1 @c Copyright (C) 1988-2015 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{} 1988-2015 Free Software Foundation, Inc.
13 Permission is granted to copy, distribute and/or modify this document
14 under the terms of the GNU Free Documentation License, Version 1.3 or
15 any later version published by the Free Software Foundation; with the
16 Invariant Sections being ``GNU General Public License'' and ``Funding
17 Free Software'', the Front-Cover texts being (a) (see below), and with
18 the Back-Cover Texts being (b) (see below). A copy of the license is
19 included in the gfdl(7) man page.
21 (a) The FSF's Front-Cover Text is:
25 (b) The FSF's Back-Cover Text is:
27 You have freedom to copy and modify this GNU Manual, like GNU
28 software. Copies published by the Free Software Foundation raise
29 funds for GNU development.
31 @c Set file name and title for the man page.
33 @settitle GNU project C and C++ compiler
35 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
36 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
37 [@option{-W}@var{warn}@dots{}] [@option{-Wpedantic}]
38 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
39 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
40 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
41 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
43 Only the most useful options are listed here; see below for the
44 remainder. @command{g++} accepts mostly the same options as @command{gcc}.
47 gpl(7), gfdl(7), fsf-funding(7),
48 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
49 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
50 @file{ld}, @file{binutils} and @file{gdb}.
53 For instructions on reporting bugs, see
57 See the Info entry for @command{gcc}, or
58 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
59 for contributors to GCC@.
64 @chapter GCC Command Options
65 @cindex GCC command options
66 @cindex command options
67 @cindex options, GCC command
69 @c man begin DESCRIPTION
70 When you invoke GCC, it normally does preprocessing, compilation,
71 assembly and linking. The ``overall options'' allow you to stop this
72 process at an intermediate stage. For example, the @option{-c} option
73 says not to run the linker. Then the output consists of object files
74 output by the assembler.
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 -fopenacc -fopenmp -fopenmp-simd @gol
172 -fms-extensions -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 -fsized-deallocation @gol
191 -fstats -ftemplate-backtrace-limit=@var{n} @gol
192 -ftemplate-depth=@var{n} @gol
193 -fno-threadsafe-statics -fuse-cxa-atexit @gol
194 -fno-weak -nostdinc++ @gol
195 -fvisibility-inlines-hidden @gol
196 -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} @gol
197 -fvtv-counts -fvtv-debug @gol
198 -fvisibility-ms-compat @gol
199 -fext-numeric-literals @gol
200 -Wabi=@var{n} -Wconversion-null -Wctor-dtor-privacy @gol
201 -Wdelete-non-virtual-dtor -Wliteral-suffix -Wnarrowing @gol
202 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
203 -Weffc++ -Wstrict-null-sentinel @gol
204 -Wno-non-template-friend -Wold-style-cast @gol
205 -Woverloaded-virtual -Wno-pmf-conversions @gol
208 @item Objective-C and Objective-C++ Language Options
209 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
210 Objective-C and Objective-C++ Dialects}.
211 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
212 -fgnu-runtime -fnext-runtime @gol
213 -fno-nil-receivers @gol
214 -fobjc-abi-version=@var{n} @gol
215 -fobjc-call-cxx-cdtors @gol
216 -fobjc-direct-dispatch @gol
217 -fobjc-exceptions @gol
220 -fobjc-std=objc1 @gol
221 -fno-local-ivars @gol
222 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
223 -freplace-objc-classes @gol
226 -Wassign-intercept @gol
227 -Wno-protocol -Wselector @gol
228 -Wstrict-selector-match @gol
229 -Wundeclared-selector}
231 @item Language Independent Options
232 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
233 @gccoptlist{-fmessage-length=@var{n} @gol
234 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
235 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
236 -fno-diagnostics-show-option -fno-diagnostics-show-caret}
238 @item Warning Options
239 @xref{Warning Options,,Options to Request or Suppress Warnings}.
240 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
241 -pedantic-errors @gol
242 -w -Wextra -Wall -Waddress -Waggregate-return @gol
243 -Waggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
245 -Wno-attributes -Wno-builtin-macro-redefined @gol
246 -Wc90-c99-compat -Wc99-c11-compat @gol
247 -Wc++-compat -Wc++11-compat -Wc++14-compat -Wcast-align -Wcast-qual @gol
248 -Wchar-subscripts -Wclobbered -Wcomment -Wconditionally-supported @gol
249 -Wconversion -Wcoverage-mismatch -Wdate-time -Wdelete-incomplete -Wno-cpp @gol
250 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
251 -Wdisabled-optimization @gol
252 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
253 -Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol
254 -Wno-endif-labels -Werror -Werror=* @gol
255 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
256 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
257 -Wformat-security -Wformat-signedness -Wformat-y2k @gol
258 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
259 -Wignored-qualifiers -Wincompatible-pointer-types @gol
260 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
261 -Winit-self -Winline -Wno-int-conversion @gol
262 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
263 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
264 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
265 -Wmain -Wmaybe-uninitialized -Wmemset-transposed-args -Wmissing-braces @gol
266 -Wmissing-field-initializers -Wmissing-include-dirs @gol
267 -Wno-multichar -Wnonnull -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
268 -Wodr -Wno-overflow -Wopenmp-simd @gol
269 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
270 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
271 -Wpointer-arith -Wno-pointer-to-int-cast @gol
272 -Wredundant-decls -Wno-return-local-addr @gol
273 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
274 -Wshift-count-negative -Wshift-count-overflow @gol
275 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
276 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
277 -Wstack-protector -Wstack-usage=@var{len} -Wstrict-aliasing @gol
278 -Wstrict-aliasing=n @gol -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
279 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]} @gol
280 -Wsuggest-final-types @gol -Wsuggest-final-methods @gol -Wsuggest-override @gol
281 -Wmissing-format-attribute @gol
282 -Wswitch -Wswitch-default -Wswitch-enum -Wswitch-bool -Wsync-nand @gol
283 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
284 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
285 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
286 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
287 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
288 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
289 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
290 -Wvla -Wvolatile-register-var -Wwrite-strings @gol
291 -Wzero-as-null-pointer-constant}
293 @item C and Objective-C-only Warning Options
294 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
295 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
296 -Wold-style-declaration -Wold-style-definition @gol
297 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
298 -Wdeclaration-after-statement -Wpointer-sign}
300 @item Debugging Options
301 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
302 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
303 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
304 -fasan-shadow-offset=@var{number} -fsanitize-undefined-trap-on-error @gol
305 -fcheck-pointer-bounds -fchkp-check-incomplete-type @gol
306 -fchkp-first-field-has-own-bounds -fchkp-narrow-bounds @gol
307 -fchkp-narrow-to-innermost-array -fchkp-optimize @gol
308 -fchkp-use-fast-string-functions -fchkp-use-nochk-string-functions @gol
309 -fchkp-use-static-bounds -fchkp-use-static-const-bounds @gol
310 -fchkp-treat-zero-dynamic-size-as-infinite -fchkp-check-read @gol
311 -fchkp-check-read -fchkp-check-write -fchkp-store-bounds @gol
312 -fchkp-instrument-calls -fchkp-instrument-marked-only @gol
313 -fchkp-use-wrappers @gol
314 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
315 -fdisable-ipa-@var{pass_name} @gol
316 -fdisable-rtl-@var{pass_name} @gol
317 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
318 -fdisable-tree-@var{pass_name} @gol
319 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
320 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
321 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
322 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
323 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
325 -fdump-statistics @gol
327 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
328 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
329 -fdump-tree-cfg -fdump-tree-alias @gol
331 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
332 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
333 -fdump-tree-gimple@r{[}-raw@r{]} @gol
334 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
335 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
336 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
337 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
338 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
339 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
340 -fdump-tree-nrv -fdump-tree-vect @gol
341 -fdump-tree-sink @gol
342 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
343 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
344 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
345 -fdump-tree-vtable-verify @gol
346 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
347 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
348 -fdump-final-insns=@var{file} @gol
349 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
350 -feliminate-dwarf2-dups -fno-eliminate-unused-debug-types @gol
351 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
352 -fenable-@var{kind}-@var{pass} @gol
353 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
354 -fdebug-types-section -fmem-report-wpa @gol
355 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
357 -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
358 -frandom-seed=@var{number} -fsched-verbose=@var{n} @gol
359 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
360 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
361 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
362 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
363 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
364 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
365 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
366 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
367 -fdebug-prefix-map=@var{old}=@var{new} @gol
368 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
369 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
370 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
371 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
372 -print-prog-name=@var{program} -print-search-dirs -Q @gol
373 -print-sysroot -print-sysroot-headers-suffix @gol
374 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
376 @item Optimization Options
377 @xref{Optimize Options,,Options that Control Optimization}.
378 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
379 -falign-jumps[=@var{n}] @gol
380 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
381 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
382 -fauto-inc-dec -fbranch-probabilities @gol
383 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
384 -fbtr-bb-exclusive -fcaller-saves @gol
385 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
386 -fcompare-elim -fcprop-registers -fcrossjumping @gol
387 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
388 -fcx-limited-range @gol
389 -fdata-sections -fdce -fdelayed-branch @gol
390 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
391 -fdevirtualize-at-ltrans -fdse @gol
392 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
393 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
394 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
395 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
396 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
397 -fif-conversion2 -findirect-inlining @gol
398 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
399 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-cp-alignment @gol
400 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference -fipa-icf @gol
401 -fira-algorithm=@var{algorithm} @gol
402 -fira-region=@var{region} -fira-hoist-pressure @gol
403 -fira-loop-pressure -fno-ira-share-save-slots @gol
404 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
405 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
406 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
407 -flive-range-shrinkage @gol
408 -floop-block -floop-interchange -floop-strip-mine @gol
409 -floop-unroll-and-jam -floop-nest-optimize @gol
410 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
411 -flto-partition=@var{alg} -flto-report -flto-report-wpa -fmerge-all-constants @gol
412 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
413 -fmove-loop-invariants -fno-branch-count-reg @gol
414 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
415 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
416 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
417 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
418 -fomit-frame-pointer -foptimize-sibling-calls @gol
419 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
420 -fprefetch-loop-arrays -fprofile-report @gol
421 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
422 -fprofile-generate=@var{path} @gol
423 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
424 -fprofile-reorder-functions @gol
425 -freciprocal-math -free -frename-registers -freorder-blocks @gol
426 -freorder-blocks-and-partition -freorder-functions @gol
427 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
428 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
429 -fsched-spec-load -fsched-spec-load-dangerous @gol
430 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
431 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
432 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
433 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
434 -fschedule-fusion @gol
435 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
436 -fselective-scheduling -fselective-scheduling2 @gol
437 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
438 -fsemantic-interposition @gol
439 -fshrink-wrap -fsignaling-nans -fsingle-precision-constant @gol
440 -fsplit-ivs-in-unroller -fsplit-wide-types -fssa-phiopt @gol
441 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
442 -fstack-protector-explicit -fstdarg-opt -fstrict-aliasing @gol
443 -fstrict-overflow -fthread-jumps -ftracer -ftree-bit-ccp @gol
444 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
445 -ftree-coalesce-inline-vars -ftree-coalesce-vars -ftree-copy-prop @gol
446 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
447 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
448 -ftree-loop-if-convert-stores -ftree-loop-im @gol
449 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
450 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
451 -ftree-loop-vectorize @gol
452 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
453 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
454 -ftree-switch-conversion -ftree-tail-merge -ftree-ter @gol
455 -ftree-vectorize -ftree-vrp @gol
456 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
457 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
458 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
459 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
460 --param @var{name}=@var{value}
461 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
463 @item Preprocessor Options
464 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
465 @gccoptlist{-A@var{question}=@var{answer} @gol
466 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
467 -C -dD -dI -dM -dN @gol
468 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
469 -idirafter @var{dir} @gol
470 -include @var{file} -imacros @var{file} @gol
471 -iprefix @var{file} -iwithprefix @var{dir} @gol
472 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
473 -imultilib @var{dir} -isysroot @var{dir} @gol
474 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
475 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
476 -remap -trigraphs -undef -U@var{macro} @gol
477 -Wp,@var{option} -Xpreprocessor @var{option} -no-integrated-cpp}
479 @item Assembler Option
480 @xref{Assembler Options,,Passing Options to the Assembler}.
481 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
484 @xref{Link Options,,Options for Linking}.
485 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
486 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
487 -s -static -static-libgcc -static-libstdc++ @gol
488 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
489 -static-libmpx -static-libmpxwrappers @gol
490 -shared -shared-libgcc -symbolic @gol
491 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
492 -u @var{symbol} -z @var{keyword}}
494 @item Directory Options
495 @xref{Directory Options,,Options for Directory Search}.
496 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
497 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
498 --sysroot=@var{dir} --no-sysroot-suffix}
500 @item Machine Dependent Options
501 @xref{Submodel Options,,Hardware Models and Configurations}.
502 @c This list is ordered alphanumerically by subsection name.
503 @c Try and put the significant identifier (CPU or system) first,
504 @c so users have a clue at guessing where the ones they want will be.
506 @emph{AArch64 Options}
507 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
508 -mgeneral-regs-only @gol
509 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
511 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
512 -mtls-dialect=desc -mtls-dialect=traditional @gol
513 -mfix-cortex-a53-835769 -mno-fix-cortex-a53-835769 @gol
514 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
516 @emph{Adapteva Epiphany Options}
517 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
518 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
519 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
520 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
521 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
522 -msplit-vecmove-early -m1reg-@var{reg}}
525 @gccoptlist{-mbarrel-shifter @gol
526 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
527 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
528 -mea -mno-mpy -mmul32x16 -mmul64 @gol
529 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
530 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
531 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
532 -mepilogue-cfi -mlong-calls -mmedium-calls -msdata @gol
533 -mucb-mcount -mvolatile-cache @gol
534 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
535 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
536 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
537 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
538 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
539 -mtune=@var{cpu} -mmultcost=@var{num} -munalign-prob-threshold=@var{probability}}
542 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
543 -mabi=@var{name} @gol
544 -mapcs-stack-check -mno-apcs-stack-check @gol
545 -mapcs-float -mno-apcs-float @gol
546 -mapcs-reentrant -mno-apcs-reentrant @gol
547 -msched-prolog -mno-sched-prolog @gol
548 -mlittle-endian -mbig-endian @gol
549 -mfloat-abi=@var{name} @gol
550 -mfp16-format=@var{name}
551 -mthumb-interwork -mno-thumb-interwork @gol
552 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
553 -mstructure-size-boundary=@var{n} @gol
554 -mabort-on-noreturn @gol
555 -mlong-calls -mno-long-calls @gol
556 -msingle-pic-base -mno-single-pic-base @gol
557 -mpic-register=@var{reg} @gol
558 -mnop-fun-dllimport @gol
559 -mpoke-function-name @gol
561 -mtpcs-frame -mtpcs-leaf-frame @gol
562 -mcaller-super-interworking -mcallee-super-interworking @gol
563 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
564 -mword-relocations @gol
565 -mfix-cortex-m3-ldrd @gol
566 -munaligned-access @gol
567 -mneon-for-64bits @gol
568 -mslow-flash-data @gol
569 -masm-syntax-unified @gol
573 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
574 -mcall-prologues -mint8 -mn_flash=@var{size} -mno-interrupts @gol
575 -mrelax -mrmw -mstrict-X -mtiny-stack -Waddr-space-convert}
577 @emph{Blackfin Options}
578 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
579 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
580 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
581 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
582 -mno-id-shared-library -mshared-library-id=@var{n} @gol
583 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
584 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
585 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
589 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
590 -msim -msdata=@var{sdata-type}}
593 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
594 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
595 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
596 -mstack-align -mdata-align -mconst-align @gol
597 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
598 -melf -maout -melinux -mlinux -sim -sim2 @gol
599 -mmul-bug-workaround -mno-mul-bug-workaround}
602 @gccoptlist{-mmac @gol
603 -mcr16cplus -mcr16c @gol
604 -msim -mint32 -mbit-ops
605 -mdata-model=@var{model}}
607 @emph{Darwin Options}
608 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
609 -arch_only -bind_at_load -bundle -bundle_loader @gol
610 -client_name -compatibility_version -current_version @gol
612 -dependency-file -dylib_file -dylinker_install_name @gol
613 -dynamic -dynamiclib -exported_symbols_list @gol
614 -filelist -flat_namespace -force_cpusubtype_ALL @gol
615 -force_flat_namespace -headerpad_max_install_names @gol
617 -image_base -init -install_name -keep_private_externs @gol
618 -multi_module -multiply_defined -multiply_defined_unused @gol
619 -noall_load -no_dead_strip_inits_and_terms @gol
620 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
621 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
622 -private_bundle -read_only_relocs -sectalign @gol
623 -sectobjectsymbols -whyload -seg1addr @gol
624 -sectcreate -sectobjectsymbols -sectorder @gol
625 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
626 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
627 -segprot -segs_read_only_addr -segs_read_write_addr @gol
628 -single_module -static -sub_library -sub_umbrella @gol
629 -twolevel_namespace -umbrella -undefined @gol
630 -unexported_symbols_list -weak_reference_mismatches @gol
631 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
632 -mkernel -mone-byte-bool}
634 @emph{DEC Alpha Options}
635 @gccoptlist{-mno-fp-regs -msoft-float @gol
636 -mieee -mieee-with-inexact -mieee-conformant @gol
637 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
638 -mtrap-precision=@var{mode} -mbuild-constants @gol
639 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
640 -mbwx -mmax -mfix -mcix @gol
641 -mfloat-vax -mfloat-ieee @gol
642 -mexplicit-relocs -msmall-data -mlarge-data @gol
643 -msmall-text -mlarge-text @gol
644 -mmemory-latency=@var{time}}
647 @gccoptlist{-msmall-model -mno-lsim}
650 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
651 -mhard-float -msoft-float @gol
652 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
653 -mdouble -mno-double @gol
654 -mmedia -mno-media -mmuladd -mno-muladd @gol
655 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
656 -mlinked-fp -mlong-calls -malign-labels @gol
657 -mlibrary-pic -macc-4 -macc-8 @gol
658 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
659 -moptimize-membar -mno-optimize-membar @gol
660 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
661 -mvliw-branch -mno-vliw-branch @gol
662 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
663 -mno-nested-cond-exec -mtomcat-stats @gol
667 @emph{GNU/Linux Options}
668 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
669 -tno-android-cc -tno-android-ld}
671 @emph{H8/300 Options}
672 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
675 @gccoptlist{-march=@var{architecture-type} @gol
676 -mdisable-fpregs -mdisable-indexing @gol
677 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
678 -mfixed-range=@var{register-range} @gol
679 -mjump-in-delay -mlinker-opt -mlong-calls @gol
680 -mlong-load-store -mno-disable-fpregs @gol
681 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
682 -mno-jump-in-delay -mno-long-load-store @gol
683 -mno-portable-runtime -mno-soft-float @gol
684 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
685 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
686 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
687 -munix=@var{unix-std} -nolibdld -static -threads}
690 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
691 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
692 -mconstant-gp -mauto-pic -mfused-madd @gol
693 -minline-float-divide-min-latency @gol
694 -minline-float-divide-max-throughput @gol
695 -mno-inline-float-divide @gol
696 -minline-int-divide-min-latency @gol
697 -minline-int-divide-max-throughput @gol
698 -mno-inline-int-divide @gol
699 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
700 -mno-inline-sqrt @gol
701 -mdwarf2-asm -mearly-stop-bits @gol
702 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
703 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
704 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
705 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
706 -msched-spec-ldc -msched-spec-control-ldc @gol
707 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
708 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
709 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
710 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
713 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
714 -msign-extend-enabled -muser-enabled}
716 @emph{M32R/D Options}
717 @gccoptlist{-m32r2 -m32rx -m32r @gol
719 -malign-loops -mno-align-loops @gol
720 -missue-rate=@var{number} @gol
721 -mbranch-cost=@var{number} @gol
722 -mmodel=@var{code-size-model-type} @gol
723 -msdata=@var{sdata-type} @gol
724 -mno-flush-func -mflush-func=@var{name} @gol
725 -mno-flush-trap -mflush-trap=@var{number} @gol
729 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
731 @emph{M680x0 Options}
732 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
733 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
734 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
735 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
736 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
737 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
738 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
739 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
743 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
744 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
745 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
746 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
747 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
750 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
751 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
752 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
753 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
756 @emph{MicroBlaze Options}
757 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
758 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
759 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
760 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
761 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
764 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
765 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
766 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
767 -mips16 -mno-mips16 -mflip-mips16 @gol
768 -minterlink-compressed -mno-interlink-compressed @gol
769 -minterlink-mips16 -mno-interlink-mips16 @gol
770 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
771 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
772 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
773 -mno-float -msingle-float -mdouble-float @gol
774 -modd-spreg -mno-odd-spreg @gol
775 -mabs=@var{mode} -mnan=@var{encoding} @gol
776 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
779 -mvirt -mno-virt @gol
781 -mmicromips -mno-micromips @gol
782 -mfpu=@var{fpu-type} @gol
783 -msmartmips -mno-smartmips @gol
784 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
785 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
786 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
787 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
788 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
789 -membedded-data -mno-embedded-data @gol
790 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
791 -mcode-readable=@var{setting} @gol
792 -msplit-addresses -mno-split-addresses @gol
793 -mexplicit-relocs -mno-explicit-relocs @gol
794 -mcheck-zero-division -mno-check-zero-division @gol
795 -mdivide-traps -mdivide-breaks @gol
796 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
797 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
798 -mfix-24k -mno-fix-24k @gol
799 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
800 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
801 -mfix-vr4120 -mno-fix-vr4120 @gol
802 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
803 -mflush-func=@var{func} -mno-flush-func @gol
804 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
805 -mfp-exceptions -mno-fp-exceptions @gol
806 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
807 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
810 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
811 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
812 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
813 -mno-base-addresses -msingle-exit -mno-single-exit}
815 @emph{MN10300 Options}
816 @gccoptlist{-mmult-bug -mno-mult-bug @gol
817 -mno-am33 -mam33 -mam33-2 -mam34 @gol
818 -mtune=@var{cpu-type} @gol
819 -mreturn-pointer-on-d0 @gol
820 -mno-crt0 -mrelax -mliw -msetlb}
823 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
825 @emph{MSP430 Options}
826 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
830 @gccoptlist{-mbig-endian -mlittle-endian @gol
831 -mreduced-regs -mfull-regs @gol
832 -mcmov -mno-cmov @gol
833 -mperf-ext -mno-perf-ext @gol
834 -mv3push -mno-v3push @gol
835 -m16bit -mno-16bit @gol
836 -misr-vector-size=@var{num} @gol
837 -mcache-block-size=@var{num} @gol
838 -march=@var{arch} @gol
839 -mcmodel=@var{code-model} @gol
842 @emph{Nios II Options}
843 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
845 -mno-bypass-cache -mbypass-cache @gol
846 -mno-cache-volatile -mcache-volatile @gol
847 -mno-fast-sw-div -mfast-sw-div @gol
848 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
849 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
850 -mcustom-fpu-cfg=@var{name} @gol
851 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name}}
853 @emph{Nvidia PTX Options}
854 @gccoptlist{-m32 -m64 -mmainkernel}
856 @emph{PDP-11 Options}
857 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
858 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
859 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
860 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
861 -mbranch-expensive -mbranch-cheap @gol
862 -munix-asm -mdec-asm}
864 @emph{picoChip Options}
865 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
866 -msymbol-as-address -mno-inefficient-warnings}
868 @emph{PowerPC Options}
869 See RS/6000 and PowerPC Options.
872 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=rl78 @gol
873 -m64bit-doubles -m32bit-doubles}
875 @emph{RS/6000 and PowerPC Options}
876 @gccoptlist{-mcpu=@var{cpu-type} @gol
877 -mtune=@var{cpu-type} @gol
878 -mcmodel=@var{code-model} @gol
880 -maltivec -mno-altivec @gol
881 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
882 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
883 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
884 -mfprnd -mno-fprnd @gol
885 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
886 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
887 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
888 -malign-power -malign-natural @gol
889 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
890 -msingle-float -mdouble-float -msimple-fpu @gol
891 -mstring -mno-string -mupdate -mno-update @gol
892 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
893 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
894 -mstrict-align -mno-strict-align -mrelocatable @gol
895 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
896 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
897 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
898 -mprioritize-restricted-insns=@var{priority} @gol
899 -msched-costly-dep=@var{dependence_type} @gol
900 -minsert-sched-nops=@var{scheme} @gol
901 -mcall-sysv -mcall-netbsd @gol
902 -maix-struct-return -msvr4-struct-return @gol
903 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
904 -mblock-move-inline-limit=@var{num} @gol
905 -misel -mno-isel @gol
906 -misel=yes -misel=no @gol
908 -mspe=yes -mspe=no @gol
910 -mgen-cell-microcode -mwarn-cell-microcode @gol
911 -mvrsave -mno-vrsave @gol
912 -mmulhw -mno-mulhw @gol
913 -mdlmzb -mno-dlmzb @gol
914 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
915 -mprototype -mno-prototype @gol
916 -msim -mmvme -mads -myellowknife -memb -msdata @gol
917 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
918 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
919 -mno-recip-precision @gol
920 -mveclibabi=@var{type} -mfriz -mno-friz @gol
921 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
922 -msave-toc-indirect -mno-save-toc-indirect @gol
923 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
924 -mcrypto -mno-crypto -mdirect-move -mno-direct-move @gol
925 -mquad-memory -mno-quad-memory @gol
926 -mquad-memory-atomic -mno-quad-memory-atomic @gol
927 -mcompat-align-parm -mno-compat-align-parm @gol
928 -mupper-regs-df -mno-upper-regs-df -mupper-regs-sf -mno-upper-regs-sf @gol
929 -mupper-regs -mno-upper-regs}
932 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
934 -mbig-endian-data -mlittle-endian-data @gol
937 -mas100-syntax -mno-as100-syntax@gol
939 -mmax-constant-size=@gol
942 -mno-warn-multiple-fast-interrupts@gol
943 -msave-acc-in-interrupts}
945 @emph{S/390 and zSeries Options}
946 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
947 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
948 -mlong-double-64 -mlong-double-128 @gol
949 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
950 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
951 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
952 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
953 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
954 -mhotpatch=@var{halfwords},@var{halfwords}}
957 @gccoptlist{-meb -mel @gol
961 -mscore5 -mscore5u -mscore7 -mscore7d}
964 @gccoptlist{-m1 -m2 -m2e @gol
965 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
967 -m4-nofpu -m4-single-only -m4-single -m4 @gol
968 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
969 -m5-64media -m5-64media-nofpu @gol
970 -m5-32media -m5-32media-nofpu @gol
971 -m5-compact -m5-compact-nofpu @gol
972 -mb -ml -mdalign -mrelax @gol
973 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
974 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
975 -mspace -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
976 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
977 -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
978 -maccumulate-outgoing-args -minvalid-symbols @gol
979 -matomic-model=@var{atomic-model} @gol
980 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
981 -mcbranch-force-delay-slot @gol
982 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
983 -mpretend-cmove -mtas}
985 @emph{Solaris 2 Options}
986 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
990 @gccoptlist{-mcpu=@var{cpu-type} @gol
991 -mtune=@var{cpu-type} @gol
992 -mcmodel=@var{code-model} @gol
993 -mmemory-model=@var{mem-model} @gol
994 -m32 -m64 -mapp-regs -mno-app-regs @gol
995 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
996 -mfpu -mno-fpu -mhard-float -msoft-float @gol
997 -mhard-quad-float -msoft-quad-float @gol
998 -mstack-bias -mno-stack-bias @gol
999 -munaligned-doubles -mno-unaligned-doubles @gol
1000 -muser-mode -mno-user-mode @gol
1001 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1002 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1003 -mcbcond -mno-cbcond @gol
1004 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
1005 -mfix-at697f -mfix-ut699}
1008 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1009 -msafe-dma -munsafe-dma @gol
1011 -msmall-mem -mlarge-mem -mstdmain @gol
1012 -mfixed-range=@var{register-range} @gol
1014 -maddress-space-conversion -mno-address-space-conversion @gol
1015 -mcache-size=@var{cache-size} @gol
1016 -matomic-updates -mno-atomic-updates}
1018 @emph{System V Options}
1019 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1021 @emph{TILE-Gx Options}
1022 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1023 -mcmodel=@var{code-model}}
1025 @emph{TILEPro Options}
1026 @gccoptlist{-mcpu=@var{cpu} -m32}
1029 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1030 -mprolog-function -mno-prolog-function -mspace @gol
1031 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1032 -mapp-regs -mno-app-regs @gol
1033 -mdisable-callt -mno-disable-callt @gol
1034 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1035 -mv850e -mv850 -mv850e3v5 @gol
1046 @gccoptlist{-mg -mgnu -munix}
1048 @emph{Visium Options}
1049 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1050 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1053 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1054 -mpointer-size=@var{size}}
1056 @emph{VxWorks Options}
1057 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1058 -Xbind-lazy -Xbind-now}
1061 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1062 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1063 -mfpmath=@var{unit} @gol
1064 -masm=@var{dialect} -mno-fancy-math-387 @gol
1065 -mno-fp-ret-in-387 -msoft-float @gol
1066 -mno-wide-multiply -mrtd -malign-double @gol
1067 -mpreferred-stack-boundary=@var{num} @gol
1068 -mincoming-stack-boundary=@var{num} @gol
1069 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1070 -mrecip -mrecip=@var{opt} @gol
1071 -mvzeroupper -mprefer-avx128 @gol
1072 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1073 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -msha @gol
1074 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma -mprefetchwt1 @gol
1075 -mclflushopt -mxsavec -mxsaves @gol
1076 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
1077 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx -mthreads @gol
1078 -mno-align-stringops -minline-all-stringops @gol
1079 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1080 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1081 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1082 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1083 -mregparm=@var{num} -msseregparm @gol
1084 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1085 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1086 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1087 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1088 -m32 -m64 -mx32 -m16 -mlarge-data-threshold=@var{num} @gol
1089 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1090 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1091 -malign-data=@var{type} -mstack-protector-guard=@var{guard}}
1093 @emph{x86 Windows Options}
1094 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1095 -mnop-fun-dllimport -mthread @gol
1096 -municode -mwin32 -mwindows -fno-set-stack-executable}
1098 @emph{Xstormy16 Options}
1101 @emph{Xtensa Options}
1102 @gccoptlist{-mconst16 -mno-const16 @gol
1103 -mfused-madd -mno-fused-madd @gol
1105 -mserialize-volatile -mno-serialize-volatile @gol
1106 -mtext-section-literals -mno-text-section-literals @gol
1107 -mtarget-align -mno-target-align @gol
1108 -mlongcalls -mno-longcalls}
1110 @emph{zSeries Options}
1111 See S/390 and zSeries Options.
1113 @item Code Generation Options
1114 @xref{Code Gen Options,,Options for Code Generation Conventions}.
1115 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
1116 -ffixed-@var{reg} -fexceptions @gol
1117 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
1118 -fasynchronous-unwind-tables @gol
1119 -fno-gnu-unique @gol
1120 -finhibit-size-directive -finstrument-functions @gol
1121 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
1122 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
1123 -fno-common -fno-ident @gol
1124 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
1125 -fno-jump-tables @gol
1126 -frecord-gcc-switches @gol
1127 -freg-struct-return -fshort-enums @gol
1128 -fshort-double -fshort-wchar @gol
1129 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
1130 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
1131 -fno-stack-limit -fsplit-stack @gol
1132 -fleading-underscore -ftls-model=@var{model} @gol
1133 -fstack-reuse=@var{reuse_level} @gol
1134 -ftrapv -fwrapv -fbounds-check @gol
1135 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
1136 -fstrict-volatile-bitfields -fsync-libcalls}
1140 @node Overall Options
1141 @section Options Controlling the Kind of Output
1143 Compilation can involve up to four stages: preprocessing, compilation
1144 proper, assembly and linking, always in that order. GCC is capable of
1145 preprocessing and compiling several files either into several
1146 assembler input files, or into one assembler input file; then each
1147 assembler input file produces an object file, and linking combines all
1148 the object files (those newly compiled, and those specified as input)
1149 into an executable file.
1151 @cindex file name suffix
1152 For any given input file, the file name suffix determines what kind of
1153 compilation is done:
1157 C source code that must be preprocessed.
1160 C source code that should not be preprocessed.
1163 C++ source code that should not be preprocessed.
1166 Objective-C source code. Note that you must link with the @file{libobjc}
1167 library to make an Objective-C program work.
1170 Objective-C source code that should not be preprocessed.
1174 Objective-C++ source code. Note that you must link with the @file{libobjc}
1175 library to make an Objective-C++ program work. Note that @samp{.M} refers
1176 to a literal capital M@.
1178 @item @var{file}.mii
1179 Objective-C++ source code that should not be preprocessed.
1182 C, C++, Objective-C or Objective-C++ header file to be turned into a
1183 precompiled header (default), or C, C++ header file to be turned into an
1184 Ada spec (via the @option{-fdump-ada-spec} switch).
1187 @itemx @var{file}.cp
1188 @itemx @var{file}.cxx
1189 @itemx @var{file}.cpp
1190 @itemx @var{file}.CPP
1191 @itemx @var{file}.c++
1193 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1194 the last two letters must both be literally @samp{x}. Likewise,
1195 @samp{.C} refers to a literal capital C@.
1199 Objective-C++ source code that must be preprocessed.
1201 @item @var{file}.mii
1202 Objective-C++ source code that should not be preprocessed.
1206 @itemx @var{file}.hp
1207 @itemx @var{file}.hxx
1208 @itemx @var{file}.hpp
1209 @itemx @var{file}.HPP
1210 @itemx @var{file}.h++
1211 @itemx @var{file}.tcc
1212 C++ header file to be turned into a precompiled header or Ada spec.
1215 @itemx @var{file}.for
1216 @itemx @var{file}.ftn
1217 Fixed form Fortran source code that should not be preprocessed.
1220 @itemx @var{file}.FOR
1221 @itemx @var{file}.fpp
1222 @itemx @var{file}.FPP
1223 @itemx @var{file}.FTN
1224 Fixed form Fortran source code that must be preprocessed (with the traditional
1227 @item @var{file}.f90
1228 @itemx @var{file}.f95
1229 @itemx @var{file}.f03
1230 @itemx @var{file}.f08
1231 Free form Fortran source code that should not be preprocessed.
1233 @item @var{file}.F90
1234 @itemx @var{file}.F95
1235 @itemx @var{file}.F03
1236 @itemx @var{file}.F08
1237 Free form Fortran source code that must be preprocessed (with the
1238 traditional preprocessor).
1243 @c FIXME: Descriptions of Java file types.
1249 @item @var{file}.ads
1250 Ada source code file that contains a library unit declaration (a
1251 declaration of a package, subprogram, or generic, or a generic
1252 instantiation), or a library unit renaming declaration (a package,
1253 generic, or subprogram renaming declaration). Such files are also
1256 @item @var{file}.adb
1257 Ada source code file containing a library unit body (a subprogram or
1258 package body). Such files are also called @dfn{bodies}.
1260 @c GCC also knows about some suffixes for languages not yet included:
1271 @itemx @var{file}.sx
1272 Assembler code that must be preprocessed.
1275 An object file to be fed straight into linking.
1276 Any file name with no recognized suffix is treated this way.
1280 You can specify the input language explicitly with the @option{-x} option:
1283 @item -x @var{language}
1284 Specify explicitly the @var{language} for the following input files
1285 (rather than letting the compiler choose a default based on the file
1286 name suffix). This option applies to all following input files until
1287 the next @option{-x} option. Possible values for @var{language} are:
1289 c c-header cpp-output
1290 c++ c++-header c++-cpp-output
1291 objective-c objective-c-header objective-c-cpp-output
1292 objective-c++ objective-c++-header objective-c++-cpp-output
1293 assembler assembler-with-cpp
1295 f77 f77-cpp-input f95 f95-cpp-input
1301 Turn off any specification of a language, so that subsequent files are
1302 handled according to their file name suffixes (as they are if @option{-x}
1303 has not been used at all).
1305 @item -pass-exit-codes
1306 @opindex pass-exit-codes
1307 Normally the @command{gcc} program exits with the code of 1 if any
1308 phase of the compiler returns a non-success return code. If you specify
1309 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1310 the numerically highest error produced by any phase returning an error
1311 indication. The C, C++, and Fortran front ends return 4 if an internal
1312 compiler error is encountered.
1315 If you only want some of the stages of compilation, you can use
1316 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1317 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1318 @command{gcc} is to stop. Note that some combinations (for example,
1319 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1324 Compile or assemble the source files, but do not link. The linking
1325 stage simply is not done. The ultimate output is in the form of an
1326 object file for each source file.
1328 By default, the object file name for a source file is made by replacing
1329 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1331 Unrecognized input files, not requiring compilation or assembly, are
1336 Stop after the stage of compilation proper; do not assemble. The output
1337 is in the form of an assembler code file for each non-assembler input
1340 By default, the assembler file name for a source file is made by
1341 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1343 Input files that don't require compilation are ignored.
1347 Stop after the preprocessing stage; do not run the compiler proper. The
1348 output is in the form of preprocessed source code, which is sent to the
1351 Input files that don't require preprocessing are ignored.
1353 @cindex output file option
1356 Place output in file @var{file}. This applies to whatever
1357 sort of output is being produced, whether it be an executable file,
1358 an object file, an assembler file or preprocessed C code.
1360 If @option{-o} is not specified, the default is to put an executable
1361 file in @file{a.out}, the object file for
1362 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1363 assembler file in @file{@var{source}.s}, a precompiled header file in
1364 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1369 Print (on standard error output) the commands executed to run the stages
1370 of compilation. Also print the version number of the compiler driver
1371 program and of the preprocessor and the compiler proper.
1375 Like @option{-v} except the commands are not executed and arguments
1376 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1377 This is useful for shell scripts to capture the driver-generated command lines.
1381 Use pipes rather than temporary files for communication between the
1382 various stages of compilation. This fails to work on some systems where
1383 the assembler is unable to read from a pipe; but the GNU assembler has
1388 Print (on the standard output) a description of the command-line options
1389 understood by @command{gcc}. If the @option{-v} option is also specified
1390 then @option{--help} is also passed on to the various processes
1391 invoked by @command{gcc}, so that they can display the command-line options
1392 they accept. If the @option{-Wextra} option has also been specified
1393 (prior to the @option{--help} option), then command-line options that
1394 have no documentation associated with them are also displayed.
1397 @opindex target-help
1398 Print (on the standard output) a description of target-specific command-line
1399 options for each tool. For some targets extra target-specific
1400 information may also be printed.
1402 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1403 Print (on the standard output) a description of the command-line
1404 options understood by the compiler that fit into all specified classes
1405 and qualifiers. These are the supported classes:
1408 @item @samp{optimizers}
1409 Display all of the optimization options supported by the
1412 @item @samp{warnings}
1413 Display all of the options controlling warning messages
1414 produced by the compiler.
1417 Display target-specific options. Unlike the
1418 @option{--target-help} option however, target-specific options of the
1419 linker and assembler are not displayed. This is because those
1420 tools do not currently support the extended @option{--help=} syntax.
1423 Display the values recognized by the @option{--param}
1426 @item @var{language}
1427 Display the options supported for @var{language}, where
1428 @var{language} is the name of one of the languages supported in this
1432 Display the options that are common to all languages.
1435 These are the supported qualifiers:
1438 @item @samp{undocumented}
1439 Display only those options that are undocumented.
1442 Display options taking an argument that appears after an equal
1443 sign in the same continuous piece of text, such as:
1444 @samp{--help=target}.
1446 @item @samp{separate}
1447 Display options taking an argument that appears as a separate word
1448 following the original option, such as: @samp{-o output-file}.
1451 Thus for example to display all the undocumented target-specific
1452 switches supported by the compiler, use:
1455 --help=target,undocumented
1458 The sense of a qualifier can be inverted by prefixing it with the
1459 @samp{^} character, so for example to display all binary warning
1460 options (i.e., ones that are either on or off and that do not take an
1461 argument) that have a description, use:
1464 --help=warnings,^joined,^undocumented
1467 The argument to @option{--help=} should not consist solely of inverted
1470 Combining several classes is possible, although this usually
1471 restricts the output so much that there is nothing to display. One
1472 case where it does work, however, is when one of the classes is
1473 @var{target}. For example, to display all the target-specific
1474 optimization options, use:
1477 --help=target,optimizers
1480 The @option{--help=} option can be repeated on the command line. Each
1481 successive use displays its requested class of options, skipping
1482 those that have already been displayed.
1484 If the @option{-Q} option appears on the command line before the
1485 @option{--help=} option, then the descriptive text displayed by
1486 @option{--help=} is changed. Instead of describing the displayed
1487 options, an indication is given as to whether the option is enabled,
1488 disabled or set to a specific value (assuming that the compiler
1489 knows this at the point where the @option{--help=} option is used).
1491 Here is a truncated example from the ARM port of @command{gcc}:
1494 % gcc -Q -mabi=2 --help=target -c
1495 The following options are target specific:
1497 -mabort-on-noreturn [disabled]
1501 The output is sensitive to the effects of previous command-line
1502 options, so for example it is possible to find out which optimizations
1503 are enabled at @option{-O2} by using:
1506 -Q -O2 --help=optimizers
1509 Alternatively you can discover which binary optimizations are enabled
1510 by @option{-O3} by using:
1513 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1514 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1515 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1518 @item -no-canonical-prefixes
1519 @opindex no-canonical-prefixes
1520 Do not expand any symbolic links, resolve references to @samp{/../}
1521 or @samp{/./}, or make the path absolute when generating a relative
1526 Display the version number and copyrights of the invoked GCC@.
1530 Invoke all subcommands under a wrapper program. The name of the
1531 wrapper program and its parameters are passed as a comma separated
1535 gcc -c t.c -wrapper gdb,--args
1539 This invokes all subprograms of @command{gcc} under
1540 @samp{gdb --args}, thus the invocation of @command{cc1} is
1541 @samp{gdb --args cc1 @dots{}}.
1543 @item -fplugin=@var{name}.so
1545 Load the plugin code in file @var{name}.so, assumed to be a
1546 shared object to be dlopen'd by the compiler. The base name of
1547 the shared object file is used to identify the plugin for the
1548 purposes of argument parsing (See
1549 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1550 Each plugin should define the callback functions specified in the
1553 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1554 @opindex fplugin-arg
1555 Define an argument called @var{key} with a value of @var{value}
1556 for the plugin called @var{name}.
1558 @item -fdump-ada-spec@r{[}-slim@r{]}
1559 @opindex fdump-ada-spec
1560 For C and C++ source and include files, generate corresponding Ada specs.
1561 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1562 GNAT User's Guide}, which provides detailed documentation on this feature.
1564 @item -fada-spec-parent=@var{unit}
1565 @opindex fada-spec-parent
1566 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1567 Ada specs as child units of parent @var{unit}.
1569 @item -fdump-go-spec=@var{file}
1570 @opindex fdump-go-spec
1571 For input files in any language, generate corresponding Go
1572 declarations in @var{file}. This generates Go @code{const},
1573 @code{type}, @code{var}, and @code{func} declarations which may be a
1574 useful way to start writing a Go interface to code written in some
1577 @include @value{srcdir}/../libiberty/at-file.texi
1581 @section Compiling C++ Programs
1583 @cindex suffixes for C++ source
1584 @cindex C++ source file suffixes
1585 C++ source files conventionally use one of the suffixes @samp{.C},
1586 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1587 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1588 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1589 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1590 files with these names and compiles them as C++ programs even if you
1591 call the compiler the same way as for compiling C programs (usually
1592 with the name @command{gcc}).
1596 However, the use of @command{gcc} does not add the C++ library.
1597 @command{g++} is a program that calls GCC and automatically specifies linking
1598 against the C++ library. It treats @samp{.c},
1599 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1600 files unless @option{-x} is used. This program is also useful when
1601 precompiling a C header file with a @samp{.h} extension for use in C++
1602 compilations. On many systems, @command{g++} is also installed with
1603 the name @command{c++}.
1605 @cindex invoking @command{g++}
1606 When you compile C++ programs, you may specify many of the same
1607 command-line options that you use for compiling programs in any
1608 language; or command-line options meaningful for C and related
1609 languages; or options that are meaningful only for C++ programs.
1610 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1611 explanations of options for languages related to C@.
1612 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1613 explanations of options that are meaningful only for C++ programs.
1615 @node C Dialect Options
1616 @section Options Controlling C Dialect
1617 @cindex dialect options
1618 @cindex language dialect options
1619 @cindex options, dialect
1621 The following options control the dialect of C (or languages derived
1622 from C, such as C++, Objective-C and Objective-C++) that the compiler
1626 @cindex ANSI support
1630 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1631 equivalent to @option{-std=c++98}.
1633 This turns off certain features of GCC that are incompatible with ISO
1634 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1635 such as the @code{asm} and @code{typeof} keywords, and
1636 predefined macros such as @code{unix} and @code{vax} that identify the
1637 type of system you are using. It also enables the undesirable and
1638 rarely used ISO trigraph feature. For the C compiler,
1639 it disables recognition of C++ style @samp{//} comments as well as
1640 the @code{inline} keyword.
1642 The alternate keywords @code{__asm__}, @code{__extension__},
1643 @code{__inline__} and @code{__typeof__} continue to work despite
1644 @option{-ansi}. You would not want to use them in an ISO C program, of
1645 course, but it is useful to put them in header files that might be included
1646 in compilations done with @option{-ansi}. Alternate predefined macros
1647 such as @code{__unix__} and @code{__vax__} are also available, with or
1648 without @option{-ansi}.
1650 The @option{-ansi} option does not cause non-ISO programs to be
1651 rejected gratuitously. For that, @option{-Wpedantic} is required in
1652 addition to @option{-ansi}. @xref{Warning Options}.
1654 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1655 option is used. Some header files may notice this macro and refrain
1656 from declaring certain functions or defining certain macros that the
1657 ISO standard doesn't call for; this is to avoid interfering with any
1658 programs that might use these names for other things.
1660 Functions that are normally built in but do not have semantics
1661 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1662 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1663 built-in functions provided by GCC}, for details of the functions
1668 Determine the language standard. @xref{Standards,,Language Standards
1669 Supported by GCC}, for details of these standard versions. This option
1670 is currently only supported when compiling C or C++.
1672 The compiler can accept several base standards, such as @samp{c90} or
1673 @samp{c++98}, and GNU dialects of those standards, such as
1674 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1675 compiler accepts all programs following that standard plus those
1676 using GNU extensions that do not contradict it. For example,
1677 @option{-std=c90} turns off certain features of GCC that are
1678 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1679 keywords, but not other GNU extensions that do not have a meaning in
1680 ISO C90, such as omitting the middle term of a @code{?:}
1681 expression. On the other hand, when a GNU dialect of a standard is
1682 specified, all features supported by the compiler are enabled, even when
1683 those features change the meaning of the base standard. As a result, some
1684 strict-conforming programs may be rejected. The particular standard
1685 is used by @option{-Wpedantic} to identify which features are GNU
1686 extensions given that version of the standard. For example
1687 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1688 comments, while @option{-std=gnu99 -Wpedantic} does not.
1690 A value for this option must be provided; possible values are
1696 Support all ISO C90 programs (certain GNU extensions that conflict
1697 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1699 @item iso9899:199409
1700 ISO C90 as modified in amendment 1.
1706 ISO C99. This standard is substantially completely supported, modulo
1707 bugs and floating-point issues
1708 (mainly but not entirely relating to optional C99 features from
1709 Annexes F and G). See
1710 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1711 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1716 ISO C11, the 2011 revision of the ISO C standard. This standard is
1717 substantially completely supported, modulo bugs, floating-point issues
1718 (mainly but not entirely relating to optional C11 features from
1719 Annexes F and G) and the optional Annexes K (Bounds-checking
1720 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1724 GNU dialect of ISO C90 (including some C99 features).
1728 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1732 GNU dialect of ISO C11. This is the default for C code.
1733 The name @samp{gnu1x} is deprecated.
1737 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1738 additional defect reports. Same as @option{-ansi} for C++ code.
1742 GNU dialect of @option{-std=c++98}. This is the default for
1747 The 2011 ISO C++ standard plus amendments.
1748 The name @samp{c++0x} is deprecated.
1752 GNU dialect of @option{-std=c++11}.
1753 The name @samp{gnu++0x} is deprecated.
1757 The 2014 ISO C++ standard plus amendments.
1758 The name @samp{c++1y} is deprecated.
1762 GNU dialect of @option{-std=c++14}.
1763 The name @samp{gnu++1y} is deprecated.
1766 The next revision of the ISO C++ standard, tentatively planned for
1767 2017. Support is highly experimental, and will almost certainly
1768 change in incompatible ways in future releases.
1771 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1772 and will almost certainly change in incompatible ways in future
1776 @item -fgnu89-inline
1777 @opindex fgnu89-inline
1778 The option @option{-fgnu89-inline} tells GCC to use the traditional
1779 GNU semantics for @code{inline} functions when in C99 mode.
1780 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1781 Using this option is roughly equivalent to adding the
1782 @code{gnu_inline} function attribute to all inline functions
1783 (@pxref{Function Attributes}).
1785 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1786 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1787 specifies the default behavior).
1788 This option is not supported in @option{-std=c90} or
1789 @option{-std=gnu90} mode.
1791 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1792 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1793 in effect for @code{inline} functions. @xref{Common Predefined
1794 Macros,,,cpp,The C Preprocessor}.
1796 @item -aux-info @var{filename}
1798 Output to the given filename prototyped declarations for all functions
1799 declared and/or defined in a translation unit, including those in header
1800 files. This option is silently ignored in any language other than C@.
1802 Besides declarations, the file indicates, in comments, the origin of
1803 each declaration (source file and line), whether the declaration was
1804 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1805 @samp{O} for old, respectively, in the first character after the line
1806 number and the colon), and whether it came from a declaration or a
1807 definition (@samp{C} or @samp{F}, respectively, in the following
1808 character). In the case of function definitions, a K&R-style list of
1809 arguments followed by their declarations is also provided, inside
1810 comments, after the declaration.
1812 @item -fallow-parameterless-variadic-functions
1813 @opindex fallow-parameterless-variadic-functions
1814 Accept variadic functions without named parameters.
1816 Although it is possible to define such a function, this is not very
1817 useful as it is not possible to read the arguments. This is only
1818 supported for C as this construct is allowed by C++.
1822 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1823 keyword, so that code can use these words as identifiers. You can use
1824 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1825 instead. @option{-ansi} implies @option{-fno-asm}.
1827 In C++, this switch only affects the @code{typeof} keyword, since
1828 @code{asm} and @code{inline} are standard keywords. You may want to
1829 use the @option{-fno-gnu-keywords} flag instead, which has the same
1830 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1831 switch only affects the @code{asm} and @code{typeof} keywords, since
1832 @code{inline} is a standard keyword in ISO C99.
1835 @itemx -fno-builtin-@var{function}
1836 @opindex fno-builtin
1837 @cindex built-in functions
1838 Don't recognize built-in functions that do not begin with
1839 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1840 functions provided by GCC}, for details of the functions affected,
1841 including those which are not built-in functions when @option{-ansi} or
1842 @option{-std} options for strict ISO C conformance are used because they
1843 do not have an ISO standard meaning.
1845 GCC normally generates special code to handle certain built-in functions
1846 more efficiently; for instance, calls to @code{alloca} may become single
1847 instructions which adjust the stack directly, and calls to @code{memcpy}
1848 may become inline copy loops. The resulting code is often both smaller
1849 and faster, but since the function calls no longer appear as such, you
1850 cannot set a breakpoint on those calls, nor can you change the behavior
1851 of the functions by linking with a different library. In addition,
1852 when a function is recognized as a built-in function, GCC may use
1853 information about that function to warn about problems with calls to
1854 that function, or to generate more efficient code, even if the
1855 resulting code still contains calls to that function. For example,
1856 warnings are given with @option{-Wformat} for bad calls to
1857 @code{printf} when @code{printf} is built in and @code{strlen} is
1858 known not to modify global memory.
1860 With the @option{-fno-builtin-@var{function}} option
1861 only the built-in function @var{function} is
1862 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1863 function is named that is not built-in in this version of GCC, this
1864 option is ignored. There is no corresponding
1865 @option{-fbuiltin-@var{function}} option; if you wish to enable
1866 built-in functions selectively when using @option{-fno-builtin} or
1867 @option{-ffreestanding}, you may define macros such as:
1870 #define abs(n) __builtin_abs ((n))
1871 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1876 @cindex hosted environment
1878 Assert that compilation targets a hosted environment. This implies
1879 @option{-fbuiltin}. A hosted environment is one in which the
1880 entire standard library is available, and in which @code{main} has a return
1881 type of @code{int}. Examples are nearly everything except a kernel.
1882 This is equivalent to @option{-fno-freestanding}.
1884 @item -ffreestanding
1885 @opindex ffreestanding
1886 @cindex hosted environment
1888 Assert that compilation targets a freestanding environment. This
1889 implies @option{-fno-builtin}. A freestanding environment
1890 is one in which the standard library may not exist, and program startup may
1891 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1892 This is equivalent to @option{-fno-hosted}.
1894 @xref{Standards,,Language Standards Supported by GCC}, for details of
1895 freestanding and hosted environments.
1899 @cindex OpenACC accelerator programming
1900 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
1901 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
1902 compiler generates accelerated code according to the OpenACC Application
1903 Programming Interface v2.0 @w{@uref{http://www.openacc.org/}}. This option
1904 implies @option{-pthread}, and thus is only supported on targets that
1905 have support for @option{-pthread}.
1907 Note that this is an experimental feature, incomplete, and subject to
1908 change in future versions of GCC. See
1909 @w{@uref{https://gcc.gnu.org/wiki/OpenACC}} for more information.
1913 @cindex OpenMP parallel
1914 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1915 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1916 compiler generates parallel code according to the OpenMP Application
1917 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
1918 implies @option{-pthread}, and thus is only supported on targets that
1919 have support for @option{-pthread}. @option{-fopenmp} implies
1920 @option{-fopenmp-simd}.
1923 @opindex fopenmp-simd
1926 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
1927 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
1932 @cindex Enable Cilk Plus
1933 Enable the usage of Cilk Plus language extension features for C/C++.
1934 When the option @option{-fcilkplus} is specified, enable the usage of
1935 the Cilk Plus Language extension features for C/C++. The present
1936 implementation follows ABI version 1.2. This is an experimental
1937 feature that is only partially complete, and whose interface may
1938 change in future versions of GCC as the official specification
1939 changes. Currently, all features but @code{_Cilk_for} have been
1944 When the option @option{-fgnu-tm} is specified, the compiler
1945 generates code for the Linux variant of Intel's current Transactional
1946 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1947 an experimental feature whose interface may change in future versions
1948 of GCC, as the official specification changes. Please note that not
1949 all architectures are supported for this feature.
1951 For more information on GCC's support for transactional memory,
1952 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1953 Transactional Memory Library}.
1955 Note that the transactional memory feature is not supported with
1956 non-call exceptions (@option{-fnon-call-exceptions}).
1958 @item -fms-extensions
1959 @opindex fms-extensions
1960 Accept some non-standard constructs used in Microsoft header files.
1962 In C++ code, this allows member names in structures to be similar
1963 to previous types declarations.
1972 Some cases of unnamed fields in structures and unions are only
1973 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1974 fields within structs/unions}, for details.
1976 Note that this option is off for all targets but x86
1977 targets using ms-abi.
1979 @item -fplan9-extensions
1980 @opindex fplan9-extensions
1981 Accept some non-standard constructs used in Plan 9 code.
1983 This enables @option{-fms-extensions}, permits passing pointers to
1984 structures with anonymous fields to functions that expect pointers to
1985 elements of the type of the field, and permits referring to anonymous
1986 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1987 struct/union fields within structs/unions}, for details. This is only
1988 supported for C, not C++.
1992 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1993 options for strict ISO C conformance) implies @option{-trigraphs}.
1995 @cindex traditional C language
1996 @cindex C language, traditional
1998 @itemx -traditional-cpp
1999 @opindex traditional-cpp
2000 @opindex traditional
2001 Formerly, these options caused GCC to attempt to emulate a pre-standard
2002 C compiler. They are now only supported with the @option{-E} switch.
2003 The preprocessor continues to support a pre-standard mode. See the GNU
2004 CPP manual for details.
2006 @item -fcond-mismatch
2007 @opindex fcond-mismatch
2008 Allow conditional expressions with mismatched types in the second and
2009 third arguments. The value of such an expression is void. This option
2010 is not supported for C++.
2012 @item -flax-vector-conversions
2013 @opindex flax-vector-conversions
2014 Allow implicit conversions between vectors with differing numbers of
2015 elements and/or incompatible element types. This option should not be
2018 @item -funsigned-char
2019 @opindex funsigned-char
2020 Let the type @code{char} be unsigned, like @code{unsigned char}.
2022 Each kind of machine has a default for what @code{char} should
2023 be. It is either like @code{unsigned char} by default or like
2024 @code{signed char} by default.
2026 Ideally, a portable program should always use @code{signed char} or
2027 @code{unsigned char} when it depends on the signedness of an object.
2028 But many programs have been written to use plain @code{char} and
2029 expect it to be signed, or expect it to be unsigned, depending on the
2030 machines they were written for. This option, and its inverse, let you
2031 make such a program work with the opposite default.
2033 The type @code{char} is always a distinct type from each of
2034 @code{signed char} or @code{unsigned char}, even though its behavior
2035 is always just like one of those two.
2038 @opindex fsigned-char
2039 Let the type @code{char} be signed, like @code{signed char}.
2041 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2042 the negative form of @option{-funsigned-char}. Likewise, the option
2043 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2045 @item -fsigned-bitfields
2046 @itemx -funsigned-bitfields
2047 @itemx -fno-signed-bitfields
2048 @itemx -fno-unsigned-bitfields
2049 @opindex fsigned-bitfields
2050 @opindex funsigned-bitfields
2051 @opindex fno-signed-bitfields
2052 @opindex fno-unsigned-bitfields
2053 These options control whether a bit-field is signed or unsigned, when the
2054 declaration does not use either @code{signed} or @code{unsigned}. By
2055 default, such a bit-field is signed, because this is consistent: the
2056 basic integer types such as @code{int} are signed types.
2059 @node C++ Dialect Options
2060 @section Options Controlling C++ Dialect
2062 @cindex compiler options, C++
2063 @cindex C++ options, command-line
2064 @cindex options, C++
2065 This section describes the command-line options that are only meaningful
2066 for C++ programs. You can also use most of the GNU compiler options
2067 regardless of what language your program is in. For example, you
2068 might compile a file @file{firstClass.C} like this:
2071 g++ -g -frepo -O -c firstClass.C
2075 In this example, only @option{-frepo} is an option meant
2076 only for C++ programs; you can use the other options with any
2077 language supported by GCC@.
2079 Here is a list of options that are @emph{only} for compiling C++ programs:
2083 @item -fabi-version=@var{n}
2084 @opindex fabi-version
2085 Use version @var{n} of the C++ ABI@. The default is version 0.
2087 Version 0 refers to the version conforming most closely to
2088 the C++ ABI specification. Therefore, the ABI obtained using version 0
2089 will change in different versions of G++ as ABI bugs are fixed.
2091 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2093 Version 2 is the version of the C++ ABI that first appeared in G++
2094 3.4, and was the default through G++ 4.9.
2096 Version 3 corrects an error in mangling a constant address as a
2099 Version 4, which first appeared in G++ 4.5, implements a standard
2100 mangling for vector types.
2102 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2103 attribute const/volatile on function pointer types, decltype of a
2104 plain decl, and use of a function parameter in the declaration of
2107 Version 6, which first appeared in G++ 4.7, corrects the promotion
2108 behavior of C++11 scoped enums and the mangling of template argument
2109 packs, const/static_cast, prefix ++ and --, and a class scope function
2110 used as a template argument.
2112 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2113 builtin type and corrects the mangling of lambdas in default argument
2116 Version 8, which first appeared in G++ 4.9, corrects the substitution
2117 behavior of function types with function-cv-qualifiers.
2119 See also @option{-Wabi}.
2121 @item -fabi-compat-version=@var{n}
2122 @opindex fabi-compat-version
2123 On targets that support strong aliases, G++
2124 works around mangling changes by creating an alias with the correct
2125 mangled name when defining a symbol with an incorrect mangled name.
2126 This switch specifies which ABI version to use for the alias.
2128 With @option{-fabi-version=0} (the default), this defaults to 2. If
2129 another ABI version is explicitly selected, this defaults to 0.
2131 The compatibility version is also set by @option{-Wabi=@var{n}}.
2133 @item -fno-access-control
2134 @opindex fno-access-control
2135 Turn off all access checking. This switch is mainly useful for working
2136 around bugs in the access control code.
2140 Check that the pointer returned by @code{operator new} is non-null
2141 before attempting to modify the storage allocated. This check is
2142 normally unnecessary because the C++ standard specifies that
2143 @code{operator new} only returns @code{0} if it is declared
2144 @code{throw()}, in which case the compiler always checks the
2145 return value even without this option. In all other cases, when
2146 @code{operator new} has a non-empty exception specification, memory
2147 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2148 @samp{new (nothrow)}.
2150 @item -fconstexpr-depth=@var{n}
2151 @opindex fconstexpr-depth
2152 Set the maximum nested evaluation depth for C++11 constexpr functions
2153 to @var{n}. A limit is needed to detect endless recursion during
2154 constant expression evaluation. The minimum specified by the standard
2157 @item -fdeduce-init-list
2158 @opindex fdeduce-init-list
2159 Enable deduction of a template type parameter as
2160 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2163 template <class T> auto forward(T t) -> decltype (realfn (t))
2170 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2174 This deduction was implemented as a possible extension to the
2175 originally proposed semantics for the C++11 standard, but was not part
2176 of the final standard, so it is disabled by default. This option is
2177 deprecated, and may be removed in a future version of G++.
2179 @item -ffriend-injection
2180 @opindex ffriend-injection
2181 Inject friend functions into the enclosing namespace, so that they are
2182 visible outside the scope of the class in which they are declared.
2183 Friend functions were documented to work this way in the old Annotated
2184 C++ Reference Manual.
2185 However, in ISO C++ a friend function that is not declared
2186 in an enclosing scope can only be found using argument dependent
2187 lookup. GCC defaults to the standard behavior.
2189 This option is for compatibility, and may be removed in a future
2192 @item -fno-elide-constructors
2193 @opindex fno-elide-constructors
2194 The C++ standard allows an implementation to omit creating a temporary
2195 that is only used to initialize another object of the same type.
2196 Specifying this option disables that optimization, and forces G++ to
2197 call the copy constructor in all cases.
2199 @item -fno-enforce-eh-specs
2200 @opindex fno-enforce-eh-specs
2201 Don't generate code to check for violation of exception specifications
2202 at run time. This option violates the C++ standard, but may be useful
2203 for reducing code size in production builds, much like defining
2204 @code{NDEBUG}. This does not give user code permission to throw
2205 exceptions in violation of the exception specifications; the compiler
2206 still optimizes based on the specifications, so throwing an
2207 unexpected exception results in undefined behavior at run time.
2209 @item -fextern-tls-init
2210 @itemx -fno-extern-tls-init
2211 @opindex fextern-tls-init
2212 @opindex fno-extern-tls-init
2213 The C++11 and OpenMP standards allow @code{thread_local} and
2214 @code{threadprivate} variables to have dynamic (runtime)
2215 initialization. To support this, any use of such a variable goes
2216 through a wrapper function that performs any necessary initialization.
2217 When the use and definition of the variable are in the same
2218 translation unit, this overhead can be optimized away, but when the
2219 use is in a different translation unit there is significant overhead
2220 even if the variable doesn't actually need dynamic initialization. If
2221 the programmer can be sure that no use of the variable in a
2222 non-defining TU needs to trigger dynamic initialization (either
2223 because the variable is statically initialized, or a use of the
2224 variable in the defining TU will be executed before any uses in
2225 another TU), they can avoid this overhead with the
2226 @option{-fno-extern-tls-init} option.
2228 On targets that support symbol aliases, the default is
2229 @option{-fextern-tls-init}. On targets that do not support symbol
2230 aliases, the default is @option{-fno-extern-tls-init}.
2233 @itemx -fno-for-scope
2235 @opindex fno-for-scope
2236 If @option{-ffor-scope} is specified, the scope of variables declared in
2237 a @i{for-init-statement} is limited to the @code{for} loop itself,
2238 as specified by the C++ standard.
2239 If @option{-fno-for-scope} is specified, the scope of variables declared in
2240 a @i{for-init-statement} extends to the end of the enclosing scope,
2241 as was the case in old versions of G++, and other (traditional)
2242 implementations of C++.
2244 If neither flag is given, the default is to follow the standard,
2245 but to allow and give a warning for old-style code that would
2246 otherwise be invalid, or have different behavior.
2248 @item -fno-gnu-keywords
2249 @opindex fno-gnu-keywords
2250 Do not recognize @code{typeof} as a keyword, so that code can use this
2251 word as an identifier. You can use the keyword @code{__typeof__} instead.
2252 @option{-ansi} implies @option{-fno-gnu-keywords}.
2254 @item -fno-implicit-templates
2255 @opindex fno-implicit-templates
2256 Never emit code for non-inline templates that are instantiated
2257 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2258 @xref{Template Instantiation}, for more information.
2260 @item -fno-implicit-inline-templates
2261 @opindex fno-implicit-inline-templates
2262 Don't emit code for implicit instantiations of inline templates, either.
2263 The default is to handle inlines differently so that compiles with and
2264 without optimization need the same set of explicit instantiations.
2266 @item -fno-implement-inlines
2267 @opindex fno-implement-inlines
2268 To save space, do not emit out-of-line copies of inline functions
2269 controlled by @code{#pragma implementation}. This causes linker
2270 errors if these functions are not inlined everywhere they are called.
2272 @item -fms-extensions
2273 @opindex fms-extensions
2274 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2275 int and getting a pointer to member function via non-standard syntax.
2277 @item -fno-nonansi-builtins
2278 @opindex fno-nonansi-builtins
2279 Disable built-in declarations of functions that are not mandated by
2280 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2281 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2284 @opindex fnothrow-opt
2285 Treat a @code{throw()} exception specification as if it were a
2286 @code{noexcept} specification to reduce or eliminate the text size
2287 overhead relative to a function with no exception specification. If
2288 the function has local variables of types with non-trivial
2289 destructors, the exception specification actually makes the
2290 function smaller because the EH cleanups for those variables can be
2291 optimized away. The semantic effect is that an exception thrown out of
2292 a function with such an exception specification results in a call
2293 to @code{terminate} rather than @code{unexpected}.
2295 @item -fno-operator-names
2296 @opindex fno-operator-names
2297 Do not treat the operator name keywords @code{and}, @code{bitand},
2298 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2299 synonyms as keywords.
2301 @item -fno-optional-diags
2302 @opindex fno-optional-diags
2303 Disable diagnostics that the standard says a compiler does not need to
2304 issue. Currently, the only such diagnostic issued by G++ is the one for
2305 a name having multiple meanings within a class.
2308 @opindex fpermissive
2309 Downgrade some diagnostics about nonconformant code from errors to
2310 warnings. Thus, using @option{-fpermissive} allows some
2311 nonconforming code to compile.
2313 @item -fno-pretty-templates
2314 @opindex fno-pretty-templates
2315 When an error message refers to a specialization of a function
2316 template, the compiler normally prints the signature of the
2317 template followed by the template arguments and any typedefs or
2318 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2319 rather than @code{void f(int)}) so that it's clear which template is
2320 involved. When an error message refers to a specialization of a class
2321 template, the compiler omits any template arguments that match
2322 the default template arguments for that template. If either of these
2323 behaviors make it harder to understand the error message rather than
2324 easier, you can use @option{-fno-pretty-templates} to disable them.
2328 Enable automatic template instantiation at link time. This option also
2329 implies @option{-fno-implicit-templates}. @xref{Template
2330 Instantiation}, for more information.
2334 Disable generation of information about every class with virtual
2335 functions for use by the C++ run-time type identification features
2336 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2337 of the language, you can save some space by using this flag. Note that
2338 exception handling uses the same information, but G++ generates it as
2339 needed. The @code{dynamic_cast} operator can still be used for casts that
2340 do not require run-time type information, i.e.@: casts to @code{void *} or to
2341 unambiguous base classes.
2343 @item -fsized-deallocation
2344 @opindex fsized-deallocation
2345 Enable the built-in global declarations
2347 void operator delete (void *, std::size_t) noexcept;
2348 void operator delete[] (void *, std::size_t) noexcept;
2350 as introduced in C++14. This is useful for user-defined replacement
2351 deallocation functions that, for example, use the size of the object
2352 to make deallocation faster. Enabled by default under
2353 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2354 warns about places that might want to add a definition.
2358 Emit statistics about front-end processing at the end of the compilation.
2359 This information is generally only useful to the G++ development team.
2361 @item -fstrict-enums
2362 @opindex fstrict-enums
2363 Allow the compiler to optimize using the assumption that a value of
2364 enumerated type can only be one of the values of the enumeration (as
2365 defined in the C++ standard; basically, a value that can be
2366 represented in the minimum number of bits needed to represent all the
2367 enumerators). This assumption may not be valid if the program uses a
2368 cast to convert an arbitrary integer value to the enumerated type.
2370 @item -ftemplate-backtrace-limit=@var{n}
2371 @opindex ftemplate-backtrace-limit
2372 Set the maximum number of template instantiation notes for a single
2373 warning or error to @var{n}. The default value is 10.
2375 @item -ftemplate-depth=@var{n}
2376 @opindex ftemplate-depth
2377 Set the maximum instantiation depth for template classes to @var{n}.
2378 A limit on the template instantiation depth is needed to detect
2379 endless recursions during template class instantiation. ANSI/ISO C++
2380 conforming programs must not rely on a maximum depth greater than 17
2381 (changed to 1024 in C++11). The default value is 900, as the compiler
2382 can run out of stack space before hitting 1024 in some situations.
2384 @item -fno-threadsafe-statics
2385 @opindex fno-threadsafe-statics
2386 Do not emit the extra code to use the routines specified in the C++
2387 ABI for thread-safe initialization of local statics. You can use this
2388 option to reduce code size slightly in code that doesn't need to be
2391 @item -fuse-cxa-atexit
2392 @opindex fuse-cxa-atexit
2393 Register destructors for objects with static storage duration with the
2394 @code{__cxa_atexit} function rather than the @code{atexit} function.
2395 This option is required for fully standards-compliant handling of static
2396 destructors, but only works if your C library supports
2397 @code{__cxa_atexit}.
2399 @item -fno-use-cxa-get-exception-ptr
2400 @opindex fno-use-cxa-get-exception-ptr
2401 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2402 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2403 if the runtime routine is not available.
2405 @item -fvisibility-inlines-hidden
2406 @opindex fvisibility-inlines-hidden
2407 This switch declares that the user does not attempt to compare
2408 pointers to inline functions or methods where the addresses of the two functions
2409 are taken in different shared objects.
