468 for @var{unit} are:
15472 Use the standard 387 floating-point coprocessor present on the majority of chips and
15473 emulated otherwise. Code compiled with this option runs almost everywhere.
15474 The temporary results are computed in 80-bit precision instead of the precision
15475 specified by the type, resulting in slightly different results compared to most
15476 of other chips. See @option{-ffloat-store} for more detailed description.
15478 This is the default choice for i386 compiler.
15481 Use scalar floating-point instructions present in the SSE instruction set.
15482 This instruction set is supported by Pentium III and newer chips,
15483 and in the AMD line
15484 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
15485 instruction set supports only single-precision arithmetic, thus the double and
15486 extended-precision arithmetic are still done using 387. A later version, present
15487 only in Pentium 4 and AMD x86-64 chips, supports double-precision
15490 For the i386 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
15491 or @option{-msse2} switches to enable SSE extensions and make this option
15492 effective. For the x86-64 compiler, these extensions are enabled by default.
15494 The resulting code should be considerably faster in the majority of cases and avoid
15495 the numerical instability problems of 387 code, but may break some existing
15496 code that expects temporaries to be 80 bits.
15498 This is the default choice for the x86-64 compiler.
15503 Attempt to utilize both instruction sets at once. This effectively doubles the
15504 amount of available registers, and on chips with separate execution units for
15505 387 and SSE the execution resources too. Use this option with care, as it is
15506 still experimental, because the GCC register allocator does not model separate
15507 functional units well, resulting in unstable performance.
15510 @item -masm=@var{dialect}
15511 @opindex masm=@var{dialect}
15512 Output assembly instructions using selected @var{dialect}. Supported
15513 choices are @samp{intel} or @samp{att} (the default). Darwin does
15514 not support @samp{intel}.
15517 @itemx -mno-ieee-fp
15519 @opindex mno-ieee-fp
15520 Control whether or not the compiler uses IEEE floating-point
15521 comparisons. These correctly handle the case where the result of a
15522 comparison is unordered.
15525 @opindex msoft-float
15526 Generate output containing library calls for floating point.
15528 @strong{Warning:} the requisite libraries are not part of GCC@.
15529 Normally the facilities of the machine's usual C compiler are used, but
15530 this can't be done directly in cross-compilation. You must make your
15531 own arrangements to provide suitable library functions for
15534 On machines where a function returns floating-point results in the 80387
15535 register stack, some floating-point opcodes may be emitted even if
15536 @option{-msoft-float} is used.
15538 @item -mno-fp-ret-in-387
15539 @opindex mno-fp-ret-in-387
15540 Do not use the FPU registers for return values of functions.
15542 The usual calling convention has functions return values of types
15543 @code{float} and @code{double} in an FPU register, even if there
15544 is no FPU@. The idea is that the operating system should emulate
15547 The option @option{-mno-fp-ret-in-387} causes such values to be returned
15548 in ordinary CPU registers instead.
15550 @item -mno-fancy-math-387
15551 @opindex mno-fancy-math-387
15552 Some 387 emulators do not support the @code{sin}, @code{cos} and
15553 @code{sqrt} instructions for the 387. Specify this option to avoid
15554 generating those instructions. This option is the default on FreeBSD,
15555 OpenBSD and NetBSD@. This option is overridden when @option{-march}
15556 indicates that the target CPU always has an FPU and so the
15557 instruction does not need emulation. These
15558 instructions are not generated unless you also use the
15559 @option{-funsafe-math-optimizations} switch.
15561 @item -malign-double
15562 @itemx -mno-align-double
15563 @opindex malign-double
15564 @opindex mno-align-double
15565 Control whether GCC aligns @code{double}, @code{long double}, and
15566 @code{long long} variables on a two-word boundary or a one-word
15567 boundary. Aligning @code{double} variables on a two-word boundary
15568 produces code that runs somewhat faster on a Pentium at the
15569 expense of more memory.
15571 On x86-64, @option{-malign-double} is enabled by default.
15573 @strong{Warning:} if you use the @option{-malign-double} switch,
15574 structures containing the above types are aligned differently than
15575 the published application binary interface specifications for the 386
15576 and are not binary compatible with structures in code compiled
15577 without that switch.
15579 @item -m96bit-long-double
15580 @itemx -m128bit-long-double
15581 @opindex m96bit-long-double
15582 @opindex m128bit-long-double
15583 These switches control the size of @code{long double} type. The i386
15584 application binary interface specifies the size to be 96 bits,
15585 so @option{-m96bit-long-double} is the default in 32-bit mode.
15587 Modern architectures (Pentium and newer) prefer @code{long double}
15588 to be aligned to an 8- or 16-byte boundary. In arrays or structures
15589 conforming to the ABI, this is not possible. So specifying
15590 @option{-m128bit-long-double} aligns @code{long double}
15591 to a 16-byte boundary by padding the @code{long double} with an additional
15594 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
15595 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
15597 Notice that neither of these options enable any extra precision over the x87
15598 standard of 80 bits for a @code{long double}.
15600 @strong{Warning:} if you override the default value for your target ABI, this
15601 changes the size of
15602 structures and arrays containing @code{long double} variables,
15603 as well as modifying the function calling convention for functions taking
15604 @code{long double}. Hence they are not binary-compatible
15605 with code compiled without that switch.
15607 @item -mlong-double-64
15608 @itemx -mlong-double-80
15609 @itemx -mlong-double-128
15610 @opindex mlong-double-64
15611 @opindex mlong-double-80
15612 @opindex mlong-double-128
15613 These switches control the size of @code{long double} type. A size
15614 of 64 bits makes the @code{long double} type equivalent to the @code{double}
15615 type. This is the default for 32-bit Bionic C library. A size
15616 of 128 bits makes the @code{long double} type equivalent to the
15617 @code{__float128} type. This is the default for 64-bit Bionic C library.
15619 @strong{Warning:} if you override the default value for your target ABI, this
15620 changes the size of
15621 structures and arrays containing @code{long double} variables,
15622 as well as modifying the function calling convention for functions taking
15623 @code{long double}. Hence they are not binary-compatible
15624 with code compiled without that switch.
15626 @item -mlarge-data-threshold=@var{threshold}
15627 @opindex mlarge-data-threshold
15628 When @option{-mcmodel=medium} is specified, data objects larger than
15629 @var{threshold} are placed in the large data section. This value must be the
15630 same across all objects linked into the binary, and defaults to 65535.
15634 Use a different function-calling convention, in which functions that
15635 take a fixed number of arguments return with the @code{ret @var{num}}
15636 instruction, which pops their arguments while returning. This saves one
15637 instruction in the caller since there is no need to pop the arguments
15640 You can specify that an individual function is called with this calling
15641 sequence with the function attribute @samp{stdcall}. You can also
15642 override the @option{-mrtd} option by using the function attribute
15643 @samp{cdecl}. @xref{Function Attributes}.
15645 @strong{Warning:} this calling convention is incompatible with the one
15646 normally used on Unix, so you cannot use it if you need to call
15647 libraries compiled with the Unix compiler.
15649 Also, you must provide function prototypes for all functions that
15650 take variable numbers of arguments (including @code{printf});
15651 otherwise incorrect code is generated for calls to those
15654 In addition, seriously incorrect code results if you call a
15655 function with too many arguments. (Normally, extra arguments are
15656 harmlessly ignored.)
15658 @item -mregparm=@var{num}
15660 Control how many registers are used to pass integer arguments. By
15661 default, no registers are used to pass arguments, and at most 3
15662 registers can be used. You can control this behavior for a specific
15663 function by using the function attribute @samp{regparm}.
15664 @xref{Function Attributes}.
15666 @strong{Warning:} if you use this switch, and
15667 @var{num} is nonzero, then you must build all modules with the same
15668 value, including any libraries. This includes the system libraries and
15672 @opindex msseregparm
15673 Use SSE register passing conventions for float and double arguments
15674 and return values. You can control this behavior for a specific
15675 function by using the function attribute @samp{sseregparm}.
15676 @xref{Function Attributes}.
15678 @strong{Warning:} if you use this switch then you must build all
15679 modules with the same value, including any libraries. This includes
15680 the system libraries and startup modules.
15682 @item -mvect8-ret-in-mem
15683 @opindex mvect8-ret-in-mem
15684 Return 8-byte vectors in memory instead of MMX registers. This is the
15685 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
15686 Studio compilers until version 12. Later compiler versions (starting
15687 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
15688 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
15689 you need to remain compatible with existing code produced by those
15690 previous compiler versions or older versions of GCC@.
15699 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
15700 is specified, the significands of results of floating-point operations are
15701 rounded to 24 bits (single precision); @option{-mpc64} rounds the
15702 significands of results of floating-point operations to 53 bits (double
15703 precision) and @option{-mpc80} rounds the significands of results of
15704 floating-point operations to 64 bits (extended double precision), which is
15705 the default. When this option is used, floating-point operations in higher
15706 precisions are not available to the programmer without setting the FPU
15707 control word explicitly.
15709 Setting the rounding of floating-point operations to less than the default
15710 80 bits can speed some programs by 2% or more. Note that some mathematical
15711 libraries assume that extended-precision (80-bit) floating-point operations
15712 are enabled by default; routines in such libraries could suffer significant
15713 loss of accuracy, typically through so-called ``catastrophic cancellation'',
15714 when this option is used to set the precision to less than extended precision.
15716 @item -mstackrealign
15717 @opindex mstackrealign
15718 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
15719 option generates an alternate prologue and epilogue that realigns the
15720 run-time stack if necessary. This supports mixing legacy codes that keep
15721 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
15722 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
15723 applicable to individual functions.
15725 @item -mpreferred-stack-boundary=@var{num}
15726 @opindex mpreferred-stack-boundary
15727 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
15728 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
15729 the default is 4 (16 bytes or 128 bits).
15731 @strong{Warning:} When generating code for the x86-64 architecture with
15732 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
15733 used to keep the stack boundary aligned to 8 byte boundary. Since
15734 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
15735 intended to be used in controlled environment where stack space is
15736 important limitation. This option will lead to wrong code when functions
15737 compiled with 16 byte stack alignment (such as functions from a standard
15738 library) are called with misaligned stack. In this case, SSE
15739 instructions may lead to misaligned memory access traps. In addition,
15740 variable arguments will be handled incorrectly for 16 byte aligned
15741 objects (including x87 long double and __int128), leading to wrong
15742 results. You must build all modules with
15743 @option{-mpreferred-stack-boundary=3}, including any libraries. This
15744 includes the system libraries and startup modules.
15746 @item -mincoming-stack-boundary=@var{num}
15747 @opindex mincoming-stack-boundary
15748 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
15749 boundary. If @option{-mincoming-stack-boundary} is not specified,
15750 the one specified by @option{-mpreferred-stack-boundary} is used.
15752 On Pentium and Pentium Pro, @code{double} and @code{long double} values
15753 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
15754 suffer significant run time performance penalties. On Pentium III, the
15755 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
15756 properly if it is not 16-byte aligned.
15758 To ensure proper alignment of this values on the stack, the stack boundary
15759 must be as aligned as that required by any value stored on the stack.
15760 Further, every function must be generated such that it keeps the stack
15761 aligned. Thus calling a function compiled with a higher preferred
15762 stack boundary from a function compiled with a lower preferred stack
15763 boundary most likely misaligns the stack. It is recommended that
15764 libraries that use callbacks always use the default setting.
15766 This extra alignment does consume extra stack space, and generally
15767 increases code size. Code that is sensitive to stack space usage, such
15768 as embedded systems and operating system kernels, may want to reduce the
15769 preferred alignment to @option{-mpreferred-stack-boundary=2}.
15793 @itemx -mno-avx512f
15796 @itemx -mno-avx512pf
15798 @itemx -mno-avx512er
15800 @itemx -mno-avx512cd
15808 @itemx -mno-clflsuhopt
15811 @itemx -mno-fsgsbase
15818 @itemx -mprefetchwt1
15819 @itemx -mno-prefetchwt1
15859 These switches enable or disable the use of instructions in the MMX, SSE,
15860 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
15861 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
15862 BMI, BMI2, FXSR, XSAVE, XSAVEOPT, LZCNT, RTM, MPX or 3DNow!@:
15863 extended instruction sets.
15864 These extensions are also available as built-in functions: see
15865 @ref{X86 Built-in Functions}, for details of the functions enabled and
15866 disabled by these switches.
15868 To generate SSE/SSE2 instructions automatically from floating-point
15869 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
15871 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
15872 generates new AVX instructions or AVX equivalence for all SSEx instructions
15875 These options enable GCC to use these extended instructions in
15876 generated code, even without @option{-mfpmath=sse}. Applications that
15877 perform run-time CPU detection must compile separate files for each
15878 supported architecture, using the appropriate flags. In particular,
15879 the file containing the CPU detection code should be compiled without
15882 @item -mdump-tune-features
15883 @opindex mdump-tune-features
15884 This option instructs GCC to dump the names of the x86 performance
15885 tuning features and default settings. The names can be used in
15886 @option{-mtune-ctrl=@var{feature-list}}.
15888 @item -mtune-ctrl=@var{feature-list}
15889 @opindex mtune-ctrl=@var{feature-list}
15890 This option is used to do fine grain control of x86 code generation features.
15891 @var{feature-list} is a comma separated list of @var{feature} names. See also
15892 @option{-mdump-tune-features}. When specified, the @var{feature} will be turned
15893 on if it is not preceded with @code{^}, otherwise, it will be turned off.
15894 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
15895 developers. Using it may lead to code paths not covered by testing and can
15896 potentially result in compiler ICEs or runtime errors.
15899 @opindex mno-default
15900 This option instructs GCC to turn off all tunable features. See also
15901 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
15905 This option instructs GCC to emit a @code{cld} instruction in the prologue
15906 of functions that use string instructions. String instructions depend on
15907 the DF flag to select between autoincrement or autodecrement mode. While the
15908 ABI specifies the DF flag to be cleared on function entry, some operating
15909 systems violate this specification by not clearing the DF flag in their
15910 exception dispatchers. The exception handler can be invoked with the DF flag
15911 set, which leads to wrong direction mode when string instructions are used.
15912 This option can be enabled by default on 32-bit x86 targets by configuring
15913 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
15914 instructions can be suppressed with the @option{-mno-cld} compiler option
15918 @opindex mvzeroupper
15919 This option instructs GCC to emit a @code{vzeroupper} instruction
15920 before a transfer of control flow out of the function to minimize
15921 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
15924 @item -mprefer-avx128
15925 @opindex mprefer-avx128
15926 This option instructs GCC to use 128-bit AVX instructions instead of
15927 256-bit AVX instructions in the auto-vectorizer.
15931 This option enables GCC to generate @code{CMPXCHG16B} instructions.
15932 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
15933 (or oword) data types.
15934 This is useful for high-resolution counters that can be updated
15935 by multiple processors (or cores). This instruction is generated as part of
15936 atomic built-in functions: see @ref{__sync Builtins} or
15937 @ref{__atomic Builtins} for details.
15941 This option enables generation of @code{SAHF} instructions in 64-bit code.
15942 Early Intel Pentium 4 CPUs with Intel 64 support,
15943 prior to the introduction of Pentium 4 G1 step in December 2005,
15944 lacked the @code{LAHF} and @code{SAHF} instructions
15945 which were supported by AMD64.
15946 These are load and store instructions, respectively, for certain status flags.
15947 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
15948 @code{drem}, and @code{remainder} built-in functions;
15949 see @ref{Other Builtins} for details.
15953 This option enables use of the @code{movbe} instruction to implement
15954 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
15958 This option enables built-in functions @code{__builtin_ia32_crc32qi},
15959 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
15960 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
15964 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
15965 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
15966 with an additional Newton-Raphson step
15967 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
15968 (and their vectorized
15969 variants) for single-precision floating-point arguments. These instructions
15970 are generated only when @option{-funsafe-math-optimizations} is enabled
15971 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
15972 Note that while the throughput of the sequence is higher than the throughput
15973 of the non-reciprocal instruction, the precision of the sequence can be
15974 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
15976 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
15977 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
15978 combination), and doesn't need @option{-mrecip}.
15980 Also note that GCC emits the above sequence with additional Newton-Raphson step
15981 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
15982 already with @option{-ffast-math} (or the above option combination), and
15983 doesn't need @option{-mrecip}.
15985 @item -mrecip=@var{opt}
15986 @opindex mrecip=opt
15987 This option controls which reciprocal estimate instructions
15988 may be used. @var{opt} is a comma-separated list of options, which may
15989 be preceded by a @samp{!} to invert the option:
15993 Enable all estimate instructions.
15996 Enable the default instructions, equivalent to @option{-mrecip}.
15999 Disable all estimate instructions, equivalent to @option{-mno-recip}.
16002 Enable the approximation for scalar division.
16005 Enable the approximation for vectorized division.
16008 Enable the approximation for scalar square root.
16011 Enable the approximation for vectorized square root.
16014 So, for example, @option{-mrecip=all,!sqrt} enables
16015 all of the reciprocal approximations, except for square root.
16017 @item -mveclibabi=@var{type}
16018 @opindex mveclibabi
16019 Specifies the ABI type to use for vectorizing intrinsics using an
16020 external library. Supported values for @var{type} are @samp{svml}
16021 for the Intel short
16022 vector math library and @samp{acml} for the AMD math core library.
16023 To use this option, both @option{-ftree-vectorize} and
16024 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
16025 ABI-compatible library must be specified at link time.
16027 GCC currently emits calls to @code{vmldExp2},
16028 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
16029 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
16030 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
16031 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
16032 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
16033 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
16034 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
16035 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
16036 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
16037 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
16038 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
16039 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
16040 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
16041 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
16042 when @option{-mveclibabi=acml} is used.
16044 @item -mabi=@var{name}
16046 Generate code for the specified calling convention. Permissible values
16047 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
16048 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
16049 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
16050 You can control this behavior for a specific function by
16051 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
16052 @xref{Function Attributes}.
16054 @item -mtls-dialect=@var{type}
16055 @opindex mtls-dialect
16056 Generate code to access thread-local storage using the @samp{gnu} or
16057 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
16058 @samp{gnu2} is more efficient, but it may add compile- and run-time
16059 requirements that cannot be satisfied on all systems.
16062 @itemx -mno-push-args
16063 @opindex mpush-args
16064 @opindex mno-push-args
16065 Use PUSH operations to store outgoing parameters. This method is shorter
16066 and usually equally fast as method using SUB/MOV operations and is enabled
16067 by default. In some cases disabling it may improve performance because of
16068 improved scheduling and reduced dependencies.
16070 @item -maccumulate-outgoing-args
16071 @opindex maccumulate-outgoing-args
16072 If enabled, the maximum amount of space required for outgoing arguments is
16073 computed in the function prologue. This is faster on most modern CPUs
16074 because of reduced dependencies, improved scheduling and reduced stack usage
16075 when the preferred stack boundary is not equal to 2. The drawback is a notable
16076 increase in code size. This switch implies @option{-mno-push-args}.
16080 Support thread-safe exception handling on MinGW. Programs that rely
16081 on thread-safe exception handling must compile and link all code with the
16082 @option{-mthreads} option. When compiling, @option{-mthreads} defines
16083 @code{-D_MT}; when linking, it links in a special thread helper library
16084 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
16086 @item -mno-align-stringops
16087 @opindex mno-align-stringops
16088 Do not align the destination of inlined string operations. This switch reduces
16089 code size and improves performance in case the destination is already aligned,
16090 but GCC doesn't know about it.
16092 @item -minline-all-stringops
16093 @opindex minline-all-stringops
16094 By default GCC inlines string operations only when the destination is
16095 known to be aligned to least a 4-byte boundary.
16096 This enables more inlining and increases code
16097 size, but may improve performance of code that depends on fast
16098 @code{memcpy}, @code{strlen},
16099 and @code{memset} for short lengths.
16101 @item -minline-stringops-dynamically
16102 @opindex minline-stringops-dynamically
16103 For string operations of unknown size, use run-time checks with
16104 inline code for small blocks and a library call for large blocks.
16106 @item -mstringop-strategy=@var{alg}
16107 @opindex mstringop-strategy=@var{alg}
16108 Override the internal decision heuristic for the particular algorithm to use
16109 for inlining string operations. The allowed values for @var{alg} are:
16115 Expand using i386 @code{rep} prefix of the specified size.
16119 @itemx unrolled_loop
16120 Expand into an inline loop.
16123 Always use a library call.
16126 @item -mmemcpy-strategy=@var{strategy}
16127 @opindex mmemcpy-strategy=@var{strategy}
16128 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
16129 should be inlined and what inline algorithm to use when the expected size
16130 of the copy operation is known. @var{strategy}
16131 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
16132 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
16133 the max byte size with which inline algorithm @var{alg} is allowed. For the last
16134 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
16135 in the list must be specified in increasing order. The minimal byte size for
16136 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
16139 @item -mmemset-strategy=@var{strategy}
16140 @opindex mmemset-strategy=@var{strategy}
16141 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
16142 @code{__builtin_memset} expansion.
16144 @item -momit-leaf-frame-pointer
16145 @opindex momit-leaf-frame-pointer
16146 Don't keep the frame pointer in a register for leaf functions. This
16147 avoids the instructions to save, set up, and restore frame pointers and
16148 makes an extra register available in leaf functions. The option
16149 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
16150 which might make debugging harder.
16152 @item -mtls-direct-seg-refs
16153 @itemx -mno-tls-direct-seg-refs
16154 @opindex mtls-direct-seg-refs
16155 Controls whether TLS variables may be accessed with offsets from the
16156 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
16157 or whether the thread base pointer must be added. Whether or not this
16158 is valid depends on the operating system, and whether it maps the
16159 segment to cover the entire TLS area.
16161 For systems that use the GNU C Library, the default is on.
16164 @itemx -mno-sse2avx
16166 Specify that the assembler should encode SSE instructions with VEX
16167 prefix. The option @option{-mavx} turns this on by default.
16172 If profiling is active (@option{-pg}), put the profiling
16173 counter call before the prologue.
16174 Note: On x86 architectures the attribute @code{ms_hook_prologue}
16175 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
16177 @item -mrecord-mcount
16178 @itemx -mno-record-mcount
16179 @opindex mrecord-mcount
16180 If profiling is active (@option{-pg}), generate a __mcount_loc section
16181 that contains pointers to each profiling call. This is useful for
16182 automatically patching and out calls.
16185 @itemx -mno-nop-mcount
16186 @opindex mnop-mcount
16187 If profiling is active (@option{-pg}), generate the calls to
16188 the profiling functions as nops. This is useful when they
16189 should be patched in later dynamically. This is likely only
16190 useful together with @option{-mrecord-mcount}.
16193 @itemx -mno-8bit-idiv
16195 On some processors, like Intel Atom, 8-bit unsigned integer divide is
16196 much faster than 32-bit/64-bit integer divide. This option generates a
16197 run-time check. If both dividend and divisor are within range of 0
16198 to 255, 8-bit unsigned integer divide is used instead of
16199 32-bit/64-bit integer divide.
16201 @item -mavx256-split-unaligned-load
16202 @itemx -mavx256-split-unaligned-store
16203 @opindex avx256-split-unaligned-load
16204 @opindex avx256-split-unaligned-store
16205 Split 32-byte AVX unaligned load and store.
16207 @item -mstack-protector-guard=@var{guard}
16208 @opindex mstack-protector-guard=@var{guard}
16209 Generate stack protection code using canary at @var{guard}. Supported
16210 locations are @samp{global} for global canary or @samp{tls} for per-thread
16211 canary in the TLS block (the default). This option has effect only when
16212 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
16216 These @samp{-m} switches are supported in addition to the above
16217 on x86-64 processors in 64-bit environments.
16228 Generate code for a 16-bit, 32-bit or 64-bit environment.
16229 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
16231 generates code that runs on any i386 system.
16233 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
16234 types to 64 bits, and generates code for the x86-64 architecture.
16235 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
16236 and @option{-mdynamic-no-pic} options.
16238 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
16240 generates code for the x86-64 architecture.
16242 The @option{-m16} option is the same as @option{-m32}, except for that
16243 it outputs the @code{.code16gcc} assembly directive at the beginning of
16244 the assembly output so that the binary can run in 16-bit mode.
16246 @item -mno-red-zone
16247 @opindex mno-red-zone
16248 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
16249 by the x86-64 ABI; it is a 128-byte area beyond the location of the
16250 stack pointer that is not modified by signal or interrupt handlers
16251 and therefore can be used for temporary data without adjusting the stack
16252 pointer. The flag @option{-mno-red-zone} disables this red zone.
16254 @item -mcmodel=small
16255 @opindex mcmodel=small
16256 Generate code for the small code model: the program and its symbols must
16257 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
16258 Programs can be statically or dynamically linked. This is the default
16261 @item -mcmodel=kernel
16262 @opindex mcmodel=kernel
16263 Generate code for the kernel code model. The kernel runs in the
16264 negative 2 GB of the address space.
16265 This model has to be used for Linux kernel code.
16267 @item -mcmodel=medium
16268 @opindex mcmodel=medium
16269 Generate code for the medium model: the program is linked in the lower 2
16270 GB of the address space. Small symbols are also placed there. Symbols
16271 with sizes larger than @option{-mlarge-data-threshold} are put into
16272 large data or BSS sections and can be located above 2GB. Programs can
16273 be statically or dynamically linked.
16275 @item -mcmodel=large
16276 @opindex mcmodel=large
16277 Generate code for the large model. This model makes no assumptions
16278 about addresses and sizes of sections.
16280 @item -maddress-mode=long
16281 @opindex maddress-mode=long
16282 Generate code for long address mode. This is only supported for 64-bit
16283 and x32 environments. It is the default address mode for 64-bit
16286 @item -maddress-mode=short
16287 @opindex maddress-mode=short
16288 Generate code for short address mode. This is only supported for 32-bit
16289 and x32 environments. It is the default address mode for 32-bit and
16293 @node i386 and x86-64 Windows Options
16294 @subsection i386 and x86-64 Windows Options
16295 @cindex i386 and x86-64 Windows Options
16297 These additional options are available for Microsoft Windows targets:
16303 specifies that a console application is to be generated, by
16304 instructing the linker to set the PE header subsystem type
16305 required for console applications.
16306 This option is available for Cygwin and MinGW targets and is
16307 enabled by default on those targets.
16311 This option is available for Cygwin and MinGW targets. It
16312 specifies that a DLL---a dynamic link library---is to be
16313 generated, enabling the selection of the required runtime
16314 startup object and entry point.
16316 @item -mnop-fun-dllimport
16317 @opindex mnop-fun-dllimport
16318 This option is available for Cygwin and MinGW targets. It
16319 specifies that the @code{dllimport} attribute should be ignored.
16323 This option is available for MinGW targets. It specifies
16324 that MinGW-specific thread support is to be used.
16328 This option is available for MinGW-w64 targets. It causes
16329 the @code{UNICODE} preprocessor macro to be predefined, and
16330 chooses Unicode-capable runtime startup code.
16334 This option is available for Cygwin and MinGW targets. It
16335 specifies that the typical Microsoft Windows predefined macros are to
16336 be set in the pre-processor, but does not influence the choice
16337 of runtime library/startup code.
16341 This option is available for Cygwin and MinGW targets. It
16342 specifies that a GUI application is to be generated by
16343 instructing the linker to set the PE header subsystem type
16346 @item -fno-set-stack-executable
16347 @opindex fno-set-stack-executable
16348 This option is available for MinGW targets. It specifies that
16349 the executable flag for the stack used by nested functions isn't
16350 set. This is necessary for binaries running in kernel mode of
16351 Microsoft Windows, as there the User32 API, which is used to set executable
16352 privileges, isn't available.
16354 @item -fwritable-relocated-rdata
16355 @opindex fno-writable-relocated-rdata
16356 This option is available for MinGW and Cygwin targets. It specifies
16357 that relocated-data in read-only section is put into .data
16358 section. This is a necessary for older runtimes not supporting
16359 modification of .rdata sections for pseudo-relocation.
16361 @item -mpe-aligned-commons
16362 @opindex mpe-aligned-commons
16363 This option is available for Cygwin and MinGW targets. It
16364 specifies that the GNU extension to the PE file format that
16365 permits the correct alignment of COMMON variables should be
16366 used when generating code. It is enabled by default if
16367 GCC detects that the target assembler found during configuration
16368 supports the feature.
16371 See also under @ref{i386 and x86-64 Options} for standard options.
16373 @node IA-64 Options
16374 @subsection IA-64 Options
16375 @cindex IA-64 Options
16377 These are the @samp{-m} options defined for the Intel IA-64 architecture.
16381 @opindex mbig-endian
16382 Generate code for a big-endian target. This is the default for HP-UX@.
16384 @item -mlittle-endian
16385 @opindex mlittle-endian
16386 Generate code for a little-endian target. This is the default for AIX5
16392 @opindex mno-gnu-as
16393 Generate (or don't) code for the GNU assembler. This is the default.
16394 @c Also, this is the default if the configure option @option{--with-gnu-as}
16400 @opindex mno-gnu-ld
16401 Generate (or don't) code for the GNU linker. This is the default.
16402 @c Also, this is the default if the configure option @option{--with-gnu-ld}
16407 Generate code that does not use a global pointer register. The result
16408 is not position independent code, and violates the IA-64 ABI@.
16410 @item -mvolatile-asm-stop
16411 @itemx -mno-volatile-asm-stop
16412 @opindex mvolatile-asm-stop
16413 @opindex mno-volatile-asm-stop
16414 Generate (or don't) a stop bit immediately before and after volatile asm
16417 @item -mregister-names
16418 @itemx -mno-register-names
16419 @opindex mregister-names
16420 @opindex mno-register-names
16421 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
16422 the stacked registers. This may make assembler output more readable.
16428 Disable (or enable) optimizations that use the small data section. This may
16429 be useful for working around optimizer bugs.
16431 @item -mconstant-gp
16432 @opindex mconstant-gp
16433 Generate code that uses a single constant global pointer value. This is
16434 useful when compiling kernel code.
16438 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
16439 This is useful when compiling firmware code.
16441 @item -minline-float-divide-min-latency
16442 @opindex minline-float-divide-min-latency
16443 Generate code for inline divides of floating-point values
16444 using the minimum latency algorithm.
16446 @item -minline-float-divide-max-throughput
16447 @opindex minline-float-divide-max-throughput
16448 Generate code for inline divides of floating-point values
16449 using the maximum throughput algorithm.
16451 @item -mno-inline-float-divide
16452 @opindex mno-inline-float-divide
16453 Do not generate inline code for divides of floating-point values.
16455 @item -minline-int-divide-min-latency
16456 @opindex minline-int-divide-min-latency
16457 Generate code for inline divides of integer values
16458 using the minimum latency algorithm.
16460 @item -minline-int-divide-max-throughput
16461 @opindex minline-int-divide-max-throughput
16462 Generate code for inline divides of integer values
16463 using the maximum throughput algorithm.
16465 @item -mno-inline-int-divide
16466 @opindex mno-inline-int-divide
16467 Do not generate inline code for divides of integer values.
16469 @item -minline-sqrt-min-latency
16470 @opindex minline-sqrt-min-latency
16471 Generate code for inline square roots
16472 using the minimum latency algorithm.
16474 @item -minline-sqrt-max-throughput
16475 @opindex minline-sqrt-max-throughput
16476 Generate code for inline square roots
16477 using the maximum throughput algorithm.