2411 The effect of this is that GCC may, effectively, mark inline methods with
2412 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2413 appear in the export table of a DSO and do not require a PLT indirection
2414 when used within the DSO@. Enabling this option can have a dramatic effect
2415 on load and link times of a DSO as it massively reduces the size of the
2416 dynamic export table when the library makes heavy use of templates.
2418 The behavior of this switch is not quite the same as marking the
2419 methods as hidden directly, because it does not affect static variables
2420 local to the function or cause the compiler to deduce that
2421 the function is defined in only one shared object.
2423 You may mark a method as having a visibility explicitly to negate the
2424 effect of the switch for that method. For example, if you do want to
2425 compare pointers to a particular inline method, you might mark it as
2426 having default visibility. Marking the enclosing class with explicit
2427 visibility has no effect.
2429 Explicitly instantiated inline methods are unaffected by this option
2430 as their linkage might otherwise cross a shared library boundary.
2431 @xref{Template Instantiation}.
2433 @item -fvisibility-ms-compat
2434 @opindex fvisibility-ms-compat
2435 This flag attempts to use visibility settings to make GCC's C++
2436 linkage model compatible with that of Microsoft Visual Studio.
2438 The flag makes these changes to GCC's linkage model:
2442 It sets the default visibility to @code{hidden}, like
2443 @option{-fvisibility=hidden}.
2446 Types, but not their members, are not hidden by default.
2449 The One Definition Rule is relaxed for types without explicit
2450 visibility specifications that are defined in more than one
2451 shared object: those declarations are permitted if they are
2452 permitted when this option is not used.
2455 In new code it is better to use @option{-fvisibility=hidden} and
2456 export those classes that are intended to be externally visible.
2457 Unfortunately it is possible for code to rely, perhaps accidentally,
2458 on the Visual Studio behavior.
2460 Among the consequences of these changes are that static data members
2461 of the same type with the same name but defined in different shared
2462 objects are different, so changing one does not change the other;
2463 and that pointers to function members defined in different shared
2464 objects may not compare equal. When this flag is given, it is a
2465 violation of the ODR to define types with the same name differently.
2467 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
2468 @opindex fvtable-verify
2469 Turn on (or off, if using @option{-fvtable-verify=none}) the security
2470 feature that verifies at run time, for every virtual call, that
2471 the vtable pointer through which the call is made is valid for the type of
2472 the object, and has not been corrupted or overwritten. If an invalid vtable
2473 pointer is detected at run time, an error is reported and execution of the
2474 program is immediately halted.
2476 This option causes run-time data structures to be built at program startup,
2477 which are used for verifying the vtable pointers.
2478 The options @samp{std} and @samp{preinit}
2479 control the timing of when these data structures are built. In both cases the
2480 data structures are built before execution reaches @code{main}. Using
2481 @option{-fvtable-verify=std} causes the data structures to be built after
2482 shared libraries have been loaded and initialized.
2483 @option{-fvtable-verify=preinit} causes them to be built before shared
2484 libraries have been loaded and initialized.
2486 If this option appears multiple times in the command line with different
2487 values specified, @samp{none} takes highest priority over both @samp{std} and
2488 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
2492 When used in conjunction with @option{-fvtable-verify=std} or
2493 @option{-fvtable-verify=preinit}, causes debug versions of the
2494 runtime functions for the vtable verification feature to be called.
2495 This flag also causes the compiler to log information about which
2496 vtable pointers it finds for each class.
2497 This information is written to a file named @file{vtv_set_ptr_data.log}
2498 in the directory named by the environment variable @env{VTV_LOGS_DIR}
2499 if that is defined or the current working directory otherwise.
2501 Note: This feature @emph{appends} data to the log file. If you want a fresh log
2502 file, be sure to delete any existing one.
2505 @opindex fvtv-counts
2506 This is a debugging flag. When used in conjunction with
2507 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
2508 causes the compiler to keep track of the total number of virtual calls
2509 it encounters and the number of verifications it inserts. It also
2510 counts the number of calls to certain run-time library functions
2511 that it inserts and logs this information for each compilation unit.
2512 The compiler writes this information to a file named
2513 @file{vtv_count_data.log} in the directory named by the environment
2514 variable @env{VTV_LOGS_DIR} if that is defined or the current working
2515 directory otherwise. It also counts the size of the vtable pointer sets
2516 for each class, and writes this information to @file{vtv_class_set_sizes.log}
2517 in the same directory.
2519 Note: This feature @emph{appends} data to the log files. To get fresh log
2520 files, be sure to delete any existing ones.
2524 Do not use weak symbol support, even if it is provided by the linker.
2525 By default, G++ uses weak symbols if they are available. This
2526 option exists only for testing, and should not be used by end-users;
2527 it results in inferior code and has no benefits. This option may
2528 be removed in a future release of G++.
2532 Do not search for header files in the standard directories specific to
2533 C++, but do still search the other standard directories. (This option
2534 is used when building the C++ library.)
2537 In addition, these optimization, warning, and code generation options
2538 have meanings only for C++ programs:
2541 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2544 When an explicit @option{-fabi-version=@var{n}} option is used, causes
2545 G++ to warn when it generates code that is probably not compatible with the
2546 vendor-neutral C++ ABI@. Since G++ now defaults to
2547 @option{-fabi-version=0}, @option{-Wabi} has no effect unless either
2548 an older ABI version is selected (with @option{-fabi-version=@var{n}})
2549 or an older compatibility version is selected (with
2550 @option{-Wabi=@var{n}} or @option{-fabi-compat-version=@var{n}}).
2552 Although an effort has been made to warn about
2553 all such cases, there are probably some cases that are not warned about,
2554 even though G++ is generating incompatible code. There may also be
2555 cases where warnings are emitted even though the code that is generated
2558 You should rewrite your code to avoid these warnings if you are
2559 concerned about the fact that code generated by G++ may not be binary
2560 compatible with code generated by other compilers.
2562 @option{-Wabi} can also be used with an explicit version number to
2563 warn about compatibility with a particular @option{-fabi-version}
2564 level, e.g. @option{-Wabi=2} to warn about changes relative to
2565 @option{-fabi-version=2}. Specifying a version number also sets
2566 @option{-fabi-compat-version=@var{n}}.
2568 The known incompatibilities in @option{-fabi-version=2} (which was the
2569 default from GCC 3.4 to 4.9) include:
2574 A template with a non-type template parameter of reference type was
2575 mangled incorrectly:
2578 template <int &> struct S @{@};
2582 This was fixed in @option{-fabi-version=3}.
2585 SIMD vector types declared using @code{__attribute ((vector_size))} were
2586 mangled in a non-standard way that does not allow for overloading of
2587 functions taking vectors of different sizes.
2589 The mangling was changed in @option{-fabi-version=4}.
2592 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2593 qualifiers, and @code{decltype} of a plain declaration was folded away.
2595 These mangling issues were fixed in @option{-fabi-version=5}.
2598 Scoped enumerators passed as arguments to a variadic function are
2599 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2600 On most targets this does not actually affect the parameter passing
2601 ABI, as there is no way to pass an argument smaller than @code{int}.
2603 Also, the ABI changed the mangling of template argument packs,
2604 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2605 a class scope function used as a template argument.
2607 These issues were corrected in @option{-fabi-version=6}.
2610 Lambdas in default argument scope were mangled incorrectly, and the
2611 ABI changed the mangling of @code{nullptr_t}.
2613 These issues were corrected in @option{-fabi-version=7}.
2616 When mangling a function type with function-cv-qualifiers, the
2617 un-qualified function type was incorrectly treated as a substitution
2620 This was fixed in @option{-fabi-version=8}.
2623 It also warns about psABI-related changes. The known psABI changes at this
2629 For SysV/x86-64, unions with @code{long double} members are
2630 passed in memory as specified in psABI. For example:
2640 @code{union U} is always passed in memory.
2644 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2645 @opindex Wctor-dtor-privacy
2646 @opindex Wno-ctor-dtor-privacy
2647 Warn when a class seems unusable because all the constructors or
2648 destructors in that class are private, and it has neither friends nor
2649 public static member functions. Also warn if there are no non-private
2650 methods, and there's at least one private member function that isn't
2651 a constructor or destructor.
2653 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2654 @opindex Wdelete-non-virtual-dtor
2655 @opindex Wno-delete-non-virtual-dtor
2656 Warn when @code{delete} is used to destroy an instance of a class that
2657 has virtual functions and non-virtual destructor. It is unsafe to delete
2658 an instance of a derived class through a pointer to a base class if the
2659 base class does not have a virtual destructor. This warning is enabled
2662 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2663 @opindex Wliteral-suffix
2664 @opindex Wno-literal-suffix
2665 Warn when a string or character literal is followed by a ud-suffix which does
2666 not begin with an underscore. As a conforming extension, GCC treats such
2667 suffixes as separate preprocessing tokens in order to maintain backwards
2668 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2672 #define __STDC_FORMAT_MACROS
2673 #include <inttypes.h>
2678 printf("My int64: %"PRId64"\n", i64);
2682 In this case, @code{PRId64} is treated as a separate preprocessing token.
2684 This warning is enabled by default.
2686 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2688 @opindex Wno-narrowing
2689 Warn when a narrowing conversion prohibited by C++11 occurs within
2693 int i = @{ 2.2 @}; // error: narrowing from double to int
2696 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2698 With @option{-std=c++11}, @option{-Wno-narrowing} suppresses for
2699 non-constants the diagnostic required by the standard. Note that this
2700 does not affect the meaning of well-formed code; narrowing conversions
2701 are still considered ill-formed in SFINAE context.
2703 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2705 @opindex Wno-noexcept
2706 Warn when a noexcept-expression evaluates to false because of a call
2707 to a function that does not have a non-throwing exception
2708 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2709 the compiler to never throw an exception.
2711 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2712 @opindex Wnon-virtual-dtor
2713 @opindex Wno-non-virtual-dtor
2714 Warn when a class has virtual functions and an accessible non-virtual
2715 destructor itself or in an accessible polymorphic base class, in which
2716 case it is possible but unsafe to delete an instance of a derived
2717 class through a pointer to the class itself or base class. This
2718 warning is automatically enabled if @option{-Weffc++} is specified.
2720 @item -Wreorder @r{(C++ and Objective-C++ only)}
2722 @opindex Wno-reorder
2723 @cindex reordering, warning
2724 @cindex warning for reordering of member initializers
2725 Warn when the order of member initializers given in the code does not
2726 match the order in which they must be executed. For instance:
2732 A(): j (0), i (1) @{ @}
2737 The compiler rearranges the member initializers for @code{i}
2738 and @code{j} to match the declaration order of the members, emitting
2739 a warning to that effect. This warning is enabled by @option{-Wall}.
2741 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2742 @opindex fext-numeric-literals
2743 @opindex fno-ext-numeric-literals
2744 Accept imaginary, fixed-point, or machine-defined
2745 literal number suffixes as GNU extensions.
2746 When this option is turned off these suffixes are treated
2747 as C++11 user-defined literal numeric suffixes.
2748 This is on by default for all pre-C++11 dialects and all GNU dialects:
2749 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2750 @option{-std=gnu++14}.
2751 This option is off by default
2752 for ISO C++11 onwards (@option{-std=c++11}, ...).
2755 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2758 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2761 Warn about violations of the following style guidelines from Scott Meyers'
2762 @cite{Effective C++} series of books:
2766 Define a copy constructor and an assignment operator for classes
2767 with dynamically-allocated memory.
2770 Prefer initialization to assignment in constructors.
2773 Have @code{operator=} return a reference to @code{*this}.
2776 Don't try to return a reference when you must return an object.
2779 Distinguish between prefix and postfix forms of increment and
2780 decrement operators.
2783 Never overload @code{&&}, @code{||}, or @code{,}.
2787 This option also enables @option{-Wnon-virtual-dtor}, which is also
2788 one of the effective C++ recommendations. However, the check is
2789 extended to warn about the lack of virtual destructor in accessible
2790 non-polymorphic bases classes too.
2792 When selecting this option, be aware that the standard library
2793 headers do not obey all of these guidelines; use @samp{grep -v}
2794 to filter out those warnings.
2796 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2797 @opindex Wstrict-null-sentinel
2798 @opindex Wno-strict-null-sentinel
2799 Warn about the use of an uncasted @code{NULL} as sentinel. When
2800 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2801 to @code{__null}. Although it is a null pointer constant rather than a
2802 null pointer, it is guaranteed to be of the same size as a pointer.
2803 But this use is not portable across different compilers.
2805 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2806 @opindex Wno-non-template-friend
2807 @opindex Wnon-template-friend
2808 Disable warnings when non-templatized friend functions are declared
2809 within a template. Since the advent of explicit template specification
2810 support in G++, if the name of the friend is an unqualified-id (i.e.,
2811 @samp{friend foo(int)}), the C++ language specification demands that the
2812 friend declare or define an ordinary, nontemplate function. (Section
2813 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2814 could be interpreted as a particular specialization of a templatized
2815 function. Because this non-conforming behavior is no longer the default
2816 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2817 check existing code for potential trouble spots and is on by default.
2818 This new compiler behavior can be turned off with
2819 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2820 but disables the helpful warning.
2822 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2823 @opindex Wold-style-cast
2824 @opindex Wno-old-style-cast
2825 Warn if an old-style (C-style) cast to a non-void type is used within
2826 a C++ program. The new-style casts (@code{dynamic_cast},
2827 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
2828 less vulnerable to unintended effects and much easier to search for.
2830 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2831 @opindex Woverloaded-virtual
2832 @opindex Wno-overloaded-virtual
2833 @cindex overloaded virtual function, warning
2834 @cindex warning for overloaded virtual function
2835 Warn when a function declaration hides virtual functions from a
2836 base class. For example, in:
2843 struct B: public A @{
2848 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2859 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2860 @opindex Wno-pmf-conversions
2861 @opindex Wpmf-conversions
2862 Disable the diagnostic for converting a bound pointer to member function
2865 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2866 @opindex Wsign-promo
2867 @opindex Wno-sign-promo
2868 Warn when overload resolution chooses a promotion from unsigned or
2869 enumerated type to a signed type, over a conversion to an unsigned type of
2870 the same size. Previous versions of G++ tried to preserve
2871 unsignedness, but the standard mandates the current behavior.
2874 @node Objective-C and Objective-C++ Dialect Options
2875 @section Options Controlling Objective-C and Objective-C++ Dialects
2877 @cindex compiler options, Objective-C and Objective-C++
2878 @cindex Objective-C and Objective-C++ options, command-line
2879 @cindex options, Objective-C and Objective-C++
2880 (NOTE: This manual does not describe the Objective-C and Objective-C++
2881 languages themselves. @xref{Standards,,Language Standards
2882 Supported by GCC}, for references.)
2884 This section describes the command-line options that are only meaningful
2885 for Objective-C and Objective-C++ programs. You can also use most of
2886 the language-independent GNU compiler options.
2887 For example, you might compile a file @file{some_class.m} like this:
2890 gcc -g -fgnu-runtime -O -c some_class.m
2894 In this example, @option{-fgnu-runtime} is an option meant only for
2895 Objective-C and Objective-C++ programs; you can use the other options with
2896 any language supported by GCC@.
2898 Note that since Objective-C is an extension of the C language, Objective-C
2899 compilations may also use options specific to the C front-end (e.g.,
2900 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2901 C++-specific options (e.g., @option{-Wabi}).
2903 Here is a list of options that are @emph{only} for compiling Objective-C
2904 and Objective-C++ programs:
2907 @item -fconstant-string-class=@var{class-name}
2908 @opindex fconstant-string-class
2909 Use @var{class-name} as the name of the class to instantiate for each
2910 literal string specified with the syntax @code{@@"@dots{}"}. The default
2911 class name is @code{NXConstantString} if the GNU runtime is being used, and
2912 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2913 @option{-fconstant-cfstrings} option, if also present, overrides the
2914 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2915 to be laid out as constant CoreFoundation strings.
2918 @opindex fgnu-runtime
2919 Generate object code compatible with the standard GNU Objective-C
2920 runtime. This is the default for most types of systems.
2922 @item -fnext-runtime
2923 @opindex fnext-runtime
2924 Generate output compatible with the NeXT runtime. This is the default
2925 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2926 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2929 @item -fno-nil-receivers
2930 @opindex fno-nil-receivers
2931 Assume that all Objective-C message dispatches (@code{[receiver
2932 message:arg]}) in this translation unit ensure that the receiver is
2933 not @code{nil}. This allows for more efficient entry points in the
2934 runtime to be used. This option is only available in conjunction with
2935 the NeXT runtime and ABI version 0 or 1.
2937 @item -fobjc-abi-version=@var{n}
2938 @opindex fobjc-abi-version
2939 Use version @var{n} of the Objective-C ABI for the selected runtime.
2940 This option is currently supported only for the NeXT runtime. In that
2941 case, Version 0 is the traditional (32-bit) ABI without support for
2942 properties and other Objective-C 2.0 additions. Version 1 is the
2943 traditional (32-bit) ABI with support for properties and other
2944 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2945 nothing is specified, the default is Version 0 on 32-bit target
2946 machines, and Version 2 on 64-bit target machines.
2948 @item -fobjc-call-cxx-cdtors
2949 @opindex fobjc-call-cxx-cdtors
2950 For each Objective-C class, check if any of its instance variables is a
2951 C++ object with a non-trivial default constructor. If so, synthesize a
2952 special @code{- (id) .cxx_construct} instance method which runs
2953 non-trivial default constructors on any such instance variables, in order,
2954 and then return @code{self}. Similarly, check if any instance variable
2955 is a C++ object with a non-trivial destructor, and if so, synthesize a
2956 special @code{- (void) .cxx_destruct} method which runs
2957 all such default destructors, in reverse order.
2959 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2960 methods thusly generated only operate on instance variables
2961 declared in the current Objective-C class, and not those inherited
2962 from superclasses. It is the responsibility of the Objective-C
2963 runtime to invoke all such methods in an object's inheritance
2964 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
2965 by the runtime immediately after a new object instance is allocated;
2966 the @code{- (void) .cxx_destruct} methods are invoked immediately
2967 before the runtime deallocates an object instance.
2969 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2970 support for invoking the @code{- (id) .cxx_construct} and
2971 @code{- (void) .cxx_destruct} methods.
2973 @item -fobjc-direct-dispatch
2974 @opindex fobjc-direct-dispatch
2975 Allow fast jumps to the message dispatcher. On Darwin this is
2976 accomplished via the comm page.
2978 @item -fobjc-exceptions
2979 @opindex fobjc-exceptions
2980 Enable syntactic support for structured exception handling in
2981 Objective-C, similar to what is offered by C++ and Java. This option
2982 is required to use the Objective-C keywords @code{@@try},
2983 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2984 @code{@@synchronized}. This option is available with both the GNU
2985 runtime and the NeXT runtime (but not available in conjunction with
2986 the NeXT runtime on Mac OS X 10.2 and earlier).
2990 Enable garbage collection (GC) in Objective-C and Objective-C++
2991 programs. This option is only available with the NeXT runtime; the
2992 GNU runtime has a different garbage collection implementation that
2993 does not require special compiler flags.
2995 @item -fobjc-nilcheck
2996 @opindex fobjc-nilcheck
2997 For the NeXT runtime with version 2 of the ABI, check for a nil
2998 receiver in method invocations before doing the actual method call.
2999 This is the default and can be disabled using
3000 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3001 checked for nil in this way no matter what this flag is set to.
3002 Currently this flag does nothing when the GNU runtime, or an older
3003 version of the NeXT runtime ABI, is used.
3005 @item -fobjc-std=objc1
3007 Conform to the language syntax of Objective-C 1.0, the language
3008 recognized by GCC 4.0. This only affects the Objective-C additions to
3009 the C/C++ language; it does not affect conformance to C/C++ standards,
3010 which is controlled by the separate C/C++ dialect option flags. When
3011 this option is used with the Objective-C or Objective-C++ compiler,
3012 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3013 This is useful if you need to make sure that your Objective-C code can
3014 be compiled with older versions of GCC@.
3016 @item -freplace-objc-classes
3017 @opindex freplace-objc-classes
3018 Emit a special marker instructing @command{ld(1)} not to statically link in
3019 the resulting object file, and allow @command{dyld(1)} to load it in at
3020 run time instead. This is used in conjunction with the Fix-and-Continue
3021 debugging mode, where the object file in question may be recompiled and
3022 dynamically reloaded in the course of program execution, without the need
3023 to restart the program itself. Currently, Fix-and-Continue functionality
3024 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3029 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3030 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3031 compile time) with static class references that get initialized at load time,
3032 which improves run-time performance. Specifying the @option{-fzero-link} flag
3033 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3034 to be retained. This is useful in Zero-Link debugging mode, since it allows
3035 for individual class implementations to be modified during program execution.
3036 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3037 regardless of command-line options.
3039 @item -fno-local-ivars
3040 @opindex fno-local-ivars
3041 @opindex flocal-ivars
3042 By default instance variables in Objective-C can be accessed as if
3043 they were local variables from within the methods of the class they're
3044 declared in. This can lead to shadowing between instance variables
3045 and other variables declared either locally inside a class method or
3046 globally with the same name. Specifying the @option{-fno-local-ivars}
3047 flag disables this behavior thus avoiding variable shadowing issues.
3049 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3050 @opindex fivar-visibility
3051 Set the default instance variable visibility to the specified option
3052 so that instance variables declared outside the scope of any access
3053 modifier directives default to the specified visibility.
3057 Dump interface declarations for all classes seen in the source file to a
3058 file named @file{@var{sourcename}.decl}.
3060 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3061 @opindex Wassign-intercept
3062 @opindex Wno-assign-intercept
3063 Warn whenever an Objective-C assignment is being intercepted by the
3066 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3067 @opindex Wno-protocol
3069 If a class is declared to implement a protocol, a warning is issued for
3070 every method in the protocol that is not implemented by the class. The
3071 default behavior is to issue a warning for every method not explicitly
3072 implemented in the class, even if a method implementation is inherited
3073 from the superclass. If you use the @option{-Wno-protocol} option, then
3074 methods inherited from the superclass are considered to be implemented,
3075 and no warning is issued for them.
3077 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3079 @opindex Wno-selector
3080 Warn if multiple methods of different types for the same selector are
3081 found during compilation. The check is performed on the list of methods
3082 in the final stage of compilation. Additionally, a check is performed
3083 for each selector appearing in a @code{@@selector(@dots{})}
3084 expression, and a corresponding method for that selector has been found
3085 during compilation. Because these checks scan the method table only at
3086 the end of compilation, these warnings are not produced if the final
3087 stage of compilation is not reached, for example because an error is
3088 found during compilation, or because the @option{-fsyntax-only} option is
3091 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3092 @opindex Wstrict-selector-match
3093 @opindex Wno-strict-selector-match
3094 Warn if multiple methods with differing argument and/or return types are
3095 found for a given selector when attempting to send a message using this
3096 selector to a receiver of type @code{id} or @code{Class}. When this flag
3097 is off (which is the default behavior), the compiler omits such warnings
3098 if any differences found are confined to types that share the same size
3101 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3102 @opindex Wundeclared-selector
3103 @opindex Wno-undeclared-selector
3104 Warn if a @code{@@selector(@dots{})} expression referring to an
3105 undeclared selector is found. A selector is considered undeclared if no
3106 method with that name has been declared before the
3107 @code{@@selector(@dots{})} expression, either explicitly in an
3108 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3109 an @code{@@implementation} section. This option always performs its
3110 checks as soon as a @code{@@selector(@dots{})} expression is found,
3111 while @option{-Wselector} only performs its checks in the final stage of
3112 compilation. This also enforces the coding style convention
3113 that methods and selectors must be declared before being used.
3115 @item -print-objc-runtime-info
3116 @opindex print-objc-runtime-info
3117 Generate C header describing the largest structure that is passed by
3122 @node Language Independent Options
3123 @section Options to Control Diagnostic Messages Formatting
3124 @cindex options to control diagnostics formatting
3125 @cindex diagnostic messages
3126 @cindex message formatting
3128 Traditionally, diagnostic messages have been formatted irrespective of
3129 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3130 options described below
3131 to control the formatting algorithm for diagnostic messages,
3132 e.g.@: how many characters per line, how often source location
3133 information should be reported. Note that some language front ends may not
3134 honor these options.
3137 @item -fmessage-length=@var{n}
3138 @opindex fmessage-length
3139 Try to format error messages so that they fit on lines of about
3140 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3141 done; each error message appears on a single line. This is the
3142 default for all front ends.
3144 @item -fdiagnostics-show-location=once
3145 @opindex fdiagnostics-show-location
3146 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3147 reporter to emit source location information @emph{once}; that is, in
3148 case the message is too long to fit on a single physical line and has to
3149 be wrapped, the source location won't be emitted (as prefix) again,
3150 over and over, in subsequent continuation lines. This is the default
3153 @item -fdiagnostics-show-location=every-line
3154 Only meaningful in line-wrapping mode. Instructs the diagnostic
3155 messages reporter to emit the same source location information (as
3156 prefix) for physical lines that result from the process of breaking
3157 a message which is too long to fit on a single line.
3159 @item -fdiagnostics-color[=@var{WHEN}]
3160 @itemx -fno-diagnostics-color
3161 @opindex fdiagnostics-color
3162 @cindex highlight, color, colour
3163 @vindex GCC_COLORS @r{environment variable}
3164 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3165 or @samp{auto}. The default depends on how the compiler has been configured,
3166 it can be any of the above @var{WHEN} options or also @samp{never}
3167 if @env{GCC_COLORS} environment variable isn't present in the environment,
3168 and @samp{auto} otherwise.
3169 @samp{auto} means to use color only when the standard error is a terminal.
3170 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3171 aliases for @option{-fdiagnostics-color=always} and
3172 @option{-fdiagnostics-color=never}, respectively.
3174 The colors are defined by the environment variable @env{GCC_COLORS}.
3175 Its value is a colon-separated list of capabilities and Select Graphic
3176 Rendition (SGR) substrings. SGR commands are interpreted by the
3177 terminal or terminal emulator. (See the section in the documentation
3178 of your text terminal for permitted values and their meanings as
3179 character attributes.) These substring values are integers in decimal
3180 representation and can be concatenated with semicolons.
3181 Common values to concatenate include
3183 @samp{4} for underline,
3185 @samp{7} for inverse,
3186 @samp{39} for default foreground color,
3187 @samp{30} to @samp{37} for foreground colors,
3188 @samp{90} to @samp{97} for 16-color mode foreground colors,
3189 @samp{38;5;0} to @samp{38;5;255}
3190 for 88-color and 256-color modes foreground colors,
3191 @samp{49} for default background color,
3192 @samp{40} to @samp{47} for background colors,
3193 @samp{100} to @samp{107} for 16-color mode background colors,
3194 and @samp{48;5;0} to @samp{48;5;255}
3195 for 88-color and 256-color modes background colors.
3197 The default @env{GCC_COLORS} is
3199 error=01;31:warning=01;35:note=01;36:caret=01;32:locus=01:quote=01
3202 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3203 @samp{01;36} is bold cyan, @samp{01;32} is bold green and
3204 @samp{01} is bold. Setting @env{GCC_COLORS} to the empty
3205 string disables colors.
3206 Supported capabilities are as follows.
3210 @vindex error GCC_COLORS @r{capability}
3211 SGR substring for error: markers.
3214 @vindex warning GCC_COLORS @r{capability}
3215 SGR substring for warning: markers.
3218 @vindex note GCC_COLORS @r{capability}
3219 SGR substring for note: markers.
3222 @vindex caret GCC_COLORS @r{capability}
3223 SGR substring for caret line.
3226 @vindex locus GCC_COLORS @r{capability}
3227 SGR substring for location information, @samp{file:line} or
3228 @samp{file:line:column} etc.
3231 @vindex quote GCC_COLORS @r{capability}
3232 SGR substring for information printed within quotes.
3235 @item -fno-diagnostics-show-option
3236 @opindex fno-diagnostics-show-option
3237 @opindex fdiagnostics-show-option
3238 By default, each diagnostic emitted includes text indicating the
3239 command-line option that directly controls the diagnostic (if such an
3240 option is known to the diagnostic machinery). Specifying the
3241 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3243 @item -fno-diagnostics-show-caret
3244 @opindex fno-diagnostics-show-caret
3245 @opindex fdiagnostics-show-caret
3246 By default, each diagnostic emitted includes the original source line
3247 and a caret '^' indicating the column. This option suppresses this
3248 information. The source line is truncated to @var{n} characters, if
3249 the @option{-fmessage-length=n} is given. When the output is done
3250 to the terminal, the width is limited to the width given by the
3251 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3255 @node Warning Options
3256 @section Options to Request or Suppress Warnings
3257 @cindex options to control warnings
3258 @cindex warning messages
3259 @cindex messages, warning
3260 @cindex suppressing warnings
3262 Warnings are diagnostic messages that report constructions that
3263 are not inherently erroneous but that are risky or suggest there
3264 may have been an error.
3266 The following language-independent options do not enable specific
3267 warnings but control the kinds of diagnostics produced by GCC@.
3270 @cindex syntax checking
3272 @opindex fsyntax-only
3273 Check the code for syntax errors, but don't do anything beyond that.
3275 @item -fmax-errors=@var{n}
3276 @opindex fmax-errors
3277 Limits the maximum number of error messages to @var{n}, at which point
3278 GCC bails out rather than attempting to continue processing the source
3279 code. If @var{n} is 0 (the default), there is no limit on the number
3280 of error messages produced. If @option{-Wfatal-errors} is also
3281 specified, then @option{-Wfatal-errors} takes precedence over this
3286 Inhibit all warning messages.
3291 Make all warnings into errors.
3296 Make the specified warning into an error. The specifier for a warning
3297 is appended; for example @option{-Werror=switch} turns the warnings
3298 controlled by @option{-Wswitch} into errors. This switch takes a
3299 negative form, to be used to negate @option{-Werror} for specific
3300 warnings; for example @option{-Wno-error=switch} makes
3301 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3304 The warning message for each controllable warning includes the
3305 option that controls the warning. That option can then be used with
3306 @option{-Werror=} and @option{-Wno-error=} as described above.
3307 (Printing of the option in the warning message can be disabled using the
3308 @option{-fno-diagnostics-show-option} flag.)
3310 Note that specifying @option{-Werror=}@var{foo} automatically implies
3311 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3314 @item -Wfatal-errors
3315 @opindex Wfatal-errors
3316 @opindex Wno-fatal-errors
3317 This option causes the compiler to abort compilation on the first error
3318 occurred rather than trying to keep going and printing further error
3323 You can request many specific warnings with options beginning with
3324 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3325 implicit declarations. Each of these specific warning options also
3326 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3327 example, @option{-Wno-implicit}. This manual lists only one of the
3328 two forms, whichever is not the default. For further
3329 language-specific options also refer to @ref{C++ Dialect Options} and
3330 @ref{Objective-C and Objective-C++ Dialect Options}.
3332 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3333 options, such as @option{-Wunused}, which may turn on further options,
3334 such as @option{-Wunused-value}. The combined effect of positive and
3335 negative forms is that more specific options have priority over less
3336 specific ones, independently of their position in the command-line. For
3337 options of the same specificity, the last one takes effect. Options
3338 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3339 as if they appeared at the end of the command-line.
3341 When an unrecognized warning option is requested (e.g.,
3342 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3343 that the option is not recognized. However, if the @option{-Wno-} form
3344 is used, the behavior is slightly different: no diagnostic is
3345 produced for @option{-Wno-unknown-warning} unless other diagnostics
3346 are being produced. This allows the use of new @option{-Wno-} options
3347 with old compilers, but if something goes wrong, the compiler
3348 warns that an unrecognized option is present.
3355 Issue all the warnings demanded by strict ISO C and ISO C++;
3356 reject all programs that use forbidden extensions, and some other
3357 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3358 version of the ISO C standard specified by any @option{-std} option used.
3360 Valid ISO C and ISO C++ programs should compile properly with or without
3361 this option (though a rare few require @option{-ansi} or a
3362 @option{-std} option specifying the required version of ISO C)@. However,
3363 without this option, certain GNU extensions and traditional C and C++
3364 features are supported as well. With this option, they are rejected.
3366 @option{-Wpedantic} does not cause warning messages for use of the
3367 alternate keywords whose names begin and end with @samp{__}. Pedantic
3368 warnings are also disabled in the expression that follows
3369 @code{__extension__}. However, only system header files should use
3370 these escape routes; application programs should avoid them.
3371 @xref{Alternate Keywords}.
3373 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3374 C conformance. They soon find that it does not do quite what they want:
3375 it finds some non-ISO practices, but not all---only those for which
3376 ISO C @emph{requires} a diagnostic, and some others for which
3377 diagnostics have been added.
3379 A feature to report any failure to conform to ISO C might be useful in
3380 some instances, but would require considerable additional work and would
3381 be quite different from @option{-Wpedantic}. We don't have plans to
3382 support such a feature in the near future.
3384 Where the standard specified with @option{-std} represents a GNU
3385 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3386 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3387 extended dialect is based. Warnings from @option{-Wpedantic} are given
3388 where they are required by the base standard. (It does not make sense
3389 for such warnings to be given only for features not in the specified GNU
3390 C dialect, since by definition the GNU dialects of C include all
3391 features the compiler supports with the given option, and there would be
3392 nothing to warn about.)
3394 @item -pedantic-errors
3395 @opindex pedantic-errors
3396 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3397 requires a diagnostic, in some cases where there is undefined behavior
3398 at compile-time and in some other cases that do not prevent compilation
3399 of programs that are valid according to the standard. This is not
3400 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3401 by this option and not enabled by the latter and vice versa.
3406 This enables all the warnings about constructions that some users
3407 consider questionable, and that are easy to avoid (or modify to
3408 prevent the warning), even in conjunction with macros. This also
3409 enables some language-specific warnings described in @ref{C++ Dialect
3410 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3412 @option{-Wall} turns on the following warning flags:
3414 @gccoptlist{-Waddress @gol
3415 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3416 -Wc++11-compat -Wc++14-compat@gol
3417 -Wchar-subscripts @gol
3418 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3419 -Wimplicit-int @r{(C and Objective-C only)} @gol
3420 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3423 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3424 -Wmaybe-uninitialized @gol
3425 -Wmissing-braces @r{(only for C/ObjC)} @gol
3432 -Wsequence-point @gol
3433 -Wsign-compare @r{(only in C++)} @gol
3434 -Wstrict-aliasing @gol
3435 -Wstrict-overflow=1 @gol
3438 -Wuninitialized @gol
3439 -Wunknown-pragmas @gol
3440 -Wunused-function @gol
3443 -Wunused-variable @gol
3444 -Wvolatile-register-var @gol
3447 Note that some warning flags are not implied by @option{-Wall}. Some of
3448 them warn about constructions that users generally do not consider
3449 questionable, but which occasionally you might wish to check for;
3450 others warn about constructions that are necessary or hard to avoid in
3451 some cases, and there is no simple way to modify the code to suppress
3452 the warning. Some of them are enabled by @option{-Wextra} but many of
3453 them must be enabled individually.
3459 This enables some extra warning flags that are not enabled by
3460 @option{-Wall}. (This option used to be called @option{-W}. The older
3461 name is still supported, but the newer name is more descriptive.)
3463 @gccoptlist{-Wclobbered @gol
3465 -Wignored-qualifiers @gol
3466 -Wmissing-field-initializers @gol
3467 -Wmissing-parameter-type @r{(C only)} @gol
3468 -Wold-style-declaration @r{(C only)} @gol
3469 -Woverride-init @gol
3472 -Wuninitialized @gol
3473 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3474 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3477 The option @option{-Wextra} also prints warning messages for the
3483 A pointer is compared against integer zero with @code{<}, @code{<=},
3484 @code{>}, or @code{>=}.
3487 (C++ only) An enumerator and a non-enumerator both appear in a
3488 conditional expression.
3491 (C++ only) Ambiguous virtual bases.
3494 (C++ only) Subscripting an array that has been declared @code{register}.
3497 (C++ only) Taking the address of a variable that has been declared
3501 (C++ only) A base class is not initialized in a derived class's copy
3506 @item -Wchar-subscripts
3507 @opindex Wchar-subscripts
3508 @opindex Wno-char-subscripts
3509 Warn if an array subscript has type @code{char}. This is a common cause
3510 of error, as programmers often forget that this type is signed on some
3512 This warning is enabled by @option{-Wall}.
3516 @opindex Wno-comment
3517 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3518 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3519 This warning is enabled by @option{-Wall}.
3521 @item -Wno-coverage-mismatch
3522 @opindex Wno-coverage-mismatch
3523 Warn if feedback profiles do not match when using the
3524 @option{-fprofile-use} option.
3525 If a source file is changed between compiling with @option{-fprofile-gen} and
3526 with @option{-fprofile-use}, the files with the profile feedback can fail
3527 to match the source file and GCC cannot use the profile feedback
3528 information. By default, this warning is enabled and is treated as an
3529 error. @option{-Wno-coverage-mismatch} can be used to disable the
3530 warning or @option{-Wno-error=coverage-mismatch} can be used to
3531 disable the error. Disabling the error for this warning can result in
3532 poorly optimized code and is useful only in the
3533 case of very minor changes such as bug fixes to an existing code-base.
3534 Completely disabling the warning is not recommended.
3537 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3539 Suppress warning messages emitted by @code{#warning} directives.
3541 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3542 @opindex Wdouble-promotion
3543 @opindex Wno-double-promotion
3544 Give a warning when a value of type @code{float} is implicitly
3545 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3546 floating-point unit implement @code{float} in hardware, but emulate
3547 @code{double} in software. On such a machine, doing computations
3548 using @code{double} values is much more expensive because of the
3549 overhead required for software emulation.
3551 It is easy to accidentally do computations with @code{double} because
3552 floating-point literals are implicitly of type @code{double}. For
3556 float area(float radius)
3558 return 3.14159 * radius * radius;
3562 the compiler performs the entire computation with @code{double}
3563 because the floating-point literal is a @code{double}.
3566 @itemx -Wformat=@var{n}
3569 @opindex ffreestanding
3570 @opindex fno-builtin
3572 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3573 the arguments supplied have types appropriate to the format string
3574 specified, and that the conversions specified in the format string make
3575 sense. This includes standard functions, and others specified by format
3576 attributes (@pxref{Function Attributes}), in the @code{printf},
3577 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3578 not in the C standard) families (or other target-specific families).
3579 Which functions are checked without format attributes having been
3580 specified depends on the standard version selected, and such checks of
3581 functions without the attribute specified are disabled by
3582 @option{-ffreestanding} or @option{-fno-builtin}.
3584 The formats are checked against the format features supported by GNU
3585 libc version 2.2. These include all ISO C90 and C99 features, as well
3586 as features from the Single Unix Specification and some BSD and GNU
3587 extensions. Other library implementations may not support all these
3588 features; GCC does not support warning about features that go beyond a
3589 particular library's limitations. However, if @option{-Wpedantic} is used
3590 with @option{-Wformat}, warnings are given about format features not
3591 in the selected standard version (but not for @code{strfmon} formats,
3592 since those are not in any version of the C standard). @xref{C Dialect
3593 Options,,Options Controlling C Dialect}.
3600 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3601 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3602 @option{-Wformat} also checks for null format arguments for several
3603 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3604 aspects of this level of format checking can be disabled by the
3605 options: @option{-Wno-format-contains-nul},
3606 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3607 @option{-Wformat} is enabled by @option{-Wall}.
3609 @item -Wno-format-contains-nul
3610 @opindex Wno-format-contains-nul
3611 @opindex Wformat-contains-nul
3612 If @option{-Wformat} is specified, do not warn about format strings that
3615 @item -Wno-format-extra-args
3616 @opindex Wno-format-extra-args
3617 @opindex Wformat-extra-args
3618 If @option{-Wformat} is specified, do not warn about excess arguments to a
3619 @code{printf} or @code{scanf} format function. The C standard specifies
3620 that such arguments are ignored.
3622 Where the unused arguments lie between used arguments that are
3623 specified with @samp{$} operand number specifications, normally
3624 warnings are still given, since the implementation could not know what
3625 type to pass to @code{va_arg} to skip the unused arguments. However,
3626 in the case of @code{scanf} formats, this option suppresses the
3627 warning if the unused arguments are all pointers, since the Single
3628 Unix Specification says that such unused arguments are allowed.
3630 @item -Wno-format-zero-length
3631 @opindex Wno-format-zero-length
3632 @opindex Wformat-zero-length
3633 If @option{-Wformat} is specified, do not warn about zero-length formats.
3634 The C standard specifies that zero-length formats are allowed.
3639 Enable @option{-Wformat} plus additional format checks. Currently
3640 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
3643 @item -Wformat-nonliteral
3644 @opindex Wformat-nonliteral
3645 @opindex Wno-format-nonliteral
3646 If @option{-Wformat} is specified, also warn if the format string is not a
3647 string literal and so cannot be checked, unless the format function
3648 takes its format arguments as a @code{va_list}.
3650 @item -Wformat-security
3651 @opindex Wformat-security
3652 @opindex Wno-format-security
3653 If @option{-Wformat} is specified, also warn about uses of format
3654 functions that represent possible security problems. At present, this
3655 warns about calls to @code{printf} and @code{scanf} functions where the
3656 format string is not a string literal and there are no format arguments,
3657 as in @code{printf (foo);}. This may be a security hole if the format
3658 string came from untrusted input and contains @samp{%n}. (This is
3659 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3660 in future warnings may be added to @option{-Wformat-security} that are not
3661 included in @option{-Wformat-nonliteral}.)
3663 @item -Wformat-signedness
3664 @opindex Wformat-signedness
3665 @opindex Wno-format-signedness
3666 If @option{-Wformat} is specified, also warn if the format string
3667 requires an unsigned argument and the argument is signed and vice versa.
3670 @opindex Wformat-y2k
3671 @opindex Wno-format-y2k
3672 If @option{-Wformat} is specified, also warn about @code{strftime}
3673 formats that may yield only a two-digit year.
3678 @opindex Wno-nonnull
3679 Warn about passing a null pointer for arguments marked as
3680 requiring a non-null value by the @code{nonnull} function attribute.
3682 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3683 can be disabled with the @option{-Wno-nonnull} option.
3685 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3687 @opindex Wno-init-self
3688 Warn about uninitialized variables that are initialized with themselves.
3689 Note this option can only be used with the @option{-Wuninitialized} option.
3691 For example, GCC warns about @code{i} being uninitialized in the
3692 following snippet only when @option{-Winit-self} has been specified:
3703 This warning is enabled by @option{-Wall} in C++.
3705 @item -Wimplicit-int @r{(C and Objective-C only)}
3706 @opindex Wimplicit-int
3707 @opindex Wno-implicit-int
3708 Warn when a declaration does not specify a type.
3709 This warning is enabled by @option{-Wall}.
3711 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3712 @opindex Wimplicit-function-declaration
3713 @opindex Wno-implicit-function-declaration
3714 Give a warning whenever a function is used before being declared. In
3715 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3716 enabled by default and it is made into an error by
3717 @option{-pedantic-errors}. This warning is also enabled by
3720 @item -Wimplicit @r{(C and Objective-C only)}
3722 @opindex Wno-implicit
3723 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3724 This warning is enabled by @option{-Wall}.
3726 @item -Wignored-qualifiers @r{(C and C++ only)}
3727 @opindex Wignored-qualifiers
3728 @opindex Wno-ignored-qualifiers
3729 Warn if the return type of a function has a type qualifier
3730 such as @code{const}. For ISO C such a type qualifier has no effect,
3731 since the value returned by a function is not an lvalue.
3732 For C++, the warning is only emitted for scalar types or @code{void}.
3733 ISO C prohibits qualified @code{void} return types on function
3734 definitions, so such return types always receive a warning
3735 even without this option.
3737 This warning is also enabled by @option{-Wextra}.
3742 Warn if the type of @code{main} is suspicious. @code{main} should be
3743 a function with external linkage, returning int, taking either zero
3744 arguments, two, or three arguments of appropriate types. This warning
3745 is enabled by default in C++ and is enabled by either @option{-Wall}
3746 or @option{-Wpedantic}.
3748 @item -Wmissing-braces
3749 @opindex Wmissing-braces
3750 @opindex Wno-missing-braces
3751 Warn if an aggregate or union initializer is not fully bracketed. In
3752 the following example, the initializer for @code{a} is not fully
3753 bracketed, but that for @code{b} is fully bracketed. This warning is
3754 enabled by @option{-Wall} in C.
3757 int a[2][2] = @{ 0, 1, 2, 3 @};
3758 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3761 This warning is enabled by @option{-Wall}.
3763 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3764 @opindex Wmissing-include-dirs
3765 @opindex Wno-missing-include-dirs
3766 Warn if a user-supplied include directory does not exist.
3769 @opindex Wparentheses
3770 @opindex Wno-parentheses
3771 Warn if parentheses are omitted in certain contexts, such
3772 as when there is an assignment in a context where a truth value
3773 is expected, or when operators are nested whose precedence people
3774 often get confused about.
3776 Also warn if a comparison like @code{x<=y<=z} appears; this is
3777 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
3778 interpretation from that of ordinary mathematical notation.
3780 Also warn about constructions where there may be confusion to which
3781 @code{if} statement an @code{else} branch belongs. Here is an example of
3796 In C/C++, every @code{else} branch belongs to the innermost possible
3797 @code{if} statement, which in this example is @code{if (b)}. This is
3798 often not what the programmer expected, as illustrated in the above
3799 example by indentation the programmer chose. When there is the
3800 potential for this confusion, GCC issues a warning when this flag
3801 is specified. To eliminate the warning, add explicit braces around
3802 the innermost @code{if} statement so there is no way the @code{else}
3803 can belong to the enclosing @code{if}. The resulting code
3820 Also warn for dangerous uses of the GNU extension to
3821 @code{?:} with omitted middle operand. When the condition
3822 in the @code{?}: operator is a boolean expression, the omitted value is
3823 always 1. Often programmers expect it to be a value computed
3824 inside the conditional expression instead.
3826 This warning is enabled by @option{-Wall}.
3828 @item -Wsequence-point
3829 @opindex Wsequence-point
3830 @opindex Wno-sequence-point
3831 Warn about code that may have undefined semantics because of violations
3832 of sequence point rules in the C and C++ standards.
3834 The C and C++ standards define the order in which expressions in a C/C++
3835 program are evaluated in terms of @dfn{sequence points}, which represent
3836 a partial ordering between the execution of parts of the program: those
3837 executed before the sequence point, and those executed after it. These
3838 occur after the evaluation of a full expression (one which is not part
3839 of a larger expression), after the evaluation of the first operand of a
3840 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3841 function is called (but after the evaluation of its arguments and the
3842 expression denoting the called function), and in certain other places.
3843 Other than as expressed by the sequence point rules, the order of
3844 evaluation of subexpressions of an expression is not specified. All
3845 these rules describe only a partial order rather than a total order,
3846 since, for example, if two functions are called within one expression
3847 with no sequence point between them, the order in which the functions
3848 are called is not specified. However, the standards committee have
3849 ruled that function calls do not overlap.
3851 It is not specified when between sequence points modifications to the
3852 values of objects take effect. Programs whose behavior depends on this
3853 have undefined behavior; the C and C++ standards specify that ``Between
3854 the previous and next sequence point an object shall have its stored
3855 value modified at most once by the evaluation of an expression.
3856 Furthermore, the prior value shall be read only to determine the value
3857 to be stored.''. If a program breaks these rules, the results on any
3858 particular implementation are entirely unpredictable.
3860 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3861 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3862 diagnosed by this option, and it may give an occasional false positive
3863 result, but in general it has been found fairly effective at detecting
3864 this sort of problem in programs.
3866 The standard is worded confusingly, therefore there is some debate
3867 over the precise meaning of the sequence point rules in subtle cases.
3868 Links to discussions of the problem, including proposed formal
3869 definitions, may be found on the GCC readings page, at
3870 @uref{http://gcc.gnu.org/@/readings.html}.
3872 This warning is enabled by @option{-Wall} for C and C++.
3874 @item -Wno-return-local-addr
3875 @opindex Wno-return-local-addr
3876 @opindex Wreturn-local-addr
3877 Do not warn about returning a pointer (or in C++, a reference) to a
3878 variable that goes out of scope after the function returns.
3881 @opindex Wreturn-type
3882 @opindex Wno-return-type
3883 Warn whenever a function is defined with a return type that defaults
3884 to @code{int}. Also warn about any @code{return} statement with no
3885 return value in a function whose return type is not @code{void}
3886 (falling off the end of the function body is considered returning
3887 without a value), and about a @code{return} statement with an
3888 expression in a function whose return type is @code{void}.
3890 For C++, a function without return type always produces a diagnostic
3891 message, even when @option{-Wno-return-type} is specified. The only
3892 exceptions are @code{main} and functions defined in system headers.
3894 This warning is enabled by @option{-Wall}.
3896 @item -Wshift-count-negative
3897 @opindex Wshift-count-negative
3898 @opindex Wno-shift-count-negative
3899 Warn if shift count is negative. This warning is enabled by default.
3901 @item -Wshift-count-overflow
3902 @opindex Wshift-count-overflow
3903 @opindex Wno-shift-count-overflow
3904 Warn if shift count >= width of type. This warning is enabled by default.
3909 Warn whenever a @code{switch} statement has an index of enumerated type
3910 and lacks a @code{case} for one or more of the named codes of that
3911 enumeration. (The presence of a @code{default} label prevents this
3912 warning.) @code{case} labels outside the enumeration range also
3913 provoke warnings when this option is used (even if there is a
3914 @code{default} label).
3915 This warning is enabled by @option{-Wall}.
3917 @item -Wswitch-default
3918 @opindex Wswitch-default
3919 @opindex Wno-switch-default
3920 Warn whenever a @code{switch} statement does not have a @code{default}
3924 @opindex Wswitch-enum
3925 @opindex Wno-switch-enum
3926 Warn whenever a @code{switch} statement has an index of enumerated type
3927 and lacks a @code{case} for one or more of the named codes of that
3928 enumeration. @code{case} labels outside the enumeration range also
3929 provoke warnings when this option is used. The only difference
3930 between @option{-Wswitch} and this option is that this option gives a
3931 warning about an omitted enumeration code even if there is a
3932 @code{default} label.
3935 @opindex Wswitch-bool
3936 @opindex Wno-switch-bool
3937 Warn whenever a @code{switch} statement has an index of boolean type.
3938 It is possible to suppress this warning by casting the controlling
3939 expression to a type other than @code{bool}. For example:
3942 switch ((int) (a == 4))
3948 This warning is enabled by default for C and C++ programs.
3950 @item -Wsync-nand @r{(C and C++ only)}
3952 @opindex Wno-sync-nand
3953 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3954 built-in functions are used. These functions changed semantics in GCC 4.4.
3958 @opindex Wno-trigraphs
3959 Warn if any trigraphs are encountered that might change the meaning of
3960 the program (trigraphs within comments are not warned about).
3961 This warning is enabled by @option{-Wall}.
3963 @item -Wunused-but-set-parameter
3964 @opindex Wunused-but-set-parameter
3965 @opindex Wno-unused-but-set-parameter
3966 Warn whenever a function parameter is assigned to, but otherwise unused
3967 (aside from its declaration).
3969 To suppress this warning use the @code{unused} attribute
3970 (@pxref{Variable Attributes}).
3972 This warning is also enabled by @option{-Wunused} together with
3975 @item -Wunused-but-set-variable
3976 @opindex Wunused-but-set-variable
3977 @opindex Wno-unused-but-set-variable
3978 Warn whenever a local variable is assigned to, but otherwise unused
3979 (aside from its declaration).
3980 This warning is enabled by @option{-Wall}.
3982 To suppress this warning use the @code{unused} attribute
3983 (@pxref{Variable Attributes}).
3985 This warning is also enabled by @option{-Wunused}, which is enabled
3988 @item -Wunused-function
3989 @opindex Wunused-function
3990 @opindex Wno-unused-function
3991 Warn whenever a static function is declared but not defined or a
3992 non-inline static function is unused.
3993 This warning is enabled by @option{-Wall}.
3995 @item -Wunused-label
3996 @opindex Wunused-label
3997 @opindex Wno-unused-label
3998 Warn whenever a label is declared but not used.
3999 This warning is enabled by @option{-Wall}.
4001 To suppress this warning use the @code{unused} attribute
4002 (@pxref{Variable Attributes}).
4004 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4005 @opindex Wunused-local-typedefs
4006 Warn when a typedef locally defined in a function is not used.
4007 This warning is enabled by @option{-Wall}.
4009 @item -Wunused-parameter
4010 @opindex Wunused-parameter
4011 @opindex Wno-unused-parameter
4012 Warn whenever a function parameter is unused aside from its declaration.
4014 To suppress this warning use the @code{unused} attribute
4015 (@pxref{Variable Attributes}).
4017 @item -Wno-unused-result
4018 @opindex Wunused-result
4019 @opindex Wno-unused-result
4020 Do not warn if a caller of a function marked with attribute
4021 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4022 its return value. The default is @option{-Wunused-result}.
4024 @item -Wunused-variable
4025 @opindex Wunused-variable
4026 @opindex Wno-unused-variable
4027 Warn whenever a local variable or non-constant static variable is unused
4028 aside from its declaration.
4029 This warning is enabled by @option{-Wall}.
4031 To suppress this warning use the @code{unused} attribute
4032 (@pxref{Variable Attributes}).
4034 @item -Wunused-value
4035 @opindex Wunused-value
4036 @opindex Wno-unused-value
4037 Warn whenever a statement computes a result that is explicitly not
4038 used. To suppress this warning cast the unused expression to
4039 @code{void}. This includes an expression-statement or the left-hand
4040 side of a comma expression that contains no side effects. For example,
4041 an expression such as @code{x[i,j]} causes a warning, while
4042 @code{x[(void)i,j]} does not.
4044 This warning is enabled by @option{-Wall}.
4049 All the above @option{-Wunused} options combined.
4051 In order to get a warning about an unused function parameter, you must
4052 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4053 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4055 @item -Wuninitialized
4056 @opindex Wuninitialized
4057 @opindex Wno-uninitialized
4058 Warn if an automatic variable is used without first being initialized
4059 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4060 warn if a non-static reference or non-static @code{const} member
4061 appears in a class without constructors.
4063 If you want to warn about code that uses the uninitialized value of the
4064 variable in its own initializer, use the @option{-Winit-self} option.
4066 These warnings occur for individual uninitialized or clobbered
4067 elements of structure, union or array variables as well as for
4068 variables that are uninitialized or clobbered as a whole. They do
4069 not occur for variables or elements declared @code{volatile}. Because
4070 these warnings depend on optimization, the exact variables or elements
4071 for which there are warnings depends on the precise optimization
4072 options and version of GCC used.
4074 Note that there may be no warning about a variable that is used only
4075 to compute a value that itself is never used, because such
4076 computations may be deleted by data flow analysis before the warnings
4079 @item -Wmaybe-uninitialized
4080 @opindex Wmaybe-uninitialized
4081 @opindex Wno-maybe-uninitialized
4082 For an automatic variable, if there exists a path from the function
4083 entry to a use of the variable that is initialized, but there exist
4084 some other paths for which the variable is not initialized, the compiler
4085 emits a warning if it cannot prove the uninitialized paths are not
4086 executed at run time. These warnings are made optional because GCC is
4087 not smart enough to see all the reasons why the code might be correct
4088 in spite of appearing to have an error. Here is one example of how
4109 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4110 always initialized, but GCC doesn't know this. To suppress the
4111 warning, you need to provide a default case with assert(0) or
4114 @cindex @code{longjmp} warnings
4115 This option also warns when a non-volatile automatic variable might be
4116 changed by a call to @code{longjmp}. These warnings as well are possible
4117 only in optimizing compilation.
4119 The compiler sees only the calls to @code{setjmp}. It cannot know
4120 where @code{longjmp} will be called; in fact, a signal handler could
4121 call it at any point in the code. As a result, you may get a warning
4122 even when there is in fact no problem because @code{longjmp} cannot
4123 in fact be called at the place that would cause a problem.
4125 Some spurious warnings can be avoided if you declare all the functions
4126 you use that never return as @code{noreturn}. @xref{Function
4129 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4131 @item -Wunknown-pragmas
4132 @opindex Wunknown-pragmas
4133 @opindex Wno-unknown-pragmas
4134 @cindex warning for unknown pragmas
4135 @cindex unknown pragmas, warning
4136 @cindex pragmas, warning of unknown
4137 Warn when a @code{#pragma} directive is encountered that is not understood by
4138 GCC@. If this command-line option is used, warnings are even issued
4139 for unknown pragmas in system header files. This is not the case if
4140 the warnings are only enabled by the @option{-Wall} command-line option.
4143 @opindex Wno-pragmas
4145 Do not warn about misuses of pragmas, such as incorrect parameters,
4146 invalid syntax, or conflicts between pragmas. See also
4147 @option{-Wunknown-pragmas}.
4149 @item -Wstrict-aliasing
4150 @opindex Wstrict-aliasing
4151 @opindex Wno-strict-aliasing
4152 This option is only active when @option{-fstrict-aliasing} is active.
4153 It warns about code that might break the strict aliasing rules that the
4154 compiler is using for optimization. The warning does not catch all
4155 cases, but does attempt to catch the more common pitfalls. It is
4156 included in @option{-Wall}.
4157 It is equivalent to @option{-Wstrict-aliasing=3}
4159 @item -Wstrict-aliasing=n
4160 @opindex Wstrict-aliasing=n
4161 This option is only active when @option{-fstrict-aliasing} is active.
4162 It warns about code that might break the strict aliasing rules that the
4163 compiler is using for optimization.
4164 Higher levels correspond to higher accuracy (fewer false positives).
4165 Higher levels also correspond to more effort, similar to the way @option{-O}
4167 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4169 Level 1: Most aggressive, quick, least accurate.
4170 Possibly useful when higher levels
4171 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4172 false negatives. However, it has many false positives.
4173 Warns for all pointer conversions between possibly incompatible types,
4174 even if never dereferenced. Runs in the front end only.
4176 Level 2: Aggressive, quick, not too precise.
4177 May still have many false positives (not as many as level 1 though),
4178 and few false negatives (but possibly more than level 1).
4179 Unlike level 1, it only warns when an address is taken. Warns about
4180 incomplete types. Runs in the front end only.
4182 Level 3 (default for @option{-Wstrict-aliasing}):
4183 Should have very few false positives and few false
4184 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4185 Takes care of the common pun+dereference pattern in the front end:
4186 @code{*(int*)&some_float}.
4187 If optimization is enabled, it also runs in the back end, where it deals
4188 with multiple statement cases using flow-sensitive points-to information.
4189 Only warns when the converted pointer is dereferenced.
4190 Does not warn about incomplete types.
4192 @item -Wstrict-overflow
4193 @itemx -Wstrict-overflow=@var{n}
4194 @opindex Wstrict-overflow
4195 @opindex Wno-strict-overflow
4196 This option is only active when @option{-fstrict-overflow} is active.
4197 It warns about cases where the compiler optimizes based on the
4198 assumption that signed overflow does not occur. Note that it does not
4199 warn about all cases where the code might overflow: it only warns
4200 about cases where the compiler implements some optimization. Thus
4201 this warning depends on the optimization level.
4203 An optimization that assumes that signed overflow does not occur is
4204 perfectly safe if the values of the variables involved are such that
4205 overflow never does, in fact, occur. Therefore this warning can
4206 easily give a false positive: a warning about code that is not
4207 actually a problem. To help focus on important issues, several
4208 warning levels are defined. No warnings are issued for the use of
4209 undefined signed overflow when estimating how many iterations a loop
4210 requires, in particular when determining whether a loop will be
4214 @item -Wstrict-overflow=1
4215 Warn about cases that are both questionable and easy to avoid. For
4216 example, with @option{-fstrict-overflow}, the compiler simplifies
4217 @code{x + 1 > x} to @code{1}. This level of
4218 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4219 are not, and must be explicitly requested.
4221 @item -Wstrict-overflow=2
4222 Also warn about other cases where a comparison is simplified to a
4223 constant. For example: @code{abs (x) >= 0}. This can only be
4224 simplified when @option{-fstrict-overflow} is in effect, because
4225 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4226 zero. @option{-Wstrict-overflow} (with no level) is the same as
4227 @option{-Wstrict-overflow=2}.
4229 @item -Wstrict-overflow=3
4230 Also warn about other cases where a comparison is simplified. For
4231 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4233 @item -Wstrict-overflow=4
4234 Also warn about other simplifications not covered by the above cases.
4235 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4237 @item -Wstrict-overflow=5
4238 Also warn about cases where the compiler reduces the magnitude of a
4239 constant involved in a comparison. For example: @code{x + 2 > y} is
4240 simplified to @code{x + 1 >= y}. This is reported only at the
4241 highest warning level because this simplification applies to many
4242 comparisons, so this warning level gives a very large number of
4246 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
4247 @opindex Wsuggest-attribute=
4248 @opindex Wno-suggest-attribute=
4249 Warn for cases where adding an attribute may be beneficial. The
4250 attributes currently supported are listed below.
4253 @item -Wsuggest-attribute=pure
4254 @itemx -Wsuggest-attribute=const
4255 @itemx -Wsuggest-attribute=noreturn
4256 @opindex Wsuggest-attribute=pure
4257 @opindex Wno-suggest-attribute=pure
4258 @opindex Wsuggest-attribute=const
4259 @opindex Wno-suggest-attribute=const
4260 @opindex Wsuggest-attribute=noreturn
4261 @opindex Wno-suggest-attribute=noreturn
4263 Warn about functions that might be candidates for attributes
4264 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
4265 functions visible in other compilation units or (in the case of @code{pure} and
4266 @code{const}) if it cannot prove that the function returns normally. A function
4267 returns normally if it doesn't contain an infinite loop or return abnormally
4268 by throwing, calling @code{abort} or trapping. This analysis requires option
4269 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
4270 higher. Higher optimization levels improve the accuracy of the analysis.
4272 @item -Wsuggest-attribute=format
4273 @itemx -Wmissing-format-attribute
4274 @opindex Wsuggest-attribute=format
4275 @opindex Wmissing-format-attribute
4276 @opindex Wno-suggest-attribute=format
4277 @opindex Wno-missing-format-attribute
4281 Warn about function pointers that might be candidates for @code{format}
4282 attributes. Note these are only possible candidates, not absolute ones.
4283 GCC guesses that function pointers with @code{format} attributes that
4284 are used in assignment, initialization, parameter passing or return
4285 statements should have a corresponding @code{format} attribute in the
4286 resulting type. I.e.@: the left-hand side of the assignment or
4287 initialization, the type of the parameter variable, or the return type
4288 of the containing function respectively should also have a @code{format}
4289 attribute to avoid the warning.
4291 GCC also warns about function definitions that might be
4292 candidates for @code{format} attributes. Again, these are only
4293 possible candidates. GCC guesses that @code{format} attributes
4294 might be appropriate for any function that calls a function like
4295 @code{vprintf} or @code{vscanf}, but this might not always be the
4296 case, and some functions for which @code{format} attributes are
4297 appropriate may not be detected.
4300 @item -Wsuggest-final-types
4301 @opindex Wno-suggest-final-types
4302 @opindex Wsuggest-final-types
4303 Warn about types with virtual methods where code quality would be improved
4304 if the type were declared with the C++11 @code{final} specifier,
4306 declared in an anonymous namespace. This allows GCC to more aggressively
4307 devirtualize the polymorphic calls. This warning is more effective with link
4308 time optimization, where the information about the class hierarchy graph is
4311 @item -Wsuggest-final-methods
4312 @opindex Wno-suggest-final-methods
4313 @opindex Wsuggest-final-methods
4314 Warn about virtual methods where code quality would be improved if the method
4315 were declared with the C++11 @code{final} specifier,
4316 or, if possible, its type were
4317 declared in an anonymous namespace or with the @code{final} specifier.
4319 more effective with link time optimization, where the information about the
4320 class hierarchy graph is more complete. It is recommended to first consider
4321 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
4324 @item -Wsuggest-override
4325 Warn about overriding virtual functions that are not marked with the override
4328 @item -Warray-bounds
4329 @itemx -Warray-bounds=@var{n}
4330 @opindex Wno-array-bounds
4331 @opindex Warray-bounds
4332 This option is only active when @option{-ftree-vrp} is active
4333 (default for @option{-O2} and above). It warns about subscripts to arrays
4334 that are always out of bounds. This warning is enabled by @option{-Wall}.
4337 @item -Warray-bounds=1
4338 This is the warning level of @option{-Warray-bounds} and is enabled
4339 by @option{-Wall}; higher levels are not, and must be explicitly requested.
4341 @item -Warray-bounds=2
4342 This warning level also warns about out of bounds access for
4343 arrays at the end of a struct and for arrays accessed through
4344 pointers. This warning level may give a larger number of
4345 false positives and is deactivated by default.
4349 @item -Wbool-compare
4350 @opindex Wno-bool-compare
4351 @opindex Wbool-compare
4352 Warn about boolean expression compared with an integer value different from
4353 @code{true}/@code{false}. For instance, the following comparison is
4358 if ((n > 1) == 2) @{ @dots{} @}
4360 This warning is enabled by @option{-Wall}.
4362 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
4363 @opindex Wno-discarded-qualifiers
4364 @opindex Wdiscarded-qualifiers
4365 Do not warn if type qualifiers on pointers are being discarded.
4366 Typically, the compiler warns if a @code{const char *} variable is
4367 passed to a function that takes a @code{char *} parameter. This option
4368 can be used to suppress such a warning.
4370 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
4371 @opindex Wno-discarded-array-qualifiers
4372 @opindex Wdiscarded-array-qualifiers
4373 Do not warn if type qualifiers on arrays which are pointer targets
4374 are being discarded. Typically, the compiler warns if a
4375 @code{const int (*)[]} variable is passed to a function that
4376 takes a @code{int (*)[]} parameter. This option can be used to
4377 suppress such a warning.
4379 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
4380 @opindex Wno-incompatible-pointer-types
4381 @opindex Wincompatible-pointer-types
4382 Do not warn when there is a conversion between pointers that have incompatible
4383 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
4384 which warns for pointer argument passing or assignment with different
4387 @item -Wno-int-conversion @r{(C and Objective-C only)}
4388 @opindex Wno-int-conversion
4389 @opindex Wint-conversion
4390 Do not warn about incompatible integer to pointer and pointer to integer
4391 conversions. This warning is about implicit conversions; for explicit
4392 conversions the warnings @option{-Wno-int-to-pointer-cast} and
4393 @option{-Wno-pointer-to-int-cast} may be used.
4395 @item -Wno-div-by-zero
4396 @opindex Wno-div-by-zero
4397 @opindex Wdiv-by-zero
4398 Do not warn about compile-time integer division by zero. Floating-point
4399 division by zero is not warned about, as it can be a legitimate way of
4400 obtaining infinities and NaNs.
4402 @item -Wsystem-headers
4403 @opindex Wsystem-headers
4404 @opindex Wno-system-headers
4405 @cindex warnings from system headers
4406 @cindex system headers, warnings from
4407 Print warning messages for constructs found in system header files.
4408 Warnings from system headers are normally suppressed, on the assumption
4409 that they usually do not indicate real problems and would only make the
4410 compiler output harder to read. Using this command-line option tells
4411 GCC to emit warnings from system headers as if they occurred in user
4412 code. However, note that using @option{-Wall} in conjunction with this
4413 option does @emph{not} warn about unknown pragmas in system
4414 headers---for that, @option{-Wunknown-pragmas} must also be used.
4417 @opindex Wtrampolines
4418 @opindex Wno-trampolines
4419 Warn about trampolines generated for pointers to nested functions.
4420 A trampoline is a small piece of data or code that is created at run
4421 time on the stack when the address of a nested function is taken, and is
4422 used to call the nested function indirectly. For some targets, it is
4423 made up of data only and thus requires no special treatment. But, for
4424 most targets, it is made up of code and thus requires the stack to be
4425 made executable in order for the program to work properly.
4428 @opindex Wfloat-equal
4429 @opindex Wno-float-equal
4430 Warn if floating-point values are used in equality comparisons.
4432 The idea behind this is that sometimes it is convenient (for the
4433 programmer) to consider floating-point values as approximations to
4434 infinitely precise real numbers. If you are doing this, then you need
4435 to compute (by analyzing the code, or in some other way) the maximum or
4436 likely maximum error that the computation introduces, and allow for it
4437 when performing comparisons (and when producing output, but that's a
4438 different problem). In particular, instead of testing for equality, you
4439 should check to see whether the two values have ranges that overlap; and
4440 this is done with the relational operators, so equality comparisons are
4443 @item -Wtraditional @r{(C and Objective-C only)}
4444 @opindex Wtraditional
4445 @opindex Wno-traditional
4446 Warn about certain constructs that behave differently in traditional and
4447 ISO C@. Also warn about ISO C constructs that have no traditional C
4448 equivalent, and/or problematic constructs that should be avoided.
4452 Macro parameters that appear within string literals in the macro body.
4453 In traditional C macro replacement takes place within string literals,
4454 but in ISO C it does not.
4457 In traditional C, some preprocessor directives did not exist.
4458 Traditional preprocessors only considered a line to be a directive
4459 if the @samp{#} appeared in column 1 on the line. Therefore
4460 @option{-Wtraditional} warns about directives that traditional C
4461 understands but ignores because the @samp{#} does not appear as the
4462 first character on the line. It also suggests you hide directives like
4463 @code{#pragma} not understood by traditional C by indenting them. Some
4464 traditional implementations do not recognize @code{#elif}, so this option
4465 suggests avoiding it altogether.
4468 A function-like macro that appears without arguments.
4471 The unary plus operator.
4474 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
4475 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
4476 constants.) Note, these suffixes appear in macros defined in the system
4477 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
4478 Use of these macros in user code might normally lead to spurious
4479 warnings, however GCC's integrated preprocessor has enough context to
4480 avoid warning in these cases.
4483 A function declared external in one block and then used after the end of
4487 A @code{switch} statement has an operand of type @code{long}.
4490 A non-@code{static} function declaration follows a @code{static} one.
4491 This construct is not accepted by some traditional C compilers.
4494 The ISO type of an integer constant has a different width or
4495 signedness from its traditional type. This warning is only issued if
4496 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
4497 typically represent bit patterns, are not warned about.
4500 Usage of ISO string concatenation is detected.
4503 Initialization of automatic aggregates.
4506 Identifier conflicts with labels. Traditional C lacks a separate
4507 namespace for labels.
4510 Initialization of unions. If the initializer is zero, the warning is
4511 omitted. This is done under the assumption that the zero initializer in
4512 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
4513 initializer warnings and relies on default initialization to zero in the
4517 Conversions by prototypes between fixed/floating-point values and vice
4518 versa. The absence of these prototypes when compiling with traditional
4519 C causes serious problems. This is a subset of the possible
4520 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
4523 Use of ISO C style function definitions. This warning intentionally is
4524 @emph{not} issued for prototype declarations or variadic functions
4525 because these ISO C features appear in your code when using
4526 libiberty's traditional C compatibility macros, @code{PARAMS} and
4527 @code{VPARAMS}. This warning is also bypassed for nested functions
4528 because that feature is already a GCC extension and thus not relevant to
4529 traditional C compatibility.
4532 @item -Wtraditional-conversion @r{(C and Objective-C only)}
4533 @opindex Wtraditional-conversion
4534 @opindex Wno-traditional-conversion
4535 Warn if a prototype causes a type conversion that is different from what
4536 would happen to the same argument in the absence of a prototype. This
4537 includes conversions of fixed point to floating and vice versa, and
4538 conversions changing the width or signedness of a fixed-point argument
4539 except when the same as the default promotion.
4541 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
4542 @opindex Wdeclaration-after-statement
4543 @opindex Wno-declaration-after-statement
4544 Warn when a declaration is found after a statement in a block. This
4545 construct, known from C++, was introduced with ISO C99 and is by default
4546 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
4551 Warn if an undefined identifier is evaluated in an @code{#if} directive.
4553 @item -Wno-endif-labels
4554 @opindex Wno-endif-labels
4555 @opindex Wendif-labels
4556 Do not warn whenever an @code{#else} or an @code{#endif} are followed by text.
4561 Warn whenever a local variable or type declaration shadows another
4562 variable, parameter, type, class member (in C++), or instance variable
4563 (in Objective-C) or whenever a built-in function is shadowed. Note
4564 that in C++, the compiler warns if a local variable shadows an
4565 explicit typedef, but not if it shadows a struct/class/enum.
4567 @item -Wno-shadow-ivar @r{(Objective-C only)}
4568 @opindex Wno-shadow-ivar
4569 @opindex Wshadow-ivar
4570 Do not warn whenever a local variable shadows an instance variable in an
4573 @item -Wlarger-than=@var{len}
4574 @opindex Wlarger-than=@var{len}
4575 @opindex Wlarger-than-@var{len}
4576 Warn whenever an object of larger than @var{len} bytes is defined.
4578 @item -Wframe-larger-than=@var{len}
4579 @opindex Wframe-larger-than
4580 Warn if the size of a function frame is larger than @var{len} bytes.
4581 The computation done to determine the stack frame size is approximate
4582 and not conservative.
4583 The actual requirements may be somewhat greater than @var{len}
4584 even if you do not get a warning. In addition, any space allocated
4585 via @code{alloca}, variable-length arrays, or related constructs
4586 is not included by the compiler when determining
4587 whether or not to issue a warning.
4589 @item -Wno-free-nonheap-object
4590 @opindex Wno-free-nonheap-object
4591 @opindex Wfree-nonheap-object
4592 Do not warn when attempting to free an object that was not allocated
4595 @item -Wstack-usage=@var{len}
4596 @opindex Wstack-usage
4597 Warn if the stack usage of a function might be larger than @var{len} bytes.
4598 The computation done to determine the stack usage is conservative.
4599 Any space allocated via @code{alloca}, variable-length arrays, or related
4600 constructs is included by the compiler when determining whether or not to
4603 The message is in keeping with the output of @option{-fstack-usage}.
4607 If the stack usage is fully static but exceeds the specified amount, it's:
4610 warning: stack usage is 1120 bytes
4613 If the stack usage is (partly) dynamic but bounded, it's:
4616 warning: stack usage might be 1648 bytes
4619 If the stack usage is (partly) dynamic and not bounded, it's:
4622 warning: stack usage might be unbounded
4626 @item -Wunsafe-loop-optimizations
4627 @opindex Wunsafe-loop-optimizations
4628 @opindex Wno-unsafe-loop-optimizations
4629 Warn if the loop cannot be optimized because the compiler cannot
4630 assume anything on the bounds of the loop indices. With
4631 @option{-funsafe-loop-optimizations} warn if the compiler makes
4634 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4635 @opindex Wno-pedantic-ms-format
4636 @opindex Wpedantic-ms-format
4637 When used in combination with @option{-Wformat}
4638 and @option{-pedantic} without GNU extensions, this option
4639 disables the warnings about non-ISO @code{printf} / @code{scanf} format
4640 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
4641 which depend on the MS runtime.
4643 @item -Wpointer-arith
4644 @opindex Wpointer-arith
4645 @opindex Wno-pointer-arith
4646 Warn about anything that depends on the ``size of'' a function type or
4647 of @code{void}. GNU C assigns these types a size of 1, for
4648 convenience in calculations with @code{void *} pointers and pointers
4649 to functions. In C++, warn also when an arithmetic operation involves
4650 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
4653 @opindex Wtype-limits
4654 @opindex Wno-type-limits
4655 Warn if a comparison is always true or always false due to the limited
4656 range of the data type, but do not warn for constant expressions. For
4657 example, warn if an unsigned variable is compared against zero with
4658 @code{<} or @code{>=}. This warning is also enabled by
4661 @item -Wbad-function-cast @r{(C and Objective-C only)}
4662 @opindex Wbad-function-cast
4663 @opindex Wno-bad-function-cast
4664 Warn when a function call is cast to a non-matching type.
4665 For example, warn if a call to a function returning an integer type
4666 is cast to a pointer type.
4668 @item -Wc90-c99-compat @r{(C and Objective-C only)}
4669 @opindex Wc90-c99-compat
4670 @opindex Wno-c90-c99-compat
4671 Warn about features not present in ISO C90, but present in ISO C99.
4672 For instance, warn about use of variable length arrays, @code{long long}
4673 type, @code{bool} type, compound literals, designated initializers, and so
4674 on. This option is independent of the standards mode. Warnings are disabled
4675 in the expression that follows @code{__extension__}.
4677 @item -Wc99-c11-compat @r{(C and Objective-C only)}
4678 @opindex Wc99-c11-compat
4679 @opindex Wno-c99-c11-compat
4680 Warn about features not present in ISO C99, but present in ISO C11.
4681 For instance, warn about use of anonymous structures and unions,
4682 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
4683 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
4684 and so on. This option is independent of the standards mode. Warnings are
4685 disabled in the expression that follows @code{__extension__}.
4687 @item -Wc++-compat @r{(C and Objective-C only)}
4688 @opindex Wc++-compat
4689 Warn about ISO C constructs that are outside of the common subset of
4690 ISO C and ISO C++, e.g.@: request for implicit conversion from
4691 @code{void *} to a pointer to non-@code{void} type.
4693 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4694 @opindex Wc++11-compat
4695 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4696 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4697 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4698 enabled by @option{-Wall}.
4700 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
4701 @opindex Wc++14-compat
4702 Warn about C++ constructs whose meaning differs between ISO C++ 2011
4703 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
4707 @opindex Wno-cast-qual
4708 Warn whenever a pointer is cast so as to remove a type qualifier from
4709 the target type. For example, warn if a @code{const char *} is cast
4710 to an ordinary @code{char *}.
4712 Also warn when making a cast that introduces a type qualifier in an
4713 unsafe way. For example, casting @code{char **} to @code{const char **}
4714 is unsafe, as in this example:
4717 /* p is char ** value. */
4718 const char **q = (const char **) p;
4719 /* Assignment of readonly string to const char * is OK. */
4721 /* Now char** pointer points to read-only memory. */
4726 @opindex Wcast-align
4727 @opindex Wno-cast-align
4728 Warn whenever a pointer is cast such that the required alignment of the
4729 target is increased. For example, warn if a @code{char *} is cast to
4730 an @code{int *} on machines where integers can only be accessed at
4731 two- or four-byte boundaries.
4733 @item -Wwrite-strings
4734 @opindex Wwrite-strings
4735 @opindex Wno-write-strings
4736 When compiling C, give string constants the type @code{const
4737 char[@var{length}]} so that copying the address of one into a
4738 non-@code{const} @code{char *} pointer produces a warning. These
4739 warnings help you find at compile time code that can try to write
4740 into a string constant, but only if you have been very careful about
4741 using @code{const} in declarations and prototypes. Otherwise, it is
4742 just a nuisance. This is why we did not make @option{-Wall} request
4745 When compiling C++, warn about the deprecated conversion from string
4746 literals to @code{char *}. This warning is enabled by default for C++
4751 @opindex Wno-clobbered
4752 Warn for variables that might be changed by @code{longjmp} or
4753 @code{vfork}. This warning is also enabled by @option{-Wextra}.
4755 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
4756 @opindex Wconditionally-supported
4757 @opindex Wno-conditionally-supported
4758 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
4761 @opindex Wconversion
4762 @opindex Wno-conversion
4763 Warn for implicit conversions that may alter a value. This includes
4764 conversions between real and integer, like @code{abs (x)} when
4765 @code{x} is @code{double}; conversions between signed and unsigned,
4766 like @code{unsigned ui = -1}; and conversions to smaller types, like
4767 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4768 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4769 changed by the conversion like in @code{abs (2.0)}. Warnings about
4770 conversions between signed and unsigned integers can be disabled by
4771 using @option{-Wno-sign-conversion}.
4773 For C++, also warn for confusing overload resolution for user-defined
4774 conversions; and conversions that never use a type conversion
4775 operator: conversions to @code{void}, the same type, a base class or a
4776 reference to them. Warnings about conversions between signed and
4777 unsigned integers are disabled by default in C++ unless
4778 @option{-Wsign-conversion} is explicitly enabled.
4780 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4781 @opindex Wconversion-null
4782 @opindex Wno-conversion-null
4783 Do not warn for conversions between @code{NULL} and non-pointer
4784 types. @option{-Wconversion-null} is enabled by default.
4786 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
4787 @opindex Wzero-as-null-pointer-constant
4788 @opindex Wno-zero-as-null-pointer-constant
4789 Warn when a literal '0' is used as null pointer constant. This can
4790 be useful to facilitate the conversion to @code{nullptr} in C++11.
4794 @opindex Wno-date-time
4795 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
4796 are encountered as they might prevent bit-wise-identical reproducible
4799 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
4800 @opindex Wdelete-incomplete
4801 @opindex Wno-delete-incomplete
4802 Warn when deleting a pointer to incomplete type, which may cause
4803 undefined behavior at runtime. This warning is enabled by default.
4805 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
4806 @opindex Wuseless-cast
4807 @opindex Wno-useless-cast
4808 Warn when an expression is casted to its own type.
4811 @opindex Wempty-body
4812 @opindex Wno-empty-body
4813 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
4814 while} statement. This warning is also enabled by @option{-Wextra}.
4816 @item -Wenum-compare
4817 @opindex Wenum-compare
4818 @opindex Wno-enum-compare
4819 Warn about a comparison between values of different enumerated types.
4820 In C++ enumeral mismatches in conditional expressions are also
4821 diagnosed and the warning is enabled by default. In C this warning is
4822 enabled by @option{-Wall}.
4824 @item -Wjump-misses-init @r{(C, Objective-C only)}
4825 @opindex Wjump-misses-init
4826 @opindex Wno-jump-misses-init
4827 Warn if a @code{goto} statement or a @code{switch} statement jumps
4828 forward across the initialization of a variable, or jumps backward to a
4829 label after the variable has been initialized. This only warns about
4830 variables that are initialized when they are declared. This warning is
4831 only supported for C and Objective-C; in C++ this sort of branch is an
4834 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4835 can be disabled with the @option{-Wno-jump-misses-init} option.
4837 @item -Wsign-compare
4838 @opindex Wsign-compare
4839 @opindex Wno-sign-compare
4840 @cindex warning for comparison of signed and unsigned values
4841 @cindex comparison of signed and unsigned values, warning
4842 @cindex signed and unsigned values, comparison warning
4843 Warn when a comparison between signed and unsigned values could produce
4844 an incorrect result when the signed value is converted to unsigned.
4845 This warning is also enabled by @option{-Wextra}; to get the other warnings
4846 of @option{-Wextra} without this warning, use @option{-Wextra -Wno-sign-compare}.
4848 @item -Wsign-conversion
4849 @opindex Wsign-conversion
4850 @opindex Wno-sign-conversion
4851 Warn for implicit conversions that may change the sign of an integer
4852 value, like assigning a signed integer expression to an unsigned
4853 integer variable. An explicit cast silences the warning. In C, this
4854 option is enabled also by @option{-Wconversion}.
4856 @item -Wfloat-conversion
4857 @opindex Wfloat-conversion
4858 @opindex Wno-float-conversion
4859 Warn for implicit conversions that reduce the precision of a real value.
4860 This includes conversions from real to integer, and from higher precision
4861 real to lower precision real values. This option is also enabled by
4862 @option{-Wconversion}.
4864 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
4865 @opindex Wsized-deallocation
4866 @opindex Wno-sized-deallocation
4867 Warn about a definition of an unsized deallocation function
4869 void operator delete (void *) noexcept;
4870 void operator delete[] (void *) noexcept;
4872 without a definition of the corresponding sized deallocation function
4874 void operator delete (void *, std::size_t) noexcept;
4875 void operator delete[] (void *, std::size_t) noexcept;
4877 or vice versa. Enabled by @option{-Wextra} along with
4878 @option{-fsized-deallocation}.
4880 @item -Wsizeof-pointer-memaccess
4881 @opindex Wsizeof-pointer-memaccess
4882 @opindex Wno-sizeof-pointer-memaccess
4883 Warn for suspicious length parameters to certain string and memory built-in
4884 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
4885 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
4886 but a pointer, and suggests a possible fix, or about
4887 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
4890 @item -Wsizeof-array-argument
4891 @opindex Wsizeof-array-argument
4892 @opindex Wno-sizeof-array-argument
4893 Warn when the @code{sizeof} operator is applied to a parameter that is
4894 declared as an array in a function definition. This warning is enabled by
4895 default for C and C++ programs.
4897 @item -Wmemset-transposed-args
4898 @opindex Wmemset-transposed-args
4899 @opindex Wno-memset-transposed-args
4900 Warn for suspicious calls to the @code{memset} built-in function, if the
4901 second argument is not zero and the third argument is zero. This warns e.g.@
4902 about @code{memset (buf, sizeof buf, 0)} where most probably
4903 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
4904 is only emitted if the third argument is literal zero, if it is some expression
4905 that is folded to zero, or e.g. a cast of zero to some type etc., it
4906 is far less likely that user has mistakenly exchanged the arguments and
4907 no warning is emitted. This warning is enabled by @option{-Wall}.
4911 @opindex Wno-address
4912 Warn about suspicious uses of memory addresses. These include using
4913 the address of a function in a conditional expression, such as
4914 @code{void func(void); if (func)}, and comparisons against the memory
4915 address of a string literal, such as @code{if (x == "abc")}. Such
4916 uses typically indicate a programmer error: the address of a function
4917 always evaluates to true, so their use in a conditional usually
4918 indicate that the programmer forgot the parentheses in a function
4919 call; and comparisons against string literals result in unspecified
4920 behavior and are not portable in C, so they usually indicate that the
4921 programmer intended to use @code{strcmp}. This warning is enabled by
4925 @opindex Wlogical-op
4926 @opindex Wno-logical-op
4927 Warn about suspicious uses of logical operators in expressions.
4928 This includes using logical operators in contexts where a
4929 bit-wise operator is likely to be expected.
4931 @item -Wlogical-not-parentheses
4932 @opindex Wlogical-not-parentheses
4933 @opindex Wno-logical-not-parentheses
4934 Warn about logical not used on the left hand side operand of a comparison.
4935 This option does not warn if the RHS operand is of a boolean type. Its
4936 purpose is to detect suspicious code like the following:
4940 if (!a > 1) @{ @dots{} @}
4943 It is possible to suppress the warning by wrapping the LHS into
4946 if ((!a) > 1) @{ @dots{} @}
4949 This warning is enabled by @option{-Wall}.
4951 @item -Waggregate-return
4952 @opindex Waggregate-return
4953 @opindex Wno-aggregate-return
4954 Warn if any functions that return structures or unions are defined or
4955 called. (In languages where you can return an array, this also elicits
4958 @item -Wno-aggressive-loop-optimizations
4959 @opindex Wno-aggressive-loop-optimizations
4960 @opindex Waggressive-loop-optimizations
4961 Warn if in a loop with constant number of iterations the compiler detects
4962 undefined behavior in some statement during one or more of the iterations.
4964 @item -Wno-attributes
4965 @opindex Wno-attributes
4966 @opindex Wattributes
4967 Do not warn if an unexpected @code{__attribute__} is used, such as
4968 unrecognized attributes, function attributes applied to variables,
4969 etc. This does not stop errors for incorrect use of supported
4972 @item -Wno-builtin-macro-redefined
4973 @opindex Wno-builtin-macro-redefined
4974 @opindex Wbuiltin-macro-redefined
4975 Do not warn if certain built-in macros are redefined. This suppresses
4976 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4977 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4979 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4980 @opindex Wstrict-prototypes
4981 @opindex Wno-strict-prototypes
4982 Warn if a function is declared or defined without specifying the
4983 argument types. (An old-style function definition is permitted without
4984 a warning if preceded by a declaration that specifies the argument
4987 @item -Wold-style-declaration @r{(C and Objective-C only)}
4988 @opindex Wold-style-declaration
4989 @opindex Wno-old-style-declaration
4990 Warn for obsolescent usages, according to the C Standard, in a
4991 declaration. For example, warn if storage-class specifiers like
4992 @code{static} are not the first things in a declaration. This warning
4993 is also enabled by @option{-Wextra}.
4995 @item -Wold-style-definition @r{(C and Objective-C only)}
4996 @opindex Wold-style-definition
4997 @opindex Wno-old-style-definition
4998 Warn if an old-style function definition is used. A warning is given
4999 even if there is a previous prototype.
5001 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
5002 @opindex Wmissing-parameter-type
5003 @opindex Wno-missing-parameter-type
5004 A function parameter is declared without a type specifier in K&R-style
5011 This warning is also enabled by @option{-Wextra}.
5013 @item -Wmissing-prototypes @r{(C and Objective-C only)}
5014 @opindex Wmissing-prototypes
5015 @opindex Wno-missing-prototypes
5016 Warn if a global function is defined without a previous prototype
5017 declaration. This warning is issued even if the definition itself
5018 provides a prototype. Use this option to detect global functions
5019 that do not have a matching prototype declaration in a header file.
5020 This option is not valid for C++ because all function declarations
5021 provide prototypes and a non-matching declaration declares an
5022 overload rather than conflict with an earlier declaration.
5023 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
5025 @item -Wmissing-declarations
5026 @opindex Wmissing-declarations
5027 @opindex Wno-missing-declarations
5028 Warn if a global function is defined without a previous declaration.
5029 Do so even if the definition itself provides a prototype.
5030 Use this option to detect global functions that are not declared in
5031 header files. In C, no warnings are issued for functions with previous
5032 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
5033 missing prototypes. In C++, no warnings are issued for function templates,
5034 or for inline functions, or for functions in anonymous namespaces.
5036 @item -Wmissing-field-initializers
5037 @opindex Wmissing-field-initializers
5038 @opindex Wno-missing-field-initializers
5042 Warn if a structure's initializer has some fields missing. For
5043 example, the following code causes such a warning, because
5044 @code{x.h} is implicitly zero:
5047 struct s @{ int f, g, h; @};
5048 struct s x = @{ 3, 4 @};
5051 This option does not warn about designated initializers, so the following
5052 modification does not trigger a warning:
5055 struct s @{ int f, g, h; @};
5056 struct s x = @{ .f = 3, .g = 4 @};
5059 In C++ this option does not warn either about the empty @{ @}
5060 initializer, for example:
5063 struct s @{ int f, g, h; @};
5067 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
5068 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
5070 @item -Wno-multichar
5071 @opindex Wno-multichar
5073 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
5074 Usually they indicate a typo in the user's code, as they have
5075 implementation-defined values, and should not be used in portable code.
5077 @item -Wnormalized@r{[}=@r{<}none@r{|}id@r{|}nfc@r{|}nfkc@r{>]}
5078 @opindex Wnormalized=
5079 @opindex Wnormalized
5080 @opindex Wno-normalized
5083 @cindex character set, input normalization
5084 In ISO C and ISO C++, two identifiers are different if they are
5085 different sequences of characters. However, sometimes when characters
5086 outside the basic ASCII character set are used, you can have two
5087 different character sequences that look the same. To avoid confusion,
5088 the ISO 10646 standard sets out some @dfn{normalization rules} which
5089 when applied ensure that two sequences that look the same are turned into
5090 the same sequence. GCC can warn you if you are using identifiers that
5091 have not been normalized; this option controls that warning.
5093 There are four levels of warning supported by GCC@. The default is
5094 @option{-Wnormalized=nfc}, which warns about any identifier that is
5095 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
5096 recommended form for most uses. It is equivalent to
5097 @option{-Wnormalized}.
5099 Unfortunately, there are some characters allowed in identifiers by
5100 ISO C and ISO C++ that, when turned into NFC, are not allowed in
5101 identifiers. That is, there's no way to use these symbols in portable
5102 ISO C or C++ and have all your identifiers in NFC@.
5103 @option{-Wnormalized=id} suppresses the warning for these characters.
5104 It is hoped that future versions of the standards involved will correct
5105 this, which is why this option is not the default.
5107 You can switch the warning off for all characters by writing
5108 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
5109 only do this if you are using some other normalization scheme (like
5110 ``D''), because otherwise you can easily create bugs that are
5111 literally impossible to see.
5113 Some characters in ISO 10646 have distinct meanings but look identical
5114 in some fonts or display methodologies, especially once formatting has
5115 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
5116 LETTER N'', displays just like a regular @code{n} that has been
5117 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
5118 normalization scheme to convert all these into a standard form as
5119 well, and GCC warns if your code is not in NFKC if you use
5120 @option{-Wnormalized=nfkc}. This warning is comparable to warning
5121 about every identifier that contains the letter O because it might be
5122 confused with the digit 0, and so is not the default, but may be
5123 useful as a local coding convention if the programming environment
5124 cannot be fixed to display these characters distinctly.
5126 @item -Wno-deprecated
5127 @opindex Wno-deprecated
5128 @opindex Wdeprecated
5129 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
5131 @item -Wno-deprecated-declarations
5132 @opindex Wno-deprecated-declarations
5133 @opindex Wdeprecated-declarations
5134 Do not warn about uses of functions (@pxref{Function Attributes}),
5135 variables (@pxref{Variable Attributes}), and types (@pxref{Type
5136 Attributes}) marked as deprecated by using the @code{deprecated}
5140 @opindex Wno-overflow
5142 Do not warn about compile-time overflow in constant expressions.
5147 Warn about One Definition Rule violations during link-time optimization.
5148 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
5151 @opindex Wopenm-simd
5152 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
5153 simd directive set by user. The @option{-fsimd-cost-model=unlimited} can
5154 be used to relax the cost model.
5156 @item -Woverride-init @r{(C and Objective-C only)}
5157 @opindex Woverride-init
5158 @opindex Wno-override-init
5162 Warn if an initialized field without side effects is overridden when
5163 using designated initializers (@pxref{Designated Inits, , Designated
5166 This warning is included in @option{-Wextra}. To get other
5167 @option{-Wextra} warnings without this one, use @option{-Wextra
5168 -Wno-override-init}.
5173 Warn if a structure is given the packed attribute, but the packed
5174 attribute has no effect on the layout or size of the structure.
5175 Such structures may be mis-aligned for little benefit. For
5176 instance, in this code, the variable @code{f.x} in @code{struct bar}
5177 is misaligned even though @code{struct bar} does not itself
5178 have the packed attribute:
5185 @} __attribute__((packed));
5193 @item -Wpacked-bitfield-compat
5194 @opindex Wpacked-bitfield-compat
5195 @opindex Wno-packed-bitfield-compat
5196 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
5197 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
5198 the change can lead to differences in the structure layout. GCC
5199 informs you when the offset of such a field has changed in GCC 4.4.
5200 For example there is no longer a 4-bit padding between field @code{a}
5201 and @code{b} in this structure:
5208 @} __attribute__ ((packed));
5211 This warning is enabled by default. Use
5212 @option{-Wno-packed-bitfield-compat} to disable this warning.
5217 Warn if padding is included in a structure, either to align an element
5218 of the structure or to align the whole structure. Sometimes when this
5219 happens it is possible to rearrange the fields of the structure to
5220 reduce the padding and so make the structure smaller.
5222 @item -Wredundant-decls
5223 @opindex Wredundant-decls
5224 @opindex Wno-redundant-decls
5225 Warn if anything is declared more than once in the same scope, even in
5226 cases where multiple declaration is valid and changes nothing.
5228 @item -Wnested-externs @r{(C and Objective-C only)}
5229 @opindex Wnested-externs
5230 @opindex Wno-nested-externs
5231 Warn if an @code{extern} declaration is encountered within a function.
5233 @item -Wno-inherited-variadic-ctor
5234 @opindex Winherited-variadic-ctor
5235 @opindex Wno-inherited-variadic-ctor
5236 Suppress warnings about use of C++11 inheriting constructors when the
5237 base class inherited from has a C variadic constructor; the warning is
5238 on by default because the ellipsis is not inherited.
5243 Warn if a function that is declared as inline cannot be inlined.
5244 Even with this option, the compiler does not warn about failures to
5245 inline functions declared in system headers.
5247 The compiler uses a variety of heuristics to determine whether or not
5248 to inline a function. For example, the compiler takes into account
5249 the size of the function being inlined and the amount of inlining
5250 that has already been done in the current function. Therefore,
5251 seemingly insignificant changes in the source program can cause the
5252 warnings produced by @option{-Winline} to appear or disappear.
5254 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
5255 @opindex Wno-invalid-offsetof
5256 @opindex Winvalid-offsetof
5257 Suppress warnings from applying the @code{offsetof} macro to a non-POD
5258 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
5259 to a non-standard-layout type is undefined. In existing C++ implementations,
5260 however, @code{offsetof} typically gives meaningful results.
5261 This flag is for users who are aware that they are
5262 writing nonportable code and who have deliberately chosen to ignore the
5265 The restrictions on @code{offsetof} may be relaxed in a future version
5266 of the C++ standard.
5268 @item -Wno-int-to-pointer-cast
5269 @opindex Wno-int-to-pointer-cast
5270 @opindex Wint-to-pointer-cast
5271 Suppress warnings from casts to pointer type of an integer of a
5272 different size. In C++, casting to a pointer type of smaller size is
5273 an error. @option{Wint-to-pointer-cast} is enabled by default.
5276 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
5277 @opindex Wno-pointer-to-int-cast
5278 @opindex Wpointer-to-int-cast
5279 Suppress warnings from casts from a pointer to an integer type of a
5283 @opindex Winvalid-pch
5284 @opindex Wno-invalid-pch
5285 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
5286 the search path but can't be used.
5290 @opindex Wno-long-long
5291 Warn if @code{long long} type is used. This is enabled by either
5292 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
5293 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
5295 @item -Wvariadic-macros
5296 @opindex Wvariadic-macros
5297 @opindex Wno-variadic-macros
5298 Warn if variadic macros are used in ISO C90 mode, or if the GNU
5299 alternate syntax is used in ISO C99 mode. This is enabled by either
5300 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
5301 messages, use @option{-Wno-variadic-macros}.
5305 @opindex Wno-varargs
5306 Warn upon questionable usage of the macros used to handle variable
5307 arguments like @code{va_start}. This is default. To inhibit the
5308 warning messages, use @option{-Wno-varargs}.
5310 @item -Wvector-operation-performance
5311 @opindex Wvector-operation-performance
5312 @opindex Wno-vector-operation-performance
5313 Warn if vector operation is not implemented via SIMD capabilities of the
5314 architecture. Mainly useful for the performance tuning.
5315 Vector operation can be implemented @code{piecewise}, which means that the
5316 scalar operation is performed on every vector element;
5317 @code{in parallel}, which means that the vector operation is implemented
5318 using scalars of wider type, which normally is more performance efficient;
5319 and @code{as a single scalar}, which means that vector fits into a
5322 @item -Wno-virtual-move-assign
5323 @opindex Wvirtual-move-assign
5324 @opindex Wno-virtual-move-assign
5325 Suppress warnings about inheriting from a virtual base with a
5326 non-trivial C++11 move assignment operator. This is dangerous because
5327 if the virtual base is reachable along more than one path, it is
5328 moved multiple times, which can mean both objects end up in the
5329 moved-from state. If the move assignment operator is written to avoid
5330 moving from a moved-from object, this warning can be disabled.
5335 Warn if variable length array is used in the code.
5336 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
5337 the variable length array.
5339 @item -Wvolatile-register-var
5340 @opindex Wvolatile-register-var
5341 @opindex Wno-volatile-register-var
5342 Warn if a register variable is declared volatile. The volatile
5343 modifier does not inhibit all optimizations that may eliminate reads
5344 and/or writes to register variables. This warning is enabled by
5347 @item -Wdisabled-optimization
5348 @opindex Wdisabled-optimization
5349 @opindex Wno-disabled-optimization
5350 Warn if a requested optimization pass is disabled. This warning does
5351 not generally indicate that there is anything wrong with your code; it
5352 merely indicates that GCC's optimizers are unable to handle the code
5353 effectively. Often, the problem is that your code is too big or too
5354 complex; GCC refuses to optimize programs when the optimization
5355 itself is likely to take inordinate amounts of time.
5357 @item -Wpointer-sign @r{(C and Objective-C only)}
5358 @opindex Wpointer-sign
5359 @opindex Wno-pointer-sign
5360 Warn for pointer argument passing or assignment with different signedness.
5361 This option is only supported for C and Objective-C@. It is implied by
5362 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
5363 @option{-Wno-pointer-sign}.
5365 @item -Wstack-protector
5366 @opindex Wstack-protector
5367 @opindex Wno-stack-protector
5368 This option is only active when @option{-fstack-protector} is active. It
5369 warns about functions that are not protected against stack smashing.
5371 @item -Woverlength-strings
5372 @opindex Woverlength-strings
5373 @opindex Wno-overlength-strings
5374 Warn about string constants that are longer than the ``minimum
5375 maximum'' length specified in the C standard. Modern compilers
5376 generally allow string constants that are much longer than the
5377 standard's minimum limit, but very portable programs should avoid
5378 using longer strings.
5380 The limit applies @emph{after} string constant concatenation, and does
5381 not count the trailing NUL@. In C90, the limit was 509 characters; in
5382 C99, it was raised to 4095. C++98 does not specify a normative
5383 minimum maximum, so we do not diagnose overlength strings in C++@.
5385 This option is implied by @option{-Wpedantic}, and can be disabled with
5386 @option{-Wno-overlength-strings}.
5388 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
5389 @opindex Wunsuffixed-float-constants
5391 Issue a warning for any floating constant that does not have
5392 a suffix. When used together with @option{-Wsystem-headers} it
5393 warns about such constants in system header files. This can be useful
5394 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
5395 from the decimal floating-point extension to C99.
5397 @item -Wno-designated-init @r{(C and Objective-C only)}
5398 Suppress warnings when a positional initializer is used to initialize
5399 a structure that has been marked with the @code{designated_init}
5404 @node Debugging Options
5405 @section Options for Debugging Your Program or GCC
5406 @cindex options, debugging
5407 @cindex debugging information options
5409 GCC has various special options that are used for debugging
5410 either your program or GCC:
5415 Produce debugging information in the operating system's native format
5416 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
5419 On most systems that use stabs format, @option{-g} enables use of extra
5420 debugging information that only GDB can use; this extra information
5421 makes debugging work better in GDB but probably makes other debuggers
5423 refuse to read the program. If you want to control for certain whether
5424 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
5425 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
5427 GCC allows you to use @option{-g} with
5428 @option{-O}. The shortcuts taken by optimized code may occasionally
5429 produce surprising results: some variables you declared may not exist
5430 at all; flow of control may briefly move where you did not expect it;
5431 some statements may not be executed because they compute constant
5432 results or their values are already at hand; some statements may
5433 execute in different places because they have been moved out of loops.
5435 Nevertheless it proves possible to debug optimized output. This makes
5436 it reasonable to use the optimizer for programs that might have bugs.
5438 The following options are useful when GCC is generated with the
5439 capability for more than one debugging format.
5442 @opindex gsplit-dwarf
5443 Separate as much dwarf debugging information as possible into a
5444 separate output file with the extension .dwo. This option allows
5445 the build system to avoid linking files with debug information. To
5446 be useful, this option requires a debugger capable of reading .dwo
5451 Produce debugging information for use by GDB@. This means to use the
5452 most expressive format available (DWARF 2, stabs, or the native format
5453 if neither of those are supported), including GDB extensions if at all
5458 Generate dwarf .debug_pubnames and .debug_pubtypes sections.
5460 @item -ggnu-pubnames
5461 @opindex ggnu-pubnames
5462 Generate .debug_pubnames and .debug_pubtypes sections in a format
5463 suitable for conversion into a GDB@ index. This option is only useful
5464 with a linker that can produce GDB@ index version 7.
5468 Produce debugging information in stabs format (if that is supported),
5469 without GDB extensions. This is the format used by DBX on most BSD
5470 systems. On MIPS, Alpha and System V Release 4 systems this option
5471 produces stabs debugging output that is not understood by DBX or SDB@.
5472 On System V Release 4 systems this option requires the GNU assembler.
5474 @item -feliminate-unused-debug-symbols
5475 @opindex feliminate-unused-debug-symbols
5476 Produce debugging information in stabs format (if that is supported),
5477 for only symbols that are actually used.
5479 @item -femit-class-debug-always
5480 @opindex femit-class-debug-always
5481 Instead of emitting debugging information for a C++ class in only one
5482 object file, emit it in all object files using the class. This option
5483 should be used only with debuggers that are unable to handle the way GCC
5484 normally emits debugging information for classes because using this
5485 option increases the size of debugging information by as much as a
5488 @item -fdebug-types-section
5489 @opindex fdebug-types-section
5490 @opindex fno-debug-types-section
5491 When using DWARF Version 4 or higher, type DIEs can be put into
5492 their own @code{.debug_types} section instead of making them part of the
5493 @code{.debug_info} section. It is more efficient to put them in a separate
5494 comdat sections since the linker can then remove duplicates.
5495 But not all DWARF consumers support @code{.debug_types} sections yet
5496 and on some objects @code{.debug_types} produces larger instead of smaller
5497 debugging information.
5501 Produce debugging information in stabs format (if that is supported),
5502 using GNU extensions understood only by the GNU debugger (GDB)@. The
5503 use of these extensions is likely to make other debuggers crash or
5504 refuse to read the program.
5508 Produce debugging information in COFF format (if that is supported).
5509 This is the format used by SDB on most System V systems prior to
5514 Produce debugging information in XCOFF format (if that is supported).
5515 This is the format used by the DBX debugger on IBM RS/6000 systems.
5519 Produce debugging information in XCOFF format (if that is supported),
5520 using GNU extensions understood only by the GNU debugger (GDB)@. The
5521 use of these extensions is likely to make other debuggers crash or
5522 refuse to read the program, and may cause assemblers other than the GNU
5523 assembler (GAS) to fail with an error.
5525 @item -gdwarf-@var{version}
5526 @opindex gdwarf-@var{version}
5527 Produce debugging information in DWARF format (if that is supported).
5528 The value of @var{version} may be either 2, 3, 4 or 5; the default version
5529 for most targets is 4. DWARF Version 5 is only experimental.
5531 Note that with DWARF Version 2, some ports require and always
5532 use some non-conflicting DWARF 3 extensions in the unwind tables.
5534 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
5535 for maximum benefit.
5537 @item -grecord-gcc-switches
5538 @opindex grecord-gcc-switches
5539 This switch causes the command-line options used to invoke the
5540 compiler that may affect code generation to be appended to the
5541 DW_AT_producer attribute in DWARF debugging information. The options
5542 are concatenated with spaces separating them from each other and from
5543 the compiler version. See also @option{-frecord-gcc-switches} for another
5544 way of storing compiler options into the object file. This is the default.
5546 @item -gno-record-gcc-switches
5547 @opindex gno-record-gcc-switches
5548 Disallow appending command-line options to the DW_AT_producer attribute
5549 in DWARF debugging information.
5551 @item -gstrict-dwarf
5552 @opindex gstrict-dwarf
5553 Disallow using extensions of later DWARF standard version than selected
5554 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
5555 DWARF extensions from later standard versions is allowed.
5557 @item -gno-strict-dwarf
5558 @opindex gno-strict-dwarf
5559 Allow using extensions of later DWARF standard version than selected with
5560 @option{-gdwarf-@var{version}}.
5562 @item -gz@r{[}=@var{type}@r{]}
5564 Produce compressed debug sections in DWARF format, if that is supported.
5565 If @var{type} is not given, the default type depends on the capabilities
5566 of the assembler and linker used. @var{type} may be one of
5567 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
5568 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
5569 compression in traditional GNU format). If the linker doesn't support
5570 writing compressed debug sections, the option is rejected. Otherwise,
5571 if the assembler does not support them, @option{-gz} is silently ignored
5572 when producing object files.
5576 Produce debugging information in Alpha/VMS debug format (if that is
5577 supported). This is the format used by DEBUG on Alpha/VMS systems.
5580 @itemx -ggdb@var{level}
5581 @itemx -gstabs@var{level}
5582 @itemx -gcoff@var{level}
5583 @itemx -gxcoff@var{level}
5584 @itemx -gvms@var{level}
5585 Request debugging information and also use @var{level} to specify how
5586 much information. The default level is 2.
5588 Level 0 produces no debug information at all. Thus, @option{-g0} negates
5591 Level 1 produces minimal information, enough for making backtraces in
5592 parts of the program that you don't plan to debug. This includes
5593 descriptions of functions and external variables, and line number
5594 tables, but no information about local variables.
5596 Level 3 includes extra information, such as all the macro definitions
5597 present in the program. Some debuggers support macro expansion when
5598 you use @option{-g3}.
5600 @option{-gdwarf-2} does not accept a concatenated debug level, because
5601 GCC used to support an option @option{-gdwarf} that meant to generate
5602 debug information in version 1 of the DWARF format (which is very
5603 different from version 2), and it would have been too confusing. That
5604 debug format is long obsolete, but the option cannot be changed now.
5605 Instead use an additional @option{-g@var{level}} option to change the
5606 debug level for DWARF.
5610 Turn off generation of debug info, if leaving out this option
5611 generates it, or turn it on at level 2 otherwise. The position of this
5612 argument in the command line does not matter; it takes effect after all
5613 other options are processed, and it does so only once, no matter how
5614 many times it is given. This is mainly intended to be used with
5615 @option{-fcompare-debug}.
5617 @item -fsanitize=address
5618 @opindex fsanitize=address
5619 Enable AddressSanitizer, a fast memory error detector.
5620 Memory access instructions are instrumented to detect
5621 out-of-bounds and use-after-free bugs.
5622 See @uref{http://code.google.com/p/address-sanitizer/} for
5623 more details. The run-time behavior can be influenced using the
5624 @env{ASAN_OPTIONS} environment variable; see
5625 @url{https://code.google.com/p/address-sanitizer/wiki/Flags#Run-time_flags} for
5626 a list of supported options.
5628 @item -fsanitize=kernel-address
5629 @opindex fsanitize=kernel-address
5630 Enable AddressSanitizer for Linux kernel.
5631 See @uref{http://code.google.com/p/address-sanitizer/wiki/AddressSanitizerForKernel} for more details.
5633 @item -fsanitize=thread
5634 @opindex fsanitize=thread
5635 Enable ThreadSanitizer, a fast data race detector.
5636 Memory access instructions are instrumented to detect
5637 data race bugs. See @uref{http://code.google.com/p/thread-sanitizer/} for more
5638 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
5639 environment variable; see
5640 @url{https://code.google.com/p/thread-sanitizer/wiki/Flags} for a list of
5643 @item -fsanitize=leak
5644 @opindex fsanitize=leak
5645 Enable LeakSanitizer, a memory leak detector.
5646 This option only matters for linking of executables and if neither
5647 @option{-fsanitize=address} nor @option{-fsanitize=thread} is used. In that
5648 case the executable is linked against a library that overrides @code{malloc}
5649 and other allocator functions. See
5650 @uref{https://code.google.com/p/address-sanitizer/wiki/LeakSanitizer} for more
5651 details. The run-time behavior can be influenced using the
5652 @env{LSAN_OPTIONS} environment variable.
5654 @item -fsanitize=undefined
5655 @opindex fsanitize=undefined
5656 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
5657 Various computations are instrumented to detect undefined behavior
5658 at runtime. Current suboptions are:
5662 @item -fsanitize=shift
5663 @opindex fsanitize=shift
5664 This option enables checking that the result of a shift operation is
5665 not undefined. Note that what exactly is considered undefined differs
5666 slightly between C and C++, as well as between ISO C90 and C99, etc.
5668 @item -fsanitize=integer-divide-by-zero
5669 @opindex fsanitize=integer-divide-by-zero
5670 Detect integer division by zero as well as @code{INT_MIN / -1} division.
5672 @item -fsanitize=unreachable
5673 @opindex fsanitize=unreachable
5674 With this option, the compiler turns the @code{__builtin_unreachable}
5675 call into a diagnostics message call instead. When reaching the
5676 @code{__builtin_unreachable} call, the behavior is undefined.
5678 @item -fsanitize=vla-bound
5679 @opindex fsanitize=vla-bound
5680 This option instructs the compiler to check that the size of a variable
5681 length array is positive.
5683 @item -fsanitize=null
5684 @opindex fsanitize=null
5685 This option enables pointer checking. Particularly, the application
5686 built with this option turned on will issue an error message when it
5687 tries to dereference a NULL pointer, or if a reference (possibly an
5688 rvalue reference) is bound to a NULL pointer, or if a method is invoked
5689 on an object pointed by a NULL pointer.
5691 @item -fsanitize=return
5692 @opindex fsanitize=return
5693 This option enables return statement checking. Programs
5694 built with this option turned on will issue an error message
5695 when the end of a non-void function is reached without actually
5696 returning a value. This option works in C++ only.
5698 @item -fsanitize=signed-integer-overflow
5699 @opindex fsanitize=signed-integer-overflow
5700 This option enables signed integer overflow checking. We check that
5701 the result of @code{+}, @code{*}, and both unary and binary @code{-}
5702 does not overflow in the signed arithmetics. Note, integer promotion
5703 rules must be taken into account. That is, the following is not an
5706 signed char a = SCHAR_MAX;
5710 @item -fsanitize=bounds
5711 @opindex fsanitize=bounds
5712 This option enables instrumentation of array bounds. Various out of bounds
5713 accesses are detected. Flexible array members, flexible array member-like
5714 arrays, and initializers of variables with static storage are not instrumented.
5716 @item -fsanitize=alignment
5717 @opindex fsanitize=alignment
5719 This option enables checking of alignment of pointers when they are
5720 dereferenced, or when a reference is bound to insufficiently aligned target,
5721 or when a method or constructor is invoked on insufficiently aligned object.
5723 @item -fsanitize=object-size
5724 @opindex fsanitize=object-size
5725 This option enables instrumentation of memory references using the
5726 @code{__builtin_object_size} function. Various out of bounds pointer
5727 accesses are detected.
5729 @item -fsanitize=float-divide-by-zero
5730 @opindex fsanitize=float-divide-by-zero
5731 Detect floating-point division by zero. Unlike other similar options,
5732 @option{-fsanitize=float-divide-by-zero} is not enabled by
5733 @option{-fsanitize=undefined}, since floating-point division by zero can
5734 be a legitimate way of obtaining infinities and NaNs.
5736 @item -fsanitize=float-cast-overflow
5737 @opindex fsanitize=float-cast-overflow
5738 This option enables floating-point type to integer conversion checking.
5739 We check that the result of the conversion does not overflow.
5740 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
5741 not enabled by @option{-fsanitize=undefined}.
5742 This option does not work well with @code{FE_INVALID} exceptions enabled.
5744 @item -fsanitize=nonnull-attribute
5745 @opindex fsanitize=nonnull-attribute
5747 This option enables instrumentation of calls, checking whether null values
5748 are not passed to arguments marked as requiring a non-null value by the
5749 @code{nonnull} function attribute.
5751 @item -fsanitize=returns-nonnull-attribute
5752 @opindex fsanitize=returns-nonnull-attribute
5754 This option enables instrumentation of return statements in functions
5755 marked with @code{returns_nonnull} function attribute, to detect returning
5756 of null values from such functions.
5758 @item -fsanitize=bool
5759 @opindex fsanitize=bool
5761 This option enables instrumentation of loads from bool. If a value other
5762 than 0/1 is loaded, a run-time error is issued.
5764 @item -fsanitize=enum
5765 @opindex fsanitize=enum
5767 This option enables instrumentation of loads from an enum type. If
5768 a value outside the range of values for the enum type is loaded,
5769 a run-time error is issued.
5771 @item -fsanitize=vptr
5772 @opindex fsanitize=vptr
5774 This option enables instrumentation of C++ member function calls, member
5775 accesses and some conversions between pointers to base and derived classes,
5776 to verify the referenced object has the correct dynamic type.
5780 While @option{-ftrapv} causes traps for signed overflows to be emitted,
5781 @option{-fsanitize=undefined} gives a diagnostic message.
5782 This currently works only for the C family of languages.
5784 @item -fno-sanitize=all
5785 @opindex fno-sanitize=all
5787 This option disables all previously enabled sanitizers.
5788 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
5791 @item -fasan-shadow-offset=@var{number}
5792 @opindex fasan-shadow-offset
5793 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
5794 It is useful for experimenting with different shadow memory layouts in
5795 Kernel AddressSanitizer.
5797 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
5798 @opindex fsanitize-recover
5799 @opindex fno-sanitize-recover
5800 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
5801 mentioned in comma-separated list of @var{opts}. Enabling this option
5802 for a sanitizer component causes it to attempt to continue
5803 running the program as if no error happened. This means multiple
5804 runtime errors can be reported in a single program run, and the exit
5805 code of the program may indicate success even when errors
5806 have been reported. The @option{-fno-sanitize-recover=} can be used to alter
5807 this behavior: only the first detected error is reported
5808 and program then exits with a non-zero exit code.
5810 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
5811 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
5812 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero} and
5813 @option{-fsanitize=kernel-address}. For these sanitizers error recovery is turned on by default.
5814 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
5815 accepted, the former enables recovery for all sanitizers that support it,
5816 the latter disables recovery for all sanitizers that support it.
5818 Syntax without explicit @var{opts} parameter is deprecated. It is equivalent to
5820 -fsanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
5823 Similarly @option{-fno-sanitize-recover} is equivalent to
5825 -fno-sanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
5828 @item -fsanitize-undefined-trap-on-error
5829 @opindex fsanitize-undefined-trap-on-error
5830 The @option{-fsanitize-undefined-trap-on-error} instructs the compiler to
5831 report undefined behavior using @code{__builtin_trap} rather than
5832 a @code{libubsan} library routine. The advantage of this is that the
5833 @code{libubsan} library is not needed and is not linked in, so this
5834 is usable even in freestanding environments.
5836 @item -fcheck-pointer-bounds
5837 @opindex fcheck-pointer-bounds
5838 @opindex fno-check-pointer-bounds
5839 Enable Pointer Bounds Checker instrumentation. Each memory reference
5840 is instrumented with checks of pointer used for memory access against
5841 bounds associated with that pointer. Generated instrumentation may
5842 be controlled by various @option{-fchkp-*} options. Currently there
5843 is only Intel MPX based implementation available, thus i386 target
5844 and @option{-mmpx} are required. MPX based instrumentation requires
5845 a runtime library to enable MPX in a hardware and handle bounds
5846 violation signals. By default when @option{-fcheck-pointer-bounds}
5847 and @option{-mmpx} options are used to link a program, the GCC driver
5848 links against @option{libmpx} runtime library. MPX based instrumentation
5849 may be used for a debugging and also it may be included into a release
5850 version to increase program security. Depending on usage you may
5851 put different requirements to runtime library. Current version
5852 of MPX runtime library is more oriented to be used as a debugging
5853 tool. MPX runtime library usage implies @option{-lpthread}. See
5854 also @option{-static-libmpx}. The runtime library behavior can be
5855 influenced using various @env{CHKP_RT_*} environment variables. See
5856 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
5859 @item -fchkp-check-incomplete-type
5860 @opindex fchkp-check-incomplete-type
5861 @opindex fno-chkp-check-incomplete-type
5862 Generate pointer bounds checks for variables with incomplete type.
5865 @item -fchkp-narrow-bounds
5866 @opindex fchkp-narrow-bounds
5867 @opindex fno-chkp-narrow-bounds
5868 Controls bounds used by Pointer Bounds Checker for pointers to object
5869 fields. If narrowing is enabled then field bounds are used. Otherwise
5870 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
5871 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
5873 @item -fchkp-first-field-has-own-bounds
5874 @opindex fchkp-first-field-has-own-bounds
5875 @opindex fno-chkp-first-field-has-own-bounds
5876 Forces Pointer Bounds Checker to use narrowed bounds for address of the
5877 first field in the structure. By default pointer to the first field has
5878 the same bounds as pointer to the whole structure.
5880 @item -fchkp-narrow-to-innermost-array
5881 @opindex fchkp-narrow-to-innermost-array
5882 @opindex fno-chkp-narrow-to-innermost-array
5883 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
5884 case of nested static arryas access. By default it is disabled and
5885 bounds of the outermost array are used.
5887 @item -fchkp-optimize
5888 @opindex fchkp-optimize
5889 @opindex fno-chkp-optimize
5890 Enables Pointer Bounds Checker optimizations. Enabled by default at
5891 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
5893 @item -fchkp-use-fast-string-functions
5894 @opindex fchkp-use-fast-string-functions
5895 @opindex fno-chkp-use-fast-string-functions
5896 Allow to use @code{*_nobnd} versions of string functions (not copying bounds)
5897 by Pointer Bounds Checker. Disabled by default.
5899 @item -fchkp-use-nochk-string-functions
5900 @opindex fchkp-use-nochk-string-functions
5901 @opindex fno-chkp-use-nochk-string-functions
5902 Allow to use @code{*_nochk} versions of string functions (not checking bounds)
5903 by Pointer Bounds Checker. Disabled by default.
5905 @item -fchkp-use-static-bounds
5906 @opindex fchkp-use-static-bounds
5907 @opindex fno-chkp-use-static-bounds
5908 Allow Pointer Bounds Checker to generate static bounds holding
5909 bounds of static variables. Enabled by default.
5911 @item -fchkp-use-static-const-bounds
5912 @opindex fchkp-use-static-const-bounds
5913 @opindex fno-chkp-use-static-const-bounds
5914 Use statically initialized bounds for constant bounds instead of
5915 generating them each time it is required. By default enabled when
5916 @option{-fchkp-use-static-bounds} is enabled.
5918 @item -fchkp-treat-zero-dynamic-size-as-infinite
5919 @opindex fchkp-treat-zero-dynamic-size-as-infinite
5920 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
5921 With this option zero size obtained dynamically for objects with
5922 incomplete type will be treated as infinite by Pointer Bounds
5923 Checker. It may be helpful if program is linked with a library
5924 missing size information for some symbols. Disabled by default.
5926 @item -fchkp-check-read
5927 @opindex fchkp-check-read
5928 @opindex fno-chkp-check-read
5929 Instructs Pointer Bounds Checker to generate checks for all read
5930 accesses to memory. Enabled by default.
5932 @item -fchkp-check-write
5933 @opindex fchkp-check-write
5934 @opindex fno-chkp-check-write
5935 Instructs Pointer Bounds Checker to generate checks for all write
5936 accesses to memory. Enabled by default.
5938 @item -fchkp-store-bounds
5939 @opindex fchkp-store-bounds
5940 @opindex fno-chkp-store-bounds
5941 Instructs Pointer Bounds Checker to generate bounds stores for
5942 pointer writes. Enabled by default.
5944 @item -fchkp-instrument-calls
5945 @opindex fchkp-instrument-calls
5946 @opindex fno-chkp-instrument-calls
5947 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
5950 @item -fchkp-instrument-marked-only
5951 @opindex fchkp-instrument-marked-only
5952 @opindex fno-chkp-instrument-marked-only
5953 Instructs Pointer Bounds Checker to instrument only functions
5954 marked with @code{bnd_instrument} attribute. Disabled by default.
5956 @item -fchkp-use-wrappers
5957 @opindex fchkp-use-wrappers
5958 @opindex fno-chkp-use-wrappers
5959 Allows Pointer Bounds Checker to replace calls to built-in function
5960 with calls to wrapper functions. When the @option{-fchkp-use-wrappers}
5961 is used to link a program, the GCC driver automatically links
5962 agains @option{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
5965 @item -fdump-final-insns@r{[}=@var{file}@r{]}
5966 @opindex fdump-final-insns
5967 Dump the final internal representation (RTL) to @var{file}. If the
5968 optional argument is omitted (or if @var{file} is @code{.}), the name
5969 of the dump file is determined by appending @code{.gkd} to the
5970 compilation output file name.
5972 @item -fcompare-debug@r{[}=@var{opts}@r{]}
5973 @opindex fcompare-debug
5974 @opindex fno-compare-debug
5975 If no error occurs during compilation, run the compiler a second time,
5976 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
5977 passed to the second compilation. Dump the final internal
5978 representation in both compilations, and print an error if they differ.
5980 If the equal sign is omitted, the default @option{-gtoggle} is used.
5982 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
5983 and nonzero, implicitly enables @option{-fcompare-debug}. If
5984 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
5985 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
5988 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
5989 is equivalent to @option{-fno-compare-debug}, which disables the dumping
5990 of the final representation and the second compilation, preventing even
5991 @env{GCC_COMPARE_DEBUG} from taking effect.
5993 To verify full coverage during @option{-fcompare-debug} testing, set
5994 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
5995 which GCC rejects as an invalid option in any actual compilation
5996 (rather than preprocessing, assembly or linking). To get just a
5997 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
5998 not overridden} will do.
6000 @item -fcompare-debug-second
6001 @opindex fcompare-debug-second
6002 This option is implicitly passed to the compiler for the second
6003 compilation requested by @option{-fcompare-debug}, along with options to
6004 silence warnings, and omitting other options that would cause
6005 side-effect compiler outputs to files or to the standard output. Dump
6006 files and preserved temporary files are renamed so as to contain the
6007 @code{.gk} additional extension during the second compilation, to avoid
6008 overwriting those generated by the first.
6010 When this option is passed to the compiler driver, it causes the
6011 @emph{first} compilation to be skipped, which makes it useful for little
6012 other than debugging the compiler proper.
6014 @item -feliminate-dwarf2-dups
6015 @opindex feliminate-dwarf2-dups
6016 Compress DWARF 2 debugging information by eliminating duplicated
6017 information about each symbol. This option only makes sense when
6018 generating DWARF 2 debugging information with @option{-gdwarf-2}.
6020 @item -femit-struct-debug-baseonly
6021 @opindex femit-struct-debug-baseonly
6022 Emit debug information for struct-like types
6023 only when the base name of the compilation source file
6024 matches the base name of file in which the struct is defined.
6026 This option substantially reduces the size of debugging information,
6027 but at significant potential loss in type information to the debugger.
6028 See @option{-femit-struct-debug-reduced} for a less aggressive option.
6029 See @option{-femit-struct-debug-detailed} for more detailed control.
6031 This option works only with DWARF 2.
6033 @item -femit-struct-debug-reduced
6034 @opindex femit-struct-debug-reduced
6035 Emit debug information for struct-like types
6036 only when the base name of the compilation source file
6037 matches the base name of file in which the type is defined,
6038 unless the struct is a template or defined in a system header.
6040 This option significantly reduces the size of debugging information,
6041 with some potential loss in type information to the debugger.
6042 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
6043 See @option{-femit-struct-debug-detailed} for more detailed control.
6045 This option works only with DWARF 2.
6047 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
6048 @opindex femit-struct-debug-detailed
6049 Specify the struct-like types
6050 for which the compiler generates debug information.
6051 The intent is to reduce duplicate struct debug information
6052 between different object files within the same program.
6054 This option is a detailed version of
6055 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
6056 which serves for most needs.
6058 A specification has the syntax@*
6059 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
6061 The optional first word limits the specification to
6062 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
6063 A struct type is used directly when it is the type of a variable, member.
6064 Indirect uses arise through pointers to structs.
6065 That is, when use of an incomplete struct is valid, the use is indirect.
6067 @samp{struct one direct; struct two * indirect;}.
6069 The optional second word limits the specification to
6070 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
6071 Generic structs are a bit complicated to explain.
6072 For C++, these are non-explicit specializations of template classes,
6073 or non-template classes within the above.
6074 Other programming languages have generics,
6075 but @option{-femit-struct-debug-detailed} does not yet implement them.
6077 The third word specifies the source files for those
6078 structs for which the compiler should emit debug information.
6079 The values @samp{none} and @samp{any} have the normal meaning.
6080 The value @samp{base} means that
6081 the base of name of the file in which the type declaration appears
6082 must match the base of the name of the main compilation file.
6083 In practice, this means that when compiling @file{foo.c}, debug information
6084 is generated for types declared in that file and @file{foo.h},
6085 but not other header files.
6086 The value @samp{sys} means those types satisfying @samp{base}
6087 or declared in system or compiler headers.
6089 You may need to experiment to determine the best settings for your application.
6091 The default is @option{-femit-struct-debug-detailed=all}.
6093 This option works only with DWARF 2.
6095 @item -fno-merge-debug-strings
6096 @opindex fmerge-debug-strings
6097 @opindex fno-merge-debug-strings
6098 Direct the linker to not merge together strings in the debugging
6099 information that are identical in different object files. Merging is
6100 not supported by all assemblers or linkers. Merging decreases the size
6101 of the debug information in the output file at the cost of increasing
6102 link processing time. Merging is enabled by default.
6104 @item -fdebug-prefix-map=@var{old}=@var{new}
6105 @opindex fdebug-prefix-map
6106 When compiling files in directory @file{@var{old}}, record debugging
6107 information describing them as in @file{@var{new}} instead.
6109 @item -fno-dwarf2-cfi-asm
6110 @opindex fdwarf2-cfi-asm
6111 @opindex fno-dwarf2-cfi-asm
6112 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
6113 instead of using GAS @code{.cfi_*} directives.
6115 @cindex @command{prof}
6118 Generate extra code to write profile information suitable for the
6119 analysis program @command{prof}. You must use this option when compiling
6120 the source files you want data about, and you must also use it when
6123 @cindex @command{gprof}
6126 Generate extra code to write profile information suitable for the
6127 analysis program @command{gprof}. You must use this option when compiling
6128 the source files you want data about, and you must also use it when
6133 Makes the compiler print out each function name as it is compiled, and
6134 print some statistics about each pass when it finishes.
6137 @opindex ftime-report
6138 Makes the compiler print some statistics about the time consumed by each
6139 pass when it finishes.
6142 @opindex fmem-report
6143 Makes the compiler print some statistics about permanent memory
6144 allocation when it finishes.
6146 @item -fmem-report-wpa
6147 @opindex fmem-report-wpa
6148 Makes the compiler print some statistics about permanent memory
6149 allocation for the WPA phase only.
6151 @item -fpre-ipa-mem-report
6152 @opindex fpre-ipa-mem-report
6153 @item -fpost-ipa-mem-report
6154 @opindex fpost-ipa-mem-report
6155 Makes the compiler print some statistics about permanent memory
6156 allocation before or after interprocedural optimization.
6158 @item -fprofile-report
6159 @opindex fprofile-report
6160 Makes the compiler print some statistics about consistency of the
6161 (estimated) profile and effect of individual passes.
6164 @opindex fstack-usage
6165 Makes the compiler output stack usage information for the program, on a
6166 per-function basis. The filename for the dump is made by appending
6167 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
6168 the output file, if explicitly specified and it is not an executable,
6169 otherwise it is the basename of the source file. An entry is made up
6174 The name of the function.
6178 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
6181 The qualifier @code{static} means that the function manipulates the stack
6182 statically: a fixed number of bytes are allocated for the frame on function
6183 entry and released on function exit; no stack adjustments are otherwise made
6184 in the function. The second field is this fixed number of bytes.
6186 The qualifier @code{dynamic} means that the function manipulates the stack
6187 dynamically: in addition to the static allocation described above, stack
6188 adjustments are made in the body of the function, for example to push/pop
6189 arguments around function calls. If the qualifier @code{bounded} is also
6190 present, the amount of these adjustments is bounded at compile time and
6191 the second field is an upper bound of the total amount of stack used by
6192 the function. If it is not present, the amount of these adjustments is
6193 not bounded at compile time and the second field only represents the
6196 @item -fprofile-arcs
6197 @opindex fprofile-arcs
6198 Add code so that program flow @dfn{arcs} are instrumented. During
6199 execution the program records how many times each branch and call is
6200 executed and how many times it is taken or returns. When the compiled
6201 program exits it saves this data to a file called
6202 @file{@var{auxname}.gcda} for each source file. The data may be used for
6203 profile-directed optimizations (@option{-fbranch-probabilities}), or for
6204 test coverage analysis (@option{-ftest-coverage}). Each object file's
6205 @var{auxname} is generated from the name of the output file, if
6206 explicitly specified and it is not the final executable, otherwise it is
6207 the basename of the source file. In both cases any suffix is removed
6208 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
6209 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
6210 @xref{Cross-profiling}.
6212 @cindex @command{gcov}
6216 This option is used to compile and link code instrumented for coverage
6217 analysis. The option is a synonym for @option{-fprofile-arcs}
6218 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
6219 linking). See the documentation for those options for more details.
6224 Compile the source files with @option{-fprofile-arcs} plus optimization
6225 and code generation options. For test coverage analysis, use the
6226 additional @option{-ftest-coverage} option. You do not need to profile
6227 every source file in a program.
6230 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
6231 (the latter implies the former).
6234 Run the program on a representative workload to generate the arc profile
6235 information. This may be repeated any number of times. You can run
6236 concurrent instances of your program, and provided that the file system
6237 supports locking, the data files will be correctly updated. Also
6238 @code{fork} calls are detected and correctly handled (double counting
6242 For profile-directed optimizations, compile the source files again with
6243 the same optimization and code generation options plus
6244 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
6245 Control Optimization}).
6248 For test coverage analysis, use @command{gcov} to produce human readable
6249 information from the @file{.gcno} and @file{.gcda} files. Refer to the
6250 @command{gcov} documentation for further information.
6254 With @option{-fprofile-arcs}, for each function of your program GCC
6255 creates a program flow graph, then finds a spanning tree for the graph.
6256 Only arcs that are not on the spanning tree have to be instrumented: the
6257 compiler adds code to count the number of times that these arcs are
6258 executed. When an arc is the only exit or only entrance to a block, the
6259 instrumentation code can be added to the block; otherwise, a new basic
6260 block must be created to hold the instrumentation code.
6263 @item -ftest-coverage
6264 @opindex ftest-coverage
6265 Produce a notes file that the @command{gcov} code-coverage utility
6266 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
6267 show program coverage. Each source file's note file is called
6268 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
6269 above for a description of @var{auxname} and instructions on how to
6270 generate test coverage data. Coverage data matches the source files
6271 more closely if you do not optimize.
6273 @item -fdbg-cnt-list
6274 @opindex fdbg-cnt-list
6275 Print the name and the counter upper bound for all debug counters.
6278 @item -fdbg-cnt=@var{counter-value-list}
6280 Set the internal debug counter upper bound. @var{counter-value-list}
6281 is a comma-separated list of @var{name}:@var{value} pairs
6282 which sets the upper bound of each debug counter @var{name} to @var{value}.
6283 All debug counters have the initial upper bound of @code{UINT_MAX};
6284 thus @code{dbg_cnt} returns true always unless the upper bound
6285 is set by this option.
6286 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
6287 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
6289 @item -fenable-@var{kind}-@var{pass}
6290 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
6294 This is a set of options that are used to explicitly disable/enable
6295 optimization passes. These options are intended for use for debugging GCC.
6296 Compiler users should use regular options for enabling/disabling
6301 @item -fdisable-ipa-@var{pass}
6302 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6303 statically invoked in the compiler multiple times, the pass name should be
6304 appended with a sequential number starting from 1.
6306 @item -fdisable-rtl-@var{pass}
6307 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
6308 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
6309 statically invoked in the compiler multiple times, the pass name should be
6310 appended with a sequential number starting from 1. @var{range-list} is a
6311 comma-separated list of function ranges or assembler names. Each range is a number
6312 pair separated by a colon. The range is inclusive in both ends. If the range
6313 is trivial, the number pair can be simplified as a single number. If the
6314 function's call graph node's @var{uid} falls within one of the specified ranges,
6315 the @var{pass} is disabled for that function. The @var{uid} is shown in the
6316 function header of a dump file, and the pass names can be dumped by using
6317 option @option{-fdump-passes}.
6319 @item -fdisable-tree-@var{pass}
6320 @itemx -fdisable-tree-@var{pass}=@var{range-list}
6321 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
6324 @item -fenable-ipa-@var{pass}
6325 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6326 statically invoked in the compiler multiple times, the pass name should be
6327 appended with a sequential number starting from 1.
6329 @item -fenable-rtl-@var{pass}
6330 @itemx -fenable-rtl-@var{pass}=@var{range-list}
6331 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
6332 description and examples.
6334 @item -fenable-tree-@var{pass}
6335 @itemx -fenable-tree-@var{pass}=@var{range-list}
6336 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
6337 of option arguments.
6341 Here are some examples showing uses of these options.
6345 # disable ccp1 for all functions
6347 # disable complete unroll for function whose cgraph node uid is 1
6348 -fenable-tree-cunroll=1
6349 # disable gcse2 for functions at the following ranges [1,1],
6350 # [300,400], and [400,1000]
6351 # disable gcse2 for functions foo and foo2
6352 -fdisable-rtl-gcse2=foo,foo2
6353 # disable early inlining
6354 -fdisable-tree-einline
6355 # disable ipa inlining
6356 -fdisable-ipa-inline
6357 # enable tree full unroll
6358 -fenable-tree-unroll
6362 @item -d@var{letters}
6363 @itemx -fdump-rtl-@var{pass}
6364 @itemx -fdump-rtl-@var{pass}=@var{filename}
6366 @opindex fdump-rtl-@var{pass}
6367 Says to make debugging dumps during compilation at times specified by
6368 @var{letters}. This is used for debugging the RTL-based passes of the
6369 compiler. The file names for most of the dumps are made by appending
6370 a pass number and a word to the @var{dumpname}, and the files are
6371 created in the directory of the output file. In case of
6372 @option{=@var{filename}} option, the dump is output on the given file
6373 instead of the pass numbered dump files. Note that the pass number is
6374 computed statically as passes get registered into the pass manager.
6375 Thus the numbering is not related to the dynamic order of execution of
6376 passes. In particular, a pass installed by a plugin could have a
6377 number over 200 even if it executed quite early. @var{dumpname} is
6378 generated from the name of the output file, if explicitly specified
6379 and it is not an executable, otherwise it is the basename of the
6380 source file. These switches may have different effects when
6381 @option{-E} is used for preprocessing.
6383 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
6384 @option{-d} option @var{letters}. Here are the possible
6385 letters for use in @var{pass} and @var{letters}, and their meanings:
6389 @item -fdump-rtl-alignments
6390 @opindex fdump-rtl-alignments
6391 Dump after branch alignments have been computed.
6393 @item -fdump-rtl-asmcons
6394 @opindex fdump-rtl-asmcons
6395 Dump after fixing rtl statements that have unsatisfied in/out constraints.
6397 @item -fdump-rtl-auto_inc_dec
6398 @opindex fdump-rtl-auto_inc_dec
6399 Dump after auto-inc-dec discovery. This pass is only run on
6400 architectures that have auto inc or auto dec instructions.
6402 @item -fdump-rtl-barriers
6403 @opindex fdump-rtl-barriers
6404 Dump after cleaning up the barrier instructions.
6406 @item -fdump-rtl-bbpart
6407 @opindex fdump-rtl-bbpart
6408 Dump after partitioning hot and cold basic blocks.
6410 @item -fdump-rtl-bbro
6411 @opindex fdump-rtl-bbro
6412 Dump after block reordering.
6414 @item -fdump-rtl-btl1
6415 @itemx -fdump-rtl-btl2
6416 @opindex fdump-rtl-btl2
6417 @opindex fdump-rtl-btl2
6418 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
6419 after the two branch
6420 target load optimization passes.
6422 @item -fdump-rtl-bypass
6423 @opindex fdump-rtl-bypass
6424 Dump after jump bypassing and control flow optimizations.
6426 @item -fdump-rtl-combine
6427 @opindex fdump-rtl-combine
6428 Dump after the RTL instruction combination pass.
6430 @item -fdump-rtl-compgotos
6431 @opindex fdump-rtl-compgotos
6432 Dump after duplicating the computed gotos.
6434 @item -fdump-rtl-ce1
6435 @itemx -fdump-rtl-ce2
6436 @itemx -fdump-rtl-ce3
6437 @opindex fdump-rtl-ce1
6438 @opindex fdump-rtl-ce2
6439 @opindex fdump-rtl-ce3
6440 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
6441 @option{-fdump-rtl-ce3} enable dumping after the three
6442 if conversion passes.
6444 @item -fdump-rtl-cprop_hardreg
6445 @opindex fdump-rtl-cprop_hardreg
6446 Dump after hard register copy propagation.
6448 @item -fdump-rtl-csa
6449 @opindex fdump-rtl-csa
6450 Dump after combining stack adjustments.
6452 @item -fdump-rtl-cse1
6453 @itemx -fdump-rtl-cse2
6454 @opindex fdump-rtl-cse1
6455 @opindex fdump-rtl-cse2
6456 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
6457 the two common subexpression elimination passes.
6459 @item -fdump-rtl-dce
6460 @opindex fdump-rtl-dce
6461 Dump after the standalone dead code elimination passes.
6463 @item -fdump-rtl-dbr
6464 @opindex fdump-rtl-dbr
6465 Dump after delayed branch scheduling.
6467 @item -fdump-rtl-dce1
6468 @itemx -fdump-rtl-dce2
6469 @opindex fdump-rtl-dce1
6470 @opindex fdump-rtl-dce2
6471 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
6472 the two dead store elimination passes.
6475 @opindex fdump-rtl-eh
6476 Dump after finalization of EH handling code.
6478 @item -fdump-rtl-eh_ranges
6479 @opindex fdump-rtl-eh_ranges
6480 Dump after conversion of EH handling range regions.
6482 @item -fdump-rtl-expand
6483 @opindex fdump-rtl-expand
6484 Dump after RTL generation.
6486 @item -fdump-rtl-fwprop1
6487 @itemx -fdump-rtl-fwprop2
6488 @opindex fdump-rtl-fwprop1
6489 @opindex fdump-rtl-fwprop2
6490 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
6491 dumping after the two forward propagation passes.
6493 @item -fdump-rtl-gcse1
6494 @itemx -fdump-rtl-gcse2
6495 @opindex fdump-rtl-gcse1
6496 @opindex fdump-rtl-gcse2
6497 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
6498 after global common subexpression elimination.