16479 @item -mno-inline-sqrt
16480 @opindex mno-inline-sqrt
16481 Do not generate inline code for @code{sqrt}.
16484 @itemx -mno-fused-madd
16485 @opindex mfused-madd
16486 @opindex mno-fused-madd
16487 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
16488 instructions. The default is to use these instructions.
16490 @item -mno-dwarf2-asm
16491 @itemx -mdwarf2-asm
16492 @opindex mno-dwarf2-asm
16493 @opindex mdwarf2-asm
16494 Don't (or do) generate assembler code for the DWARF 2 line number debugging
16495 info. This may be useful when not using the GNU assembler.
16497 @item -mearly-stop-bits
16498 @itemx -mno-early-stop-bits
16499 @opindex mearly-stop-bits
16500 @opindex mno-early-stop-bits
16501 Allow stop bits to be placed earlier than immediately preceding the
16502 instruction that triggered the stop bit. This can improve instruction
16503 scheduling, but does not always do so.
16505 @item -mfixed-range=@var{register-range}
16506 @opindex mfixed-range
16507 Generate code treating the given register range as fixed registers.
16508 A fixed register is one that the register allocator cannot use. This is
16509 useful when compiling kernel code. A register range is specified as
16510 two registers separated by a dash. Multiple register ranges can be
16511 specified separated by a comma.
16513 @item -mtls-size=@var{tls-size}
16515 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
16518 @item -mtune=@var{cpu-type}
16520 Tune the instruction scheduling for a particular CPU, Valid values are
16521 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
16522 and @samp{mckinley}.
16528 Generate code for a 32-bit or 64-bit environment.
16529 The 32-bit environment sets int, long and pointer to 32 bits.
16530 The 64-bit environment sets int to 32 bits and long and pointer
16531 to 64 bits. These are HP-UX specific flags.
16533 @item -mno-sched-br-data-spec
16534 @itemx -msched-br-data-spec
16535 @opindex mno-sched-br-data-spec
16536 @opindex msched-br-data-spec
16537 (Dis/En)able data speculative scheduling before reload.
16538 This results in generation of @code{ld.a} instructions and
16539 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16540 The default is 'disable'.
16542 @item -msched-ar-data-spec
16543 @itemx -mno-sched-ar-data-spec
16544 @opindex msched-ar-data-spec
16545 @opindex mno-sched-ar-data-spec
16546 (En/Dis)able data speculative scheduling after reload.
16547 This results in generation of @code{ld.a} instructions and
16548 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16549 The default is 'enable'.
16551 @item -mno-sched-control-spec
16552 @itemx -msched-control-spec
16553 @opindex mno-sched-control-spec
16554 @opindex msched-control-spec
16555 (Dis/En)able control speculative scheduling. This feature is
16556 available only during region scheduling (i.e.@: before reload).
16557 This results in generation of the @code{ld.s} instructions and
16558 the corresponding check instructions @code{chk.s}.
16559 The default is 'disable'.
16561 @item -msched-br-in-data-spec
16562 @itemx -mno-sched-br-in-data-spec
16563 @opindex msched-br-in-data-spec
16564 @opindex mno-sched-br-in-data-spec
16565 (En/Dis)able speculative scheduling of the instructions that
16566 are dependent on the data speculative loads before reload.
16567 This is effective only with @option{-msched-br-data-spec} enabled.
16568 The default is 'enable'.
16570 @item -msched-ar-in-data-spec
16571 @itemx -mno-sched-ar-in-data-spec
16572 @opindex msched-ar-in-data-spec
16573 @opindex mno-sched-ar-in-data-spec
16574 (En/Dis)able speculative scheduling of the instructions that
16575 are dependent on the data speculative loads after reload.
16576 This is effective only with @option{-msched-ar-data-spec} enabled.
16577 The default is 'enable'.
16579 @item -msched-in-control-spec
16580 @itemx -mno-sched-in-control-spec
16581 @opindex msched-in-control-spec
16582 @opindex mno-sched-in-control-spec
16583 (En/Dis)able speculative scheduling of the instructions that
16584 are dependent on the control speculative loads.
16585 This is effective only with @option{-msched-control-spec} enabled.
16586 The default is 'enable'.
16588 @item -mno-sched-prefer-non-data-spec-insns
16589 @itemx -msched-prefer-non-data-spec-insns
16590 @opindex mno-sched-prefer-non-data-spec-insns
16591 @opindex msched-prefer-non-data-spec-insns
16592 If enabled, data-speculative instructions are chosen for schedule
16593 only if there are no other choices at the moment. This makes
16594 the use of the data speculation much more conservative.
16595 The default is 'disable'.
16597 @item -mno-sched-prefer-non-control-spec-insns
16598 @itemx -msched-prefer-non-control-spec-insns
16599 @opindex mno-sched-prefer-non-control-spec-insns
16600 @opindex msched-prefer-non-control-spec-insns
16601 If enabled, control-speculative instructions are chosen for schedule
16602 only if there are no other choices at the moment. This makes
16603 the use of the control speculation much more conservative.
16604 The default is 'disable'.
16606 @item -mno-sched-count-spec-in-critical-path
16607 @itemx -msched-count-spec-in-critical-path
16608 @opindex mno-sched-count-spec-in-critical-path
16609 @opindex msched-count-spec-in-critical-path
16610 If enabled, speculative dependencies are considered during
16611 computation of the instructions priorities. This makes the use of the
16612 speculation a bit more conservative.
16613 The default is 'disable'.
16615 @item -msched-spec-ldc
16616 @opindex msched-spec-ldc
16617 Use a simple data speculation check. This option is on by default.
16619 @item -msched-control-spec-ldc
16620 @opindex msched-spec-ldc
16621 Use a simple check for control speculation. This option is on by default.
16623 @item -msched-stop-bits-after-every-cycle
16624 @opindex msched-stop-bits-after-every-cycle
16625 Place a stop bit after every cycle when scheduling. This option is on
16628 @item -msched-fp-mem-deps-zero-cost
16629 @opindex msched-fp-mem-deps-zero-cost
16630 Assume that floating-point stores and loads are not likely to cause a conflict
16631 when placed into the same instruction group. This option is disabled by
16634 @item -msel-sched-dont-check-control-spec
16635 @opindex msel-sched-dont-check-control-spec
16636 Generate checks for control speculation in selective scheduling.
16637 This flag is disabled by default.
16639 @item -msched-max-memory-insns=@var{max-insns}
16640 @opindex msched-max-memory-insns
16641 Limit on the number of memory insns per instruction group, giving lower
16642 priority to subsequent memory insns attempting to schedule in the same
16643 instruction group. Frequently useful to prevent cache bank conflicts.
16644 The default value is 1.
16646 @item -msched-max-memory-insns-hard-limit
16647 @opindex msched-max-memory-insns-hard-limit
16648 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
16649 disallowing more than that number in an instruction group.
16650 Otherwise, the limit is ``soft'', meaning that non-memory operations
16651 are preferred when the limit is reached, but memory operations may still
16657 @subsection LM32 Options
16658 @cindex LM32 options
16660 These @option{-m} options are defined for the LatticeMico32 architecture:
16663 @item -mbarrel-shift-enabled
16664 @opindex mbarrel-shift-enabled
16665 Enable barrel-shift instructions.
16667 @item -mdivide-enabled
16668 @opindex mdivide-enabled
16669 Enable divide and modulus instructions.
16671 @item -mmultiply-enabled
16672 @opindex multiply-enabled
16673 Enable multiply instructions.
16675 @item -msign-extend-enabled
16676 @opindex msign-extend-enabled
16677 Enable sign extend instructions.
16679 @item -muser-enabled
16680 @opindex muser-enabled
16681 Enable user-defined instructions.
16686 @subsection M32C Options
16687 @cindex M32C options
16690 @item -mcpu=@var{name}
16692 Select the CPU for which code is generated. @var{name} may be one of
16693 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
16694 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
16695 the M32C/80 series.
16699 Specifies that the program will be run on the simulator. This causes
16700 an alternate runtime library to be linked in which supports, for
16701 example, file I/O@. You must not use this option when generating
16702 programs that will run on real hardware; you must provide your own
16703 runtime library for whatever I/O functions are needed.
16705 @item -memregs=@var{number}
16707 Specifies the number of memory-based pseudo-registers GCC uses
16708 during code generation. These pseudo-registers are used like real
16709 registers, so there is a tradeoff between GCC's ability to fit the
16710 code into available registers, and the performance penalty of using
16711 memory instead of registers. Note that all modules in a program must
16712 be compiled with the same value for this option. Because of that, you
16713 must not use this option with GCC's default runtime libraries.
16717 @node M32R/D Options
16718 @subsection M32R/D Options
16719 @cindex M32R/D options
16721 These @option{-m} options are defined for Renesas M32R/D architectures:
16726 Generate code for the M32R/2@.
16730 Generate code for the M32R/X@.
16734 Generate code for the M32R@. This is the default.
16736 @item -mmodel=small
16737 @opindex mmodel=small
16738 Assume all objects live in the lower 16MB of memory (so that their addresses
16739 can be loaded with the @code{ld24} instruction), and assume all subroutines
16740 are reachable with the @code{bl} instruction.
16741 This is the default.
16743 The addressability of a particular object can be set with the
16744 @code{model} attribute.
16746 @item -mmodel=medium
16747 @opindex mmodel=medium
16748 Assume objects may be anywhere in the 32-bit address space (the compiler
16749 generates @code{seth/add3} instructions to load their addresses), and
16750 assume all subroutines are reachable with the @code{bl} instruction.
16752 @item -mmodel=large
16753 @opindex mmodel=large
16754 Assume objects may be anywhere in the 32-bit address space (the compiler
16755 generates @code{seth/add3} instructions to load their addresses), and
16756 assume subroutines may not be reachable with the @code{bl} instruction
16757 (the compiler generates the much slower @code{seth/add3/jl}
16758 instruction sequence).
16761 @opindex msdata=none
16762 Disable use of the small data area. Variables are put into
16763 one of @samp{.data}, @samp{.bss}, or @samp{.rodata} (unless the
16764 @code{section} attribute has been specified).
16765 This is the default.
16767 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
16768 Objects may be explicitly put in the small data area with the
16769 @code{section} attribute using one of these sections.
16771 @item -msdata=sdata
16772 @opindex msdata=sdata
16773 Put small global and static data in the small data area, but do not
16774 generate special code to reference them.
16777 @opindex msdata=use
16778 Put small global and static data in the small data area, and generate
16779 special instructions to reference them.
16783 @cindex smaller data references
16784 Put global and static objects less than or equal to @var{num} bytes
16785 into the small data or BSS sections instead of the normal data or BSS
16786 sections. The default value of @var{num} is 8.
16787 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
16788 for this option to have any effect.
16790 All modules should be compiled with the same @option{-G @var{num}} value.
16791 Compiling with different values of @var{num} may or may not work; if it
16792 doesn't the linker gives an error message---incorrect code is not
16797 Makes the M32R-specific code in the compiler display some statistics
16798 that might help in debugging programs.
16800 @item -malign-loops
16801 @opindex malign-loops
16802 Align all loops to a 32-byte boundary.
16804 @item -mno-align-loops
16805 @opindex mno-align-loops
16806 Do not enforce a 32-byte alignment for loops. This is the default.
16808 @item -missue-rate=@var{number}
16809 @opindex missue-rate=@var{number}
16810 Issue @var{number} instructions per cycle. @var{number} can only be 1
16813 @item -mbranch-cost=@var{number}
16814 @opindex mbranch-cost=@var{number}
16815 @var{number} can only be 1 or 2. If it is 1 then branches are
16816 preferred over conditional code, if it is 2, then the opposite applies.
16818 @item -mflush-trap=@var{number}
16819 @opindex mflush-trap=@var{number}
16820 Specifies the trap number to use to flush the cache. The default is
16821 12. Valid numbers are between 0 and 15 inclusive.
16823 @item -mno-flush-trap
16824 @opindex mno-flush-trap
16825 Specifies that the cache cannot be flushed by using a trap.
16827 @item -mflush-func=@var{name}
16828 @opindex mflush-func=@var{name}
16829 Specifies the name of the operating system function to call to flush
16830 the cache. The default is @emph{_flush_cache}, but a function call
16831 is only used if a trap is not available.
16833 @item -mno-flush-func
16834 @opindex mno-flush-func
16835 Indicates that there is no OS function for flushing the cache.
16839 @node M680x0 Options
16840 @subsection M680x0 Options
16841 @cindex M680x0 options
16843 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
16844 The default settings depend on which architecture was selected when
16845 the compiler was configured; the defaults for the most common choices
16849 @item -march=@var{arch}
16851 Generate code for a specific M680x0 or ColdFire instruction set
16852 architecture. Permissible values of @var{arch} for M680x0
16853 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
16854 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
16855 architectures are selected according to Freescale's ISA classification
16856 and the permissible values are: @samp{isaa}, @samp{isaaplus},
16857 @samp{isab} and @samp{isac}.
16859 GCC defines a macro @samp{__mcf@var{arch}__} whenever it is generating
16860 code for a ColdFire target. The @var{arch} in this macro is one of the
16861 @option{-march} arguments given above.
16863 When used together, @option{-march} and @option{-mtune} select code
16864 that runs on a family of similar processors but that is optimized
16865 for a particular microarchitecture.
16867 @item -mcpu=@var{cpu}
16869 Generate code for a specific M680x0 or ColdFire processor.
16870 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
16871 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
16872 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
16873 below, which also classifies the CPUs into families:
16875 @multitable @columnfractions 0.20 0.80
16876 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
16877 @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}
16878 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
16879 @item @samp{5206e} @tab @samp{5206e}
16880 @item @samp{5208} @tab @samp{5207} @samp{5208}
16881 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
16882 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
16883 @item @samp{5216} @tab @samp{5214} @samp{5216}
16884 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
16885 @item @samp{5225} @tab @samp{5224} @samp{5225}
16886 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
16887 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
16888 @item @samp{5249} @tab @samp{5249}
16889 @item @samp{5250} @tab @samp{5250}
16890 @item @samp{5271} @tab @samp{5270} @samp{5271}
16891 @item @samp{5272} @tab @samp{5272}
16892 @item @samp{5275} @tab @samp{5274} @samp{5275}
16893 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
16894 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
16895 @item @samp{5307} @tab @samp{5307}
16896 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
16897 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
16898 @item @samp{5407} @tab @samp{5407}
16899 @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}
16902 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
16903 @var{arch} is compatible with @var{cpu}. Other combinations of
16904 @option{-mcpu} and @option{-march} are rejected.
16906 GCC defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
16907 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
16908 where the value of @var{family} is given by the table above.
16910 @item -mtune=@var{tune}
16912 Tune the code for a particular microarchitecture within the
16913 constraints set by @option{-march} and @option{-mcpu}.
16914 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
16915 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
16916 and @samp{cpu32}. The ColdFire microarchitectures
16917 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
16919 You can also use @option{-mtune=68020-40} for code that needs
16920 to run relatively well on 68020, 68030 and 68040 targets.
16921 @option{-mtune=68020-60} is similar but includes 68060 targets
16922 as well. These two options select the same tuning decisions as
16923 @option{-m68020-40} and @option{-m68020-60} respectively.
16925 GCC defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
16926 when tuning for 680x0 architecture @var{arch}. It also defines
16927 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
16928 option is used. If GCC is tuning for a range of architectures,
16929 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
16930 it defines the macros for every architecture in the range.
16932 GCC also defines the macro @samp{__m@var{uarch}__} when tuning for
16933 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
16934 of the arguments given above.
16940 Generate output for a 68000. This is the default
16941 when the compiler is configured for 68000-based systems.
16942 It is equivalent to @option{-march=68000}.
16944 Use this option for microcontrollers with a 68000 or EC000 core,
16945 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
16949 Generate output for a 68010. This is the default
16950 when the compiler is configured for 68010-based systems.
16951 It is equivalent to @option{-march=68010}.
16957 Generate output for a 68020. This is the default
16958 when the compiler is configured for 68020-based systems.
16959 It is equivalent to @option{-march=68020}.
16963 Generate output for a 68030. This is the default when the compiler is
16964 configured for 68030-based systems. It is equivalent to
16965 @option{-march=68030}.
16969 Generate output for a 68040. This is the default when the compiler is
16970 configured for 68040-based systems. It is equivalent to
16971 @option{-march=68040}.
16973 This option inhibits the use of 68881/68882 instructions that have to be
16974 emulated by software on the 68040. Use this option if your 68040 does not
16975 have code to emulate those instructions.
16979 Generate output for a 68060. This is the default when the compiler is
16980 configured for 68060-based systems. It is equivalent to
16981 @option{-march=68060}.
16983 This option inhibits the use of 68020 and 68881/68882 instructions that
16984 have to be emulated by software on the 68060. Use this option if your 68060
16985 does not have code to emulate those instructions.
16989 Generate output for a CPU32. This is the default
16990 when the compiler is configured for CPU32-based systems.
16991 It is equivalent to @option{-march=cpu32}.
16993 Use this option for microcontrollers with a
16994 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
16995 68336, 68340, 68341, 68349 and 68360.
16999 Generate output for a 520X ColdFire CPU@. This is the default
17000 when the compiler is configured for 520X-based systems.
17001 It is equivalent to @option{-mcpu=5206}, and is now deprecated
17002 in favor of that option.
17004 Use this option for microcontroller with a 5200 core, including
17005 the MCF5202, MCF5203, MCF5204 and MCF5206.
17009 Generate output for a 5206e ColdFire CPU@. The option is now
17010 deprecated in favor of the equivalent @option{-mcpu=5206e}.
17014 Generate output for a member of the ColdFire 528X family.
17015 The option is now deprecated in favor of the equivalent
17016 @option{-mcpu=528x}.
17020 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
17021 in favor of the equivalent @option{-mcpu=5307}.
17025 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
17026 in favor of the equivalent @option{-mcpu=5407}.
17030 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
17031 This includes use of hardware floating-point instructions.
17032 The option is equivalent to @option{-mcpu=547x}, and is now
17033 deprecated in favor of that option.
17037 Generate output for a 68040, without using any of the new instructions.
17038 This results in code that can run relatively efficiently on either a
17039 68020/68881 or a 68030 or a 68040. The generated code does use the
17040 68881 instructions that are emulated on the 68040.
17042 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
17046 Generate output for a 68060, without using any of the new instructions.
17047 This results in code that can run relatively efficiently on either a
17048 68020/68881 or a 68030 or a 68040. The generated code does use the
17049 68881 instructions that are emulated on the 68060.
17051 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
17055 @opindex mhard-float
17057 Generate floating-point instructions. This is the default for 68020
17058 and above, and for ColdFire devices that have an FPU@. It defines the
17059 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
17060 on ColdFire targets.
17063 @opindex msoft-float
17064 Do not generate floating-point instructions; use library calls instead.
17065 This is the default for 68000, 68010, and 68832 targets. It is also
17066 the default for ColdFire devices that have no FPU.
17072 Generate (do not generate) ColdFire hardware divide and remainder
17073 instructions. If @option{-march} is used without @option{-mcpu},
17074 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
17075 architectures. Otherwise, the default is taken from the target CPU
17076 (either the default CPU, or the one specified by @option{-mcpu}). For
17077 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
17078 @option{-mcpu=5206e}.
17080 GCC defines the macro @samp{__mcfhwdiv__} when this option is enabled.
17084 Consider type @code{int} to be 16 bits wide, like @code{short int}.
17085 Additionally, parameters passed on the stack are also aligned to a
17086 16-bit boundary even on targets whose API mandates promotion to 32-bit.
17090 Do not consider type @code{int} to be 16 bits wide. This is the default.
17093 @itemx -mno-bitfield
17094 @opindex mnobitfield
17095 @opindex mno-bitfield
17096 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
17097 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
17101 Do use the bit-field instructions. The @option{-m68020} option implies
17102 @option{-mbitfield}. This is the default if you use a configuration
17103 designed for a 68020.
17107 Use a different function-calling convention, in which functions
17108 that take a fixed number of arguments return with the @code{rtd}
17109 instruction, which pops their arguments while returning. This
17110 saves one instruction in the caller since there is no need to pop
17111 the arguments there.
17113 This calling convention is incompatible with the one normally
17114 used on Unix, so you cannot use it if you need to call libraries
17115 compiled with the Unix compiler.
17117 Also, you must provide function prototypes for all functions that
17118 take variable numbers of arguments (including @code{printf});
17119 otherwise incorrect code is generated for calls to those
17122 In addition, seriously incorrect code results if you call a
17123 function with too many arguments. (Normally, extra arguments are
17124 harmlessly ignored.)
17126 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
17127 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
17131 Do not use the calling conventions selected by @option{-mrtd}.
17132 This is the default.
17135 @itemx -mno-align-int
17136 @opindex malign-int
17137 @opindex mno-align-int
17138 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
17139 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
17140 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
17141 Aligning variables on 32-bit boundaries produces code that runs somewhat
17142 faster on processors with 32-bit busses at the expense of more memory.
17144 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
17145 aligns structures containing the above types differently than
17146 most published application binary interface specifications for the m68k.
17150 Use the pc-relative addressing mode of the 68000 directly, instead of
17151 using a global offset table. At present, this option implies @option{-fpic},
17152 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
17153 not presently supported with @option{-mpcrel}, though this could be supported for
17154 68020 and higher processors.
17156 @item -mno-strict-align
17157 @itemx -mstrict-align
17158 @opindex mno-strict-align
17159 @opindex mstrict-align
17160 Do not (do) assume that unaligned memory references are handled by
17164 Generate code that allows the data segment to be located in a different
17165 area of memory from the text segment. This allows for execute-in-place in
17166 an environment without virtual memory management. This option implies
17169 @item -mno-sep-data
17170 Generate code that assumes that the data segment follows the text segment.
17171 This is the default.
17173 @item -mid-shared-library
17174 Generate code that supports shared libraries via the library ID method.
17175 This allows for execute-in-place and shared libraries in an environment
17176 without virtual memory management. This option implies @option{-fPIC}.
17178 @item -mno-id-shared-library
17179 Generate code that doesn't assume ID-based shared libraries are being used.
17180 This is the default.
17182 @item -mshared-library-id=n
17183 Specifies the identification number of the ID-based shared library being
17184 compiled. Specifying a value of 0 generates more compact code; specifying
17185 other values forces the allocation of that number to the current
17186 library, but is no more space- or time-efficient than omitting this option.
17192 When generating position-independent code for ColdFire, generate code
17193 that works if the GOT has more than 8192 entries. This code is
17194 larger and slower than code generated without this option. On M680x0
17195 processors, this option is not needed; @option{-fPIC} suffices.
17197 GCC normally uses a single instruction to load values from the GOT@.
17198 While this is relatively efficient, it only works if the GOT
17199 is smaller than about 64k. Anything larger causes the linker
17200 to report an error such as:
17202 @cindex relocation truncated to fit (ColdFire)
17204 relocation truncated to fit: R_68K_GOT16O foobar
17207 If this happens, you should recompile your code with @option{-mxgot}.
17208 It should then work with very large GOTs. However, code generated with
17209 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
17210 the value of a global symbol.
17212 Note that some linkers, including newer versions of the GNU linker,
17213 can create multiple GOTs and sort GOT entries. If you have such a linker,
17214 you should only need to use @option{-mxgot} when compiling a single
17215 object file that accesses more than 8192 GOT entries. Very few do.
17217 These options have no effect unless GCC is generating
17218 position-independent code.
17222 @node MCore Options
17223 @subsection MCore Options
17224 @cindex MCore options
17226 These are the @samp{-m} options defined for the Motorola M*Core
17232 @itemx -mno-hardlit
17234 @opindex mno-hardlit
17235 Inline constants into the code stream if it can be done in two
17236 instructions or less.
17242 Use the divide instruction. (Enabled by default).
17244 @item -mrelax-immediate
17245 @itemx -mno-relax-immediate
17246 @opindex mrelax-immediate
17247 @opindex mno-relax-immediate
17248 Allow arbitrary-sized immediates in bit operations.
17250 @item -mwide-bitfields
17251 @itemx -mno-wide-bitfields
17252 @opindex mwide-bitfields
17253 @opindex mno-wide-bitfields
17254 Always treat bit-fields as @code{int}-sized.
17256 @item -m4byte-functions
17257 @itemx -mno-4byte-functions
17258 @opindex m4byte-functions
17259 @opindex mno-4byte-functions
17260 Force all functions to be aligned to a 4-byte boundary.
17262 @item -mcallgraph-data
17263 @itemx -mno-callgraph-data
17264 @opindex mcallgraph-data
17265 @opindex mno-callgraph-data
17266 Emit callgraph information.
17269 @itemx -mno-slow-bytes
17270 @opindex mslow-bytes
17271 @opindex mno-slow-bytes
17272 Prefer word access when reading byte quantities.
17274 @item -mlittle-endian
17275 @itemx -mbig-endian
17276 @opindex mlittle-endian
17277 @opindex mbig-endian
17278 Generate code for a little-endian target.
17284 Generate code for the 210 processor.
17288 Assume that runtime support has been provided and so omit the
17289 simulator library (@file{libsim.a)} from the linker command line.
17291 @item -mstack-increment=@var{size}
17292 @opindex mstack-increment
17293 Set the maximum amount for a single stack increment operation. Large
17294 values can increase the speed of programs that contain functions
17295 that need a large amount of stack space, but they can also trigger a
17296 segmentation fault if the stack is extended too much. The default
17302 @subsection MeP Options
17303 @cindex MeP options
17309 Enables the @code{abs} instruction, which is the absolute difference
17310 between two registers.
17314 Enables all the optional instructions---average, multiply, divide, bit
17315 operations, leading zero, absolute difference, min/max, clip, and
17321 Enables the @code{ave} instruction, which computes the average of two
17324 @item -mbased=@var{n}
17326 Variables of size @var{n} bytes or smaller are placed in the
17327 @code{.based} section by default. Based variables use the @code{$tp}
17328 register as a base register, and there is a 128-byte limit to the
17329 @code{.based} section.
17333 Enables the bit operation instructions---bit test (@code{btstm}), set
17334 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
17335 test-and-set (@code{tas}).
17337 @item -mc=@var{name}
17339 Selects which section constant data is placed in. @var{name} may
17340 be @code{tiny}, @code{near}, or @code{far}.
17344 Enables the @code{clip} instruction. Note that @code{-mclip} is not
17345 useful unless you also provide @code{-mminmax}.
17347 @item -mconfig=@var{name}
17349 Selects one of the built-in core configurations. Each MeP chip has
17350 one or more modules in it; each module has a core CPU and a variety of
17351 coprocessors, optional instructions, and peripherals. The
17352 @code{MeP-Integrator} tool, not part of GCC, provides these
17353 configurations through this option; using this option is the same as
17354 using all the corresponding command-line options. The default
17355 configuration is @code{default}.
17359 Enables the coprocessor instructions. By default, this is a 32-bit
17360 coprocessor. Note that the coprocessor is normally enabled via the
17361 @code{-mconfig=} option.
17365 Enables the 32-bit coprocessor's instructions.
17369 Enables the 64-bit coprocessor's instructions.
17373 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
17377 Causes constant variables to be placed in the @code{.near} section.
17381 Enables the @code{div} and @code{divu} instructions.
17385 Generate big-endian code.
17389 Generate little-endian code.
17391 @item -mio-volatile
17392 @opindex mio-volatile
17393 Tells the compiler that any variable marked with the @code{io}
17394 attribute is to be considered volatile.
17398 Causes variables to be assigned to the @code{.far} section by default.
17402 Enables the @code{leadz} (leading zero) instruction.
17406 Causes variables to be assigned to the @code{.near} section by default.
17410 Enables the @code{min} and @code{max} instructions.
17414 Enables the multiplication and multiply-accumulate instructions.
17418 Disables all the optional instructions enabled by @code{-mall-opts}.
17422 Enables the @code{repeat} and @code{erepeat} instructions, used for
17423 low-overhead looping.
17427 Causes all variables to default to the @code{.tiny} section. Note
17428 that there is a 65536-byte limit to this section. Accesses to these
17429 variables use the @code{%gp} base register.
17433 Enables the saturation instructions. Note that the compiler does not
17434 currently generate these itself, but this option is included for
17435 compatibility with other tools, like @code{as}.
17439 Link the SDRAM-based runtime instead of the default ROM-based runtime.
17443 Link the simulator run-time libraries.
17447 Link the simulator runtime libraries, excluding built-in support
17448 for reset and exception vectors and tables.
17452 Causes all functions to default to the @code{.far} section. Without
17453 this option, functions default to the @code{.near} section.
17455 @item -mtiny=@var{n}
17457 Variables that are @var{n} bytes or smaller are allocated to the
17458 @code{.tiny} section. These variables use the @code{$gp} base
17459 register. The default for this option is 4, but note that there's a
17460 65536-byte limit to the @code{.tiny} section.
17464 @node MicroBlaze Options
17465 @subsection MicroBlaze Options
17466 @cindex MicroBlaze Options
17471 @opindex msoft-float
17472 Use software emulation for floating point (default).
17475 @opindex mhard-float
17476 Use hardware floating-point instructions.
17480 Do not optimize block moves, use @code{memcpy}.
17482 @item -mno-clearbss
17483 @opindex mno-clearbss
17484 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
17486 @item -mcpu=@var{cpu-type}
17488 Use features of, and schedule code for, the given CPU.
17489 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
17490 where @var{X} is a major version, @var{YY} is the minor version, and
17491 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
17492 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
17494 @item -mxl-soft-mul
17495 @opindex mxl-soft-mul
17496 Use software multiply emulation (default).
17498 @item -mxl-soft-div
17499 @opindex mxl-soft-div
17500 Use software emulation for divides (default).
17502 @item -mxl-barrel-shift
17503 @opindex mxl-barrel-shift
17504 Use the hardware barrel shifter.
17506 @item -mxl-pattern-compare
17507 @opindex mxl-pattern-compare
17508 Use pattern compare instructions.
17510 @item -msmall-divides
17511 @opindex msmall-divides
17512 Use table lookup optimization for small signed integer divisions.
17514 @item -mxl-stack-check
17515 @opindex mxl-stack-check
17516 This option is deprecated. Use @option{-fstack-check} instead.
17519 @opindex mxl-gp-opt
17520 Use GP-relative @code{.sdata}/@code{.sbss} sections.
17522 @item -mxl-multiply-high
17523 @opindex mxl-multiply-high
17524 Use multiply high instructions for high part of 32x32 multiply.
17526 @item -mxl-float-convert
17527 @opindex mxl-float-convert
17528 Use hardware floating-point conversion instructions.
17530 @item -mxl-float-sqrt
17531 @opindex mxl-float-sqrt
17532 Use hardware floating-point square root instruction.
17535 @opindex mbig-endian
17536 Generate code for a big-endian target.
17538 @item -mlittle-endian
17539 @opindex mlittle-endian
17540 Generate code for a little-endian target.
17543 @opindex mxl-reorder
17544 Use reorder instructions (swap and byte reversed load/store).
17546 @item -mxl-mode-@var{app-model}
17547 Select application model @var{app-model}. Valid models are
17550 normal executable (default), uses startup code @file{crt0.o}.
17553 for use with Xilinx Microprocessor Debugger (XMD) based
17554 software intrusive debug agent called xmdstub. This uses startup file
17555 @file{crt1.o} and sets the start address of the program to 0x800.
17558 for applications that are loaded using a bootloader.