6500 @item -fdump-rtl-init-regs
6501 @opindex fdump-rtl-init-regs
6502 Dump after the initialization of the registers.
6504 @item -fdump-rtl-initvals
6505 @opindex fdump-rtl-initvals
6506 Dump after the computation of the initial value sets.
6508 @item -fdump-rtl-into_cfglayout
6509 @opindex fdump-rtl-into_cfglayout
6510 Dump after converting to cfglayout mode.
6512 @item -fdump-rtl-ira
6513 @opindex fdump-rtl-ira
6514 Dump after iterated register allocation.
6516 @item -fdump-rtl-jump
6517 @opindex fdump-rtl-jump
6518 Dump after the second jump optimization.
6520 @item -fdump-rtl-loop2
6521 @opindex fdump-rtl-loop2
6522 @option{-fdump-rtl-loop2} enables dumping after the rtl
6523 loop optimization passes.
6525 @item -fdump-rtl-mach
6526 @opindex fdump-rtl-mach
6527 Dump after performing the machine dependent reorganization pass, if that
6530 @item -fdump-rtl-mode_sw
6531 @opindex fdump-rtl-mode_sw
6532 Dump after removing redundant mode switches.
6534 @item -fdump-rtl-rnreg
6535 @opindex fdump-rtl-rnreg
6536 Dump after register renumbering.
6538 @item -fdump-rtl-outof_cfglayout
6539 @opindex fdump-rtl-outof_cfglayout
6540 Dump after converting from cfglayout mode.
6542 @item -fdump-rtl-peephole2
6543 @opindex fdump-rtl-peephole2
6544 Dump after the peephole pass.
6546 @item -fdump-rtl-postreload
6547 @opindex fdump-rtl-postreload
6548 Dump after post-reload optimizations.
6550 @item -fdump-rtl-pro_and_epilogue
6551 @opindex fdump-rtl-pro_and_epilogue
6552 Dump after generating the function prologues and epilogues.
6554 @item -fdump-rtl-sched1
6555 @itemx -fdump-rtl-sched2
6556 @opindex fdump-rtl-sched1
6557 @opindex fdump-rtl-sched2
6558 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
6559 after the basic block scheduling passes.
6561 @item -fdump-rtl-ree
6562 @opindex fdump-rtl-ree
6563 Dump after sign/zero extension elimination.
6565 @item -fdump-rtl-seqabstr
6566 @opindex fdump-rtl-seqabstr
6567 Dump after common sequence discovery.
6569 @item -fdump-rtl-shorten
6570 @opindex fdump-rtl-shorten
6571 Dump after shortening branches.
6573 @item -fdump-rtl-sibling
6574 @opindex fdump-rtl-sibling
6575 Dump after sibling call optimizations.
6577 @item -fdump-rtl-split1
6578 @itemx -fdump-rtl-split2
6579 @itemx -fdump-rtl-split3
6580 @itemx -fdump-rtl-split4
6581 @itemx -fdump-rtl-split5
6582 @opindex fdump-rtl-split1
6583 @opindex fdump-rtl-split2
6584 @opindex fdump-rtl-split3
6585 @opindex fdump-rtl-split4
6586 @opindex fdump-rtl-split5
6587 These options enable dumping after five rounds of
6588 instruction splitting.
6590 @item -fdump-rtl-sms
6591 @opindex fdump-rtl-sms
6592 Dump after modulo scheduling. This pass is only run on some
6595 @item -fdump-rtl-stack
6596 @opindex fdump-rtl-stack
6597 Dump after conversion from GCC's ``flat register file'' registers to the
6598 x87's stack-like registers. This pass is only run on x86 variants.
6600 @item -fdump-rtl-subreg1
6601 @itemx -fdump-rtl-subreg2
6602 @opindex fdump-rtl-subreg1
6603 @opindex fdump-rtl-subreg2
6604 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
6605 the two subreg expansion passes.
6607 @item -fdump-rtl-unshare
6608 @opindex fdump-rtl-unshare
6609 Dump after all rtl has been unshared.
6611 @item -fdump-rtl-vartrack
6612 @opindex fdump-rtl-vartrack
6613 Dump after variable tracking.
6615 @item -fdump-rtl-vregs
6616 @opindex fdump-rtl-vregs
6617 Dump after converting virtual registers to hard registers.
6619 @item -fdump-rtl-web
6620 @opindex fdump-rtl-web
6621 Dump after live range splitting.
6623 @item -fdump-rtl-regclass
6624 @itemx -fdump-rtl-subregs_of_mode_init
6625 @itemx -fdump-rtl-subregs_of_mode_finish
6626 @itemx -fdump-rtl-dfinit
6627 @itemx -fdump-rtl-dfinish
6628 @opindex fdump-rtl-regclass
6629 @opindex fdump-rtl-subregs_of_mode_init
6630 @opindex fdump-rtl-subregs_of_mode_finish
6631 @opindex fdump-rtl-dfinit
6632 @opindex fdump-rtl-dfinish
6633 These dumps are defined but always produce empty files.
6636 @itemx -fdump-rtl-all
6638 @opindex fdump-rtl-all
6639 Produce all the dumps listed above.
6643 Annotate the assembler output with miscellaneous debugging information.
6647 Dump all macro definitions, at the end of preprocessing, in addition to
6652 Produce a core dump whenever an error occurs.
6656 Annotate the assembler output with a comment indicating which
6657 pattern and alternative is used. The length of each instruction is
6662 Dump the RTL in the assembler output as a comment before each instruction.
6663 Also turns on @option{-dp} annotation.
6667 Just generate RTL for a function instead of compiling it. Usually used
6668 with @option{-fdump-rtl-expand}.
6672 @opindex fdump-noaddr
6673 When doing debugging dumps, suppress address output. This makes it more
6674 feasible to use diff on debugging dumps for compiler invocations with
6675 different compiler binaries and/or different
6676 text / bss / data / heap / stack / dso start locations.
6679 @opindex freport-bug
6680 Collect and dump debug information into temporary file if ICE in C/C++
6683 @item -fdump-unnumbered
6684 @opindex fdump-unnumbered
6685 When doing debugging dumps, suppress instruction numbers and address output.
6686 This makes it more feasible to use diff on debugging dumps for compiler
6687 invocations with different options, in particular with and without
6690 @item -fdump-unnumbered-links
6691 @opindex fdump-unnumbered-links
6692 When doing debugging dumps (see @option{-d} option above), suppress
6693 instruction numbers for the links to the previous and next instructions
6696 @item -fdump-translation-unit @r{(C++ only)}
6697 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
6698 @opindex fdump-translation-unit
6699 Dump a representation of the tree structure for the entire translation
6700 unit to a file. The file name is made by appending @file{.tu} to the
6701 source file name, and the file is created in the same directory as the
6702 output file. If the @samp{-@var{options}} form is used, @var{options}
6703 controls the details of the dump as described for the
6704 @option{-fdump-tree} options.
6706 @item -fdump-class-hierarchy @r{(C++ only)}
6707 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
6708 @opindex fdump-class-hierarchy
6709 Dump a representation of each class's hierarchy and virtual function
6710 table layout to a file. The file name is made by appending
6711 @file{.class} to the source file name, and the file is created in the
6712 same directory as the output file. If the @samp{-@var{options}} form
6713 is used, @var{options} controls the details of the dump as described
6714 for the @option{-fdump-tree} options.
6716 @item -fdump-ipa-@var{switch}
6718 Control the dumping at various stages of inter-procedural analysis
6719 language tree to a file. The file name is generated by appending a
6720 switch specific suffix to the source file name, and the file is created
6721 in the same directory as the output file. The following dumps are
6726 Enables all inter-procedural analysis dumps.
6729 Dumps information about call-graph optimization, unused function removal,
6730 and inlining decisions.
6733 Dump after function inlining.
6738 @opindex fdump-passes
6739 Dump the list of optimization passes that are turned on and off by
6740 the current command-line options.
6742 @item -fdump-statistics-@var{option}
6743 @opindex fdump-statistics
6744 Enable and control dumping of pass statistics in a separate file. The
6745 file name is generated by appending a suffix ending in
6746 @samp{.statistics} to the source file name, and the file is created in
6747 the same directory as the output file. If the @samp{-@var{option}}
6748 form is used, @samp{-stats} causes counters to be summed over the
6749 whole compilation unit while @samp{-details} dumps every event as
6750 the passes generate them. The default with no option is to sum
6751 counters for each function compiled.
6753 @item -fdump-tree-@var{switch}
6754 @itemx -fdump-tree-@var{switch}-@var{options}
6755 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
6757 Control the dumping at various stages of processing the intermediate
6758 language tree to a file. The file name is generated by appending a
6759 switch-specific suffix to the source file name, and the file is
6760 created in the same directory as the output file. In case of
6761 @option{=@var{filename}} option, the dump is output on the given file
6762 instead of the auto named dump files. If the @samp{-@var{options}}
6763 form is used, @var{options} is a list of @samp{-} separated options
6764 which control the details of the dump. Not all options are applicable
6765 to all dumps; those that are not meaningful are ignored. The
6766 following options are available
6770 Print the address of each node. Usually this is not meaningful as it
6771 changes according to the environment and source file. Its primary use
6772 is for tying up a dump file with a debug environment.
6774 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
6775 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
6776 use working backward from mangled names in the assembly file.
6778 When dumping front-end intermediate representations, inhibit dumping
6779 of members of a scope or body of a function merely because that scope
6780 has been reached. Only dump such items when they are directly reachable
6783 When dumping pretty-printed trees, this option inhibits dumping the
6784 bodies of control structures.
6786 When dumping RTL, print the RTL in slim (condensed) form instead of
6787 the default LISP-like representation.
6789 Print a raw representation of the tree. By default, trees are
6790 pretty-printed into a C-like representation.
6792 Enable more detailed dumps (not honored by every dump option). Also
6793 include information from the optimization passes.
6795 Enable dumping various statistics about the pass (not honored by every dump
6798 Enable showing basic block boundaries (disabled in raw dumps).
6800 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
6801 dump a representation of the control flow graph suitable for viewing with
6802 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
6803 the file is pretty-printed as a subgraph, so that GraphViz can render them
6804 all in a single plot.
6806 This option currently only works for RTL dumps, and the RTL is always
6807 dumped in slim form.
6809 Enable showing virtual operands for every statement.
6811 Enable showing line numbers for statements.
6813 Enable showing the unique ID (@code{DECL_UID}) for each variable.
6815 Enable showing the tree dump for each statement.
6817 Enable showing the EH region number holding each statement.
6819 Enable showing scalar evolution analysis details.
6821 Enable showing optimization information (only available in certain
6824 Enable showing missed optimization information (only available in certain
6827 Enable other detailed optimization information (only available in
6829 @item =@var{filename}
6830 Instead of an auto named dump file, output into the given file
6831 name. The file names @file{stdout} and @file{stderr} are treated
6832 specially and are considered already open standard streams. For
6836 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
6837 -fdump-tree-pre=stderr file.c
6840 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
6841 output on to @file{stderr}. If two conflicting dump filenames are
6842 given for the same pass, then the latter option overrides the earlier
6846 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
6847 and @option{lineno}.
6850 Turn on all optimization options, i.e., @option{optimized},
6851 @option{missed}, and @option{note}.
6854 The following tree dumps are possible:
6858 @opindex fdump-tree-original
6859 Dump before any tree based optimization, to @file{@var{file}.original}.
6862 @opindex fdump-tree-optimized
6863 Dump after all tree based optimization, to @file{@var{file}.optimized}.
6866 @opindex fdump-tree-gimple
6867 Dump each function before and after the gimplification pass to a file. The
6868 file name is made by appending @file{.gimple} to the source file name.
6871 @opindex fdump-tree-cfg
6872 Dump the control flow graph of each function to a file. The file name is
6873 made by appending @file{.cfg} to the source file name.
6876 @opindex fdump-tree-ch
6877 Dump each function after copying loop headers. The file name is made by
6878 appending @file{.ch} to the source file name.
6881 @opindex fdump-tree-ssa
6882 Dump SSA related information to a file. The file name is made by appending
6883 @file{.ssa} to the source file name.
6886 @opindex fdump-tree-alias
6887 Dump aliasing information for each function. The file name is made by
6888 appending @file{.alias} to the source file name.
6891 @opindex fdump-tree-ccp
6892 Dump each function after CCP@. The file name is made by appending
6893 @file{.ccp} to the source file name.
6896 @opindex fdump-tree-storeccp
6897 Dump each function after STORE-CCP@. The file name is made by appending
6898 @file{.storeccp} to the source file name.
6901 @opindex fdump-tree-pre
6902 Dump trees after partial redundancy elimination. The file name is made
6903 by appending @file{.pre} to the source file name.
6906 @opindex fdump-tree-fre
6907 Dump trees after full redundancy elimination. The file name is made
6908 by appending @file{.fre} to the source file name.
6911 @opindex fdump-tree-copyprop
6912 Dump trees after copy propagation. The file name is made
6913 by appending @file{.copyprop} to the source file name.
6915 @item store_copyprop
6916 @opindex fdump-tree-store_copyprop
6917 Dump trees after store copy-propagation. The file name is made
6918 by appending @file{.store_copyprop} to the source file name.
6921 @opindex fdump-tree-dce
6922 Dump each function after dead code elimination. The file name is made by
6923 appending @file{.dce} to the source file name.
6926 @opindex fdump-tree-sra
6927 Dump each function after performing scalar replacement of aggregates. The
6928 file name is made by appending @file{.sra} to the source file name.
6931 @opindex fdump-tree-sink
6932 Dump each function after performing code sinking. The file name is made
6933 by appending @file{.sink} to the source file name.
6936 @opindex fdump-tree-dom
6937 Dump each function after applying dominator tree optimizations. The file
6938 name is made by appending @file{.dom} to the source file name.
6941 @opindex fdump-tree-dse
6942 Dump each function after applying dead store elimination. The file
6943 name is made by appending @file{.dse} to the source file name.
6946 @opindex fdump-tree-phiopt
6947 Dump each function after optimizing PHI nodes into straightline code. The file
6948 name is made by appending @file{.phiopt} to the source file name.
6951 @opindex fdump-tree-forwprop
6952 Dump each function after forward propagating single use variables. The file
6953 name is made by appending @file{.forwprop} to the source file name.
6956 @opindex fdump-tree-copyrename
6957 Dump each function after applying the copy rename optimization. The file
6958 name is made by appending @file{.copyrename} to the source file name.
6961 @opindex fdump-tree-nrv
6962 Dump each function after applying the named return value optimization on
6963 generic trees. The file name is made by appending @file{.nrv} to the source
6967 @opindex fdump-tree-vect
6968 Dump each function after applying vectorization of loops. The file name is
6969 made by appending @file{.vect} to the source file name.
6972 @opindex fdump-tree-slp
6973 Dump each function after applying vectorization of basic blocks. The file name
6974 is made by appending @file{.slp} to the source file name.
6977 @opindex fdump-tree-vrp
6978 Dump each function after Value Range Propagation (VRP). The file name
6979 is made by appending @file{.vrp} to the source file name.
6982 @opindex fdump-tree-all
6983 Enable all the available tree dumps with the flags provided in this option.
6987 @itemx -fopt-info-@var{options}
6988 @itemx -fopt-info-@var{options}=@var{filename}
6990 Controls optimization dumps from various optimization passes. If the
6991 @samp{-@var{options}} form is used, @var{options} is a list of
6992 @samp{-} separated option keywords to select the dump details and
6995 The @var{options} can be divided into two groups: options describing the
6996 verbosity of the dump, and options describing which optimizations
6997 should be included. The options from both the groups can be freely
6998 mixed as they are non-overlapping. However, in case of any conflicts,
6999 the later options override the earlier options on the command
7002 The following options control the dump verbosity:
7006 Print information when an optimization is successfully applied. It is
7007 up to a pass to decide which information is relevant. For example, the
7008 vectorizer passes print the source location of loops which are
7009 successfully vectorized.
7011 Print information about missed optimizations. Individual passes
7012 control which information to include in the output.
7014 Print verbose information about optimizations, such as certain
7015 transformations, more detailed messages about decisions etc.
7017 Print detailed optimization information. This includes
7018 @samp{optimized}, @samp{missed}, and @samp{note}.
7021 One or more of the following option keywords can be used to describe a
7022 group of optimizations:
7026 Enable dumps from all interprocedural optimizations.
7028 Enable dumps from all loop optimizations.
7030 Enable dumps from all inlining optimizations.
7032 Enable dumps from all vectorization optimizations.
7034 Enable dumps from all optimizations. This is a superset of
7035 the optimization groups listed above.
7039 omitted, it defaults to @samp{optimized-optall}, which means to dump all
7040 info about successful optimizations from all the passes.
7042 If the @var{filename} is provided, then the dumps from all the
7043 applicable optimizations are concatenated into the @var{filename}.
7044 Otherwise the dump is output onto @file{stderr}. Though multiple
7045 @option{-fopt-info} options are accepted, only one of them can include
7046 a @var{filename}. If other filenames are provided then all but the
7047 first such option are ignored.
7049 Note that the output @var{filename} is overwritten
7050 in case of multiple translation units. If a combined output from
7051 multiple translation units is desired, @file{stderr} should be used
7054 In the following example, the optimization info is output to
7063 gcc -O3 -fopt-info-missed=missed.all
7067 outputs missed optimization report from all the passes into
7068 @file{missed.all}, and this one:
7071 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
7075 prints information about missed optimization opportunities from
7076 vectorization passes on @file{stderr}.
7077 Note that @option{-fopt-info-vec-missed} is equivalent to
7078 @option{-fopt-info-missed-vec}.
7082 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
7086 outputs information about missed optimizations as well as
7087 optimized locations from all the inlining passes into
7093 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
7097 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
7098 in conflict since only one output file is allowed. In this case, only
7099 the first option takes effect and the subsequent options are
7100 ignored. Thus only @file{vec.miss} is produced which contains
7101 dumps from the vectorizer about missed opportunities.
7103 @item -frandom-seed=@var{number}
7104 @opindex frandom-seed
7105 This option provides a seed that GCC uses in place of
7106 random numbers in generating certain symbol names
7107 that have to be different in every compiled file. It is also used to
7108 place unique stamps in coverage data files and the object files that
7109 produce them. You can use the @option{-frandom-seed} option to produce
7110 reproducibly identical object files.
7112 The @var{number} should be different for every file you compile.
7114 @item -fsched-verbose=@var{n}
7115 @opindex fsched-verbose
7116 On targets that use instruction scheduling, this option controls the
7117 amount of debugging output the scheduler prints. This information is
7118 written to standard error, unless @option{-fdump-rtl-sched1} or
7119 @option{-fdump-rtl-sched2} is specified, in which case it is output
7120 to the usual dump listing file, @file{.sched1} or @file{.sched2}
7121 respectively. However for @var{n} greater than nine, the output is
7122 always printed to standard error.
7124 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
7125 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
7126 For @var{n} greater than one, it also output basic block probabilities,
7127 detailed ready list information and unit/insn info. For @var{n} greater
7128 than two, it includes RTL at abort point, control-flow and regions info.
7129 And for @var{n} over four, @option{-fsched-verbose} also includes
7133 @itemx -save-temps=cwd
7135 Store the usual ``temporary'' intermediate files permanently; place them
7136 in the current directory and name them based on the source file. Thus,
7137 compiling @file{foo.c} with @option{-c -save-temps} produces files
7138 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
7139 preprocessed @file{foo.i} output file even though the compiler now
7140 normally uses an integrated preprocessor.
7142 When used in combination with the @option{-x} command-line option,
7143 @option{-save-temps} is sensible enough to avoid over writing an
7144 input source file with the same extension as an intermediate file.
7145 The corresponding intermediate file may be obtained by renaming the
7146 source file before using @option{-save-temps}.
7148 If you invoke GCC in parallel, compiling several different source
7149 files that share a common base name in different subdirectories or the
7150 same source file compiled for multiple output destinations, it is
7151 likely that the different parallel compilers will interfere with each
7152 other, and overwrite the temporary files. For instance:
7155 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
7156 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
7159 may result in @file{foo.i} and @file{foo.o} being written to
7160 simultaneously by both compilers.
7162 @item -save-temps=obj
7163 @opindex save-temps=obj
7164 Store the usual ``temporary'' intermediate files permanently. If the
7165 @option{-o} option is used, the temporary files are based on the
7166 object file. If the @option{-o} option is not used, the
7167 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
7172 gcc -save-temps=obj -c foo.c
7173 gcc -save-temps=obj -c bar.c -o dir/xbar.o
7174 gcc -save-temps=obj foobar.c -o dir2/yfoobar
7178 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
7179 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
7180 @file{dir2/yfoobar.o}.
7182 @item -time@r{[}=@var{file}@r{]}
7184 Report the CPU time taken by each subprocess in the compilation
7185 sequence. For C source files, this is the compiler proper and assembler
7186 (plus the linker if linking is done).
7188 Without the specification of an output file, the output looks like this:
7195 The first number on each line is the ``user time'', that is time spent
7196 executing the program itself. The second number is ``system time'',
7197 time spent executing operating system routines on behalf of the program.
7198 Both numbers are in seconds.
7200 With the specification of an output file, the output is appended to the
7201 named file, and it looks like this:
7204 0.12 0.01 cc1 @var{options}
7205 0.00 0.01 as @var{options}
7208 The ``user time'' and the ``system time'' are moved before the program
7209 name, and the options passed to the program are displayed, so that one
7210 can later tell what file was being compiled, and with which options.
7212 @item -fvar-tracking
7213 @opindex fvar-tracking
7214 Run variable tracking pass. It computes where variables are stored at each
7215 position in code. Better debugging information is then generated
7216 (if the debugging information format supports this information).
7218 It is enabled by default when compiling with optimization (@option{-Os},
7219 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7220 the debug info format supports it.
7222 @item -fvar-tracking-assignments
7223 @opindex fvar-tracking-assignments
7224 @opindex fno-var-tracking-assignments
7225 Annotate assignments to user variables early in the compilation and
7226 attempt to carry the annotations over throughout the compilation all the
7227 way to the end, in an attempt to improve debug information while
7228 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7230 It can be enabled even if var-tracking is disabled, in which case
7231 annotations are created and maintained, but discarded at the end.
7232 By default, this flag is enabled together with @option{-fvar-tracking},
7233 except when selective scheduling is enabled.
7235 @item -fvar-tracking-assignments-toggle
7236 @opindex fvar-tracking-assignments-toggle
7237 @opindex fno-var-tracking-assignments-toggle
7238 Toggle @option{-fvar-tracking-assignments}, in the same way that
7239 @option{-gtoggle} toggles @option{-g}.
7241 @item -print-file-name=@var{library}
7242 @opindex print-file-name
7243 Print the full absolute name of the library file @var{library} that
7244 would be used when linking---and don't do anything else. With this
7245 option, GCC does not compile or link anything; it just prints the
7248 @item -print-multi-directory
7249 @opindex print-multi-directory
7250 Print the directory name corresponding to the multilib selected by any
7251 other switches present in the command line. This directory is supposed
7252 to exist in @env{GCC_EXEC_PREFIX}.
7254 @item -print-multi-lib
7255 @opindex print-multi-lib
7256 Print the mapping from multilib directory names to compiler switches
7257 that enable them. The directory name is separated from the switches by
7258 @samp{;}, and each switch starts with an @samp{@@} instead of the
7259 @samp{-}, without spaces between multiple switches. This is supposed to
7260 ease shell processing.
7262 @item -print-multi-os-directory
7263 @opindex print-multi-os-directory
7264 Print the path to OS libraries for the selected
7265 multilib, relative to some @file{lib} subdirectory. If OS libraries are
7266 present in the @file{lib} subdirectory and no multilibs are used, this is
7267 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
7268 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
7269 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
7270 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
7272 @item -print-multiarch
7273 @opindex print-multiarch
7274 Print the path to OS libraries for the selected multiarch,
7275 relative to some @file{lib} subdirectory.
7277 @item -print-prog-name=@var{program}
7278 @opindex print-prog-name
7279 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
7281 @item -print-libgcc-file-name
7282 @opindex print-libgcc-file-name
7283 Same as @option{-print-file-name=libgcc.a}.
7285 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
7286 but you do want to link with @file{libgcc.a}. You can do:
7289 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
7292 @item -print-search-dirs
7293 @opindex print-search-dirs
7294 Print the name of the configured installation directory and a list of
7295 program and library directories @command{gcc} searches---and don't do anything else.
7297 This is useful when @command{gcc} prints the error message
7298 @samp{installation problem, cannot exec cpp0: No such file or directory}.
7299 To resolve this you either need to put @file{cpp0} and the other compiler
7300 components where @command{gcc} expects to find them, or you can set the environment
7301 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
7302 Don't forget the trailing @samp{/}.
7303 @xref{Environment Variables}.
7305 @item -print-sysroot
7306 @opindex print-sysroot
7307 Print the target sysroot directory that is used during
7308 compilation. This is the target sysroot specified either at configure
7309 time or using the @option{--sysroot} option, possibly with an extra
7310 suffix that depends on compilation options. If no target sysroot is
7311 specified, the option prints nothing.
7313 @item -print-sysroot-headers-suffix
7314 @opindex print-sysroot-headers-suffix
7315 Print the suffix added to the target sysroot when searching for
7316 headers, or give an error if the compiler is not configured with such
7317 a suffix---and don't do anything else.
7320 @opindex dumpmachine
7321 Print the compiler's target machine (for example,
7322 @samp{i686-pc-linux-gnu})---and don't do anything else.
7325 @opindex dumpversion
7326 Print the compiler version (for example, @code{3.0})---and don't do
7331 Print the compiler's built-in specs---and don't do anything else. (This
7332 is used when GCC itself is being built.) @xref{Spec Files}.
7334 @item -fno-eliminate-unused-debug-types
7335 @opindex feliminate-unused-debug-types
7336 @opindex fno-eliminate-unused-debug-types
7337 Normally, when producing DWARF 2 output, GCC avoids producing debug symbol
7338 output for types that are nowhere used in the source file being compiled.
7339 Sometimes it is useful to have GCC emit debugging
7340 information for all types declared in a compilation
7341 unit, regardless of whether or not they are actually used
7342 in that compilation unit, for example
7343 if, in the debugger, you want to cast a value to a type that is
7344 not actually used in your program (but is declared). More often,
7345 however, this results in a significant amount of wasted space.
7348 @node Optimize Options
7349 @section Options That Control Optimization
7350 @cindex optimize options
7351 @cindex options, optimization
7353 These options control various sorts of optimizations.
7355 Without any optimization option, the compiler's goal is to reduce the
7356 cost of compilation and to make debugging produce the expected
7357 results. Statements are independent: if you stop the program with a
7358 breakpoint between statements, you can then assign a new value to any
7359 variable or change the program counter to any other statement in the
7360 function and get exactly the results you expect from the source
7363 Turning on optimization flags makes the compiler attempt to improve
7364 the performance and/or code size at the expense of compilation time
7365 and possibly the ability to debug the program.
7367 The compiler performs optimization based on the knowledge it has of the
7368 program. Compiling multiple files at once to a single output file mode allows
7369 the compiler to use information gained from all of the files when compiling
7372 Not all optimizations are controlled directly by a flag. Only
7373 optimizations that have a flag are listed in this section.
7375 Most optimizations are only enabled if an @option{-O} level is set on
7376 the command line. Otherwise they are disabled, even if individual
7377 optimization flags are specified.
7379 Depending on the target and how GCC was configured, a slightly different
7380 set of optimizations may be enabled at each @option{-O} level than
7381 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
7382 to find out the exact set of optimizations that are enabled at each level.
7383 @xref{Overall Options}, for examples.
7390 Optimize. Optimizing compilation takes somewhat more time, and a lot
7391 more memory for a large function.
7393 With @option{-O}, the compiler tries to reduce code size and execution
7394 time, without performing any optimizations that take a great deal of
7397 @option{-O} turns on the following optimization flags:
7400 -fbranch-count-reg @gol
7401 -fcombine-stack-adjustments @gol
7403 -fcprop-registers @gol
7406 -fdelayed-branch @gol
7408 -fforward-propagate @gol
7409 -fguess-branch-probability @gol
7410 -fif-conversion2 @gol
7411 -fif-conversion @gol
7412 -finline-functions-called-once @gol
7413 -fipa-pure-const @gol
7415 -fipa-reference @gol
7416 -fmerge-constants @gol
7417 -fmove-loop-invariants @gol
7419 -fsplit-wide-types @gol
7424 -ftree-copy-prop @gol
7425 -ftree-copyrename @gol
7427 -ftree-dominator-opts @gol
7429 -ftree-forwprop @gol
7439 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
7440 where doing so does not interfere with debugging.
7444 Optimize even more. GCC performs nearly all supported optimizations
7445 that do not involve a space-speed tradeoff.
7446 As compared to @option{-O}, this option increases both compilation time
7447 and the performance of the generated code.
7449 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7450 also turns on the following optimization flags:
7451 @gccoptlist{-fthread-jumps @gol
7452 -falign-functions -falign-jumps @gol
7453 -falign-loops -falign-labels @gol
7456 -fcse-follow-jumps -fcse-skip-blocks @gol
7457 -fdelete-null-pointer-checks @gol
7458 -fdevirtualize -fdevirtualize-speculatively @gol
7459 -fexpensive-optimizations @gol
7460 -fgcse -fgcse-lm @gol
7461 -fhoist-adjacent-loads @gol
7462 -finline-small-functions @gol
7463 -findirect-inlining @gol
7465 -fipa-cp-alignment @gol
7468 -fisolate-erroneous-paths-dereference @gol
7470 -foptimize-sibling-calls @gol
7471 -foptimize-strlen @gol
7472 -fpartial-inlining @gol
7474 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
7475 -frerun-cse-after-loop @gol
7476 -fsched-interblock -fsched-spec @gol
7477 -fschedule-insns -fschedule-insns2 @gol
7478 -fstrict-aliasing -fstrict-overflow @gol
7479 -ftree-builtin-call-dce @gol
7480 -ftree-switch-conversion -ftree-tail-merge @gol
7485 Please note the warning under @option{-fgcse} about
7486 invoking @option{-O2} on programs that use computed gotos.
7490 Optimize yet more. @option{-O3} turns on all optimizations specified
7491 by @option{-O2} and also turns on the @option{-finline-functions},
7492 @option{-funswitch-loops}, @option{-fpredictive-commoning},
7493 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
7494 @option{-ftree-loop-distribute-patterns},
7495 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
7496 @option{-ftree-partial-pre} and @option{-fipa-cp-clone} options.
7500 Reduce compilation time and make debugging produce the expected
7501 results. This is the default.
7505 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7506 do not typically increase code size. It also performs further
7507 optimizations designed to reduce code size.
7509 @option{-Os} disables the following optimization flags:
7510 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7511 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
7512 -fprefetch-loop-arrays}
7516 Disregard strict standards compliance. @option{-Ofast} enables all
7517 @option{-O3} optimizations. It also enables optimizations that are not
7518 valid for all standard-compliant programs.
7519 It turns on @option{-ffast-math} and the Fortran-specific
7520 @option{-fno-protect-parens} and @option{-fstack-arrays}.
7524 Optimize debugging experience. @option{-Og} enables optimizations
7525 that do not interfere with debugging. It should be the optimization
7526 level of choice for the standard edit-compile-debug cycle, offering
7527 a reasonable level of optimization while maintaining fast compilation
7528 and a good debugging experience.
7530 If you use multiple @option{-O} options, with or without level numbers,
7531 the last such option is the one that is effective.
7534 Options of the form @option{-f@var{flag}} specify machine-independent
7535 flags. Most flags have both positive and negative forms; the negative
7536 form of @option{-ffoo} is @option{-fno-foo}. In the table
7537 below, only one of the forms is listed---the one you typically
7538 use. You can figure out the other form by either removing @samp{no-}
7541 The following options control specific optimizations. They are either
7542 activated by @option{-O} options or are related to ones that are. You
7543 can use the following flags in the rare cases when ``fine-tuning'' of
7544 optimizations to be performed is desired.
7547 @item -fno-defer-pop
7548 @opindex fno-defer-pop
7549 Always pop the arguments to each function call as soon as that function
7550 returns. For machines that must pop arguments after a function call,
7551 the compiler normally lets arguments accumulate on the stack for several
7552 function calls and pops them all at once.
7554 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7556 @item -fforward-propagate
7557 @opindex fforward-propagate
7558 Perform a forward propagation pass on RTL@. The pass tries to combine two
7559 instructions and checks if the result can be simplified. If loop unrolling
7560 is active, two passes are performed and the second is scheduled after
7563 This option is enabled by default at optimization levels @option{-O},
7564 @option{-O2}, @option{-O3}, @option{-Os}.
7566 @item -ffp-contract=@var{style}
7567 @opindex ffp-contract
7568 @option{-ffp-contract=off} disables floating-point expression contraction.
7569 @option{-ffp-contract=fast} enables floating-point expression contraction
7570 such as forming of fused multiply-add operations if the target has
7571 native support for them.
7572 @option{-ffp-contract=on} enables floating-point expression contraction
7573 if allowed by the language standard. This is currently not implemented
7574 and treated equal to @option{-ffp-contract=off}.
7576 The default is @option{-ffp-contract=fast}.
7578 @item -fomit-frame-pointer
7579 @opindex fomit-frame-pointer
7580 Don't keep the frame pointer in a register for functions that
7581 don't need one. This avoids the instructions to save, set up and
7582 restore frame pointers; it also makes an extra register available
7583 in many functions. @strong{It also makes debugging impossible on
7586 On some machines, such as the VAX, this flag has no effect, because
7587 the standard calling sequence automatically handles the frame pointer
7588 and nothing is saved by pretending it doesn't exist. The
7589 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
7590 whether a target machine supports this flag. @xref{Registers,,Register
7591 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
7593 The default setting (when not optimizing for
7594 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets is
7595 @option{-fomit-frame-pointer}. You can configure GCC with the
7596 @option{--enable-frame-pointer} configure option to change the default.
7598 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7600 @item -foptimize-sibling-calls
7601 @opindex foptimize-sibling-calls
7602 Optimize sibling and tail recursive calls.
7604 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7606 @item -foptimize-strlen
7607 @opindex foptimize-strlen
7608 Optimize various standard C string functions (e.g. @code{strlen},
7609 @code{strchr} or @code{strcpy}) and
7610 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7612 Enabled at levels @option{-O2}, @option{-O3}.
7616 Do not expand any functions inline apart from those marked with
7617 the @code{always_inline} attribute. This is the default when not
7620 Single functions can be exempted from inlining by marking them
7621 with the @code{noinline} attribute.
7623 @item -finline-small-functions
7624 @opindex finline-small-functions
7625 Integrate functions into their callers when their body is smaller than expected
7626 function call code (so overall size of program gets smaller). The compiler
7627 heuristically decides which functions are simple enough to be worth integrating
7628 in this way. This inlining applies to all functions, even those not declared
7631 Enabled at level @option{-O2}.
7633 @item -findirect-inlining
7634 @opindex findirect-inlining
7635 Inline also indirect calls that are discovered to be known at compile
7636 time thanks to previous inlining. This option has any effect only
7637 when inlining itself is turned on by the @option{-finline-functions}
7638 or @option{-finline-small-functions} options.
7640 Enabled at level @option{-O2}.
7642 @item -finline-functions
7643 @opindex finline-functions
7644 Consider all functions for inlining, even if they are not declared inline.
7645 The compiler heuristically decides which functions are worth integrating
7648 If all calls to a given function are integrated, and the function is
7649 declared @code{static}, then the function is normally not output as
7650 assembler code in its own right.
7652 Enabled at level @option{-O3}.
7654 @item -finline-functions-called-once
7655 @opindex finline-functions-called-once
7656 Consider all @code{static} functions called once for inlining into their
7657 caller even if they are not marked @code{inline}. If a call to a given
7658 function is integrated, then the function is not output as assembler code
7661 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7663 @item -fearly-inlining
7664 @opindex fearly-inlining
7665 Inline functions marked by @code{always_inline} and functions whose body seems
7666 smaller than the function call overhead early before doing
7667 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7668 makes profiling significantly cheaper and usually inlining faster on programs
7669 having large chains of nested wrapper functions.
7675 Perform interprocedural scalar replacement of aggregates, removal of
7676 unused parameters and replacement of parameters passed by reference
7677 by parameters passed by value.
7679 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7681 @item -finline-limit=@var{n}
7682 @opindex finline-limit
7683 By default, GCC limits the size of functions that can be inlined. This flag
7684 allows coarse control of this limit. @var{n} is the size of functions that
7685 can be inlined in number of pseudo instructions.
7687 Inlining is actually controlled by a number of parameters, which may be
7688 specified individually by using @option{--param @var{name}=@var{value}}.
7689 The @option{-finline-limit=@var{n}} option sets some of these parameters
7693 @item max-inline-insns-single
7694 is set to @var{n}/2.
7695 @item max-inline-insns-auto
7696 is set to @var{n}/2.
7699 See below for a documentation of the individual
7700 parameters controlling inlining and for the defaults of these parameters.
7702 @emph{Note:} there may be no value to @option{-finline-limit} that results
7703 in default behavior.
7705 @emph{Note:} pseudo instruction represents, in this particular context, an
7706 abstract measurement of function's size. In no way does it represent a count
7707 of assembly instructions and as such its exact meaning might change from one
7708 release to an another.
7710 @item -fno-keep-inline-dllexport
7711 @opindex fno-keep-inline-dllexport
7712 This is a more fine-grained version of @option{-fkeep-inline-functions},
7713 which applies only to functions that are declared using the @code{dllexport}
7714 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
7717 @item -fkeep-inline-functions
7718 @opindex fkeep-inline-functions
7719 In C, emit @code{static} functions that are declared @code{inline}
7720 into the object file, even if the function has been inlined into all
7721 of its callers. This switch does not affect functions using the
7722 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7723 inline functions into the object file.
7725 @item -fkeep-static-consts
7726 @opindex fkeep-static-consts
7727 Emit variables declared @code{static const} when optimization isn't turned
7728 on, even if the variables aren't referenced.
7730 GCC enables this option by default. If you want to force the compiler to
7731 check if a variable is referenced, regardless of whether or not
7732 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7734 @item -fmerge-constants
7735 @opindex fmerge-constants
7736 Attempt to merge identical constants (string constants and floating-point
7737 constants) across compilation units.
7739 This option is the default for optimized compilation if the assembler and
7740 linker support it. Use @option{-fno-merge-constants} to inhibit this
7743 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7745 @item -fmerge-all-constants
7746 @opindex fmerge-all-constants
7747 Attempt to merge identical constants and identical variables.
7749 This option implies @option{-fmerge-constants}. In addition to
7750 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7751 arrays or initialized constant variables with integral or floating-point
7752 types. Languages like C or C++ require each variable, including multiple
7753 instances of the same variable in recursive calls, to have distinct locations,
7754 so using this option results in non-conforming
7757 @item -fmodulo-sched
7758 @opindex fmodulo-sched
7759 Perform swing modulo scheduling immediately before the first scheduling
7760 pass. This pass looks at innermost loops and reorders their
7761 instructions by overlapping different iterations.
7763 @item -fmodulo-sched-allow-regmoves
7764 @opindex fmodulo-sched-allow-regmoves
7765 Perform more aggressive SMS-based modulo scheduling with register moves
7766 allowed. By setting this flag certain anti-dependences edges are
7767 deleted, which triggers the generation of reg-moves based on the
7768 life-range analysis. This option is effective only with
7769 @option{-fmodulo-sched} enabled.
7771 @item -fno-branch-count-reg
7772 @opindex fno-branch-count-reg
7773 Do not use ``decrement and branch'' instructions on a count register,
7774 but instead generate a sequence of instructions that decrement a
7775 register, compare it against zero, then branch based upon the result.
7776 This option is only meaningful on architectures that support such
7777 instructions, which include x86, PowerPC, IA-64 and S/390.
7779 Enabled by default at @option{-O1} and higher.
7781 The default is @option{-fbranch-count-reg}.
7783 @item -fno-function-cse
7784 @opindex fno-function-cse
7785 Do not put function addresses in registers; make each instruction that
7786 calls a constant function contain the function's address explicitly.
7788 This option results in less efficient code, but some strange hacks
7789 that alter the assembler output may be confused by the optimizations
7790 performed when this option is not used.
7792 The default is @option{-ffunction-cse}
7794 @item -fno-zero-initialized-in-bss
7795 @opindex fno-zero-initialized-in-bss
7796 If the target supports a BSS section, GCC by default puts variables that
7797 are initialized to zero into BSS@. This can save space in the resulting
7800 This option turns off this behavior because some programs explicitly
7801 rely on variables going to the data section---e.g., so that the
7802 resulting executable can find the beginning of that section and/or make
7803 assumptions based on that.
7805 The default is @option{-fzero-initialized-in-bss}.
7807 @item -fthread-jumps
7808 @opindex fthread-jumps
7809 Perform optimizations that check to see if a jump branches to a
7810 location where another comparison subsumed by the first is found. If
7811 so, the first branch is redirected to either the destination of the
7812 second branch or a point immediately following it, depending on whether
7813 the condition is known to be true or false.
7815 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7817 @item -fsplit-wide-types
7818 @opindex fsplit-wide-types
7819 When using a type that occupies multiple registers, such as @code{long
7820 long} on a 32-bit system, split the registers apart and allocate them
7821 independently. This normally generates better code for those types,
7822 but may make debugging more difficult.
7824 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7827 @item -fcse-follow-jumps
7828 @opindex fcse-follow-jumps
7829 In common subexpression elimination (CSE), scan through jump instructions
7830 when the target of the jump is not reached by any other path. For
7831 example, when CSE encounters an @code{if} statement with an
7832 @code{else} clause, CSE follows the jump when the condition
7835 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7837 @item -fcse-skip-blocks
7838 @opindex fcse-skip-blocks
7839 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
7840 follow jumps that conditionally skip over blocks. When CSE
7841 encounters a simple @code{if} statement with no else clause,
7842 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
7843 body of the @code{if}.
7845 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7847 @item -frerun-cse-after-loop
7848 @opindex frerun-cse-after-loop
7849 Re-run common subexpression elimination after loop optimizations are
7852 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7856 Perform a global common subexpression elimination pass.
7857 This pass also performs global constant and copy propagation.
7859 @emph{Note:} When compiling a program using computed gotos, a GCC
7860 extension, you may get better run-time performance if you disable
7861 the global common subexpression elimination pass by adding
7862 @option{-fno-gcse} to the command line.
7864 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7868 When @option{-fgcse-lm} is enabled, global common subexpression elimination
7869 attempts to move loads that are only killed by stores into themselves. This
7870 allows a loop containing a load/store sequence to be changed to a load outside
7871 the loop, and a copy/store within the loop.
7873 Enabled by default when @option{-fgcse} is enabled.
7877 When @option{-fgcse-sm} is enabled, a store motion pass is run after
7878 global common subexpression elimination. This pass attempts to move
7879 stores out of loops. When used in conjunction with @option{-fgcse-lm},
7880 loops containing a load/store sequence can be changed to a load before
7881 the loop and a store after the loop.
7883 Not enabled at any optimization level.
7887 When @option{-fgcse-las} is enabled, the global common subexpression
7888 elimination pass eliminates redundant loads that come after stores to the
7889 same memory location (both partial and full redundancies).
7891 Not enabled at any optimization level.
7893 @item -fgcse-after-reload
7894 @opindex fgcse-after-reload
7895 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
7896 pass is performed after reload. The purpose of this pass is to clean up
7899 @item -faggressive-loop-optimizations
7900 @opindex faggressive-loop-optimizations
7901 This option tells the loop optimizer to use language constraints to
7902 derive bounds for the number of iterations of a loop. This assumes that
7903 loop code does not invoke undefined behavior by for example causing signed
7904 integer overflows or out-of-bound array accesses. The bounds for the
7905 number of iterations of a loop are used to guide loop unrolling and peeling
7906 and loop exit test optimizations.
7907 This option is enabled by default.
7909 @item -funsafe-loop-optimizations
7910 @opindex funsafe-loop-optimizations
7911 This option tells the loop optimizer to assume that loop indices do not
7912 overflow, and that loops with nontrivial exit condition are not
7913 infinite. This enables a wider range of loop optimizations even if
7914 the loop optimizer itself cannot prove that these assumptions are valid.
7915 If you use @option{-Wunsafe-loop-optimizations}, the compiler warns you
7916 if it finds this kind of loop.
7918 @item -fcrossjumping
7919 @opindex fcrossjumping
7920 Perform cross-jumping transformation.
7921 This transformation unifies equivalent code and saves code size. The
7922 resulting code may or may not perform better than without cross-jumping.
7924 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7926 @item -fauto-inc-dec
7927 @opindex fauto-inc-dec
7928 Combine increments or decrements of addresses with memory accesses.
7929 This pass is always skipped on architectures that do not have
7930 instructions to support this. Enabled by default at @option{-O} and
7931 higher on architectures that support this.
7935 Perform dead code elimination (DCE) on RTL@.
7936 Enabled by default at @option{-O} and higher.
7940 Perform dead store elimination (DSE) on RTL@.
7941 Enabled by default at @option{-O} and higher.
7943 @item -fif-conversion
7944 @opindex fif-conversion
7945 Attempt to transform conditional jumps into branch-less equivalents. This
7946 includes use of conditional moves, min, max, set flags and abs instructions, and
7947 some tricks doable by standard arithmetics. The use of conditional execution
7948 on chips where it is available is controlled by @option{-fif-conversion2}.
7950 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7952 @item -fif-conversion2
7953 @opindex fif-conversion2
7954 Use conditional execution (where available) to transform conditional jumps into
7955 branch-less equivalents.
7957 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7959 @item -fdeclone-ctor-dtor
7960 @opindex fdeclone-ctor-dtor
7961 The C++ ABI requires multiple entry points for constructors and
7962 destructors: one for a base subobject, one for a complete object, and
7963 one for a virtual destructor that calls operator delete afterwards.
7964 For a hierarchy with virtual bases, the base and complete variants are
7965 clones, which means two copies of the function. With this option, the
7966 base and complete variants are changed to be thunks that call a common
7969 Enabled by @option{-Os}.
7971 @item -fdelete-null-pointer-checks
7972 @opindex fdelete-null-pointer-checks
7973 Assume that programs cannot safely dereference null pointers, and that
7974 no code or data element resides there. This enables simple constant
7975 folding optimizations at all optimization levels. In addition, other
7976 optimization passes in GCC use this flag to control global dataflow
7977 analyses that eliminate useless checks for null pointers; these assume
7978 that if a pointer is checked after it has already been dereferenced,
7981 Note however that in some environments this assumption is not true.
7982 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
7983 for programs that depend on that behavior.
7985 Some targets, especially embedded ones, disable this option at all levels.
7986 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
7987 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
7988 are enabled independently at different optimization levels.
7990 @item -fdevirtualize
7991 @opindex fdevirtualize
7992 Attempt to convert calls to virtual functions to direct calls. This
7993 is done both within a procedure and interprocedurally as part of
7994 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
7995 propagation (@option{-fipa-cp}).
7996 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7998 @item -fdevirtualize-speculatively
7999 @opindex fdevirtualize-speculatively
8000 Attempt to convert calls to virtual functions to speculative direct calls.
8001 Based on the analysis of the type inheritance graph, determine for a given call
8002 the set of likely targets. If the set is small, preferably of size 1, change
8003 the call into a conditional deciding between direct and indirect calls. The
8004 speculative calls enable more optimizations, such as inlining. When they seem
8005 useless after further optimization, they are converted back into original form.
8007 @item -fdevirtualize-at-ltrans
8008 @opindex fdevirtualize-at-ltrans
8009 Stream extra information needed for aggressive devirtualization when running
8010 the link-time optimizer in local transformation mode.
8011 This option enables more devirtualization but
8012 significantly increases the size of streamed data. For this reason it is
8013 disabled by default.
8015 @item -fexpensive-optimizations
8016 @opindex fexpensive-optimizations
8017 Perform a number of minor optimizations that are relatively expensive.
8019 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8023 Attempt to remove redundant extension instructions. This is especially
8024 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
8025 registers after writing to their lower 32-bit half.
8027 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
8028 @option{-O3}, @option{-Os}.
8030 @item -fno-lifetime-dse
8031 @opindex fno-lifetime-dse
8032 In C++ the value of an object is only affected by changes within its
8033 lifetime: when the constructor begins, the object has an indeterminate
8034 value, and any changes during the lifetime of the object are dead when
8035 the object is destroyed. Normally dead store elimination will take
8036 advantage of this; if your code relies on the value of the object
8037 storage persisting beyond the lifetime of the object, you can use this
8038 flag to disable this optimization.
8040 @item -flive-range-shrinkage
8041 @opindex flive-range-shrinkage
8042 Attempt to decrease register pressure through register live range
8043 shrinkage. This is helpful for fast processors with small or moderate
8046 @item -fira-algorithm=@var{algorithm}
8047 @opindex fira-algorithm
8048 Use the specified coloring algorithm for the integrated register
8049 allocator. The @var{algorithm} argument can be @samp{priority}, which
8050 specifies Chow's priority coloring, or @samp{CB}, which specifies
8051 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
8052 for all architectures, but for those targets that do support it, it is
8053 the default because it generates better code.
8055 @item -fira-region=@var{region}
8056 @opindex fira-region
8057 Use specified regions for the integrated register allocator. The
8058 @var{region} argument should be one of the following:
8063 Use all loops as register allocation regions.
8064 This can give the best results for machines with a small and/or
8065 irregular register set.
8068 Use all loops except for loops with small register pressure
8069 as the regions. This value usually gives
8070 the best results in most cases and for most architectures,
8071 and is enabled by default when compiling with optimization for speed
8072 (@option{-O}, @option{-O2}, @dots{}).
8075 Use all functions as a single region.
8076 This typically results in the smallest code size, and is enabled by default for
8077 @option{-Os} or @option{-O0}.
8081 @item -fira-hoist-pressure
8082 @opindex fira-hoist-pressure
8083 Use IRA to evaluate register pressure in the code hoisting pass for
8084 decisions to hoist expressions. This option usually results in smaller
8085 code, but it can slow the compiler down.
8087 This option is enabled at level @option{-Os} for all targets.
8089 @item -fira-loop-pressure
8090 @opindex fira-loop-pressure
8091 Use IRA to evaluate register pressure in loops for decisions to move
8092 loop invariants. This option usually results in generation
8093 of faster and smaller code on machines with large register files (>= 32
8094 registers), but it can slow the compiler down.
8096 This option is enabled at level @option{-O3} for some targets.
8098 @item -fno-ira-share-save-slots
8099 @opindex fno-ira-share-save-slots
8100 Disable sharing of stack slots used for saving call-used hard
8101 registers living through a call. Each hard register gets a
8102 separate stack slot, and as a result function stack frames are
8105 @item -fno-ira-share-spill-slots
8106 @opindex fno-ira-share-spill-slots
8107 Disable sharing of stack slots allocated for pseudo-registers. Each
8108 pseudo-register that does not get a hard register gets a separate
8109 stack slot, and as a result function stack frames are larger.
8111 @item -fira-verbose=@var{n}
8112 @opindex fira-verbose
8113 Control the verbosity of the dump file for the integrated register allocator.
8114 The default value is 5. If the value @var{n} is greater or equal to 10,
8115 the dump output is sent to stderr using the same format as @var{n} minus 10.
8119 Enable CFG-sensitive rematerialization in LRA. Instead of loading
8120 values of spilled pseudos, LRA tries to rematerialize (recalculate)
8121 values if it is profitable.
8123 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8125 @item -fdelayed-branch
8126 @opindex fdelayed-branch
8127 If supported for the target machine, attempt to reorder instructions
8128 to exploit instruction slots available after delayed branch
8131 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8133 @item -fschedule-insns
8134 @opindex fschedule-insns
8135 If supported for the target machine, attempt to reorder instructions to
8136 eliminate execution stalls due to required data being unavailable. This
8137 helps machines that have slow floating point or memory load instructions
8138 by allowing other instructions to be issued until the result of the load
8139 or floating-point instruction is required.
8141 Enabled at levels @option{-O2}, @option{-O3}.
8143 @item -fschedule-insns2
8144 @opindex fschedule-insns2
8145 Similar to @option{-fschedule-insns}, but requests an additional pass of
8146 instruction scheduling after register allocation has been done. This is
8147 especially useful on machines with a relatively small number of
8148 registers and where memory load instructions take more than one cycle.
8150 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8152 @item -fno-sched-interblock
8153 @opindex fno-sched-interblock
8154 Don't schedule instructions across basic blocks. This is normally
8155 enabled by default when scheduling before register allocation, i.e.@:
8156 with @option{-fschedule-insns} or at @option{-O2} or higher.
8158 @item -fno-sched-spec
8159 @opindex fno-sched-spec
8160 Don't allow speculative motion of non-load instructions. This is normally
8161 enabled by default when scheduling before register allocation, i.e.@:
8162 with @option{-fschedule-insns} or at @option{-O2} or higher.
8164 @item -fsched-pressure
8165 @opindex fsched-pressure
8166 Enable register pressure sensitive insn scheduling before register
8167 allocation. This only makes sense when scheduling before register
8168 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
8169 @option{-O2} or higher. Usage of this option can improve the
8170 generated code and decrease its size by preventing register pressure
8171 increase above the number of available hard registers and subsequent
8172 spills in register allocation.
8174 @item -fsched-spec-load
8175 @opindex fsched-spec-load
8176 Allow speculative motion of some load instructions. This only makes
8177 sense when scheduling before register allocation, i.e.@: with
8178 @option{-fschedule-insns} or at @option{-O2} or higher.
8180 @item -fsched-spec-load-dangerous
8181 @opindex fsched-spec-load-dangerous
8182 Allow speculative motion of more load instructions. This only makes
8183 sense when scheduling before register allocation, i.e.@: with
8184 @option{-fschedule-insns} or at @option{-O2} or higher.
8186 @item -fsched-stalled-insns
8187 @itemx -fsched-stalled-insns=@var{n}
8188 @opindex fsched-stalled-insns
8189 Define how many insns (if any) can be moved prematurely from the queue
8190 of stalled insns into the ready list during the second scheduling pass.
8191 @option{-fno-sched-stalled-insns} means that no insns are moved
8192 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
8193 on how many queued insns can be moved prematurely.
8194 @option{-fsched-stalled-insns} without a value is equivalent to
8195 @option{-fsched-stalled-insns=1}.
8197 @item -fsched-stalled-insns-dep
8198 @itemx -fsched-stalled-insns-dep=@var{n}
8199 @opindex fsched-stalled-insns-dep
8200 Define how many insn groups (cycles) are examined for a dependency
8201 on a stalled insn that is a candidate for premature removal from the queue
8202 of stalled insns. This has an effect only during the second scheduling pass,
8203 and only if @option{-fsched-stalled-insns} is used.
8204 @option{-fno-sched-stalled-insns-dep} is equivalent to
8205 @option{-fsched-stalled-insns-dep=0}.
8206 @option{-fsched-stalled-insns-dep} without a value is equivalent to
8207 @option{-fsched-stalled-insns-dep=1}.
8209 @item -fsched2-use-superblocks
8210 @opindex fsched2-use-superblocks
8211 When scheduling after register allocation, use superblock scheduling.
8212 This allows motion across basic block boundaries,
8213 resulting in faster schedules. This option is experimental, as not all machine
8214 descriptions used by GCC model the CPU closely enough to avoid unreliable
8215 results from the algorithm.
8217 This only makes sense when scheduling after register allocation, i.e.@: with
8218 @option{-fschedule-insns2} or at @option{-O2} or higher.
8220 @item -fsched-group-heuristic
8221 @opindex fsched-group-heuristic
8222 Enable the group heuristic in the scheduler. This heuristic favors
8223 the instruction that belongs to a schedule group. This is enabled
8224 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8225 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8227 @item -fsched-critical-path-heuristic
8228 @opindex fsched-critical-path-heuristic
8229 Enable the critical-path heuristic in the scheduler. This heuristic favors
8230 instructions on the critical path. This is enabled by default when
8231 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8232 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8234 @item -fsched-spec-insn-heuristic
8235 @opindex fsched-spec-insn-heuristic
8236 Enable the speculative instruction heuristic in the scheduler. This
8237 heuristic favors speculative instructions with greater dependency weakness.
8238 This is enabled by default when scheduling is enabled, i.e.@:
8239 with @option{-fschedule-insns} or @option{-fschedule-insns2}
8240 or at @option{-O2} or higher.
8242 @item -fsched-rank-heuristic
8243 @opindex fsched-rank-heuristic
8244 Enable the rank heuristic in the scheduler. This heuristic favors
8245 the instruction belonging to a basic block with greater size or frequency.
8246 This is enabled by default when scheduling is enabled, i.e.@:
8247 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8248 at @option{-O2} or higher.
8250 @item -fsched-last-insn-heuristic
8251 @opindex fsched-last-insn-heuristic
8252 Enable the last-instruction heuristic in the scheduler. This heuristic
8253 favors the instruction that is less dependent on the last instruction
8254 scheduled. This is enabled by default when scheduling is enabled,
8255 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8256 at @option{-O2} or higher.
8258 @item -fsched-dep-count-heuristic
8259 @opindex fsched-dep-count-heuristic
8260 Enable the dependent-count heuristic in the scheduler. This heuristic
8261 favors the instruction that has more instructions depending on it.
8262 This is enabled by default when scheduling is enabled, i.e.@:
8263 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8264 at @option{-O2} or higher.
8266 @item -freschedule-modulo-scheduled-loops
8267 @opindex freschedule-modulo-scheduled-loops
8268 Modulo scheduling is performed before traditional scheduling. If a loop
8269 is modulo scheduled, later scheduling passes may change its schedule.
8270 Use this option to control that behavior.
8272 @item -fselective-scheduling
8273 @opindex fselective-scheduling
8274 Schedule instructions using selective scheduling algorithm. Selective
8275 scheduling runs instead of the first scheduler pass.
8277 @item -fselective-scheduling2
8278 @opindex fselective-scheduling2
8279 Schedule instructions using selective scheduling algorithm. Selective
8280 scheduling runs instead of the second scheduler pass.
8282 @item -fsel-sched-pipelining
8283 @opindex fsel-sched-pipelining
8284 Enable software pipelining of innermost loops during selective scheduling.
8285 This option has no effect unless one of @option{-fselective-scheduling} or
8286 @option{-fselective-scheduling2} is turned on.
8288 @item -fsel-sched-pipelining-outer-loops
8289 @opindex fsel-sched-pipelining-outer-loops
8290 When pipelining loops during selective scheduling, also pipeline outer loops.
8291 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
8293 @item -fsemantic-interposition
8294 @opindex fsemantic-interposition
8295 Some object formats, like ELF, allow interposing of symbols by the
8297 This means that for symbols exported from the DSO, the compiler cannot perform
8298 interprocedural propagation, inlining and other optimizations in anticipation
8299 that the function or variable in question may change. While this feature is
8300 useful, for example, to rewrite memory allocation functions by a debugging
8301 implementation, it is expensive in the terms of code quality.
8302 With @option{-fno-semantic-interposition} the compiler assumes that
8303 if interposition happens for functions the overwriting function will have
8304 precisely the same semantics (and side effects).
8305 Similarly if interposition happens
8306 for variables, the constructor of the variable will be the same. The flag
8307 has no effect for functions explicitly declared inline
8308 (where it is never allowed for interposition to change semantics)
8309 and for symbols explicitly declared weak.
8312 @opindex fshrink-wrap
8313 Emit function prologues only before parts of the function that need it,
8314 rather than at the top of the function. This flag is enabled by default at
8315 @option{-O} and higher.
8317 @item -fcaller-saves
8318 @opindex fcaller-saves
8319 Enable allocation of values to registers that are clobbered by
8320 function calls, by emitting extra instructions to save and restore the
8321 registers around such calls. Such allocation is done only when it
8322 seems to result in better code.
8324 This option is always enabled by default on certain machines, usually
8325 those which have no call-preserved registers to use instead.
8327 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8329 @item -fcombine-stack-adjustments
8330 @opindex fcombine-stack-adjustments
8331 Tracks stack adjustments (pushes and pops) and stack memory references
8332 and then tries to find ways to combine them.
8334 Enabled by default at @option{-O1} and higher.
8338 Use caller save registers for allocation if those registers are not used by
8339 any called function. In that case it is not necessary to save and restore
8340 them around calls. This is only possible if called functions are part of
8341 same compilation unit as current function and they are compiled before it.
8343 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8345 @item -fconserve-stack
8346 @opindex fconserve-stack
8347 Attempt to minimize stack usage. The compiler attempts to use less
8348 stack space, even if that makes the program slower. This option
8349 implies setting the @option{large-stack-frame} parameter to 100
8350 and the @option{large-stack-frame-growth} parameter to 400.
8352 @item -ftree-reassoc
8353 @opindex ftree-reassoc
8354 Perform reassociation on trees. This flag is enabled by default
8355 at @option{-O} and higher.
8359 Perform partial redundancy elimination (PRE) on trees. This flag is
8360 enabled by default at @option{-O2} and @option{-O3}.
8362 @item -ftree-partial-pre
8363 @opindex ftree-partial-pre
8364 Make partial redundancy elimination (PRE) more aggressive. This flag is
8365 enabled by default at @option{-O3}.
8367 @item -ftree-forwprop
8368 @opindex ftree-forwprop
8369 Perform forward propagation on trees. This flag is enabled by default
8370 at @option{-O} and higher.
8374 Perform full redundancy elimination (FRE) on trees. The difference
8375 between FRE and PRE is that FRE only considers expressions
8376 that are computed on all paths leading to the redundant computation.
8377 This analysis is faster than PRE, though it exposes fewer redundancies.
8378 This flag is enabled by default at @option{-O} and higher.
8380 @item -ftree-phiprop
8381 @opindex ftree-phiprop
8382 Perform hoisting of loads from conditional pointers on trees. This
8383 pass is enabled by default at @option{-O} and higher.
8385 @item -fhoist-adjacent-loads
8386 @opindex fhoist-adjacent-loads
8387 Speculatively hoist loads from both branches of an if-then-else if the
8388 loads are from adjacent locations in the same structure and the target
8389 architecture has a conditional move instruction. This flag is enabled
8390 by default at @option{-O2} and higher.
8392 @item -ftree-copy-prop
8393 @opindex ftree-copy-prop
8394 Perform copy propagation on trees. This pass eliminates unnecessary
8395 copy operations. This flag is enabled by default at @option{-O} and
8398 @item -fipa-pure-const
8399 @opindex fipa-pure-const
8400 Discover which functions are pure or constant.
8401 Enabled by default at @option{-O} and higher.
8403 @item -fipa-reference
8404 @opindex fipa-reference
8405 Discover which static variables do not escape the
8407 Enabled by default at @option{-O} and higher.
8411 Perform interprocedural pointer analysis and interprocedural modification
8412 and reference analysis. This option can cause excessive memory and
8413 compile-time usage on large compilation units. It is not enabled by
8414 default at any optimization level.
8417 @opindex fipa-profile
8418 Perform interprocedural profile propagation. The functions called only from
8419 cold functions are marked as cold. Also functions executed once (such as
8420 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8421 functions and loop less parts of functions executed once are then optimized for
8423 Enabled by default at @option{-O} and higher.
8427 Perform interprocedural constant propagation.
8428 This optimization analyzes the program to determine when values passed
8429 to functions are constants and then optimizes accordingly.
8430 This optimization can substantially increase performance
8431 if the application has constants passed to functions.
8432 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8434 @item -fipa-cp-clone
8435 @opindex fipa-cp-clone
8436 Perform function cloning to make interprocedural constant propagation stronger.
8437 When enabled, interprocedural constant propagation performs function cloning
8438 when externally visible function can be called with constant arguments.
8439 Because this optimization can create multiple copies of functions,
8440 it may significantly increase code size
8441 (see @option{--param ipcp-unit-growth=@var{value}}).
8442 This flag is enabled by default at @option{-O3}.
8444 @item -fipa-cp-alignment
8445 @opindex -fipa-cp-alignment
8446 When enabled, this optimization propagates alignment of function
8447 parameters to support better vectorization and string operations.
8449 This flag is enabled by default at @option{-O2} and @option{-Os}. It
8450 requires that @option{-fipa-cp} is enabled.
8454 Perform Identical Code Folding for functions and read-only variables.
8455 The optimization reduces code size and may disturb unwind stacks by replacing
8456 a function by equivalent one with a different name. The optimization works
8457 more effectively with link time optimization enabled.
8459 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8460 works on different levels and thus the optimizations are not same - there are
8461 equivalences that are found only by GCC and equivalences found only by Gold.
8463 This flag is enabled by default at @option{-O2} and @option{-Os}.
8465 @item -fisolate-erroneous-paths-dereference
8466 @opindex fisolate-erroneous-paths-dereference
8467 Detect paths that trigger erroneous or undefined behavior due to
8468 dereferencing a null pointer. Isolate those paths from the main control
8469 flow and turn the statement with erroneous or undefined behavior into a trap.
8470 This flag is enabled by default at @option{-O2} and higher.
8472 @item -fisolate-erroneous-paths-attribute
8473 @opindex fisolate-erroneous-paths-attribute
8474 Detect paths that trigger erroneous or undefined behavior due a null value
8475 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
8476 attribute. Isolate those paths from the main control flow and turn the
8477 statement with erroneous or undefined behavior into a trap. This is not
8478 currently enabled, but may be enabled by @option{-O2} in the future.
8482 Perform forward store motion on trees. This flag is
8483 enabled by default at @option{-O} and higher.
8485 @item -ftree-bit-ccp
8486 @opindex ftree-bit-ccp
8487 Perform sparse conditional bit constant propagation on trees and propagate
8488 pointer alignment information.
8489 This pass only operates on local scalar variables and is enabled by default
8490 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8494 Perform sparse conditional constant propagation (CCP) on trees. This
8495 pass only operates on local scalar variables and is enabled by default
8496 at @option{-O} and higher.
8499 @opindex fssa-phiopt
8500 Perform pattern matching on SSA PHI nodes to optimize conditional
8501 code. This pass is enabled by default at @option{-O} and higher.
8503 @item -ftree-switch-conversion
8504 @opindex ftree-switch-conversion
8505 Perform conversion of simple initializations in a switch to
8506 initializations from a scalar array. This flag is enabled by default
8507 at @option{-O2} and higher.
8509 @item -ftree-tail-merge
8510 @opindex ftree-tail-merge
8511 Look for identical code sequences. When found, replace one with a jump to the
8512 other. This optimization is known as tail merging or cross jumping. This flag
8513 is enabled by default at @option{-O2} and higher. The compilation time
8515 be limited using @option{max-tail-merge-comparisons} parameter and
8516 @option{max-tail-merge-iterations} parameter.
8520 Perform dead code elimination (DCE) on trees. This flag is enabled by
8521 default at @option{-O} and higher.
8523 @item -ftree-builtin-call-dce
8524 @opindex ftree-builtin-call-dce
8525 Perform conditional dead code elimination (DCE) for calls to built-in functions
8526 that may set @code{errno} but are otherwise side-effect free. This flag is
8527 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8530 @item -ftree-dominator-opts
8531 @opindex ftree-dominator-opts
8532 Perform a variety of simple scalar cleanups (constant/copy
8533 propagation, redundancy elimination, range propagation and expression
8534 simplification) based on a dominator tree traversal. This also
8535 performs jump threading (to reduce jumps to jumps). This flag is
8536 enabled by default at @option{-O} and higher.
8540 Perform dead store elimination (DSE) on trees. A dead store is a store into
8541 a memory location that is later overwritten by another store without
8542 any intervening loads. In this case the earlier store can be deleted. This
8543 flag is enabled by default at @option{-O} and higher.
8547 Perform loop header copying on trees. This is beneficial since it increases
8548 effectiveness of code motion optimizations. It also saves one jump. This flag
8549 is enabled by default at @option{-O} and higher. It is not enabled
8550 for @option{-Os}, since it usually increases code size.
8552 @item -ftree-loop-optimize
8553 @opindex ftree-loop-optimize
8554 Perform loop optimizations on trees. This flag is enabled by default
8555 at @option{-O} and higher.
8557 @item -ftree-loop-linear
8558 @opindex ftree-loop-linear
8559 Perform loop interchange transformations on tree. Same as
8560 @option{-floop-interchange}. To use this code transformation, GCC has
8561 to be configured with @option{--with-isl} to enable the Graphite loop
8562 transformation infrastructure.
8564 @item -floop-interchange
8565 @opindex floop-interchange
8566 Perform loop interchange transformations on loops. Interchanging two
8567 nested loops switches the inner and outer loops. For example, given a
8572 A(J, I) = A(J, I) * C
8577 loop interchange transforms the loop as if it were written:
8581 A(J, I) = A(J, I) * C
8585 which can be beneficial when @code{N} is larger than the caches,
8586 because in Fortran, the elements of an array are stored in memory
8587 contiguously by column, and the original loop iterates over rows,
8588 potentially creating at each access a cache miss. This optimization
8589 applies to all the languages supported by GCC and is not limited to
8590 Fortran. To use this code transformation, GCC has to be configured
8591 with @option{--with-isl} to enable the Graphite loop transformation
8594 @item -floop-strip-mine
8595 @opindex floop-strip-mine
8596 Perform loop strip mining transformations on loops. Strip mining
8597 splits a loop into two nested loops. The outer loop has strides
8598 equal to the strip size and the inner loop has strides of the
8599 original loop within a strip. The strip length can be changed
8600 using the @option{loop-block-tile-size} parameter. For example,
8608 loop strip mining transforms the loop as if it were written:
8611 DO I = II, min (II + 50, N)
8616 This optimization applies to all the languages supported by GCC and is
8617 not limited to Fortran. To use this code transformation, GCC has to
8618 be configured with @option{--with-isl} to enable the Graphite loop
8619 transformation infrastructure.
8622 @opindex floop-block
8623 Perform loop blocking transformations on loops. Blocking strip mines
8624 each loop in the loop nest such that the memory accesses of the
8625 element loops fit inside caches. The strip length can be changed
8626 using the @option{loop-block-tile-size} parameter. For example, given
8631 A(J, I) = B(I) + C(J)
8636 loop blocking transforms the loop as if it were written:
8640 DO I = II, min (II + 50, N)
8641 DO J = JJ, min (JJ + 50, M)
8642 A(J, I) = B(I) + C(J)
8648 which can be beneficial when @code{M} is larger than the caches,
8649 because the innermost loop iterates over a smaller amount of data
8650 which can be kept in the caches. This optimization applies to all the
8651 languages supported by GCC and is not limited to Fortran. To use this
8652 code transformation, GCC has to be configured with @option{--with-isl}
8653 to enable the Graphite loop transformation infrastructure.
8655 @item -fgraphite-identity
8656 @opindex fgraphite-identity
8657 Enable the identity transformation for graphite. For every SCoP we generate
8658 the polyhedral representation and transform it back to gimple. Using
8659 @option{-fgraphite-identity} we can check the costs or benefits of the
8660 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8661 are also performed by the code generator ISL, like index splitting and
8662 dead code elimination in loops.
8664 @item -floop-nest-optimize
8665 @opindex floop-nest-optimize
8666 Enable the ISL based loop nest optimizer. This is a generic loop nest
8667 optimizer based on the Pluto optimization algorithms. It calculates a loop
8668 structure optimized for data-locality and parallelism. This option
8671 @item -floop-unroll-and-jam
8672 @opindex floop-unroll-and-jam
8673 Enable unroll and jam for the ISL based loop nest optimizer. The unroll
8674 factor can be changed using the @option{loop-unroll-jam-size} parameter.
8675 The unrolled dimension (counting from the most inner one) can be changed
8676 using the @option{loop-unroll-jam-depth} parameter. .
8678 @item -floop-parallelize-all
8679 @opindex floop-parallelize-all
8680 Use the Graphite data dependence analysis to identify loops that can
8681 be parallelized. Parallelize all the loops that can be analyzed to
8682 not contain loop carried dependences without checking that it is
8683 profitable to parallelize the loops.
8685 @item -fcheck-data-deps
8686 @opindex fcheck-data-deps
8687 Compare the results of several data dependence analyzers. This option
8688 is used for debugging the data dependence analyzers.
8690 @item -ftree-loop-if-convert
8691 @opindex ftree-loop-if-convert
8692 Attempt to transform conditional jumps in the innermost loops to
8693 branch-less equivalents. The intent is to remove control-flow from
8694 the innermost loops in order to improve the ability of the
8695 vectorization pass to handle these loops. This is enabled by default
8696 if vectorization is enabled.
8698 @item -ftree-loop-if-convert-stores
8699 @opindex ftree-loop-if-convert-stores
8700 Attempt to also if-convert conditional jumps containing memory writes.
8701 This transformation can be unsafe for multi-threaded programs as it
8702 transforms conditional memory writes into unconditional memory writes.
8705 for (i = 0; i < N; i++)
8711 for (i = 0; i < N; i++)
8712 A[i] = cond ? expr : A[i];
8714 potentially producing data races.
8716 @item -ftree-loop-distribution
8717 @opindex ftree-loop-distribution
8718 Perform loop distribution. This flag can improve cache performance on
8719 big loop bodies and allow further loop optimizations, like
8720 parallelization or vectorization, to take place. For example, the loop
8737 @item -ftree-loop-distribute-patterns
8738 @opindex ftree-loop-distribute-patterns
8739 Perform loop distribution of patterns that can be code generated with
8740 calls to a library. This flag is enabled by default at @option{-O3}.
8742 This pass distributes the initialization loops and generates a call to
8743 memset zero. For example, the loop
8759 and the initialization loop is transformed into a call to memset zero.
8761 @item -ftree-loop-im
8762 @opindex ftree-loop-im
8763 Perform loop invariant motion on trees. This pass moves only invariants that
8764 are hard to handle at RTL level (function calls, operations that expand to
8765 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8766 operands of conditions that are invariant out of the loop, so that we can use
8767 just trivial invariantness analysis in loop unswitching. The pass also includes
8770 @item -ftree-loop-ivcanon
8771 @opindex ftree-loop-ivcanon
8772 Create a canonical counter for number of iterations in loops for which
8773 determining number of iterations requires complicated analysis. Later
8774 optimizations then may determine the number easily. Useful especially
8775 in connection with unrolling.
8779 Perform induction variable optimizations (strength reduction, induction
8780 variable merging and induction variable elimination) on trees.
8782 @item -ftree-parallelize-loops=n
8783 @opindex ftree-parallelize-loops
8784 Parallelize loops, i.e., split their iteration space to run in n threads.
8785 This is only possible for loops whose iterations are independent
8786 and can be arbitrarily reordered. The optimization is only
8787 profitable on multiprocessor machines, for loops that are CPU-intensive,
8788 rather than constrained e.g.@: by memory bandwidth. This option
8789 implies @option{-pthread}, and thus is only supported on targets
8790 that have support for @option{-pthread}.
8794 Perform function-local points-to analysis on trees. This flag is
8795 enabled by default at @option{-O} and higher.
8799 Perform scalar replacement of aggregates. This pass replaces structure
8800 references with scalars to prevent committing structures to memory too
8801 early. This flag is enabled by default at @option{-O} and higher.
8803 @item -ftree-copyrename
8804 @opindex ftree-copyrename
8805 Perform copy renaming on trees. This pass attempts to rename compiler
8806 temporaries to other variables at copy locations, usually resulting in
8807 variable names which more closely resemble the original variables. This flag
8808 is enabled by default at @option{-O} and higher.
8810 @item -ftree-coalesce-inlined-vars
8811 @opindex ftree-coalesce-inlined-vars
8812 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8813 combine small user-defined variables too, but only if they are inlined
8814 from other functions. It is a more limited form of
8815 @option{-ftree-coalesce-vars}. This may harm debug information of such
8816 inlined variables, but it keeps variables of the inlined-into
8817 function apart from each other, such that they are more likely to
8818 contain the expected values in a debugging session.
8820 @item -ftree-coalesce-vars
8821 @opindex ftree-coalesce-vars
8822 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8823 combine small user-defined variables too, instead of just compiler
8824 temporaries. This may severely limit the ability to debug an optimized
8825 program compiled with @option{-fno-var-tracking-assignments}. In the
8826 negated form, this flag prevents SSA coalescing of user variables,
8827 including inlined ones. This option is enabled by default.
8831 Perform temporary expression replacement during the SSA->normal phase. Single
8832 use/single def temporaries are replaced at their use location with their
8833 defining expression. This results in non-GIMPLE code, but gives the expanders
8834 much more complex trees to work on resulting in better RTL generation. This is
8835 enabled by default at @option{-O} and higher.
8839 Perform straight-line strength reduction on trees. This recognizes related
8840 expressions involving multiplications and replaces them by less expensive
8841 calculations when possible. This is enabled by default at @option{-O} and
8844 @item -ftree-vectorize
8845 @opindex ftree-vectorize
8846 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8847 and @option{-ftree-slp-vectorize} if not explicitly specified.
8849 @item -ftree-loop-vectorize
8850 @opindex ftree-loop-vectorize
8851 Perform loop vectorization on trees. This flag is enabled by default at
8852 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8854 @item -ftree-slp-vectorize
8855 @opindex ftree-slp-vectorize
8856 Perform basic block vectorization on trees. This flag is enabled by default at
8857 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8859 @item -fvect-cost-model=@var{model}
8860 @opindex fvect-cost-model
8861 Alter the cost model used for vectorization. The @var{model} argument
8862 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
8863 With the @samp{unlimited} model the vectorized code-path is assumed
8864 to be profitable while with the @samp{dynamic} model a runtime check
8865 guards the vectorized code-path to enable it only for iteration
8866 counts that will likely execute faster than when executing the original
8867 scalar loop. The @samp{cheap} model disables vectorization of
8868 loops where doing so would be cost prohibitive for example due to
8869 required runtime checks for data dependence or alignment but otherwise
8870 is equal to the @samp{dynamic} model.
8871 The default cost model depends on other optimization flags and is
8872 either @samp{dynamic} or @samp{cheap}.
8874 @item -fsimd-cost-model=@var{model}
8875 @opindex fsimd-cost-model
8876 Alter the cost model used for vectorization of loops marked with the OpenMP
8877 or Cilk Plus simd directive. The @var{model} argument should be one of
8878 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
8879 have the same meaning as described in @option{-fvect-cost-model} and by
8880 default a cost model defined with @option{-fvect-cost-model} is used.
8884 Perform Value Range Propagation on trees. This is similar to the
8885 constant propagation pass, but instead of values, ranges of values are
8886 propagated. This allows the optimizers to remove unnecessary range
8887 checks like array bound checks and null pointer checks. This is
8888 enabled by default at @option{-O2} and higher. Null pointer check
8889 elimination is only done if @option{-fdelete-null-pointer-checks} is
8892 @item -fsplit-ivs-in-unroller
8893 @opindex fsplit-ivs-in-unroller
8894 Enables expression of values of induction variables in later iterations
8895 of the unrolled loop using the value in the first iteration. This breaks
8896 long dependency chains, thus improving efficiency of the scheduling passes.
8898 A combination of @option{-fweb} and CSE is often sufficient to obtain the
8899 same effect. However, that is not reliable in cases where the loop body
8900 is more complicated than a single basic block. It also does not work at all
8901 on some architectures due to restrictions in the CSE pass.
8903 This optimization is enabled by default.
8905 @item -fvariable-expansion-in-unroller
8906 @opindex fvariable-expansion-in-unroller
8907 With this option, the compiler creates multiple copies of some
8908 local variables when unrolling a loop, which can result in superior code.
8910 @item -fpartial-inlining
8911 @opindex fpartial-inlining
8912 Inline parts of functions. This option has any effect only
8913 when inlining itself is turned on by the @option{-finline-functions}
8914 or @option{-finline-small-functions} options.
8916 Enabled at level @option{-O2}.
8918 @item -fpredictive-commoning
8919 @opindex fpredictive-commoning
8920 Perform predictive commoning optimization, i.e., reusing computations
8921 (especially memory loads and stores) performed in previous
8922 iterations of loops.
8924 This option is enabled at level @option{-O3}.
8926 @item -fprefetch-loop-arrays
8927 @opindex fprefetch-loop-arrays
8928 If supported by the target machine, generate instructions to prefetch
8929 memory to improve the performance of loops that access large arrays.
8931 This option may generate better or worse code; results are highly
8932 dependent on the structure of loops within the source code.
8934 Disabled at level @option{-Os}.
8937 @itemx -fno-peephole2
8938 @opindex fno-peephole
8939 @opindex fno-peephole2
8940 Disable any machine-specific peephole optimizations. The difference
8941 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
8942 are implemented in the compiler; some targets use one, some use the
8943 other, a few use both.
8945 @option{-fpeephole} is enabled by default.
8946 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8948 @item -fno-guess-branch-probability
8949 @opindex fno-guess-branch-probability
8950 Do not guess branch probabilities using heuristics.
8952 GCC uses heuristics to guess branch probabilities if they are
8953 not provided by profiling feedback (@option{-fprofile-arcs}). These
8954 heuristics are based on the control flow graph. If some branch probabilities
8955 are specified by @code{__builtin_expect}, then the heuristics are
8956 used to guess branch probabilities for the rest of the control flow graph,
8957 taking the @code{__builtin_expect} info into account. The interactions
8958 between the heuristics and @code{__builtin_expect} can be complex, and in
8959 some cases, it may be useful to disable the heuristics so that the effects
8960 of @code{__builtin_expect} are easier to understand.
8962 The default is @option{-fguess-branch-probability} at levels
8963 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8965 @item -freorder-blocks
8966 @opindex freorder-blocks
8967 Reorder basic blocks in the compiled function in order to reduce number of
8968 taken branches and improve code locality.
8970 Enabled at levels @option{-O2}, @option{-O3}.
8972 @item -freorder-blocks-and-partition
8973 @opindex freorder-blocks-and-partition
8974 In addition to reordering basic blocks in the compiled function, in order
8975 to reduce number of taken branches, partitions hot and cold basic blocks
8976 into separate sections of the assembly and .o files, to improve
8977 paging and cache locality performance.
8979 This optimization is automatically turned off in the presence of
8980 exception handling, for linkonce sections, for functions with a user-defined
8981 section attribute and on any architecture that does not support named
8984 Enabled for x86 at levels @option{-O2}, @option{-O3}.
8986 @item -freorder-functions
8987 @opindex freorder-functions
8988 Reorder functions in the object file in order to
8989 improve code locality. This is implemented by using special
8990 subsections @code{.text.hot} for most frequently executed functions and
8991 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
8992 the linker so object file format must support named sections and linker must
8993 place them in a reasonable way.
8995 Also profile feedback must be available to make this option effective. See
8996 @option{-fprofile-arcs} for details.
8998 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9000 @item -fstrict-aliasing
9001 @opindex fstrict-aliasing
9002 Allow the compiler to assume the strictest aliasing rules applicable to
9003 the language being compiled. For C (and C++), this activates
9004 optimizations based on the type of expressions. In particular, an
9005 object of one type is assumed never to reside at the same address as an
9006 object of a different type, unless the types are almost the same. For
9007 example, an @code{unsigned int} can alias an @code{int}, but not a
9008 @code{void*} or a @code{double}. A character type may alias any other
9011 @anchor{Type-punning}Pay special attention to code like this:
9024 The practice of reading from a different union member than the one most
9025 recently written to (called ``type-punning'') is common. Even with
9026 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
9027 is accessed through the union type. So, the code above works as
9028 expected. @xref{Structures unions enumerations and bit-fields
9029 implementation}. However, this code might not:
9040 Similarly, access by taking the address, casting the resulting pointer
9041 and dereferencing the result has undefined behavior, even if the cast
9042 uses a union type, e.g.:
9046 return ((union a_union *) &d)->i;
9050 The @option{-fstrict-aliasing} option is enabled at levels
9051 @option{-O2}, @option{-O3}, @option{-Os}.
9053 @item -fstrict-overflow
9054 @opindex fstrict-overflow
9055 Allow the compiler to assume strict signed overflow rules, depending
9056 on the language being compiled. For C (and C++) this means that
9057 overflow when doing arithmetic with signed numbers is undefined, which
9058 means that the compiler may assume that it does not happen. This
9059 permits various optimizations. For example, the compiler assumes
9060 that an expression like @code{i + 10 > i} is always true for
9061 signed @code{i}. This assumption is only valid if signed overflow is
9062 undefined, as the expression is false if @code{i + 10} overflows when
9063 using twos complement arithmetic. When this option is in effect any
9064 attempt to determine whether an operation on signed numbers
9065 overflows must be written carefully to not actually involve overflow.
9067 This option also allows the compiler to assume strict pointer
9068 semantics: given a pointer to an object, if adding an offset to that
9069 pointer does not produce a pointer to the same object, the addition is
9070 undefined. This permits the compiler to conclude that @code{p + u >
9071 p} is always true for a pointer @code{p} and unsigned integer
9072 @code{u}. This assumption is only valid because pointer wraparound is
9073 undefined, as the expression is false if @code{p + u} overflows using
9074 twos complement arithmetic.
9076 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
9077 that integer signed overflow is fully defined: it wraps. When
9078 @option{-fwrapv} is used, there is no difference between
9079 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
9080 integers. With @option{-fwrapv} certain types of overflow are
9081 permitted. For example, if the compiler gets an overflow when doing
9082 arithmetic on constants, the overflowed value can still be used with
9083 @option{-fwrapv}, but not otherwise.
9085 The @option{-fstrict-overflow} option is enabled at levels
9086 @option{-O2}, @option{-O3}, @option{-Os}.
9088 @item -falign-functions
9089 @itemx -falign-functions=@var{n}
9090 @opindex falign-functions
9091 Align the start of functions to the next power-of-two greater than
9092 @var{n}, skipping up to @var{n} bytes. For instance,
9093 @option{-falign-functions=32} aligns functions to the next 32-byte
9094 boundary, but @option{-falign-functions=24} aligns to the next
9095 32-byte boundary only if this can be done by skipping 23 bytes or less.
9097 @option{-fno-align-functions} and @option{-falign-functions=1} are
9098 equivalent and mean that functions are not aligned.
9100 Some assemblers only support this flag when @var{n} is a power of two;
9101 in that case, it is rounded up.
9103 If @var{n} is not specified or is zero, use a machine-dependent default.
9105 Enabled at levels @option{-O2}, @option{-O3}.
9107 @item -falign-labels
9108 @itemx -falign-labels=@var{n}
9109 @opindex falign-labels
9110 Align all branch targets to a power-of-two boundary, skipping up to
9111 @var{n} bytes like @option{-falign-functions}. This option can easily
9112 make code slower, because it must insert dummy operations for when the
9113 branch target is reached in the usual flow of the code.
9115 @option{-fno-align-labels} and @option{-falign-labels=1} are
9116 equivalent and mean that labels are not aligned.
9118 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
9119 are greater than this value, then their values are used instead.
9121 If @var{n} is not specified or is zero, use a machine-dependent default
9122 which is very likely to be @samp{1}, meaning no alignment.
9124 Enabled at levels @option{-O2}, @option{-O3}.
9127 @itemx -falign-loops=@var{n}
9128 @opindex falign-loops
9129 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
9130 like @option{-falign-functions}. If the loops are
9131 executed many times, this makes up for any execution of the dummy
9134 @option{-fno-align-loops} and @option{-falign-loops=1} are
9135 equivalent and mean that loops are not aligned.
9137 If @var{n} is not specified or is zero, use a machine-dependent default.
9139 Enabled at levels @option{-O2}, @option{-O3}.
9142 @itemx -falign-jumps=@var{n}
9143 @opindex falign-jumps
9144 Align branch targets to a power-of-two boundary, for branch targets
9145 where the targets can only be reached by jumping, skipping up to @var{n}
9146 bytes like @option{-falign-functions}. In this case, no dummy operations
9149 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
9150 equivalent and mean that loops are not aligned.
9152 If @var{n} is not specified or is zero, use a machine-dependent default.
9154 Enabled at levels @option{-O2}, @option{-O3}.
9156 @item -funit-at-a-time
9157 @opindex funit-at-a-time
9158 This option is left for compatibility reasons. @option{-funit-at-a-time}
9159 has no effect, while @option{-fno-unit-at-a-time} implies
9160 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
9164 @item -fno-toplevel-reorder
9165 @opindex fno-toplevel-reorder
9166 Do not reorder top-level functions, variables, and @code{asm}
9167 statements. Output them in the same order that they appear in the
9168 input file. When this option is used, unreferenced static variables
9169 are not removed. This option is intended to support existing code
9170 that relies on a particular ordering. For new code, it is better to
9171 use attributes when possible.
9173 Enabled at level @option{-O0}. When disabled explicitly, it also implies
9174 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
9179 Constructs webs as commonly used for register allocation purposes and assign
9180 each web individual pseudo register. This allows the register allocation pass
9181 to operate on pseudos directly, but also strengthens several other optimization
9182 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
9183 however, make debugging impossible, since variables no longer stay in a
9186 Enabled by default with @option{-funroll-loops}.
9188 @item -fwhole-program
9189 @opindex fwhole-program
9190 Assume that the current compilation unit represents the whole program being
9191 compiled. All public functions and variables with the exception of @code{main}
9192 and those merged by attribute @code{externally_visible} become static functions
9193 and in effect are optimized more aggressively by interprocedural optimizers.
9195 This option should not be used in combination with @option{-flto}.
9196 Instead relying on a linker plugin should provide safer and more precise
9199 @item -flto[=@var{n}]
9201 This option runs the standard link-time optimizer. When invoked
9202 with source code, it generates GIMPLE (one of GCC's internal
9203 representations) and writes it to special ELF sections in the object
9204 file. When the object files are linked together, all the function
9205 bodies are read from these ELF sections and instantiated as if they
9206 had been part of the same translation unit.
9208 To use the link-time optimizer, @option{-flto} and optimization
9209 options should be specified at compile time and during the final link.
9213 gcc -c -O2 -flto foo.c
9214 gcc -c -O2 -flto bar.c
9215 gcc -o myprog -flto -O2 foo.o bar.o
9218 The first two invocations to GCC save a bytecode representation
9219 of GIMPLE into special ELF sections inside @file{foo.o} and
9220 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
9221 @file{foo.o} and @file{bar.o}, merges the two files into a single
9222 internal image, and compiles the result as usual. Since both
9223 @file{foo.o} and @file{bar.o} are merged into a single image, this
9224 causes all the interprocedural analyses and optimizations in GCC to
9225 work across the two files as if they were a single one. This means,
9226 for example, that the inliner is able to inline functions in
9227 @file{bar.o} into functions in @file{foo.o} and vice-versa.
9229 Another (simpler) way to enable link-time optimization is:
9232 gcc -o myprog -flto -O2 foo.c bar.c
9235 The above generates bytecode for @file{foo.c} and @file{bar.c},
9236 merges them together into a single GIMPLE representation and optimizes
9237 them as usual to produce @file{myprog}.
9239 The only important thing to keep in mind is that to enable link-time
9240 optimizations you need to use the GCC driver to perform the link-step.
9241 GCC then automatically performs link-time optimization if any of the
9242 objects involved were compiled with the @option{-flto}. You generally
9243 should specify the optimization options to be used for link-time
9244 optimization though GCC tries to be clever at guessing an
9245 optimization level to use from the options used at compile-time
9246 if you fail to specify one at link-time. You can always override
9247 the automatic decision to do link-time optimization at link-time
9248 by passing @option{-fno-lto} to the link command.
9250 To make whole program optimization effective, it is necessary to make
9251 certain whole program assumptions. The compiler needs to know
9252 what functions and variables can be accessed by libraries and runtime
9253 outside of the link-time optimized unit. When supported by the linker,
9254 the linker plugin (see @option{-fuse-linker-plugin}) passes information
9255 to the compiler about used and externally visible symbols. When
9256 the linker plugin is not available, @option{-fwhole-program} should be
9257 used to allow the compiler to make these assumptions, which leads
9258 to more aggressive optimization decisions.
9260 When @option{-fuse-linker-plugin} is not enabled then, when a file is
9261 compiled with @option{-flto}, the generated object file is larger than
9262 a regular object file because it contains GIMPLE bytecodes and the usual
9263 final code (see @option{-ffat-lto-objects}. This means that
9264 object files with LTO information can be linked as normal object
9265 files; if @option{-fno-lto} is passed to the linker, no
9266 interprocedural optimizations are applied. Note that when
9267 @option{-fno-fat-lto-objects} is enabled the compile-stage is faster
9268 but you cannot perform a regular, non-LTO link on them.
9270 Additionally, the optimization flags used to compile individual files
9271 are not necessarily related to those used at link time. For instance,
9274 gcc -c -O0 -ffat-lto-objects -flto foo.c
9275 gcc -c -O0 -ffat-lto-objects -flto bar.c
9276 gcc -o myprog -O3 foo.o bar.o
9279 This produces individual object files with unoptimized assembler
9280 code, but the resulting binary @file{myprog} is optimized at
9281 @option{-O3}. If, instead, the final binary is generated with
9282 @option{-fno-lto}, then @file{myprog} is not optimized.
9284 When producing the final binary, GCC only
9285 applies link-time optimizations to those files that contain bytecode.
9286 Therefore, you can mix and match object files and libraries with
9287 GIMPLE bytecodes and final object code. GCC automatically selects
9288 which files to optimize in LTO mode and which files to link without
9291 There are some code generation flags preserved by GCC when
9292 generating bytecodes, as they need to be used during the final link
9293 stage. Generally options specified at link-time override those
9294 specified at compile-time.
9296 If you do not specify an optimization level option @option{-O} at
9297 link-time then GCC computes one based on the optimization levels
9298 used when compiling the object files. The highest optimization
9301 Currently, the following options and their setting are take from
9302 the first object file that explicitely specified it:
9303 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
9304 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
9305 and all the @option{-m} target flags.
9307 Certain ABI changing flags are required to match in all compilation-units
9308 and trying to override this at link-time with a conflicting value
9309 is ignored. This includes options such as @option{-freg-struct-return}
9310 and @option{-fpcc-struct-return}.
9312 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
9313 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
9314 are passed through to the link stage and merged conservatively for
9315 conflicting translation units. Specifically
9316 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
9317 precedence and for example @option{-ffp-contract=off} takes precedence
9318 over @option{-ffp-contract=fast}. You can override them at linke-time.
9320 It is recommended that you compile all the files participating in the
9321 same link with the same options and also specify those options at
9324 If LTO encounters objects with C linkage declared with incompatible
9325 types in separate translation units to be linked together (undefined
9326 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
9327 issued. The behavior is still undefined at run time. Similar
9328 diagnostics may be raised for other languages.
9330 Another feature of LTO is that it is possible to apply interprocedural
9331 optimizations on files written in different languages:
9336 gfortran -c -flto baz.f90
9337 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
9340 Notice that the final link is done with @command{g++} to get the C++
9341 runtime libraries and @option{-lgfortran} is added to get the Fortran
9342 runtime libraries. In general, when mixing languages in LTO mode, you
9343 should use the same link command options as when mixing languages in a
9344 regular (non-LTO) compilation.
9346 If object files containing GIMPLE bytecode are stored in a library archive, say
9347 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
9348 are using a linker with plugin support. To create static libraries suitable
9349 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
9350 and @command{ranlib};
9351 to show the symbols of object files with GIMPLE bytecode, use
9352 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
9353 and @command{nm} have been compiled with plugin support. At link time, use the the
9354 flag @option{-fuse-linker-plugin} to ensure that the library participates in
9355 the LTO optimization process:
9358 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
9361 With the linker plugin enabled, the linker extracts the needed
9362 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
9363 to make them part of the aggregated GIMPLE image to be optimized.
9365 If you are not using a linker with plugin support and/or do not
9366 enable the linker plugin, then the objects inside @file{libfoo.a}
9367 are extracted and linked as usual, but they do not participate
9368 in the LTO optimization process. In order to make a static library suitable
9369 for both LTO optimization and usual linkage, compile its object files with
9370 @option{-flto} @option{-ffat-lto-objects}.
9372 Link-time optimizations do not require the presence of the whole program to
9373 operate. If the program does not require any symbols to be exported, it is
9374 possible to combine @option{-flto} and @option{-fwhole-program} to allow
9375 the interprocedural optimizers to use more aggressive assumptions which may
9376 lead to improved optimization opportunities.
9377 Use of @option{-fwhole-program} is not needed when linker plugin is
9378 active (see @option{-fuse-linker-plugin}).
9380 The current implementation of LTO makes no
9381 attempt to generate bytecode that is portable between different
9382 types of hosts. The bytecode files are versioned and there is a
9383 strict version check, so bytecode files generated in one version of
9384 GCC do not work with an older or newer version of GCC.
9386 Link-time optimization does not work well with generation of debugging
9387 information. Combining @option{-flto} with
9388 @option{-g} is currently experimental and expected to produce unexpected
9391 If you specify the optional @var{n}, the optimization and code
9392 generation done at link time is executed in parallel using @var{n}
9393 parallel jobs by utilizing an installed @command{make} program. The
9394 environment variable @env{MAKE} may be used to override the program
9395 used. The default value for @var{n} is 1.
9397 You can also specify @option{-flto=jobserver} to use GNU make's
9398 job server mode to determine the number of parallel jobs. This
9399 is useful when the Makefile calling GCC is already executing in parallel.
9400 You must prepend a @samp{+} to the command recipe in the parent Makefile
9401 for this to work. This option likely only works if @env{MAKE} is
9404 @item -flto-partition=@var{alg}
9405 @opindex flto-partition
9406 Specify the partitioning algorithm used by the link-time optimizer.
9407 The value is either @samp{1to1} to specify a partitioning mirroring
9408 the original source files or @samp{balanced} to specify partitioning
9409 into equally sized chunks (whenever possible) or @samp{max} to create
9410 new partition for every symbol where possible. Specifying @samp{none}
9411 as an algorithm disables partitioning and streaming completely.
9412 The default value is @samp{balanced}. While @samp{1to1} can be used
9413 as an workaround for various code ordering issues, the @samp{max}
9414 partitioning is intended for internal testing only.
9415 The value @samp{one} specifies that exactly one partition should be
9416 used while the value @samp{none} bypasses partitioning and executes
9417 the link-time optimization step directly from the WPA phase.
9419 @item -flto-odr-type-merging
9420 @opindex flto-odr-type-merging
9421 Enable streaming of mangled types names of C++ types and their unification
9422 at linktime. This increases size of LTO object files, but enable
9423 diagnostics about One Definition Rule violations.
9425 @item -flto-compression-level=@var{n}
9426 @opindex flto-compression-level
9427 This option specifies the level of compression used for intermediate
9428 language written to LTO object files, and is only meaningful in
9429 conjunction with LTO mode (@option{-flto}). Valid
9430 values are 0 (no compression) to 9 (maximum compression). Values
9431 outside this range are clamped to either 0 or 9. If the option is not
9432 given, a default balanced compression setting is used.
9435 @opindex flto-report
9436 Prints a report with internal details on the workings of the link-time
9437 optimizer. The contents of this report vary from version to version.
9438 It is meant to be useful to GCC developers when processing object
9439 files in LTO mode (via @option{-flto}).
9441 Disabled by default.
9443 @item -flto-report-wpa
9444 @opindex flto-report-wpa
9445 Like @option{-flto-report}, but only print for the WPA phase of Link
9448 @item -fuse-linker-plugin
9449 @opindex fuse-linker-plugin
9450 Enables the use of a linker plugin during link-time optimization. This
9451 option relies on plugin support in the linker, which is available in gold
9452 or in GNU ld 2.21 or newer.
9454 This option enables the extraction of object files with GIMPLE bytecode out
9455 of library archives. This improves the quality of optimization by exposing
9456 more code to the link-time optimizer. This information specifies what
9457 symbols can be accessed externally (by non-LTO object or during dynamic
9458 linking). Resulting code quality improvements on binaries (and shared
9459 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
9460 See @option{-flto} for a description of the effect of this flag and how to
9463 This option is enabled by default when LTO support in GCC is enabled
9464 and GCC was configured for use with
9465 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9467 @item -ffat-lto-objects
9468 @opindex ffat-lto-objects
9469 Fat LTO objects are object files that contain both the intermediate language
9470 and the object code. This makes them usable for both LTO linking and normal
9471 linking. This option is effective only when compiling with @option{-flto}
9472 and is ignored at link time.
9474 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9475 requires the complete toolchain to be aware of LTO. It requires a linker with
9476 linker plugin support for basic functionality. Additionally,
9477 @command{nm}, @command{ar} and @command{ranlib}
9478 need to support linker plugins to allow a full-featured build environment
9479 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9480 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9481 to these tools. With non fat LTO makefiles need to be modified to use them.
9483 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9486 @item -fcompare-elim
9487 @opindex fcompare-elim
9488 After register allocation and post-register allocation instruction splitting,
9489 identify arithmetic instructions that compute processor flags similar to a
9490 comparison operation based on that arithmetic. If possible, eliminate the
9491 explicit comparison operation.
9493 This pass only applies to certain targets that cannot explicitly represent
9494 the comparison operation before register allocation is complete.
9496 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9498 @item -fcprop-registers
9499 @opindex fcprop-registers
9500 After register allocation and post-register allocation instruction splitting,
9501 perform a copy-propagation pass to try to reduce scheduling dependencies
9502 and occasionally eliminate the copy.
9504 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9506 @item -fprofile-correction
9507 @opindex fprofile-correction
9508 Profiles collected using an instrumented binary for multi-threaded programs may
9509 be inconsistent due to missed counter updates. When this option is specified,
9510 GCC uses heuristics to correct or smooth out such inconsistencies. By
9511 default, GCC emits an error message when an inconsistent profile is detected.
9513 @item -fprofile-dir=@var{path}
9514 @opindex fprofile-dir
9516 Set the directory to search for the profile data files in to @var{path}.
9517 This option affects only the profile data generated by
9518 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
9519 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
9520 and its related options. Both absolute and relative paths can be used.
9521 By default, GCC uses the current directory as @var{path}, thus the
9522 profile data file appears in the same directory as the object file.
9524 @item -fprofile-generate
9525 @itemx -fprofile-generate=@var{path}
9526 @opindex fprofile-generate
9528 Enable options usually used for instrumenting application to produce
9529 profile useful for later recompilation with profile feedback based
9530 optimization. You must use @option{-fprofile-generate} both when
9531 compiling and when linking your program.
9533 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
9535 If @var{path} is specified, GCC looks at the @var{path} to find
9536 the profile feedback data files. See @option{-fprofile-dir}.
9539 @itemx -fprofile-use=@var{path}
9540 @opindex fprofile-use
9541 Enable profile feedback-directed optimizations,
9542 and the following optimizations
9543 which are generally profitable only with profile feedback available:
9544 @option{-fbranch-probabilities}, @option{-fvpt},
9545 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9546 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
9548 By default, GCC emits an error message if the feedback profiles do not
9549 match the source code. This error can be turned into a warning by using
9550 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9553 If @var{path} is specified, GCC looks at the @var{path} to find
9554 the profile feedback data files. See @option{-fprofile-dir}.
9556 @item -fauto-profile
9557 @itemx -fauto-profile=@var{path}
9558 @opindex fauto-profile
9559 Enable sampling-based feedback-directed optimizations,
9560 and the following optimizations
9561 which are generally profitable only with profile feedback available:
9562 @option{-fbranch-probabilities}, @option{-fvpt},
9563 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9564 @option{-ftree-vectorize},
9565 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
9566 @option{-fpredictive-commoning}, @option{-funswitch-loops},
9567 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
9569 @var{path} is the name of a file containing AutoFDO profile information.
9570 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
9572 Producing an AutoFDO profile data file requires running your program
9573 with the @command{perf} utility on a supported GNU/Linux target system.
9574 For more information, see @uref{https://perf.wiki.kernel.org/}.
9578 perf record -e br_inst_retired:near_taken -b -o perf.data \
9582 Then use the @command{create_gcov} tool to convert the raw profile data
9583 to a format that can be used by GCC.@ You must also supply the
9584 unstripped binary for your program to this tool.
9585 See @uref{https://github.com/google/autofdo}.
9589 create_gcov --binary=your_program.unstripped --profile=perf.data \
9594 The following options control compiler behavior regarding floating-point
9595 arithmetic. These options trade off between speed and
9596 correctness. All must be specifically enabled.
9600 @opindex ffloat-store
9601 Do not store floating-point variables in registers, and inhibit other
9602 options that might change whether a floating-point value is taken from a
9605 @cindex floating-point precision
9606 This option prevents undesirable excess precision on machines such as
9607 the 68000 where the floating registers (of the 68881) keep more
9608 precision than a @code{double} is supposed to have. Similarly for the
9609 x86 architecture. For most programs, the excess precision does only
9610 good, but a few programs rely on the precise definition of IEEE floating
9611 point. Use @option{-ffloat-store} for such programs, after modifying
9612 them to store all pertinent intermediate computations into variables.
9614 @item -fexcess-precision=@var{style}
9615 @opindex fexcess-precision
9616 This option allows further control over excess precision on machines
9617 where floating-point registers have more precision than the IEEE
9618 @code{float} and @code{double} types and the processor does not
9619 support operations rounding to those types. By default,
9620 @option{-fexcess-precision=fast} is in effect; this means that
9621 operations are carried out in the precision of the registers and that
9622 it is unpredictable when rounding to the types specified in the source
9623 code takes place. When compiling C, if
9624 @option{-fexcess-precision=standard} is specified then excess
9625 precision follows the rules specified in ISO C99; in particular,
9626 both casts and assignments cause values to be rounded to their
9627 semantic types (whereas @option{-ffloat-store} only affects
9628 assignments). This option is enabled by default for C if a strict
9629 conformance option such as @option{-std=c99} is used.
9632 @option{-fexcess-precision=standard} is not implemented for languages
9633 other than C, and has no effect if
9634 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
9635 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
9636 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9637 semantics apply without excess precision, and in the latter, rounding
9642 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9643 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9644 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
9646 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9648 This option is not turned on by any @option{-O} option besides
9649 @option{-Ofast} since it can result in incorrect output for programs
9650 that depend on an exact implementation of IEEE or ISO rules/specifications
9651 for math functions. It may, however, yield faster code for programs
9652 that do not require the guarantees of these specifications.
9654 @item -fno-math-errno
9655 @opindex fno-math-errno
9656 Do not set @code{errno} after calling math functions that are executed
9657 with a single instruction, e.g., @code{sqrt}. A program that relies on
9658 IEEE exceptions for math error handling may want to use this flag
9659 for speed while maintaining IEEE arithmetic compatibility.
9661 This option is not turned on by any @option{-O} option since
9662 it can result in incorrect output for programs that depend on
9663 an exact implementation of IEEE or ISO rules/specifications for
9664 math functions. It may, however, yield faster code for programs
9665 that do not require the guarantees of these specifications.
9667 The default is @option{-fmath-errno}.
9669 On Darwin systems, the math library never sets @code{errno}. There is
9670 therefore no reason for the compiler to consider the possibility that
9671 it might, and @option{-fno-math-errno} is the default.
9673 @item -funsafe-math-optimizations
9674 @opindex funsafe-math-optimizations
9676 Allow optimizations for floating-point arithmetic that (a) assume
9677 that arguments and results are valid and (b) may violate IEEE or
9678 ANSI standards. When used at link-time, it may include libraries
9679 or startup files that change the default FPU control word or other
9680 similar optimizations.
9682 This option is not turned on by any @option{-O} option since
9683 it can result in incorrect output for programs that depend on
9684 an exact implementation of IEEE or ISO rules/specifications for
9685 math functions. It may, however, yield faster code for programs
9686 that do not require the guarantees of these specifications.
9687 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9688 @option{-fassociative-math} and @option{-freciprocal-math}.
9690 The default is @option{-fno-unsafe-math-optimizations}.
9692 @item -fassociative-math
9693 @opindex fassociative-math
9695 Allow re-association of operands in series of floating-point operations.
9696 This violates the ISO C and C++ language standard by possibly changing
9697 computation result. NOTE: re-ordering may change the sign of zero as
9698 well as ignore NaNs and inhibit or create underflow or overflow (and
9699 thus cannot be used on code that relies on rounding behavior like
9700 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9701 and thus may not be used when ordered comparisons are required.
9702 This option requires that both @option{-fno-signed-zeros} and
9703 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9704 much sense with @option{-frounding-math}. For Fortran the option
9705 is automatically enabled when both @option{-fno-signed-zeros} and
9706 @option{-fno-trapping-math} are in effect.
9708 The default is @option{-fno-associative-math}.
9710 @item -freciprocal-math
9711 @opindex freciprocal-math
9713 Allow the reciprocal of a value to be used instead of dividing by
9714 the value if this enables optimizations. For example @code{x / y}
9715 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9716 is subject to common subexpression elimination. Note that this loses
9717 precision and increases the number of flops operating on the value.
9719 The default is @option{-fno-reciprocal-math}.
9721 @item -ffinite-math-only
9722 @opindex ffinite-math-only
9723 Allow optimizations for floating-point arithmetic that assume
9724 that arguments and results are not NaNs or +-Infs.
9726 This option is not turned on by any @option{-O} option since
9727 it can result in incorrect output for programs that depend on
9728 an exact implementation of IEEE or ISO rules/specifications for
9729 math functions. It may, however, yield faster code for programs
9730 that do not require the guarantees of these specifications.
9732 The default is @option{-fno-finite-math-only}.
9734 @item -fno-signed-zeros
9735 @opindex fno-signed-zeros
9736 Allow optimizations for floating-point arithmetic that ignore the
9737 signedness of zero. IEEE arithmetic specifies the behavior of
9738 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9739 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9740 This option implies that the sign of a zero result isn't significant.
9742 The default is @option{-fsigned-zeros}.
9744 @item -fno-trapping-math
9745 @opindex fno-trapping-math
9746 Compile code assuming that floating-point operations cannot generate
9747 user-visible traps. These traps include division by zero, overflow,
9748 underflow, inexact result and invalid operation. This option requires
9749 that @option{-fno-signaling-nans} be in effect. Setting this option may
9750 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9752 This option should never be turned on by any @option{-O} option since
9753 it can result in incorrect output for programs that depend on
9754 an exact implementation of IEEE or ISO rules/specifications for
9757 The default is @option{-ftrapping-math}.
9759 @item -frounding-math
9760 @opindex frounding-math
9761 Disable transformations and optimizations that assume default floating-point
9762 rounding behavior. This is round-to-zero for all floating point
9763 to integer conversions, and round-to-nearest for all other arithmetic
9764 truncations. This option should be specified for programs that change
9765 the FP rounding mode dynamically, or that may be executed with a
9766 non-default rounding mode. This option disables constant folding of
9767 floating-point expressions at compile time (which may be affected by
9768 rounding mode) and arithmetic transformations that are unsafe in the
9769 presence of sign-dependent rounding modes.
9771 The default is @option{-fno-rounding-math}.
9773 This option is experimental and does not currently guarantee to
9774 disable all GCC optimizations that are affected by rounding mode.
9775 Future versions of GCC may provide finer control of this setting
9776 using C99's @code{FENV_ACCESS} pragma. This command-line option
9777 will be used to specify the default state for @code{FENV_ACCESS}.
9779 @item -fsignaling-nans
9780 @opindex fsignaling-nans
9781 Compile code assuming that IEEE signaling NaNs may generate user-visible
9782 traps during floating-point operations. Setting this option disables
9783 optimizations that may change the number of exceptions visible with
9784 signaling NaNs. This option implies @option{-ftrapping-math}.
9786 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9789 The default is @option{-fno-signaling-nans}.
9791 This option is experimental and does not currently guarantee to
9792 disable all GCC optimizations that affect signaling NaN behavior.
9794 @item -fsingle-precision-constant
9795 @opindex fsingle-precision-constant
9796 Treat floating-point constants as single precision instead of
9797 implicitly converting them to double-precision constants.
9799 @item -fcx-limited-range
9800 @opindex fcx-limited-range
9801 When enabled, this option states that a range reduction step is not
9802 needed when performing complex division. Also, there is no checking
9803 whether the result of a complex multiplication or division is @code{NaN
9804 + I*NaN}, with an attempt to rescue the situation in that case. The
9805 default is @option{-fno-cx-limited-range}, but is enabled by
9806 @option{-ffast-math}.
9808 This option controls the default setting of the ISO C99
9809 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9812 @item -fcx-fortran-rules
9813 @opindex fcx-fortran-rules
9814 Complex multiplication and division follow Fortran rules. Range
9815 reduction is done as part of complex division, but there is no checking
9816 whether the result of a complex multiplication or division is @code{NaN
9817 + I*NaN}, with an attempt to rescue the situation in that case.
9819 The default is @option{-fno-cx-fortran-rules}.
9823 The following options control optimizations that may improve
9824 performance, but are not enabled by any @option{-O} options. This
9825 section includes experimental options that may produce broken code.
9828 @item -fbranch-probabilities
9829 @opindex fbranch-probabilities
9830 After running a program compiled with @option{-fprofile-arcs}
9831 (@pxref{Debugging Options,, Options for Debugging Your Program or
9832 @command{gcc}}), you can compile it a second time using
9833 @option{-fbranch-probabilities}, to improve optimizations based on
9834 the number of times each branch was taken. When a program
9835 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9836 counts to a file called @file{@var{sourcename}.gcda} for each source
9837 file. The information in this data file is very dependent on the
9838 structure of the generated code, so you must use the same source code
9839 and the same optimization options for both compilations.
9841 With @option{-fbranch-probabilities}, GCC puts a
9842 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9843 These can be used to improve optimization. Currently, they are only
9844 used in one place: in @file{reorg.c}, instead of guessing which path a
9845 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9846 exactly determine which path is taken more often.
9848 @item -fprofile-values
9849 @opindex fprofile-values
9850 If combined with @option{-fprofile-arcs}, it adds code so that some
9851 data about values of expressions in the program is gathered.
9853 With @option{-fbranch-probabilities}, it reads back the data gathered
9854 from profiling values of expressions for usage in optimizations.
9856 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9858 @item -fprofile-reorder-functions
9859 @opindex fprofile-reorder-functions
9860 Function reordering based on profile instrumentation collects
9861 first time of execution of a function and orders these functions
9864 Enabled with @option{-fprofile-use}.
9868 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9869 to add code to gather information about values of expressions.
9871 With @option{-fbranch-probabilities}, it reads back the data gathered
9872 and actually performs the optimizations based on them.
9873 Currently the optimizations include specialization of division operations
9874 using the knowledge about the value of the denominator.
9876 @item -frename-registers
9877 @opindex frename-registers
9878 Attempt to avoid false dependencies in scheduled code by making use
9879 of registers left over after register allocation. This optimization
9880 most benefits processors with lots of registers. Depending on the
9881 debug information format adopted by the target, however, it can
9882 make debugging impossible, since variables no longer stay in
9883 a ``home register''.
9885 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
9887 @item -fschedule-fusion
9888 @opindex fschedule-fusion
9889 Performs a target dependent pass over the instruction stream to schedule
9890 instructions of same type together because target machine can execute them
9891 more efficiently if they are adjacent to each other in the instruction flow.
9893 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9897 Perform tail duplication to enlarge superblock size. This transformation
9898 simplifies the control flow of the function allowing other optimizations to do
9901 Enabled with @option{-fprofile-use}.
9903 @item -funroll-loops
9904 @opindex funroll-loops
9905 Unroll loops whose number of iterations can be determined at compile time or
9906 upon entry to the loop. @option{-funroll-loops} implies
9907 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
9908 It also turns on complete loop peeling (i.e.@: complete removal of loops with
9909 a small constant number of iterations). This option makes code larger, and may
9910 or may not make it run faster.
9912 Enabled with @option{-fprofile-use}.
9914 @item -funroll-all-loops
9915 @opindex funroll-all-loops
9916 Unroll all loops, even if their number of iterations is uncertain when
9917 the loop is entered. This usually makes programs run more slowly.
9918 @option{-funroll-all-loops} implies the same options as
9919 @option{-funroll-loops}.
9922 @opindex fpeel-loops
9923 Peels loops for which there is enough information that they do not
9924 roll much (from profile feedback). It also turns on complete loop peeling
9925 (i.e.@: complete removal of loops with small constant number of iterations).
9927 Enabled with @option{-fprofile-use}.
9929 @item -fmove-loop-invariants
9930 @opindex fmove-loop-invariants
9931 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
9932 at level @option{-O1}
9934 @item -funswitch-loops
9935 @opindex funswitch-loops
9936 Move branches with loop invariant conditions out of the loop, with duplicates
9937 of the loop on both branches (modified according to result of the condition).
9939 @item -ffunction-sections
9940 @itemx -fdata-sections
9941 @opindex ffunction-sections
9942 @opindex fdata-sections
9943 Place each function or data item into its own section in the output
9944 file if the target supports arbitrary sections. The name of the
9945 function or the name of the data item determines the section's name
9948 Use these options on systems where the linker can perform optimizations
9949 to improve locality of reference in the instruction space. Most systems
9950 using the ELF object format and SPARC processors running Solaris 2 have
9951 linkers with such optimizations. AIX may have these optimizations in
9954 Only use these options when there are significant benefits from doing
9955 so. When you specify these options, the assembler and linker
9956 create larger object and executable files and are also slower.
9957 You cannot use @command{gprof} on all systems if you
9958 specify this option, and you may have problems with debugging if
9959 you specify both this option and @option{-g}.
9961 @item -fbranch-target-load-optimize
9962 @opindex fbranch-target-load-optimize
9963 Perform branch target register load optimization before prologue / epilogue
9965 The use of target registers can typically be exposed only during reload,
9966 thus hoisting loads out of loops and doing inter-block scheduling needs
9967 a separate optimization pass.
9969 @item -fbranch-target-load-optimize2
9970 @opindex fbranch-target-load-optimize2
9971 Perform branch target register load optimization after prologue / epilogue
9974 @item -fbtr-bb-exclusive
9975 @opindex fbtr-bb-exclusive
9976 When performing branch target register load optimization, don't reuse
9977 branch target registers within any basic block.
9979 @item -fstack-protector
9980 @opindex fstack-protector
9981 Emit extra code to check for buffer overflows, such as stack smashing
9982 attacks. This is done by adding a guard variable to functions with
9983 vulnerable objects. This includes functions that call @code{alloca}, and
9984 functions with buffers larger than 8 bytes. The guards are initialized
9985 when a function is entered and then checked when the function exits.
9986 If a guard check fails, an error message is printed and the program exits.
9988 @item -fstack-protector-all
9989 @opindex fstack-protector-all
9990 Like @option{-fstack-protector} except that all functions are protected.
9992 @item -fstack-protector-strong
9993 @opindex fstack-protector-strong
9994 Like @option{-fstack-protector} but includes additional functions to
9995 be protected --- those that have local array definitions, or have
9996 references to local frame addresses.
9998 @item -fstack-protector-explicit
9999 @opindex fstack-protector-explicit
10000 Like @option{-fstack-protector} but only protects those functions which
10001 have the @code{stack_protect} attribute
10004 @opindex fstdarg-opt
10005 Optimize the prologue of variadic argument functions with respect to usage of
10008 @item -fsection-anchors
10009 @opindex fsection-anchors
10010 Try to reduce the number of symbolic address calculations by using
10011 shared ``anchor'' symbols to address nearby objects. This transformation
10012 can help to reduce the number of GOT entries and GOT accesses on some
10015 For example, the implementation of the following function @code{foo}:
10018 static int a, b, c;
10019 int foo (void) @{ return a + b + c; @}
10023 usually calculates the addresses of all three variables, but if you
10024 compile it with @option{-fsection-anchors}, it accesses the variables
10025 from a common anchor point instead. The effect is similar to the
10026 following pseudocode (which isn't valid C):
10031 register int *xr = &x;
10032 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
10036 Not all targets support this option.
10038 @item --param @var{name}=@var{value}
10040 In some places, GCC uses various constants to control the amount of
10041 optimization that is done. For example, GCC does not inline functions
10042 that contain more than a certain number of instructions. You can
10043 control some of these constants on the command line using the
10044 @option{--param} option.
10046 The names of specific parameters, and the meaning of the values, are
10047 tied to the internals of the compiler, and are subject to change
10048 without notice in future releases.
10050 In each case, the @var{value} is an integer. The allowable choices for
10054 @item predictable-branch-outcome
10055 When branch is predicted to be taken with probability lower than this threshold
10056 (in percent), then it is considered well predictable. The default is 10.
10058 @item max-crossjump-edges
10059 The maximum number of incoming edges to consider for cross-jumping.
10060 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
10061 the number of edges incoming to each block. Increasing values mean
10062 more aggressive optimization, making the compilation time increase with
10063 probably small improvement in executable size.
10065 @item min-crossjump-insns
10066 The minimum number of instructions that must be matched at the end
10067 of two blocks before cross-jumping is performed on them. This
10068 value is ignored in the case where all instructions in the block being
10069 cross-jumped from are matched. The default value is 5.
10071 @item max-grow-copy-bb-insns
10072 The maximum code size expansion factor when copying basic blocks
10073 instead of jumping. The expansion is relative to a jump instruction.
10074 The default value is 8.
10076 @item max-goto-duplication-insns
10077 The maximum number of instructions to duplicate to a block that jumps
10078 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
10079 passes, GCC factors computed gotos early in the compilation process,
10080 and unfactors them as late as possible. Only computed jumps at the
10081 end of a basic blocks with no more than max-goto-duplication-insns are
10082 unfactored. The default value is 8.
10084 @item max-delay-slot-insn-search
10085 The maximum number of instructions to consider when looking for an
10086 instruction to fill a delay slot. If more than this arbitrary number of
10087 instructions are searched, the time savings from filling the delay slot
10088 are minimal, so stop searching. Increasing values mean more
10089 aggressive optimization, making the compilation time increase with probably
10090 small improvement in execution time.
10092 @item max-delay-slot-live-search
10093 When trying to fill delay slots, the maximum number of instructions to
10094 consider when searching for a block with valid live register
10095 information. Increasing this arbitrarily chosen value means more
10096 aggressive optimization, increasing the compilation time. This parameter
10097 should be removed when the delay slot code is rewritten to maintain the
10098 control-flow graph.
10100 @item max-gcse-memory
10101 The approximate maximum amount of memory that can be allocated in
10102 order to perform the global common subexpression elimination
10103 optimization. If more memory than specified is required, the
10104 optimization is not done.
10106 @item max-gcse-insertion-ratio
10107 If the ratio of expression insertions to deletions is larger than this value
10108 for any expression, then RTL PRE inserts or removes the expression and thus
10109 leaves partially redundant computations in the instruction stream. The default value is 20.
10111 @item max-pending-list-length
10112 The maximum number of pending dependencies scheduling allows
10113 before flushing the current state and starting over. Large functions
10114 with few branches or calls can create excessively large lists which
10115 needlessly consume memory and resources.
10117 @item max-modulo-backtrack-attempts
10118 The maximum number of backtrack attempts the scheduler should make
10119 when modulo scheduling a loop. Larger values can exponentially increase
10122 @item max-inline-insns-single
10123 Several parameters control the tree inliner used in GCC@.
10124 This number sets the maximum number of instructions (counted in GCC's
10125 internal representation) in a single function that the tree inliner
10126 considers for inlining. This only affects functions declared
10127 inline and methods implemented in a class declaration (C++).
10128 The default value is 400.
10130 @item max-inline-insns-auto
10131 When you use @option{-finline-functions} (included in @option{-O3}),
10132 a lot of functions that would otherwise not be considered for inlining
10133 by the compiler are investigated. To those functions, a different
10134 (more restrictive) limit compared to functions declared inline can
10136 The default value is 40.
10138 @item inline-min-speedup
10139 When estimated performance improvement of caller + callee runtime exceeds this
10140 threshold (in precent), the function can be inlined regardless the limit on
10141 @option{--param max-inline-insns-single} and @option{--param
10142 max-inline-insns-auto}.
10144 @item large-function-insns
10145 The limit specifying really large functions. For functions larger than this
10146 limit after inlining, inlining is constrained by
10147 @option{--param large-function-growth}. This parameter is useful primarily
10148 to avoid extreme compilation time caused by non-linear algorithms used by the
10150 The default value is 2700.
10152 @item large-function-growth
10153 Specifies maximal growth of large function caused by inlining in percents.
10154 The default value is 100 which limits large function growth to 2.0 times
10157 @item large-unit-insns
10158 The limit specifying large translation unit. Growth caused by inlining of
10159 units larger than this limit is limited by @option{--param inline-unit-growth}.
10160 For small units this might be too tight.
10161 For example, consider a unit consisting of function A
10162 that is inline and B that just calls A three times. If B is small relative to
10163 A, the growth of unit is 300\% and yet such inlining is very sane. For very
10164 large units consisting of small inlineable functions, however, the overall unit
10165 growth limit is needed to avoid exponential explosion of code size. Thus for
10166 smaller units, the size is increased to @option{--param large-unit-insns}
10167 before applying @option{--param inline-unit-growth}. The default is 10000.
10169 @item inline-unit-growth
10170 Specifies maximal overall growth of the compilation unit caused by inlining.
10171 The default value is 15 which limits unit growth to 1.15 times the original
10172 size. Cold functions (either marked cold via an attribute or by profile
10173 feedback) are not accounted into the unit size.
10175 @item ipcp-unit-growth
10176 Specifies maximal overall growth of the compilation unit caused by
10177 interprocedural constant propagation. The default value is 10 which limits
10178 unit growth to 1.1 times the original size.
10180 @item large-stack-frame
10181 The limit specifying large stack frames. While inlining the algorithm is trying
10182 to not grow past this limit too much. The default value is 256 bytes.
10184 @item large-stack-frame-growth
10185 Specifies maximal growth of large stack frames caused by inlining in percents.
10186 The default value is 1000 which limits large stack frame growth to 11 times
10189 @item max-inline-insns-recursive
10190 @itemx max-inline-insns-recursive-auto
10191 Specifies the maximum number of instructions an out-of-line copy of a
10192 self-recursive inline
10193 function can grow into by performing recursive inlining.
10195 @option{--param max-inline-insns-recursive} applies to functions
10197 For functions not declared inline, recursive inlining
10198 happens only when @option{-finline-functions} (included in @option{-O3}) is
10199 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
10200 default value is 450.
10202 @item max-inline-recursive-depth
10203 @itemx max-inline-recursive-depth-auto
10204 Specifies the maximum recursion depth used for recursive inlining.
10206 @option{--param max-inline-recursive-depth} applies to functions
10207 declared inline. For functions not declared inline, recursive inlining
10208 happens only when @option{-finline-functions} (included in @option{-O3}) is
10209 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
10210 default value is 8.
10212 @item min-inline-recursive-probability
10213 Recursive inlining is profitable only for function having deep recursion
10214 in average and can hurt for function having little recursion depth by
10215 increasing the prologue size or complexity of function body to other
10218 When profile feedback is available (see @option{-fprofile-generate}) the actual
10219 recursion depth can be guessed from probability that function recurses via a
10220 given call expression. This parameter limits inlining only to call expressions
10221 whose probability exceeds the given threshold (in percents).
10222 The default value is 10.
10224 @item early-inlining-insns
10225 Specify growth that the early inliner can make. In effect it increases
10226 the amount of inlining for code having a large abstraction penalty.
10227 The default value is 14.
10229 @item max-early-inliner-iterations
10230 Limit of iterations of the early inliner. This basically bounds
10231 the number of nested indirect calls the early inliner can resolve.
10232 Deeper chains are still handled by late inlining.
10234 @item comdat-sharing-probability
10235 Probability (in percent) that C++ inline function with comdat visibility
10236 are shared across multiple compilation units. The default value is 20.
10238 @item profile-func-internal-id
10239 A parameter to control whether to use function internal id in profile
10240 database lookup. If the value is 0, the compiler uses an id that
10241 is based on function assembler name and filename, which makes old profile
10242 data more tolerant to source changes such as function reordering etc.
10243 The default value is 0.
10245 @item min-vect-loop-bound
10246 The minimum number of iterations under which loops are not vectorized
10247 when @option{-ftree-vectorize} is used. The number of iterations after
10248 vectorization needs to be greater than the value specified by this option
10249 to allow vectorization. The default value is 0.
10251 @item gcse-cost-distance-ratio
10252 Scaling factor in calculation of maximum distance an expression
10253 can be moved by GCSE optimizations. This is currently supported only in the
10254 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
10255 is with simple expressions, i.e., the expressions that have cost
10256 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
10257 hoisting of simple expressions. The default value is 10.
10259 @item gcse-unrestricted-cost
10260 Cost, roughly measured as the cost of a single typical machine
10261 instruction, at which GCSE optimizations do not constrain
10262 the distance an expression can travel. This is currently
10263 supported only in the code hoisting pass. The lesser the cost,
10264 the more aggressive code hoisting is. Specifying 0
10265 allows all expressions to travel unrestricted distances.
10266 The default value is 3.
10268 @item max-hoist-depth
10269 The depth of search in the dominator tree for expressions to hoist.
10270 This is used to avoid quadratic behavior in hoisting algorithm.
10271 The value of 0 does not limit on the search, but may slow down compilation
10272 of huge functions. The default value is 30.
10274 @item max-tail-merge-comparisons
10275 The maximum amount of similar bbs to compare a bb with. This is used to
10276 avoid quadratic behavior in tree tail merging. The default value is 10.
10278 @item max-tail-merge-iterations
10279 The maximum amount of iterations of the pass over the function. This is used to
10280 limit compilation time in tree tail merging. The default value is 2.
10282 @item max-unrolled-insns
10283 The maximum number of instructions that a loop may have to be unrolled.
10284 If a loop is unrolled, this parameter also determines how many times
10285 the loop code is unrolled.
10287 @item max-average-unrolled-insns
10288 The maximum number of instructions biased by probabilities of their execution
10289 that a loop may have to be unrolled. If a loop is unrolled,
10290 this parameter also determines how many times the loop code is unrolled.
10292 @item max-unroll-times
10293 The maximum number of unrollings of a single loop.
10295 @item max-peeled-insns
10296 The maximum number of instructions that a loop may have to be peeled.
10297 If a loop is peeled, this parameter also determines how many times
10298 the loop code is peeled.
10300 @item max-peel-times
10301 The maximum number of peelings of a single loop.
10303 @item max-peel-branches
10304 The maximum number of branches on the hot path through the peeled sequence.
10306 @item max-completely-peeled-insns
10307 The maximum number of insns of a completely peeled loop.
10309 @item max-completely-peel-times
10310 The maximum number of iterations of a loop to be suitable for complete peeling.
10312 @item max-completely-peel-loop-nest-depth
10313 The maximum depth of a loop nest suitable for complete peeling.
10315 @item max-unswitch-insns
10316 The maximum number of insns of an unswitched loop.
10318 @item max-unswitch-level
10319 The maximum number of branches unswitched in a single loop.
10321 @item lim-expensive
10322 The minimum cost of an expensive expression in the loop invariant motion.
10324 @item iv-consider-all-candidates-bound
10325 Bound on number of candidates for induction variables, below which
10326 all candidates are considered for each use in induction variable
10327 optimizations. If there are more candidates than this,
10328 only the most relevant ones are considered to avoid quadratic time complexity.
10330 @item iv-max-considered-uses
10331 The induction variable optimizations give up on loops that contain more
10332 induction variable uses.
10334 @item iv-always-prune-cand-set-bound
10335 If the number of candidates in the set is smaller than this value,
10336 always try to remove unnecessary ivs from the set
10337 when adding a new one.
10339 @item scev-max-expr-size
10340 Bound on size of expressions used in the scalar evolutions analyzer.
10341 Large expressions slow the analyzer.
10343 @item scev-max-expr-complexity
10344 Bound on the complexity of the expressions in the scalar evolutions analyzer.
10345 Complex expressions slow the analyzer.
10347 @item omega-max-vars
10348 The maximum number of variables in an Omega constraint system.
10349 The default value is 128.
10351 @item omega-max-geqs
10352 The maximum number of inequalities in an Omega constraint system.
10353 The default value is 256.
10355 @item omega-max-eqs
10356 The maximum number of equalities in an Omega constraint system.
10357 The default value is 128.
10359 @item omega-max-wild-cards
10360 The maximum number of wildcard variables that the Omega solver is
10361 able to insert. The default value is 18.
10363 @item omega-hash-table-size
10364 The size of the hash table in the Omega solver. The default value is
10367 @item omega-max-keys
10368 The maximal number of keys used by the Omega solver. The default
10371 @item omega-eliminate-redundant-constraints
10372 When set to 1, use expensive methods to eliminate all redundant
10373 constraints. The default value is 0.
10375 @item vect-max-version-for-alignment-checks
10376 The maximum number of run-time checks that can be performed when
10377 doing loop versioning for alignment in the vectorizer.
10379 @item vect-max-version-for-alias-checks
10380 The maximum number of run-time checks that can be performed when
10381 doing loop versioning for alias in the vectorizer.
10383 @item vect-max-peeling-for-alignment
10384 The maximum number of loop peels to enhance access alignment
10385 for vectorizer. Value -1 means 'no limit'.
10387 @item max-iterations-to-track
10388 The maximum number of iterations of a loop the brute-force algorithm
10389 for analysis of the number of iterations of the loop tries to evaluate.
10391 @item hot-bb-count-ws-permille
10392 A basic block profile count is considered hot if it contributes to
10393 the given permillage (i.e. 0...1000) of the entire profiled execution.
10395 @item hot-bb-frequency-fraction
10396 Select fraction of the entry block frequency of executions of basic block in
10397 function given basic block needs to have to be considered hot.
10399 @item max-predicted-iterations
10400 The maximum number of loop iterations we predict statically. This is useful
10401 in cases where a function contains a single loop with known bound and
10402 another loop with unknown bound.
10403 The known number of iterations is predicted correctly, while
10404 the unknown number of iterations average to roughly 10. This means that the
10405 loop without bounds appears artificially cold relative to the other one.
10407 @item builtin-expect-probability
10408 Control the probability of the expression having the specified value. This
10409 parameter takes a percentage (i.e. 0 ... 100) as input.
10410 The default probability of 90 is obtained empirically.
10412 @item align-threshold
10414 Select fraction of the maximal frequency of executions of a basic block in
10415 a function to align the basic block.
10417 @item align-loop-iterations
10419 A loop expected to iterate at least the selected number of iterations is
10422 @item tracer-dynamic-coverage
10423 @itemx tracer-dynamic-coverage-feedback
10425 This value is used to limit superblock formation once the given percentage of
10426 executed instructions is covered. This limits unnecessary code size
10429 The @option{tracer-dynamic-coverage-feedback} is used only when profile
10430 feedback is available. The real profiles (as opposed to statically estimated
10431 ones) are much less balanced allowing the threshold to be larger value.
10433 @item tracer-max-code-growth
10434 Stop tail duplication once code growth has reached given percentage. This is
10435 a rather artificial limit, as most of the duplicates are eliminated later in
10436 cross jumping, so it may be set to much higher values than is the desired code
10439 @item tracer-min-branch-ratio
10441 Stop reverse growth when the reverse probability of best edge is less than this
10442 threshold (in percent).
10444 @item tracer-min-branch-ratio
10445 @itemx tracer-min-branch-ratio-feedback
10447 Stop forward growth if the best edge has probability lower than this
10450 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
10451 compilation for profile feedback and one for compilation without. The value
10452 for compilation with profile feedback needs to be more conservative (higher) in
10453 order to make tracer effective.
10455 @item max-cse-path-length
10457 The maximum number of basic blocks on path that CSE considers.
10460 @item max-cse-insns
10461 The maximum number of instructions CSE processes before flushing.
10462 The default is 1000.
10464 @item ggc-min-expand
10466 GCC uses a garbage collector to manage its own memory allocation. This
10467 parameter specifies the minimum percentage by which the garbage
10468 collector's heap should be allowed to expand between collections.
10469 Tuning this may improve compilation speed; it has no effect on code
10472 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10473 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10474 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10475 GCC is not able to calculate RAM on a particular platform, the lower
10476 bound of 30% is used. Setting this parameter and
10477 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10478 every opportunity. This is extremely slow, but can be useful for
10481 @item ggc-min-heapsize
10483 Minimum size of the garbage collector's heap before it begins bothering
10484 to collect garbage. The first collection occurs after the heap expands
10485 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10486 tuning this may improve compilation speed, and has no effect on code
10489 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10490 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10491 with a lower bound of 4096 (four megabytes) and an upper bound of
10492 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10493 particular platform, the lower bound is used. Setting this parameter
10494 very large effectively disables garbage collection. Setting this
10495 parameter and @option{ggc-min-expand} to zero causes a full collection
10496 to occur at every opportunity.
10498 @item max-reload-search-insns
10499 The maximum number of instruction reload should look backward for equivalent
10500 register. Increasing values mean more aggressive optimization, making the
10501 compilation time increase with probably slightly better performance.
10502 The default value is 100.
10504 @item max-cselib-memory-locations
10505 The maximum number of memory locations cselib should take into account.
10506 Increasing values mean more aggressive optimization, making the compilation time
10507 increase with probably slightly better performance. The default value is 500.
10509 @item reorder-blocks-duplicate
10510 @itemx reorder-blocks-duplicate-feedback
10512 Used by the basic block reordering pass to decide whether to use unconditional
10513 branch or duplicate the code on its destination. Code is duplicated when its
10514 estimated size is smaller than this value multiplied by the estimated size of
10515 unconditional jump in the hot spots of the program.
10517 The @option{reorder-block-duplicate-feedback} is used only when profile
10518 feedback is available. It may be set to higher values than
10519 @option{reorder-block-duplicate} since information about the hot spots is more
10522 @item max-sched-ready-insns
10523 The maximum number of instructions ready to be issued the scheduler should
10524 consider at any given time during the first scheduling pass. Increasing
10525 values mean more thorough searches, making the compilation time increase
10526 with probably little benefit. The default value is 100.
10528 @item max-sched-region-blocks
10529 The maximum number of blocks in a region to be considered for
10530 interblock scheduling. The default value is 10.
10532 @item max-pipeline-region-blocks
10533 The maximum number of blocks in a region to be considered for
10534 pipelining in the selective scheduler. The default value is 15.
10536 @item max-sched-region-insns
10537 The maximum number of insns in a region to be considered for
10538 interblock scheduling. The default value is 100.
10540 @item max-pipeline-region-insns
10541 The maximum number of insns in a region to be considered for
10542 pipelining in the selective scheduler. The default value is 200.
10544 @item min-spec-prob
10545 The minimum probability (in percents) of reaching a source block
10546 for interblock speculative scheduling. The default value is 40.
10548 @item max-sched-extend-regions-iters
10549 The maximum number of iterations through CFG to extend regions.
10550 A value of 0 (the default) disables region extensions.
10552 @item max-sched-insn-conflict-delay
10553 The maximum conflict delay for an insn to be considered for speculative motion.
10554 The default value is 3.
10556 @item sched-spec-prob-cutoff
10557 The minimal probability of speculation success (in percents), so that
10558 speculative insns are scheduled.
10559 The default value is 40.
10561 @item sched-spec-state-edge-prob-cutoff
10562 The minimum probability an edge must have for the scheduler to save its
10564 The default value is 10.
10566 @item sched-mem-true-dep-cost
10567 Minimal distance (in CPU cycles) between store and load targeting same
10568 memory locations. The default value is 1.
10570 @item selsched-max-lookahead
10571 The maximum size of the lookahead window of selective scheduling. It is a
10572 depth of search for available instructions.
10573 The default value is 50.
10575 @item selsched-max-sched-times
10576 The maximum number of times that an instruction is scheduled during
10577 selective scheduling. This is the limit on the number of iterations
10578 through which the instruction may be pipelined. The default value is 2.
10580 @item selsched-max-insns-to-rename
10581 The maximum number of best instructions in the ready list that are considered
10582 for renaming in the selective scheduler. The default value is 2.
10585 The minimum value of stage count that swing modulo scheduler
10586 generates. The default value is 2.
10588 @item max-last-value-rtl
10589 The maximum size measured as number of RTLs that can be recorded in an expression
10590 in combiner for a pseudo register as last known value of that register. The default
10593 @item max-combine-insns
10594 The maximum number of instructions the RTL combiner tries to combine.
10595 The default value is 2 at @option{-Og} and 4 otherwise.
10597 @item integer-share-limit
10598 Small integer constants can use a shared data structure, reducing the
10599 compiler's memory usage and increasing its speed. This sets the maximum
10600 value of a shared integer constant. The default value is 256.
10602 @item ssp-buffer-size
10603 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10604 protection when @option{-fstack-protection} is used.
10606 @item min-size-for-stack-sharing
10607 The minimum size of variables taking part in stack slot sharing when not
10608 optimizing. The default value is 32.
10610 @item max-jump-thread-duplication-stmts
10611 Maximum number of statements allowed in a block that needs to be
10612 duplicated when threading jumps.
10614 @item max-fields-for-field-sensitive
10615 Maximum number of fields in a structure treated in
10616 a field sensitive manner during pointer analysis. The default is zero
10617 for @option{-O0} and @option{-O1},
10618 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10620 @item prefetch-latency
10621 Estimate on average number of instructions that are executed before
10622 prefetch finishes. The distance prefetched ahead is proportional
10623 to this constant. Increasing this number may also lead to less
10624 streams being prefetched (see @option{simultaneous-prefetches}).
10626 @item simultaneous-prefetches
10627 Maximum number of prefetches that can run at the same time.
10629 @item l1-cache-line-size
10630 The size of cache line in L1 cache, in bytes.
10632 @item l1-cache-size
10633 The size of L1 cache, in kilobytes.
10635 @item l2-cache-size
10636 The size of L2 cache, in kilobytes.
10638 @item min-insn-to-prefetch-ratio
10639 The minimum ratio between the number of instructions and the
10640 number of prefetches to enable prefetching in a loop.
10642 @item prefetch-min-insn-to-mem-ratio
10643 The minimum ratio between the number of instructions and the
10644 number of memory references to enable prefetching in a loop.