17559 This model uses startup file @file{crt2.o} which does not contain a processor
17560 reset vector handler. This is suitable for transferring control on a
17561 processor reset to the bootloader rather than the application.
17564 for applications that do not require any of the
17565 MicroBlaze vectors. This option may be useful for applications running
17566 within a monitoring application. This model uses @file{crt3.o} as a startup file.
17569 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
17570 @option{-mxl-mode-@var{app-model}}.
17575 @subsection MIPS Options
17576 @cindex MIPS options
17582 Generate big-endian code.
17586 Generate little-endian code. This is the default for @samp{mips*el-*-*}
17589 @item -march=@var{arch}
17591 Generate code that runs on @var{arch}, which can be the name of a
17592 generic MIPS ISA, or the name of a particular processor.
17594 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
17595 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
17596 @samp{mips64}, @samp{mips64r2}, @samp{mips64r3} and @samp{mips64r5}.
17597 The processor names are:
17598 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
17599 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
17600 @samp{5kc}, @samp{5kf},
17602 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
17603 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
17604 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
17605 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
17606 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
17607 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
17609 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
17610 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
17613 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
17614 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
17615 @samp{rm7000}, @samp{rm9000},
17616 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
17619 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
17620 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
17621 @samp{xlr} and @samp{xlp}.
17622 The special value @samp{from-abi} selects the
17623 most compatible architecture for the selected ABI (that is,
17624 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
17626 The native Linux/GNU toolchain also supports the value @samp{native},
17627 which selects the best architecture option for the host processor.
17628 @option{-march=native} has no effect if GCC does not recognize
17631 In processor names, a final @samp{000} can be abbreviated as @samp{k}
17632 (for example, @option{-march=r2k}). Prefixes are optional, and
17633 @samp{vr} may be written @samp{r}.
17635 Names of the form @samp{@var{n}f2_1} refer to processors with
17636 FPUs clocked at half the rate of the core, names of the form
17637 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
17638 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
17639 processors with FPUs clocked a ratio of 3:2 with respect to the core.
17640 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
17641 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
17642 accepted as synonyms for @samp{@var{n}f1_1}.
17644 GCC defines two macros based on the value of this option. The first
17645 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
17646 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
17647 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
17648 For example, @option{-march=r2000} sets @samp{_MIPS_ARCH}
17649 to @samp{"r2000"} and defines the macro @samp{_MIPS_ARCH_R2000}.
17651 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
17652 above. In other words, it has the full prefix and does not
17653 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
17654 the macro names the resolved architecture (either @samp{"mips1"} or
17655 @samp{"mips3"}). It names the default architecture when no
17656 @option{-march} option is given.
17658 @item -mtune=@var{arch}
17660 Optimize for @var{arch}. Among other things, this option controls
17661 the way instructions are scheduled, and the perceived cost of arithmetic
17662 operations. The list of @var{arch} values is the same as for
17665 When this option is not used, GCC optimizes for the processor
17666 specified by @option{-march}. By using @option{-march} and
17667 @option{-mtune} together, it is possible to generate code that
17668 runs on a family of processors, but optimize the code for one
17669 particular member of that family.
17671 @option{-mtune} defines the macros @samp{_MIPS_TUNE} and
17672 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
17673 @option{-march} ones described above.
17677 Equivalent to @option{-march=mips1}.
17681 Equivalent to @option{-march=mips2}.
17685 Equivalent to @option{-march=mips3}.
17689 Equivalent to @option{-march=mips4}.
17693 Equivalent to @option{-march=mips32}.
17697 Equivalent to @option{-march=mips32r3}.
17701 Equivalent to @option{-march=mips32r5}.
17705 Equivalent to @option{-march=mips64}.
17709 Equivalent to @option{-march=mips64r2}.
17713 Equivalent to @option{-march=mips64r3}.
17717 Equivalent to @option{-march=mips64r5}.
17722 @opindex mno-mips16
17723 Generate (do not generate) MIPS16 code. If GCC is targeting a
17724 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
17726 MIPS16 code generation can also be controlled on a per-function basis
17727 by means of @code{mips16} and @code{nomips16} attributes.
17728 @xref{Function Attributes}, for more information.
17730 @item -mflip-mips16
17731 @opindex mflip-mips16
17732 Generate MIPS16 code on alternating functions. This option is provided
17733 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
17734 not intended for ordinary use in compiling user code.
17736 @item -minterlink-compressed
17737 @item -mno-interlink-compressed
17738 @opindex minterlink-compressed
17739 @opindex mno-interlink-compressed
17740 Require (do not require) that code using the standard (uncompressed) MIPS ISA
17741 be link-compatible with MIPS16 and microMIPS code, and vice versa.
17743 For example, code using the standard ISA encoding cannot jump directly
17744 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
17745 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
17746 knows that the target of the jump is not compressed.
17748 @item -minterlink-mips16
17749 @itemx -mno-interlink-mips16
17750 @opindex minterlink-mips16
17751 @opindex mno-interlink-mips16
17752 Aliases of @option{-minterlink-compressed} and
17753 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
17754 and are retained for backwards compatibility.
17766 Generate code for the given ABI@.
17768 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
17769 generates 64-bit code when you select a 64-bit architecture, but you
17770 can use @option{-mgp32} to get 32-bit code instead.
17772 For information about the O64 ABI, see
17773 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
17775 GCC supports a variant of the o32 ABI in which floating-point registers
17776 are 64 rather than 32 bits wide. You can select this combination with
17777 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
17778 and @code{mfhc1} instructions and is therefore only supported for
17779 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
17781 The register assignments for arguments and return values remain the
17782 same, but each scalar value is passed in a single 64-bit register
17783 rather than a pair of 32-bit registers. For example, scalar
17784 floating-point values are returned in @samp{$f0} only, not a
17785 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
17786 remains the same in that the even-numbered double-precision registers
17789 Two additional variants of the o32 ABI are supported to enable
17790 a transition from 32-bit to 64-bit registers. These are FPXX
17791 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
17792 The FPXX extension mandates that all code must execute correctly
17793 when run using 32-bit or 64-bit registers. The code can be interlinked
17794 with either FP32 or FP64, but not both.
17795 The FP64A extension is similar to the FP64 extension but forbids the
17796 use of odd-numbered single-precision registers. This can be used
17797 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
17798 processors and allows both FP32 and FP64A code to interlink and
17799 run in the same process without changing FPU modes.
17802 @itemx -mno-abicalls
17804 @opindex mno-abicalls
17805 Generate (do not generate) code that is suitable for SVR4-style
17806 dynamic objects. @option{-mabicalls} is the default for SVR4-based
17811 Generate (do not generate) code that is fully position-independent,
17812 and that can therefore be linked into shared libraries. This option
17813 only affects @option{-mabicalls}.
17815 All @option{-mabicalls} code has traditionally been position-independent,
17816 regardless of options like @option{-fPIC} and @option{-fpic}. However,
17817 as an extension, the GNU toolchain allows executables to use absolute
17818 accesses for locally-binding symbols. It can also use shorter GP
17819 initialization sequences and generate direct calls to locally-defined
17820 functions. This mode is selected by @option{-mno-shared}.
17822 @option{-mno-shared} depends on binutils 2.16 or higher and generates
17823 objects that can only be linked by the GNU linker. However, the option
17824 does not affect the ABI of the final executable; it only affects the ABI
17825 of relocatable objects. Using @option{-mno-shared} generally makes
17826 executables both smaller and quicker.
17828 @option{-mshared} is the default.
17834 Assume (do not assume) that the static and dynamic linkers
17835 support PLTs and copy relocations. This option only affects
17836 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
17837 has no effect without @option{-msym32}.
17839 You can make @option{-mplt} the default by configuring
17840 GCC with @option{--with-mips-plt}. The default is
17841 @option{-mno-plt} otherwise.
17847 Lift (do not lift) the usual restrictions on the size of the global
17850 GCC normally uses a single instruction to load values from the GOT@.
17851 While this is relatively efficient, it only works if the GOT
17852 is smaller than about 64k. Anything larger causes the linker
17853 to report an error such as:
17855 @cindex relocation truncated to fit (MIPS)
17857 relocation truncated to fit: R_MIPS_GOT16 foobar
17860 If this happens, you should recompile your code with @option{-mxgot}.
17861 This works with very large GOTs, although the code is also
17862 less efficient, since it takes three instructions to fetch the
17863 value of a global symbol.
17865 Note that some linkers can create multiple GOTs. If you have such a
17866 linker, you should only need to use @option{-mxgot} when a single object
17867 file accesses more than 64k's worth of GOT entries. Very few do.
17869 These options have no effect unless GCC is generating position
17874 Assume that general-purpose registers are 32 bits wide.
17878 Assume that general-purpose registers are 64 bits wide.
17882 Assume that floating-point registers are 32 bits wide.
17886 Assume that floating-point registers are 64 bits wide.
17890 Do not assume the width of floating-point registers.
17893 @opindex mhard-float
17894 Use floating-point coprocessor instructions.
17897 @opindex msoft-float
17898 Do not use floating-point coprocessor instructions. Implement
17899 floating-point calculations using library calls instead.
17903 Equivalent to @option{-msoft-float}, but additionally asserts that the
17904 program being compiled does not perform any floating-point operations.
17905 This option is presently supported only by some bare-metal MIPS
17906 configurations, where it may select a special set of libraries
17907 that lack all floating-point support (including, for example, the
17908 floating-point @code{printf} formats).
17909 If code compiled with @code{-mno-float} accidentally contains
17910 floating-point operations, it is likely to suffer a link-time
17911 or run-time failure.
17913 @item -msingle-float
17914 @opindex msingle-float
17915 Assume that the floating-point coprocessor only supports single-precision
17918 @item -mdouble-float
17919 @opindex mdouble-float
17920 Assume that the floating-point coprocessor supports double-precision
17921 operations. This is the default.
17924 @itemx -mno-odd-spreg
17925 @opindex modd-spreg
17926 @opindex mno-odd-spreg
17927 Enable the use of odd-numbered single-precision floating-point registers
17928 for the o32 ABI. This is the default for processors that are known to
17929 support these registers. When using the o32 FPXX ABI, @code{-mno-odd-spreg}
17933 @itemx -mabs=legacy
17935 @opindex mabs=legacy
17936 These options control the treatment of the special not-a-number (NaN)
17937 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
17938 @code{neg.@i{fmt}} machine instructions.
17940 By default or when the @option{-mabs=legacy} is used the legacy
17941 treatment is selected. In this case these instructions are considered
17942 arithmetic and avoided where correct operation is required and the
17943 input operand might be a NaN. A longer sequence of instructions that
17944 manipulate the sign bit of floating-point datum manually is used
17945 instead unless the @option{-ffinite-math-only} option has also been
17948 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
17949 this case these instructions are considered non-arithmetic and therefore
17950 operating correctly in all cases, including in particular where the
17951 input operand is a NaN. These instructions are therefore always used
17952 for the respective operations.
17955 @itemx -mnan=legacy
17957 @opindex mnan=legacy
17958 These options control the encoding of the special not-a-number (NaN)
17959 IEEE 754 floating-point data.
17961 The @option{-mnan=legacy} option selects the legacy encoding. In this
17962 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
17963 significand field being 0, whereas signalling NaNs (sNaNs) are denoted
17964 by the first bit of their trailing significand field being 1.
17966 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
17967 this case qNaNs are denoted by the first bit of their trailing
17968 significand field being 1, whereas sNaNs are denoted by the first bit of
17969 their trailing significand field being 0.
17971 The default is @option{-mnan=legacy} unless GCC has been configured with
17972 @option{--with-nan=2008}.
17978 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
17979 implement atomic memory built-in functions. When neither option is
17980 specified, GCC uses the instructions if the target architecture
17983 @option{-mllsc} is useful if the runtime environment can emulate the
17984 instructions and @option{-mno-llsc} can be useful when compiling for
17985 nonstandard ISAs. You can make either option the default by
17986 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
17987 respectively. @option{--with-llsc} is the default for some
17988 configurations; see the installation documentation for details.
17994 Use (do not use) revision 1 of the MIPS DSP ASE@.
17995 @xref{MIPS DSP Built-in Functions}. This option defines the
17996 preprocessor macro @samp{__mips_dsp}. It also defines
17997 @samp{__mips_dsp_rev} to 1.
18003 Use (do not use) revision 2 of the MIPS DSP ASE@.
18004 @xref{MIPS DSP Built-in Functions}. This option defines the
18005 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
18006 It also defines @samp{__mips_dsp_rev} to 2.
18009 @itemx -mno-smartmips
18010 @opindex msmartmips
18011 @opindex mno-smartmips
18012 Use (do not use) the MIPS SmartMIPS ASE.
18014 @item -mpaired-single
18015 @itemx -mno-paired-single
18016 @opindex mpaired-single
18017 @opindex mno-paired-single
18018 Use (do not use) paired-single floating-point instructions.
18019 @xref{MIPS Paired-Single Support}. This option requires
18020 hardware floating-point support to be enabled.
18026 Use (do not use) MIPS Digital Media Extension instructions.
18027 This option can only be used when generating 64-bit code and requires
18028 hardware floating-point support to be enabled.
18033 @opindex mno-mips3d
18034 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
18035 The option @option{-mips3d} implies @option{-mpaired-single}.
18038 @itemx -mno-micromips
18039 @opindex mmicromips
18040 @opindex mno-mmicromips
18041 Generate (do not generate) microMIPS code.
18043 MicroMIPS code generation can also be controlled on a per-function basis
18044 by means of @code{micromips} and @code{nomicromips} attributes.
18045 @xref{Function Attributes}, for more information.
18051 Use (do not use) MT Multithreading instructions.
18057 Use (do not use) the MIPS MCU ASE instructions.
18063 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
18069 Use (do not use) the MIPS Virtualization Application Specific instructions.
18075 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
18079 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
18080 an explanation of the default and the way that the pointer size is
18085 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
18087 The default size of @code{int}s, @code{long}s and pointers depends on
18088 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
18089 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
18090 32-bit @code{long}s. Pointers are the same size as @code{long}s,
18091 or the same size as integer registers, whichever is smaller.
18097 Assume (do not assume) that all symbols have 32-bit values, regardless
18098 of the selected ABI@. This option is useful in combination with
18099 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
18100 to generate shorter and faster references to symbolic addresses.
18104 Put definitions of externally-visible data in a small data section
18105 if that data is no bigger than @var{num} bytes. GCC can then generate
18106 more efficient accesses to the data; see @option{-mgpopt} for details.
18108 The default @option{-G} option depends on the configuration.
18110 @item -mlocal-sdata
18111 @itemx -mno-local-sdata
18112 @opindex mlocal-sdata
18113 @opindex mno-local-sdata
18114 Extend (do not extend) the @option{-G} behavior to local data too,
18115 such as to static variables in C@. @option{-mlocal-sdata} is the
18116 default for all configurations.
18118 If the linker complains that an application is using too much small data,
18119 you might want to try rebuilding the less performance-critical parts with
18120 @option{-mno-local-sdata}. You might also want to build large
18121 libraries with @option{-mno-local-sdata}, so that the libraries leave
18122 more room for the main program.
18124 @item -mextern-sdata
18125 @itemx -mno-extern-sdata
18126 @opindex mextern-sdata
18127 @opindex mno-extern-sdata
18128 Assume (do not assume) that externally-defined data is in
18129 a small data section if the size of that data is within the @option{-G} limit.
18130 @option{-mextern-sdata} is the default for all configurations.
18132 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
18133 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
18134 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
18135 is placed in a small data section. If @var{Var} is defined by another
18136 module, you must either compile that module with a high-enough
18137 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
18138 definition. If @var{Var} is common, you must link the application
18139 with a high-enough @option{-G} setting.
18141 The easiest way of satisfying these restrictions is to compile
18142 and link every module with the same @option{-G} option. However,
18143 you may wish to build a library that supports several different
18144 small data limits. You can do this by compiling the library with
18145 the highest supported @option{-G} setting and additionally using
18146 @option{-mno-extern-sdata} to stop the library from making assumptions
18147 about externally-defined data.
18153 Use (do not use) GP-relative accesses for symbols that are known to be
18154 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
18155 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
18158 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
18159 might not hold the value of @code{_gp}. For example, if the code is
18160 part of a library that might be used in a boot monitor, programs that
18161 call boot monitor routines pass an unknown value in @code{$gp}.
18162 (In such situations, the boot monitor itself is usually compiled
18163 with @option{-G0}.)
18165 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
18166 @option{-mno-extern-sdata}.
18168 @item -membedded-data
18169 @itemx -mno-embedded-data
18170 @opindex membedded-data
18171 @opindex mno-embedded-data
18172 Allocate variables to the read-only data section first if possible, then
18173 next in the small data section if possible, otherwise in data. This gives
18174 slightly slower code than the default, but reduces the amount of RAM required
18175 when executing, and thus may be preferred for some embedded systems.
18177 @item -muninit-const-in-rodata
18178 @itemx -mno-uninit-const-in-rodata
18179 @opindex muninit-const-in-rodata
18180 @opindex mno-uninit-const-in-rodata
18181 Put uninitialized @code{const} variables in the read-only data section.
18182 This option is only meaningful in conjunction with @option{-membedded-data}.
18184 @item -mcode-readable=@var{setting}
18185 @opindex mcode-readable
18186 Specify whether GCC may generate code that reads from executable sections.
18187 There are three possible settings:
18190 @item -mcode-readable=yes
18191 Instructions may freely access executable sections. This is the
18194 @item -mcode-readable=pcrel
18195 MIPS16 PC-relative load instructions can access executable sections,
18196 but other instructions must not do so. This option is useful on 4KSc
18197 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
18198 It is also useful on processors that can be configured to have a dual
18199 instruction/data SRAM interface and that, like the M4K, automatically
18200 redirect PC-relative loads to the instruction RAM.
18202 @item -mcode-readable=no
18203 Instructions must not access executable sections. This option can be
18204 useful on targets that are configured to have a dual instruction/data
18205 SRAM interface but that (unlike the M4K) do not automatically redirect
18206 PC-relative loads to the instruction RAM.
18209 @item -msplit-addresses
18210 @itemx -mno-split-addresses
18211 @opindex msplit-addresses
18212 @opindex mno-split-addresses
18213 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
18214 relocation operators. This option has been superseded by
18215 @option{-mexplicit-relocs} but is retained for backwards compatibility.
18217 @item -mexplicit-relocs
18218 @itemx -mno-explicit-relocs
18219 @opindex mexplicit-relocs
18220 @opindex mno-explicit-relocs
18221 Use (do not use) assembler relocation operators when dealing with symbolic
18222 addresses. The alternative, selected by @option{-mno-explicit-relocs},
18223 is to use assembler macros instead.
18225 @option{-mexplicit-relocs} is the default if GCC was configured
18226 to use an assembler that supports relocation operators.
18228 @item -mcheck-zero-division
18229 @itemx -mno-check-zero-division
18230 @opindex mcheck-zero-division
18231 @opindex mno-check-zero-division
18232 Trap (do not trap) on integer division by zero.
18234 The default is @option{-mcheck-zero-division}.
18236 @item -mdivide-traps
18237 @itemx -mdivide-breaks
18238 @opindex mdivide-traps
18239 @opindex mdivide-breaks
18240 MIPS systems check for division by zero by generating either a
18241 conditional trap or a break instruction. Using traps results in
18242 smaller code, but is only supported on MIPS II and later. Also, some
18243 versions of the Linux kernel have a bug that prevents trap from
18244 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
18245 allow conditional traps on architectures that support them and
18246 @option{-mdivide-breaks} to force the use of breaks.
18248 The default is usually @option{-mdivide-traps}, but this can be
18249 overridden at configure time using @option{--with-divide=breaks}.
18250 Divide-by-zero checks can be completely disabled using
18251 @option{-mno-check-zero-division}.
18256 @opindex mno-memcpy
18257 Force (do not force) the use of @code{memcpy()} for non-trivial block
18258 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
18259 most constant-sized copies.
18262 @itemx -mno-long-calls
18263 @opindex mlong-calls
18264 @opindex mno-long-calls
18265 Disable (do not disable) use of the @code{jal} instruction. Calling
18266 functions using @code{jal} is more efficient but requires the caller
18267 and callee to be in the same 256 megabyte segment.
18269 This option has no effect on abicalls code. The default is
18270 @option{-mno-long-calls}.
18276 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
18277 instructions, as provided by the R4650 ISA@.
18283 Enable (disable) use of the @code{madd} and @code{msub} integer
18284 instructions. The default is @option{-mimadd} on architectures
18285 that support @code{madd} and @code{msub} except for the 74k
18286 architecture where it was found to generate slower code.
18289 @itemx -mno-fused-madd
18290 @opindex mfused-madd
18291 @opindex mno-fused-madd
18292 Enable (disable) use of the floating-point multiply-accumulate
18293 instructions, when they are available. The default is
18294 @option{-mfused-madd}.
18296 On the R8000 CPU when multiply-accumulate instructions are used,
18297 the intermediate product is calculated to infinite precision
18298 and is not subject to the FCSR Flush to Zero bit. This may be
18299 undesirable in some circumstances. On other processors the result
18300 is numerically identical to the equivalent computation using
18301 separate multiply, add, subtract and negate instructions.
18305 Tell the MIPS assembler to not run its preprocessor over user
18306 assembler files (with a @samp{.s} suffix) when assembling them.
18311 @opindex mno-fix-24k
18312 Work around the 24K E48 (lost data on stores during refill) errata.
18313 The workarounds are implemented by the assembler rather than by GCC@.
18316 @itemx -mno-fix-r4000
18317 @opindex mfix-r4000
18318 @opindex mno-fix-r4000
18319 Work around certain R4000 CPU errata:
18322 A double-word or a variable shift may give an incorrect result if executed
18323 immediately after starting an integer division.
18325 A double-word or a variable shift may give an incorrect result if executed
18326 while an integer multiplication is in progress.
18328 An integer division may give an incorrect result if started in a delay slot
18329 of a taken branch or a jump.
18333 @itemx -mno-fix-r4400
18334 @opindex mfix-r4400
18335 @opindex mno-fix-r4400
18336 Work around certain R4400 CPU errata:
18339 A double-word or a variable shift may give an incorrect result if executed
18340 immediately after starting an integer division.
18344 @itemx -mno-fix-r10000
18345 @opindex mfix-r10000
18346 @opindex mno-fix-r10000
18347 Work around certain R10000 errata:
18350 @code{ll}/@code{sc} sequences may not behave atomically on revisions
18351 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
18354 This option can only be used if the target architecture supports
18355 branch-likely instructions. @option{-mfix-r10000} is the default when
18356 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
18360 @itemx -mno-fix-rm7000
18361 @opindex mfix-rm7000
18362 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
18363 workarounds are implemented by the assembler rather than by GCC@.
18366 @itemx -mno-fix-vr4120
18367 @opindex mfix-vr4120
18368 Work around certain VR4120 errata:
18371 @code{dmultu} does not always produce the correct result.
18373 @code{div} and @code{ddiv} do not always produce the correct result if one
18374 of the operands is negative.
18376 The workarounds for the division errata rely on special functions in
18377 @file{libgcc.a}. At present, these functions are only provided by
18378 the @code{mips64vr*-elf} configurations.
18380 Other VR4120 errata require a NOP to be inserted between certain pairs of
18381 instructions. These errata are handled by the assembler, not by GCC itself.
18384 @opindex mfix-vr4130
18385 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
18386 workarounds are implemented by the assembler rather than by GCC,
18387 although GCC avoids using @code{mflo} and @code{mfhi} if the
18388 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
18389 instructions are available instead.
18392 @itemx -mno-fix-sb1
18394 Work around certain SB-1 CPU core errata.
18395 (This flag currently works around the SB-1 revision 2
18396 ``F1'' and ``F2'' floating-point errata.)
18398 @item -mr10k-cache-barrier=@var{setting}
18399 @opindex mr10k-cache-barrier
18400 Specify whether GCC should insert cache barriers to avoid the
18401 side-effects of speculation on R10K processors.
18403 In common with many processors, the R10K tries to predict the outcome
18404 of a conditional branch and speculatively executes instructions from
18405 the ``taken'' branch. It later aborts these instructions if the
18406 predicted outcome is wrong. However, on the R10K, even aborted
18407 instructions can have side effects.
18409 This problem only affects kernel stores and, depending on the system,
18410 kernel loads. As an example, a speculatively-executed store may load
18411 the target memory into cache and mark the cache line as dirty, even if
18412 the store itself is later aborted. If a DMA operation writes to the
18413 same area of memory before the ``dirty'' line is flushed, the cached
18414 data overwrites the DMA-ed data. See the R10K processor manual
18415 for a full description, including other potential problems.
18417 One workaround is to insert cache barrier instructions before every memory
18418 access that might be speculatively executed and that might have side
18419 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
18420 controls GCC's implementation of this workaround. It assumes that
18421 aborted accesses to any byte in the following regions does not have
18426 the memory occupied by the current function's stack frame;
18429 the memory occupied by an incoming stack argument;
18432 the memory occupied by an object with a link-time-constant address.
18435 It is the kernel's responsibility to ensure that speculative
18436 accesses to these regions are indeed safe.
18438 If the input program contains a function declaration such as:
18444 then the implementation of @code{foo} must allow @code{j foo} and
18445 @code{jal foo} to be executed speculatively. GCC honors this
18446 restriction for functions it compiles itself. It expects non-GCC
18447 functions (such as hand-written assembly code) to do the same.
18449 The option has three forms:
18452 @item -mr10k-cache-barrier=load-store
18453 Insert a cache barrier before a load or store that might be
18454 speculatively executed and that might have side effects even
18457 @item -mr10k-cache-barrier=store
18458 Insert a cache barrier before a store that might be speculatively
18459 executed and that might have side effects even if aborted.
18461 @item -mr10k-cache-barrier=none
18462 Disable the insertion of cache barriers. This is the default setting.
18465 @item -mflush-func=@var{func}
18466 @itemx -mno-flush-func
18467 @opindex mflush-func
18468 Specifies the function to call to flush the I and D caches, or to not
18469 call any such function. If called, the function must take the same
18470 arguments as the common @code{_flush_func()}, that is, the address of the
18471 memory range for which the cache is being flushed, the size of the
18472 memory range, and the number 3 (to flush both caches). The default
18473 depends on the target GCC was configured for, but commonly is either
18474 @samp{_flush_func} or @samp{__cpu_flush}.
18476 @item mbranch-cost=@var{num}
18477 @opindex mbranch-cost
18478 Set the cost of branches to roughly @var{num} ``simple'' instructions.
18479 This cost is only a heuristic and is not guaranteed to produce
18480 consistent results across releases. A zero cost redundantly selects
18481 the default, which is based on the @option{-mtune} setting.
18483 @item -mbranch-likely
18484 @itemx -mno-branch-likely
18485 @opindex mbranch-likely
18486 @opindex mno-branch-likely
18487 Enable or disable use of Branch Likely instructions, regardless of the
18488 default for the selected architecture. By default, Branch Likely
18489 instructions may be generated if they are supported by the selected
18490 architecture. An exception is for the MIPS32 and MIPS64 architectures
18491 and processors that implement those architectures; for those, Branch
18492 Likely instructions are not be generated by default because the MIPS32
18493 and MIPS64 architectures specifically deprecate their use.
18495 @item -mfp-exceptions
18496 @itemx -mno-fp-exceptions
18497 @opindex mfp-exceptions
18498 Specifies whether FP exceptions are enabled. This affects how
18499 FP instructions are scheduled for some processors.
18500 The default is that FP exceptions are
18503 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
18504 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
18507 @item -mvr4130-align
18508 @itemx -mno-vr4130-align
18509 @opindex mvr4130-align
18510 The VR4130 pipeline is two-way superscalar, but can only issue two
18511 instructions together if the first one is 8-byte aligned. When this
18512 option is enabled, GCC aligns pairs of instructions that it
18513 thinks should execute in parallel.
18515 This option only has an effect when optimizing for the VR4130.
18516 It normally makes code faster, but at the expense of making it bigger.
18517 It is enabled by default at optimization level @option{-O3}.
18522 Enable (disable) generation of @code{synci} instructions on
18523 architectures that support it. The @code{synci} instructions (if
18524 enabled) are generated when @code{__builtin___clear_cache()} is
18527 This option defaults to @code{-mno-synci}, but the default can be
18528 overridden by configuring with @code{--with-synci}.
18530 When compiling code for single processor systems, it is generally safe
18531 to use @code{synci}. However, on many multi-core (SMP) systems, it
18532 does not invalidate the instruction caches on all cores and may lead
18533 to undefined behavior.
18535 @item -mrelax-pic-calls
18536 @itemx -mno-relax-pic-calls
18537 @opindex mrelax-pic-calls
18538 Try to turn PIC calls that are normally dispatched via register
18539 @code{$25} into direct calls. This is only possible if the linker can
18540 resolve the destination at link-time and if the destination is within
18541 range for a direct call.
18543 @option{-mrelax-pic-calls} is the default if GCC was configured to use
18544 an assembler and a linker that support the @code{.reloc} assembly
18545 directive and @code{-mexplicit-relocs} is in effect. With
18546 @code{-mno-explicit-relocs}, this optimization can be performed by the
18547 assembler and the linker alone without help from the compiler.
18549 @item -mmcount-ra-address
18550 @itemx -mno-mcount-ra-address
18551 @opindex mmcount-ra-address
18552 @opindex mno-mcount-ra-address
18553 Emit (do not emit) code that allows @code{_mcount} to modify the
18554 calling function's return address. When enabled, this option extends
18555 the usual @code{_mcount} interface with a new @var{ra-address}
18556 parameter, which has type @code{intptr_t *} and is passed in register
18557 @code{$12}. @code{_mcount} can then modify the return address by
18558 doing both of the following:
18561 Returning the new address in register @code{$31}.
18563 Storing the new address in @code{*@var{ra-address}},
18564 if @var{ra-address} is nonnull.
18567 The default is @option{-mno-mcount-ra-address}.
18572 @subsection MMIX Options
18573 @cindex MMIX Options
18575 These options are defined for the MMIX:
18579 @itemx -mno-libfuncs
18581 @opindex mno-libfuncs
18582 Specify that intrinsic library functions are being compiled, passing all
18583 values in registers, no matter the size.
18586 @itemx -mno-epsilon
18588 @opindex mno-epsilon
18589 Generate floating-point comparison instructions that compare with respect
18590 to the @code{rE} epsilon register.
18592 @item -mabi=mmixware
18594 @opindex mabi=mmixware
18596 Generate code that passes function parameters and return values that (in
18597 the called function) are seen as registers @code{$0} and up, as opposed to
18598 the GNU ABI which uses global registers @code{$231} and up.
18600 @item -mzero-extend
18601 @itemx -mno-zero-extend
18602 @opindex mzero-extend
18603 @opindex mno-zero-extend
18604 When reading data from memory in sizes shorter than 64 bits, use (do not
18605 use) zero-extending load instructions by default, rather than
18606 sign-extending ones.
18609 @itemx -mno-knuthdiv
18611 @opindex mno-knuthdiv
18612 Make the result of a division yielding a remainder have the same sign as
18613 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
18614 remainder follows the sign of the dividend. Both methods are
18615 arithmetically valid, the latter being almost exclusively used.
18617 @item -mtoplevel-symbols
18618 @itemx -mno-toplevel-symbols
18619 @opindex mtoplevel-symbols
18620 @opindex mno-toplevel-symbols
18621 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
18622 code can be used with the @code{PREFIX} assembly directive.