10646 @item use-canonical-types
10647 Whether the compiler should use the ``canonical'' type system. By
10648 default, this should always be 1, which uses a more efficient internal
10649 mechanism for comparing types in C++ and Objective-C++. However, if
10650 bugs in the canonical type system are causing compilation failures,
10651 set this value to 0 to disable canonical types.
10653 @item switch-conversion-max-branch-ratio
10654 Switch initialization conversion refuses to create arrays that are
10655 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10656 branches in the switch.
10658 @item max-partial-antic-length
10659 Maximum length of the partial antic set computed during the tree
10660 partial redundancy elimination optimization (@option{-ftree-pre}) when
10661 optimizing at @option{-O3} and above. For some sorts of source code
10662 the enhanced partial redundancy elimination optimization can run away,
10663 consuming all of the memory available on the host machine. This
10664 parameter sets a limit on the length of the sets that are computed,
10665 which prevents the runaway behavior. Setting a value of 0 for
10666 this parameter allows an unlimited set length.
10668 @item sccvn-max-scc-size
10669 Maximum size of a strongly connected component (SCC) during SCCVN
10670 processing. If this limit is hit, SCCVN processing for the whole
10671 function is not done and optimizations depending on it are
10672 disabled. The default maximum SCC size is 10000.
10674 @item sccvn-max-alias-queries-per-access
10675 Maximum number of alias-oracle queries we perform when looking for
10676 redundancies for loads and stores. If this limit is hit the search
10677 is aborted and the load or store is not considered redundant. The
10678 number of queries is algorithmically limited to the number of
10679 stores on all paths from the load to the function entry.
10680 The default maxmimum number of queries is 1000.
10682 @item ira-max-loops-num
10683 IRA uses regional register allocation by default. If a function
10684 contains more loops than the number given by this parameter, only at most
10685 the given number of the most frequently-executed loops form regions
10686 for regional register allocation. The default value of the
10689 @item ira-max-conflict-table-size
10690 Although IRA uses a sophisticated algorithm to compress the conflict
10691 table, the table can still require excessive amounts of memory for
10692 huge functions. If the conflict table for a function could be more
10693 than the size in MB given by this parameter, the register allocator
10694 instead uses a faster, simpler, and lower-quality
10695 algorithm that does not require building a pseudo-register conflict table.
10696 The default value of the parameter is 2000.
10698 @item ira-loop-reserved-regs
10699 IRA can be used to evaluate more accurate register pressure in loops
10700 for decisions to move loop invariants (see @option{-O3}). The number
10701 of available registers reserved for some other purposes is given
10702 by this parameter. The default value of the parameter is 2, which is
10703 the minimal number of registers needed by typical instructions.
10704 This value is the best found from numerous experiments.
10706 @item lra-inheritance-ebb-probability-cutoff
10707 LRA tries to reuse values reloaded in registers in subsequent insns.
10708 This optimization is called inheritance. EBB is used as a region to
10709 do this optimization. The parameter defines a minimal fall-through
10710 edge probability in percentage used to add BB to inheritance EBB in
10711 LRA. The default value of the parameter is 40. The value was chosen
10712 from numerous runs of SPEC2000 on x86-64.
10714 @item loop-invariant-max-bbs-in-loop
10715 Loop invariant motion can be very expensive, both in compilation time and
10716 in amount of needed compile-time memory, with very large loops. Loops
10717 with more basic blocks than this parameter won't have loop invariant
10718 motion optimization performed on them. The default value of the
10719 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10721 @item loop-max-datarefs-for-datadeps
10722 Building data dapendencies is expensive for very large loops. This
10723 parameter limits the number of data references in loops that are
10724 considered for data dependence analysis. These large loops are no
10725 handled by the optimizations using loop data dependencies.
10726 The default value is 1000.
10728 @item max-vartrack-size
10729 Sets a maximum number of hash table slots to use during variable
10730 tracking dataflow analysis of any function. If this limit is exceeded
10731 with variable tracking at assignments enabled, analysis for that
10732 function is retried without it, after removing all debug insns from
10733 the function. If the limit is exceeded even without debug insns, var
10734 tracking analysis is completely disabled for the function. Setting
10735 the parameter to zero makes it unlimited.
10737 @item max-vartrack-expr-depth
10738 Sets a maximum number of recursion levels when attempting to map
10739 variable names or debug temporaries to value expressions. This trades
10740 compilation time for more complete debug information. If this is set too
10741 low, value expressions that are available and could be represented in
10742 debug information may end up not being used; setting this higher may
10743 enable the compiler to find more complex debug expressions, but compile
10744 time and memory use may grow. The default is 12.
10746 @item min-nondebug-insn-uid
10747 Use uids starting at this parameter for nondebug insns. The range below
10748 the parameter is reserved exclusively for debug insns created by
10749 @option{-fvar-tracking-assignments}, but debug insns may get
10750 (non-overlapping) uids above it if the reserved range is exhausted.
10752 @item ipa-sra-ptr-growth-factor
10753 IPA-SRA replaces a pointer to an aggregate with one or more new
10754 parameters only when their cumulative size is less or equal to
10755 @option{ipa-sra-ptr-growth-factor} times the size of the original
10758 @item sra-max-scalarization-size-Ospeed
10759 @item sra-max-scalarization-size-Osize
10760 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10761 replace scalar parts of aggregates with uses of independent scalar
10762 variables. These parameters control the maximum size, in storage units,
10763 of aggregate which is considered for replacement when compiling for
10765 (@option{sra-max-scalarization-size-Ospeed}) or size
10766 (@option{sra-max-scalarization-size-Osize}) respectively.
10768 @item tm-max-aggregate-size
10769 When making copies of thread-local variables in a transaction, this
10770 parameter specifies the size in bytes after which variables are
10771 saved with the logging functions as opposed to save/restore code
10772 sequence pairs. This option only applies when using
10775 @item graphite-max-nb-scop-params
10776 To avoid exponential effects in the Graphite loop transforms, the
10777 number of parameters in a Static Control Part (SCoP) is bounded. The
10778 default value is 10 parameters. A variable whose value is unknown at
10779 compilation time and defined outside a SCoP is a parameter of the SCoP.
10781 @item graphite-max-bbs-per-function
10782 To avoid exponential effects in the detection of SCoPs, the size of
10783 the functions analyzed by Graphite is bounded. The default value is
10786 @item loop-block-tile-size
10787 Loop blocking or strip mining transforms, enabled with
10788 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10789 loop in the loop nest by a given number of iterations. The strip
10790 length can be changed using the @option{loop-block-tile-size}
10791 parameter. The default value is 51 iterations.
10793 @item loop-unroll-jam-size
10794 Specify the unroll factor for the @option{-floop-unroll-and-jam}. The
10795 default value is 4.
10797 @item loop-unroll-jam-depth
10798 Specify the dimension to be unrolled (counting from the most inner loop)
10799 for the @option{-floop-unroll-and-jam}. The default value is 2.
10801 @item ipa-cp-value-list-size
10802 IPA-CP attempts to track all possible values and types passed to a function's
10803 parameter in order to propagate them and perform devirtualization.
10804 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10805 stores per one formal parameter of a function.
10807 @item ipa-cp-eval-threshold
10808 IPA-CP calculates its own score of cloning profitability heuristics
10809 and performs those cloning opportunities with scores that exceed
10810 @option{ipa-cp-eval-threshold}.
10812 @item ipa-max-agg-items
10813 IPA-CP is also capable to propagate a number of scalar values passed
10814 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10815 number of such values per one parameter.
10817 @item ipa-cp-loop-hint-bonus
10818 When IPA-CP determines that a cloning candidate would make the number
10819 of iterations of a loop known, it adds a bonus of
10820 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10823 @item ipa-cp-array-index-hint-bonus
10824 When IPA-CP determines that a cloning candidate would make the index of
10825 an array access known, it adds a bonus of
10826 @option{ipa-cp-array-index-hint-bonus} to the profitability
10827 score of the candidate.
10829 @item ipa-max-aa-steps
10830 During its analysis of function bodies, IPA-CP employs alias analysis
10831 in order to track values pointed to by function parameters. In order
10832 not spend too much time analyzing huge functions, it gives up and
10833 consider all memory clobbered after examining
10834 @option{ipa-max-aa-steps} statements modifying memory.
10836 @item lto-partitions
10837 Specify desired number of partitions produced during WHOPR compilation.
10838 The number of partitions should exceed the number of CPUs used for compilation.
10839 The default value is 32.
10841 @item lto-minpartition
10842 Size of minimal partition for WHOPR (in estimated instructions).
10843 This prevents expenses of splitting very small programs into too many
10846 @item cxx-max-namespaces-for-diagnostic-help
10847 The maximum number of namespaces to consult for suggestions when C++
10848 name lookup fails for an identifier. The default is 1000.
10850 @item sink-frequency-threshold
10851 The maximum relative execution frequency (in percents) of the target block
10852 relative to a statement's original block to allow statement sinking of a
10853 statement. Larger numbers result in more aggressive statement sinking.
10854 The default value is 75. A small positive adjustment is applied for
10855 statements with memory operands as those are even more profitable so sink.
10857 @item max-stores-to-sink
10858 The maximum number of conditional stores paires that can be sunk. Set to 0
10859 if either vectorization (@option{-ftree-vectorize}) or if-conversion
10860 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
10862 @item allow-store-data-races
10863 Allow optimizers to introduce new data races on stores.
10864 Set to 1 to allow, otherwise to 0. This option is enabled by default
10865 at optimization level @option{-Ofast}.
10867 @item case-values-threshold
10868 The smallest number of different values for which it is best to use a
10869 jump-table instead of a tree of conditional branches. If the value is
10870 0, use the default for the machine. The default is 0.
10872 @item tree-reassoc-width
10873 Set the maximum number of instructions executed in parallel in
10874 reassociated tree. This parameter overrides target dependent
10875 heuristics used by default if has non zero value.
10877 @item sched-pressure-algorithm
10878 Choose between the two available implementations of
10879 @option{-fsched-pressure}. Algorithm 1 is the original implementation
10880 and is the more likely to prevent instructions from being reordered.
10881 Algorithm 2 was designed to be a compromise between the relatively
10882 conservative approach taken by algorithm 1 and the rather aggressive
10883 approach taken by the default scheduler. It relies more heavily on
10884 having a regular register file and accurate register pressure classes.
10885 See @file{haifa-sched.c} in the GCC sources for more details.
10887 The default choice depends on the target.
10889 @item max-slsr-cand-scan
10890 Set the maximum number of existing candidates that are considered when
10891 seeking a basis for a new straight-line strength reduction candidate.
10894 Enable buffer overflow detection for global objects. This kind
10895 of protection is enabled by default if you are using
10896 @option{-fsanitize=address} option.
10897 To disable global objects protection use @option{--param asan-globals=0}.
10900 Enable buffer overflow detection for stack objects. This kind of
10901 protection is enabled by default when using@option{-fsanitize=address}.
10902 To disable stack protection use @option{--param asan-stack=0} option.
10904 @item asan-instrument-reads
10905 Enable buffer overflow detection for memory reads. This kind of
10906 protection is enabled by default when using @option{-fsanitize=address}.
10907 To disable memory reads protection use
10908 @option{--param asan-instrument-reads=0}.
10910 @item asan-instrument-writes
10911 Enable buffer overflow detection for memory writes. This kind of
10912 protection is enabled by default when using @option{-fsanitize=address}.
10913 To disable memory writes protection use
10914 @option{--param asan-instrument-writes=0} option.
10916 @item asan-memintrin
10917 Enable detection for built-in functions. This kind of protection
10918 is enabled by default when using @option{-fsanitize=address}.
10919 To disable built-in functions protection use
10920 @option{--param asan-memintrin=0}.
10922 @item asan-use-after-return
10923 Enable detection of use-after-return. This kind of protection
10924 is enabled by default when using @option{-fsanitize=address} option.
10925 To disable use-after-return detection use
10926 @option{--param asan-use-after-return=0}.
10928 @item asan-instrumentation-with-call-threshold
10929 If number of memory accesses in function being instrumented
10930 is greater or equal to this number, use callbacks instead of inline checks.
10931 E.g. to disable inline code use
10932 @option{--param asan-instrumentation-with-call-threshold=0}.
10934 @item chkp-max-ctor-size
10935 Static constructors generated by Pointer Bounds Checker may become very
10936 large and significantly increase compile time at optimization level
10937 @option{-O1} and higher. This parameter is a maximum nubmer of statements
10938 in a single generated constructor. Default value is 5000.
10940 @item max-fsm-thread-path-insns
10941 Maximum number of instructions to copy when duplicating blocks on a
10942 finite state automaton jump thread path. The default is 100.
10944 @item max-fsm-thread-length
10945 Maximum number of basic blocks on a finite state automaton jump thread
10946 path. The default is 10.
10948 @item max-fsm-thread-paths
10949 Maximum number of new jump thread paths to create for a finite state
10950 automaton. The default is 50.
10955 @node Preprocessor Options
10956 @section Options Controlling the Preprocessor
10957 @cindex preprocessor options
10958 @cindex options, preprocessor
10960 These options control the C preprocessor, which is run on each C source
10961 file before actual compilation.
10963 If you use the @option{-E} option, nothing is done except preprocessing.
10964 Some of these options make sense only together with @option{-E} because
10965 they cause the preprocessor output to be unsuitable for actual
10969 @item -Wp,@var{option}
10971 You can use @option{-Wp,@var{option}} to bypass the compiler driver
10972 and pass @var{option} directly through to the preprocessor. If
10973 @var{option} contains commas, it is split into multiple options at the
10974 commas. However, many options are modified, translated or interpreted
10975 by the compiler driver before being passed to the preprocessor, and
10976 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
10977 interface is undocumented and subject to change, so whenever possible
10978 you should avoid using @option{-Wp} and let the driver handle the
10981 @item -Xpreprocessor @var{option}
10982 @opindex Xpreprocessor
10983 Pass @var{option} as an option to the preprocessor. You can use this to
10984 supply system-specific preprocessor options that GCC does not
10987 If you want to pass an option that takes an argument, you must use
10988 @option{-Xpreprocessor} twice, once for the option and once for the argument.
10990 @item -no-integrated-cpp
10991 @opindex no-integrated-cpp
10992 Perform preprocessing as a separate pass before compilation.
10993 By default, GCC performs preprocessing as an integrated part of
10994 input tokenization and parsing.
10995 If this option is provided, the appropriate language front end
10996 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
10997 and Objective-C, respectively) is instead invoked twice,
10998 once for preprocessing only and once for actual compilation
10999 of the preprocessed input.
11000 This option may be useful in conjunction with the @option{-B} or
11001 @option{-wrapper} options to specify an alternate preprocessor or
11002 perform additional processing of the program source between
11003 normal preprocessing and compilation.
11006 @include cppopts.texi
11008 @node Assembler Options
11009 @section Passing Options to the Assembler
11011 @c prevent bad page break with this line
11012 You can pass options to the assembler.
11015 @item -Wa,@var{option}
11017 Pass @var{option} as an option to the assembler. If @var{option}
11018 contains commas, it is split into multiple options at the commas.
11020 @item -Xassembler @var{option}
11021 @opindex Xassembler
11022 Pass @var{option} as an option to the assembler. You can use this to
11023 supply system-specific assembler options that GCC does not
11026 If you want to pass an option that takes an argument, you must use
11027 @option{-Xassembler} twice, once for the option and once for the argument.
11032 @section Options for Linking
11033 @cindex link options
11034 @cindex options, linking
11036 These options come into play when the compiler links object files into
11037 an executable output file. They are meaningless if the compiler is
11038 not doing a link step.
11042 @item @var{object-file-name}
11043 A file name that does not end in a special recognized suffix is
11044 considered to name an object file or library. (Object files are
11045 distinguished from libraries by the linker according to the file
11046 contents.) If linking is done, these object files are used as input
11055 If any of these options is used, then the linker is not run, and
11056 object file names should not be used as arguments. @xref{Overall
11060 @opindex fuse-ld=bfd
11061 Use the @command{bfd} linker instead of the default linker.
11063 @item -fuse-ld=gold
11064 @opindex fuse-ld=gold
11065 Use the @command{gold} linker instead of the default linker.
11068 @item -l@var{library}
11069 @itemx -l @var{library}
11071 Search the library named @var{library} when linking. (The second
11072 alternative with the library as a separate argument is only for
11073 POSIX compliance and is not recommended.)
11075 It makes a difference where in the command you write this option; the
11076 linker searches and processes libraries and object files in the order they
11077 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
11078 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
11079 to functions in @samp{z}, those functions may not be loaded.
11081 The linker searches a standard list of directories for the library,
11082 which is actually a file named @file{lib@var{library}.a}. The linker
11083 then uses this file as if it had been specified precisely by name.
11085 The directories searched include several standard system directories
11086 plus any that you specify with @option{-L}.
11088 Normally the files found this way are library files---archive files
11089 whose members are object files. The linker handles an archive file by
11090 scanning through it for members which define symbols that have so far
11091 been referenced but not defined. But if the file that is found is an
11092 ordinary object file, it is linked in the usual fashion. The only
11093 difference between using an @option{-l} option and specifying a file name
11094 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
11095 and searches several directories.
11099 You need this special case of the @option{-l} option in order to
11100 link an Objective-C or Objective-C++ program.
11102 @item -nostartfiles
11103 @opindex nostartfiles
11104 Do not use the standard system startup files when linking.
11105 The standard system libraries are used normally, unless @option{-nostdlib}
11106 or @option{-nodefaultlibs} is used.
11108 @item -nodefaultlibs
11109 @opindex nodefaultlibs
11110 Do not use the standard system libraries when linking.
11111 Only the libraries you specify are passed to the linker, and options
11112 specifying linkage of the system libraries, such as @option{-static-libgcc}
11113 or @option{-shared-libgcc}, are ignored.
11114 The standard startup files are used normally, unless @option{-nostartfiles}
11117 The compiler may generate calls to @code{memcmp},
11118 @code{memset}, @code{memcpy} and @code{memmove}.
11119 These entries are usually resolved by entries in
11120 libc. These entry points should be supplied through some other
11121 mechanism when this option is specified.
11125 Do not use the standard system startup files or libraries when linking.
11126 No startup files and only the libraries you specify are passed to
11127 the linker, and options specifying linkage of the system libraries, such as
11128 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
11130 The compiler may generate calls to @code{memcmp}, @code{memset},
11131 @code{memcpy} and @code{memmove}.
11132 These entries are usually resolved by entries in
11133 libc. These entry points should be supplied through some other
11134 mechanism when this option is specified.
11136 @cindex @option{-lgcc}, use with @option{-nostdlib}
11137 @cindex @option{-nostdlib} and unresolved references
11138 @cindex unresolved references and @option{-nostdlib}
11139 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
11140 @cindex @option{-nodefaultlibs} and unresolved references
11141 @cindex unresolved references and @option{-nodefaultlibs}
11142 One of the standard libraries bypassed by @option{-nostdlib} and
11143 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
11144 which GCC uses to overcome shortcomings of particular machines, or special
11145 needs for some languages.
11146 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
11147 Collection (GCC) Internals},
11148 for more discussion of @file{libgcc.a}.)
11149 In most cases, you need @file{libgcc.a} even when you want to avoid
11150 other standard libraries. In other words, when you specify @option{-nostdlib}
11151 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
11152 This ensures that you have no unresolved references to internal GCC
11153 library subroutines.
11154 (An example of such an internal subroutine is @code{__main}, used to ensure C++
11155 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
11156 GNU Compiler Collection (GCC) Internals}.)
11160 Produce a position independent executable on targets that support it.
11161 For predictable results, you must also specify the same set of options
11162 used for compilation (@option{-fpie}, @option{-fPIE},
11163 or model suboptions) when you specify this linker option.
11167 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
11168 that support it. This instructs the linker to add all symbols, not
11169 only used ones, to the dynamic symbol table. This option is needed
11170 for some uses of @code{dlopen} or to allow obtaining backtraces
11171 from within a program.
11175 Remove all symbol table and relocation information from the executable.
11179 On systems that support dynamic linking, this prevents linking with the shared
11180 libraries. On other systems, this option has no effect.
11184 Produce a shared object which can then be linked with other objects to
11185 form an executable. Not all systems support this option. For predictable
11186 results, you must also specify the same set of options used for compilation
11187 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
11188 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
11189 needs to build supplementary stub code for constructors to work. On
11190 multi-libbed systems, @samp{gcc -shared} must select the correct support
11191 libraries to link against. Failing to supply the correct flags may lead
11192 to subtle defects. Supplying them in cases where they are not necessary
11195 @item -shared-libgcc
11196 @itemx -static-libgcc
11197 @opindex shared-libgcc
11198 @opindex static-libgcc
11199 On systems that provide @file{libgcc} as a shared library, these options
11200 force the use of either the shared or static version, respectively.
11201 If no shared version of @file{libgcc} was built when the compiler was
11202 configured, these options have no effect.
11204 There are several situations in which an application should use the
11205 shared @file{libgcc} instead of the static version. The most common
11206 of these is when the application wishes to throw and catch exceptions
11207 across different shared libraries. In that case, each of the libraries
11208 as well as the application itself should use the shared @file{libgcc}.
11210 Therefore, the G++ and GCJ drivers automatically add
11211 @option{-shared-libgcc} whenever you build a shared library or a main
11212 executable, because C++ and Java programs typically use exceptions, so
11213 this is the right thing to do.
11215 If, instead, you use the GCC driver to create shared libraries, you may
11216 find that they are not always linked with the shared @file{libgcc}.
11217 If GCC finds, at its configuration time, that you have a non-GNU linker
11218 or a GNU linker that does not support option @option{--eh-frame-hdr},
11219 it links the shared version of @file{libgcc} into shared libraries
11220 by default. Otherwise, it takes advantage of the linker and optimizes
11221 away the linking with the shared version of @file{libgcc}, linking with
11222 the static version of libgcc by default. This allows exceptions to
11223 propagate through such shared libraries, without incurring relocation
11224 costs at library load time.
11226 However, if a library or main executable is supposed to throw or catch
11227 exceptions, you must link it using the G++ or GCJ driver, as appropriate
11228 for the languages used in the program, or using the option
11229 @option{-shared-libgcc}, such that it is linked with the shared
11232 @item -static-libasan
11233 @opindex static-libasan
11234 When the @option{-fsanitize=address} option is used to link a program,
11235 the GCC driver automatically links against @option{libasan}. If
11236 @file{libasan} is available as a shared library, and the @option{-static}
11237 option is not used, then this links against the shared version of
11238 @file{libasan}. The @option{-static-libasan} option directs the GCC
11239 driver to link @file{libasan} statically, without necessarily linking
11240 other libraries statically.
11242 @item -static-libtsan
11243 @opindex static-libtsan
11244 When the @option{-fsanitize=thread} option is used to link a program,
11245 the GCC driver automatically links against @option{libtsan}. If
11246 @file{libtsan} is available as a shared library, and the @option{-static}
11247 option is not used, then this links against the shared version of
11248 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
11249 driver to link @file{libtsan} statically, without necessarily linking
11250 other libraries statically.
11252 @item -static-liblsan
11253 @opindex static-liblsan
11254 When the @option{-fsanitize=leak} option is used to link a program,
11255 the GCC driver automatically links against @option{liblsan}. If
11256 @file{liblsan} is available as a shared library, and the @option{-static}
11257 option is not used, then this links against the shared version of
11258 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
11259 driver to link @file{liblsan} statically, without necessarily linking
11260 other libraries statically.
11262 @item -static-libubsan
11263 @opindex static-libubsan
11264 When the @option{-fsanitize=undefined} option is used to link a program,
11265 the GCC driver automatically links against @option{libubsan}. If
11266 @file{libubsan} is available as a shared library, and the @option{-static}
11267 option is not used, then this links against the shared version of
11268 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
11269 driver to link @file{libubsan} statically, without necessarily linking
11270 other libraries statically.
11272 @item -static-libmpx
11273 @opindex static-libmpx
11274 When @option{-fcheck-pointer bounds} and @option{-mmpx} options are
11275 used to link a program, the GCC driver automatically links against
11276 @option{libmpx}. If @file{libmpx} is available as a shared library,
11277 and the @option{-static} option is not used, then this links against
11278 the shared version of @file{libmpx}. The @option{-static-libmpx}
11279 option directs the GCC driver to link @file{libmpx} statically,
11280 without necessarily linking other libraries statically.
11282 @item -static-libmpxwrappers
11283 @opindex static-libmpxwrappers
11284 When @option{-fcheck-pointer bounds}, @option{-mmpx} options are used and
11285 @option{-fno-chkp-use-wrappers} option is not used to link a program, the
11286 GCC driver automatically links against @option{libmpxwrappers}. If
11287 @file{libmpxwrappers} is available as a shared library, and the
11288 @option{-static} option is not used, then this links against the shared
11289 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
11290 option directs the GCC driver to link @file{libmpxwrappers} statically,
11291 without necessarily linking other libraries statically.
11293 @item -static-libstdc++
11294 @opindex static-libstdc++
11295 When the @command{g++} program is used to link a C++ program, it
11296 normally automatically links against @option{libstdc++}. If
11297 @file{libstdc++} is available as a shared library, and the
11298 @option{-static} option is not used, then this links against the
11299 shared version of @file{libstdc++}. That is normally fine. However, it
11300 is sometimes useful to freeze the version of @file{libstdc++} used by
11301 the program without going all the way to a fully static link. The
11302 @option{-static-libstdc++} option directs the @command{g++} driver to
11303 link @file{libstdc++} statically, without necessarily linking other
11304 libraries statically.
11308 Bind references to global symbols when building a shared object. Warn
11309 about any unresolved references (unless overridden by the link editor
11310 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
11313 @item -T @var{script}
11315 @cindex linker script
11316 Use @var{script} as the linker script. This option is supported by most
11317 systems using the GNU linker. On some targets, such as bare-board
11318 targets without an operating system, the @option{-T} option may be required
11319 when linking to avoid references to undefined symbols.
11321 @item -Xlinker @var{option}
11323 Pass @var{option} as an option to the linker. You can use this to
11324 supply system-specific linker options that GCC does not recognize.
11326 If you want to pass an option that takes a separate argument, you must use
11327 @option{-Xlinker} twice, once for the option and once for the argument.
11328 For example, to pass @option{-assert definitions}, you must write
11329 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
11330 @option{-Xlinker "-assert definitions"}, because this passes the entire
11331 string as a single argument, which is not what the linker expects.
11333 When using the GNU linker, it is usually more convenient to pass
11334 arguments to linker options using the @option{@var{option}=@var{value}}
11335 syntax than as separate arguments. For example, you can specify
11336 @option{-Xlinker -Map=output.map} rather than
11337 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
11338 this syntax for command-line options.
11340 @item -Wl,@var{option}
11342 Pass @var{option} as an option to the linker. If @var{option} contains
11343 commas, it is split into multiple options at the commas. You can use this
11344 syntax to pass an argument to the option.
11345 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
11346 linker. When using the GNU linker, you can also get the same effect with
11347 @option{-Wl,-Map=output.map}.
11349 @item -u @var{symbol}
11351 Pretend the symbol @var{symbol} is undefined, to force linking of
11352 library modules to define it. You can use @option{-u} multiple times with
11353 different symbols to force loading of additional library modules.
11355 @item -z @var{keyword}
11357 @option{-z} is passed directly on to the linker along with the keyword
11358 @var{keyword}. See the section in the documentation of your linker for
11359 permitted values and their meanings.
11362 @node Directory Options
11363 @section Options for Directory Search
11364 @cindex directory options
11365 @cindex options, directory search
11366 @cindex search path
11368 These options specify directories to search for header files, for
11369 libraries and for parts of the compiler:
11374 Add the directory @var{dir} to the head of the list of directories to be
11375 searched for header files. This can be used to override a system header
11376 file, substituting your own version, since these directories are
11377 searched before the system header file directories. However, you should
11378 not use this option to add directories that contain vendor-supplied
11379 system header files (use @option{-isystem} for that). If you use more than
11380 one @option{-I} option, the directories are scanned in left-to-right
11381 order; the standard system directories come after.
11383 If a standard system include directory, or a directory specified with
11384 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
11385 option is ignored. The directory is still searched but as a
11386 system directory at its normal position in the system include chain.
11387 This is to ensure that GCC's procedure to fix buggy system headers and
11388 the ordering for the @code{include_next} directive are not inadvertently changed.
11389 If you really need to change the search order for system directories,
11390 use the @option{-nostdinc} and/or @option{-isystem} options.
11392 @item -iplugindir=@var{dir}
11393 @opindex iplugindir=
11394 Set the directory to search for plugins that are passed
11395 by @option{-fplugin=@var{name}} instead of
11396 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
11397 to be used by the user, but only passed by the driver.
11399 @item -iquote@var{dir}
11401 Add the directory @var{dir} to the head of the list of directories to
11402 be searched for header files only for the case of @code{#include
11403 "@var{file}"}; they are not searched for @code{#include <@var{file}>},
11404 otherwise just like @option{-I}.
11408 Add directory @var{dir} to the list of directories to be searched
11411 @item -B@var{prefix}
11413 This option specifies where to find the executables, libraries,
11414 include files, and data files of the compiler itself.
11416 The compiler driver program runs one or more of the subprograms
11417 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
11418 @var{prefix} as a prefix for each program it tries to run, both with and
11419 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
11421 For each subprogram to be run, the compiler driver first tries the
11422 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
11423 is not specified, the driver tries two standard prefixes,
11424 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
11425 those results in a file name that is found, the unmodified program
11426 name is searched for using the directories specified in your
11427 @env{PATH} environment variable.
11429 The compiler checks to see if the path provided by the @option{-B}
11430 refers to a directory, and if necessary it adds a directory
11431 separator character at the end of the path.
11433 @option{-B} prefixes that effectively specify directory names also apply
11434 to libraries in the linker, because the compiler translates these
11435 options into @option{-L} options for the linker. They also apply to
11436 include files in the preprocessor, because the compiler translates these
11437 options into @option{-isystem} options for the preprocessor. In this case,
11438 the compiler appends @samp{include} to the prefix.
11440 The runtime support file @file{libgcc.a} can also be searched for using
11441 the @option{-B} prefix, if needed. If it is not found there, the two
11442 standard prefixes above are tried, and that is all. The file is left
11443 out of the link if it is not found by those means.
11445 Another way to specify a prefix much like the @option{-B} prefix is to use
11446 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
11449 As a special kludge, if the path provided by @option{-B} is
11450 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
11451 9, then it is replaced by @file{[dir/]include}. This is to help
11452 with boot-strapping the compiler.
11454 @item -specs=@var{file}
11456 Process @var{file} after the compiler reads in the standard @file{specs}
11457 file, in order to override the defaults which the @command{gcc} driver
11458 program uses when determining what switches to pass to @command{cc1},
11459 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
11460 @option{-specs=@var{file}} can be specified on the command line, and they
11461 are processed in order, from left to right.
11463 @item --sysroot=@var{dir}
11465 Use @var{dir} as the logical root directory for headers and libraries.
11466 For example, if the compiler normally searches for headers in
11467 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
11468 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
11470 If you use both this option and the @option{-isysroot} option, then
11471 the @option{--sysroot} option applies to libraries, but the
11472 @option{-isysroot} option applies to header files.
11474 The GNU linker (beginning with version 2.16) has the necessary support
11475 for this option. If your linker does not support this option, the
11476 header file aspect of @option{--sysroot} still works, but the
11477 library aspect does not.
11479 @item --no-sysroot-suffix
11480 @opindex no-sysroot-suffix
11481 For some targets, a suffix is added to the root directory specified
11482 with @option{--sysroot}, depending on the other options used, so that
11483 headers may for example be found in
11484 @file{@var{dir}/@var{suffix}/usr/include} instead of
11485 @file{@var{dir}/usr/include}. This option disables the addition of
11490 This option has been deprecated. Please use @option{-iquote} instead for
11491 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
11492 Any directories you specify with @option{-I} options before the @option{-I-}
11493 option are searched only for the case of @code{#include "@var{file}"};
11494 they are not searched for @code{#include <@var{file}>}.
11496 If additional directories are specified with @option{-I} options after
11497 the @option{-I-}, these directories are searched for all @code{#include}
11498 directives. (Ordinarily @emph{all} @option{-I} directories are used
11501 In addition, the @option{-I-} option inhibits the use of the current
11502 directory (where the current input file came from) as the first search
11503 directory for @code{#include "@var{file}"}. There is no way to
11504 override this effect of @option{-I-}. With @option{-I.} you can specify
11505 searching the directory that is current when the compiler is
11506 invoked. That is not exactly the same as what the preprocessor does
11507 by default, but it is often satisfactory.
11509 @option{-I-} does not inhibit the use of the standard system directories
11510 for header files. Thus, @option{-I-} and @option{-nostdinc} are
11517 @section Specifying Subprocesses and the Switches to Pass to Them
11520 @command{gcc} is a driver program. It performs its job by invoking a
11521 sequence of other programs to do the work of compiling, assembling and
11522 linking. GCC interprets its command-line parameters and uses these to
11523 deduce which programs it should invoke, and which command-line options
11524 it ought to place on their command lines. This behavior is controlled
11525 by @dfn{spec strings}. In most cases there is one spec string for each
11526 program that GCC can invoke, but a few programs have multiple spec
11527 strings to control their behavior. The spec strings built into GCC can
11528 be overridden by using the @option{-specs=} command-line switch to specify
11531 @dfn{Spec files} are plaintext files that are used to construct spec
11532 strings. They consist of a sequence of directives separated by blank
11533 lines. The type of directive is determined by the first non-whitespace
11534 character on the line, which can be one of the following:
11537 @item %@var{command}
11538 Issues a @var{command} to the spec file processor. The commands that can
11542 @item %include <@var{file}>
11543 @cindex @code{%include}
11544 Search for @var{file} and insert its text at the current point in the
11547 @item %include_noerr <@var{file}>
11548 @cindex @code{%include_noerr}
11549 Just like @samp{%include}, but do not generate an error message if the include
11550 file cannot be found.
11552 @item %rename @var{old_name} @var{new_name}
11553 @cindex @code{%rename}
11554 Rename the spec string @var{old_name} to @var{new_name}.
11558 @item *[@var{spec_name}]:
11559 This tells the compiler to create, override or delete the named spec
11560 string. All lines after this directive up to the next directive or
11561 blank line are considered to be the text for the spec string. If this
11562 results in an empty string then the spec is deleted. (Or, if the
11563 spec did not exist, then nothing happens.) Otherwise, if the spec
11564 does not currently exist a new spec is created. If the spec does
11565 exist then its contents are overridden by the text of this
11566 directive, unless the first character of that text is the @samp{+}
11567 character, in which case the text is appended to the spec.
11569 @item [@var{suffix}]:
11570 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
11571 and up to the next directive or blank line are considered to make up the
11572 spec string for the indicated suffix. When the compiler encounters an
11573 input file with the named suffix, it processes the spec string in
11574 order to work out how to compile that file. For example:
11578 z-compile -input %i
11581 This says that any input file whose name ends in @samp{.ZZ} should be
11582 passed to the program @samp{z-compile}, which should be invoked with the
11583 command-line switch @option{-input} and with the result of performing the
11584 @samp{%i} substitution. (See below.)
11586 As an alternative to providing a spec string, the text following a
11587 suffix directive can be one of the following:
11590 @item @@@var{language}
11591 This says that the suffix is an alias for a known @var{language}. This is
11592 similar to using the @option{-x} command-line switch to GCC to specify a
11593 language explicitly. For example:
11600 Says that .ZZ files are, in fact, C++ source files.
11603 This causes an error messages saying:
11606 @var{name} compiler not installed on this system.
11610 GCC already has an extensive list of suffixes built into it.
11611 This directive adds an entry to the end of the list of suffixes, but
11612 since the list is searched from the end backwards, it is effectively
11613 possible to override earlier entries using this technique.
11617 GCC has the following spec strings built into it. Spec files can
11618 override these strings or create their own. Note that individual
11619 targets can also add their own spec strings to this list.
11622 asm Options to pass to the assembler
11623 asm_final Options to pass to the assembler post-processor
11624 cpp Options to pass to the C preprocessor
11625 cc1 Options to pass to the C compiler
11626 cc1plus Options to pass to the C++ compiler
11627 endfile Object files to include at the end of the link
11628 link Options to pass to the linker
11629 lib Libraries to include on the command line to the linker
11630 libgcc Decides which GCC support library to pass to the linker
11631 linker Sets the name of the linker
11632 predefines Defines to be passed to the C preprocessor
11633 signed_char Defines to pass to CPP to say whether @code{char} is signed
11635 startfile Object files to include at the start of the link
11638 Here is a small example of a spec file:
11641 %rename lib old_lib
11644 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
11647 This example renames the spec called @samp{lib} to @samp{old_lib} and
11648 then overrides the previous definition of @samp{lib} with a new one.
11649 The new definition adds in some extra command-line options before
11650 including the text of the old definition.
11652 @dfn{Spec strings} are a list of command-line options to be passed to their
11653 corresponding program. In addition, the spec strings can contain
11654 @samp{%}-prefixed sequences to substitute variable text or to
11655 conditionally insert text into the command line. Using these constructs
11656 it is possible to generate quite complex command lines.
11658 Here is a table of all defined @samp{%}-sequences for spec
11659 strings. Note that spaces are not generated automatically around the
11660 results of expanding these sequences. Therefore you can concatenate them
11661 together or combine them with constant text in a single argument.
11665 Substitute one @samp{%} into the program name or argument.
11668 Substitute the name of the input file being processed.
11671 Substitute the basename of the input file being processed.
11672 This is the substring up to (and not including) the last period
11673 and not including the directory.
11676 This is the same as @samp{%b}, but include the file suffix (text after
11680 Marks the argument containing or following the @samp{%d} as a
11681 temporary file name, so that that file is deleted if GCC exits
11682 successfully. Unlike @samp{%g}, this contributes no text to the
11685 @item %g@var{suffix}
11686 Substitute a file name that has suffix @var{suffix} and is chosen
11687 once per compilation, and mark the argument in the same way as
11688 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
11689 name is now chosen in a way that is hard to predict even when previously
11690 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
11691 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
11692 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
11693 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
11694 was simply substituted with a file name chosen once per compilation,
11695 without regard to any appended suffix (which was therefore treated
11696 just like ordinary text), making such attacks more likely to succeed.
11698 @item %u@var{suffix}
11699 Like @samp{%g}, but generates a new temporary file name
11700 each time it appears instead of once per compilation.
11702 @item %U@var{suffix}
11703 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
11704 new one if there is no such last file name. In the absence of any
11705 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
11706 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
11707 involves the generation of two distinct file names, one
11708 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
11709 simply substituted with a file name chosen for the previous @samp{%u},
11710 without regard to any appended suffix.
11712 @item %j@var{suffix}
11713 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
11714 writable, and if @option{-save-temps} is not used;
11715 otherwise, substitute the name
11716 of a temporary file, just like @samp{%u}. This temporary file is not
11717 meant for communication between processes, but rather as a junk
11718 disposal mechanism.
11720 @item %|@var{suffix}
11721 @itemx %m@var{suffix}
11722 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
11723 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
11724 all. These are the two most common ways to instruct a program that it
11725 should read from standard input or write to standard output. If you
11726 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
11727 construct: see for example @file{f/lang-specs.h}.
11729 @item %.@var{SUFFIX}
11730 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
11731 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
11732 terminated by the next space or %.
11735 Marks the argument containing or following the @samp{%w} as the
11736 designated output file of this compilation. This puts the argument
11737 into the sequence of arguments that @samp{%o} substitutes.
11740 Substitutes the names of all the output files, with spaces
11741 automatically placed around them. You should write spaces
11742 around the @samp{%o} as well or the results are undefined.
11743 @samp{%o} is for use in the specs for running the linker.
11744 Input files whose names have no recognized suffix are not compiled
11745 at all, but they are included among the output files, so they are
11749 Substitutes the suffix for object files. Note that this is
11750 handled specially when it immediately follows @samp{%g, %u, or %U},
11751 because of the need for those to form complete file names. The
11752 handling is such that @samp{%O} is treated exactly as if it had already
11753 been substituted, except that @samp{%g, %u, and %U} do not currently
11754 support additional @var{suffix} characters following @samp{%O} as they do
11755 following, for example, @samp{.o}.
11758 Substitutes the standard macro predefinitions for the
11759 current target machine. Use this when running @command{cpp}.
11762 Like @samp{%p}, but puts @samp{__} before and after the name of each
11763 predefined macro, except for macros that start with @samp{__} or with
11764 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
11768 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
11769 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
11770 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
11771 and @option{-imultilib} as necessary.
11774 Current argument is the name of a library or startup file of some sort.
11775 Search for that file in a standard list of directories and substitute
11776 the full name found. The current working directory is included in the
11777 list of directories scanned.
11780 Current argument is the name of a linker script. Search for that file
11781 in the current list of directories to scan for libraries. If the file
11782 is located insert a @option{--script} option into the command line
11783 followed by the full path name found. If the file is not found then
11784 generate an error message. Note: the current working directory is not
11788 Print @var{str} as an error message. @var{str} is terminated by a newline.
11789 Use this when inconsistent options are detected.
11791 @item %(@var{name})
11792 Substitute the contents of spec string @var{name} at this point.
11794 @item %x@{@var{option}@}
11795 Accumulate an option for @samp{%X}.
11798 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
11802 Output the accumulated assembler options specified by @option{-Wa}.
11805 Output the accumulated preprocessor options specified by @option{-Wp}.
11808 Process the @code{asm} spec. This is used to compute the
11809 switches to be passed to the assembler.
11812 Process the @code{asm_final} spec. This is a spec string for
11813 passing switches to an assembler post-processor, if such a program is
11817 Process the @code{link} spec. This is the spec for computing the
11818 command line passed to the linker. Typically it makes use of the
11819 @samp{%L %G %S %D and %E} sequences.
11822 Dump out a @option{-L} option for each directory that GCC believes might
11823 contain startup files. If the target supports multilibs then the
11824 current multilib directory is prepended to each of these paths.
11827 Process the @code{lib} spec. This is a spec string for deciding which
11828 libraries are included on the command line to the linker.
11831 Process the @code{libgcc} spec. This is a spec string for deciding
11832 which GCC support library is included on the command line to the linker.
11835 Process the @code{startfile} spec. This is a spec for deciding which
11836 object files are the first ones passed to the linker. Typically
11837 this might be a file named @file{crt0.o}.
11840 Process the @code{endfile} spec. This is a spec string that specifies
11841 the last object files that are passed to the linker.
11844 Process the @code{cpp} spec. This is used to construct the arguments
11845 to be passed to the C preprocessor.
11848 Process the @code{cc1} spec. This is used to construct the options to be
11849 passed to the actual C compiler (@command{cc1}).
11852 Process the @code{cc1plus} spec. This is used to construct the options to be
11853 passed to the actual C++ compiler (@command{cc1plus}).
11856 Substitute the variable part of a matched option. See below.
11857 Note that each comma in the substituted string is replaced by
11861 Remove all occurrences of @code{-S} from the command line. Note---this
11862 command is position dependent. @samp{%} commands in the spec string
11863 before this one see @code{-S}, @samp{%} commands in the spec string
11864 after this one do not.
11866 @item %:@var{function}(@var{args})
11867 Call the named function @var{function}, passing it @var{args}.
11868 @var{args} is first processed as a nested spec string, then split
11869 into an argument vector in the usual fashion. The function returns
11870 a string which is processed as if it had appeared literally as part
11871 of the current spec.
11873 The following built-in spec functions are provided:
11876 @item @code{getenv}
11877 The @code{getenv} spec function takes two arguments: an environment
11878 variable name and a string. If the environment variable is not
11879 defined, a fatal error is issued. Otherwise, the return value is the
11880 value of the environment variable concatenated with the string. For
11881 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
11884 %:getenv(TOPDIR /include)
11887 expands to @file{/path/to/top/include}.
11889 @item @code{if-exists}
11890 The @code{if-exists} spec function takes one argument, an absolute
11891 pathname to a file. If the file exists, @code{if-exists} returns the
11892 pathname. Here is a small example of its usage:
11896 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
11899 @item @code{if-exists-else}
11900 The @code{if-exists-else} spec function is similar to the @code{if-exists}
11901 spec function, except that it takes two arguments. The first argument is
11902 an absolute pathname to a file. If the file exists, @code{if-exists-else}
11903 returns the pathname. If it does not exist, it returns the second argument.
11904 This way, @code{if-exists-else} can be used to select one file or another,
11905 based on the existence of the first. Here is a small example of its usage:
11909 crt0%O%s %:if-exists(crti%O%s) \
11910 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
11913 @item @code{replace-outfile}
11914 The @code{replace-outfile} spec function takes two arguments. It looks for the
11915 first argument in the outfiles array and replaces it with the second argument. Here
11916 is a small example of its usage:
11919 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
11922 @item @code{remove-outfile}
11923 The @code{remove-outfile} spec function takes one argument. It looks for the
11924 first argument in the outfiles array and removes it. Here is a small example
11928 %:remove-outfile(-lm)
11931 @item @code{pass-through-libs}
11932 The @code{pass-through-libs} spec function takes any number of arguments. It
11933 finds any @option{-l} options and any non-options ending in @file{.a} (which it
11934 assumes are the names of linker input library archive files) and returns a
11935 result containing all the found arguments each prepended by
11936 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
11937 intended to be passed to the LTO linker plugin.
11940 %:pass-through-libs(%G %L %G)
11943 @item @code{print-asm-header}
11944 The @code{print-asm-header} function takes no arguments and simply
11945 prints a banner like:
11951 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
11954 It is used to separate compiler options from assembler options
11955 in the @option{--target-help} output.
11958 @item %@{@code{S}@}
11959 Substitutes the @code{-S} switch, if that switch is given to GCC@.
11960 If that switch is not specified, this substitutes nothing. Note that
11961 the leading dash is omitted when specifying this option, and it is
11962 automatically inserted if the substitution is performed. Thus the spec
11963 string @samp{%@{foo@}} matches the command-line option @option{-foo}
11964 and outputs the command-line option @option{-foo}.
11966 @item %W@{@code{S}@}
11967 Like %@{@code{S}@} but mark last argument supplied within as a file to be
11968 deleted on failure.
11970 @item %@{@code{S}*@}
11971 Substitutes all the switches specified to GCC whose names start
11972 with @code{-S}, but which also take an argument. This is used for
11973 switches like @option{-o}, @option{-D}, @option{-I}, etc.
11974 GCC considers @option{-o foo} as being
11975 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
11976 text, including the space. Thus two arguments are generated.
11978 @item %@{@code{S}*&@code{T}*@}
11979 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
11980 (the order of @code{S} and @code{T} in the spec is not significant).
11981 There can be any number of ampersand-separated variables; for each the
11982 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
11984 @item %@{@code{S}:@code{X}@}
11985 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
11987 @item %@{!@code{S}:@code{X}@}
11988 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
11990 @item %@{@code{S}*:@code{X}@}
11991 Substitutes @code{X} if one or more switches whose names start with
11992 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
11993 once, no matter how many such switches appeared. However, if @code{%*}
11994 appears somewhere in @code{X}, then @code{X} is substituted once
11995 for each matching switch, with the @code{%*} replaced by the part of
11996 that switch matching the @code{*}.
11998 If @code{%*} appears as the last part of a spec sequence then a space
11999 is added after the end of the last substitution. If there is more
12000 text in the sequence, however, then a space is not generated. This
12001 allows the @code{%*} substitution to be used as part of a larger
12002 string. For example, a spec string like this:
12005 %@{mcu=*:--script=%*/memory.ld@}
12009 when matching an option like @option{-mcu=newchip} produces:
12012 --script=newchip/memory.ld
12015 @item %@{.@code{S}:@code{X}@}
12016 Substitutes @code{X}, if processing a file with suffix @code{S}.
12018 @item %@{!.@code{S}:@code{X}@}
12019 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
12021 @item %@{,@code{S}:@code{X}@}
12022 Substitutes @code{X}, if processing a file for language @code{S}.
12024 @item %@{!,@code{S}:@code{X}@}
12025 Substitutes @code{X}, if not processing a file for language @code{S}.
12027 @item %@{@code{S}|@code{P}:@code{X}@}
12028 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
12029 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
12030 @code{*} sequences as well, although they have a stronger binding than
12031 the @samp{|}. If @code{%*} appears in @code{X}, all of the
12032 alternatives must be starred, and only the first matching alternative
12035 For example, a spec string like this:
12038 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
12042 outputs the following command-line options from the following input
12043 command-line options:
12048 -d fred.c -foo -baz -boggle
12049 -d jim.d -bar -baz -boggle
12052 @item %@{S:X; T:Y; :D@}
12054 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
12055 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
12056 be as many clauses as you need. This may be combined with @code{.},
12057 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
12062 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
12063 construct may contain other nested @samp{%} constructs or spaces, or
12064 even newlines. They are processed as usual, as described above.
12065 Trailing white space in @code{X} is ignored. White space may also
12066 appear anywhere on the left side of the colon in these constructs,
12067 except between @code{.} or @code{*} and the corresponding word.
12069 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
12070 handled specifically in these constructs. If another value of
12071 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
12072 @option{-W} switch is found later in the command line, the earlier
12073 switch value is ignored, except with @{@code{S}*@} where @code{S} is
12074 just one letter, which passes all matching options.
12076 The character @samp{|} at the beginning of the predicate text is used to
12077 indicate that a command should be piped to the following command, but
12078 only if @option{-pipe} is specified.
12080 It is built into GCC which switches take arguments and which do not.
12081 (You might think it would be useful to generalize this to allow each
12082 compiler's spec to say which switches take arguments. But this cannot
12083 be done in a consistent fashion. GCC cannot even decide which input
12084 files have been specified without knowing which switches take arguments,
12085 and it must know which input files to compile in order to tell which
12088 GCC also knows implicitly that arguments starting in @option{-l} are to be
12089 treated as compiler output files, and passed to the linker in their
12090 proper position among the other output files.
12092 @c man begin OPTIONS
12094 @node Target Options
12095 @section Specifying Target Machine and Compiler Version
12096 @cindex target options
12097 @cindex cross compiling
12098 @cindex specifying machine version
12099 @cindex specifying compiler version and target machine
12100 @cindex compiler version, specifying
12101 @cindex target machine, specifying
12103 The usual way to run GCC is to run the executable called @command{gcc}, or
12104 @command{@var{machine}-gcc} when cross-compiling, or
12105 @command{@var{machine}-gcc-@var{version}} to run a version other than the
12106 one that was installed last.
12108 @node Submodel Options
12109 @section Hardware Models and Configurations
12110 @cindex submodel options
12111 @cindex specifying hardware config
12112 @cindex hardware models and configurations, specifying
12113 @cindex machine dependent options
12115 Each target machine types can have its own
12116 special options, starting with @samp{-m}, to choose among various
12117 hardware models or configurations---for example, 68010 vs 68020,
12118 floating coprocessor or none. A single installed version of the
12119 compiler can compile for any model or configuration, according to the
12122 Some configurations of the compiler also support additional special
12123 options, usually for compatibility with other compilers on the same
12126 @c This list is ordered alphanumerically by subsection name.
12127 @c It should be the same order and spelling as these options are listed
12128 @c in Machine Dependent Options
12131 * AArch64 Options::
12132 * Adapteva Epiphany Options::
12136 * Blackfin Options::
12141 * DEC Alpha Options::
12144 * GNU/Linux Options::
12154 * MicroBlaze Options::
12157 * MN10300 Options::
12161 * Nios II Options::
12162 * Nvidia PTX Options::
12164 * picoChip Options::
12165 * PowerPC Options::
12167 * RS/6000 and PowerPC Options::
12169 * S/390 and zSeries Options::
12172 * Solaris 2 Options::
12175 * System V Options::
12176 * TILE-Gx Options::
12177 * TILEPro Options::
12182 * VxWorks Options::
12184 * x86 Windows Options::
12185 * Xstormy16 Options::
12187 * zSeries Options::
12190 @node AArch64 Options
12191 @subsection AArch64 Options
12192 @cindex AArch64 Options
12194 These options are defined for AArch64 implementations:
12198 @item -mabi=@var{name}
12200 Generate code for the specified data model. Permissible values
12201 are @samp{ilp32} for SysV-like data model where int, long int and pointer
12202 are 32-bit, and @samp{lp64} for SysV-like data model where int is 32-bit,
12203 but long int and pointer are 64-bit.
12205 The default depends on the specific target configuration. Note that
12206 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
12207 entire program with the same ABI, and link with a compatible set of libraries.
12210 @opindex mbig-endian
12211 Generate big-endian code. This is the default when GCC is configured for an
12212 @samp{aarch64_be-*-*} target.
12214 @item -mgeneral-regs-only
12215 @opindex mgeneral-regs-only
12216 Generate code which uses only the general registers.
12218 @item -mlittle-endian
12219 @opindex mlittle-endian
12220 Generate little-endian code. This is the default when GCC is configured for an
12221 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
12223 @item -mcmodel=tiny
12224 @opindex mcmodel=tiny
12225 Generate code for the tiny code model. The program and its statically defined
12226 symbols must be within 1GB of each other. Pointers are 64 bits. Programs can
12227 be statically or dynamically linked. This model is not fully implemented and
12228 mostly treated as @samp{small}.
12230 @item -mcmodel=small
12231 @opindex mcmodel=small
12232 Generate code for the small code model. The program and its statically defined
12233 symbols must be within 4GB of each other. Pointers are 64 bits. Programs can
12234 be statically or dynamically linked. This is the default code model.
12236 @item -mcmodel=large
12237 @opindex mcmodel=large
12238 Generate code for the large code model. This makes no assumptions about
12239 addresses and sizes of sections. Pointers are 64 bits. Programs can be
12240 statically linked only.
12242 @item -mstrict-align
12243 @opindex mstrict-align
12244 Do not assume that unaligned memory references are handled by the system.
12246 @item -momit-leaf-frame-pointer
12247 @itemx -mno-omit-leaf-frame-pointer
12248 @opindex momit-leaf-frame-pointer
12249 @opindex mno-omit-leaf-frame-pointer
12250 Omit or keep the frame pointer in leaf functions. The former behaviour is the
12253 @item -mtls-dialect=desc
12254 @opindex mtls-dialect=desc
12255 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
12256 of TLS variables. This is the default.
12258 @item -mtls-dialect=traditional
12259 @opindex mtls-dialect=traditional
12260 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
12263 @item -mfix-cortex-a53-835769
12264 @itemx -mno-fix-cortex-a53-835769
12265 @opindex mfix-cortex-a53-835769
12266 @opindex mno-fix-cortex-a53-835769
12267 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
12268 This involves inserting a NOP instruction between memory instructions and
12269 64-bit integer multiply-accumulate instructions.
12271 @item -march=@var{name}
12273 Specify the name of the target architecture, optionally suffixed by one or
12274 more feature modifiers. This option has the form
12275 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
12276 only permissible value for @var{arch} is @samp{armv8-a}. The permissible
12277 values for @var{feature} are documented in the sub-section below.
12279 Where conflicting feature modifiers are specified, the right-most feature is
12282 GCC uses this name to determine what kind of instructions it can emit when
12283 generating assembly code.
12285 Where @option{-march} is specified without either of @option{-mtune}
12286 or @option{-mcpu} also being specified, the code is tuned to perform
12287 well across a range of target processors implementing the target
12290 @item -mtune=@var{name}
12292 Specify the name of the target processor for which GCC should tune the
12293 performance of the code. Permissible values for this option are:
12294 @samp{generic}, @samp{cortex-a53}, @samp{cortex-a57},
12295 @samp{cortex-a72}, @samp{thunderx}, @samp{xgene1}.
12297 Additionally, this option can specify that GCC should tune the performance
12298 of the code for a big.LITTLE system. Permissible values for this
12299 option are: @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}.
12301 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
12302 are specified, the code is tuned to perform well across a range
12303 of target processors.
12305 This option cannot be suffixed by feature modifiers.
12307 @item -mcpu=@var{name}
12309 Specify the name of the target processor, optionally suffixed by one or more
12310 feature modifiers. This option has the form
12311 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
12312 permissible values for @var{cpu} are the same as those available for
12315 The permissible values for @var{feature} are documented in the sub-section
12318 Where conflicting feature modifiers are specified, the right-most feature is
12321 GCC uses this name to determine what kind of instructions it can emit when
12322 generating assembly code (as if by @option{-march}) and to determine
12323 the target processor for which to tune for performance (as if
12324 by @option{-mtune}). Where this option is used in conjunction
12325 with @option{-march} or @option{-mtune}, those options take precedence
12326 over the appropriate part of this option.
12329 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
12330 @cindex @option{-march} feature modifiers
12331 @cindex @option{-mcpu} feature modifiers
12332 Feature modifiers used with @option{-march} and @option{-mcpu} can be one
12337 Enable CRC extension.
12339 Enable Crypto extension. This implies Advanced SIMD is enabled.
12341 Enable floating-point instructions.
12343 Enable Advanced SIMD instructions. This implies floating-point instructions
12344 are enabled. This is the default for all current possible values for options
12345 @option{-march} and @option{-mcpu=}.
12348 @node Adapteva Epiphany Options
12349 @subsection Adapteva Epiphany Options
12351 These @samp{-m} options are defined for Adapteva Epiphany:
12354 @item -mhalf-reg-file
12355 @opindex mhalf-reg-file
12356 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
12357 That allows code to run on hardware variants that lack these registers.
12359 @item -mprefer-short-insn-regs
12360 @opindex mprefer-short-insn-regs
12361 Preferrentially allocate registers that allow short instruction generation.
12362 This can result in increased instruction count, so this may either reduce or
12363 increase overall code size.
12365 @item -mbranch-cost=@var{num}
12366 @opindex mbranch-cost
12367 Set the cost of branches to roughly @var{num} ``simple'' instructions.
12368 This cost is only a heuristic and is not guaranteed to produce
12369 consistent results across releases.
12373 Enable the generation of conditional moves.
12375 @item -mnops=@var{num}
12377 Emit @var{num} NOPs before every other generated instruction.
12379 @item -mno-soft-cmpsf
12380 @opindex mno-soft-cmpsf
12381 For single-precision floating-point comparisons, emit an @code{fsub} instruction
12382 and test the flags. This is faster than a software comparison, but can
12383 get incorrect results in the presence of NaNs, or when two different small
12384 numbers are compared such that their difference is calculated as zero.
12385 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
12386 software comparisons.
12388 @item -mstack-offset=@var{num}
12389 @opindex mstack-offset
12390 Set the offset between the top of the stack and the stack pointer.
12391 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
12392 can be used by leaf functions without stack allocation.
12393 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
12394 Note also that this option changes the ABI; compiling a program with a
12395 different stack offset than the libraries have been compiled with
12396 generally does not work.
12397 This option can be useful if you want to evaluate if a different stack
12398 offset would give you better code, but to actually use a different stack
12399 offset to build working programs, it is recommended to configure the
12400 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
12402 @item -mno-round-nearest
12403 @opindex mno-round-nearest
12404 Make the scheduler assume that the rounding mode has been set to
12405 truncating. The default is @option{-mround-nearest}.
12408 @opindex mlong-calls
12409 If not otherwise specified by an attribute, assume all calls might be beyond
12410 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
12411 function address into a register before performing a (otherwise direct) call.
12412 This is the default.
12414 @item -mshort-calls
12415 @opindex short-calls
12416 If not otherwise specified by an attribute, assume all direct calls are
12417 in the range of the @code{b} / @code{bl} instructions, so use these instructions
12418 for direct calls. The default is @option{-mlong-calls}.
12422 Assume addresses can be loaded as 16-bit unsigned values. This does not
12423 apply to function addresses for which @option{-mlong-calls} semantics
12426 @item -mfp-mode=@var{mode}
12428 Set the prevailing mode of the floating-point unit.
12429 This determines the floating-point mode that is provided and expected
12430 at function call and return time. Making this mode match the mode you
12431 predominantly need at function start can make your programs smaller and
12432 faster by avoiding unnecessary mode switches.
12434 @var{mode} can be set to one the following values:
12438 Any mode at function entry is valid, and retained or restored when
12439 the function returns, and when it calls other functions.
12440 This mode is useful for compiling libraries or other compilation units
12441 you might want to incorporate into different programs with different
12442 prevailing FPU modes, and the convenience of being able to use a single
12443 object file outweighs the size and speed overhead for any extra
12444 mode switching that might be needed, compared with what would be needed
12445 with a more specific choice of prevailing FPU mode.
12448 This is the mode used for floating-point calculations with
12449 truncating (i.e.@: round towards zero) rounding mode. That includes
12450 conversion from floating point to integer.
12452 @item round-nearest
12453 This is the mode used for floating-point calculations with
12454 round-to-nearest-or-even rounding mode.
12457 This is the mode used to perform integer calculations in the FPU, e.g.@:
12458 integer multiply, or integer multiply-and-accumulate.
12461 The default is @option{-mfp-mode=caller}
12463 @item -mnosplit-lohi
12464 @itemx -mno-postinc
12465 @itemx -mno-postmodify
12466 @opindex mnosplit-lohi
12467 @opindex mno-postinc
12468 @opindex mno-postmodify
12469 Code generation tweaks that disable, respectively, splitting of 32-bit
12470 loads, generation of post-increment addresses, and generation of
12471 post-modify addresses. The defaults are @option{msplit-lohi},
12472 @option{-mpost-inc}, and @option{-mpost-modify}.
12474 @item -mnovect-double
12475 @opindex mno-vect-double
12476 Change the preferred SIMD mode to SImode. The default is
12477 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
12479 @item -max-vect-align=@var{num}
12480 @opindex max-vect-align
12481 The maximum alignment for SIMD vector mode types.
12482 @var{num} may be 4 or 8. The default is 8.
12483 Note that this is an ABI change, even though many library function
12484 interfaces are unaffected if they don't use SIMD vector modes
12485 in places that affect size and/or alignment of relevant types.
12487 @item -msplit-vecmove-early
12488 @opindex msplit-vecmove-early
12489 Split vector moves into single word moves before reload. In theory this
12490 can give better register allocation, but so far the reverse seems to be
12491 generally the case.
12493 @item -m1reg-@var{reg}
12495 Specify a register to hold the constant @minus{}1, which makes loading small negative
12496 constants and certain bitmasks faster.
12497 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
12498 which specify use of that register as a fixed register,
12499 and @samp{none}, which means that no register is used for this
12500 purpose. The default is @option{-m1reg-none}.
12505 @subsection ARC Options
12506 @cindex ARC options
12508 The following options control the architecture variant for which code
12511 @c architecture variants
12514 @item -mbarrel-shifter
12515 @opindex mbarrel-shifter
12516 Generate instructions supported by barrel shifter. This is the default
12517 unless @option{-mcpu=ARC601} is in effect.
12519 @item -mcpu=@var{cpu}
12521 Set architecture type, register usage, and instruction scheduling
12522 parameters for @var{cpu}. There are also shortcut alias options
12523 available for backward compatibility and convenience. Supported
12524 values for @var{cpu} are
12530 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
12534 Compile for ARC601. Alias: @option{-mARC601}.
12539 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
12540 This is the default when configured with @option{--with-cpu=arc700}@.
12545 @itemx -mdpfp-compact
12546 @opindex mdpfp-compact
12547 FPX: Generate Double Precision FPX instructions, tuned for the compact
12551 @opindex mdpfp-fast
12552 FPX: Generate Double Precision FPX instructions, tuned for the fast
12555 @item -mno-dpfp-lrsr
12556 @opindex mno-dpfp-lrsr
12557 Disable LR and SR instructions from using FPX extension aux registers.
12561 Generate Extended arithmetic instructions. Currently only
12562 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
12563 supported. This is always enabled for @option{-mcpu=ARC700}.
12567 Do not generate mpy instructions for ARC700.
12571 Generate 32x16 bit multiply and mac instructions.
12575 Generate mul64 and mulu64 instructions. Only valid for @option{-mcpu=ARC600}.
12579 Generate norm instruction. This is the default if @option{-mcpu=ARC700}
12584 @itemx -mspfp-compact
12585 @opindex mspfp-compact
12586 FPX: Generate Single Precision FPX instructions, tuned for the compact
12590 @opindex mspfp-fast
12591 FPX: Generate Single Precision FPX instructions, tuned for the fast
12596 Enable generation of ARC SIMD instructions via target-specific
12597 builtins. Only valid for @option{-mcpu=ARC700}.
12600 @opindex msoft-float
12601 This option ignored; it is provided for compatibility purposes only.
12602 Software floating point code is emitted by default, and this default
12603 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
12604 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
12605 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
12609 Generate swap instructions.
12613 The following options are passed through to the assembler, and also
12614 define preprocessor macro symbols.
12616 @c Flags used by the assembler, but for which we define preprocessor
12617 @c macro symbols as well.
12620 @opindex mdsp-packa
12621 Passed down to the assembler to enable the DSP Pack A extensions.
12622 Also sets the preprocessor symbol @code{__Xdsp_packa}.
12626 Passed down to the assembler to enable the dual viterbi butterfly
12627 extension. Also sets the preprocessor symbol @code{__Xdvbf}.
12629 @c ARC700 4.10 extension instruction
12632 Passed down to the assembler to enable the Locked Load/Store
12633 Conditional extension. Also sets the preprocessor symbol
12638 Passed down to the assembler. Also sets the preprocessor symbol
12639 @code{__Xxmac_d16}.
12643 Passed down to the assembler. Also sets the preprocessor symbol
12646 @c ARC700 4.10 extension instruction
12649 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
12650 extension instruction. Also sets the preprocessor symbol
12653 @c ARC700 4.10 extension instruction
12656 Passed down to the assembler to enable the swap byte ordering
12657 extension instruction. Also sets the preprocessor symbol
12661 @opindex mtelephony
12662 Passed down to the assembler to enable dual and single operand
12663 instructions for telephony. Also sets the preprocessor symbol
12664 @code{__Xtelephony}.
12668 Passed down to the assembler to enable the XY Memory extension. Also
12669 sets the preprocessor symbol @code{__Xxy}.
12673 The following options control how the assembly code is annotated:
12675 @c Assembly annotation options
12679 Annotate assembler instructions with estimated addresses.
12681 @item -mannotate-align
12682 @opindex mannotate-align
12683 Explain what alignment considerations lead to the decision to make an
12684 instruction short or long.
12688 The following options are passed through to the linker:
12690 @c options passed through to the linker
12694 Passed through to the linker, to specify use of the @code{arclinux} emulation.
12695 This option is enabled by default in tool chains built for
12696 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
12697 when profiling is not requested.
12699 @item -marclinux_prof
12700 @opindex marclinux_prof
12701 Passed through to the linker, to specify use of the
12702 @code{arclinux_prof} emulation. This option is enabled by default in
12703 tool chains built for @w{@code{arc-linux-uclibc}} and
12704 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
12708 The following options control the semantics of generated code:
12710 @c semantically relevant code generation options
12712 @item -mepilogue-cfi
12713 @opindex mepilogue-cfi
12714 Enable generation of call frame information for epilogues.
12716 @item -mno-epilogue-cfi
12717 @opindex mno-epilogue-cfi
12718 Disable generation of call frame information for epilogues.
12721 @opindex mlong-calls
12722 Generate call insns as register indirect calls, thus providing access
12723 to the full 32-bit address range.
12725 @item -mmedium-calls
12726 @opindex mmedium-calls
12727 Don't use less than 25 bit addressing range for calls, which is the
12728 offset available for an unconditional branch-and-link
12729 instruction. Conditional execution of function calls is suppressed, to
12730 allow use of the 25-bit range, rather than the 21-bit range with
12731 conditional branch-and-link. This is the default for tool chains built
12732 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
12736 Do not generate sdata references. This is the default for tool chains
12737 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
12741 @opindex mucb-mcount
12742 Instrument with mcount calls as used in UCB code. I.e. do the
12743 counting in the callee, not the caller. By default ARC instrumentation
12744 counts in the caller.
12746 @item -mvolatile-cache
12747 @opindex mvolatile-cache
12748 Use ordinarily cached memory accesses for volatile references. This is the
12751 @item -mno-volatile-cache
12752 @opindex mno-volatile-cache
12753 Enable cache bypass for volatile references.
12757 The following options fine tune code generation:
12758 @c code generation tuning options
12761 @opindex malign-call
12762 Do alignment optimizations for call instructions.
12764 @item -mauto-modify-reg
12765 @opindex mauto-modify-reg
12766 Enable the use of pre/post modify with register displacement.
12768 @item -mbbit-peephole
12769 @opindex mbbit-peephole
12770 Enable bbit peephole2.
12774 This option disables a target-specific pass in @file{arc_reorg} to
12775 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
12776 generation driven by the combiner pass.
12778 @item -mcase-vector-pcrel
12779 @opindex mcase-vector-pcrel
12780 Use pc-relative switch case tables - this enables case table shortening.
12781 This is the default for @option{-Os}.
12783 @item -mcompact-casesi
12784 @opindex mcompact-casesi
12785 Enable compact casesi pattern.
12786 This is the default for @option{-Os}.
12788 @item -mno-cond-exec
12789 @opindex mno-cond-exec
12790 Disable ARCompact specific pass to generate conditional execution instructions.
12791 Due to delay slot scheduling and interactions between operand numbers,
12792 literal sizes, instruction lengths, and the support for conditional execution,
12793 the target-independent pass to generate conditional execution is often lacking,
12794 so the ARC port has kept a special pass around that tries to find more
12795 conditional execution generating opportunities after register allocation,
12796 branch shortening, and delay slot scheduling have been done. This pass
12797 generally, but not always, improves performance and code size, at the cost of
12798 extra compilation time, which is why there is an option to switch it off.
12799 If you have a problem with call instructions exceeding their allowable
12800 offset range because they are conditionalized, you should consider using
12801 @option{-mmedium-calls} instead.
12803 @item -mearly-cbranchsi
12804 @opindex mearly-cbranchsi
12805 Enable pre-reload use of the cbranchsi pattern.
12807 @item -mexpand-adddi
12808 @opindex mexpand-adddi
12809 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
12810 @code{add.f}, @code{adc} etc.
12812 @item -mindexed-loads
12813 @opindex mindexed-loads
12814 Enable the use of indexed loads. This can be problematic because some
12815 optimizers then assume that indexed stores exist, which is not
12820 Enable Local Register Allocation. This is still experimental for ARC,
12821 so by default the compiler uses standard reload
12822 (i.e. @option{-mno-lra}).
12824 @item -mlra-priority-none
12825 @opindex mlra-priority-none
12826 Don't indicate any priority for target registers.
12828 @item -mlra-priority-compact
12829 @opindex mlra-priority-compact
12830 Indicate target register priority for r0..r3 / r12..r15.
12832 @item -mlra-priority-noncompact
12833 @opindex mlra-priority-noncompact
12834 Reduce target regsiter priority for r0..r3 / r12..r15.
12836 @item -mno-millicode
12837 @opindex mno-millicode
12838 When optimizing for size (using @option{-Os}), prologues and epilogues
12839 that have to save or restore a large number of registers are often
12840 shortened by using call to a special function in libgcc; this is
12841 referred to as a @emph{millicode} call. As these calls can pose
12842 performance issues, and/or cause linking issues when linking in a
12843 nonstandard way, this option is provided to turn off millicode call
12847 @opindex mmixed-code
12848 Tweak register allocation to help 16-bit instruction generation.
12849 This generally has the effect of decreasing the average instruction size
12850 while increasing the instruction count.