18626 Generate an executable in the ELF format, rather than the default
18627 @samp{mmo} format used by the @command{mmix} simulator.
18629 @item -mbranch-predict
18630 @itemx -mno-branch-predict
18631 @opindex mbranch-predict
18632 @opindex mno-branch-predict
18633 Use (do not use) the probable-branch instructions, when static branch
18634 prediction indicates a probable branch.
18636 @item -mbase-addresses
18637 @itemx -mno-base-addresses
18638 @opindex mbase-addresses
18639 @opindex mno-base-addresses
18640 Generate (do not generate) code that uses @emph{base addresses}. Using a
18641 base address automatically generates a request (handled by the assembler
18642 and the linker) for a constant to be set up in a global register. The
18643 register is used for one or more base address requests within the range 0
18644 to 255 from the value held in the register. The generally leads to short
18645 and fast code, but the number of different data items that can be
18646 addressed is limited. This means that a program that uses lots of static
18647 data may require @option{-mno-base-addresses}.
18649 @item -msingle-exit
18650 @itemx -mno-single-exit
18651 @opindex msingle-exit
18652 @opindex mno-single-exit
18653 Force (do not force) generated code to have a single exit point in each
18657 @node MN10300 Options
18658 @subsection MN10300 Options
18659 @cindex MN10300 options
18661 These @option{-m} options are defined for Matsushita MN10300 architectures:
18666 Generate code to avoid bugs in the multiply instructions for the MN10300
18667 processors. This is the default.
18669 @item -mno-mult-bug
18670 @opindex mno-mult-bug
18671 Do not generate code to avoid bugs in the multiply instructions for the
18672 MN10300 processors.
18676 Generate code using features specific to the AM33 processor.
18680 Do not generate code using features specific to the AM33 processor. This
18685 Generate code using features specific to the AM33/2.0 processor.
18689 Generate code using features specific to the AM34 processor.
18691 @item -mtune=@var{cpu-type}
18693 Use the timing characteristics of the indicated CPU type when
18694 scheduling instructions. This does not change the targeted processor
18695 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
18696 @samp{am33-2} or @samp{am34}.
18698 @item -mreturn-pointer-on-d0
18699 @opindex mreturn-pointer-on-d0
18700 When generating a function that returns a pointer, return the pointer
18701 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
18702 only in @code{a0}, and attempts to call such functions without a prototype
18703 result in errors. Note that this option is on by default; use
18704 @option{-mno-return-pointer-on-d0} to disable it.
18708 Do not link in the C run-time initialization object file.
18712 Indicate to the linker that it should perform a relaxation optimization pass
18713 to shorten branches, calls and absolute memory addresses. This option only
18714 has an effect when used on the command line for the final link step.
18716 This option makes symbolic debugging impossible.
18720 Allow the compiler to generate @emph{Long Instruction Word}
18721 instructions if the target is the @samp{AM33} or later. This is the
18722 default. This option defines the preprocessor macro @samp{__LIW__}.
18726 Do not allow the compiler to generate @emph{Long Instruction Word}
18727 instructions. This option defines the preprocessor macro
18732 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
18733 instructions if the target is the @samp{AM33} or later. This is the
18734 default. This option defines the preprocessor macro @samp{__SETLB__}.
18738 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
18739 instructions. This option defines the preprocessor macro
18740 @samp{__NO_SETLB__}.
18744 @node Moxie Options
18745 @subsection Moxie Options
18746 @cindex Moxie Options
18752 Generate big-endian code. This is the default for @samp{moxie-*-*}
18757 Generate little-endian code.
18761 Do not link in the C run-time initialization object file.
18765 @node MSP430 Options
18766 @subsection MSP430 Options
18767 @cindex MSP430 Options
18769 These options are defined for the MSP430:
18775 Force assembly output to always use hex constants. Normally such
18776 constants are signed decimals, but this option is available for
18777 testsuite and/or aesthetic purposes.
18781 Select the MCU to target. This is used to create a C preprocessor
18782 symbol based upon the MCU name, converted to upper case and pre- and
18783 post- fixed with @code{__}. This in turn will be used by the
18784 @code{msp430.h} header file to select an MCU specific supplimentary
18787 The option also sets the ISA to use. If the MCU name is one that is
18788 known to only support the 430 ISA then that is selected, otherwise the
18789 430X ISA is selected. A generic MCU name of @code{msp430} can also be
18790 used to select the 430 ISA. Similarly the generic @code{msp430x} MCU
18791 name will select the 430X ISA.
18793 In addition an MCU specific linker script will be added to the linker
18794 command line. The script's name is the name of the MCU with
18795 @code{.ld} appended. Thus specifying @option{-mmcu=xxx} on the gcc
18796 command line will define the C preprocessor symbol @code{__XXX__} and
18797 cause the linker to search for a script called @file{xxx.ld}.
18799 This option is also passed on to the assembler.
18803 Specifies the ISA to use. Accepted values are @code{msp430},
18804 @code{msp430x} and @code{msp430xv2}. This option is deprecated. The
18805 @option{-mmcu=} option should be used to select the ISA.
18809 Link to the simulator runtime libraries and linker script. Overrides
18810 any scripts that would be selected by the @option{-mmcu=} option.
18814 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
18818 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
18822 This option is passed to the assembler and linker, and allows the
18823 linker to perform certain optimizations that cannot be done until
18828 Describes the type of hardware multiply supported by the target.
18829 Accepted values are @code{none} for no hardware multiply, @code{16bit}
18830 for the original 16-bit-only multiply supported by early MCUs.
18831 @code{32bit} for the 16/32-bit multiply supported by later MCUs and
18832 @code{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
18833 A value of @code{auto} can also be given. This tells GCC to deduce
18834 the hardware multiply support based upon the MCU name provided by the
18835 @option{-mmcu} option. If no @option{-mmcu} option is specified then
18836 @code{32bit} hardware multiply support is assumed. @code{auto} is the
18839 Hardware multiplies are normally performed by calling a library
18840 routine. This saves space in the generated code. When compiling at
18841 @code{-O3} or higher however the hardware multiplier is invoked
18842 inline. This makes for bigger, but faster code.
18844 The hardware multiply routines disable interrupts whilst running and
18845 restore the previous interrupt state when they finish. This makes
18846 them safe to use inside interrupt handlers as well as in normal code.
18850 Enable the use of a minimum runtime environment - no static
18851 initializers or constructors. This is intended for memory-constrained
18852 devices. The compiler will include special symbols in some objects
18853 that tell the linker and runtime which code fragments are required.
18857 @node NDS32 Options
18858 @subsection NDS32 Options
18859 @cindex NDS32 Options
18861 These options are defined for NDS32 implementations:
18866 @opindex mbig-endian
18867 Generate code in big-endian mode.
18869 @item -mlittle-endian
18870 @opindex mlittle-endian
18871 Generate code in little-endian mode.
18873 @item -mreduced-regs
18874 @opindex mreduced-regs
18875 Use reduced-set registers for register allocation.
18878 @opindex mfull-regs
18879 Use full-set registers for register allocation.
18883 Generate conditional move instructions.
18887 Do not generate conditional move instructions.
18891 Generate performance extension instructions.
18893 @item -mno-perf-ext
18894 @opindex mno-perf-ext
18895 Do not generate performance extension instructions.
18899 Generate v3 push25/pop25 instructions.
18902 @opindex mno-v3push
18903 Do not generate v3 push25/pop25 instructions.
18907 Generate 16-bit instructions.
18910 @opindex mno-16-bit
18911 Do not generate 16-bit instructions.
18914 @opindex mgp-direct
18915 Generate GP base instructions directly.
18917 @item -mno-gp-direct
18918 @opindex mno-gp-direct
18919 Do no generate GP base instructions directly.
18921 @item -misr-vector-size=@var{num}
18922 @opindex misr-vector-size
18923 Specify the size of each interrupt vector, which must be 4 or 16.
18925 @item -mcache-block-size=@var{num}
18926 @opindex mcache-block-size
18927 Specify the size of each cache block,
18928 which must be a power of 2 between 4 and 512.
18930 @item -march=@var{arch}
18932 Specify the name of the target architecture.
18934 @item -mforce-fp-as-gp
18935 @opindex mforce-fp-as-gp
18936 Prevent $fp being allocated during register allocation so that compiler
18937 is able to force performing fp-as-gp optimization.
18939 @item -mforbid-fp-as-gp
18940 @opindex mforbid-fp-as-gp
18941 Forbid using $fp to access static and global variables.
18942 This option strictly forbids fp-as-gp optimization
18943 regardless of @option{-mforce-fp-as-gp}.
18947 Use special directives to guide linker doing ex9 optimization.
18950 @opindex mctor-dtor
18951 Enable constructor/destructor feature.
18955 Guide linker to relax instructions.
18959 @node Nios II Options
18960 @subsection Nios II Options
18961 @cindex Nios II options
18962 @cindex Altera Nios II options
18964 These are the options defined for the Altera Nios II processor.
18970 @cindex smaller data references
18971 Put global and static objects less than or equal to @var{num} bytes
18972 into the small data or BSS sections instead of the normal data or BSS
18973 sections. The default value of @var{num} is 8.
18979 Generate (do not generate) GP-relative accesses for objects in the
18980 small data or BSS sections. The default is @option{-mgpopt} except
18981 when @option{-fpic} or @option{-fPIC} is specified to generate
18982 position-independent code. Note that the Nios II ABI does not permit
18983 GP-relative accesses from shared libraries.
18985 You may need to specify @option{-mno-gpopt} explicitly when building
18986 programs that include large amounts of small data, including large
18987 GOT data sections. In this case, the 16-bit offset for GP-relative
18988 addressing may not be large enough to allow access to the entire
18989 small data section.
18995 Generate little-endian (default) or big-endian (experimental) code,
18998 @item -mbypass-cache
18999 @itemx -mno-bypass-cache
19000 @opindex mno-bypass-cache
19001 @opindex mbypass-cache
19002 Force all load and store instructions to always bypass cache by
19003 using I/O variants of the instructions. The default is not to
19006 @item -mno-cache-volatile
19007 @itemx -mcache-volatile
19008 @opindex mcache-volatile
19009 @opindex mno-cache-volatile
19010 Volatile memory access bypass the cache using the I/O variants of
19011 the load and store instructions. The default is not to bypass the cache.
19013 @item -mno-fast-sw-div
19014 @itemx -mfast-sw-div
19015 @opindex mno-fast-sw-div
19016 @opindex mfast-sw-div
19017 Do not use table-based fast divide for small numbers. The default
19018 is to use the fast divide at @option{-O3} and above.
19022 @itemx -mno-hw-mulx
19026 @opindex mno-hw-mul
19028 @opindex mno-hw-mulx
19030 @opindex mno-hw-div
19032 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
19033 instructions by the compiler. The default is to emit @code{mul}
19034 and not emit @code{div} and @code{mulx}.
19036 @item -mcustom-@var{insn}=@var{N}
19037 @itemx -mno-custom-@var{insn}
19038 @opindex mcustom-@var{insn}
19039 @opindex mno-custom-@var{insn}
19040 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
19041 custom instruction with encoding @var{N} when generating code that uses
19042 @var{insn}. For example, @code{-mcustom-fadds=253} generates custom
19043 instruction 253 for single-precision floating-point add operations instead
19044 of the default behavior of using a library call.
19046 The following values of @var{insn} are supported. Except as otherwise
19047 noted, floating-point operations are expected to be implemented with
19048 normal IEEE 754 semantics and correspond directly to the C operators or the
19049 equivalent GCC built-in functions (@pxref{Other Builtins}).
19051 Single-precision floating point:
19054 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
19055 Binary arithmetic operations.
19061 Unary absolute value.
19063 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
19064 Comparison operations.
19066 @item @samp{fmins}, @samp{fmaxs}
19067 Floating-point minimum and maximum. These instructions are only
19068 generated if @option{-ffinite-math-only} is specified.
19070 @item @samp{fsqrts}
19071 Unary square root operation.
19073 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
19074 Floating-point trigonometric and exponential functions. These instructions
19075 are only generated if @option{-funsafe-math-optimizations} is also specified.
19079 Double-precision floating point:
19082 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
19083 Binary arithmetic operations.
19089 Unary absolute value.
19091 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
19092 Comparison operations.
19094 @item @samp{fmind}, @samp{fmaxd}
19095 Double-precision minimum and maximum. These instructions are only
19096 generated if @option{-ffinite-math-only} is specified.
19098 @item @samp{fsqrtd}
19099 Unary square root operation.
19101 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
19102 Double-precision trigonometric and exponential functions. These instructions
19103 are only generated if @option{-funsafe-math-optimizations} is also specified.
19109 @item @samp{fextsd}
19110 Conversion from single precision to double precision.
19112 @item @samp{ftruncds}
19113 Conversion from double precision to single precision.
19115 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
19116 Conversion from floating point to signed or unsigned integer types, with
19117 truncation towards zero.
19120 Conversion from single-precision floating point to signed integer,
19121 rounding to the nearest integer and ties away from zero.
19122 This corresponds to the @code{__builtin_lroundf} function when
19123 @option{-fno-math-errno} is used.
19125 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
19126 Conversion from signed or unsigned integer types to floating-point types.
19130 In addition, all of the following transfer instructions for internal
19131 registers X and Y must be provided to use any of the double-precision
19132 floating-point instructions. Custom instructions taking two
19133 double-precision source operands expect the first operand in the
19134 64-bit register X. The other operand (or only operand of a unary
19135 operation) is given to the custom arithmetic instruction with the
19136 least significant half in source register @var{src1} and the most
19137 significant half in @var{src2}. A custom instruction that returns a
19138 double-precision result returns the most significant 32 bits in the
19139 destination register and the other half in 32-bit register Y.
19140 GCC automatically generates the necessary code sequences to write
19141 register X and/or read register Y when double-precision floating-point
19142 instructions are used.
19147 Write @var{src1} into the least significant half of X and @var{src2} into
19148 the most significant half of X.
19151 Write @var{src1} into Y.
19153 @item @samp{frdxhi}, @samp{frdxlo}
19154 Read the most or least (respectively) significant half of X and store it in
19158 Read the value of Y and store it into @var{dest}.
19161 Note that you can gain more local control over generation of Nios II custom
19162 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
19163 and @code{target("no-custom-@var{insn}")} function attributes
19164 (@pxref{Function Attributes})
19165 or pragmas (@pxref{Function Specific Option Pragmas}).
19167 @item -mcustom-fpu-cfg=@var{name}
19168 @opindex mcustom-fpu-cfg
19170 This option enables a predefined, named set of custom instruction encodings
19171 (see @option{-mcustom-@var{insn}} above).
19172 Currently, the following sets are defined:
19174 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
19175 @gccoptlist{-mcustom-fmuls=252 @gol
19176 -mcustom-fadds=253 @gol
19177 -mcustom-fsubs=254 @gol
19178 -fsingle-precision-constant}
19180 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
19181 @gccoptlist{-mcustom-fmuls=252 @gol
19182 -mcustom-fadds=253 @gol
19183 -mcustom-fsubs=254 @gol
19184 -mcustom-fdivs=255 @gol
19185 -fsingle-precision-constant}
19187 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
19188 @gccoptlist{-mcustom-floatus=243 @gol
19189 -mcustom-fixsi=244 @gol
19190 -mcustom-floatis=245 @gol
19191 -mcustom-fcmpgts=246 @gol
19192 -mcustom-fcmples=249 @gol
19193 -mcustom-fcmpeqs=250 @gol
19194 -mcustom-fcmpnes=251 @gol
19195 -mcustom-fmuls=252 @gol
19196 -mcustom-fadds=253 @gol
19197 -mcustom-fsubs=254 @gol
19198 -mcustom-fdivs=255 @gol
19199 -fsingle-precision-constant}
19201 Custom instruction assignments given by individual
19202 @option{-mcustom-@var{insn}=} options override those given by
19203 @option{-mcustom-fpu-cfg=}, regardless of the
19204 order of the options on the command line.
19206 Note that you can gain more local control over selection of a FPU
19207 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
19208 function attribute (@pxref{Function Attributes})
19209 or pragma (@pxref{Function Specific Option Pragmas}).
19213 These additional @samp{-m} options are available for the Altera Nios II
19214 ELF (bare-metal) target:
19220 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
19221 startup and termination code, and is typically used in conjunction with
19222 @option{-msys-crt0=} to specify the location of the alternate startup code
19223 provided by the HAL BSP.
19227 Link with a limited version of the C library, @option{-lsmallc}, rather than
19230 @item -msys-crt0=@var{startfile}
19232 @var{startfile} is the file name of the startfile (crt0) to use
19233 when linking. This option is only useful in conjunction with @option{-mhal}.
19235 @item -msys-lib=@var{systemlib}
19237 @var{systemlib} is the library name of the library that provides
19238 low-level system calls required by the C library,
19239 e.g. @code{read} and @code{write}.
19240 This option is typically used to link with a library provided by a HAL BSP.
19244 @node PDP-11 Options
19245 @subsection PDP-11 Options
19246 @cindex PDP-11 Options
19248 These options are defined for the PDP-11:
19253 Use hardware FPP floating point. This is the default. (FIS floating
19254 point on the PDP-11/40 is not supported.)
19257 @opindex msoft-float
19258 Do not use hardware floating point.
19262 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
19266 Return floating-point results in memory. This is the default.
19270 Generate code for a PDP-11/40.
19274 Generate code for a PDP-11/45. This is the default.
19278 Generate code for a PDP-11/10.
19280 @item -mbcopy-builtin
19281 @opindex mbcopy-builtin
19282 Use inline @code{movmemhi} patterns for copying memory. This is the
19287 Do not use inline @code{movmemhi} patterns for copying memory.
19293 Use 16-bit @code{int}. This is the default.
19299 Use 32-bit @code{int}.
19302 @itemx -mno-float32
19304 @opindex mno-float32
19305 Use 64-bit @code{float}. This is the default.
19308 @itemx -mno-float64
19310 @opindex mno-float64
19311 Use 32-bit @code{float}.
19315 Use @code{abshi2} pattern. This is the default.
19319 Do not use @code{abshi2} pattern.
19321 @item -mbranch-expensive
19322 @opindex mbranch-expensive
19323 Pretend that branches are expensive. This is for experimenting with
19324 code generation only.
19326 @item -mbranch-cheap
19327 @opindex mbranch-cheap
19328 Do not pretend that branches are expensive. This is the default.
19332 Use Unix assembler syntax. This is the default when configured for
19333 @samp{pdp11-*-bsd}.
19337 Use DEC assembler syntax. This is the default when configured for any
19338 PDP-11 target other than @samp{pdp11-*-bsd}.
19341 @node picoChip Options
19342 @subsection picoChip Options
19343 @cindex picoChip options
19345 These @samp{-m} options are defined for picoChip implementations:
19349 @item -mae=@var{ae_type}
19351 Set the instruction set, register set, and instruction scheduling
19352 parameters for array element type @var{ae_type}. Supported values
19353 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
19355 @option{-mae=ANY} selects a completely generic AE type. Code
19356 generated with this option runs on any of the other AE types. The
19357 code is not as efficient as it would be if compiled for a specific
19358 AE type, and some types of operation (e.g., multiplication) do not
19359 work properly on all types of AE.
19361 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
19362 for compiled code, and is the default.
19364 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
19365 option may suffer from poor performance of byte (char) manipulation,
19366 since the DSP AE does not provide hardware support for byte load/stores.
19368 @item -msymbol-as-address
19369 Enable the compiler to directly use a symbol name as an address in a
19370 load/store instruction, without first loading it into a
19371 register. Typically, the use of this option generates larger
19372 programs, which run faster than when the option isn't used. However, the
19373 results vary from program to program, so it is left as a user option,
19374 rather than being permanently enabled.
19376 @item -mno-inefficient-warnings
19377 Disables warnings about the generation of inefficient code. These
19378 warnings can be generated, for example, when compiling code that
19379 performs byte-level memory operations on the MAC AE type. The MAC AE has
19380 no hardware support for byte-level memory operations, so all byte
19381 load/stores must be synthesized from word load/store operations. This is
19382 inefficient and a warning is generated to indicate
19383 that you should rewrite the code to avoid byte operations, or to target
19384 an AE type that has the necessary hardware support. This option disables
19389 @node PowerPC Options
19390 @subsection PowerPC Options
19391 @cindex PowerPC options
19393 These are listed under @xref{RS/6000 and PowerPC Options}.
19396 @subsection RL78 Options
19397 @cindex RL78 Options
19403 Links in additional target libraries to support operation within a
19410 Specifies the type of hardware multiplication support to be used. The
19411 default is @code{none}, which uses software multiplication functions.
19412 The @code{g13} option is for the hardware multiply/divide peripheral
19413 only on the RL78/G13 targets. The @code{rl78} option is for the
19414 standard hardware multiplication defined in the RL78 software manual.
19416 @item -m64bit-doubles
19417 @itemx -m32bit-doubles
19418 @opindex m64bit-doubles
19419 @opindex m32bit-doubles
19420 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
19421 or 32 bits (@option{-m32bit-doubles}) in size. The default is
19422 @option{-m32bit-doubles}.
19426 @node RS/6000 and PowerPC Options
19427 @subsection IBM RS/6000 and PowerPC Options
19428 @cindex RS/6000 and PowerPC Options
19429 @cindex IBM RS/6000 and PowerPC Options
19431 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
19433 @item -mpowerpc-gpopt
19434 @itemx -mno-powerpc-gpopt
19435 @itemx -mpowerpc-gfxopt
19436 @itemx -mno-powerpc-gfxopt
19439 @itemx -mno-powerpc64
19443 @itemx -mno-popcntb
19445 @itemx -mno-popcntd
19454 @itemx -mno-hard-dfp
19455 @opindex mpowerpc-gpopt
19456 @opindex mno-powerpc-gpopt
19457 @opindex mpowerpc-gfxopt
19458 @opindex mno-powerpc-gfxopt
19459 @opindex mpowerpc64
19460 @opindex mno-powerpc64
19464 @opindex mno-popcntb
19466 @opindex mno-popcntd
19472 @opindex mno-mfpgpr
19474 @opindex mno-hard-dfp
19475 You use these options to specify which instructions are available on the
19476 processor you are using. The default value of these options is
19477 determined when configuring GCC@. Specifying the
19478 @option{-mcpu=@var{cpu_type}} overrides the specification of these
19479 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
19480 rather than the options listed above.
19482 Specifying @option{-mpowerpc-gpopt} allows
19483 GCC to use the optional PowerPC architecture instructions in the
19484 General Purpose group, including floating-point square root. Specifying
19485 @option{-mpowerpc-gfxopt} allows GCC to
19486 use the optional PowerPC architecture instructions in the Graphics
19487 group, including floating-point select.
19489 The @option{-mmfcrf} option allows GCC to generate the move from
19490 condition register field instruction implemented on the POWER4
19491 processor and other processors that support the PowerPC V2.01
19493 The @option{-mpopcntb} option allows GCC to generate the popcount and
19494 double-precision FP reciprocal estimate instruction implemented on the
19495 POWER5 processor and other processors that support the PowerPC V2.02
19497 The @option{-mpopcntd} option allows GCC to generate the popcount
19498 instruction implemented on the POWER7 processor and other processors
19499 that support the PowerPC V2.06 architecture.
19500 The @option{-mfprnd} option allows GCC to generate the FP round to
19501 integer instructions implemented on the POWER5+ processor and other
19502 processors that support the PowerPC V2.03 architecture.
19503 The @option{-mcmpb} option allows GCC to generate the compare bytes
19504 instruction implemented on the POWER6 processor and other processors
19505 that support the PowerPC V2.05 architecture.
19506 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
19507 general-purpose register instructions implemented on the POWER6X
19508 processor and other processors that support the extended PowerPC V2.05
19510 The @option{-mhard-dfp} option allows GCC to generate the decimal
19511 floating-point instructions implemented on some POWER processors.
19513 The @option{-mpowerpc64} option allows GCC to generate the additional
19514 64-bit instructions that are found in the full PowerPC64 architecture
19515 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
19516 @option{-mno-powerpc64}.
19518 @item -mcpu=@var{cpu_type}
19520 Set architecture type, register usage, and
19521 instruction scheduling parameters for machine type @var{cpu_type}.
19522 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
19523 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
19524 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
19525 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
19526 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
19527 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
19528 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
19529 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
19530 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
19531 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8}, @samp{powerpc},
19532 @samp{powerpc64}, and @samp{rs64}.
19534 @option{-mcpu=powerpc}, and @option{-mcpu=powerpc64} specify pure 32-bit
19535 PowerPC and 64-bit PowerPC architecture machine
19536 types, with an appropriate, generic processor model assumed for
19537 scheduling purposes.
19539 The other options specify a specific processor. Code generated under
19540 those options runs best on that processor, and may not run at all on
19543 The @option{-mcpu} options automatically enable or disable the
19546 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
19547 -mpopcntb -mpopcntd -mpowerpc64 @gol
19548 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
19549 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
19550 -mcrypto -mdirect-move -mpower8-fusion -mpower8-vector @gol
19551 -mquad-memory -mquad-memory-atomic}
19553 The particular options set for any particular CPU varies between
19554 compiler versions, depending on what setting seems to produce optimal
19555 code for that CPU; it doesn't necessarily reflect the actual hardware's
19556 capabilities. If you wish to set an individual option to a particular
19557 value, you may specify it after the @option{-mcpu} option, like
19558 @option{-mcpu=970 -mno-altivec}.
19560 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
19561 not enabled or disabled by the @option{-mcpu} option at present because
19562 AIX does not have full support for these options. You may still
19563 enable or disable them individually if you're sure it'll work in your
19566 @item -mtune=@var{cpu_type}
19568 Set the instruction scheduling parameters for machine type
19569 @var{cpu_type}, but do not set the architecture type or register usage,
19570 as @option{-mcpu=@var{cpu_type}} does. The same
19571 values for @var{cpu_type} are used for @option{-mtune} as for
19572 @option{-mcpu}. If both are specified, the code generated uses the
19573 architecture and registers set by @option{-mcpu}, but the
19574 scheduling parameters set by @option{-mtune}.
19576 @item -mcmodel=small
19577 @opindex mcmodel=small
19578 Generate PowerPC64 code for the small model: The TOC is limited to
19581 @item -mcmodel=medium
19582 @opindex mcmodel=medium
19583 Generate PowerPC64 code for the medium model: The TOC and other static
19584 data may be up to a total of 4G in size.
19586 @item -mcmodel=large
19587 @opindex mcmodel=large
19588 Generate PowerPC64 code for the large model: The TOC may be up to 4G
19589 in size. Other data and code is only limited by the 64-bit address
19593 @itemx -mno-altivec
19595 @opindex mno-altivec
19596 Generate code that uses (does not use) AltiVec instructions, and also
19597 enable the use of built-in functions that allow more direct access to
19598 the AltiVec instruction set. You may also need to set
19599 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
19602 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
19603 @option{-maltivec=be}, the element order for Altivec intrinsics such
19604 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert} will
19605 match array element order corresponding to the endianness of the
19606 target. That is, element zero identifies the leftmost element in a
19607 vector register when targeting a big-endian platform, and identifies
19608 the rightmost element in a vector register when targeting a
19609 little-endian platform.
19612 @opindex maltivec=be
19613 Generate Altivec instructions using big-endian element order,
19614 regardless of whether the target is big- or little-endian. This is
19615 the default when targeting a big-endian platform.
19617 The element order is used to interpret element numbers in Altivec
19618 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19619 @code{vec_insert}. By default, these will match array element order
19620 corresponding to the endianness for the target.
19623 @opindex maltivec=le
19624 Generate Altivec instructions using little-endian element order,
19625 regardless of whether the target is big- or little-endian. This is
19626 the default when targeting a little-endian platform. This option is
19627 currently ignored when targeting a big-endian platform.
19629 The element order is used to interpret element numbers in Altivec
19630 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19631 @code{vec_insert}. By default, these will match array element order
19632 corresponding to the endianness for the target.
19637 @opindex mno-vrsave
19638 Generate VRSAVE instructions when generating AltiVec code.
19640 @item -mgen-cell-microcode
19641 @opindex mgen-cell-microcode
19642 Generate Cell microcode instructions.
19644 @item -mwarn-cell-microcode
19645 @opindex mwarn-cell-microcode
19646 Warn when a Cell microcode instruction is emitted. An example
19647 of a Cell microcode instruction is a variable shift.
19650 @opindex msecure-plt
19651 Generate code that allows @command{ld} and @command{ld.so}
19652 to build executables and shared
19653 libraries with non-executable @code{.plt} and @code{.got} sections.
19655 32-bit SYSV ABI option.
19659 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
19661 requires @code{.plt} and @code{.got}
19662 sections that are both writable and executable.
19663 This is a PowerPC 32-bit SYSV ABI option.
19669 This switch enables or disables the generation of ISEL instructions.
19671 @item -misel=@var{yes/no}
19672 This switch has been deprecated. Use @option{-misel} and
19673 @option{-mno-isel} instead.
19679 This switch enables or disables the generation of SPE simd
19685 @opindex mno-paired
19686 This switch enables or disables the generation of PAIRED simd
19689 @item -mspe=@var{yes/no}
19690 This option has been deprecated. Use @option{-mspe} and
19691 @option{-mno-spe} instead.
19697 Generate code that uses (does not use) vector/scalar (VSX)
19698 instructions, and also enable the use of built-in functions that allow
19699 more direct access to the VSX instruction set.
19704 @opindex mno-crypto
19705 Enable the use (disable) of the built-in functions that allow direct
19706 access to the cryptographic instructions that were added in version
19707 2.07 of the PowerPC ISA.
19709 @item -mdirect-move
19710 @itemx -mno-direct-move
19711 @opindex mdirect-move
19712 @opindex mno-direct-move
19713 Generate code that uses (does not use) the instructions to move data
19714 between the general purpose registers and the vector/scalar (VSX)
19715 registers that were added in version 2.07 of the PowerPC ISA.
19717 @item -mpower8-fusion
19718 @itemx -mno-power8-fusion
19719 @opindex mpower8-fusion
19720 @opindex mno-power8-fusion
19721 Generate code that keeps (does not keeps) some integer operations
19722 adjacent so that the instructions can be fused together on power8 and
19725 @item -mpower8-vector
19726 @itemx -mno-power8-vector
19727 @opindex mpower8-vector
19728 @opindex mno-power8-vector
19729 Generate code that uses (does not use) the vector and scalar
19730 instructions that were added in version 2.07 of the PowerPC ISA. Also
19731 enable the use of built-in functions that allow more direct access to
19732 the vector instructions.
19734 @item -mquad-memory
19735 @itemx -mno-quad-memory
19736 @opindex mquad-memory
19737 @opindex mno-quad-memory
19738 Generate code that uses (does not use) the non-atomic quad word memory
19739 instructions. The @option{-mquad-memory} option requires use of
19742 @item -mquad-memory-atomic
19743 @itemx -mno-quad-memory-atomic
19744 @opindex mquad-memory-atomic
19745 @opindex mno-quad-memory-atomic
19746 Generate code that uses (does not use) the atomic quad word memory
19747 instructions. The @option{-mquad-memory-atomic} option requires use of
19750 @item -mupper-regs-df
19751 @itemx -mno-upper-regs-df
19752 @opindex mupper-regs-df
19753 @opindex mno-upper-regs-df
19754 Generate code that uses (does not use) the scalar double precision
19755 instructions that target all 64 registers in the vector/scalar
19756 floating point register set that were added in version 2.06 of the
19757 PowerPC ISA. The @option{-mupper-regs-df} turned on by default if you
19758 use either of the @option{-mcpu=power7}, @option{-mcpu=power8}, or
19759 @option{-mvsx} options.
19761 @item -mupper-regs-sf
19762 @itemx -mno-upper-regs-sf
19763 @opindex mupper-regs-sf
19764 @opindex mno-upper-regs-sf
19765 Generate code that uses (does not use) the scalar single precision
19766 instructions that target all 64 registers in the vector/scalar
19767 floating point register set that were added in version 2.07 of the
19768 PowerPC ISA. The @option{-mupper-regs-sf} turned on by default if you
19769 use either of the @option{-mcpu=power8}, or @option{-mpower8-vector}
19773 @itemx -mno-upper-regs
19774 @opindex mupper-regs
19775 @opindex mno-upper-regs
19776 Generate code that uses (does not use) the scalar
19777 instructions that target all 64 registers in the vector/scalar
19778 floating point register set, depending on the model of the machine.
19780 If the @option{-mno-upper-regs} option was used, it will turn off both
19781 @option{-mupper-regs-sf} and @option{-mupper-regs-df} options.
19783 @item -mfloat-gprs=@var{yes/single/double/no}
19784 @itemx -mfloat-gprs
19785 @opindex mfloat-gprs
19786 This switch enables or disables the generation of floating-point
19787 operations on the general-purpose registers for architectures that
19790 The argument @var{yes} or @var{single} enables the use of
19791 single-precision floating-point operations.
19793 The argument @var{double} enables the use of single and
19794 double-precision floating-point operations.
19796 The argument @var{no} disables floating-point operations on the
19797 general-purpose registers.
19799 This option is currently only available on the MPC854x.
19805 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
19806 targets (including GNU/Linux). The 32-bit environment sets int, long
19807 and pointer to 32 bits and generates code that runs on any PowerPC
19808 variant. The 64-bit environment sets int to 32 bits and long and
19809 pointer to 64 bits, and generates code for PowerPC64, as for
19810 @option{-mpowerpc64}.
19813 @itemx -mno-fp-in-toc
19814 @itemx -mno-sum-in-toc
19815 @itemx -mminimal-toc
19817 @opindex mno-fp-in-toc
19818 @opindex mno-sum-in-toc
19819 @opindex mminimal-toc
19820 Modify generation of the TOC (Table Of Contents), which is created for
19821 every executable file. The @option{-mfull-toc} option is selected by
19822 default. In that case, GCC allocates at least one TOC entry for
19823 each unique non-automatic variable reference in your program. GCC
19824 also places floating-point constants in the TOC@. However, only
19825 16,384 entries are available in the TOC@.
19827 If you receive a linker error message that saying you have overflowed
19828 the available TOC space, you can reduce the amount of TOC space used
19829 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
19830 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
19831 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
19832 generate code to calculate the sum of an address and a constant at
19833 run time instead of putting that sum into the TOC@. You may specify one
19834 or both of these options. Each causes GCC to produce very slightly
19835 slower and larger code at the expense of conserving TOC space.
19837 If you still run out of space in the TOC even when you specify both of
19838 these options, specify @option{-mminimal-toc} instead. This option causes
19839 GCC to make only one TOC entry for every file. When you specify this
19840 option, GCC produces code that is slower and larger but which
19841 uses extremely little TOC space. You may wish to use this option
19842 only on files that contain less frequently-executed code.
19848 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
19849 @code{long} type, and the infrastructure needed to support them.
19850 Specifying @option{-maix64} implies @option{-mpowerpc64},
19851 while @option{-maix32} disables the 64-bit ABI and
19852 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
19855 @itemx -mno-xl-compat
19856 @opindex mxl-compat
19857 @opindex mno-xl-compat
19858 Produce code that conforms more closely to IBM XL compiler semantics
19859 when using AIX-compatible ABI@. Pass floating-point arguments to
19860 prototyped functions beyond the register save area (RSA) on the stack
19861 in addition to argument FPRs. Do not assume that most significant
19862 double in 128-bit long double value is properly rounded when comparing
19863 values and converting to double. Use XL symbol names for long double
19866 The AIX calling convention was extended but not initially documented to
19867 handle an obscure K&R C case of calling a function that takes the
19868 address of its arguments with fewer arguments than declared. IBM XL
19869 compilers access floating-point arguments that do not fit in the
19870 RSA from the stack when a subroutine is compiled without
19871 optimization. Because always storing floating-point arguments on the
19872 stack is inefficient and rarely needed, this option is not enabled by
19873 default and only is necessary when calling subroutines compiled by IBM
19874 XL compilers without optimization.
19878 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
19879 application written to use message passing with special startup code to
19880 enable the application to run. The system must have PE installed in the
19881 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
19882 must be overridden with the @option{-specs=} option to specify the
19883 appropriate directory location. The Parallel Environment does not
19884 support threads, so the @option{-mpe} option and the @option{-pthread}
19885 option are incompatible.
19887 @item -malign-natural
19888 @itemx -malign-power
19889 @opindex malign-natural
19890 @opindex malign-power
19891 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
19892 @option{-malign-natural} overrides the ABI-defined alignment of larger
19893 types, such as floating-point doubles, on their natural size-based boundary.
19894 The option @option{-malign-power} instructs GCC to follow the ABI-specified
19895 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
19897 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
19901 @itemx -mhard-float
19902 @opindex msoft-float
19903 @opindex mhard-float
19904 Generate code that does not use (uses) the floating-point register set.
19905 Software floating-point emulation is provided if you use the
19906 @option{-msoft-float} option, and pass the option to GCC when linking.
19908 @item -msingle-float
19909 @itemx -mdouble-float
19910 @opindex msingle-float
19911 @opindex mdouble-float
19912 Generate code for single- or double-precision floating-point operations.
19913 @option{-mdouble-float} implies @option{-msingle-float}.
19916 @opindex msimple-fpu
19917 Do not generate @code{sqrt} and @code{div} instructions for hardware
19918 floating-point unit.
19920 @item -mfpu=@var{name}
19922 Specify type of floating-point unit. Valid values for @var{name} are
19923 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
19924 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
19925 @samp{sp_full} (equivalent to @option{-msingle-float}),
19926 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
19929 @opindex mxilinx-fpu
19930 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
19933 @itemx -mno-multiple
19935 @opindex mno-multiple
19936 Generate code that uses (does not use) the load multiple word
19937 instructions and the store multiple word instructions. These
19938 instructions are generated by default on POWER systems, and not
19939 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
19940 PowerPC systems, since those instructions do not work when the
19941 processor is in little-endian mode. The exceptions are PPC740 and
19942 PPC750 which permit these instructions in little-endian mode.
19947 @opindex mno-string
19948 Generate code that uses (does not use) the load string instructions
19949 and the store string word instructions to save multiple registers and
19950 do small block moves. These instructions are generated by default on
19951 POWER systems, and not generated on PowerPC systems. Do not use
19952 @option{-mstring} on little-endian PowerPC systems, since those
19953 instructions do not work when the processor is in little-endian mode.
19954 The exceptions are PPC740 and PPC750 which permit these instructions
19955 in little-endian mode.
19960 @opindex mno-update
19961 Generate code that uses (does not use) the load or store instructions
19962 that update the base register to the address of the calculated memory
19963 location. These instructions are generated by default. If you use
19964 @option{-mno-update}, there is a small window between the time that the
19965 stack pointer is updated and the address of the previous frame is
19966 stored, which means code that walks the stack frame across interrupts or
19967 signals may get corrupted data.
19969 @item -mavoid-indexed-addresses
19970 @itemx -mno-avoid-indexed-addresses
19971 @opindex mavoid-indexed-addresses
19972 @opindex mno-avoid-indexed-addresses
19973 Generate code that tries to avoid (not avoid) the use of indexed load
19974 or store instructions. These instructions can incur a performance
19975 penalty on Power6 processors in certain situations, such as when
19976 stepping through large arrays that cross a 16M boundary. This option
19977 is enabled by default when targeting Power6 and disabled otherwise.
19980 @itemx -mno-fused-madd
19981 @opindex mfused-madd
19982 @opindex mno-fused-madd
19983 Generate code that uses (does not use) the floating-point multiply and
19984 accumulate instructions. These instructions are generated by default
19985 if hardware floating point is used. The machine-dependent
19986 @option{-mfused-madd} option is now mapped to the machine-independent
19987 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
19988 mapped to @option{-ffp-contract=off}.
19994 Generate code that uses (does not use) the half-word multiply and
19995 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
19996 These instructions are generated by default when targeting those
20003 Generate code that uses (does not use) the string-search @samp{dlmzb}
20004 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
20005 generated by default when targeting those processors.
20007 @item -mno-bit-align
20009 @opindex mno-bit-align
20010 @opindex mbit-align
20011 On System V.4 and embedded PowerPC systems do not (do) force structures
20012 and unions that contain bit-fields to be aligned to the base type of the
20015 For example, by default a structure containing nothing but 8
20016 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
20017 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
20018 the structure is aligned to a 1-byte boundary and is 1 byte in
20021 @item -mno-strict-align
20022 @itemx -mstrict-align
20023 @opindex mno-strict-align
20024 @opindex mstrict-align
20025 On System V.4 and embedded PowerPC systems do not (do) assume that
20026 unaligned memory references are handled by the system.
20028 @item -mrelocatable
20029 @itemx -mno-relocatable
20030 @opindex mrelocatable
20031 @opindex mno-relocatable
20032 Generate code that allows (does not allow) a static executable to be
20033 relocated to a different address at run time. A simple embedded
20034 PowerPC system loader should relocate the entire contents of
20035 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
20036 a table of 32-bit addresses generated by this option. For this to
20037 work, all objects linked together must be compiled with
20038 @option{-mrelocatable} or @option{-mrelocatable-lib}.
20039 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
20041 @item -mrelocatable-lib
20042 @itemx -mno-relocatable-lib
20043 @opindex mrelocatable-lib
20044 @opindex mno-relocatable-lib
20045 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
20046 @code{.fixup} section to allow static executables to be relocated at
20047 run time, but @option{-mrelocatable-lib} does not use the smaller stack
20048 alignment of @option{-mrelocatable}. Objects compiled with
20049 @option{-mrelocatable-lib} may be linked with objects compiled with
20050 any combination of the @option{-mrelocatable} options.
20056 On System V.4 and embedded PowerPC systems do not (do) assume that
20057 register 2 contains a pointer to a global area pointing to the addresses
20058 used in the program.
20061 @itemx -mlittle-endian
20063 @opindex mlittle-endian
20064 On System V.4 and embedded PowerPC systems compile code for the
20065 processor in little-endian mode. The @option{-mlittle-endian} option is
20066 the same as @option{-mlittle}.
20069 @itemx -mbig-endian
20071 @opindex mbig-endian
20072 On System V.4 and embedded PowerPC systems compile code for the
20073 processor in big-endian mode. The @option{-mbig-endian} option is
20074 the same as @option{-mbig}.
20076 @item -mdynamic-no-pic
20077 @opindex mdynamic-no-pic
20078 On Darwin and Mac OS X systems, compile code so that it is not
20079 relocatable, but that its external references are relocatable. The
20080 resulting code is suitable for applications, but not shared
20083 @item -msingle-pic-base
20084 @opindex msingle-pic-base
20085 Treat the register used for PIC addressing as read-only, rather than
20086 loading it in the prologue for each function. The runtime system is
20087 responsible for initializing this register with an appropriate value
20088 before execution begins.
20090 @item -mprioritize-restricted-insns=@var{priority}
20091 @opindex mprioritize-restricted-insns
20092 This option controls the priority that is assigned to
20093 dispatch-slot restricted instructions during the second scheduling
20094 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
20095 or @samp{2} to assign no, highest, or second-highest (respectively)
20096 priority to dispatch-slot restricted
20099 @item -msched-costly-dep=@var{dependence_type}
20100 @opindex msched-costly-dep
20101 This option controls which dependences are considered costly
20102 by the target during instruction scheduling. The argument
20103 @var{dependence_type} takes one of the following values:
20107 No dependence is costly.
20110 All dependences are costly.
20112 @item @samp{true_store_to_load}
20113 A true dependence from store to load is costly.
20115 @item @samp{store_to_load}
20116 Any dependence from store to load is costly.
20119 Any dependence for which the latency is greater than or equal to
20120 @var{number} is costly.
20123 @item -minsert-sched-nops=@var{scheme}
20124 @opindex minsert-sched-nops
20125 This option controls which NOP insertion scheme is used during
20126 the second scheduling pass. The argument @var{scheme} takes one of the
20134 Pad with NOPs any dispatch group that has vacant issue slots,
20135 according to the scheduler's grouping.
20137 @item @samp{regroup_exact}
20138 Insert NOPs to force costly dependent insns into
20139 separate groups. Insert exactly as many NOPs as needed to force an insn
20140 to a new group, according to the estimated processor grouping.
20143 Insert NOPs to force costly dependent insns into
20144 separate groups. Insert @var{number} NOPs to force an insn to a new group.
20148 @opindex mcall-sysv
20149 On System V.4 and embedded PowerPC systems compile code using calling
20150 conventions that adhere to the March 1995 draft of the System V
20151 Application Binary Interface, PowerPC processor supplement. This is the
20152 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
20154 @item -mcall-sysv-eabi
20156 @opindex mcall-sysv-eabi
20157 @opindex mcall-eabi
20158 Specify both @option{-mcall-sysv} and @option{-meabi} options.
20160 @item -mcall-sysv-noeabi
20161 @opindex mcall-sysv-noeabi
20162 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
20164 @item -mcall-aixdesc
20166 On System V.4 and embedded PowerPC systems compile code for the AIX
20170 @opindex mcall-linux
20171 On System V.4 and embedded PowerPC systems compile code for the
20172 Linux-based GNU system.
20174 @item -mcall-freebsd
20175 @opindex mcall-freebsd
20176 On System V.4 and embedded PowerPC systems compile code for the
20177 FreeBSD operating system.
20179 @item -mcall-netbsd
20180 @opindex mcall-netbsd
20181 On System V.4 and embedded PowerPC systems compile code for the
20182 NetBSD operating system.
20184 @item -mcall-openbsd
20185 @opindex mcall-netbsd
20186 On System V.4 and embedded PowerPC systems compile code for the
20187 OpenBSD operating system.
20189 @item -maix-struct-return
20190 @opindex maix-struct-return
20191 Return all structures in memory (as specified by the AIX ABI)@.
20193 @item -msvr4-struct-return
20194 @opindex msvr4-struct-return
20195 Return structures smaller than 8 bytes in registers (as specified by the
20198 @item -mabi=@var{abi-type}
20200 Extend the current ABI with a particular extension, or remove such extension.
20201 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
20202 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble},
20203 @var{elfv1}, @var{elfv2}@.
20207 Extend the current ABI with SPE ABI extensions. This does not change
20208 the default ABI, instead it adds the SPE ABI extensions to the current
20212 @opindex mabi=no-spe
20213 Disable Book-E SPE ABI extensions for the current ABI@.
20215 @item -mabi=ibmlongdouble
20216 @opindex mabi=ibmlongdouble
20217 Change the current ABI to use IBM extended-precision long double.
20218 This is a PowerPC 32-bit SYSV ABI option.
20220 @item -mabi=ieeelongdouble
20221 @opindex mabi=ieeelongdouble
20222 Change the current ABI to use IEEE extended-precision long double.
20223 This is a PowerPC 32-bit Linux ABI option.
20226 @opindex mabi=elfv1
20227 Change the current ABI to use the ELFv1 ABI.
20228 This is the default ABI for big-endian PowerPC 64-bit Linux.
20229 Overriding the default ABI requires special system support and is
20230 likely to fail in spectacular ways.
20233 @opindex mabi=elfv2
20234 Change the current ABI to use the ELFv2 ABI.
20235 This is the default ABI for little-endian PowerPC 64-bit Linux.
20236 Overriding the default ABI requires special system support and is
20237 likely to fail in spectacular ways.
20240 @itemx -mno-prototype
20241 @opindex mprototype
20242 @opindex mno-prototype
20243 On System V.4 and embedded PowerPC systems assume that all calls to
20244 variable argument functions are properly prototyped. Otherwise, the
20245 compiler must insert an instruction before every non-prototyped call to
20246 set or clear bit 6 of the condition code register (@var{CR}) to
20247 indicate whether floating-point values are passed in the floating-point
20248 registers in case the function takes variable arguments. With
20249 @option{-mprototype}, only calls to prototyped variable argument functions
20250 set or clear the bit.
20254 On embedded PowerPC systems, assume that the startup module is called
20255 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
20256 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
20261 On embedded PowerPC systems, assume that the startup module is called
20262 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
20267 On embedded PowerPC systems, assume that the startup module is called
20268 @file{crt0.o} and the standard C libraries are @file{libads.a} and
20271 @item -myellowknife
20272 @opindex myellowknife
20273 On embedded PowerPC systems, assume that the startup module is called
20274 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
20279 On System V.4 and embedded PowerPC systems, specify that you are
20280 compiling for a VxWorks system.
20284 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
20285 header to indicate that @samp{eabi} extended relocations are used.
20291 On System V.4 and embedded PowerPC systems do (do not) adhere to the
20292 Embedded Applications Binary Interface (EABI), which is a set of
20293 modifications to the System V.4 specifications. Selecting @option{-meabi}
20294 means that the stack is aligned to an 8-byte boundary, a function
20295 @code{__eabi} is called from @code{main} to set up the EABI
20296 environment, and the @option{-msdata} option can use both @code{r2} and
20297 @code{r13} to point to two separate small data areas. Selecting
20298 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
20299 no EABI initialization function is called from @code{main}, and the
20300 @option{-msdata} option only uses @code{r13} to point to a single
20301 small data area. The @option{-meabi} option is on by default if you
20302 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
20305 @opindex msdata=eabi
20306 On System V.4 and embedded PowerPC systems, put small initialized
20307 @code{const} global and static data in the @samp{.sdata2} section, which
20308 is pointed to by register @code{r2}. Put small initialized
20309 non-@code{const} global and static data in the @samp{.sdata} section,
20310 which is pointed to by register @code{r13}. Put small uninitialized
20311 global and static data in the @samp{.sbss} section, which is adjacent to
20312 the @samp{.sdata} section. The @option{-msdata=eabi} option is
20313 incompatible with the @option{-mrelocatable} option. The
20314 @option{-msdata=eabi} option also sets the @option{-memb} option.
20317 @opindex msdata=sysv
20318 On System V.4 and embedded PowerPC systems, put small global and static
20319 data in the @samp{.sdata} section, which is pointed to by register
20320 @code{r13}. Put small uninitialized global and static data in the
20321 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
20322 The @option{-msdata=sysv} option is incompatible with the
20323 @option{-mrelocatable} option.
20325 @item -msdata=default
20327 @opindex msdata=default
20329 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
20330 compile code the same as @option{-msdata=eabi}, otherwise compile code the
20331 same as @option{-msdata=sysv}.
20334 @opindex msdata=data
20335 On System V.4 and embedded PowerPC systems, put small global
20336 data in the @samp{.sdata} section. Put small uninitialized global
20337 data in the @samp{.sbss} section. Do not use register @code{r13}
20338 to address small data however. This is the default behavior unless
20339 other @option{-msdata} options are used.
20343 @opindex msdata=none
20345 On embedded PowerPC systems, put all initialized global and static data
20346 in the @samp{.data} section, and all uninitialized data in the
20347 @samp{.bss} section.
20349 @item -mblock-move-inline-limit=@var{num}
20350 @opindex mblock-move-inline-limit
20351 Inline all block moves (such as calls to @code{memcpy} or structure
20352 copies) less than or equal to @var{num} bytes. The minimum value for
20353 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
20354 targets. The default value is target-specific.
20358 @cindex smaller data references (PowerPC)
20359 @cindex .sdata/.sdata2 references (PowerPC)
20360 On embedded PowerPC systems, put global and static items less than or
20361 equal to @var{num} bytes into the small data or BSS sections instead of
20362 the normal data or BSS section. By default, @var{num} is 8. The
20363 @option{-G @var{num}} switch is also passed to the linker.
20364 All modules should be compiled with the same @option{-G @var{num}} value.
20367 @itemx -mno-regnames
20369 @opindex mno-regnames
20370 On System V.4 and embedded PowerPC systems do (do not) emit register
20371 names in the assembly language output using symbolic forms.
20374 @itemx -mno-longcall
20376 @opindex mno-longcall
20377 By default assume that all calls are far away so that a longer and more
20378 expensive calling sequence is required. This is required for calls
20379 farther than 32 megabytes (33,554,432 bytes) from the current location.
20380 A short call is generated if the compiler knows
20381 the call cannot be that far away. This setting can be overridden by
20382 the @code{shortcall} function attribute, or by @code{#pragma
20385 Some linkers are capable of detecting out-of-range calls and generating
20386 glue code on the fly. On these systems, long calls are unnecessary and
20387 generate slower code. As of this writing, the AIX linker can do this,
20388 as can the GNU linker for PowerPC/64. It is planned to add this feature
20389 to the GNU linker for 32-bit PowerPC systems as well.
20391 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
20392 callee, L42}, plus a @dfn{branch island} (glue code). The two target
20393 addresses represent the callee and the branch island. The
20394 Darwin/PPC linker prefers the first address and generates a @code{bl
20395 callee} if the PPC @code{bl} instruction reaches the callee directly;
20396 otherwise, the linker generates @code{bl L42} to call the branch
20397 island. The branch island is appended to the body of the
20398 calling function; it computes the full 32-bit address of the callee
20401 On Mach-O (Darwin) systems, this option directs the compiler emit to
20402 the glue for every direct call, and the Darwin linker decides whether
20403 to use or discard it.
20405 In the future, GCC may ignore all longcall specifications
20406 when the linker is known to generate glue.
20408 @item -mtls-markers
20409 @itemx -mno-tls-markers
20410 @opindex mtls-markers
20411 @opindex mno-tls-markers
20412 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
20413 specifying the function argument. The relocation allows the linker to
20414 reliably associate function call with argument setup instructions for
20415 TLS optimization, which in turn allows GCC to better schedule the
20420 Adds support for multithreading with the @dfn{pthreads} library.
20421 This option sets flags for both the preprocessor and linker.
20426 This option enables use of the reciprocal estimate and
20427 reciprocal square root estimate instructions with additional
20428 Newton-Raphson steps to increase precision instead of doing a divide or
20429 square root and divide for floating-point arguments. You should use
20430 the @option{-ffast-math} option when using @option{-mrecip} (or at
20431 least @option{-funsafe-math-optimizations},
20432 @option{-finite-math-only}, @option{-freciprocal-math} and
20433 @option{-fno-trapping-math}). Note that while the throughput of the
20434 sequence is generally higher than the throughput of the non-reciprocal
20435 instruction, the precision of the sequence can be decreased by up to 2
20436 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
20439 @item -mrecip=@var{opt}
20440 @opindex mrecip=opt
20441 This option controls which reciprocal estimate instructions
20442 may be used. @var{opt} is a comma-separated list of options, which may
20443 be preceded by a @code{!} to invert the option:
20444 @code{all}: enable all estimate instructions,
20445 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
20446 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
20447 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
20448 @code{divf}: enable the single-precision reciprocal approximation instructions;
20449 @code{divd}: enable the double-precision reciprocal approximation instructions;
20450 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
20451 @code{rsqrtf}: enable the single-precision reciprocal square root approximation instructions;
20452 @code{rsqrtd}: enable the double-precision reciprocal square root approximation instructions;
20454 So, for example, @option{-mrecip=all,!rsqrtd} enables
20455 all of the reciprocal estimate instructions, except for the
20456 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
20457 which handle the double-precision reciprocal square root calculations.
20459 @item -mrecip-precision
20460 @itemx -mno-recip-precision
20461 @opindex mrecip-precision
20462 Assume (do not assume) that the reciprocal estimate instructions
20463 provide higher-precision estimates than is mandated by the PowerPC
20464 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
20465 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
20466 The double-precision square root estimate instructions are not generated by
20467 default on low-precision machines, since they do not provide an
20468 estimate that converges after three steps.
20470 @item -mveclibabi=@var{type}
20471 @opindex mveclibabi
20472 Specifies the ABI type to use for vectorizing intrinsics using an
20473 external library. The only type supported at present is @code{mass},
20474 which specifies to use IBM's Mathematical Acceleration Subsystem
20475 (MASS) libraries for vectorizing intrinsics using external libraries.
20476 GCC currently emits calls to @code{acosd2}, @code{acosf4},
20477 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
20478 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
20479 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
20480 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
20481 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
20482 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
20483 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
20484 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
20485 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
20486 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
20487 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
20488 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
20489 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
20490 for power7. Both @option{-ftree-vectorize} and
20491 @option{-funsafe-math-optimizations} must also be enabled. The MASS
20492 libraries must be specified at link time.
20497 Generate (do not generate) the @code{friz} instruction when the
20498 @option{-funsafe-math-optimizations} option is used to optimize
20499 rounding of floating-point values to 64-bit integer and back to floating
20500 point. The @code{friz} instruction does not return the same value if
20501 the floating-point number is too large to fit in an integer.
20503 @item -mpointers-to-nested-functions
20504 @itemx -mno-pointers-to-nested-functions
20505 @opindex mpointers-to-nested-functions
20506 Generate (do not generate) code to load up the static chain register
20507 (@var{r11}) when calling through a pointer on AIX and 64-bit Linux
20508 systems where a function pointer points to a 3-word descriptor giving
20509 the function address, TOC value to be loaded in register @var{r2}, and
20510 static chain value to be loaded in register @var{r11}. The
20511 @option{-mpointers-to-nested-functions} is on by default. You cannot
20512 call through pointers to nested functions or pointers
20513 to functions compiled in other languages that use the static chain if
20514 you use the @option{-mno-pointers-to-nested-functions}.
20516 @item -msave-toc-indirect
20517 @itemx -mno-save-toc-indirect
20518 @opindex msave-toc-indirect
20519 Generate (do not generate) code to save the TOC value in the reserved
20520 stack location in the function prologue if the function calls through
20521 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
20522 saved in the prologue, it is saved just before the call through the
20523 pointer. The @option{-mno-save-toc-indirect} option is the default.
20525 @item -mcompat-align-parm
20526 @itemx -mno-compat-align-parm
20527 @opindex mcompat-align-parm
20528 Generate (do not generate) code to pass structure parameters with a
20529 maximum alignment of 64 bits, for compatibility with older versions
20532 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
20533 structure parameter on a 128-bit boundary when that structure contained
20534 a member requiring 128-bit alignment. This is corrected in more
20535 recent versions of GCC. This option may be used to generate code
20536 that is compatible with functions compiled with older versions of
20539 The @option{-mno-compat-align-parm} option is the default.
20543 @subsection RX Options
20546 These command-line options are defined for RX targets:
20549 @item -m64bit-doubles
20550 @itemx -m32bit-doubles
20551 @opindex m64bit-doubles
20552 @opindex m32bit-doubles
20553 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
20554 or 32 bits (@option{-m32bit-doubles}) in size. The default is
20555 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
20556 works on 32-bit values, which is why the default is
20557 @option{-m32bit-doubles}.
20563 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
20564 floating-point hardware. The default is enabled for the @var{RX600}
20565 series and disabled for the @var{RX200} series.
20567 Floating-point instructions are only generated for 32-bit floating-point
20568 values, however, so the FPU hardware is not used for doubles if the
20569 @option{-m64bit-doubles} option is used.
20571 @emph{Note} If the @option{-fpu} option is enabled then
20572 @option{-funsafe-math-optimizations} is also enabled automatically.
20573 This is because the RX FPU instructions are themselves unsafe.
20575 @item -mcpu=@var{name}
20577 Selects the type of RX CPU to be targeted. Currently three types are
20578 supported, the generic @var{RX600} and @var{RX200} series hardware and
20579 the specific @var{RX610} CPU. The default is @var{RX600}.
20581 The only difference between @var{RX600} and @var{RX610} is that the
20582 @var{RX610} does not support the @code{MVTIPL} instruction.
20584 The @var{RX200} series does not have a hardware floating-point unit
20585 and so @option{-nofpu} is enabled by default when this type is
20588 @item -mbig-endian-data
20589 @itemx -mlittle-endian-data
20590 @opindex mbig-endian-data
20591 @opindex mlittle-endian-data
20592 Store data (but not code) in the big-endian format. The default is
20593 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
20596 @item -msmall-data-limit=@var{N}
20597 @opindex msmall-data-limit
20598 Specifies the maximum size in bytes of global and static variables
20599 which can be placed into the small data area. Using the small data
20600 area can lead to smaller and faster code, but the size of area is
20601 limited and it is up to the programmer to ensure that the area does
20602 not overflow. Also when the small data area is used one of the RX's
20603 registers (usually @code{r13}) is reserved for use pointing to this
20604 area, so it is no longer available for use by the compiler. This
20605 could result in slower and/or larger code if variables are pushed onto
20606 the stack instead of being held in this register.
20608 Note, common variables (variables that have not been initialized) and
20609 constants are not placed into the small data area as they are assigned
20610 to other sections in the output executable.
20612 The default value is zero, which disables this feature. Note, this
20613 feature is not enabled by default with higher optimization levels
20614 (@option{-O2} etc) because of the potentially detrimental effects of
20615 reserving a register. It is up to the programmer to experiment and
20616 discover whether this feature is of benefit to their program. See the
20617 description of the @option{-mpid} option for a description of how the
20618 actual register to hold the small data area pointer is chosen.
20624 Use the simulator runtime. The default is to use the libgloss
20625 board-specific runtime.
20627 @item -mas100-syntax
20628 @itemx -mno-as100-syntax
20629 @opindex mas100-syntax
20630 @opindex mno-as100-syntax
20631 When generating assembler output use a syntax that is compatible with
20632 Renesas's AS100 assembler. This syntax can also be handled by the GAS
20633 assembler, but it has some restrictions so it is not generated by default.
20635 @item -mmax-constant-size=@var{N}
20636 @opindex mmax-constant-size
20637 Specifies the maximum size, in bytes, of a constant that can be used as
20638 an operand in a RX instruction. Although the RX instruction set does
20639 allow constants of up to 4 bytes in length to be used in instructions,
20640 a longer value equates to a longer instruction. Thus in some
20641 circumstances it can be beneficial to restrict the size of constants
20642 that are used in instructions. Constants that are too big are instead
20643 placed into a constant pool and referenced via register indirection.
20645 The value @var{N} can be between 0 and 4. A value of 0 (the default)
20646 or 4 means that constants of any size are allowed.
20650 Enable linker relaxation. Linker relaxation is a process whereby the
20651 linker attempts to reduce the size of a program by finding shorter
20652 versions of various instructions. Disabled by default.
20654 @item -mint-register=@var{N}
20655 @opindex mint-register
20656 Specify the number of registers to reserve for fast interrupt handler
20657 functions. The value @var{N} can be between 0 and 4. A value of 1
20658 means that register @code{r13} is reserved for the exclusive use
20659 of fast interrupt handlers. A value of 2 reserves @code{r13} and
20660 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
20661 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
20662 A value of 0, the default, does not reserve any registers.
20664 @item -msave-acc-in-interrupts
20665 @opindex msave-acc-in-interrupts
20666 Specifies that interrupt handler functions should preserve the
20667 accumulator register. This is only necessary if normal code might use
20668 the accumulator register, for example because it performs 64-bit
20669 multiplications. The default is to ignore the accumulator as this
20670 makes the interrupt handlers faster.