12854 Enable 'q' instruction alternatives.
12855 This is the default for @option{-Os}.
12859 Enable Rcq constraint handling - most short code generation depends on this.
12860 This is the default.
12864 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
12865 This is the default.
12867 @item -msize-level=@var{level}
12868 @opindex msize-level
12869 Fine-tune size optimization with regards to instruction lengths and alignment.
12870 The recognized values for @var{level} are:
12873 No size optimization. This level is deprecated and treated like @samp{1}.
12876 Short instructions are used opportunistically.
12879 In addition, alignment of loops and of code after barriers are dropped.
12882 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
12886 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
12887 the behavior when this is not set is equivalent to level @samp{1}.
12889 @item -mtune=@var{cpu}
12891 Set instruction scheduling parameters for @var{cpu}, overriding any implied
12892 by @option{-mcpu=}.
12894 Supported values for @var{cpu} are
12898 Tune for ARC600 cpu.
12901 Tune for ARC601 cpu.
12904 Tune for ARC700 cpu with standard multiplier block.
12907 Tune for ARC700 cpu with XMAC block.
12910 Tune for ARC725D cpu.
12913 Tune for ARC750D cpu.
12917 @item -mmultcost=@var{num}
12919 Cost to assume for a multiply instruction, with @samp{4} being equal to a
12920 normal instruction.
12922 @item -munalign-prob-threshold=@var{probability}
12923 @opindex munalign-prob-threshold
12924 Set probability threshold for unaligning branches.
12925 When tuning for @samp{ARC700} and optimizing for speed, branches without
12926 filled delay slot are preferably emitted unaligned and long, unless
12927 profiling indicates that the probability for the branch to be taken
12928 is below @var{probability}. @xref{Cross-profiling}.
12929 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
12933 The following options are maintained for backward compatibility, but
12934 are now deprecated and will be removed in a future release:
12936 @c Deprecated options
12944 @opindex mbig-endian
12947 Compile code for big endian targets. Use of these options is now
12948 deprecated. Users wanting big-endian code, should use the
12949 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
12950 building the tool chain, for which big-endian is the default.
12952 @item -mlittle-endian
12953 @opindex mlittle-endian
12956 Compile code for little endian targets. Use of these options is now
12957 deprecated. Users wanting little-endian code should use the
12958 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
12959 building the tool chain, for which little-endian is the default.
12961 @item -mbarrel_shifter
12962 @opindex mbarrel_shifter
12963 Replaced by @option{-mbarrel-shifter}.
12965 @item -mdpfp_compact
12966 @opindex mdpfp_compact
12967 Replaced by @option{-mdpfp-compact}.
12970 @opindex mdpfp_fast
12971 Replaced by @option{-mdpfp-fast}.
12974 @opindex mdsp_packa
12975 Replaced by @option{-mdsp-packa}.
12979 Replaced by @option{-mea}.
12983 Replaced by @option{-mmac-24}.
12987 Replaced by @option{-mmac-d16}.
12989 @item -mspfp_compact
12990 @opindex mspfp_compact
12991 Replaced by @option{-mspfp-compact}.
12994 @opindex mspfp_fast
12995 Replaced by @option{-mspfp-fast}.
12997 @item -mtune=@var{cpu}
12999 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
13000 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
13001 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
13003 @item -multcost=@var{num}
13005 Replaced by @option{-mmultcost}.
13010 @subsection ARM Options
13011 @cindex ARM options
13013 These @samp{-m} options are defined for the ARM port:
13016 @item -mabi=@var{name}
13018 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
13019 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
13022 @opindex mapcs-frame
13023 Generate a stack frame that is compliant with the ARM Procedure Call
13024 Standard for all functions, even if this is not strictly necessary for
13025 correct execution of the code. Specifying @option{-fomit-frame-pointer}
13026 with this option causes the stack frames not to be generated for
13027 leaf functions. The default is @option{-mno-apcs-frame}.
13028 This option is deprecated.
13032 This is a synonym for @option{-mapcs-frame} and is deprecated.
13035 @c not currently implemented
13036 @item -mapcs-stack-check
13037 @opindex mapcs-stack-check
13038 Generate code to check the amount of stack space available upon entry to
13039 every function (that actually uses some stack space). If there is
13040 insufficient space available then either the function
13041 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
13042 called, depending upon the amount of stack space required. The runtime
13043 system is required to provide these functions. The default is
13044 @option{-mno-apcs-stack-check}, since this produces smaller code.
13046 @c not currently implemented
13048 @opindex mapcs-float
13049 Pass floating-point arguments using the floating-point registers. This is
13050 one of the variants of the APCS@. This option is recommended if the
13051 target hardware has a floating-point unit or if a lot of floating-point
13052 arithmetic is going to be performed by the code. The default is
13053 @option{-mno-apcs-float}, since the size of integer-only code is
13054 slightly increased if @option{-mapcs-float} is used.
13056 @c not currently implemented
13057 @item -mapcs-reentrant
13058 @opindex mapcs-reentrant
13059 Generate reentrant, position-independent code. The default is
13060 @option{-mno-apcs-reentrant}.
13063 @item -mthumb-interwork
13064 @opindex mthumb-interwork
13065 Generate code that supports calling between the ARM and Thumb
13066 instruction sets. Without this option, on pre-v5 architectures, the
13067 two instruction sets cannot be reliably used inside one program. The
13068 default is @option{-mno-thumb-interwork}, since slightly larger code
13069 is generated when @option{-mthumb-interwork} is specified. In AAPCS
13070 configurations this option is meaningless.
13072 @item -mno-sched-prolog
13073 @opindex mno-sched-prolog
13074 Prevent the reordering of instructions in the function prologue, or the
13075 merging of those instruction with the instructions in the function's
13076 body. This means that all functions start with a recognizable set
13077 of instructions (or in fact one of a choice from a small set of
13078 different function prologues), and this information can be used to
13079 locate the start of functions inside an executable piece of code. The
13080 default is @option{-msched-prolog}.
13082 @item -mfloat-abi=@var{name}
13083 @opindex mfloat-abi
13084 Specifies which floating-point ABI to use. Permissible values
13085 are: @samp{soft}, @samp{softfp} and @samp{hard}.
13087 Specifying @samp{soft} causes GCC to generate output containing
13088 library calls for floating-point operations.
13089 @samp{softfp} allows the generation of code using hardware floating-point
13090 instructions, but still uses the soft-float calling conventions.
13091 @samp{hard} allows generation of floating-point instructions
13092 and uses FPU-specific calling conventions.
13094 The default depends on the specific target configuration. Note that
13095 the hard-float and soft-float ABIs are not link-compatible; you must
13096 compile your entire program with the same ABI, and link with a
13097 compatible set of libraries.
13099 @item -mlittle-endian
13100 @opindex mlittle-endian
13101 Generate code for a processor running in little-endian mode. This is
13102 the default for all standard configurations.
13105 @opindex mbig-endian
13106 Generate code for a processor running in big-endian mode; the default is
13107 to compile code for a little-endian processor.
13109 @item -march=@var{name}
13111 This specifies the name of the target ARM architecture. GCC uses this
13112 name to determine what kind of instructions it can emit when generating
13113 assembly code. This option can be used in conjunction with or instead
13114 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
13115 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
13116 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
13117 @samp{armv6}, @samp{armv6j},
13118 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
13119 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m}, @samp{armv7e-m},
13120 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc},
13121 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
13123 @option{-march=armv7ve} is the armv7-a architecture with virtualization
13126 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
13127 architecture together with the optional CRC32 extensions.
13129 @option{-march=native} causes the compiler to auto-detect the architecture
13130 of the build computer. At present, this feature is only supported on
13131 GNU/Linux, and not all architectures are recognized. If the auto-detect
13132 is unsuccessful the option has no effect.
13134 @item -mtune=@var{name}
13136 This option specifies the name of the target ARM processor for
13137 which GCC should tune the performance of the code.
13138 For some ARM implementations better performance can be obtained by using
13140 Permissible names are: @samp{arm2}, @samp{arm250},
13141 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
13142 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
13143 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
13144 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
13146 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
13147 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
13148 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
13149 @samp{strongarm1110},
13150 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
13151 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
13152 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
13153 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
13154 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
13155 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
13156 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
13157 @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8}, @samp{cortex-a9},
13158 @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a53},
13159 @samp{cortex-a57}, @samp{cortex-a72},
13161 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-m7},
13166 @samp{cortex-m0plus},
13167 @samp{cortex-m1.small-multiply},
13168 @samp{cortex-m0.small-multiply},
13169 @samp{cortex-m0plus.small-multiply},
13170 @samp{marvell-pj4},
13171 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
13172 @samp{fa526}, @samp{fa626},
13173 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
13176 Additionally, this option can specify that GCC should tune the performance
13177 of the code for a big.LITTLE system. Permissible names are:
13178 @samp{cortex-a15.cortex-a7}, @samp{cortex-a57.cortex-a53},
13179 @samp{cortex-a72.cortex-a53}.
13181 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
13182 performance for a blend of processors within architecture @var{arch}.
13183 The aim is to generate code that run well on the current most popular
13184 processors, balancing between optimizations that benefit some CPUs in the
13185 range, and avoiding performance pitfalls of other CPUs. The effects of
13186 this option may change in future GCC versions as CPU models come and go.
13188 @option{-mtune=native} causes the compiler to auto-detect the CPU
13189 of the build computer. At present, this feature is only supported on
13190 GNU/Linux, and not all architectures are recognized. If the auto-detect is
13191 unsuccessful the option has no effect.
13193 @item -mcpu=@var{name}
13195 This specifies the name of the target ARM processor. GCC uses this name
13196 to derive the name of the target ARM architecture (as if specified
13197 by @option{-march}) and the ARM processor type for which to tune for
13198 performance (as if specified by @option{-mtune}). Where this option
13199 is used in conjunction with @option{-march} or @option{-mtune},
13200 those options take precedence over the appropriate part of this option.
13202 Permissible names for this option are the same as those for
13205 @option{-mcpu=generic-@var{arch}} is also permissible, and is
13206 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
13207 See @option{-mtune} for more information.
13209 @option{-mcpu=native} causes the compiler to auto-detect the CPU
13210 of the build computer. At present, this feature is only supported on
13211 GNU/Linux, and not all architectures are recognized. If the auto-detect
13212 is unsuccessful the option has no effect.
13214 @item -mfpu=@var{name}
13216 This specifies what floating-point hardware (or hardware emulation) is
13217 available on the target. Permissible names are: @samp{vfp}, @samp{vfpv3},
13218 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
13219 @samp{vfpv3xd-fp16}, @samp{neon}, @samp{neon-fp16}, @samp{vfpv4},
13220 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
13221 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
13222 @samp{fp-armv8}, @samp{neon-fp-armv8}, and @samp{crypto-neon-fp-armv8}.
13224 If @option{-msoft-float} is specified this specifies the format of
13225 floating-point values.
13227 If the selected floating-point hardware includes the NEON extension
13228 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
13229 operations are not generated by GCC's auto-vectorization pass unless
13230 @option{-funsafe-math-optimizations} is also specified. This is
13231 because NEON hardware does not fully implement the IEEE 754 standard for
13232 floating-point arithmetic (in particular denormal values are treated as
13233 zero), so the use of NEON instructions may lead to a loss of precision.
13235 @item -mfp16-format=@var{name}
13236 @opindex mfp16-format
13237 Specify the format of the @code{__fp16} half-precision floating-point type.
13238 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
13239 the default is @samp{none}, in which case the @code{__fp16} type is not
13240 defined. @xref{Half-Precision}, for more information.
13242 @item -mstructure-size-boundary=@var{n}
13243 @opindex mstructure-size-boundary
13244 The sizes of all structures and unions are rounded up to a multiple
13245 of the number of bits set by this option. Permissible values are 8, 32
13246 and 64. The default value varies for different toolchains. For the COFF
13247 targeted toolchain the default value is 8. A value of 64 is only allowed
13248 if the underlying ABI supports it.
13250 Specifying a larger number can produce faster, more efficient code, but
13251 can also increase the size of the program. Different values are potentially
13252 incompatible. Code compiled with one value cannot necessarily expect to
13253 work with code or libraries compiled with another value, if they exchange
13254 information using structures or unions.
13256 @item -mabort-on-noreturn
13257 @opindex mabort-on-noreturn
13258 Generate a call to the function @code{abort} at the end of a
13259 @code{noreturn} function. It is executed if the function tries to
13263 @itemx -mno-long-calls
13264 @opindex mlong-calls
13265 @opindex mno-long-calls
13266 Tells the compiler to perform function calls by first loading the
13267 address of the function into a register and then performing a subroutine
13268 call on this register. This switch is needed if the target function
13269 lies outside of the 64-megabyte addressing range of the offset-based
13270 version of subroutine call instruction.
13272 Even if this switch is enabled, not all function calls are turned
13273 into long calls. The heuristic is that static functions, functions
13274 that have the @code{short_call} attribute, functions that are inside
13275 the scope of a @code{#pragma no_long_calls} directive, and functions whose
13276 definitions have already been compiled within the current compilation
13277 unit are not turned into long calls. The exceptions to this rule are
13278 that weak function definitions, functions with the @code{long_call}
13279 attribute or the @code{section} attribute, and functions that are within
13280 the scope of a @code{#pragma long_calls} directive are always
13281 turned into long calls.
13283 This feature is not enabled by default. Specifying
13284 @option{-mno-long-calls} restores the default behavior, as does
13285 placing the function calls within the scope of a @code{#pragma
13286 long_calls_off} directive. Note these switches have no effect on how
13287 the compiler generates code to handle function calls via function
13290 @item -msingle-pic-base
13291 @opindex msingle-pic-base
13292 Treat the register used for PIC addressing as read-only, rather than
13293 loading it in the prologue for each function. The runtime system is
13294 responsible for initializing this register with an appropriate value
13295 before execution begins.
13297 @item -mpic-register=@var{reg}
13298 @opindex mpic-register
13299 Specify the register to be used for PIC addressing.
13300 For standard PIC base case, the default is any suitable register
13301 determined by compiler. For single PIC base case, the default is
13302 @samp{R9} if target is EABI based or stack-checking is enabled,
13303 otherwise the default is @samp{R10}.
13305 @item -mpic-data-is-text-relative
13306 @opindex mpic-data-is-text-relative
13307 Assume that each data segments are relative to text segment at load time.
13308 Therefore, it permits addressing data using PC-relative operations.
13309 This option is on by default for targets other than VxWorks RTP.
13311 @item -mpoke-function-name
13312 @opindex mpoke-function-name
13313 Write the name of each function into the text section, directly
13314 preceding the function prologue. The generated code is similar to this:
13318 .ascii "arm_poke_function_name", 0
13321 .word 0xff000000 + (t1 - t0)
13322 arm_poke_function_name
13324 stmfd sp!, @{fp, ip, lr, pc@}
13328 When performing a stack backtrace, code can inspect the value of
13329 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
13330 location @code{pc - 12} and the top 8 bits are set, then we know that
13331 there is a function name embedded immediately preceding this location
13332 and has length @code{((pc[-3]) & 0xff000000)}.
13339 Select between generating code that executes in ARM and Thumb
13340 states. The default for most configurations is to generate code
13341 that executes in ARM state, but the default can be changed by
13342 configuring GCC with the @option{--with-mode=}@var{state}
13346 @opindex mtpcs-frame
13347 Generate a stack frame that is compliant with the Thumb Procedure Call
13348 Standard for all non-leaf functions. (A leaf function is one that does
13349 not call any other functions.) The default is @option{-mno-tpcs-frame}.
13351 @item -mtpcs-leaf-frame
13352 @opindex mtpcs-leaf-frame
13353 Generate a stack frame that is compliant with the Thumb Procedure Call
13354 Standard for all leaf functions. (A leaf function is one that does
13355 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
13357 @item -mcallee-super-interworking
13358 @opindex mcallee-super-interworking
13359 Gives all externally visible functions in the file being compiled an ARM
13360 instruction set header which switches to Thumb mode before executing the
13361 rest of the function. This allows these functions to be called from
13362 non-interworking code. This option is not valid in AAPCS configurations
13363 because interworking is enabled by default.
13365 @item -mcaller-super-interworking
13366 @opindex mcaller-super-interworking
13367 Allows calls via function pointers (including virtual functions) to
13368 execute correctly regardless of whether the target code has been
13369 compiled for interworking or not. There is a small overhead in the cost
13370 of executing a function pointer if this option is enabled. This option
13371 is not valid in AAPCS configurations because interworking is enabled
13374 @item -mtp=@var{name}
13376 Specify the access model for the thread local storage pointer. The valid
13377 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
13378 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
13379 (supported in the arm6k architecture), and @samp{auto}, which uses the
13380 best available method for the selected processor. The default setting is
13383 @item -mtls-dialect=@var{dialect}
13384 @opindex mtls-dialect
13385 Specify the dialect to use for accessing thread local storage. Two
13386 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
13387 @samp{gnu} dialect selects the original GNU scheme for supporting
13388 local and global dynamic TLS models. The @samp{gnu2} dialect
13389 selects the GNU descriptor scheme, which provides better performance
13390 for shared libraries. The GNU descriptor scheme is compatible with
13391 the original scheme, but does require new assembler, linker and
13392 library support. Initial and local exec TLS models are unaffected by
13393 this option and always use the original scheme.
13395 @item -mword-relocations
13396 @opindex mword-relocations
13397 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
13398 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
13399 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
13402 @item -mfix-cortex-m3-ldrd
13403 @opindex mfix-cortex-m3-ldrd
13404 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
13405 with overlapping destination and base registers are used. This option avoids
13406 generating these instructions. This option is enabled by default when
13407 @option{-mcpu=cortex-m3} is specified.
13409 @item -munaligned-access
13410 @itemx -mno-unaligned-access
13411 @opindex munaligned-access
13412 @opindex mno-unaligned-access
13413 Enables (or disables) reading and writing of 16- and 32- bit values
13414 from addresses that are not 16- or 32- bit aligned. By default
13415 unaligned access is disabled for all pre-ARMv6 and all ARMv6-M
13416 architectures, and enabled for all other architectures. If unaligned
13417 access is not enabled then words in packed data structures are
13418 accessed a byte at a time.
13420 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
13421 generated object file to either true or false, depending upon the
13422 setting of this option. If unaligned access is enabled then the
13423 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
13426 @item -mneon-for-64bits
13427 @opindex mneon-for-64bits
13428 Enables using Neon to handle scalar 64-bits operations. This is
13429 disabled by default since the cost of moving data from core registers
13432 @item -mslow-flash-data
13433 @opindex mslow-flash-data
13434 Assume loading data from flash is slower than fetching instruction.
13435 Therefore literal load is minimized for better performance.
13436 This option is only supported when compiling for ARMv7 M-profile and
13439 @item -masm-syntax-unified
13440 @opindex masm-syntax-unified
13441 Assume inline assembler is using unified asm syntax. The default is
13442 currently off which implies divided syntax. Currently this option is
13443 available only for Thumb1 and has no effect on ARM state and Thumb2.
13444 However, this may change in future releases of GCC. Divided syntax
13445 should be considered deprecated.
13447 @item -mrestrict-it
13448 @opindex mrestrict-it
13449 Restricts generation of IT blocks to conform to the rules of ARMv8.
13450 IT blocks can only contain a single 16-bit instruction from a select
13451 set of instructions. This option is on by default for ARMv8 Thumb mode.
13455 @subsection AVR Options
13456 @cindex AVR Options
13458 These options are defined for AVR implementations:
13461 @item -mmcu=@var{mcu}
13463 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
13465 The default for this option is@tie{}@samp{avr2}.
13467 GCC supports the following AVR devices and ISAs:
13469 @include avr-mmcu.texi
13471 @item -maccumulate-args
13472 @opindex maccumulate-args
13473 Accumulate outgoing function arguments and acquire/release the needed
13474 stack space for outgoing function arguments once in function
13475 prologue/epilogue. Without this option, outgoing arguments are pushed
13476 before calling a function and popped afterwards.
13478 Popping the arguments after the function call can be expensive on
13479 AVR so that accumulating the stack space might lead to smaller
13480 executables because arguments need not to be removed from the
13481 stack after such a function call.
13483 This option can lead to reduced code size for functions that perform
13484 several calls to functions that get their arguments on the stack like
13485 calls to printf-like functions.
13487 @item -mbranch-cost=@var{cost}
13488 @opindex mbranch-cost
13489 Set the branch costs for conditional branch instructions to
13490 @var{cost}. Reasonable values for @var{cost} are small, non-negative
13491 integers. The default branch cost is 0.
13493 @item -mcall-prologues
13494 @opindex mcall-prologues
13495 Functions prologues/epilogues are expanded as calls to appropriate
13496 subroutines. Code size is smaller.
13500 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
13501 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
13502 and @code{long long} is 4 bytes. Please note that this option does not
13503 conform to the C standards, but it results in smaller code
13506 @item -mn-flash=@var{num}
13508 Assume that the flash memory has a size of
13509 @var{num} times 64@tie{}KiB.
13511 @item -mno-interrupts
13512 @opindex mno-interrupts
13513 Generated code is not compatible with hardware interrupts.
13514 Code size is smaller.
13518 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
13519 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
13520 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
13521 the assembler's command line and the @option{--relax} option to the
13522 linker's command line.
13524 Jump relaxing is performed by the linker because jump offsets are not
13525 known before code is located. Therefore, the assembler code generated by the
13526 compiler is the same, but the instructions in the executable may
13527 differ from instructions in the assembler code.
13529 Relaxing must be turned on if linker stubs are needed, see the
13530 section on @code{EIND} and linker stubs below.
13534 Assume that the device supports the Read-Modify-Write
13535 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
13539 Treat the stack pointer register as an 8-bit register,
13540 i.e.@: assume the high byte of the stack pointer is zero.
13541 In general, you don't need to set this option by hand.
13543 This option is used internally by the compiler to select and
13544 build multilibs for architectures @code{avr2} and @code{avr25}.
13545 These architectures mix devices with and without @code{SPH}.
13546 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
13547 the compiler driver adds or removes this option from the compiler
13548 proper's command line, because the compiler then knows if the device
13549 or architecture has an 8-bit stack pointer and thus no @code{SPH}
13554 Use address register @code{X} in a way proposed by the hardware. This means
13555 that @code{X} is only used in indirect, post-increment or
13556 pre-decrement addressing.
13558 Without this option, the @code{X} register may be used in the same way
13559 as @code{Y} or @code{Z} which then is emulated by additional
13561 For example, loading a value with @code{X+const} addressing with a
13562 small non-negative @code{const < 64} to a register @var{Rn} is
13566 adiw r26, const ; X += const
13567 ld @var{Rn}, X ; @var{Rn} = *X
13568 sbiw r26, const ; X -= const
13572 @opindex mtiny-stack
13573 Only change the lower 8@tie{}bits of the stack pointer.
13575 @item -Waddr-space-convert
13576 @opindex Waddr-space-convert
13577 Warn about conversions between address spaces in the case where the
13578 resulting address space is not contained in the incoming address space.
13581 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
13582 @cindex @code{EIND}
13583 Pointers in the implementation are 16@tie{}bits wide.
13584 The address of a function or label is represented as word address so
13585 that indirect jumps and calls can target any code address in the
13586 range of 64@tie{}Ki words.
13588 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
13589 bytes of program memory space, there is a special function register called
13590 @code{EIND} that serves as most significant part of the target address
13591 when @code{EICALL} or @code{EIJMP} instructions are used.
13593 Indirect jumps and calls on these devices are handled as follows by
13594 the compiler and are subject to some limitations:
13599 The compiler never sets @code{EIND}.
13602 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
13603 instructions or might read @code{EIND} directly in order to emulate an
13604 indirect call/jump by means of a @code{RET} instruction.
13607 The compiler assumes that @code{EIND} never changes during the startup
13608 code or during the application. In particular, @code{EIND} is not
13609 saved/restored in function or interrupt service routine
13613 For indirect calls to functions and computed goto, the linker
13614 generates @emph{stubs}. Stubs are jump pads sometimes also called
13615 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
13616 The stub contains a direct jump to the desired address.
13619 Linker relaxation must be turned on so that the linker generates
13620 the stubs correctly in all situations. See the compiler option
13621 @option{-mrelax} and the linker option @option{--relax}.
13622 There are corner cases where the linker is supposed to generate stubs
13623 but aborts without relaxation and without a helpful error message.
13626 The default linker script is arranged for code with @code{EIND = 0}.
13627 If code is supposed to work for a setup with @code{EIND != 0}, a custom
13628 linker script has to be used in order to place the sections whose
13629 name start with @code{.trampolines} into the segment where @code{EIND}
13633 The startup code from libgcc never sets @code{EIND}.
13634 Notice that startup code is a blend of code from libgcc and AVR-LibC.
13635 For the impact of AVR-LibC on @code{EIND}, see the
13636 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
13639 It is legitimate for user-specific startup code to set up @code{EIND}
13640 early, for example by means of initialization code located in
13641 section @code{.init3}. Such code runs prior to general startup code
13642 that initializes RAM and calls constructors, but after the bit
13643 of startup code from AVR-LibC that sets @code{EIND} to the segment
13644 where the vector table is located.
13646 #include <avr/io.h>
13649 __attribute__((section(".init3"),naked,used,no_instrument_function))
13650 init3_set_eind (void)
13652 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
13653 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
13658 The @code{__trampolines_start} symbol is defined in the linker script.
13661 Stubs are generated automatically by the linker if
13662 the following two conditions are met:
13665 @item The address of a label is taken by means of the @code{gs} modifier
13666 (short for @emph{generate stubs}) like so:
13668 LDI r24, lo8(gs(@var{func}))
13669 LDI r25, hi8(gs(@var{func}))
13671 @item The final location of that label is in a code segment
13672 @emph{outside} the segment where the stubs are located.
13676 The compiler emits such @code{gs} modifiers for code labels in the
13677 following situations:
13679 @item Taking address of a function or code label.
13680 @item Computed goto.
13681 @item If prologue-save function is used, see @option{-mcall-prologues}
13682 command-line option.
13683 @item Switch/case dispatch tables. If you do not want such dispatch
13684 tables you can specify the @option{-fno-jump-tables} command-line option.
13685 @item C and C++ constructors/destructors called during startup/shutdown.
13686 @item If the tools hit a @code{gs()} modifier explained above.
13690 Jumping to non-symbolic addresses like so is @emph{not} supported:
13695 /* Call function at word address 0x2 */
13696 return ((int(*)(void)) 0x2)();
13700 Instead, a stub has to be set up, i.e.@: the function has to be called
13701 through a symbol (@code{func_4} in the example):
13706 extern int func_4 (void);
13708 /* Call function at byte address 0x4 */
13713 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
13714 Alternatively, @code{func_4} can be defined in the linker script.
13717 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
13718 @cindex @code{RAMPD}
13719 @cindex @code{RAMPX}
13720 @cindex @code{RAMPY}
13721 @cindex @code{RAMPZ}
13722 Some AVR devices support memories larger than the 64@tie{}KiB range
13723 that can be accessed with 16-bit pointers. To access memory locations
13724 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
13725 register is used as high part of the address:
13726 The @code{X}, @code{Y}, @code{Z} address register is concatenated
13727 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
13728 register, respectively, to get a wide address. Similarly,
13729 @code{RAMPD} is used together with direct addressing.
13733 The startup code initializes the @code{RAMP} special function
13734 registers with zero.
13737 If a @ref{AVR Named Address Spaces,named address space} other than
13738 generic or @code{__flash} is used, then @code{RAMPZ} is set
13739 as needed before the operation.
13742 If the device supports RAM larger than 64@tie{}KiB and the compiler
13743 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
13744 is reset to zero after the operation.
13747 If the device comes with a specific @code{RAMP} register, the ISR
13748 prologue/epilogue saves/restores that SFR and initializes it with
13749 zero in case the ISR code might (implicitly) use it.
13752 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
13753 If you use inline assembler to read from locations outside the
13754 16-bit address range and change one of the @code{RAMP} registers,
13755 you must reset it to zero after the access.
13759 @subsubsection AVR Built-in Macros
13761 GCC defines several built-in macros so that the user code can test
13762 for the presence or absence of features. Almost any of the following
13763 built-in macros are deduced from device capabilities and thus
13764 triggered by the @option{-mmcu=} command-line option.
13766 For even more AVR-specific built-in macros see
13767 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
13772 Build-in macro that resolves to a decimal number that identifies the
13773 architecture and depends on the @option{-mmcu=@var{mcu}} option.
13774 Possible values are:
13776 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
13777 @code{4}, @code{5}, @code{51}, @code{6}
13779 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
13780 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
13784 @code{100}, @code{102}, @code{104},
13785 @code{105}, @code{106}, @code{107}
13787 for @var{mcu}=@code{avrtiny}, @code{avrxmega2}, @code{avrxmega4},
13788 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
13789 If @var{mcu} specifies a device, this built-in macro is set
13790 accordingly. For example, with @option{-mmcu=atmega8} the macro is
13791 defined to @code{4}.
13793 @item __AVR_@var{Device}__
13794 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
13795 the device's name. For example, @option{-mmcu=atmega8} defines the
13796 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
13797 @code{__AVR_ATtiny261A__}, etc.
13799 The built-in macros' names follow
13800 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
13801 the device name as from the AVR user manual. The difference between
13802 @var{Device} in the built-in macro and @var{device} in
13803 @option{-mmcu=@var{device}} is that the latter is always lowercase.
13805 If @var{device} is not a device but only a core architecture like
13806 @samp{avr51}, this macro is not defined.
13808 @item __AVR_DEVICE_NAME__
13809 Setting @option{-mmcu=@var{device}} defines this built-in macro to
13810 the device's name. For example, with @option{-mmcu=atmega8} the macro
13811 is defined to @code{atmega8}.
13813 If @var{device} is not a device but only a core architecture like
13814 @samp{avr51}, this macro is not defined.
13816 @item __AVR_XMEGA__
13817 The device / architecture belongs to the XMEGA family of devices.
13819 @item __AVR_HAVE_ELPM__
13820 The device has the the @code{ELPM} instruction.
13822 @item __AVR_HAVE_ELPMX__
13823 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
13824 R@var{n},Z+} instructions.
13826 @item __AVR_HAVE_MOVW__
13827 The device has the @code{MOVW} instruction to perform 16-bit
13828 register-register moves.
13830 @item __AVR_HAVE_LPMX__
13831 The device has the @code{LPM R@var{n},Z} and
13832 @code{LPM R@var{n},Z+} instructions.
13834 @item __AVR_HAVE_MUL__
13835 The device has a hardware multiplier.
13837 @item __AVR_HAVE_JMP_CALL__
13838 The device has the @code{JMP} and @code{CALL} instructions.
13839 This is the case for devices with at least 16@tie{}KiB of program
13842 @item __AVR_HAVE_EIJMP_EICALL__
13843 @itemx __AVR_3_BYTE_PC__
13844 The device has the @code{EIJMP} and @code{EICALL} instructions.
13845 This is the case for devices with more than 128@tie{}KiB of program memory.
13846 This also means that the program counter
13847 (PC) is 3@tie{}bytes wide.
13849 @item __AVR_2_BYTE_PC__
13850 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
13851 with up to 128@tie{}KiB of program memory.
13853 @item __AVR_HAVE_8BIT_SP__
13854 @itemx __AVR_HAVE_16BIT_SP__
13855 The stack pointer (SP) register is treated as 8-bit respectively
13856 16-bit register by the compiler.
13857 The definition of these macros is affected by @option{-mtiny-stack}.
13859 @item __AVR_HAVE_SPH__
13861 The device has the SPH (high part of stack pointer) special function
13862 register or has an 8-bit stack pointer, respectively.
13863 The definition of these macros is affected by @option{-mmcu=} and
13864 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
13867 @item __AVR_HAVE_RAMPD__
13868 @itemx __AVR_HAVE_RAMPX__
13869 @itemx __AVR_HAVE_RAMPY__
13870 @itemx __AVR_HAVE_RAMPZ__
13871 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
13872 @code{RAMPZ} special function register, respectively.
13874 @item __NO_INTERRUPTS__
13875 This macro reflects the @option{-mno-interrupts} command line option.
13877 @item __AVR_ERRATA_SKIP__
13878 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
13879 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
13880 instructions because of a hardware erratum. Skip instructions are
13881 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
13882 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
13885 @item __AVR_ISA_RMW__
13886 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
13888 @item __AVR_SFR_OFFSET__=@var{offset}
13889 Instructions that can address I/O special function registers directly
13890 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
13891 address as if addressed by an instruction to access RAM like @code{LD}
13892 or @code{STS}. This offset depends on the device architecture and has
13893 to be subtracted from the RAM address in order to get the
13894 respective I/O@tie{}address.
13896 @item __WITH_AVRLIBC__
13897 The compiler is configured to be used together with AVR-Libc.
13898 See the @option{--with-avrlibc} configure option.
13902 @node Blackfin Options
13903 @subsection Blackfin Options
13904 @cindex Blackfin Options
13907 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
13909 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
13910 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
13911 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
13912 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
13913 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
13914 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
13915 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
13916 @samp{bf561}, @samp{bf592}.
13918 The optional @var{sirevision} specifies the silicon revision of the target
13919 Blackfin processor. Any workarounds available for the targeted silicon revision
13920 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
13921 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
13922 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
13923 hexadecimal digits representing the major and minor numbers in the silicon
13924 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
13925 is not defined. If @var{sirevision} is @samp{any}, the
13926 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
13927 If this optional @var{sirevision} is not used, GCC assumes the latest known
13928 silicon revision of the targeted Blackfin processor.
13930 GCC defines a preprocessor macro for the specified @var{cpu}.
13931 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
13932 provided by libgloss to be linked in if @option{-msim} is not given.
13934 Without this option, @samp{bf532} is used as the processor by default.
13936 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
13937 only the preprocessor macro is defined.
13941 Specifies that the program will be run on the simulator. This causes
13942 the simulator BSP provided by libgloss to be linked in. This option
13943 has effect only for @samp{bfin-elf} toolchain.
13944 Certain other options, such as @option{-mid-shared-library} and
13945 @option{-mfdpic}, imply @option{-msim}.
13947 @item -momit-leaf-frame-pointer
13948 @opindex momit-leaf-frame-pointer
13949 Don't keep the frame pointer in a register for leaf functions. This
13950 avoids the instructions to save, set up and restore frame pointers and
13951 makes an extra register available in leaf functions. The option
13952 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
13953 which might make debugging harder.
13955 @item -mspecld-anomaly
13956 @opindex mspecld-anomaly
13957 When enabled, the compiler ensures that the generated code does not
13958 contain speculative loads after jump instructions. If this option is used,
13959 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
13961 @item -mno-specld-anomaly
13962 @opindex mno-specld-anomaly
13963 Don't generate extra code to prevent speculative loads from occurring.
13965 @item -mcsync-anomaly
13966 @opindex mcsync-anomaly
13967 When enabled, the compiler ensures that the generated code does not
13968 contain CSYNC or SSYNC instructions too soon after conditional branches.
13969 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
13971 @item -mno-csync-anomaly
13972 @opindex mno-csync-anomaly
13973 Don't generate extra code to prevent CSYNC or SSYNC instructions from
13974 occurring too soon after a conditional branch.
13978 When enabled, the compiler is free to take advantage of the knowledge that
13979 the entire program fits into the low 64k of memory.
13982 @opindex mno-low-64k
13983 Assume that the program is arbitrarily large. This is the default.
13985 @item -mstack-check-l1
13986 @opindex mstack-check-l1
13987 Do stack checking using information placed into L1 scratchpad memory by the
13990 @item -mid-shared-library
13991 @opindex mid-shared-library
13992 Generate code that supports shared libraries via the library ID method.
13993 This allows for execute in place and shared libraries in an environment
13994 without virtual memory management. This option implies @option{-fPIC}.
13995 With a @samp{bfin-elf} target, this option implies @option{-msim}.
13997 @item -mno-id-shared-library
13998 @opindex mno-id-shared-library
13999 Generate code that doesn't assume ID-based shared libraries are being used.
14000 This is the default.
14002 @item -mleaf-id-shared-library
14003 @opindex mleaf-id-shared-library
14004 Generate code that supports shared libraries via the library ID method,
14005 but assumes that this library or executable won't link against any other
14006 ID shared libraries. That allows the compiler to use faster code for jumps
14009 @item -mno-leaf-id-shared-library
14010 @opindex mno-leaf-id-shared-library
14011 Do not assume that the code being compiled won't link against any ID shared
14012 libraries. Slower code is generated for jump and call insns.
14014 @item -mshared-library-id=n
14015 @opindex mshared-library-id
14016 Specifies the identification number of the ID-based shared library being
14017 compiled. Specifying a value of 0 generates more compact code; specifying
14018 other values forces the allocation of that number to the current
14019 library but is no more space- or time-efficient than omitting this option.
14023 Generate code that allows the data segment to be located in a different
14024 area of memory from the text segment. This allows for execute in place in
14025 an environment without virtual memory management by eliminating relocations
14026 against the text section.
14028 @item -mno-sep-data
14029 @opindex mno-sep-data
14030 Generate code that assumes that the data segment follows the text segment.
14031 This is the default.
14034 @itemx -mno-long-calls
14035 @opindex mlong-calls
14036 @opindex mno-long-calls
14037 Tells the compiler to perform function calls by first loading the
14038 address of the function into a register and then performing a subroutine
14039 call on this register. This switch is needed if the target function
14040 lies outside of the 24-bit addressing range of the offset-based
14041 version of subroutine call instruction.
14043 This feature is not enabled by default. Specifying
14044 @option{-mno-long-calls} restores the default behavior. Note these
14045 switches have no effect on how the compiler generates code to handle
14046 function calls via function pointers.
14050 Link with the fast floating-point library. This library relaxes some of
14051 the IEEE floating-point standard's rules for checking inputs against
14052 Not-a-Number (NAN), in the interest of performance.
14055 @opindex minline-plt
14056 Enable inlining of PLT entries in function calls to functions that are
14057 not known to bind locally. It has no effect without @option{-mfdpic}.
14060 @opindex mmulticore
14061 Build a standalone application for multicore Blackfin processors.
14062 This option causes proper start files and link scripts supporting
14063 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
14064 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
14066 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
14067 selects the one-application-per-core programming model. Without
14068 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
14069 programming model is used. In this model, the main function of Core B
14070 should be named as @code{coreb_main}.
14072 If this option is not used, the single-core application programming
14077 Build a standalone application for Core A of BF561 when using
14078 the one-application-per-core programming model. Proper start files
14079 and link scripts are used to support Core A, and the macro
14080 @code{__BFIN_COREA} is defined.
14081 This option can only be used in conjunction with @option{-mmulticore}.
14085 Build a standalone application for Core B of BF561 when using
14086 the one-application-per-core programming model. Proper start files
14087 and link scripts are used to support Core B, and the macro
14088 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
14089 should be used instead of @code{main}.
14090 This option can only be used in conjunction with @option{-mmulticore}.
14094 Build a standalone application for SDRAM. Proper start files and
14095 link scripts are used to put the application into SDRAM, and the macro
14096 @code{__BFIN_SDRAM} is defined.
14097 The loader should initialize SDRAM before loading the application.
14101 Assume that ICPLBs are enabled at run time. This has an effect on certain
14102 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
14103 are enabled; for standalone applications the default is off.
14107 @subsection C6X Options
14108 @cindex C6X Options
14111 @item -march=@var{name}
14113 This specifies the name of the target architecture. GCC uses this
14114 name to determine what kind of instructions it can emit when generating
14115 assembly code. Permissible names are: @samp{c62x},
14116 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
14119 @opindex mbig-endian
14120 Generate code for a big-endian target.
14122 @item -mlittle-endian
14123 @opindex mlittle-endian
14124 Generate code for a little-endian target. This is the default.
14128 Choose startup files and linker script suitable for the simulator.
14130 @item -msdata=default
14131 @opindex msdata=default
14132 Put small global and static data in the @code{.neardata} section,
14133 which is pointed to by register @code{B14}. Put small uninitialized
14134 global and static data in the @code{.bss} section, which is adjacent
14135 to the @code{.neardata} section. Put small read-only data into the
14136 @code{.rodata} section. The corresponding sections used for large
14137 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
14140 @opindex msdata=all
14141 Put all data, not just small objects, into the sections reserved for
14142 small data, and use addressing relative to the @code{B14} register to
14146 @opindex msdata=none
14147 Make no use of the sections reserved for small data, and use absolute
14148 addresses to access all data. Put all initialized global and static
14149 data in the @code{.fardata} section, and all uninitialized data in the
14150 @code{.far} section. Put all constant data into the @code{.const}
14155 @subsection CRIS Options
14156 @cindex CRIS Options
14158 These options are defined specifically for the CRIS ports.
14161 @item -march=@var{architecture-type}
14162 @itemx -mcpu=@var{architecture-type}
14165 Generate code for the specified architecture. The choices for
14166 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
14167 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
14168 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
14171 @item -mtune=@var{architecture-type}
14173 Tune to @var{architecture-type} everything applicable about the generated
14174 code, except for the ABI and the set of available instructions. The
14175 choices for @var{architecture-type} are the same as for
14176 @option{-march=@var{architecture-type}}.
14178 @item -mmax-stack-frame=@var{n}
14179 @opindex mmax-stack-frame
14180 Warn when the stack frame of a function exceeds @var{n} bytes.
14186 The options @option{-metrax4} and @option{-metrax100} are synonyms for
14187 @option{-march=v3} and @option{-march=v8} respectively.
14189 @item -mmul-bug-workaround
14190 @itemx -mno-mul-bug-workaround
14191 @opindex mmul-bug-workaround
14192 @opindex mno-mul-bug-workaround
14193 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
14194 models where it applies. This option is active by default.
14198 Enable CRIS-specific verbose debug-related information in the assembly
14199 code. This option also has the effect of turning off the @samp{#NO_APP}
14200 formatted-code indicator to the assembler at the beginning of the
14205 Do not use condition-code results from previous instruction; always emit
14206 compare and test instructions before use of condition codes.
14208 @item -mno-side-effects
14209 @opindex mno-side-effects
14210 Do not emit instructions with side effects in addressing modes other than
14213 @item -mstack-align
14214 @itemx -mno-stack-align
14215 @itemx -mdata-align
14216 @itemx -mno-data-align
14217 @itemx -mconst-align
14218 @itemx -mno-const-align
14219 @opindex mstack-align
14220 @opindex mno-stack-align
14221 @opindex mdata-align
14222 @opindex mno-data-align
14223 @opindex mconst-align
14224 @opindex mno-const-align
14225 These options (@samp{no-} options) arrange (eliminate arrangements) for the
14226 stack frame, individual data and constants to be aligned for the maximum
14227 single data access size for the chosen CPU model. The default is to
14228 arrange for 32-bit alignment. ABI details such as structure layout are
14229 not affected by these options.
14237 Similar to the stack- data- and const-align options above, these options
14238 arrange for stack frame, writable data and constants to all be 32-bit,
14239 16-bit or 8-bit aligned. The default is 32-bit alignment.
14241 @item -mno-prologue-epilogue
14242 @itemx -mprologue-epilogue
14243 @opindex mno-prologue-epilogue
14244 @opindex mprologue-epilogue
14245 With @option{-mno-prologue-epilogue}, the normal function prologue and
14246 epilogue which set up the stack frame are omitted and no return
14247 instructions or return sequences are generated in the code. Use this
14248 option only together with visual inspection of the compiled code: no
14249 warnings or errors are generated when call-saved registers must be saved,
14250 or storage for local variables needs to be allocated.
14254 @opindex mno-gotplt
14256 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
14257 instruction sequences that load addresses for functions from the PLT part
14258 of the GOT rather than (traditional on other architectures) calls to the
14259 PLT@. The default is @option{-mgotplt}.
14263 Legacy no-op option only recognized with the cris-axis-elf and
14264 cris-axis-linux-gnu targets.
14268 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
14272 This option, recognized for the cris-axis-elf, arranges
14273 to link with input-output functions from a simulator library. Code,
14274 initialized data and zero-initialized data are allocated consecutively.
14278 Like @option{-sim}, but pass linker options to locate initialized data at
14279 0x40000000 and zero-initialized data at 0x80000000.
14283 @subsection CR16 Options
14284 @cindex CR16 Options
14286 These options are defined specifically for the CR16 ports.
14292 Enable the use of multiply-accumulate instructions. Disabled by default.
14296 @opindex mcr16cplus
14298 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
14303 Links the library libsim.a which is in compatible with simulator. Applicable
14304 to ELF compiler only.
14308 Choose integer type as 32-bit wide.
14312 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
14314 @item -mdata-model=@var{model}
14315 @opindex mdata-model
14316 Choose a data model. The choices for @var{model} are @samp{near},
14317 @samp{far} or @samp{medium}. @samp{medium} is default.
14318 However, @samp{far} is not valid with @option{-mcr16c}, as the
14319 CR16C architecture does not support the far data model.
14322 @node Darwin Options
14323 @subsection Darwin Options
14324 @cindex Darwin options
14326 These options are defined for all architectures running the Darwin operating
14329 FSF GCC on Darwin does not create ``fat'' object files; it creates
14330 an object file for the single architecture that GCC was built to
14331 target. Apple's GCC on Darwin does create ``fat'' files if multiple
14332 @option{-arch} options are used; it does so by running the compiler or
14333 linker multiple times and joining the results together with
14336 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
14337 @samp{i686}) is determined by the flags that specify the ISA
14338 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
14339 @option{-force_cpusubtype_ALL} option can be used to override this.
14341 The Darwin tools vary in their behavior when presented with an ISA
14342 mismatch. The assembler, @file{as}, only permits instructions to
14343 be used that are valid for the subtype of the file it is generating,
14344 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
14345 The linker for shared libraries, @file{/usr/bin/libtool}, fails
14346 and prints an error if asked to create a shared library with a less
14347 restrictive subtype than its input files (for instance, trying to put
14348 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
14349 for executables, @command{ld}, quietly gives the executable the most
14350 restrictive subtype of any of its input files.
14355 Add the framework directory @var{dir} to the head of the list of
14356 directories to be searched for header files. These directories are
14357 interleaved with those specified by @option{-I} options and are
14358 scanned in a left-to-right order.
14360 A framework directory is a directory with frameworks in it. A
14361 framework is a directory with a @file{Headers} and/or
14362 @file{PrivateHeaders} directory contained directly in it that ends
14363 in @file{.framework}. The name of a framework is the name of this
14364 directory excluding the @file{.framework}. Headers associated with
14365 the framework are found in one of those two directories, with
14366 @file{Headers} being searched first. A subframework is a framework
14367 directory that is in a framework's @file{Frameworks} directory.
14368 Includes of subframework headers can only appear in a header of a
14369 framework that contains the subframework, or in a sibling subframework
14370 header. Two subframeworks are siblings if they occur in the same
14371 framework. A subframework should not have the same name as a
14372 framework; a warning is issued if this is violated. Currently a
14373 subframework cannot have subframeworks; in the future, the mechanism
14374 may be extended to support this. The standard frameworks can be found
14375 in @file{/System/Library/Frameworks} and
14376 @file{/Library/Frameworks}. An example include looks like
14377 @code{#include <Framework/header.h>}, where @file{Framework} denotes
14378 the name of the framework and @file{header.h} is found in the
14379 @file{PrivateHeaders} or @file{Headers} directory.
14381 @item -iframework@var{dir}
14382 @opindex iframework
14383 Like @option{-F} except the directory is a treated as a system
14384 directory. The main difference between this @option{-iframework} and
14385 @option{-F} is that with @option{-iframework} the compiler does not
14386 warn about constructs contained within header files found via
14387 @var{dir}. This option is valid only for the C family of languages.
14391 Emit debugging information for symbols that are used. For stabs
14392 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
14393 This is by default ON@.
14397 Emit debugging information for all symbols and types.
14399 @item -mmacosx-version-min=@var{version}
14400 The earliest version of MacOS X that this executable will run on
14401 is @var{version}. Typical values of @var{version} include @code{10.1},
14402 @code{10.2}, and @code{10.3.9}.
14404 If the compiler was built to use the system's headers by default,
14405 then the default for this option is the system version on which the
14406 compiler is running, otherwise the default is to make choices that
14407 are compatible with as many systems and code bases as possible.
14411 Enable kernel development mode. The @option{-mkernel} option sets
14412 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
14413 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
14414 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
14415 applicable. This mode also sets @option{-mno-altivec},
14416 @option{-msoft-float}, @option{-fno-builtin} and
14417 @option{-mlong-branch} for PowerPC targets.
14419 @item -mone-byte-bool
14420 @opindex mone-byte-bool
14421 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
14422 By default @code{sizeof(bool)} is @code{4} when compiling for
14423 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
14424 option has no effect on x86.
14426 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
14427 to generate code that is not binary compatible with code generated
14428 without that switch. Using this switch may require recompiling all
14429 other modules in a program, including system libraries. Use this
14430 switch to conform to a non-default data model.
14432 @item -mfix-and-continue
14433 @itemx -ffix-and-continue
14434 @itemx -findirect-data
14435 @opindex mfix-and-continue
14436 @opindex ffix-and-continue
14437 @opindex findirect-data
14438 Generate code suitable for fast turnaround development, such as to
14439 allow GDB to dynamically load @file{.o} files into already-running
14440 programs. @option{-findirect-data} and @option{-ffix-and-continue}
14441 are provided for backwards compatibility.
14445 Loads all members of static archive libraries.
14446 See man ld(1) for more information.
14448 @item -arch_errors_fatal
14449 @opindex arch_errors_fatal
14450 Cause the errors having to do with files that have the wrong architecture
14453 @item -bind_at_load
14454 @opindex bind_at_load
14455 Causes the output file to be marked such that the dynamic linker will
14456 bind all undefined references when the file is loaded or launched.
14460 Produce a Mach-o bundle format file.
14461 See man ld(1) for more information.
14463 @item -bundle_loader @var{executable}
14464 @opindex bundle_loader
14465 This option specifies the @var{executable} that will load the build
14466 output file being linked. See man ld(1) for more information.
14469 @opindex dynamiclib
14470 When passed this option, GCC produces a dynamic library instead of
14471 an executable when linking, using the Darwin @file{libtool} command.
14473 @item -force_cpusubtype_ALL
14474 @opindex force_cpusubtype_ALL
14475 This causes GCC's output file to have the @samp{ALL} subtype, instead of
14476 one controlled by the @option{-mcpu} or @option{-march} option.
14478 @item -allowable_client @var{client_name}
14479 @itemx -client_name
14480 @itemx -compatibility_version
14481 @itemx -current_version
14483 @itemx -dependency-file
14485 @itemx -dylinker_install_name
14487 @itemx -exported_symbols_list
14490 @itemx -flat_namespace
14491 @itemx -force_flat_namespace
14492 @itemx -headerpad_max_install_names
14495 @itemx -install_name
14496 @itemx -keep_private_externs
14497 @itemx -multi_module
14498 @itemx -multiply_defined
14499 @itemx -multiply_defined_unused
14502 @itemx -no_dead_strip_inits_and_terms
14503 @itemx -nofixprebinding
14504 @itemx -nomultidefs
14506 @itemx -noseglinkedit
14507 @itemx -pagezero_size
14509 @itemx -prebind_all_twolevel_modules
14510 @itemx -private_bundle
14512 @itemx -read_only_relocs
14514 @itemx -sectobjectsymbols
14518 @itemx -sectobjectsymbols
14521 @itemx -segs_read_only_addr
14523 @itemx -segs_read_write_addr
14524 @itemx -seg_addr_table
14525 @itemx -seg_addr_table_filename
14526 @itemx -seglinkedit
14528 @itemx -segs_read_only_addr
14529 @itemx -segs_read_write_addr
14530 @itemx -single_module
14532 @itemx -sub_library
14534 @itemx -sub_umbrella
14535 @itemx -twolevel_namespace
14538 @itemx -unexported_symbols_list
14539 @itemx -weak_reference_mismatches
14540 @itemx -whatsloaded
14541 @opindex allowable_client
14542 @opindex client_name
14543 @opindex compatibility_version
14544 @opindex current_version
14545 @opindex dead_strip
14546 @opindex dependency-file
14547 @opindex dylib_file
14548 @opindex dylinker_install_name
14550 @opindex exported_symbols_list
14552 @opindex flat_namespace
14553 @opindex force_flat_namespace
14554 @opindex headerpad_max_install_names
14555 @opindex image_base
14557 @opindex install_name
14558 @opindex keep_private_externs
14559 @opindex multi_module
14560 @opindex multiply_defined
14561 @opindex multiply_defined_unused
14562 @opindex noall_load
14563 @opindex no_dead_strip_inits_and_terms
14564 @opindex nofixprebinding
14565 @opindex nomultidefs
14567 @opindex noseglinkedit
14568 @opindex pagezero_size
14570 @opindex prebind_all_twolevel_modules
14571 @opindex private_bundle
14572 @opindex read_only_relocs
14574 @opindex sectobjectsymbols
14577 @opindex sectcreate
14578 @opindex sectobjectsymbols
14581 @opindex segs_read_only_addr
14582 @opindex segs_read_write_addr
14583 @opindex seg_addr_table
14584 @opindex seg_addr_table_filename
14585 @opindex seglinkedit
14587 @opindex segs_read_only_addr
14588 @opindex segs_read_write_addr
14589 @opindex single_module
14591 @opindex sub_library
14592 @opindex sub_umbrella
14593 @opindex twolevel_namespace
14596 @opindex unexported_symbols_list
14597 @opindex weak_reference_mismatches
14598 @opindex whatsloaded
14599 These options are passed to the Darwin linker. The Darwin linker man page
14600 describes them in detail.
14603 @node DEC Alpha Options
14604 @subsection DEC Alpha Options
14606 These @samp{-m} options are defined for the DEC Alpha implementations:
14609 @item -mno-soft-float
14610 @itemx -msoft-float
14611 @opindex mno-soft-float
14612 @opindex msoft-float
14613 Use (do not use) the hardware floating-point instructions for
14614 floating-point operations. When @option{-msoft-float} is specified,
14615 functions in @file{libgcc.a} are used to perform floating-point
14616 operations. Unless they are replaced by routines that emulate the
14617 floating-point operations, or compiled in such a way as to call such
14618 emulations routines, these routines issue floating-point
14619 operations. If you are compiling for an Alpha without floating-point
14620 operations, you must ensure that the library is built so as not to call
14623 Note that Alpha implementations without floating-point operations are
14624 required to have floating-point registers.
14627 @itemx -mno-fp-regs
14629 @opindex mno-fp-regs
14630 Generate code that uses (does not use) the floating-point register set.
14631 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
14632 register set is not used, floating-point operands are passed in integer
14633 registers as if they were integers and floating-point results are passed
14634 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
14635 so any function with a floating-point argument or return value called by code
14636 compiled with @option{-mno-fp-regs} must also be compiled with that
14639 A typical use of this option is building a kernel that does not use,
14640 and hence need not save and restore, any floating-point registers.
14644 The Alpha architecture implements floating-point hardware optimized for
14645 maximum performance. It is mostly compliant with the IEEE floating-point
14646 standard. However, for full compliance, software assistance is
14647 required. This option generates code fully IEEE-compliant code
14648 @emph{except} that the @var{inexact-flag} is not maintained (see below).
14649 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
14650 defined during compilation. The resulting code is less efficient but is
14651 able to correctly support denormalized numbers and exceptional IEEE
14652 values such as not-a-number and plus/minus infinity. Other Alpha
14653 compilers call this option @option{-ieee_with_no_inexact}.
14655 @item -mieee-with-inexact
14656 @opindex mieee-with-inexact
14657 This is like @option{-mieee} except the generated code also maintains
14658 the IEEE @var{inexact-flag}. Turning on this option causes the
14659 generated code to implement fully-compliant IEEE math. In addition to
14660 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
14661 macro. On some Alpha implementations the resulting code may execute
14662 significantly slower than the code generated by default. Since there is
14663 very little code that depends on the @var{inexact-flag}, you should
14664 normally not specify this option. Other Alpha compilers call this
14665 option @option{-ieee_with_inexact}.
14667 @item -mfp-trap-mode=@var{trap-mode}
14668 @opindex mfp-trap-mode
14669 This option controls what floating-point related traps are enabled.
14670 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
14671 The trap mode can be set to one of four values:
14675 This is the default (normal) setting. The only traps that are enabled
14676 are the ones that cannot be disabled in software (e.g., division by zero
14680 In addition to the traps enabled by @samp{n}, underflow traps are enabled
14684 Like @samp{u}, but the instructions are marked to be safe for software
14685 completion (see Alpha architecture manual for details).
14688 Like @samp{su}, but inexact traps are enabled as well.
14691 @item -mfp-rounding-mode=@var{rounding-mode}
14692 @opindex mfp-rounding-mode
14693 Selects the IEEE rounding mode. Other Alpha compilers call this option
14694 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
14699 Normal IEEE rounding mode. Floating-point numbers are rounded towards
14700 the nearest machine number or towards the even machine number in case
14704 Round towards minus infinity.
14707 Chopped rounding mode. Floating-point numbers are rounded towards zero.
14710 Dynamic rounding mode. A field in the floating-point control register
14711 (@var{fpcr}, see Alpha architecture reference manual) controls the
14712 rounding mode in effect. The C library initializes this register for
14713 rounding towards plus infinity. Thus, unless your program modifies the
14714 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
14717 @item -mtrap-precision=@var{trap-precision}
14718 @opindex mtrap-precision
14719 In the Alpha architecture, floating-point traps are imprecise. This
14720 means without software assistance it is impossible to recover from a
14721 floating trap and program execution normally needs to be terminated.
14722 GCC can generate code that can assist operating system trap handlers
14723 in determining the exact location that caused a floating-point trap.
14724 Depending on the requirements of an application, different levels of
14725 precisions can be selected:
14729 Program precision. This option is the default and means a trap handler
14730 can only identify which program caused a floating-point exception.
14733 Function precision. The trap handler can determine the function that
14734 caused a floating-point exception.
14737 Instruction precision. The trap handler can determine the exact
14738 instruction that caused a floating-point exception.
14741 Other Alpha compilers provide the equivalent options called
14742 @option{-scope_safe} and @option{-resumption_safe}.
14744 @item -mieee-conformant
14745 @opindex mieee-conformant
14746 This option marks the generated code as IEEE conformant. You must not
14747 use this option unless you also specify @option{-mtrap-precision=i} and either
14748 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
14749 is to emit the line @samp{.eflag 48} in the function prologue of the
14750 generated assembly file.
14752 @item -mbuild-constants
14753 @opindex mbuild-constants
14754 Normally GCC examines a 32- or 64-bit integer constant to
14755 see if it can construct it from smaller constants in two or three
14756 instructions. If it cannot, it outputs the constant as a literal and
14757 generates code to load it from the data segment at run time.
14759 Use this option to require GCC to construct @emph{all} integer constants
14760 using code, even if it takes more instructions (the maximum is six).
14762 You typically use this option to build a shared library dynamic
14763 loader. Itself a shared library, it must relocate itself in memory
14764 before it can find the variables and constants in its own data segment.
14782 Indicate whether GCC should generate code to use the optional BWX,
14783 CIX, FIX and MAX instruction sets. The default is to use the instruction
14784 sets supported by the CPU type specified via @option{-mcpu=} option or that
14785 of the CPU on which GCC was built if none is specified.
14788 @itemx -mfloat-ieee
14789 @opindex mfloat-vax
14790 @opindex mfloat-ieee
14791 Generate code that uses (does not use) VAX F and G floating-point
14792 arithmetic instead of IEEE single and double precision.
14794 @item -mexplicit-relocs
14795 @itemx -mno-explicit-relocs
14796 @opindex mexplicit-relocs
14797 @opindex mno-explicit-relocs
14798 Older Alpha assemblers provided no way to generate symbol relocations
14799 except via assembler macros. Use of these macros does not allow
14800 optimal instruction scheduling. GNU binutils as of version 2.12
14801 supports a new syntax that allows the compiler to explicitly mark
14802 which relocations should apply to which instructions. This option
14803 is mostly useful for debugging, as GCC detects the capabilities of
14804 the assembler when it is built and sets the default accordingly.
14807 @itemx -mlarge-data
14808 @opindex msmall-data
14809 @opindex mlarge-data
14810 When @option{-mexplicit-relocs} is in effect, static data is
14811 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
14812 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
14813 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
14814 16-bit relocations off of the @code{$gp} register. This limits the
14815 size of the small data area to 64KB, but allows the variables to be
14816 directly accessed via a single instruction.
14818 The default is @option{-mlarge-data}. With this option the data area
14819 is limited to just below 2GB@. Programs that require more than 2GB of
14820 data must use @code{malloc} or @code{mmap} to allocate the data in the
14821 heap instead of in the program's data segment.
14823 When generating code for shared libraries, @option{-fpic} implies
14824 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
14827 @itemx -mlarge-text
14828 @opindex msmall-text
14829 @opindex mlarge-text
14830 When @option{-msmall-text} is used, the compiler assumes that the
14831 code of the entire program (or shared library) fits in 4MB, and is
14832 thus reachable with a branch instruction. When @option{-msmall-data}
14833 is used, the compiler can assume that all local symbols share the
14834 same @code{$gp} value, and thus reduce the number of instructions
14835 required for a function call from 4 to 1.
14837 The default is @option{-mlarge-text}.
14839 @item -mcpu=@var{cpu_type}
14841 Set the instruction set and instruction scheduling parameters for
14842 machine type @var{cpu_type}. You can specify either the @samp{EV}
14843 style name or the corresponding chip number. GCC supports scheduling
14844 parameters for the EV4, EV5 and EV6 family of processors and
14845 chooses the default values for the instruction set from the processor
14846 you specify. If you do not specify a processor type, GCC defaults
14847 to the processor on which the compiler was built.
14849 Supported values for @var{cpu_type} are
14855 Schedules as an EV4 and has no instruction set extensions.
14859 Schedules as an EV5 and has no instruction set extensions.
14863 Schedules as an EV5 and supports the BWX extension.
14868 Schedules as an EV5 and supports the BWX and MAX extensions.
14872 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
14876 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
14879 Native toolchains also support the value @samp{native},
14880 which selects the best architecture option for the host processor.
14881 @option{-mcpu=native} has no effect if GCC does not recognize
14884 @item -mtune=@var{cpu_type}
14886 Set only the instruction scheduling parameters for machine type
14887 @var{cpu_type}. The instruction set is not changed.
14889 Native toolchains also support the value @samp{native},
14890 which selects the best architecture option for the host processor.
14891 @option{-mtune=native} has no effect if GCC does not recognize
14894 @item -mmemory-latency=@var{time}
14895 @opindex mmemory-latency
14896 Sets the latency the scheduler should assume for typical memory
14897 references as seen by the application. This number is highly
14898 dependent on the memory access patterns used by the application
14899 and the size of the external cache on the machine.
14901 Valid options for @var{time} are
14905 A decimal number representing clock cycles.
14911 The compiler contains estimates of the number of clock cycles for
14912 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
14913 (also called Dcache, Scache, and Bcache), as well as to main memory.
14914 Note that L3 is only valid for EV5.
14920 @subsection FR30 Options
14921 @cindex FR30 Options
14923 These options are defined specifically for the FR30 port.
14927 @item -msmall-model
14928 @opindex msmall-model
14929 Use the small address space model. This can produce smaller code, but
14930 it does assume that all symbolic values and addresses fit into a
14935 Assume that runtime support has been provided and so there is no need
14936 to include the simulator library (@file{libsim.a}) on the linker
14942 @subsection FRV Options
14943 @cindex FRV Options
14949 Only use the first 32 general-purpose registers.
14954 Use all 64 general-purpose registers.
14959 Use only the first 32 floating-point registers.
14964 Use all 64 floating-point registers.
14967 @opindex mhard-float
14969 Use hardware instructions for floating-point operations.
14972 @opindex msoft-float
14974 Use library routines for floating-point operations.
14979 Dynamically allocate condition code registers.
14984 Do not try to dynamically allocate condition code registers, only
14985 use @code{icc0} and @code{fcc0}.
14990 Change ABI to use double word insns.
14995 Do not use double word instructions.
15000 Use floating-point double instructions.
15003 @opindex mno-double
15005 Do not use floating-point double instructions.
15010 Use media instructions.
15015 Do not use media instructions.
15020 Use multiply and add/subtract instructions.
15023 @opindex mno-muladd
15025 Do not use multiply and add/subtract instructions.
15030 Select the FDPIC ABI, which uses function descriptors to represent
15031 pointers to functions. Without any PIC/PIE-related options, it
15032 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
15033 assumes GOT entries and small data are within a 12-bit range from the
15034 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
15035 are computed with 32 bits.
15036 With a @samp{bfin-elf} target, this option implies @option{-msim}.
15039 @opindex minline-plt
15041 Enable inlining of PLT entries in function calls to functions that are
15042 not known to bind locally. It has no effect without @option{-mfdpic}.
15043 It's enabled by default if optimizing for speed and compiling for
15044 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
15045 optimization option such as @option{-O3} or above is present in the
15051 Assume a large TLS segment when generating thread-local code.
15056 Do not assume a large TLS segment when generating thread-local code.
15061 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
15062 that is known to be in read-only sections. It's enabled by default,
15063 except for @option{-fpic} or @option{-fpie}: even though it may help
15064 make the global offset table smaller, it trades 1 instruction for 4.
15065 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
15066 one of which may be shared by multiple symbols, and it avoids the need
15067 for a GOT entry for the referenced symbol, so it's more likely to be a
15068 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
15070 @item -multilib-library-pic
15071 @opindex multilib-library-pic
15073 Link with the (library, not FD) pic libraries. It's implied by
15074 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
15075 @option{-fpic} without @option{-mfdpic}. You should never have to use
15079 @opindex mlinked-fp
15081 Follow the EABI requirement of always creating a frame pointer whenever
15082 a stack frame is allocated. This option is enabled by default and can
15083 be disabled with @option{-mno-linked-fp}.
15086 @opindex mlong-calls
15088 Use indirect addressing to call functions outside the current
15089 compilation unit. This allows the functions to be placed anywhere
15090 within the 32-bit address space.
15092 @item -malign-labels
15093 @opindex malign-labels
15095 Try to align labels to an 8-byte boundary by inserting NOPs into the
15096 previous packet. This option only has an effect when VLIW packing
15097 is enabled. It doesn't create new packets; it merely adds NOPs to
15100 @item -mlibrary-pic
15101 @opindex mlibrary-pic
15103 Generate position-independent EABI code.
15108 Use only the first four media accumulator registers.
15113 Use all eight media accumulator registers.
15118 Pack VLIW instructions.
15123 Do not pack VLIW instructions.
15126 @opindex mno-eflags
15128 Do not mark ABI switches in e_flags.
15131 @opindex mcond-move
15133 Enable the use of conditional-move instructions (default).
15135 This switch is mainly for debugging the compiler and will likely be removed
15136 in a future version.
15138 @item -mno-cond-move
15139 @opindex mno-cond-move
15141 Disable the use of conditional-move instructions.
15143 This switch is mainly for debugging the compiler and will likely be removed
15144 in a future version.
15149 Enable the use of conditional set instructions (default).
15151 This switch is mainly for debugging the compiler and will likely be removed
15152 in a future version.
15157 Disable the use of conditional set instructions.
15159 This switch is mainly for debugging the compiler and will likely be removed
15160 in a future version.
15163 @opindex mcond-exec
15165 Enable the use of conditional execution (default).
15167 This switch is mainly for debugging the compiler and will likely be removed
15168 in a future version.
15170 @item -mno-cond-exec
15171 @opindex mno-cond-exec
15173 Disable the use of conditional execution.
15175 This switch is mainly for debugging the compiler and will likely be removed
15176 in a future version.
15178 @item -mvliw-branch
15179 @opindex mvliw-branch
15181 Run a pass to pack branches into VLIW instructions (default).
15183 This switch is mainly for debugging the compiler and will likely be removed
15184 in a future version.
15186 @item -mno-vliw-branch
15187 @opindex mno-vliw-branch
15189 Do not run a pass to pack branches into VLIW instructions.
15191 This switch is mainly for debugging the compiler and will likely be removed
15192 in a future version.
15194 @item -mmulti-cond-exec
15195 @opindex mmulti-cond-exec
15197 Enable optimization of @code{&&} and @code{||} in conditional execution
15200 This switch is mainly for debugging the compiler and will likely be removed
15201 in a future version.
15203 @item -mno-multi-cond-exec
15204 @opindex mno-multi-cond-exec
15206 Disable optimization of @code{&&} and @code{||} in conditional execution.
15208 This switch is mainly for debugging the compiler and will likely be removed
15209 in a future version.
15211 @item -mnested-cond-exec
15212 @opindex mnested-cond-exec
15214 Enable nested conditional execution optimizations (default).
15216 This switch is mainly for debugging the compiler and will likely be removed
15217 in a future version.
15219 @item -mno-nested-cond-exec
15220 @opindex mno-nested-cond-exec
15222 Disable nested conditional execution optimizations.
15224 This switch is mainly for debugging the compiler and will likely be removed
15225 in a future version.
15227 @item -moptimize-membar
15228 @opindex moptimize-membar
15230 This switch removes redundant @code{membar} instructions from the
15231 compiler-generated code. It is enabled by default.
15233 @item -mno-optimize-membar
15234 @opindex mno-optimize-membar
15236 This switch disables the automatic removal of redundant @code{membar}
15237 instructions from the generated code.
15239 @item -mtomcat-stats
15240 @opindex mtomcat-stats
15242 Cause gas to print out tomcat statistics.
15244 @item -mcpu=@var{cpu}
15247 Select the processor type for which to generate code. Possible values are
15248 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
15249 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
15253 @node GNU/Linux Options
15254 @subsection GNU/Linux Options
15256 These @samp{-m} options are defined for GNU/Linux targets:
15261 Use the GNU C library. This is the default except
15262 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
15266 Use uClibc C library. This is the default on
15267 @samp{*-*-linux-*uclibc*} targets.
15271 Use Bionic C library. This is the default on
15272 @samp{*-*-linux-*android*} targets.
15276 Compile code compatible with Android platform. This is the default on
15277 @samp{*-*-linux-*android*} targets.
15279 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
15280 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
15281 this option makes the GCC driver pass Android-specific options to the linker.
15282 Finally, this option causes the preprocessor macro @code{__ANDROID__}
15285 @item -tno-android-cc
15286 @opindex tno-android-cc
15287 Disable compilation effects of @option{-mandroid}, i.e., do not enable
15288 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
15289 @option{-fno-rtti} by default.