20676 Enables the generation of position independent data. When enabled any
20677 access to constant data is done via an offset from a base address
20678 held in a register. This allows the location of constant data to be
20679 determined at run time without requiring the executable to be
20680 relocated, which is a benefit to embedded applications with tight
20681 memory constraints. Data that can be modified is not affected by this
20684 Note, using this feature reserves a register, usually @code{r13}, for
20685 the constant data base address. This can result in slower and/or
20686 larger code, especially in complicated functions.
20688 The actual register chosen to hold the constant data base address
20689 depends upon whether the @option{-msmall-data-limit} and/or the
20690 @option{-mint-register} command-line options are enabled. Starting
20691 with register @code{r13} and proceeding downwards, registers are
20692 allocated first to satisfy the requirements of @option{-mint-register},
20693 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
20694 is possible for the small data area register to be @code{r8} if both
20695 @option{-mint-register=4} and @option{-mpid} are specified on the
20698 By default this feature is not enabled. The default can be restored
20699 via the @option{-mno-pid} command-line option.
20701 @item -mno-warn-multiple-fast-interrupts
20702 @itemx -mwarn-multiple-fast-interrupts
20703 @opindex mno-warn-multiple-fast-interrupts
20704 @opindex mwarn-multiple-fast-interrupts
20705 Prevents GCC from issuing a warning message if it finds more than one
20706 fast interrupt handler when it is compiling a file. The default is to
20707 issue a warning for each extra fast interrupt handler found, as the RX
20708 only supports one such interrupt.
20712 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
20713 has special significance to the RX port when used with the
20714 @code{interrupt} function attribute. This attribute indicates a
20715 function intended to process fast interrupts. GCC ensures
20716 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
20717 and/or @code{r13} and only provided that the normal use of the
20718 corresponding registers have been restricted via the
20719 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
20722 @node S/390 and zSeries Options
20723 @subsection S/390 and zSeries Options
20724 @cindex S/390 and zSeries Options
20726 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
20730 @itemx -msoft-float
20731 @opindex mhard-float
20732 @opindex msoft-float
20733 Use (do not use) the hardware floating-point instructions and registers
20734 for floating-point operations. When @option{-msoft-float} is specified,
20735 functions in @file{libgcc.a} are used to perform floating-point
20736 operations. When @option{-mhard-float} is specified, the compiler
20737 generates IEEE floating-point instructions. This is the default.
20740 @itemx -mno-hard-dfp
20742 @opindex mno-hard-dfp
20743 Use (do not use) the hardware decimal-floating-point instructions for
20744 decimal-floating-point operations. When @option{-mno-hard-dfp} is
20745 specified, functions in @file{libgcc.a} are used to perform
20746 decimal-floating-point operations. When @option{-mhard-dfp} is
20747 specified, the compiler generates decimal-floating-point hardware
20748 instructions. This is the default for @option{-march=z9-ec} or higher.
20750 @item -mlong-double-64
20751 @itemx -mlong-double-128
20752 @opindex mlong-double-64
20753 @opindex mlong-double-128
20754 These switches control the size of @code{long double} type. A size
20755 of 64 bits makes the @code{long double} type equivalent to the @code{double}
20756 type. This is the default.
20759 @itemx -mno-backchain
20760 @opindex mbackchain
20761 @opindex mno-backchain
20762 Store (do not store) the address of the caller's frame as backchain pointer
20763 into the callee's stack frame.
20764 A backchain may be needed to allow debugging using tools that do not understand
20765 DWARF 2 call frame information.
20766 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
20767 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
20768 the backchain is placed into the topmost word of the 96/160 byte register
20771 In general, code compiled with @option{-mbackchain} is call-compatible with
20772 code compiled with @option{-mmo-backchain}; however, use of the backchain
20773 for debugging purposes usually requires that the whole binary is built with
20774 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
20775 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20776 to build a linux kernel use @option{-msoft-float}.
20778 The default is to not maintain the backchain.
20780 @item -mpacked-stack
20781 @itemx -mno-packed-stack
20782 @opindex mpacked-stack
20783 @opindex mno-packed-stack
20784 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
20785 specified, the compiler uses the all fields of the 96/160 byte register save
20786 area only for their default purpose; unused fields still take up stack space.
20787 When @option{-mpacked-stack} is specified, register save slots are densely
20788 packed at the top of the register save area; unused space is reused for other
20789 purposes, allowing for more efficient use of the available stack space.
20790 However, when @option{-mbackchain} is also in effect, the topmost word of
20791 the save area is always used to store the backchain, and the return address
20792 register is always saved two words below the backchain.
20794 As long as the stack frame backchain is not used, code generated with
20795 @option{-mpacked-stack} is call-compatible with code generated with
20796 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
20797 S/390 or zSeries generated code that uses the stack frame backchain at run
20798 time, not just for debugging purposes. Such code is not call-compatible
20799 with code compiled with @option{-mpacked-stack}. Also, note that the
20800 combination of @option{-mbackchain},
20801 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20802 to build a linux kernel use @option{-msoft-float}.
20804 The default is to not use the packed stack layout.
20807 @itemx -mno-small-exec
20808 @opindex msmall-exec
20809 @opindex mno-small-exec
20810 Generate (or do not generate) code using the @code{bras} instruction
20811 to do subroutine calls.
20812 This only works reliably if the total executable size does not
20813 exceed 64k. The default is to use the @code{basr} instruction instead,
20814 which does not have this limitation.
20820 When @option{-m31} is specified, generate code compliant to the
20821 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
20822 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
20823 particular to generate 64-bit instructions. For the @samp{s390}
20824 targets, the default is @option{-m31}, while the @samp{s390x}
20825 targets default to @option{-m64}.
20831 When @option{-mzarch} is specified, generate code using the
20832 instructions available on z/Architecture.
20833 When @option{-mesa} is specified, generate code using the
20834 instructions available on ESA/390. Note that @option{-mesa} is
20835 not possible with @option{-m64}.
20836 When generating code compliant to the GNU/Linux for S/390 ABI,
20837 the default is @option{-mesa}. When generating code compliant
20838 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
20844 Generate (or do not generate) code using the @code{mvcle} instruction
20845 to perform block moves. When @option{-mno-mvcle} is specified,
20846 use a @code{mvc} loop instead. This is the default unless optimizing for
20853 Print (or do not print) additional debug information when compiling.
20854 The default is to not print debug information.
20856 @item -march=@var{cpu-type}
20858 Generate code that runs on @var{cpu-type}, which is the name of a system
20859 representing a certain processor type. Possible values for
20860 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
20861 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
20862 When generating code using the instructions available on z/Architecture,
20863 the default is @option{-march=z900}. Otherwise, the default is
20864 @option{-march=g5}.
20866 @item -mtune=@var{cpu-type}
20868 Tune to @var{cpu-type} everything applicable about the generated code,
20869 except for the ABI and the set of available instructions.
20870 The list of @var{cpu-type} values is the same as for @option{-march}.
20871 The default is the value used for @option{-march}.
20874 @itemx -mno-tpf-trace
20875 @opindex mtpf-trace
20876 @opindex mno-tpf-trace
20877 Generate code that adds (does not add) in TPF OS specific branches to trace
20878 routines in the operating system. This option is off by default, even
20879 when compiling for the TPF OS@.
20882 @itemx -mno-fused-madd
20883 @opindex mfused-madd
20884 @opindex mno-fused-madd
20885 Generate code that uses (does not use) the floating-point multiply and
20886 accumulate instructions. These instructions are generated by default if
20887 hardware floating point is used.
20889 @item -mwarn-framesize=@var{framesize}
20890 @opindex mwarn-framesize
20891 Emit a warning if the current function exceeds the given frame size. Because
20892 this is a compile-time check it doesn't need to be a real problem when the program
20893 runs. It is intended to identify functions that most probably cause
20894 a stack overflow. It is useful to be used in an environment with limited stack
20895 size e.g.@: the linux kernel.
20897 @item -mwarn-dynamicstack
20898 @opindex mwarn-dynamicstack
20899 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
20900 arrays. This is generally a bad idea with a limited stack size.
20902 @item -mstack-guard=@var{stack-guard}
20903 @itemx -mstack-size=@var{stack-size}
20904 @opindex mstack-guard
20905 @opindex mstack-size
20906 If these options are provided the S/390 back end emits additional instructions in
20907 the function prologue that trigger a trap if the stack size is @var{stack-guard}
20908 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
20909 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
20910 the frame size of the compiled function is chosen.
20911 These options are intended to be used to help debugging stack overflow problems.
20912 The additionally emitted code causes only little overhead and hence can also be
20913 used in production-like systems without greater performance degradation. The given
20914 values have to be exact powers of 2 and @var{stack-size} has to be greater than
20915 @var{stack-guard} without exceeding 64k.
20916 In order to be efficient the extra code makes the assumption that the stack starts
20917 at an address aligned to the value given by @var{stack-size}.
20918 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
20920 @item -mhotpatch[=@var{halfwords}]
20921 @itemx -mno-hotpatch
20923 If the hotpatch option is enabled, a ``hot-patching'' function
20924 prologue is generated for all functions in the compilation unit.
20925 The funtion label is prepended with the given number of two-byte
20926 Nop instructions (@var{halfwords}, maximum 1000000) or 12 Nop
20927 instructions if no argument is present. Functions with a
20928 hot-patching prologue are never inlined automatically, and a
20929 hot-patching prologue is never generated for functions
20930 that are explicitly inline.
20932 This option can be overridden for individual functions with the
20933 @code{hotpatch} attribute.
20936 @node Score Options
20937 @subsection Score Options
20938 @cindex Score Options
20940 These options are defined for Score implementations:
20945 Compile code for big-endian mode. This is the default.
20949 Compile code for little-endian mode.
20953 Disable generation of @code{bcnz} instructions.
20957 Enable generation of unaligned load and store instructions.
20961 Enable the use of multiply-accumulate instructions. Disabled by default.
20965 Specify the SCORE5 as the target architecture.
20969 Specify the SCORE5U of the target architecture.
20973 Specify the SCORE7 as the target architecture. This is the default.
20977 Specify the SCORE7D as the target architecture.
20981 @subsection SH Options
20983 These @samp{-m} options are defined for the SH implementations:
20988 Generate code for the SH1.
20992 Generate code for the SH2.
20995 Generate code for the SH2e.
20999 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
21000 that the floating-point unit is not used.
21002 @item -m2a-single-only
21003 @opindex m2a-single-only
21004 Generate code for the SH2a-FPU, in such a way that no double-precision
21005 floating-point operations are used.
21008 @opindex m2a-single
21009 Generate code for the SH2a-FPU assuming the floating-point unit is in
21010 single-precision mode by default.
21014 Generate code for the SH2a-FPU assuming the floating-point unit is in
21015 double-precision mode by default.
21019 Generate code for the SH3.
21023 Generate code for the SH3e.
21027 Generate code for the SH4 without a floating-point unit.
21029 @item -m4-single-only
21030 @opindex m4-single-only
21031 Generate code for the SH4 with a floating-point unit that only
21032 supports single-precision arithmetic.
21036 Generate code for the SH4 assuming the floating-point unit is in
21037 single-precision mode by default.
21041 Generate code for the SH4.
21045 Generate code for SH4-100.
21047 @item -m4-100-nofpu
21048 @opindex m4-100-nofpu
21049 Generate code for SH4-100 in such a way that the
21050 floating-point unit is not used.
21052 @item -m4-100-single
21053 @opindex m4-100-single
21054 Generate code for SH4-100 assuming the floating-point unit is in
21055 single-precision mode by default.
21057 @item -m4-100-single-only
21058 @opindex m4-100-single-only
21059 Generate code for SH4-100 in such a way that no double-precision
21060 floating-point operations are used.
21064 Generate code for SH4-200.
21066 @item -m4-200-nofpu
21067 @opindex m4-200-nofpu
21068 Generate code for SH4-200 without in such a way that the
21069 floating-point unit is not used.
21071 @item -m4-200-single
21072 @opindex m4-200-single
21073 Generate code for SH4-200 assuming the floating-point unit is in
21074 single-precision mode by default.
21076 @item -m4-200-single-only
21077 @opindex m4-200-single-only
21078 Generate code for SH4-200 in such a way that no double-precision
21079 floating-point operations are used.
21083 Generate code for SH4-300.
21085 @item -m4-300-nofpu
21086 @opindex m4-300-nofpu
21087 Generate code for SH4-300 without in such a way that the
21088 floating-point unit is not used.
21090 @item -m4-300-single
21091 @opindex m4-300-single
21092 Generate code for SH4-300 in such a way that no double-precision
21093 floating-point operations are used.
21095 @item -m4-300-single-only
21096 @opindex m4-300-single-only
21097 Generate code for SH4-300 in such a way that no double-precision
21098 floating-point operations are used.
21102 Generate code for SH4-340 (no MMU, no FPU).
21106 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
21111 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
21112 floating-point unit is not used.
21114 @item -m4a-single-only
21115 @opindex m4a-single-only
21116 Generate code for the SH4a, in such a way that no double-precision
21117 floating-point operations are used.
21120 @opindex m4a-single
21121 Generate code for the SH4a assuming the floating-point unit is in
21122 single-precision mode by default.
21126 Generate code for the SH4a.
21130 Same as @option{-m4a-nofpu}, except that it implicitly passes
21131 @option{-dsp} to the assembler. GCC doesn't generate any DSP
21132 instructions at the moment.
21135 @opindex m5-32media
21136 Generate 32-bit code for SHmedia.
21138 @item -m5-32media-nofpu
21139 @opindex m5-32media-nofpu
21140 Generate 32-bit code for SHmedia in such a way that the
21141 floating-point unit is not used.
21144 @opindex m5-64media
21145 Generate 64-bit code for SHmedia.
21147 @item -m5-64media-nofpu
21148 @opindex m5-64media-nofpu
21149 Generate 64-bit code for SHmedia in such a way that the
21150 floating-point unit is not used.
21153 @opindex m5-compact
21154 Generate code for SHcompact.
21156 @item -m5-compact-nofpu
21157 @opindex m5-compact-nofpu
21158 Generate code for SHcompact in such a way that the
21159 floating-point unit is not used.
21163 Compile code for the processor in big-endian mode.
21167 Compile code for the processor in little-endian mode.
21171 Align doubles at 64-bit boundaries. Note that this changes the calling
21172 conventions, and thus some functions from the standard C library do
21173 not work unless you recompile it first with @option{-mdalign}.
21177 Shorten some address references at link time, when possible; uses the
21178 linker option @option{-relax}.
21182 Use 32-bit offsets in @code{switch} tables. The default is to use
21187 Enable the use of bit manipulation instructions on SH2A.
21191 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
21192 alignment constraints.
21196 Comply with the calling conventions defined by Renesas.
21199 @opindex mno-renesas
21200 Comply with the calling conventions defined for GCC before the Renesas
21201 conventions were available. This option is the default for all
21202 targets of the SH toolchain.
21205 @opindex mnomacsave
21206 Mark the @code{MAC} register as call-clobbered, even if
21207 @option{-mrenesas} is given.
21213 Control the IEEE compliance of floating-point comparisons, which affects the
21214 handling of cases where the result of a comparison is unordered. By default
21215 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
21216 enabled @option{-mno-ieee} is implicitly set, which results in faster
21217 floating-point greater-equal and less-equal comparisons. The implcit settings
21218 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
21220 @item -minline-ic_invalidate
21221 @opindex minline-ic_invalidate
21222 Inline code to invalidate instruction cache entries after setting up
21223 nested function trampolines.
21224 This option has no effect if @option{-musermode} is in effect and the selected
21225 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
21227 If the selected code generation option does not allow the use of the @code{icbi}
21228 instruction, and @option{-musermode} is not in effect, the inlined code
21229 manipulates the instruction cache address array directly with an associative
21230 write. This not only requires privileged mode at run time, but it also
21231 fails if the cache line had been mapped via the TLB and has become unmapped.
21235 Dump instruction size and location in the assembly code.
21238 @opindex mpadstruct
21239 This option is deprecated. It pads structures to multiple of 4 bytes,
21240 which is incompatible with the SH ABI@.
21242 @item -matomic-model=@var{model}
21243 @opindex matomic-model=@var{model}
21244 Sets the model of atomic operations and additional parameters as a comma
21245 separated list. For details on the atomic built-in functions see
21246 @ref{__atomic Builtins}. The following models and parameters are supported:
21251 Disable compiler generated atomic sequences and emit library calls for atomic
21252 operations. This is the default if the target is not @code{sh*-*-linux*}.
21255 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
21256 built-in functions. The generated atomic sequences require additional support
21257 from the interrupt/exception handling code of the system and are only suitable
21258 for SH3* and SH4* single-core systems. This option is enabled by default when
21259 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
21260 this option will also partially utilize the hardware atomic instructions
21261 @code{movli.l} and @code{movco.l} to create more efficient code, unless
21262 @samp{strict} is specified.
21265 Generate software atomic sequences that use a variable in the thread control
21266 block. This is a variation of the gUSA sequences which can also be used on
21267 SH1* and SH2* targets. The generated atomic sequences require additional
21268 support from the interrupt/exception handling code of the system and are only
21269 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
21270 parameter has to be specified as well.
21273 Generate software atomic sequences that temporarily disable interrupts by
21274 setting @code{SR.IMASK = 1111}. This model works only when the program runs
21275 in privileged mode and is only suitable for single-core systems. Additional
21276 support from the interrupt/exception handling code of the system is not
21277 required. This model is enabled by default when the target is
21278 @code{sh*-*-linux*} and SH1* or SH2*.
21281 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
21282 instructions only. This is only available on SH4A and is suitable for
21283 multi-core systems. Since the hardware instructions support only 32 bit atomic
21284 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
21285 Code compiled with this option will also be compatible with other software
21286 atomic model interrupt/exception handling systems if executed on an SH4A
21287 system. Additional support from the interrupt/exception handling code of the
21288 system is not required for this model.
21291 This parameter specifies the offset in bytes of the variable in the thread
21292 control block structure that should be used by the generated atomic sequences
21293 when the @samp{soft-tcb} model has been selected. For other models this
21294 parameter is ignored. The specified value must be an integer multiple of four
21295 and in the range 0-1020.
21298 This parameter prevents mixed usage of multiple atomic models, even though they
21299 would be compatible, and will make the compiler generate atomic sequences of the
21300 specified model only.
21306 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
21307 Notice that depending on the particular hardware and software configuration
21308 this can degrade overall performance due to the operand cache line flushes
21309 that are implied by the @code{tas.b} instruction. On multi-core SH4A
21310 processors the @code{tas.b} instruction must be used with caution since it
21311 can result in data corruption for certain cache configurations.
21314 @opindex mprefergot
21315 When generating position-independent code, emit function calls using
21316 the Global Offset Table instead of the Procedure Linkage Table.
21319 @itemx -mno-usermode
21321 @opindex mno-usermode
21322 Don't allow (allow) the compiler generating privileged mode code. Specifying
21323 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
21324 inlined code would not work in user mode. @option{-musermode} is the default
21325 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
21326 @option{-musermode} has no effect, since there is no user mode.
21328 @item -multcost=@var{number}
21329 @opindex multcost=@var{number}
21330 Set the cost to assume for a multiply insn.
21332 @item -mdiv=@var{strategy}
21333 @opindex mdiv=@var{strategy}
21334 Set the division strategy to be used for integer division operations.
21335 For SHmedia @var{strategy} can be one of:
21340 Performs the operation in floating point. This has a very high latency,
21341 but needs only a few instructions, so it might be a good choice if
21342 your code has enough easily-exploitable ILP to allow the compiler to
21343 schedule the floating-point instructions together with other instructions.
21344 Division by zero causes a floating-point exception.
21347 Uses integer operations to calculate the inverse of the divisor,
21348 and then multiplies the dividend with the inverse. This strategy allows
21349 CSE and hoisting of the inverse calculation. Division by zero calculates
21350 an unspecified result, but does not trap.
21353 A variant of @samp{inv} where, if no CSE or hoisting opportunities
21354 have been found, or if the entire operation has been hoisted to the same
21355 place, the last stages of the inverse calculation are intertwined with the
21356 final multiply to reduce the overall latency, at the expense of using a few
21357 more instructions, and thus offering fewer scheduling opportunities with
21361 Calls a library function that usually implements the @samp{inv:minlat}
21363 This gives high code density for @code{m5-*media-nofpu} compilations.
21366 Uses a different entry point of the same library function, where it
21367 assumes that a pointer to a lookup table has already been set up, which
21368 exposes the pointer load to CSE and code hoisting optimizations.
21373 Use the @samp{inv} algorithm for initial
21374 code generation, but if the code stays unoptimized, revert to the @samp{call},
21375 @samp{call2}, or @samp{fp} strategies, respectively. Note that the
21376 potentially-trapping side effect of division by zero is carried by a
21377 separate instruction, so it is possible that all the integer instructions
21378 are hoisted out, but the marker for the side effect stays where it is.
21379 A recombination to floating-point operations or a call is not possible
21384 Variants of the @samp{inv:minlat} strategy. In the case
21385 that the inverse calculation is not separated from the multiply, they speed
21386 up division where the dividend fits into 20 bits (plus sign where applicable)
21387 by inserting a test to skip a number of operations in this case; this test
21388 slows down the case of larger dividends. @samp{inv20u} assumes the case of a such
21389 a small dividend to be unlikely, and @samp{inv20l} assumes it to be likely.
21393 For targets other than SHmedia @var{strategy} can be one of:
21398 Calls a library function that uses the single-step division instruction
21399 @code{div1} to perform the operation. Division by zero calculates an
21400 unspecified result and does not trap. This is the default except for SH4,
21401 SH2A and SHcompact.
21404 Calls a library function that performs the operation in double precision
21405 floating point. Division by zero causes a floating-point exception. This is
21406 the default for SHcompact with FPU. Specifying this for targets that do not
21407 have a double precision FPU will default to @code{call-div1}.
21410 Calls a library function that uses a lookup table for small divisors and
21411 the @code{div1} instruction with case distinction for larger divisors. Division
21412 by zero calculates an unspecified result and does not trap. This is the default
21413 for SH4. Specifying this for targets that do not have dynamic shift
21414 instructions will default to @code{call-div1}.
21418 When a division strategy has not been specified the default strategy will be
21419 selected based on the current target. For SH2A the default strategy is to
21420 use the @code{divs} and @code{divu} instructions instead of library function
21423 @item -maccumulate-outgoing-args
21424 @opindex maccumulate-outgoing-args
21425 Reserve space once for outgoing arguments in the function prologue rather
21426 than around each call. Generally beneficial for performance and size. Also
21427 needed for unwinding to avoid changing the stack frame around conditional code.
21429 @item -mdivsi3_libfunc=@var{name}
21430 @opindex mdivsi3_libfunc=@var{name}
21431 Set the name of the library function used for 32-bit signed division to
21433 This only affects the name used in the @samp{call} and @samp{inv:call}
21434 division strategies, and the compiler still expects the same
21435 sets of input/output/clobbered registers as if this option were not present.
21437 @item -mfixed-range=@var{register-range}
21438 @opindex mfixed-range
21439 Generate code treating the given register range as fixed registers.
21440 A fixed register is one that the register allocator can not use. This is
21441 useful when compiling kernel code. A register range is specified as
21442 two registers separated by a dash. Multiple register ranges can be
21443 specified separated by a comma.
21445 @item -mindexed-addressing
21446 @opindex mindexed-addressing
21447 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
21448 This is only safe if the hardware and/or OS implement 32-bit wrap-around
21449 semantics for the indexed addressing mode. The architecture allows the
21450 implementation of processors with 64-bit MMU, which the OS could use to
21451 get 32-bit addressing, but since no current hardware implementation supports
21452 this or any other way to make the indexed addressing mode safe to use in
21453 the 32-bit ABI, the default is @option{-mno-indexed-addressing}.
21455 @item -mgettrcost=@var{number}
21456 @opindex mgettrcost=@var{number}
21457 Set the cost assumed for the @code{gettr} instruction to @var{number}.
21458 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
21462 Assume @code{pt*} instructions won't trap. This generally generates
21463 better-scheduled code, but is unsafe on current hardware.
21464 The current architecture
21465 definition says that @code{ptabs} and @code{ptrel} trap when the target
21467 This has the unintentional effect of making it unsafe to schedule these
21468 instructions before a branch, or hoist them out of a loop. For example,
21469 @code{__do_global_ctors}, a part of @file{libgcc}
21470 that runs constructors at program
21471 startup, calls functions in a list which is delimited by @minus{}1. With the
21472 @option{-mpt-fixed} option, the @code{ptabs} is done before testing against @minus{}1.
21473 That means that all the constructors run a bit more quickly, but when
21474 the loop comes to the end of the list, the program crashes because @code{ptabs}
21475 loads @minus{}1 into a target register.
21477 Since this option is unsafe for any
21478 hardware implementing the current architecture specification, the default
21479 is @option{-mno-pt-fixed}. Unless specified explicitly with
21480 @option{-mgettrcost}, @option{-mno-pt-fixed} also implies @option{-mgettrcost=100};
21481 this deters register allocation from using target registers for storing
21484 @item -minvalid-symbols
21485 @opindex minvalid-symbols
21486 Assume symbols might be invalid. Ordinary function symbols generated by
21487 the compiler are always valid to load with
21488 @code{movi}/@code{shori}/@code{ptabs} or
21489 @code{movi}/@code{shori}/@code{ptrel},
21490 but with assembler and/or linker tricks it is possible
21491 to generate symbols that cause @code{ptabs} or @code{ptrel} to trap.
21492 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
21493 It prevents cross-basic-block CSE, hoisting and most scheduling
21494 of symbol loads. The default is @option{-mno-invalid-symbols}.
21496 @item -mbranch-cost=@var{num}
21497 @opindex mbranch-cost=@var{num}
21498 Assume @var{num} to be the cost for a branch instruction. Higher numbers
21499 make the compiler try to generate more branch-free code if possible.
21500 If not specified the value is selected depending on the processor type that
21501 is being compiled for.
21504 @itemx -mno-zdcbranch
21505 @opindex mzdcbranch
21506 @opindex mno-zdcbranch
21507 Assume (do not assume) that zero displacement conditional branch instructions
21508 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
21509 compiler will try to prefer zero displacement branch code sequences. This is
21510 enabled by default when generating code for SH4 and SH4A. It can be explicitly
21511 disabled by specifying @option{-mno-zdcbranch}.
21514 @itemx -mno-fused-madd
21515 @opindex mfused-madd
21516 @opindex mno-fused-madd
21517 Generate code that uses (does not use) the floating-point multiply and
21518 accumulate instructions. These instructions are generated by default
21519 if hardware floating point is used. The machine-dependent
21520 @option{-mfused-madd} option is now mapped to the machine-independent
21521 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
21522 mapped to @option{-ffp-contract=off}.
21528 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
21529 and cosine approximations. The option @code{-mfsca} must be used in
21530 combination with @code{-funsafe-math-optimizations}. It is enabled by default
21531 when generating code for SH4A. Using @code{-mno-fsca} disables sine and cosine
21532 approximations even if @code{-funsafe-math-optimizations} is in effect.
21538 Allow or disallow the compiler to emit the @code{fsrra} instruction for
21539 reciprocal square root approximations. The option @code{-mfsrra} must be used
21540 in combination with @code{-funsafe-math-optimizations} and
21541 @code{-ffinite-math-only}. It is enabled by default when generating code for
21542 SH4A. Using @code{-mno-fsrra} disables reciprocal square root approximations
21543 even if @code{-funsafe-math-optimizations} and @code{-ffinite-math-only} are
21546 @item -mpretend-cmove
21547 @opindex mpretend-cmove
21548 Prefer zero-displacement conditional branches for conditional move instruction
21549 patterns. This can result in faster code on the SH4 processor.
21553 @node Solaris 2 Options
21554 @subsection Solaris 2 Options
21555 @cindex Solaris 2 options
21557 These @samp{-m} options are supported on Solaris 2:
21560 @item -mclear-hwcap
21561 @opindex mclear-hwcap
21562 @option{-mclear-hwcap} tells the compiler to remove the hardware
21563 capabilities generated by the Solaris assembler. This is only necessary
21564 when object files use ISA extensions not supported by the current
21565 machine, but check at runtime whether or not to use them.
21567 @item -mimpure-text
21568 @opindex mimpure-text
21569 @option{-mimpure-text}, used in addition to @option{-shared}, tells
21570 the compiler to not pass @option{-z text} to the linker when linking a
21571 shared object. Using this option, you can link position-dependent
21572 code into a shared object.
21574 @option{-mimpure-text} suppresses the ``relocations remain against
21575 allocatable but non-writable sections'' linker error message.
21576 However, the necessary relocations trigger copy-on-write, and the
21577 shared object is not actually shared across processes. Instead of
21578 using @option{-mimpure-text}, you should compile all source code with
21579 @option{-fpic} or @option{-fPIC}.
21583 These switches are supported in addition to the above on Solaris 2:
21588 Add support for multithreading using the POSIX threads library. This
21589 option sets flags for both the preprocessor and linker. This option does
21590 not affect the thread safety of object code produced by the compiler or
21591 that of libraries supplied with it.
21595 This is a synonym for @option{-pthreads}.
21598 @node SPARC Options
21599 @subsection SPARC Options
21600 @cindex SPARC options
21602 These @samp{-m} options are supported on the SPARC:
21605 @item -mno-app-regs
21607 @opindex mno-app-regs
21609 Specify @option{-mapp-regs} to generate output using the global registers
21610 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
21611 global register 1, each global register 2 through 4 is then treated as an
21612 allocable register that is clobbered by function calls. This is the default.
21614 To be fully SVR4 ABI-compliant at the cost of some performance loss,
21615 specify @option{-mno-app-regs}. You should compile libraries and system
21616 software with this option.
21622 With @option{-mflat}, the compiler does not generate save/restore instructions
21623 and uses a ``flat'' or single register window model. This model is compatible
21624 with the regular register window model. The local registers and the input
21625 registers (0--5) are still treated as ``call-saved'' registers and are
21626 saved on the stack as needed.
21628 With @option{-mno-flat} (the default), the compiler generates save/restore
21629 instructions (except for leaf functions). This is the normal operating mode.
21632 @itemx -mhard-float
21634 @opindex mhard-float
21635 Generate output containing floating-point instructions. This is the
21639 @itemx -msoft-float
21641 @opindex msoft-float
21642 Generate output containing library calls for floating point.
21643 @strong{Warning:} the requisite libraries are not available for all SPARC
21644 targets. Normally the facilities of the machine's usual C compiler are
21645 used, but this cannot be done directly in cross-compilation. You must make
21646 your own arrangements to provide suitable library functions for
21647 cross-compilation. The embedded targets @samp{sparc-*-aout} and
21648 @samp{sparclite-*-*} do provide software floating-point support.
21650 @option{-msoft-float} changes the calling convention in the output file;
21651 therefore, it is only useful if you compile @emph{all} of a program with
21652 this option. In particular, you need to compile @file{libgcc.a}, the
21653 library that comes with GCC, with @option{-msoft-float} in order for
21656 @item -mhard-quad-float
21657 @opindex mhard-quad-float
21658 Generate output containing quad-word (long double) floating-point
21661 @item -msoft-quad-float
21662 @opindex msoft-quad-float
21663 Generate output containing library calls for quad-word (long double)
21664 floating-point instructions. The functions called are those specified
21665 in the SPARC ABI@. This is the default.
21667 As of this writing, there are no SPARC implementations that have hardware
21668 support for the quad-word floating-point instructions. They all invoke
21669 a trap handler for one of these instructions, and then the trap handler
21670 emulates the effect of the instruction. Because of the trap handler overhead,
21671 this is much slower than calling the ABI library routines. Thus the
21672 @option{-msoft-quad-float} option is the default.
21674 @item -mno-unaligned-doubles
21675 @itemx -munaligned-doubles
21676 @opindex mno-unaligned-doubles
21677 @opindex munaligned-doubles
21678 Assume that doubles have 8-byte alignment. This is the default.
21680 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
21681 alignment only if they are contained in another type, or if they have an
21682 absolute address. Otherwise, it assumes they have 4-byte alignment.
21683 Specifying this option avoids some rare compatibility problems with code
21684 generated by other compilers. It is not the default because it results
21685 in a performance loss, especially for floating-point code.
21688 @itemx -mno-user-mode
21689 @opindex muser-mode
21690 @opindex mno-user-mode
21691 Do not generate code that can only run in supervisor mode. This is relevant
21692 only for the @code{casa} instruction emitted for the LEON3 processor. The
21693 default is @option{-mno-user-mode}.
21695 @item -mno-faster-structs
21696 @itemx -mfaster-structs
21697 @opindex mno-faster-structs
21698 @opindex mfaster-structs
21699 With @option{-mfaster-structs}, the compiler assumes that structures
21700 should have 8-byte alignment. This enables the use of pairs of
21701 @code{ldd} and @code{std} instructions for copies in structure
21702 assignment, in place of twice as many @code{ld} and @code{st} pairs.
21703 However, the use of this changed alignment directly violates the SPARC
21704 ABI@. Thus, it's intended only for use on targets where the developer
21705 acknowledges that their resulting code is not directly in line with
21706 the rules of the ABI@.
21708 @item -mcpu=@var{cpu_type}
21710 Set the instruction set, register set, and instruction scheduling parameters
21711 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
21712 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
21713 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
21714 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
21715 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21716 @samp{niagara3} and @samp{niagara4}.
21718 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
21719 which selects the best architecture option for the host processor.
21720 @option{-mcpu=native} has no effect if GCC does not recognize
21723 Default instruction scheduling parameters are used for values that select
21724 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
21725 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
21727 Here is a list of each supported architecture and their supported
21735 supersparc, hypersparc, leon, leon3
21738 f930, f934, sparclite86x
21744 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
21747 By default (unless configured otherwise), GCC generates code for the V7
21748 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
21749 additionally optimizes it for the Cypress CY7C602 chip, as used in the
21750 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
21751 SPARCStation 1, 2, IPX etc.
21753 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
21754 architecture. The only difference from V7 code is that the compiler emits
21755 the integer multiply and integer divide instructions which exist in SPARC-V8
21756 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
21757 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
21760 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
21761 the SPARC architecture. This adds the integer multiply, integer divide step
21762 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
21763 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
21764 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
21765 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
21766 MB86934 chip, which is the more recent SPARClite with FPU@.
21768 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
21769 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
21770 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
21771 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
21772 optimizes it for the TEMIC SPARClet chip.
21774 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
21775 architecture. This adds 64-bit integer and floating-point move instructions,
21776 3 additional floating-point condition code registers and conditional move
21777 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
21778 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
21779 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
21780 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
21781 @option{-mcpu=niagara}, the compiler additionally optimizes it for
21782 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
21783 additionally optimizes it for Sun UltraSPARC T2 chips. With
21784 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
21785 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
21786 additionally optimizes it for Sun UltraSPARC T4 chips.
21788 @item -mtune=@var{cpu_type}
21790 Set the instruction scheduling parameters for machine type
21791 @var{cpu_type}, but do not set the instruction set or register set that the
21792 option @option{-mcpu=@var{cpu_type}} does.
21794 The same values for @option{-mcpu=@var{cpu_type}} can be used for
21795 @option{-mtune=@var{cpu_type}}, but the only useful values are those
21796 that select a particular CPU implementation. Those are @samp{cypress},
21797 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{leon3},
21798 @samp{leon3v7}, @samp{f930}, @samp{f934}, @samp{sparclite86x}, @samp{tsc701},
21799 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21800 @samp{niagara3} and @samp{niagara4}. With native Solaris and GNU/Linux
21801 toolchains, @samp{native} can also be used.
21806 @opindex mno-v8plus
21807 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
21808 difference from the V8 ABI is that the global and out registers are
21809 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
21810 mode for all SPARC-V9 processors.
21816 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
21817 Visual Instruction Set extensions. The default is @option{-mno-vis}.
21823 With @option{-mvis2}, GCC generates code that takes advantage of
21824 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
21825 default is @option{-mvis2} when targeting a cpu that supports such
21826 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
21827 also sets @option{-mvis}.
21833 With @option{-mvis3}, GCC generates code that takes advantage of
21834 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
21835 default is @option{-mvis3} when targeting a cpu that supports such
21836 instructions, such as niagara-3 and later. Setting @option{-mvis3}
21837 also sets @option{-mvis2} and @option{-mvis}.
21842 @opindex mno-cbcond
21843 With @option{-mcbcond}, GCC generates code that takes advantage of
21844 compare-and-branch instructions, as defined in the Sparc Architecture 2011.
21845 The default is @option{-mcbcond} when targeting a cpu that supports such
21846 instructions, such as niagara-4 and later.
21852 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
21853 population count instruction. The default is @option{-mpopc}
21854 when targeting a cpu that supports such instructions, such as Niagara-2 and
21861 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
21862 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
21863 when targeting a cpu that supports such instructions, such as Niagara-3 and
21867 @opindex mfix-at697f
21868 Enable the documented workaround for the single erratum of the Atmel AT697F
21869 processor (which corresponds to erratum #13 of the AT697E processor).
21872 @opindex mfix-ut699
21873 Enable the documented workarounds for the floating-point errata and the data
21874 cache nullify errata of the UT699 processor.
21877 These @samp{-m} options are supported in addition to the above
21878 on SPARC-V9 processors in 64-bit environments:
21885 Generate code for a 32-bit or 64-bit environment.
21886 The 32-bit environment sets int, long and pointer to 32 bits.
21887 The 64-bit environment sets int to 32 bits and long and pointer
21890 @item -mcmodel=@var{which}
21892 Set the code model to one of
21896 The Medium/Low code model: 64-bit addresses, programs
21897 must be linked in the low 32 bits of memory. Programs can be statically
21898 or dynamically linked.
21901 The Medium/Middle code model: 64-bit addresses, programs
21902 must be linked in the low 44 bits of memory, the text and data segments must
21903 be less than 2GB in size and the data segment must be located within 2GB of
21907 The Medium/Anywhere code model: 64-bit addresses, programs
21908 may be linked anywhere in memory, the text and data segments must be less
21909 than 2GB in size and the data segment must be located within 2GB of the
21913 The Medium/Anywhere code model for embedded systems:
21914 64-bit addresses, the text and data segments must be less than 2GB in
21915 size, both starting anywhere in memory (determined at link time). The
21916 global register %g4 points to the base of the data segment. Programs
21917 are statically linked and PIC is not supported.
21920 @item -mmemory-model=@var{mem-model}
21921 @opindex mmemory-model
21922 Set the memory model in force on the processor to one of
21926 The default memory model for the processor and operating system.
21929 Relaxed Memory Order
21932 Partial Store Order
21938 Sequential Consistency
21941 These memory models are formally defined in Appendix D of the Sparc V9
21942 architecture manual, as set in the processor's @code{PSTATE.MM} field.
21945 @itemx -mno-stack-bias
21946 @opindex mstack-bias
21947 @opindex mno-stack-bias
21948 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
21949 frame pointer if present, are offset by @minus{}2047 which must be added back
21950 when making stack frame references. This is the default in 64-bit mode.
21951 Otherwise, assume no such offset is present.
21955 @subsection SPU Options
21956 @cindex SPU options
21958 These @samp{-m} options are supported on the SPU:
21962 @itemx -merror-reloc
21963 @opindex mwarn-reloc
21964 @opindex merror-reloc
21966 The loader for SPU does not handle dynamic relocations. By default, GCC
21967 gives an error when it generates code that requires a dynamic
21968 relocation. @option{-mno-error-reloc} disables the error,
21969 @option{-mwarn-reloc} generates a warning instead.
21972 @itemx -munsafe-dma
21974 @opindex munsafe-dma
21976 Instructions that initiate or test completion of DMA must not be
21977 reordered with respect to loads and stores of the memory that is being
21979 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
21980 memory accesses, but that can lead to inefficient code in places where the
21981 memory is known to not change. Rather than mark the memory as volatile,
21982 you can use @option{-msafe-dma} to tell the compiler to treat
21983 the DMA instructions as potentially affecting all memory.
21985 @item -mbranch-hints
21986 @opindex mbranch-hints
21988 By default, GCC generates a branch hint instruction to avoid
21989 pipeline stalls for always-taken or probably-taken branches. A hint
21990 is not generated closer than 8 instructions away from its branch.
21991 There is little reason to disable them, except for debugging purposes,
21992 or to make an object a little bit smaller.
21996 @opindex msmall-mem
21997 @opindex mlarge-mem
21999 By default, GCC generates code assuming that addresses are never larger
22000 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
22001 a full 32-bit address.
22006 By default, GCC links against startup code that assumes the SPU-style
22007 main function interface (which has an unconventional parameter list).
22008 With @option{-mstdmain}, GCC links your program against startup
22009 code that assumes a C99-style interface to @code{main}, including a
22010 local copy of @code{argv} strings.
22012 @item -mfixed-range=@var{register-range}
22013 @opindex mfixed-range
22014 Generate code treating the given register range as fixed registers.
22015 A fixed register is one that the register allocator cannot use. This is
22016 useful when compiling kernel code. A register range is specified as
22017 two registers separated by a dash. Multiple register ranges can be
22018 specified separated by a comma.
22024 Compile code assuming that pointers to the PPU address space accessed
22025 via the @code{__ea} named address space qualifier are either 32 or 64
22026 bits wide. The default is 32 bits. As this is an ABI-changing option,
22027 all object code in an executable must be compiled with the same setting.
22029 @item -maddress-space-conversion
22030 @itemx -mno-address-space-conversion
22031 @opindex maddress-space-conversion
22032 @opindex mno-address-space-conversion
22033 Allow/disallow treating the @code{__ea} address space as superset
22034 of the generic address space. This enables explicit type casts
22035 between @code{__ea} and generic pointer as well as implicit
22036 conversions of generic pointers to @code{__ea} pointers. The
22037 default is to allow address space pointer conversions.
22039 @item -mcache-size=@var{cache-size}
22040 @opindex mcache-size
22041 This option controls the version of libgcc that the compiler links to an
22042 executable and selects a software-managed cache for accessing variables
22043 in the @code{__ea} address space with a particular cache size. Possible
22044 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
22045 and @samp{128}. The default cache size is 64KB.
22047 @item -matomic-updates
22048 @itemx -mno-atomic-updates
22049 @opindex matomic-updates
22050 @opindex mno-atomic-updates
22051 This option controls the version of libgcc that the compiler links to an
22052 executable and selects whether atomic updates to the software-managed
22053 cache of PPU-side variables are used. If you use atomic updates, changes
22054 to a PPU variable from SPU code using the @code{__ea} named address space
22055 qualifier do not interfere with changes to other PPU variables residing
22056 in the same cache line from PPU code. If you do not use atomic updates,
22057 such interference may occur; however, writing back cache lines is
22058 more efficient. The default behavior is to use atomic updates.
22061 @itemx -mdual-nops=@var{n}
22062 @opindex mdual-nops
22063 By default, GCC inserts nops to increase dual issue when it expects
22064 it to increase performance. @var{n} can be a value from 0 to 10. A
22065 smaller @var{n} inserts fewer nops. 10 is the default, 0 is the
22066 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
22068 @item -mhint-max-nops=@var{n}
22069 @opindex mhint-max-nops
22070 Maximum number of nops to insert for a branch hint. A branch hint must
22071 be at least 8 instructions away from the branch it is affecting. GCC
22072 inserts up to @var{n} nops to enforce this, otherwise it does not
22073 generate the branch hint.
22075 @item -mhint-max-distance=@var{n}
22076 @opindex mhint-max-distance
22077 The encoding of the branch hint instruction limits the hint to be within
22078 256 instructions of the branch it is affecting. By default, GCC makes
22079 sure it is within 125.
22082 @opindex msafe-hints
22083 Work around a hardware bug that causes the SPU to stall indefinitely.
22084 By default, GCC inserts the @code{hbrp} instruction to make sure
22085 this stall won't happen.
22089 @node System V Options
22090 @subsection Options for System V
22092 These additional options are available on System V Release 4 for
22093 compatibility with other compilers on those systems:
22098 Create a shared object.
22099 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
22103 Identify the versions of each tool used by the compiler, in a
22104 @code{.ident} assembler directive in the output.
22108 Refrain from adding @code{.ident} directives to the output file (this is
22111 @item -YP,@var{dirs}
22113 Search the directories @var{dirs}, and no others, for libraries
22114 specified with @option{-l}.
22116 @item -Ym,@var{dir}
22118 Look in the directory @var{dir} to find the M4 preprocessor.
22119 The assembler uses this option.
22120 @c This is supposed to go with a -Yd for predefined M4 macro files, but
22121 @c the generic assembler that comes with Solaris takes just -Ym.
22124 @node TILE-Gx Options
22125 @subsection TILE-Gx Options
22126 @cindex TILE-Gx options
22128 These @samp{-m} options are supported on the TILE-Gx:
22131 @item -mcmodel=small
22132 @opindex mcmodel=small
22133 Generate code for the small model. The distance for direct calls is
22134 limited to 500M in either direction. PC-relative addresses are 32
22135 bits. Absolute addresses support the full address range.
22137 @item -mcmodel=large
22138 @opindex mcmodel=large
22139 Generate code for the large model. There is no limitation on call
22140 distance, pc-relative addresses, or absolute addresses.
22142 @item -mcpu=@var{name}
22144 Selects the type of CPU to be targeted. Currently the only supported
22145 type is @samp{tilegx}.
22151 Generate code for a 32-bit or 64-bit environment. The 32-bit
22152 environment sets int, long, and pointer to 32 bits. The 64-bit
22153 environment sets int to 32 bits and long and pointer to 64 bits.
22156 @itemx -mlittle-endian
22157 @opindex mbig-endian
22158 @opindex mlittle-endian
22159 Generate code in big/little endian mode, respectively.
22162 @node TILEPro Options
22163 @subsection TILEPro Options
22164 @cindex TILEPro options
22166 These @samp{-m} options are supported on the TILEPro:
22169 @item -mcpu=@var{name}
22171 Selects the type of CPU to be targeted. Currently the only supported
22172 type is @samp{tilepro}.
22176 Generate code for a 32-bit environment, which sets int, long, and
22177 pointer to 32 bits. This is the only supported behavior so the flag
22178 is essentially ignored.
22182 @subsection V850 Options
22183 @cindex V850 Options
22185 These @samp{-m} options are defined for V850 implementations:
22189 @itemx -mno-long-calls
22190 @opindex mlong-calls
22191 @opindex mno-long-calls
22192 Treat all calls as being far away (near). If calls are assumed to be
22193 far away, the compiler always loads the function's address into a
22194 register, and calls indirect through the pointer.
22200 Do not optimize (do optimize) basic blocks that use the same index
22201 pointer 4 or more times to copy pointer into the @code{ep} register, and
22202 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
22203 option is on by default if you optimize.
22205 @item -mno-prolog-function
22206 @itemx -mprolog-function
22207 @opindex mno-prolog-function
22208 @opindex mprolog-function
22209 Do not use (do use) external functions to save and restore registers
22210 at the prologue and epilogue of a function. The external functions
22211 are slower, but use less code space if more than one function saves
22212 the same number of registers. The @option{-mprolog-function} option
22213 is on by default if you optimize.
22217 Try to make the code as small as possible. At present, this just turns
22218 on the @option{-mep} and @option{-mprolog-function} options.
22220 @item -mtda=@var{n}
22222 Put static or global variables whose size is @var{n} bytes or less into
22223 the tiny data area that register @code{ep} points to. The tiny data
22224 area can hold up to 256 bytes in total (128 bytes for byte references).
22226 @item -msda=@var{n}
22228 Put static or global variables whose size is @var{n} bytes or less into
22229 the small data area that register @code{gp} points to. The small data
22230 area can hold up to 64 kilobytes.
22232 @item -mzda=@var{n}
22234 Put static or global variables whose size is @var{n} bytes or less into
22235 the first 32 kilobytes of memory.
22239 Specify that the target processor is the V850.
22243 Specify that the target processor is the V850E3V5. The preprocessor
22244 constant @samp{__v850e3v5__} is defined if this option is used.
22248 Specify that the target processor is the V850E3V5. This is an alias for
22249 the @option{-mv850e3v5} option.
22253 Specify that the target processor is the V850E2V3. The preprocessor
22254 constant @samp{__v850e2v3__} is defined if this option is used.
22258 Specify that the target processor is the V850E2. The preprocessor
22259 constant @samp{__v850e2__} is defined if this option is used.
22263 Specify that the target processor is the V850E1. The preprocessor
22264 constants @samp{__v850e1__} and @samp{__v850e__} are defined if
22265 this option is used.
22269 Specify that the target processor is the V850ES. This is an alias for
22270 the @option{-mv850e1} option.
22274 Specify that the target processor is the V850E@. The preprocessor
22275 constant @samp{__v850e__} is defined if this option is used.
22277 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
22278 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
22279 are defined then a default target processor is chosen and the
22280 relevant @samp{__v850*__} preprocessor constant is defined.
22282 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
22283 defined, regardless of which processor variant is the target.
22285 @item -mdisable-callt
22286 @itemx -mno-disable-callt
22287 @opindex mdisable-callt
22288 @opindex mno-disable-callt
22289 This option suppresses generation of the @code{CALLT} instruction for the
22290 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
22293 This option is enabled by default when the RH850 ABI is
22294 in use (see @option{-mrh850-abi}), and disabled by default when the
22295 GCC ABI is in use. If @code{CALLT} instructions are being generated
22296 then the C preprocessor symbol @code{__V850_CALLT__} will be defined.
22302 Pass on (or do not pass on) the @option{-mrelax} command line option
22306 @itemx -mno-long-jumps
22307 @opindex mlong-jumps
22308 @opindex mno-long-jumps
22309 Disable (or re-enable) the generation of PC-relative jump instructions.
22312 @itemx -mhard-float
22313 @opindex msoft-float
22314 @opindex mhard-float
22315 Disable (or re-enable) the generation of hardware floating point
22316 instructions. This option is only significant when the target
22317 architecture is @samp{V850E2V3} or higher. If hardware floating point
22318 instructions are being generated then the C preprocessor symbol
22319 @code{__FPU_OK__} will be defined, otherwise the symbol
22320 @code{__NO_FPU__} will be defined.
22324 Enables the use of the e3v5 LOOP instruction. The use of this
22325 instruction is not enabled by default when the e3v5 architecture is
22326 selected because its use is still experimental.
22330 @opindex mrh850-abi
22332 Enables support for the RH850 version of the V850 ABI. This is the
22333 default. With this version of the ABI the following rules apply:
22337 Integer sized structures and unions are returned via a memory pointer
22338 rather than a register.
22341 Large structures and unions (more than 8 bytes in size) are passed by
22345 Functions are aligned to 16-bit boundaries.
22348 The @option{-m8byte-align} command line option is supported.
22351 The @option{-mdisable-callt} command line option is enabled by
22352 default. The @option{-mno-disable-callt} command line option is not
22356 When this version of the ABI is enabled the C preprocessor symbol
22357 @code{__V850_RH850_ABI__} is defined.
22361 Enables support for the old GCC version of the V850 ABI. With this
22362 version of the ABI the following rules apply:
22366 Integer sized structures and unions are returned in register @code{r10}.
22369 Large structures and unions (more than 8 bytes in size) are passed by
22373 Functions are aligned to 32-bit boundaries, unless optimizing for
22377 The @option{-m8byte-align} command line option is not supported.
22380 The @option{-mdisable-callt} command line option is supported but not
22381 enabled by default.
22384 When this version of the ABI is enabled the C preprocessor symbol
22385 @code{__V850_GCC_ABI__} is defined.
22387 @item -m8byte-align
22388 @itemx -mno-8byte-align
22389 @opindex m8byte-align
22390 @opindex mno-8byte-align
22391 Enables support for @code{doubles} and @code{long long} types to be
22392 aligned on 8-byte boundaries. The default is to restrict the
22393 alignment of all objects to at most 4-bytes. When
22394 @option{-m8byte-align} is in effect the C preprocessor symbol
22395 @code{__V850_8BYTE_ALIGN__} will be defined.
22398 @opindex mbig-switch
22399 Generate code suitable for big switch tables. Use this option only if
22400 the assembler/linker complain about out of range branches within a switch
22405 This option causes r2 and r5 to be used in the code generated by
22406 the compiler. This setting is the default.
22408 @item -mno-app-regs
22409 @opindex mno-app-regs
22410 This option causes r2 and r5 to be treated as fixed registers.
22415 @subsection VAX Options
22416 @cindex VAX options
22418 These @samp{-m} options are defined for the VAX:
22423 Do not output certain jump instructions (@code{aobleq} and so on)
22424 that the Unix assembler for the VAX cannot handle across long
22429 Do output those jump instructions, on the assumption that the
22430 GNU assembler is being used.
22434 Output code for G-format floating-point numbers instead of D-format.
22438 @subsection VMS Options
22440 These @samp{-m} options are defined for the VMS implementations:
22443 @item -mvms-return-codes
22444 @opindex mvms-return-codes
22445 Return VMS condition codes from @code{main}. The default is to return POSIX-style
22446 condition (e.g.@ error) codes.
22448 @item -mdebug-main=@var{prefix}
22449 @opindex mdebug-main=@var{prefix}
22450 Flag the first routine whose name starts with @var{prefix} as the main
22451 routine for the debugger.
22455 Default to 64-bit memory allocation routines.
22457 @item -mpointer-size=@var{size}
22458 @opindex -mpointer-size=@var{size}
22459 Set the default size of pointers. Possible options for @var{size} are
22460 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
22461 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
22462 The later option disables @code{pragma pointer_size}.
22465 @node VxWorks Options
22466 @subsection VxWorks Options
22467 @cindex VxWorks Options
22469 The options in this section are defined for all VxWorks targets.
22470 Options specific to the target hardware are listed with the other
22471 options for that target.
22476 GCC can generate code for both VxWorks kernels and real time processes
22477 (RTPs). This option switches from the former to the latter. It also
22478 defines the preprocessor macro @code{__RTP__}.
22481 @opindex non-static
22482 Link an RTP executable against shared libraries rather than static
22483 libraries. The options @option{-static} and @option{-shared} can
22484 also be used for RTPs (@pxref{Link Options}); @option{-static}
22491 These options are passed down to the linker. They are defined for
22492 compatibility with Diab.
22495 @opindex Xbind-lazy
22496 Enable lazy binding of function calls. This option is equivalent to
22497 @option{-Wl,-z,now} and is defined for compatibility with Diab.
22501 Disable lazy binding of function calls. This option is the default and
22502 is defined for compatibility with Diab.
22505 @node x86-64 Options
22506 @subsection x86-64 Options
22507 @cindex x86-64 options
22509 These are listed under @xref{i386 and x86-64 Options}.
22511 @node Xstormy16 Options
22512 @subsection Xstormy16 Options
22513 @cindex Xstormy16 Options
22515 These options are defined for Xstormy16:
22520 Choose startup files and linker script suitable for the simulator.
22523 @node Xtensa Options
22524 @subsection Xtensa Options
22525 @cindex Xtensa Options
22527 These options are supported for Xtensa targets:
22531 @itemx -mno-const16
22533 @opindex mno-const16
22534 Enable or disable use of @code{CONST16} instructions for loading
22535 constant values. The @code{CONST16} instruction is currently not a
22536 standard option from Tensilica. When enabled, @code{CONST16}
22537 instructions are always used in place of the standard @code{L32R}
22538 instructions. The use of @code{CONST16} is enabled by default only if
22539 the @code{L32R} instruction is not available.
22542 @itemx -mno-fused-madd
22543 @opindex mfused-madd
22544 @opindex mno-fused-madd
22545 Enable or disable use of fused multiply/add and multiply/subtract
22546 instructions in the floating-point option. This has no effect if the
22547 floating-point option is not also enabled. Disabling fused multiply/add
22548 and multiply/subtract instructions forces the compiler to use separate
22549 instructions for the multiply and add/subtract operations. This may be
22550 desirable in some cases where strict IEEE 754-compliant results are
22551 required: the fused multiply add/subtract instructions do not round the
22552 intermediate result, thereby producing results with @emph{more} bits of
22553 precision than specified by the IEEE standard. Disabling fused multiply
22554 add/subtract instructions also ensures that the program output is not
22555 sensitive to the compiler's ability to combine multiply and add/subtract
22558 @item -mserialize-volatile
22559 @itemx -mno-serialize-volatile
22560 @opindex mserialize-volatile
22561 @opindex mno-serialize-volatile
22562 When this option is enabled, GCC inserts @code{MEMW} instructions before
22563 @code{volatile} memory references to guarantee sequential consistency.
22564 The default is @option{-mserialize-volatile}. Use
22565 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
22567 @item -mforce-no-pic
22568 @opindex mforce-no-pic
22569 For targets, like GNU/Linux, where all user-mode Xtensa code must be
22570 position-independent code (PIC), this option disables PIC for compiling
22573 @item -mtext-section-literals
22574 @itemx -mno-text-section-literals
22575 @opindex mtext-section-literals
22576 @opindex mno-text-section-literals
22577 Control the treatment of literal pools. The default is
22578 @option{-mno-text-section-literals}, which places literals in a separate
22579 section in the output file. This allows the literal pool to be placed
22580 in a data RAM/ROM, and it also allows the linker to combine literal
22581 pools from separate object files to remove redundant literals and
22582 improve code size. With @option{-mtext-section-literals}, the literals
22583 are interspersed in the text section in order to keep them as close as
22584 possible to their references. This may be necessary for large assembly
22587 @item -mtarget-align
22588 @itemx -mno-target-align
22589 @opindex mtarget-align
22590 @opindex mno-target-align
22591 When this option is enabled, GCC instructs the assembler to
22592 automatically align instructions to reduce branch penalties at the
22593 expense of some code density. The assembler attempts to widen density
22594 instructions to align branch targets and the instructions following call
22595 instructions. If there are not enough preceding safe density
22596 instructions to align a target, no widening is performed. The
22597 default is @option{-mtarget-align}. These options do not affect the
22598 treatment of auto-aligned instructions like @code{LOOP}, which the
22599 assembler always aligns, either by widening density instructions or
22600 by inserting NOP instructions.
22603 @itemx -mno-longcalls
22604 @opindex mlongcalls
22605 @opindex mno-longcalls
22606 When this option is enabled, GCC instructs the assembler to translate
22607 direct calls to indirect calls unless it can determine that the target
22608 of a direct call is in the range allowed by the call instruction. This
22609 translation typically occurs for calls to functions in other source
22610 files. Specifically, the assembler translates a direct @code{CALL}
22611 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
22612 The default is @option{-mno-longcalls}. This option should be used in
22613 programs where the call target can potentially be out of range. This
22614 option is implemented in the assembler, not the compiler, so the
22615 assembly code generated by GCC still shows direct call
22616 instructions---look at the disassembled object code to see the actual
22617 instructions. Note that the assembler uses an indirect call for
22618 every cross-file call, not just those that really are out of range.
22621 @node zSeries Options
22622 @subsection zSeries Options
22623 @cindex zSeries options
22625 These are listed under @xref{S/390 and zSeries Options}.
22627 @node Code Gen Options
22628 @section Options for Code Generation Conventions
22629 @cindex code generation conventions
22630 @cindex options, code generation
22631 @cindex run-time options
22633 These machine-independent options control the interface conventions
22634 used in code generation.
22636 Most of them have both positive and negative forms; the negative form
22637 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
22638 one of the forms is listed---the one that is not the default. You
22639 can figure out the other form by either removing @samp{no-} or adding
22643 @item -fbounds-check
22644 @opindex fbounds-check
22645 For front ends that support it, generate additional code to check that
22646 indices used to access arrays are within the declared range. This is
22647 currently only supported by the Java and Fortran front ends, where
22648 this option defaults to true and false respectively.
22650 @item -fstack-reuse=@var{reuse-level}
22651 @opindex fstack_reuse
22652 This option controls stack space reuse for user declared local/auto variables
22653 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
22654 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
22655 local variables and temporaries, @samp{named_vars} enables the reuse only for
22656 user defined local variables with names, and @samp{none} disables stack reuse
22657 completely. The default value is @samp{all}. The option is needed when the
22658 program extends the lifetime of a scoped local variable or a compiler generated
22659 temporary beyond the end point defined by the language. When a lifetime of
22660 a variable ends, and if the variable lives in memory, the optimizing compiler
22661 has the freedom to reuse its stack space with other temporaries or scoped
22662 local variables whose live range does not overlap with it. Legacy code extending
22663 local lifetime will likely to break with the stack reuse optimization.
22682 if (*p == 10) // out of scope use of local1
22693 A(int k) : i(k), j(k) @{ @}
22700 void foo(const A& ar)
22707 foo(A(10)); // temp object's lifetime ends when foo returns
22713 ap->i+= 10; // ap references out of scope temp whose space
22714 // is reused with a. What is the value of ap->i?
22719 The lifetime of a compiler generated temporary is well defined by the C++
22720 standard. When a lifetime of a temporary ends, and if the temporary lives
22721 in memory, the optimizing compiler has the freedom to reuse its stack
22722 space with other temporaries or scoped local variables whose live range
22723 does not overlap with it. However some of the legacy code relies on
22724 the behavior of older compilers in which temporaries' stack space is
22725 not reused, the aggressive stack reuse can lead to runtime errors. This
22726 option is used to control the temporary stack reuse optimization.
22730 This option generates traps for signed overflow on addition, subtraction,
22731 multiplication operations.
22735 This option instructs the compiler to assume that signed arithmetic
22736 overflow of addition, subtraction and multiplication wraps around
22737 using twos-complement representation. This flag enables some optimizations
22738 and disables others. This option is enabled by default for the Java
22739 front end, as required by the Java language specification.
22742 @opindex fexceptions
22743 Enable exception handling. Generates extra code needed to propagate
22744 exceptions. For some targets, this implies GCC generates frame
22745 unwind information for all functions, which can produce significant data
22746 size overhead, although it does not affect execution. If you do not
22747 specify this option, GCC enables it by default for languages like
22748 C++ that normally require exception handling, and disables it for
22749 languages like C that do not normally require it. However, you may need
22750 to enable this option when compiling C code that needs to interoperate
22751 properly with exception handlers written in C++. You may also wish to
22752 disable this option if you are compiling older C++ programs that don't
22753 use exception handling.
22755 @item -fnon-call-exceptions
22756 @opindex fnon-call-exceptions
22757 Generate code that allows trapping instructions to throw exceptions.
22758 Note that this requires platform-specific runtime support that does
22759 not exist everywhere. Moreover, it only allows @emph{trapping}
22760 instructions to throw exceptions, i.e.@: memory references or floating-point
22761 instructions. It does not allow exceptions to be thrown from
22762 arbitrary signal handlers such as @code{SIGALRM}.
22764 @item -fdelete-dead-exceptions
22765 @opindex fdelete-dead-exceptions
22766 Consider that instructions that may throw exceptions but don't otherwise
22767 contribute to the execution of the program can be optimized away.
22768 This option is enabled by default for the Ada front end, as permitted by
22769 the Ada language specification.
22770 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
22772 @item -funwind-tables
22773 @opindex funwind-tables
22774 Similar to @option{-fexceptions}, except that it just generates any needed
22775 static data, but does not affect the generated code in any other way.
22776 You normally do not need to enable this option; instead, a language processor
22777 that needs this handling enables it on your behalf.
22779 @item -fasynchronous-unwind-tables
22780 @opindex fasynchronous-unwind-tables
22781 Generate unwind table in DWARF 2 format, if supported by target machine. The
22782 table is exact at each instruction boundary, so it can be used for stack
22783 unwinding from asynchronous events (such as debugger or garbage collector).
22785 @item -fno-gnu-unique
22786 @opindex fno-gnu-unique
22787 On systems with recent GNU assembler and C library, the C++ compiler
22788 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
22789 of template static data members and static local variables in inline
22790 functions are unique even in the presence of @code{RTLD_LOCAL}; this
22791 is necessary to avoid problems with a library used by two different
22792 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
22793 therefore disagreeing with the other one about the binding of the
22794 symbol. But this causes @code{dlclose} to be ignored for affected
22795 DSOs; if your program relies on reinitialization of a DSO via
22796 @code{dlclose} and @code{dlopen}, you can use
22797 @option{-fno-gnu-unique}.
22799 @item -fpcc-struct-return
22800 @opindex fpcc-struct-return
22801 Return ``short'' @code{struct} and @code{union} values in memory like
22802 longer ones, rather than in registers. This convention is less
22803 efficient, but it has the advantage of allowing intercallability between
22804 GCC-compiled files and files compiled with other compilers, particularly
22805 the Portable C Compiler (pcc).
22807 The precise convention for returning structures in memory depends
22808 on the target configuration macros.
22810 Short structures and unions are those whose size and alignment match
22811 that of some integer type.
22813 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
22814 switch is not binary compatible with code compiled with the
22815 @option{-freg-struct-return} switch.
22816 Use it to conform to a non-default application binary interface.
22818 @item -freg-struct-return
22819 @opindex freg-struct-return
22820 Return @code{struct} and @code{union} values in registers when possible.
22821 This is more efficient for small structures than
22822 @option{-fpcc-struct-return}.
22824 If you specify neither @option{-fpcc-struct-return} nor
22825 @option{-freg-struct-return}, GCC defaults to whichever convention is
22826 standard for the target. If there is no standard convention, GCC
22827 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
22828 the principal compiler. In those cases, we can choose the standard, and
22829 we chose the more efficient register return alternative.
22831 @strong{Warning:} code compiled with the @option{-freg-struct-return}
22832 switch is not binary compatible with code compiled with the
22833 @option{-fpcc-struct-return} switch.
22834 Use it to conform to a non-default application binary interface.
22836 @item -fshort-enums
22837 @opindex fshort-enums
22838 Allocate to an @code{enum} type only as many bytes as it needs for the
22839 declared range of possible values. Specifically, the @code{enum} type
22840 is equivalent to the smallest integer type that has enough room.
22842 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
22843 code that is not binary compatible with code generated without that switch.
22844 Use it to conform to a non-default application binary interface.
22846 @item -fshort-double
22847 @opindex fshort-double
22848 Use the same size for @code{double} as for @code{float}.
22850 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
22851 code that is not binary compatible with code generated without that switch.
22852 Use it to conform to a non-default application binary interface.
22854 @item -fshort-wchar
22855 @opindex fshort-wchar
22856 Override the underlying type for @samp{wchar_t} to be @samp{short
22857 unsigned int} instead of the default for the target. This option is
22858 useful for building programs to run under WINE@.
22860 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
22861 code that is not binary compatible with code generated without that switch.
22862 Use it to conform to a non-default application binary interface.
22865 @opindex fno-common
22866 In C code, controls the placement of uninitialized global variables.
22867 Unix C compilers have traditionally permitted multiple definitions of
22868 such variables in different compilation units by placing the variables
22870 This is the behavior specified by @option{-fcommon}, and is the default
22871 for GCC on most targets.
22872 On the other hand, this behavior is not required by ISO C, and on some
22873 targets may carry a speed or code size penalty on variable references.
22874 The @option{-fno-common} option specifies that the compiler should place
22875 uninitialized global variables in the data section of the object file,
22876 rather than generating them as common blocks.
22877 This has the effect that if the same variable is declared
22878 (without @code{extern}) in two different compilations,
22879 you get a multiple-definition error when you link them.
22880 In this case, you must compile with @option{-fcommon} instead.
22881 Compiling with @option{-fno-common} is useful on targets for which
22882 it provides better performance, or if you wish to verify that the
22883 program will work on other systems that always treat uninitialized
22884 variable declarations this way.
22888 Ignore the @samp{#ident} directive.
22890 @item -finhibit-size-directive
22891 @opindex finhibit-size-directive
22892 Don't output a @code{.size} assembler directive, or anything else that
22893 would cause trouble if the function is split in the middle, and the
22894 two halves are placed at locations far apart in memory. This option is
22895 used when compiling @file{crtstuff.c}; you should not need to use it
22898 @item -fverbose-asm
22899 @opindex fverbose-asm
22900 Put extra commentary information in the generated assembly code to
22901 make it more readable. This option is generally only of use to those
22902 who actually need to read the generated assembly code (perhaps while
22903 debugging the compiler itself).
22905 @option{-fno-verbose-asm}, the default, causes the
22906 extra information to be omitted and is useful when comparing two assembler
22909 @item -frecord-gcc-switches
22910 @opindex frecord-gcc-switches
22911 This switch causes the command line used to invoke the
22912 compiler to be recorded into the object file that is being created.
22913 This switch is only implemented on some targets and the exact format
22914 of the recording is target and binary file format dependent, but it
22915 usually takes the form of a section containing ASCII text. This
22916 switch is related to the @option{-fverbose-asm} switch, but that
22917 switch only records information in the assembler output file as
22918 comments, so it never reaches the object file.
22919 See also @option{-grecord-gcc-switches} for another
22920 way of storing compiler options into the object file.
22924 @cindex global offset table
22926 Generate position-independent code (PIC) suitable for use in a shared
22927 library, if supported for the target machine. Such code accesses all
22928 constant addresses through a global offset table (GOT)@. The dynamic
22929 loader resolves the GOT entries when the program starts (the dynamic
22930 loader is not part of GCC; it is part of the operating system). If
22931 the GOT size for the linked executable exceeds a machine-specific
22932 maximum size, you get an error message from the linker indicating that
22933 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
22934 instead. (These maximums are 8k on the SPARC and 32k
22935 on the m68k and RS/6000. The 386 has no such limit.)
22937 Position-independent code requires special support, and therefore works
22938 only on certain machines. For the 386, GCC supports PIC for System V
22939 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
22940 position-independent.
22942 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
22947 If supported for the target machine, emit position-independent code,
22948 suitable for dynamic linking and avoiding any limit on the size of the
22949 global offset table. This option makes a difference on the m68k,
22950 PowerPC and SPARC@.
22952 Position-independent code requires special support, and therefore works
22953 only on certain machines.
22955 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
22962 These options are similar to @option{-fpic} and @option{-fPIC}, but
22963 generated position independent code can be only linked into executables.
22964 Usually these options are used when @option{-pie} GCC option is
22965 used during linking.
22967 @option{-fpie} and @option{-fPIE} both define the macros
22968 @code{__pie__} and @code{__PIE__}. The macros have the value 1
22969 for @option{-fpie} and 2 for @option{-fPIE}.
22971 @item -fno-jump-tables
22972 @opindex fno-jump-tables
22973 Do not use jump tables for switch statements even where it would be
22974 more efficient than other code generation strategies. This option is
22975 of use in conjunction with @option{-fpic} or @option{-fPIC} for
22976 building code that forms part of a dynamic linker and cannot
22977 reference the address of a jump table. On some targets, jump tables
22978 do not require a GOT and this option is not needed.
22980 @item -ffixed-@var{reg}
22982 Treat the register named @var{reg} as a fixed register; generated code
22983 should never refer to it (except perhaps as a stack pointer, frame
22984 pointer or in some other fixed role).
22986 @var{reg} must be the name of a register. The register names accepted
22987 are machine-specific and are defined in the @code{REGISTER_NAMES}
22988 macro in the machine description macro file.
22990 This flag does not have a negative form, because it specifies a
22993 @item -fcall-used-@var{reg}
22994 @opindex fcall-used
22995 Treat the register named @var{reg} as an allocable register that is
22996 clobbered by function calls. It may be allocated for temporaries or
22997 variables that do not live across a call. Functions compiled this way
22998 do not save and restore the register @var{reg}.
23000 It is an error to use this flag with the frame pointer or stack pointer.
23001 Use of this flag for other registers that have fixed pervasive roles in
23002 the machine's execution model produces disastrous results.
23004 This flag does not have a negative form, because it specifies a
23007 @item -fcall-saved-@var{reg}
23008 @opindex fcall-saved
23009 Treat the register named @var{reg} as an allocable register saved by
23010 functions. It may be allocated even for temporaries or variables that
23011 live across a call. Functions compiled this way save and restore
23012 the register @var{reg} if they use it.
23014 It is an error to use this flag with the frame pointer or stack pointer.
23015 Use of this flag for other registers that have fixed pervasive roles in
23016 the machine's execution model produces disastrous results.
23018 A different sort of disaster results from the use of this flag for
23019 a register in which function values may be returned.
23021 This flag does not have a negative form, because it specifies a
23024 @item -fpack-struct[=@var{n}]
23025 @opindex fpack-struct
23026 Without a value specified, pack all structure members together without
23027 holes. When a value is specified (which must be a small power of two), pack
23028 structure members according to this value, representing the maximum
23029 alignment (that is, objects with default alignment requirements larger than
23030 this are output potentially unaligned at the next fitting location.
23032 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
23033 code that is not binary compatible with code generated without that switch.
23034 Additionally, it makes the code suboptimal.
23035 Use it to conform to a non-default application binary interface.
23037 @item -finstrument-functions
23038 @opindex finstrument-functions
23039 Generate instrumentation calls for entry and exit to functions. Just
23040 after function entry and just before function exit, the following
23041 profiling functions are called with the address of the current
23042 function and its call site. (On some platforms,
23043 @code{__builtin_return_address} does not work beyond the current
23044 function, so the call site information may not be available to the
23045 profiling functions otherwise.)
23048 void __cyg_profile_func_enter (void *this_fn,
23050 void __cyg_profile_func_exit (void *this_fn,
23054 The first argument is the address of the start of the current function,
23055 which may be looked up exactly in the symbol table.
23057 This instrumentation is also done for functions expanded inline in other
23058 functions. The profiling calls indicate where, conceptually, the
23059 inline function is entered and exited. This means that addressable
23060 versions of such functions must be available. If all your uses of a
23061 function are expanded inline, this may mean an additional expansion of
23062 code size. If you use @samp{extern inline} in your C code, an
23063 addressable version of such functions must be provided. (This is
23064 normally the case anyway, but if you get lucky and the optimizer always
23065 expands the functions inline, you might have gotten away without
23066 providing static copies.)
23068 A function may be given the attribute @code{no_instrument_function}, in
23069 which case this instrumentation is not done. This can be used, for
23070 example, for the profiling functions listed above, high-priority
23071 interrupt routines, and any functions from which the profiling functions
23072 cannot safely be called (perhaps signal handlers, if the profiling
23073 routines generate output or allocate memory).
23075 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
23076 @opindex finstrument-functions-exclude-file-list
23078 Set the list of functions that are excluded from instrumentation (see
23079 the description of @code{-finstrument-functions}). If the file that
23080 contains a function definition matches with one of @var{file}, then
23081 that function is not instrumented. The match is done on substrings:
23082 if the @var{file} parameter is a substring of the file name, it is
23083 considered to be a match.
23088 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
23092 excludes any inline function defined in files whose pathnames
23093 contain @code{/bits/stl} or @code{include/sys}.
23095 If, for some reason, you want to include letter @code{','} in one of
23096 @var{sym}, write @code{'\,'}. For example,
23097 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
23098 (note the single quote surrounding the option).
23100 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
23101 @opindex finstrument-functions-exclude-function-list
23103 This is similar to @code{-finstrument-functions-exclude-file-list},
23104 but this option sets the list of function names to be excluded from
23105 instrumentation. The function name to be matched is its user-visible
23106 name, such as @code{vector<int> blah(const vector<int> &)}, not the
23107 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
23108 match is done on substrings: if the @var{sym} parameter is a substring
23109 of the function name, it is considered to be a match. For C99 and C++
23110 extended identifiers, the function name must be given in UTF-8, not
23111 using universal character names.
23113 @item -fstack-check
23114 @opindex fstack-check
23115 Generate code to verify that you do not go beyond the boundary of the
23116 stack. You should specify this flag if you are running in an
23117 environment with multiple threads, but you only rarely need to specify it in
23118 a single-threaded environment since stack overflow is automatically
23119 detected on nearly all systems if there is only one stack.
23121 Note that this switch does not actually cause checking to be done; the
23122 operating system or the language runtime must do that. The switch causes
23123 generation of code to ensure that they see the stack being extended.
23125 You can additionally specify a string parameter: @code{no} means no
23126 checking, @code{generic} means force the use of old-style checking,
23127 @code{specific} means use the best checking method and is equivalent
23128 to bare @option{-fstack-check}.
23130 Old-style checking is a generic mechanism that requires no specific
23131 target support in the compiler but comes with the following drawbacks:
23135 Modified allocation strategy for large objects: they are always
23136 allocated dynamically if their size exceeds a fixed threshold.
23139 Fixed limit on the size of the static frame of functions: when it is
23140 topped by a particular function, stack checking is not reliable and
23141 a warning is issued by the compiler.
23144 Inefficiency: because of both the modified allocation strategy and the
23145 generic implementation, code performance is hampered.
23148 Note that old-style stack checking is also the fallback method for
23149 @code{specific} if no target support has been added in the compiler.
23151 @item -fstack-limit-register=@var{reg}
23152 @itemx -fstack-limit-symbol=@var{sym}
23153 @itemx -fno-stack-limit
23154 @opindex fstack-limit-register
23155 @opindex fstack-limit-symbol
23156 @opindex fno-stack-limit
23157 Generate code to ensure that the stack does not grow beyond a certain value,
23158 either the value of a register or the address of a symbol. If a larger
23159 stack is required, a signal is raised at run time. For most targets,
23160 the signal is raised before the stack overruns the boundary, so
23161 it is possible to catch the signal without taking special precautions.
23163 For instance, if the stack starts at absolute address @samp{0x80000000}
23164 and grows downwards, you can use the flags
23165 @option{-fstack-limit-symbol=__stack_limit} and
23166 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
23167 of 128KB@. Note that this may only work with the GNU linker.
23169 @item -fsplit-stack
23170 @opindex fsplit-stack
23171 Generate code to automatically split the stack before it overflows.
23172 The resulting program has a discontiguous stack which can only
23173 overflow if the program is unable to allocate any more memory. This
23174 is most useful when running threaded programs, as it is no longer
23175 necessary to calculate a good stack size to use for each thread. This
23176 is currently only implemented for the i386 and x86_64 back ends running
23179 When code compiled with @option{-fsplit-stack} calls code compiled
23180 without @option{-fsplit-stack}, there may not be much stack space
23181 available for the latter code to run. If compiling all code,
23182 including library code, with @option{-fsplit-stack} is not an option,
23183 then the linker can fix up these calls so that the code compiled
23184 without @option{-fsplit-stack} always has a large stack. Support for
23185 this is implemented in the gold linker in GNU binutils release 2.21
23188 @item -fleading-underscore
23189 @opindex fleading-underscore
23190 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
23191 change the way C symbols are represented in the object file. One use
23192 is to help link with legacy assembly code.
23194 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
23195 generate code that is not binary compatible with code generated without that
23196 switch. Use it to conform to a non-default application binary interface.
23197 Not all targets provide complete support for this switch.
23199 @item -ftls-model=@var{model}
23200 @opindex ftls-model
23201 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
23202 The @var{model} argument should be one of @code{global-dynamic},
23203 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
23204 Note that the choice is subject to optimization: the compiler may use
23205 a more efficient model for symbols not visible outside of the translation
23206 unit, or if @option{-fpic} is not given on the command line.
23208 The default without @option{-fpic} is @code{initial-exec}; with
23209 @option{-fpic} the default is @code{global-dynamic}.
23211 @item -fvisibility=@var{default|internal|hidden|protected}
23212 @opindex fvisibility
23213 Set the default ELF image symbol visibility to the specified option---all
23214 symbols are marked with this unless overridden within the code.
23215 Using this feature can very substantially improve linking and
23216 load times of shared object libraries, produce more optimized
23217 code, provide near-perfect API export and prevent symbol clashes.
23218 It is @strong{strongly} recommended that you use this in any shared objects
23221 Despite the nomenclature, @code{default} always means public; i.e.,
23222 available to be linked against from outside the shared object.
23223 @code{protected} and @code{internal} are pretty useless in real-world
23224 usage so the only other commonly used option is @code{hidden}.
23225 The default if @option{-fvisibility} isn't specified is
23226 @code{default}, i.e., make every
23227 symbol public---this causes the same behavior as previous versions of
23230 A good explanation of the benefits offered by ensuring ELF
23231 symbols have the correct visibility is given by ``How To Write
23232 Shared Libraries'' by Ulrich Drepper (which can be found at
23233 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
23234 solution made possible by this option to marking things hidden when
23235 the default is public is to make the default hidden and mark things
23236 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
23237 and @code{__attribute__ ((visibility("default")))} instead of
23238 @code{__declspec(dllexport)} you get almost identical semantics with
23239 identical syntax. This is a great boon to those working with
23240 cross-platform projects.
23242 For those adding visibility support to existing code, you may find
23243 @samp{#pragma GCC visibility} of use. This works by you enclosing
23244 the declarations you wish to set visibility for with (for example)
23245 @samp{#pragma GCC visibility push(hidden)} and
23246 @samp{#pragma GCC visibility pop}.
23247 Bear in mind that symbol visibility should be viewed @strong{as
23248 part of the API interface contract} and thus all new code should
23249 always specify visibility when it is not the default; i.e., declarations
23250 only for use within the local DSO should @strong{always} be marked explicitly
23251 as hidden as so to avoid PLT indirection overheads---making this
23252 abundantly clear also aids readability and self-documentation of the code.
23253 Note that due to ISO C++ specification requirements, @code{operator new} and
23254 @code{operator delete} must always be of default visibility.
23256 Be aware that headers from outside your project, in particular system
23257 headers and headers from any other library you use, may not be
23258 expecting to be compiled with visibility other than the default. You
23259 may need to explicitly say @samp{#pragma GCC visibility push(default)}
23260 before including any such headers.
23262 @samp{extern} declarations are not affected by @option{-fvisibility}, so
23263 a lot of code can be recompiled with @option{-fvisibility=hidden} with
23264 no modifications. However, this means that calls to @code{extern}
23265 functions with no explicit visibility use the PLT, so it is more
23266 effective to use @code{__attribute ((visibility))} and/or
23267 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
23268 declarations should be treated as hidden.
23270 Note that @option{-fvisibility} does affect C++ vague linkage
23271 entities. This means that, for instance, an exception class that is
23272 be thrown between DSOs must be explicitly marked with default
23273 visibility so that the @samp{type_info} nodes are unified between
23276 An overview of these techniques, their benefits and how to use them
23277 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
23279 @item -fstrict-volatile-bitfields
23280 @opindex fstrict-volatile-bitfields
23281 This option should be used if accesses to volatile bit-fields (or other
23282 structure fields, although the compiler usually honors those types
23283 anyway) should use a single access of the width of the
23284 field's type, aligned to a natural alignment if possible. For
23285 example, targets with memory-mapped peripheral registers might require
23286 all such accesses to be 16 bits wide; with this flag you can
23287 declare all peripheral bit-fields as @code{unsigned short} (assuming short
23288 is 16 bits on these targets) to force GCC to use 16-bit accesses
23289 instead of, perhaps, a more efficient 32-bit access.
23291 If this option is disabled, the compiler uses the most efficient
23292 instruction. In the previous example, that might be a 32-bit load
23293 instruction, even though that accesses bytes that do not contain
23294 any portion of the bit-field, or memory-mapped registers unrelated to
23295 the one being updated.
23297 In some cases, such as when the @code{packed} attribute is applied to a
23298 structure field, it may not be possible to access the field with a single
23299 read or write that is correctly aligned for the target machine. In this
23300 case GCC falls back to generating multiple accesses rather than code that
23301 will fault or truncate the result at run time.
23303 Note: Due to restrictions of the C/C++11 memory model, write accesses are
23304 not allowed to touch non bit-field members. It is therefore recommended
23305 to define all bits of the field's type as bit-field members.
23307 The default value of this option is determined by the application binary
23308 interface for the target processor.
23310 @item -fsync-libcalls
23311 @opindex fsync-libcalls
23312 This option controls whether any out-of-line instance of the @code{__sync}
23313 family of functions may be used to implement the C++11 @code{__atomic}
23314 family of functions.
23316 The default value of this option is enabled, thus the only useful form
23317 of the option is @option{-fno-sync-libcalls}. This option is used in
23318 the implementation of the @file{libatomic} runtime library.
23324 @node Environment Variables
23325 @section Environment Variables Affecting GCC
23326 @cindex environment variables
23328 @c man begin ENVIRONMENT
23329 This section describes several environment variables that affect how GCC
23330 operates. Some of them work by specifying directories or prefixes to use
23331 when searching for various kinds of files. Some are used to specify other
23332 aspects of the compilation environment.
23334 Note that you can also specify places to search using options such as
23335 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
23336 take precedence over places specified using environment variables, which
23337 in turn take precedence over those specified by the configuration of GCC@.
23338 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
23339 GNU Compiler Collection (GCC) Internals}.
23344 @c @itemx LC_COLLATE
23346 @c @itemx LC_MONETARY
23347 @c @itemx LC_NUMERIC
23352 @c @findex LC_COLLATE
23353 @findex LC_MESSAGES
23354 @c @findex LC_MONETARY
23355 @c @findex LC_NUMERIC
23359 These environment variables control the way that GCC uses
23360 localization information which allows GCC to work with different
23361 national conventions. GCC inspects the locale categories
23362 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
23363 so. These locale categories can be set to any value supported by your
23364 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
23365 Kingdom encoded in UTF-8.
23367 The @env{LC_CTYPE} environment variable specifies character
23368 classification. GCC uses it to determine the character boundaries in
23369 a string; this is needed for some multibyte encodings that contain quote
23370 and escape characters that are otherwise interpreted as a string
23373 The @env{LC_MESSAGES} environment variable specifies the language to
23374 use in diagnostic messages.
23376 If the @env{LC_ALL} environment variable is set, it overrides the value
23377 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
23378 and @env{LC_MESSAGES} default to the value of the @env{LANG}
23379 environment variable. If none of these variables are set, GCC
23380 defaults to traditional C English behavior.
23384 If @env{TMPDIR} is set, it specifies the directory to use for temporary
23385 files. GCC uses temporary files to hold the output of one stage of
23386 compilation which is to be used as input to the next stage: for example,
23387 the output of the preprocessor, which is the input to the compiler
23390 @item GCC_COMPARE_DEBUG
23391 @findex GCC_COMPARE_DEBUG
23392 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
23393 @option{-fcompare-debug} to the compiler driver. See the documentation
23394 of this option for more details.
23396 @item GCC_EXEC_PREFIX
23397 @findex GCC_EXEC_PREFIX
23398 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
23399 names of the subprograms executed by the compiler. No slash is added
23400 when this prefix is combined with the name of a subprogram, but you can
23401 specify a prefix that ends with a slash if you wish.
23403 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
23404 an appropriate prefix to use based on the pathname it is invoked with.
23406 If GCC cannot find the subprogram using the specified prefix, it
23407 tries looking in the usual places for the subprogram.
23409 The default value of @env{GCC_EXEC_PREFIX} is
23410 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
23411 the installed compiler. In many cases @var{prefix} is the value
23412 of @code{prefix} when you ran the @file{configure} script.
23414 Other prefixes specified with @option{-B} take precedence over this prefix.
23416 This prefix is also used for finding files such as @file{crt0.o} that are
23419 In addition, the prefix is used in an unusual way in finding the
23420 directories to search for header files. For each of the standard
23421 directories whose name normally begins with @samp{/usr/local/lib/gcc}
23422 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
23423 replacing that beginning with the specified prefix to produce an
23424 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
23425 @file{foo/bar} just before it searches the standard directory
23426 @file{/usr/local/lib/bar}.
23427 If a standard directory begins with the configured
23428 @var{prefix} then the value of @var{prefix} is replaced by
23429 @env{GCC_EXEC_PREFIX} when looking for header files.
23431 @item COMPILER_PATH
23432 @findex COMPILER_PATH
23433 The value of @env{COMPILER_PATH} is a colon-separated list of
23434 directories, much like @env{PATH}. GCC tries the directories thus
23435 specified when searching for subprograms, if it can't find the
23436 subprograms using @env{GCC_EXEC_PREFIX}.
23439 @findex LIBRARY_PATH
23440 The value of @env{LIBRARY_PATH} is a colon-separated list of
23441 directories, much like @env{PATH}. When configured as a native compiler,
23442 GCC tries the directories thus specified when searching for special
23443 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
23444 using GCC also uses these directories when searching for ordinary
23445 libraries for the @option{-l} option (but directories specified with
23446 @option{-L} come first).
23450 @cindex locale definition
23451 This variable is used to pass locale information to the compiler. One way in
23452 which this information is used is to determine the character set to be used
23453 when character literals, string literals and comments are parsed in C and C++.
23454 When the compiler is configured to allow multibyte characters,
23455 the following values for @env{LANG} are recognized:
23459 Recognize JIS characters.
23461 Recognize SJIS characters.
23463 Recognize EUCJP characters.
23466 If @env{LANG} is not defined, or if it has some other value, then the
23467 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
23468 recognize and translate multibyte characters.
23472 Some additional environment variables affect the behavior of the
23475 @include cppenv.texi
23479 @node Precompiled Headers
23480 @section Using Precompiled Headers
23481 @cindex precompiled headers
23482 @cindex speed of compilation
23484 Often large projects have many header files that are included in every
23485 source file. The time the compiler takes to process these header files
23486 over and over again can account for nearly all of the time required to
23487 build the project. To make builds faster, GCC allows you to
23488 @dfn{precompile} a header file.
23490 To create a precompiled header file, simply compile it as you would any
23491 other file, if necessary using the @option{-x} option to make the driver
23492 treat it as a C or C++ header file. You may want to use a
23493 tool like @command{make} to keep the precompiled header up-to-date when
23494 the headers it contains change.
23496 A precompiled header file is searched for when @code{#include} is
23497 seen in the compilation. As it searches for the included file
23498 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
23499 compiler looks for a precompiled header in each directory just before it
23500 looks for the include file in that directory. The name searched for is
23501 the name specified in the @code{#include} with @samp{.gch} appended. If
23502 the precompiled header file can't be used, it is ignored.
23504 For instance, if you have @code{#include "all.h"}, and you have
23505 @file{all.h.gch} in the same directory as @file{all.h}, then the
23506 precompiled header file is used if possible, and the original
23507 header is used otherwise.
23509 Alternatively, you might decide to put the precompiled header file in a
23510 directory and use @option{-I} to ensure that directory is searched
23511 before (or instead of) the directory containing the original header.
23512 Then, if you want to check that the precompiled header file is always
23513 used, you can put a file of the same name as the original header in this
23514 directory containing an @code{#error} command.
23516 This also works with @option{-include}. So yet another way to use
23517 precompiled headers, good for projects not designed with precompiled
23518 header files in mind, is to simply take most of the header files used by
23519 a project, include them from another header file, precompile that header
23520 file, and @option{-include} the precompiled header. If the header files
23521 have guards against multiple inclusion, they are skipped because
23522 they've already been included (in the precompiled header).
23524 If you need to precompile the same header file for different
23525 languages, targets, or compiler options, you can instead make a
23526 @emph{directory} named like @file{all.h.gch}, and put each precompiled
23527 header in the directory, perhaps using @option{-o}. It doesn't matter
23528 what you call the files in the directory; every precompiled header in
23529 the directory is considered. The first precompiled header
23530 encountered in the directory that is valid for this compilation is
23531 used; they're searched in no particular order.
23533 There are many other possibilities, limited only by your imagination,
23534 good sense, and the constraints of your build system.
23536 A precompiled header file can be used only when these conditions apply:
23540 Only one precompiled header can be used in a particular compilation.
23543 A precompiled header can't be used once the first C token is seen. You
23544 can have preprocessor directives before a precompiled header; you cannot
23545 include a precompiled header from inside another header.
23548 The precompiled header file must be produced for the same language as
23549 the current compilation. You can't use a C precompiled header for a C++
23553 The precompiled header file must have been produced by the same compiler
23554 binary as the current compilation is using.
23557 Any macros defined before the precompiled header is included must
23558 either be defined in the same way as when the precompiled header was
23559 generated, or must not affect the precompiled header, which usually
23560 means that they don't appear in the precompiled header at all.
23562 The @option{-D} option is one way to define a macro before a
23563 precompiled header is included; using a @code{#define} can also do it.
23564 There are also some options that define macros implicitly, like
23565 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
23568 @item If debugging information is output when using the precompiled
23569 header, using @option{-g} or similar, the same kind of debugging information
23570 must have been output when building the precompiled header. However,
23571 a precompiled header built using @option{-g} can be used in a compilation
23572 when no debugging information is being output.
23574 @item The same @option{-m} options must generally be used when building
23575 and using the precompiled header. @xref{Submodel Options},
23576 for any cases where this rule is relaxed.
23578 @item Each of the following options must be the same when building and using
23579 the precompiled header:
23581 @gccoptlist{-fexceptions}
23584 Some other command-line options starting with @option{-f},
23585 @option{-p}, or @option{-O} must be defined in the same way as when
23586 the precompiled header was generated. At present, it's not clear
23587 which options are safe to change and which are not; the safest choice
23588 is to use exactly the same options when generating and using the
23589 precompiled header. The following are known to be safe:
23591 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
23592 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
23593 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
23598 For all of these except the last, the compiler automatically
23599 ignores the precompiled header if the conditions aren't met. If you
23600 find an option combination that doesn't work and doesn't cause the
23601 precompiled header to be ignored, please consider filing a bug report,
23604 If you do use differing options when generating and using the
23605 precompiled header, the actual behavior is a mixture of the
23606 behavior for the options. For instance, if you use @option{-g} to
23607 generate the precompiled header but not when using it, you may or may
23608 not get debugging information for routines in the precompiled header.