15291 @item -tno-android-ld
15292 @opindex tno-android-ld
15293 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
15294 linking options to the linker.
15298 @node H8/300 Options
15299 @subsection H8/300 Options
15301 These @samp{-m} options are defined for the H8/300 implementations:
15306 Shorten some address references at link time, when possible; uses the
15307 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
15308 ld, Using ld}, for a fuller description.
15312 Generate code for the H8/300H@.
15316 Generate code for the H8S@.
15320 Generate code for the H8S and H8/300H in the normal mode. This switch
15321 must be used either with @option{-mh} or @option{-ms}.
15325 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
15329 Extended registers are stored on stack before execution of function
15330 with monitor attribute. Default option is @option{-mexr}.
15331 This option is valid only for H8S targets.
15335 Extended registers are not stored on stack before execution of function
15336 with monitor attribute. Default option is @option{-mno-exr}.
15337 This option is valid only for H8S targets.
15341 Make @code{int} data 32 bits by default.
15344 @opindex malign-300
15345 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
15346 The default for the H8/300H and H8S is to align longs and floats on
15348 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
15349 This option has no effect on the H8/300.
15353 @subsection HPPA Options
15354 @cindex HPPA Options
15356 These @samp{-m} options are defined for the HPPA family of computers:
15359 @item -march=@var{architecture-type}
15361 Generate code for the specified architecture. The choices for
15362 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
15363 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
15364 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
15365 architecture option for your machine. Code compiled for lower numbered
15366 architectures runs on higher numbered architectures, but not the
15369 @item -mpa-risc-1-0
15370 @itemx -mpa-risc-1-1
15371 @itemx -mpa-risc-2-0
15372 @opindex mpa-risc-1-0
15373 @opindex mpa-risc-1-1
15374 @opindex mpa-risc-2-0
15375 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
15377 @item -mjump-in-delay
15378 @opindex mjump-in-delay
15379 This option is ignored and provided for compatibility purposes only.
15381 @item -mdisable-fpregs
15382 @opindex mdisable-fpregs
15383 Prevent floating-point registers from being used in any manner. This is
15384 necessary for compiling kernels that perform lazy context switching of
15385 floating-point registers. If you use this option and attempt to perform
15386 floating-point operations, the compiler aborts.
15388 @item -mdisable-indexing
15389 @opindex mdisable-indexing
15390 Prevent the compiler from using indexing address modes. This avoids some
15391 rather obscure problems when compiling MIG generated code under MACH@.
15393 @item -mno-space-regs
15394 @opindex mno-space-regs
15395 Generate code that assumes the target has no space registers. This allows
15396 GCC to generate faster indirect calls and use unscaled index address modes.
15398 Such code is suitable for level 0 PA systems and kernels.
15400 @item -mfast-indirect-calls
15401 @opindex mfast-indirect-calls
15402 Generate code that assumes calls never cross space boundaries. This
15403 allows GCC to emit code that performs faster indirect calls.
15405 This option does not work in the presence of shared libraries or nested
15408 @item -mfixed-range=@var{register-range}
15409 @opindex mfixed-range
15410 Generate code treating the given register range as fixed registers.
15411 A fixed register is one that the register allocator cannot use. This is
15412 useful when compiling kernel code. A register range is specified as
15413 two registers separated by a dash. Multiple register ranges can be
15414 specified separated by a comma.
15416 @item -mlong-load-store
15417 @opindex mlong-load-store
15418 Generate 3-instruction load and store sequences as sometimes required by
15419 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
15422 @item -mportable-runtime
15423 @opindex mportable-runtime
15424 Use the portable calling conventions proposed by HP for ELF systems.
15428 Enable the use of assembler directives only GAS understands.
15430 @item -mschedule=@var{cpu-type}
15432 Schedule code according to the constraints for the machine type
15433 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
15434 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
15435 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
15436 proper scheduling option for your machine. The default scheduling is
15440 @opindex mlinker-opt
15441 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
15442 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
15443 linkers in which they give bogus error messages when linking some programs.
15446 @opindex msoft-float
15447 Generate output containing library calls for floating point.
15448 @strong{Warning:} the requisite libraries are not available for all HPPA
15449 targets. Normally the facilities of the machine's usual C compiler are
15450 used, but this cannot be done directly in cross-compilation. You must make
15451 your own arrangements to provide suitable library functions for
15454 @option{-msoft-float} changes the calling convention in the output file;
15455 therefore, it is only useful if you compile @emph{all} of a program with
15456 this option. In particular, you need to compile @file{libgcc.a}, the
15457 library that comes with GCC, with @option{-msoft-float} in order for
15462 Generate the predefine, @code{_SIO}, for server IO@. The default is
15463 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
15464 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
15465 options are available under HP-UX and HI-UX@.
15469 Use options specific to GNU @command{ld}.
15470 This passes @option{-shared} to @command{ld} when
15471 building a shared library. It is the default when GCC is configured,
15472 explicitly or implicitly, with the GNU linker. This option does not
15473 affect which @command{ld} is called; it only changes what parameters
15474 are passed to that @command{ld}.
15475 The @command{ld} that is called is determined by the
15476 @option{--with-ld} configure option, GCC's program search path, and
15477 finally by the user's @env{PATH}. The linker used by GCC can be printed
15478 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
15479 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15483 Use options specific to HP @command{ld}.
15484 This passes @option{-b} to @command{ld} when building
15485 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
15486 links. It is the default when GCC is configured, explicitly or
15487 implicitly, with the HP linker. This option does not affect
15488 which @command{ld} is called; it only changes what parameters are passed to that
15490 The @command{ld} that is called is determined by the @option{--with-ld}
15491 configure option, GCC's program search path, and finally by the user's
15492 @env{PATH}. The linker used by GCC can be printed using @samp{which
15493 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
15494 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15497 @opindex mno-long-calls
15498 Generate code that uses long call sequences. This ensures that a call
15499 is always able to reach linker generated stubs. The default is to generate
15500 long calls only when the distance from the call site to the beginning
15501 of the function or translation unit, as the case may be, exceeds a
15502 predefined limit set by the branch type being used. The limits for
15503 normal calls are 7,600,000 and 240,000 bytes, respectively for the
15504 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
15507 Distances are measured from the beginning of functions when using the
15508 @option{-ffunction-sections} option, or when using the @option{-mgas}
15509 and @option{-mno-portable-runtime} options together under HP-UX with
15512 It is normally not desirable to use this option as it degrades
15513 performance. However, it may be useful in large applications,
15514 particularly when partial linking is used to build the application.
15516 The types of long calls used depends on the capabilities of the
15517 assembler and linker, and the type of code being generated. The
15518 impact on systems that support long absolute calls, and long pic
15519 symbol-difference or pc-relative calls should be relatively small.
15520 However, an indirect call is used on 32-bit ELF systems in pic code
15521 and it is quite long.
15523 @item -munix=@var{unix-std}
15525 Generate compiler predefines and select a startfile for the specified
15526 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
15527 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
15528 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
15529 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
15530 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
15533 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
15534 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
15535 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
15536 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
15537 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
15538 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
15540 It is @emph{important} to note that this option changes the interfaces
15541 for various library routines. It also affects the operational behavior
15542 of the C library. Thus, @emph{extreme} care is needed in using this
15545 Library code that is intended to operate with more than one UNIX
15546 standard must test, set and restore the variable @code{__xpg4_extended_mask}
15547 as appropriate. Most GNU software doesn't provide this capability.
15551 Suppress the generation of link options to search libdld.sl when the
15552 @option{-static} option is specified on HP-UX 10 and later.
15556 The HP-UX implementation of setlocale in libc has a dependency on
15557 libdld.sl. There isn't an archive version of libdld.sl. Thus,
15558 when the @option{-static} option is specified, special link options
15559 are needed to resolve this dependency.
15561 On HP-UX 10 and later, the GCC driver adds the necessary options to
15562 link with libdld.sl when the @option{-static} option is specified.
15563 This causes the resulting binary to be dynamic. On the 64-bit port,
15564 the linkers generate dynamic binaries by default in any case. The
15565 @option{-nolibdld} option can be used to prevent the GCC driver from
15566 adding these link options.
15570 Add support for multithreading with the @dfn{dce thread} library
15571 under HP-UX@. This option sets flags for both the preprocessor and
15575 @node IA-64 Options
15576 @subsection IA-64 Options
15577 @cindex IA-64 Options
15579 These are the @samp{-m} options defined for the Intel IA-64 architecture.
15583 @opindex mbig-endian
15584 Generate code for a big-endian target. This is the default for HP-UX@.
15586 @item -mlittle-endian
15587 @opindex mlittle-endian
15588 Generate code for a little-endian target. This is the default for AIX5
15594 @opindex mno-gnu-as
15595 Generate (or don't) code for the GNU assembler. This is the default.
15596 @c Also, this is the default if the configure option @option{--with-gnu-as}
15602 @opindex mno-gnu-ld
15603 Generate (or don't) code for the GNU linker. This is the default.
15604 @c Also, this is the default if the configure option @option{--with-gnu-ld}
15609 Generate code that does not use a global pointer register. The result
15610 is not position independent code, and violates the IA-64 ABI@.
15612 @item -mvolatile-asm-stop
15613 @itemx -mno-volatile-asm-stop
15614 @opindex mvolatile-asm-stop
15615 @opindex mno-volatile-asm-stop
15616 Generate (or don't) a stop bit immediately before and after volatile asm
15619 @item -mregister-names
15620 @itemx -mno-register-names
15621 @opindex mregister-names
15622 @opindex mno-register-names
15623 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
15624 the stacked registers. This may make assembler output more readable.
15630 Disable (or enable) optimizations that use the small data section. This may
15631 be useful for working around optimizer bugs.
15633 @item -mconstant-gp
15634 @opindex mconstant-gp
15635 Generate code that uses a single constant global pointer value. This is
15636 useful when compiling kernel code.
15640 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
15641 This is useful when compiling firmware code.
15643 @item -minline-float-divide-min-latency
15644 @opindex minline-float-divide-min-latency
15645 Generate code for inline divides of floating-point values
15646 using the minimum latency algorithm.
15648 @item -minline-float-divide-max-throughput
15649 @opindex minline-float-divide-max-throughput
15650 Generate code for inline divides of floating-point values
15651 using the maximum throughput algorithm.
15653 @item -mno-inline-float-divide
15654 @opindex mno-inline-float-divide
15655 Do not generate inline code for divides of floating-point values.
15657 @item -minline-int-divide-min-latency
15658 @opindex minline-int-divide-min-latency
15659 Generate code for inline divides of integer values
15660 using the minimum latency algorithm.
15662 @item -minline-int-divide-max-throughput
15663 @opindex minline-int-divide-max-throughput
15664 Generate code for inline divides of integer values
15665 using the maximum throughput algorithm.
15667 @item -mno-inline-int-divide
15668 @opindex mno-inline-int-divide
15669 Do not generate inline code for divides of integer values.
15671 @item -minline-sqrt-min-latency
15672 @opindex minline-sqrt-min-latency
15673 Generate code for inline square roots
15674 using the minimum latency algorithm.
15676 @item -minline-sqrt-max-throughput
15677 @opindex minline-sqrt-max-throughput
15678 Generate code for inline square roots
15679 using the maximum throughput algorithm.
15681 @item -mno-inline-sqrt
15682 @opindex mno-inline-sqrt
15683 Do not generate inline code for @code{sqrt}.
15686 @itemx -mno-fused-madd
15687 @opindex mfused-madd
15688 @opindex mno-fused-madd
15689 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
15690 instructions. The default is to use these instructions.
15692 @item -mno-dwarf2-asm
15693 @itemx -mdwarf2-asm
15694 @opindex mno-dwarf2-asm
15695 @opindex mdwarf2-asm
15696 Don't (or do) generate assembler code for the DWARF 2 line number debugging
15697 info. This may be useful when not using the GNU assembler.
15699 @item -mearly-stop-bits
15700 @itemx -mno-early-stop-bits
15701 @opindex mearly-stop-bits
15702 @opindex mno-early-stop-bits
15703 Allow stop bits to be placed earlier than immediately preceding the
15704 instruction that triggered the stop bit. This can improve instruction
15705 scheduling, but does not always do so.
15707 @item -mfixed-range=@var{register-range}
15708 @opindex mfixed-range
15709 Generate code treating the given register range as fixed registers.
15710 A fixed register is one that the register allocator cannot use. This is
15711 useful when compiling kernel code. A register range is specified as
15712 two registers separated by a dash. Multiple register ranges can be
15713 specified separated by a comma.
15715 @item -mtls-size=@var{tls-size}
15717 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
15720 @item -mtune=@var{cpu-type}
15722 Tune the instruction scheduling for a particular CPU, Valid values are
15723 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
15724 and @samp{mckinley}.
15730 Generate code for a 32-bit or 64-bit environment.
15731 The 32-bit environment sets int, long and pointer to 32 bits.
15732 The 64-bit environment sets int to 32 bits and long and pointer
15733 to 64 bits. These are HP-UX specific flags.
15735 @item -mno-sched-br-data-spec
15736 @itemx -msched-br-data-spec
15737 @opindex mno-sched-br-data-spec
15738 @opindex msched-br-data-spec
15739 (Dis/En)able data speculative scheduling before reload.
15740 This results in generation of @code{ld.a} instructions and
15741 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
15742 The default is 'disable'.
15744 @item -msched-ar-data-spec
15745 @itemx -mno-sched-ar-data-spec
15746 @opindex msched-ar-data-spec
15747 @opindex mno-sched-ar-data-spec
15748 (En/Dis)able data speculative scheduling after reload.
15749 This results in generation of @code{ld.a} instructions and
15750 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
15751 The default is 'enable'.
15753 @item -mno-sched-control-spec
15754 @itemx -msched-control-spec
15755 @opindex mno-sched-control-spec
15756 @opindex msched-control-spec
15757 (Dis/En)able control speculative scheduling. This feature is
15758 available only during region scheduling (i.e.@: before reload).
15759 This results in generation of the @code{ld.s} instructions and
15760 the corresponding check instructions @code{chk.s}.
15761 The default is 'disable'.
15763 @item -msched-br-in-data-spec
15764 @itemx -mno-sched-br-in-data-spec
15765 @opindex msched-br-in-data-spec
15766 @opindex mno-sched-br-in-data-spec
15767 (En/Dis)able speculative scheduling of the instructions that
15768 are dependent on the data speculative loads before reload.
15769 This is effective only with @option{-msched-br-data-spec} enabled.
15770 The default is 'enable'.
15772 @item -msched-ar-in-data-spec
15773 @itemx -mno-sched-ar-in-data-spec
15774 @opindex msched-ar-in-data-spec
15775 @opindex mno-sched-ar-in-data-spec
15776 (En/Dis)able speculative scheduling of the instructions that
15777 are dependent on the data speculative loads after reload.
15778 This is effective only with @option{-msched-ar-data-spec} enabled.
15779 The default is 'enable'.
15781 @item -msched-in-control-spec
15782 @itemx -mno-sched-in-control-spec
15783 @opindex msched-in-control-spec
15784 @opindex mno-sched-in-control-spec
15785 (En/Dis)able speculative scheduling of the instructions that
15786 are dependent on the control speculative loads.
15787 This is effective only with @option{-msched-control-spec} enabled.
15788 The default is 'enable'.
15790 @item -mno-sched-prefer-non-data-spec-insns
15791 @itemx -msched-prefer-non-data-spec-insns
15792 @opindex mno-sched-prefer-non-data-spec-insns
15793 @opindex msched-prefer-non-data-spec-insns
15794 If enabled, data-speculative instructions are chosen for schedule
15795 only if there are no other choices at the moment. This makes
15796 the use of the data speculation much more conservative.
15797 The default is 'disable'.
15799 @item -mno-sched-prefer-non-control-spec-insns
15800 @itemx -msched-prefer-non-control-spec-insns
15801 @opindex mno-sched-prefer-non-control-spec-insns
15802 @opindex msched-prefer-non-control-spec-insns
15803 If enabled, control-speculative instructions are chosen for schedule
15804 only if there are no other choices at the moment. This makes
15805 the use of the control speculation much more conservative.
15806 The default is 'disable'.
15808 @item -mno-sched-count-spec-in-critical-path
15809 @itemx -msched-count-spec-in-critical-path
15810 @opindex mno-sched-count-spec-in-critical-path
15811 @opindex msched-count-spec-in-critical-path
15812 If enabled, speculative dependencies are considered during
15813 computation of the instructions priorities. This makes the use of the
15814 speculation a bit more conservative.
15815 The default is 'disable'.
15817 @item -msched-spec-ldc
15818 @opindex msched-spec-ldc
15819 Use a simple data speculation check. This option is on by default.
15821 @item -msched-control-spec-ldc
15822 @opindex msched-spec-ldc
15823 Use a simple check for control speculation. This option is on by default.
15825 @item -msched-stop-bits-after-every-cycle
15826 @opindex msched-stop-bits-after-every-cycle
15827 Place a stop bit after every cycle when scheduling. This option is on
15830 @item -msched-fp-mem-deps-zero-cost
15831 @opindex msched-fp-mem-deps-zero-cost
15832 Assume that floating-point stores and loads are not likely to cause a conflict
15833 when placed into the same instruction group. This option is disabled by
15836 @item -msel-sched-dont-check-control-spec
15837 @opindex msel-sched-dont-check-control-spec
15838 Generate checks for control speculation in selective scheduling.
15839 This flag is disabled by default.
15841 @item -msched-max-memory-insns=@var{max-insns}
15842 @opindex msched-max-memory-insns
15843 Limit on the number of memory insns per instruction group, giving lower
15844 priority to subsequent memory insns attempting to schedule in the same
15845 instruction group. Frequently useful to prevent cache bank conflicts.
15846 The default value is 1.
15848 @item -msched-max-memory-insns-hard-limit
15849 @opindex msched-max-memory-insns-hard-limit
15850 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
15851 disallowing more than that number in an instruction group.
15852 Otherwise, the limit is ``soft'', meaning that non-memory operations
15853 are preferred when the limit is reached, but memory operations may still
15859 @subsection LM32 Options
15860 @cindex LM32 options
15862 These @option{-m} options are defined for the LatticeMico32 architecture:
15865 @item -mbarrel-shift-enabled
15866 @opindex mbarrel-shift-enabled
15867 Enable barrel-shift instructions.
15869 @item -mdivide-enabled
15870 @opindex mdivide-enabled
15871 Enable divide and modulus instructions.
15873 @item -mmultiply-enabled
15874 @opindex multiply-enabled
15875 Enable multiply instructions.
15877 @item -msign-extend-enabled
15878 @opindex msign-extend-enabled
15879 Enable sign extend instructions.
15881 @item -muser-enabled
15882 @opindex muser-enabled
15883 Enable user-defined instructions.
15888 @subsection M32C Options
15889 @cindex M32C options
15892 @item -mcpu=@var{name}
15894 Select the CPU for which code is generated. @var{name} may be one of
15895 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
15896 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
15897 the M32C/80 series.
15901 Specifies that the program will be run on the simulator. This causes
15902 an alternate runtime library to be linked in which supports, for
15903 example, file I/O@. You must not use this option when generating
15904 programs that will run on real hardware; you must provide your own
15905 runtime library for whatever I/O functions are needed.
15907 @item -memregs=@var{number}
15909 Specifies the number of memory-based pseudo-registers GCC uses
15910 during code generation. These pseudo-registers are used like real
15911 registers, so there is a tradeoff between GCC's ability to fit the
15912 code into available registers, and the performance penalty of using
15913 memory instead of registers. Note that all modules in a program must
15914 be compiled with the same value for this option. Because of that, you
15915 must not use this option with GCC's default runtime libraries.
15919 @node M32R/D Options
15920 @subsection M32R/D Options
15921 @cindex M32R/D options
15923 These @option{-m} options are defined for Renesas M32R/D architectures:
15928 Generate code for the M32R/2@.
15932 Generate code for the M32R/X@.
15936 Generate code for the M32R@. This is the default.
15938 @item -mmodel=small
15939 @opindex mmodel=small
15940 Assume all objects live in the lower 16MB of memory (so that their addresses
15941 can be loaded with the @code{ld24} instruction), and assume all subroutines
15942 are reachable with the @code{bl} instruction.
15943 This is the default.
15945 The addressability of a particular object can be set with the
15946 @code{model} attribute.
15948 @item -mmodel=medium
15949 @opindex mmodel=medium
15950 Assume objects may be anywhere in the 32-bit address space (the compiler
15951 generates @code{seth/add3} instructions to load their addresses), and
15952 assume all subroutines are reachable with the @code{bl} instruction.
15954 @item -mmodel=large
15955 @opindex mmodel=large
15956 Assume objects may be anywhere in the 32-bit address space (the compiler
15957 generates @code{seth/add3} instructions to load their addresses), and
15958 assume subroutines may not be reachable with the @code{bl} instruction
15959 (the compiler generates the much slower @code{seth/add3/jl}
15960 instruction sequence).
15963 @opindex msdata=none
15964 Disable use of the small data area. Variables are put into
15965 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
15966 @code{section} attribute has been specified).
15967 This is the default.
15969 The small data area consists of sections @code{.sdata} and @code{.sbss}.
15970 Objects may be explicitly put in the small data area with the
15971 @code{section} attribute using one of these sections.
15973 @item -msdata=sdata
15974 @opindex msdata=sdata
15975 Put small global and static data in the small data area, but do not
15976 generate special code to reference them.
15979 @opindex msdata=use
15980 Put small global and static data in the small data area, and generate
15981 special instructions to reference them.
15985 @cindex smaller data references
15986 Put global and static objects less than or equal to @var{num} bytes
15987 into the small data or BSS sections instead of the normal data or BSS
15988 sections. The default value of @var{num} is 8.
15989 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
15990 for this option to have any effect.
15992 All modules should be compiled with the same @option{-G @var{num}} value.
15993 Compiling with different values of @var{num} may or may not work; if it
15994 doesn't the linker gives an error message---incorrect code is not
15999 Makes the M32R-specific code in the compiler display some statistics
16000 that might help in debugging programs.
16002 @item -malign-loops
16003 @opindex malign-loops
16004 Align all loops to a 32-byte boundary.
16006 @item -mno-align-loops
16007 @opindex mno-align-loops
16008 Do not enforce a 32-byte alignment for loops. This is the default.
16010 @item -missue-rate=@var{number}
16011 @opindex missue-rate=@var{number}
16012 Issue @var{number} instructions per cycle. @var{number} can only be 1
16015 @item -mbranch-cost=@var{number}
16016 @opindex mbranch-cost=@var{number}
16017 @var{number} can only be 1 or 2. If it is 1 then branches are
16018 preferred over conditional code, if it is 2, then the opposite applies.
16020 @item -mflush-trap=@var{number}
16021 @opindex mflush-trap=@var{number}
16022 Specifies the trap number to use to flush the cache. The default is
16023 12. Valid numbers are between 0 and 15 inclusive.
16025 @item -mno-flush-trap
16026 @opindex mno-flush-trap
16027 Specifies that the cache cannot be flushed by using a trap.
16029 @item -mflush-func=@var{name}
16030 @opindex mflush-func=@var{name}
16031 Specifies the name of the operating system function to call to flush
16032 the cache. The default is @samp{_flush_cache}, but a function call
16033 is only used if a trap is not available.
16035 @item -mno-flush-func
16036 @opindex mno-flush-func
16037 Indicates that there is no OS function for flushing the cache.
16041 @node M680x0 Options
16042 @subsection M680x0 Options
16043 @cindex M680x0 options
16045 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
16046 The default settings depend on which architecture was selected when
16047 the compiler was configured; the defaults for the most common choices
16051 @item -march=@var{arch}
16053 Generate code for a specific M680x0 or ColdFire instruction set
16054 architecture. Permissible values of @var{arch} for M680x0
16055 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
16056 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
16057 architectures are selected according to Freescale's ISA classification
16058 and the permissible values are: @samp{isaa}, @samp{isaaplus},
16059 @samp{isab} and @samp{isac}.
16061 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
16062 code for a ColdFire target. The @var{arch} in this macro is one of the
16063 @option{-march} arguments given above.
16065 When used together, @option{-march} and @option{-mtune} select code
16066 that runs on a family of similar processors but that is optimized
16067 for a particular microarchitecture.
16069 @item -mcpu=@var{cpu}
16071 Generate code for a specific M680x0 or ColdFire processor.
16072 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
16073 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
16074 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
16075 below, which also classifies the CPUs into families:
16077 @multitable @columnfractions 0.20 0.80
16078 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
16079 @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}
16080 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
16081 @item @samp{5206e} @tab @samp{5206e}
16082 @item @samp{5208} @tab @samp{5207} @samp{5208}
16083 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
16084 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
16085 @item @samp{5216} @tab @samp{5214} @samp{5216}
16086 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
16087 @item @samp{5225} @tab @samp{5224} @samp{5225}
16088 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
16089 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
16090 @item @samp{5249} @tab @samp{5249}
16091 @item @samp{5250} @tab @samp{5250}
16092 @item @samp{5271} @tab @samp{5270} @samp{5271}
16093 @item @samp{5272} @tab @samp{5272}
16094 @item @samp{5275} @tab @samp{5274} @samp{5275}
16095 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
16096 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
16097 @item @samp{5307} @tab @samp{5307}
16098 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
16099 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
16100 @item @samp{5407} @tab @samp{5407}
16101 @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}
16104 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
16105 @var{arch} is compatible with @var{cpu}. Other combinations of
16106 @option{-mcpu} and @option{-march} are rejected.
16108 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
16109 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
16110 where the value of @var{family} is given by the table above.
16112 @item -mtune=@var{tune}
16114 Tune the code for a particular microarchitecture within the
16115 constraints set by @option{-march} and @option{-mcpu}.
16116 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
16117 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
16118 and @samp{cpu32}. The ColdFire microarchitectures
16119 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
16121 You can also use @option{-mtune=68020-40} for code that needs
16122 to run relatively well on 68020, 68030 and 68040 targets.
16123 @option{-mtune=68020-60} is similar but includes 68060 targets
16124 as well. These two options select the same tuning decisions as
16125 @option{-m68020-40} and @option{-m68020-60} respectively.
16127 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
16128 when tuning for 680x0 architecture @var{arch}. It also defines
16129 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
16130 option is used. If GCC is tuning for a range of architectures,
16131 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
16132 it defines the macros for every architecture in the range.
16134 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
16135 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
16136 of the arguments given above.
16142 Generate output for a 68000. This is the default
16143 when the compiler is configured for 68000-based systems.
16144 It is equivalent to @option{-march=68000}.
16146 Use this option for microcontrollers with a 68000 or EC000 core,
16147 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
16151 Generate output for a 68010. This is the default
16152 when the compiler is configured for 68010-based systems.
16153 It is equivalent to @option{-march=68010}.
16159 Generate output for a 68020. This is the default
16160 when the compiler is configured for 68020-based systems.
16161 It is equivalent to @option{-march=68020}.
16165 Generate output for a 68030. This is the default when the compiler is
16166 configured for 68030-based systems. It is equivalent to
16167 @option{-march=68030}.
16171 Generate output for a 68040. This is the default when the compiler is
16172 configured for 68040-based systems. It is equivalent to
16173 @option{-march=68040}.
16175 This option inhibits the use of 68881/68882 instructions that have to be
16176 emulated by software on the 68040. Use this option if your 68040 does not
16177 have code to emulate those instructions.
16181 Generate output for a 68060. This is the default when the compiler is
16182 configured for 68060-based systems. It is equivalent to
16183 @option{-march=68060}.
16185 This option inhibits the use of 68020 and 68881/68882 instructions that
16186 have to be emulated by software on the 68060. Use this option if your 68060
16187 does not have code to emulate those instructions.
16191 Generate output for a CPU32. This is the default
16192 when the compiler is configured for CPU32-based systems.
16193 It is equivalent to @option{-march=cpu32}.
16195 Use this option for microcontrollers with a
16196 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
16197 68336, 68340, 68341, 68349 and 68360.
16201 Generate output for a 520X ColdFire CPU@. This is the default
16202 when the compiler is configured for 520X-based systems.
16203 It is equivalent to @option{-mcpu=5206}, and is now deprecated
16204 in favor of that option.
16206 Use this option for microcontroller with a 5200 core, including
16207 the MCF5202, MCF5203, MCF5204 and MCF5206.
16211 Generate output for a 5206e ColdFire CPU@. The option is now
16212 deprecated in favor of the equivalent @option{-mcpu=5206e}.
16216 Generate output for a member of the ColdFire 528X family.
16217 The option is now deprecated in favor of the equivalent
16218 @option{-mcpu=528x}.
16222 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
16223 in favor of the equivalent @option{-mcpu=5307}.
16227 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
16228 in favor of the equivalent @option{-mcpu=5407}.
16232 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
16233 This includes use of hardware floating-point instructions.
16234 The option is equivalent to @option{-mcpu=547x}, and is now
16235 deprecated in favor of that option.
16239 Generate output for a 68040, without using any of the new instructions.
16240 This results in code that can run relatively efficiently on either a
16241 68020/68881 or a 68030 or a 68040. The generated code does use the
16242 68881 instructions that are emulated on the 68040.
16244 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
16248 Generate output for a 68060, without using any of the new instructions.
16249 This results in code that can run relatively efficiently on either a
16250 68020/68881 or a 68030 or a 68040. The generated code does use the
16251 68881 instructions that are emulated on the 68060.
16253 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
16257 @opindex mhard-float
16259 Generate floating-point instructions. This is the default for 68020
16260 and above, and for ColdFire devices that have an FPU@. It defines the
16261 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
16262 on ColdFire targets.
16265 @opindex msoft-float
16266 Do not generate floating-point instructions; use library calls instead.
16267 This is the default for 68000, 68010, and 68832 targets. It is also
16268 the default for ColdFire devices that have no FPU.
16274 Generate (do not generate) ColdFire hardware divide and remainder
16275 instructions. If @option{-march} is used without @option{-mcpu},
16276 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
16277 architectures. Otherwise, the default is taken from the target CPU
16278 (either the default CPU, or the one specified by @option{-mcpu}). For
16279 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
16280 @option{-mcpu=5206e}.
16282 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
16286 Consider type @code{int} to be 16 bits wide, like @code{short int}.
16287 Additionally, parameters passed on the stack are also aligned to a
16288 16-bit boundary even on targets whose API mandates promotion to 32-bit.
16292 Do not consider type @code{int} to be 16 bits wide. This is the default.
16295 @itemx -mno-bitfield
16296 @opindex mnobitfield
16297 @opindex mno-bitfield
16298 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
16299 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
16303 Do use the bit-field instructions. The @option{-m68020} option implies
16304 @option{-mbitfield}. This is the default if you use a configuration
16305 designed for a 68020.
16309 Use a different function-calling convention, in which functions
16310 that take a fixed number of arguments return with the @code{rtd}
16311 instruction, which pops their arguments while returning. This
16312 saves one instruction in the caller since there is no need to pop
16313 the arguments there.
16315 This calling convention is incompatible with the one normally
16316 used on Unix, so you cannot use it if you need to call libraries
16317 compiled with the Unix compiler.
16319 Also, you must provide function prototypes for all functions that
16320 take variable numbers of arguments (including @code{printf});
16321 otherwise incorrect code is generated for calls to those
16324 In addition, seriously incorrect code results if you call a
16325 function with too many arguments. (Normally, extra arguments are
16326 harmlessly ignored.)
16328 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
16329 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
16333 Do not use the calling conventions selected by @option{-mrtd}.
16334 This is the default.
16337 @itemx -mno-align-int
16338 @opindex malign-int
16339 @opindex mno-align-int
16340 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
16341 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
16342 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
16343 Aligning variables on 32-bit boundaries produces code that runs somewhat
16344 faster on processors with 32-bit busses at the expense of more memory.
16346 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
16347 aligns structures containing the above types differently than
16348 most published application binary interface specifications for the m68k.
16352 Use the pc-relative addressing mode of the 68000 directly, instead of
16353 using a global offset table. At present, this option implies @option{-fpic},
16354 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
16355 not presently supported with @option{-mpcrel}, though this could be supported for
16356 68020 and higher processors.
16358 @item -mno-strict-align
16359 @itemx -mstrict-align
16360 @opindex mno-strict-align
16361 @opindex mstrict-align
16362 Do not (do) assume that unaligned memory references are handled by
16366 Generate code that allows the data segment to be located in a different
16367 area of memory from the text segment. This allows for execute-in-place in
16368 an environment without virtual memory management. This option implies
16371 @item -mno-sep-data
16372 Generate code that assumes that the data segment follows the text segment.
16373 This is the default.
16375 @item -mid-shared-library
16376 Generate code that supports shared libraries via the library ID method.
16377 This allows for execute-in-place and shared libraries in an environment
16378 without virtual memory management. This option implies @option{-fPIC}.
16380 @item -mno-id-shared-library
16381 Generate code that doesn't assume ID-based shared libraries are being used.
16382 This is the default.
16384 @item -mshared-library-id=n
16385 Specifies the identification number of the ID-based shared library being
16386 compiled. Specifying a value of 0 generates more compact code; specifying
16387 other values forces the allocation of that number to the current
16388 library, but is no more space- or time-efficient than omitting this option.
16394 When generating position-independent code for ColdFire, generate code
16395 that works if the GOT has more than 8192 entries. This code is
16396 larger and slower than code generated without this option. On M680x0
16397 processors, this option is not needed; @option{-fPIC} suffices.
16399 GCC normally uses a single instruction to load values from the GOT@.
16400 While this is relatively efficient, it only works if the GOT
16401 is smaller than about 64k. Anything larger causes the linker
16402 to report an error such as:
16404 @cindex relocation truncated to fit (ColdFire)
16406 relocation truncated to fit: R_68K_GOT16O foobar
16409 If this happens, you should recompile your code with @option{-mxgot}.
16410 It should then work with very large GOTs. However, code generated with
16411 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
16412 the value of a global symbol.
16414 Note that some linkers, including newer versions of the GNU linker,
16415 can create multiple GOTs and sort GOT entries. If you have such a linker,
16416 you should only need to use @option{-mxgot} when compiling a single
16417 object file that accesses more than 8192 GOT entries. Very few do.
16419 These options have no effect unless GCC is generating
16420 position-independent code.
16424 @node MCore Options
16425 @subsection MCore Options
16426 @cindex MCore options
16428 These are the @samp{-m} options defined for the Motorola M*Core
16434 @itemx -mno-hardlit
16436 @opindex mno-hardlit
16437 Inline constants into the code stream if it can be done in two
16438 instructions or less.
16444 Use the divide instruction. (Enabled by default).
16446 @item -mrelax-immediate
16447 @itemx -mno-relax-immediate
16448 @opindex mrelax-immediate
16449 @opindex mno-relax-immediate
16450 Allow arbitrary-sized immediates in bit operations.
16452 @item -mwide-bitfields
16453 @itemx -mno-wide-bitfields
16454 @opindex mwide-bitfields
16455 @opindex mno-wide-bitfields
16456 Always treat bit-fields as @code{int}-sized.
16458 @item -m4byte-functions
16459 @itemx -mno-4byte-functions
16460 @opindex m4byte-functions
16461 @opindex mno-4byte-functions
16462 Force all functions to be aligned to a 4-byte boundary.
16464 @item -mcallgraph-data
16465 @itemx -mno-callgraph-data
16466 @opindex mcallgraph-data
16467 @opindex mno-callgraph-data
16468 Emit callgraph information.
16471 @itemx -mno-slow-bytes
16472 @opindex mslow-bytes
16473 @opindex mno-slow-bytes
16474 Prefer word access when reading byte quantities.
16476 @item -mlittle-endian
16477 @itemx -mbig-endian
16478 @opindex mlittle-endian
16479 @opindex mbig-endian
16480 Generate code for a little-endian target.
16486 Generate code for the 210 processor.
16490 Assume that runtime support has been provided and so omit the
16491 simulator library (@file{libsim.a)} from the linker command line.
16493 @item -mstack-increment=@var{size}
16494 @opindex mstack-increment
16495 Set the maximum amount for a single stack increment operation. Large
16496 values can increase the speed of programs that contain functions
16497 that need a large amount of stack space, but they can also trigger a
16498 segmentation fault if the stack is extended too much. The default
16504 @subsection MeP Options
16505 @cindex MeP options
16511 Enables the @code{abs} instruction, which is the absolute difference
16512 between two registers.
16516 Enables all the optional instructions---average, multiply, divide, bit
16517 operations, leading zero, absolute difference, min/max, clip, and
16523 Enables the @code{ave} instruction, which computes the average of two
16526 @item -mbased=@var{n}
16528 Variables of size @var{n} bytes or smaller are placed in the
16529 @code{.based} section by default. Based variables use the @code{$tp}
16530 register as a base register, and there is a 128-byte limit to the
16531 @code{.based} section.
16535 Enables the bit operation instructions---bit test (@code{btstm}), set
16536 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
16537 test-and-set (@code{tas}).
16539 @item -mc=@var{name}
16541 Selects which section constant data is placed in. @var{name} may
16542 be @samp{tiny}, @samp{near}, or @samp{far}.
16546 Enables the @code{clip} instruction. Note that @option{-mclip} is not
16547 useful unless you also provide @option{-mminmax}.
16549 @item -mconfig=@var{name}
16551 Selects one of the built-in core configurations. Each MeP chip has
16552 one or more modules in it; each module has a core CPU and a variety of
16553 coprocessors, optional instructions, and peripherals. The
16554 @code{MeP-Integrator} tool, not part of GCC, provides these
16555 configurations through this option; using this option is the same as
16556 using all the corresponding command-line options. The default
16557 configuration is @samp{default}.
16561 Enables the coprocessor instructions. By default, this is a 32-bit
16562 coprocessor. Note that the coprocessor is normally enabled via the
16563 @option{-mconfig=} option.
16567 Enables the 32-bit coprocessor's instructions.
16571 Enables the 64-bit coprocessor's instructions.
16575 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
16579 Causes constant variables to be placed in the @code{.near} section.
16583 Enables the @code{div} and @code{divu} instructions.
16587 Generate big-endian code.
16591 Generate little-endian code.
16593 @item -mio-volatile
16594 @opindex mio-volatile
16595 Tells the compiler that any variable marked with the @code{io}
16596 attribute is to be considered volatile.
16600 Causes variables to be assigned to the @code{.far} section by default.
16604 Enables the @code{leadz} (leading zero) instruction.
16608 Causes variables to be assigned to the @code{.near} section by default.
16612 Enables the @code{min} and @code{max} instructions.
16616 Enables the multiplication and multiply-accumulate instructions.
16620 Disables all the optional instructions enabled by @option{-mall-opts}.
16624 Enables the @code{repeat} and @code{erepeat} instructions, used for
16625 low-overhead looping.
16629 Causes all variables to default to the @code{.tiny} section. Note
16630 that there is a 65536-byte limit to this section. Accesses to these
16631 variables use the @code{%gp} base register.
16635 Enables the saturation instructions. Note that the compiler does not
16636 currently generate these itself, but this option is included for
16637 compatibility with other tools, like @code{as}.
16641 Link the SDRAM-based runtime instead of the default ROM-based runtime.
16645 Link the simulator run-time libraries.
16649 Link the simulator runtime libraries, excluding built-in support
16650 for reset and exception vectors and tables.
16654 Causes all functions to default to the @code{.far} section. Without
16655 this option, functions default to the @code{.near} section.
16657 @item -mtiny=@var{n}
16659 Variables that are @var{n} bytes or smaller are allocated to the
16660 @code{.tiny} section. These variables use the @code{$gp} base
16661 register. The default for this option is 4, but note that there's a
16662 65536-byte limit to the @code{.tiny} section.
16666 @node MicroBlaze Options
16667 @subsection MicroBlaze Options
16668 @cindex MicroBlaze Options
16673 @opindex msoft-float
16674 Use software emulation for floating point (default).
16677 @opindex mhard-float
16678 Use hardware floating-point instructions.
16682 Do not optimize block moves, use @code{memcpy}.
16684 @item -mno-clearbss
16685 @opindex mno-clearbss
16686 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
16688 @item -mcpu=@var{cpu-type}
16690 Use features of, and schedule code for, the given CPU.
16691 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
16692 where @var{X} is a major version, @var{YY} is the minor version, and
16693 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
16694 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
16696 @item -mxl-soft-mul
16697 @opindex mxl-soft-mul
16698 Use software multiply emulation (default).
16700 @item -mxl-soft-div
16701 @opindex mxl-soft-div
16702 Use software emulation for divides (default).
16704 @item -mxl-barrel-shift
16705 @opindex mxl-barrel-shift
16706 Use the hardware barrel shifter.
16708 @item -mxl-pattern-compare
16709 @opindex mxl-pattern-compare
16710 Use pattern compare instructions.
16712 @item -msmall-divides
16713 @opindex msmall-divides
16714 Use table lookup optimization for small signed integer divisions.
16716 @item -mxl-stack-check
16717 @opindex mxl-stack-check
16718 This option is deprecated. Use @option{-fstack-check} instead.
16721 @opindex mxl-gp-opt
16722 Use GP-relative @code{.sdata}/@code{.sbss} sections.
16724 @item -mxl-multiply-high
16725 @opindex mxl-multiply-high
16726 Use multiply high instructions for high part of 32x32 multiply.
16728 @item -mxl-float-convert
16729 @opindex mxl-float-convert
16730 Use hardware floating-point conversion instructions.
16732 @item -mxl-float-sqrt
16733 @opindex mxl-float-sqrt
16734 Use hardware floating-point square root instruction.
16737 @opindex mbig-endian
16738 Generate code for a big-endian target.
16740 @item -mlittle-endian
16741 @opindex mlittle-endian
16742 Generate code for a little-endian target.
16745 @opindex mxl-reorder
16746 Use reorder instructions (swap and byte reversed load/store).
16748 @item -mxl-mode-@var{app-model}
16749 Select application model @var{app-model}. Valid models are
16752 normal executable (default), uses startup code @file{crt0.o}.
16755 for use with Xilinx Microprocessor Debugger (XMD) based
16756 software intrusive debug agent called xmdstub. This uses startup file
16757 @file{crt1.o} and sets the start address of the program to 0x800.
16760 for applications that are loaded using a bootloader.
16761 This model uses startup file @file{crt2.o} which does not contain a processor
16762 reset vector handler. This is suitable for transferring control on a
16763 processor reset to the bootloader rather than the application.
16766 for applications that do not require any of the
16767 MicroBlaze vectors. This option may be useful for applications running
16768 within a monitoring application. This model uses @file{crt3.o} as a startup file.
16771 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
16772 @option{-mxl-mode-@var{app-model}}.
16777 @subsection MIPS Options
16778 @cindex MIPS options
16784 Generate big-endian code.
16788 Generate little-endian code. This is the default for @samp{mips*el-*-*}
16791 @item -march=@var{arch}
16793 Generate code that runs on @var{arch}, which can be the name of a
16794 generic MIPS ISA, or the name of a particular processor.
16796 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
16797 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
16798 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
16799 @samp{mips64r5} and @samp{mips64r6}.
16800 The processor names are:
16801 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
16802 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
16803 @samp{5kc}, @samp{5kf},
16805 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
16806 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
16807 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
16808 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
16809 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
16810 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
16812 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
16813 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
16816 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
16817 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
16818 @samp{rm7000}, @samp{rm9000},
16819 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
16822 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
16823 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
16824 @samp{xlr} and @samp{xlp}.
16825 The special value @samp{from-abi} selects the
16826 most compatible architecture for the selected ABI (that is,
16827 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
16829 The native Linux/GNU toolchain also supports the value @samp{native},
16830 which selects the best architecture option for the host processor.
16831 @option{-march=native} has no effect if GCC does not recognize
16834 In processor names, a final @samp{000} can be abbreviated as @samp{k}
16835 (for example, @option{-march=r2k}). Prefixes are optional, and
16836 @samp{vr} may be written @samp{r}.
16838 Names of the form @samp{@var{n}f2_1} refer to processors with
16839 FPUs clocked at half the rate of the core, names of the form
16840 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
16841 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
16842 processors with FPUs clocked a ratio of 3:2 with respect to the core.
16843 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
16844 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
16845 accepted as synonyms for @samp{@var{n}f1_1}.
16847 GCC defines two macros based on the value of this option. The first
16848 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
16849 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
16850 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
16851 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
16852 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
16854 Note that the @code{_MIPS_ARCH} macro uses the processor names given
16855 above. In other words, it has the full prefix and does not
16856 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
16857 the macro names the resolved architecture (either @code{"mips1"} or
16858 @code{"mips3"}). It names the default architecture when no
16859 @option{-march} option is given.
16861 @item -mtune=@var{arch}
16863 Optimize for @var{arch}. Among other things, this option controls
16864 the way instructions are scheduled, and the perceived cost of arithmetic
16865 operations. The list of @var{arch} values is the same as for
16868 When this option is not used, GCC optimizes for the processor
16869 specified by @option{-march}. By using @option{-march} and
16870 @option{-mtune} together, it is possible to generate code that
16871 runs on a family of processors, but optimize the code for one
16872 particular member of that family.
16874 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
16875 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
16876 @option{-march} ones described above.
16880 Equivalent to @option{-march=mips1}.
16884 Equivalent to @option{-march=mips2}.
16888 Equivalent to @option{-march=mips3}.
16892 Equivalent to @option{-march=mips4}.
16896 Equivalent to @option{-march=mips32}.
16900 Equivalent to @option{-march=mips32r3}.
16904 Equivalent to @option{-march=mips32r5}.
16908 Equivalent to @option{-march=mips32r6}.
16912 Equivalent to @option{-march=mips64}.
16916 Equivalent to @option{-march=mips64r2}.
16920 Equivalent to @option{-march=mips64r3}.
16924 Equivalent to @option{-march=mips64r5}.
16928 Equivalent to @option{-march=mips64r6}.
16933 @opindex mno-mips16
16934 Generate (do not generate) MIPS16 code. If GCC is targeting a
16935 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
16937 MIPS16 code generation can also be controlled on a per-function basis
16938 by means of @code{mips16} and @code{nomips16} attributes.
16939 @xref{Function Attributes}, for more information.
16941 @item -mflip-mips16
16942 @opindex mflip-mips16
16943 Generate MIPS16 code on alternating functions. This option is provided
16944 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
16945 not intended for ordinary use in compiling user code.
16947 @item -minterlink-compressed
16948 @item -mno-interlink-compressed
16949 @opindex minterlink-compressed
16950 @opindex mno-interlink-compressed
16951 Require (do not require) that code using the standard (uncompressed) MIPS ISA
16952 be link-compatible with MIPS16 and microMIPS code, and vice versa.
16954 For example, code using the standard ISA encoding cannot jump directly
16955 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
16956 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
16957 knows that the target of the jump is not compressed.
16959 @item -minterlink-mips16
16960 @itemx -mno-interlink-mips16
16961 @opindex minterlink-mips16
16962 @opindex mno-interlink-mips16
16963 Aliases of @option{-minterlink-compressed} and
16964 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
16965 and are retained for backwards compatibility.
16977 Generate code for the given ABI@.
16979 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
16980 generates 64-bit code when you select a 64-bit architecture, but you
16981 can use @option{-mgp32} to get 32-bit code instead.
16983 For information about the O64 ABI, see
16984 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
16986 GCC supports a variant of the o32 ABI in which floating-point registers
16987 are 64 rather than 32 bits wide. You can select this combination with
16988 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
16989 and @code{mfhc1} instructions and is therefore only supported for
16990 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
16992 The register assignments for arguments and return values remain the
16993 same, but each scalar value is passed in a single 64-bit register
16994 rather than a pair of 32-bit registers. For example, scalar
16995 floating-point values are returned in @samp{$f0} only, not a
16996 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
16997 remains the same in that the even-numbered double-precision registers
17000 Two additional variants of the o32 ABI are supported to enable
17001 a transition from 32-bit to 64-bit registers. These are FPXX
17002 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
17003 The FPXX extension mandates that all code must execute correctly
17004 when run using 32-bit or 64-bit registers. The code can be interlinked
17005 with either FP32 or FP64, but not both.
17006 The FP64A extension is similar to the FP64 extension but forbids the
17007 use of odd-numbered single-precision registers. This can be used
17008 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
17009 processors and allows both FP32 and FP64A code to interlink and
17010 run in the same process without changing FPU modes.
17013 @itemx -mno-abicalls
17015 @opindex mno-abicalls
17016 Generate (do not generate) code that is suitable for SVR4-style
17017 dynamic objects. @option{-mabicalls} is the default for SVR4-based
17022 Generate (do not generate) code that is fully position-independent,
17023 and that can therefore be linked into shared libraries. This option
17024 only affects @option{-mabicalls}.
17026 All @option{-mabicalls} code has traditionally been position-independent,
17027 regardless of options like @option{-fPIC} and @option{-fpic}. However,
17028 as an extension, the GNU toolchain allows executables to use absolute
17029 accesses for locally-binding symbols. It can also use shorter GP
17030 initialization sequences and generate direct calls to locally-defined
17031 functions. This mode is selected by @option{-mno-shared}.
17033 @option{-mno-shared} depends on binutils 2.16 or higher and generates
17034 objects that can only be linked by the GNU linker. However, the option
17035 does not affect the ABI of the final executable; it only affects the ABI
17036 of relocatable objects. Using @option{-mno-shared} generally makes
17037 executables both smaller and quicker.
17039 @option{-mshared} is the default.
17045 Assume (do not assume) that the static and dynamic linkers
17046 support PLTs and copy relocations. This option only affects
17047 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
17048 has no effect without @option{-msym32}.
17050 You can make @option{-mplt} the default by configuring
17051 GCC with @option{--with-mips-plt}. The default is
17052 @option{-mno-plt} otherwise.
17058 Lift (do not lift) the usual restrictions on the size of the global
17061 GCC normally uses a single instruction to load values from the GOT@.
17062 While this is relatively efficient, it only works if the GOT
17063 is smaller than about 64k. Anything larger causes the linker
17064 to report an error such as:
17066 @cindex relocation truncated to fit (MIPS)
17068 relocation truncated to fit: R_MIPS_GOT16 foobar
17071 If this happens, you should recompile your code with @option{-mxgot}.
17072 This works with very large GOTs, although the code is also
17073 less efficient, since it takes three instructions to fetch the
17074 value of a global symbol.
17076 Note that some linkers can create multiple GOTs. If you have such a
17077 linker, you should only need to use @option{-mxgot} when a single object
17078 file accesses more than 64k's worth of GOT entries. Very few do.
17080 These options have no effect unless GCC is generating position
17085 Assume that general-purpose registers are 32 bits wide.
17089 Assume that general-purpose registers are 64 bits wide.
17093 Assume that floating-point registers are 32 bits wide.
17097 Assume that floating-point registers are 64 bits wide.
17101 Do not assume the width of floating-point registers.
17104 @opindex mhard-float
17105 Use floating-point coprocessor instructions.
17108 @opindex msoft-float
17109 Do not use floating-point coprocessor instructions. Implement
17110 floating-point calculations using library calls instead.
17114 Equivalent to @option{-msoft-float}, but additionally asserts that the
17115 program being compiled does not perform any floating-point operations.
17116 This option is presently supported only by some bare-metal MIPS
17117 configurations, where it may select a special set of libraries
17118 that lack all floating-point support (including, for example, the
17119 floating-point @code{printf} formats).
17120 If code compiled with @option{-mno-float} accidentally contains
17121 floating-point operations, it is likely to suffer a link-time
17122 or run-time failure.
17124 @item -msingle-float
17125 @opindex msingle-float
17126 Assume that the floating-point coprocessor only supports single-precision
17129 @item -mdouble-float
17130 @opindex mdouble-float
17131 Assume that the floating-point coprocessor supports double-precision
17132 operations. This is the default.
17135 @itemx -mno-odd-spreg
17136 @opindex modd-spreg
17137 @opindex mno-odd-spreg
17138 Enable the use of odd-numbered single-precision floating-point registers
17139 for the o32 ABI. This is the default for processors that are known to
17140 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
17144 @itemx -mabs=legacy
17146 @opindex mabs=legacy
17147 These options control the treatment of the special not-a-number (NaN)
17148 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
17149 @code{neg.@i{fmt}} machine instructions.
17151 By default or when the @option{-mabs=legacy} is used the legacy
17152 treatment is selected. In this case these instructions are considered
17153 arithmetic and avoided where correct operation is required and the
17154 input operand might be a NaN. A longer sequence of instructions that
17155 manipulate the sign bit of floating-point datum manually is used
17156 instead unless the @option{-ffinite-math-only} option has also been
17159 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
17160 this case these instructions are considered non-arithmetic and therefore
17161 operating correctly in all cases, including in particular where the
17162 input operand is a NaN. These instructions are therefore always used
17163 for the respective operations.
17166 @itemx -mnan=legacy
17168 @opindex mnan=legacy
17169 These options control the encoding of the special not-a-number (NaN)
17170 IEEE 754 floating-point data.
17172 The @option{-mnan=legacy} option selects the legacy encoding. In this
17173 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
17174 significand field being 0, whereas signalling NaNs (sNaNs) are denoted
17175 by the first bit of their trailing significand field being 1.
17177 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
17178 this case qNaNs are denoted by the first bit of their trailing
17179 significand field being 1, whereas sNaNs are denoted by the first bit of
17180 their trailing significand field being 0.
17182 The default is @option{-mnan=legacy} unless GCC has been configured with
17183 @option{--with-nan=2008}.
17189 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
17190 implement atomic memory built-in functions. When neither option is
17191 specified, GCC uses the instructions if the target architecture
17194 @option{-mllsc} is useful if the runtime environment can emulate the
17195 instructions and @option{-mno-llsc} can be useful when compiling for
17196 nonstandard ISAs. You can make either option the default by
17197 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
17198 respectively. @option{--with-llsc} is the default for some
17199 configurations; see the installation documentation for details.
17205 Use (do not use) revision 1 of the MIPS DSP ASE@.
17206 @xref{MIPS DSP Built-in Functions}. This option defines the
17207 preprocessor macro @code{__mips_dsp}. It also defines
17208 @code{__mips_dsp_rev} to 1.
17214 Use (do not use) revision 2 of the MIPS DSP ASE@.
17215 @xref{MIPS DSP Built-in Functions}. This option defines the
17216 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
17217 It also defines @code{__mips_dsp_rev} to 2.
17220 @itemx -mno-smartmips
17221 @opindex msmartmips
17222 @opindex mno-smartmips
17223 Use (do not use) the MIPS SmartMIPS ASE.
17225 @item -mpaired-single
17226 @itemx -mno-paired-single
17227 @opindex mpaired-single
17228 @opindex mno-paired-single
17229 Use (do not use) paired-single floating-point instructions.
17230 @xref{MIPS Paired-Single Support}. This option requires
17231 hardware floating-point support to be enabled.
17237 Use (do not use) MIPS Digital Media Extension instructions.
17238 This option can only be used when generating 64-bit code and requires
17239 hardware floating-point support to be enabled.
17244 @opindex mno-mips3d
17245 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
17246 The option @option{-mips3d} implies @option{-mpaired-single}.
17249 @itemx -mno-micromips
17250 @opindex mmicromips
17251 @opindex mno-mmicromips
17252 Generate (do not generate) microMIPS code.
17254 MicroMIPS code generation can also be controlled on a per-function basis
17255 by means of @code{micromips} and @code{nomicromips} attributes.
17256 @xref{Function Attributes}, for more information.
17262 Use (do not use) MT Multithreading instructions.
17268 Use (do not use) the MIPS MCU ASE instructions.
17274 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
17280 Use (do not use) the MIPS Virtualization Application Specific instructions.
17286 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
17290 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
17291 an explanation of the default and the way that the pointer size is
17296 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
17298 The default size of @code{int}s, @code{long}s and pointers depends on
17299 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
17300 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
17301 32-bit @code{long}s. Pointers are the same size as @code{long}s,
17302 or the same size as integer registers, whichever is smaller.
17308 Assume (do not assume) that all symbols have 32-bit values, regardless
17309 of the selected ABI@. This option is useful in combination with
17310 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
17311 to generate shorter and faster references to symbolic addresses.
17315 Put definitions of externally-visible data in a small data section
17316 if that data is no bigger than @var{num} bytes. GCC can then generate
17317 more efficient accesses to the data; see @option{-mgpopt} for details.
17319 The default @option{-G} option depends on the configuration.
17321 @item -mlocal-sdata
17322 @itemx -mno-local-sdata
17323 @opindex mlocal-sdata
17324 @opindex mno-local-sdata
17325 Extend (do not extend) the @option{-G} behavior to local data too,
17326 such as to static variables in C@. @option{-mlocal-sdata} is the
17327 default for all configurations.
17329 If the linker complains that an application is using too much small data,
17330 you might want to try rebuilding the less performance-critical parts with
17331 @option{-mno-local-sdata}. You might also want to build large
17332 libraries with @option{-mno-local-sdata}, so that the libraries leave
17333 more room for the main program.
17335 @item -mextern-sdata
17336 @itemx -mno-extern-sdata
17337 @opindex mextern-sdata
17338 @opindex mno-extern-sdata
17339 Assume (do not assume) that externally-defined data is in
17340 a small data section if the size of that data is within the @option{-G} limit.
17341 @option{-mextern-sdata} is the default for all configurations.
17343 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
17344 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
17345 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
17346 is placed in a small data section. If @var{Var} is defined by another
17347 module, you must either compile that module with a high-enough
17348 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
17349 definition. If @var{Var} is common, you must link the application
17350 with a high-enough @option{-G} setting.
17352 The easiest way of satisfying these restrictions is to compile
17353 and link every module with the same @option{-G} option. However,
17354 you may wish to build a library that supports several different
17355 small data limits. You can do this by compiling the library with
17356 the highest supported @option{-G} setting and additionally using
17357 @option{-mno-extern-sdata} to stop the library from making assumptions
17358 about externally-defined data.
17364 Use (do not use) GP-relative accesses for symbols that are known to be
17365 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
17366 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
17369 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
17370 might not hold the value of @code{_gp}. For example, if the code is
17371 part of a library that might be used in a boot monitor, programs that
17372 call boot monitor routines pass an unknown value in @code{$gp}.
17373 (In such situations, the boot monitor itself is usually compiled
17374 with @option{-G0}.)
17376 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
17377 @option{-mno-extern-sdata}.
17379 @item -membedded-data
17380 @itemx -mno-embedded-data
17381 @opindex membedded-data
17382 @opindex mno-embedded-data
17383 Allocate variables to the read-only data section first if possible, then
17384 next in the small data section if possible, otherwise in data. This gives
17385 slightly slower code than the default, but reduces the amount of RAM required
17386 when executing, and thus may be preferred for some embedded systems.
17388 @item -muninit-const-in-rodata
17389 @itemx -mno-uninit-const-in-rodata
17390 @opindex muninit-const-in-rodata
17391 @opindex mno-uninit-const-in-rodata
17392 Put uninitialized @code{const} variables in the read-only data section.
17393 This option is only meaningful in conjunction with @option{-membedded-data}.
17395 @item -mcode-readable=@var{setting}
17396 @opindex mcode-readable
17397 Specify whether GCC may generate code that reads from executable sections.
17398 There are three possible settings:
17401 @item -mcode-readable=yes
17402 Instructions may freely access executable sections. This is the
17405 @item -mcode-readable=pcrel
17406 MIPS16 PC-relative load instructions can access executable sections,
17407 but other instructions must not do so. This option is useful on 4KSc
17408 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
17409 It is also useful on processors that can be configured to have a dual
17410 instruction/data SRAM interface and that, like the M4K, automatically
17411 redirect PC-relative loads to the instruction RAM.
17413 @item -mcode-readable=no
17414 Instructions must not access executable sections. This option can be
17415 useful on targets that are configured to have a dual instruction/data
17416 SRAM interface but that (unlike the M4K) do not automatically redirect
17417 PC-relative loads to the instruction RAM.
17420 @item -msplit-addresses
17421 @itemx -mno-split-addresses
17422 @opindex msplit-addresses
17423 @opindex mno-split-addresses
17424 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
17425 relocation operators. This option has been superseded by
17426 @option{-mexplicit-relocs} but is retained for backwards compatibility.
17428 @item -mexplicit-relocs
17429 @itemx -mno-explicit-relocs
17430 @opindex mexplicit-relocs
17431 @opindex mno-explicit-relocs
17432 Use (do not use) assembler relocation operators when dealing with symbolic
17433 addresses. The alternative, selected by @option{-mno-explicit-relocs},
17434 is to use assembler macros instead.
17436 @option{-mexplicit-relocs} is the default if GCC was configured
17437 to use an assembler that supports relocation operators.
17439 @item -mcheck-zero-division
17440 @itemx -mno-check-zero-division
17441 @opindex mcheck-zero-division
17442 @opindex mno-check-zero-division
17443 Trap (do not trap) on integer division by zero.
17445 The default is @option{-mcheck-zero-division}.
17447 @item -mdivide-traps
17448 @itemx -mdivide-breaks
17449 @opindex mdivide-traps
17450 @opindex mdivide-breaks
17451 MIPS systems check for division by zero by generating either a
17452 conditional trap or a break instruction. Using traps results in
17453 smaller code, but is only supported on MIPS II and later. Also, some
17454 versions of the Linux kernel have a bug that prevents trap from
17455 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
17456 allow conditional traps on architectures that support them and
17457 @option{-mdivide-breaks} to force the use of breaks.
17459 The default is usually @option{-mdivide-traps}, but this can be
17460 overridden at configure time using @option{--with-divide=breaks}.
17461 Divide-by-zero checks can be completely disabled using
17462 @option{-mno-check-zero-division}.
17467 @opindex mno-memcpy
17468 Force (do not force) the use of @code{memcpy} for non-trivial block
17469 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
17470 most constant-sized copies.
17473 @itemx -mno-long-calls
17474 @opindex mlong-calls
17475 @opindex mno-long-calls
17476 Disable (do not disable) use of the @code{jal} instruction. Calling
17477 functions using @code{jal} is more efficient but requires the caller
17478 and callee to be in the same 256 megabyte segment.
17480 This option has no effect on abicalls code. The default is
17481 @option{-mno-long-calls}.
17487 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
17488 instructions, as provided by the R4650 ISA@.
17494 Enable (disable) use of the @code{madd} and @code{msub} integer
17495 instructions. The default is @option{-mimadd} on architectures
17496 that support @code{madd} and @code{msub} except for the 74k
17497 architecture where it was found to generate slower code.
17500 @itemx -mno-fused-madd
17501 @opindex mfused-madd
17502 @opindex mno-fused-madd
17503 Enable (disable) use of the floating-point multiply-accumulate
17504 instructions, when they are available. The default is
17505 @option{-mfused-madd}.
17507 On the R8000 CPU when multiply-accumulate instructions are used,
17508 the intermediate product is calculated to infinite precision
17509 and is not subject to the FCSR Flush to Zero bit. This may be
17510 undesirable in some circumstances. On other processors the result
17511 is numerically identical to the equivalent computation using
17512 separate multiply, add, subtract and negate instructions.
17516 Tell the MIPS assembler to not run its preprocessor over user
17517 assembler files (with a @samp{.s} suffix) when assembling them.
17522 @opindex mno-fix-24k
17523 Work around the 24K E48 (lost data on stores during refill) errata.
17524 The workarounds are implemented by the assembler rather than by GCC@.
17527 @itemx -mno-fix-r4000
17528 @opindex mfix-r4000
17529 @opindex mno-fix-r4000
17530 Work around certain R4000 CPU errata:
17533 A double-word or a variable shift may give an incorrect result if executed
17534 immediately after starting an integer division.
17536 A double-word or a variable shift may give an incorrect result if executed
17537 while an integer multiplication is in progress.
17539 An integer division may give an incorrect result if started in a delay slot
17540 of a taken branch or a jump.
17544 @itemx -mno-fix-r4400
17545 @opindex mfix-r4400
17546 @opindex mno-fix-r4400
17547 Work around certain R4400 CPU errata:
17550 A double-word or a variable shift may give an incorrect result if executed
17551 immediately after starting an integer division.
17555 @itemx -mno-fix-r10000
17556 @opindex mfix-r10000
17557 @opindex mno-fix-r10000
17558 Work around certain R10000 errata:
17561 @code{ll}/@code{sc} sequences may not behave atomically on revisions
17562 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
17565 This option can only be used if the target architecture supports
17566 branch-likely instructions. @option{-mfix-r10000} is the default when
17567 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
17571 @itemx -mno-fix-rm7000
17572 @opindex mfix-rm7000
17573 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
17574 workarounds are implemented by the assembler rather than by GCC@.
17577 @itemx -mno-fix-vr4120
17578 @opindex mfix-vr4120
17579 Work around certain VR4120 errata:
17582 @code{dmultu} does not always produce the correct result.
17584 @code{div} and @code{ddiv} do not always produce the correct result if one
17585 of the operands is negative.
17587 The workarounds for the division errata rely on special functions in
17588 @file{libgcc.a}. At present, these functions are only provided by
17589 the @code{mips64vr*-elf} configurations.
17591 Other VR4120 errata require a NOP to be inserted between certain pairs of
17592 instructions. These errata are handled by the assembler, not by GCC itself.
17595 @opindex mfix-vr4130
17596 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
17597 workarounds are implemented by the assembler rather than by GCC,
17598 although GCC avoids using @code{mflo} and @code{mfhi} if the
17599 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
17600 instructions are available instead.
17603 @itemx -mno-fix-sb1
17605 Work around certain SB-1 CPU core errata.
17606 (This flag currently works around the SB-1 revision 2
17607 ``F1'' and ``F2'' floating-point errata.)
17609 @item -mr10k-cache-barrier=@var{setting}
17610 @opindex mr10k-cache-barrier
17611 Specify whether GCC should insert cache barriers to avoid the
17612 side-effects of speculation on R10K processors.
17614 In common with many processors, the R10K tries to predict the outcome
17615 of a conditional branch and speculatively executes instructions from
17616 the ``taken'' branch. It later aborts these instructions if the
17617 predicted outcome is wrong. However, on the R10K, even aborted
17618 instructions can have side effects.
17620 This problem only affects kernel stores and, depending on the system,
17621 kernel loads. As an example, a speculatively-executed store may load
17622 the target memory into cache and mark the cache line as dirty, even if
17623 the store itself is later aborted. If a DMA operation writes to the
17624 same area of memory before the ``dirty'' line is flushed, the cached
17625 data overwrites the DMA-ed data. See the R10K processor manual
17626 for a full description, including other potential problems.
17628 One workaround is to insert cache barrier instructions before every memory
17629 access that might be speculatively executed and that might have side
17630 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
17631 controls GCC's implementation of this workaround. It assumes that
17632 aborted accesses to any byte in the following regions does not have
17637 the memory occupied by the current function's stack frame;
17640 the memory occupied by an incoming stack argument;
17643 the memory occupied by an object with a link-time-constant address.
17646 It is the kernel's responsibility to ensure that speculative
17647 accesses to these regions are indeed safe.
17649 If the input program contains a function declaration such as:
17655 then the implementation of @code{foo} must allow @code{j foo} and
17656 @code{jal foo} to be executed speculatively. GCC honors this
17657 restriction for functions it compiles itself. It expects non-GCC
17658 functions (such as hand-written assembly code) to do the same.
17660 The option has three forms:
17663 @item -mr10k-cache-barrier=load-store
17664 Insert a cache barrier before a load or store that might be
17665 speculatively executed and that might have side effects even
17668 @item -mr10k-cache-barrier=store
17669 Insert a cache barrier before a store that might be speculatively
17670 executed and that might have side effects even if aborted.
17672 @item -mr10k-cache-barrier=none
17673 Disable the insertion of cache barriers. This is the default setting.
17676 @item -mflush-func=@var{func}
17677 @itemx -mno-flush-func
17678 @opindex mflush-func
17679 Specifies the function to call to flush the I and D caches, or to not
17680 call any such function. If called, the function must take the same
17681 arguments as the common @code{_flush_func}, that is, the address of the
17682 memory range for which the cache is being flushed, the size of the
17683 memory range, and the number 3 (to flush both caches). The default
17684 depends on the target GCC was configured for, but commonly is either
17685 @code{_flush_func} or @code{__cpu_flush}.
17687 @item mbranch-cost=@var{num}
17688 @opindex mbranch-cost
17689 Set the cost of branches to roughly @var{num} ``simple'' instructions.
17690 This cost is only a heuristic and is not guaranteed to produce
17691 consistent results across releases. A zero cost redundantly selects
17692 the default, which is based on the @option{-mtune} setting.
17694 @item -mbranch-likely
17695 @itemx -mno-branch-likely
17696 @opindex mbranch-likely
17697 @opindex mno-branch-likely
17698 Enable or disable use of Branch Likely instructions, regardless of the
17699 default for the selected architecture. By default, Branch Likely
17700 instructions may be generated if they are supported by the selected
17701 architecture. An exception is for the MIPS32 and MIPS64 architectures
17702 and processors that implement those architectures; for those, Branch
17703 Likely instructions are not be generated by default because the MIPS32
17704 and MIPS64 architectures specifically deprecate their use.
17706 @item -mfp-exceptions
17707 @itemx -mno-fp-exceptions
17708 @opindex mfp-exceptions
17709 Specifies whether FP exceptions are enabled. This affects how
17710 FP instructions are scheduled for some processors.
17711 The default is that FP exceptions are
17714 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
17715 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
17718 @item -mvr4130-align
17719 @itemx -mno-vr4130-align
17720 @opindex mvr4130-align
17721 The VR4130 pipeline is two-way superscalar, but can only issue two
17722 instructions together if the first one is 8-byte aligned. When this
17723 option is enabled, GCC aligns pairs of instructions that it
17724 thinks should execute in parallel.
17726 This option only has an effect when optimizing for the VR4130.
17727 It normally makes code faster, but at the expense of making it bigger.
17728 It is enabled by default at optimization level @option{-O3}.
17733 Enable (disable) generation of @code{synci} instructions on
17734 architectures that support it. The @code{synci} instructions (if
17735 enabled) are generated when @code{__builtin___clear_cache} is
17738 This option defaults to @option{-mno-synci}, but the default can be
17739 overridden by configuring GCC with @option{--with-synci}.
17741 When compiling code for single processor systems, it is generally safe
17742 to use @code{synci}. However, on many multi-core (SMP) systems, it
17743 does not invalidate the instruction caches on all cores and may lead
17744 to undefined behavior.
17746 @item -mrelax-pic-calls
17747 @itemx -mno-relax-pic-calls
17748 @opindex mrelax-pic-calls
17749 Try to turn PIC calls that are normally dispatched via register
17750 @code{$25} into direct calls. This is only possible if the linker can
17751 resolve the destination at link-time and if the destination is within
17752 range for a direct call.
17754 @option{-mrelax-pic-calls} is the default if GCC was configured to use
17755 an assembler and a linker that support the @code{.reloc} assembly
17756 directive and @option{-mexplicit-relocs} is in effect. With
17757 @option{-mno-explicit-relocs}, this optimization can be performed by the
17758 assembler and the linker alone without help from the compiler.
17760 @item -mmcount-ra-address
17761 @itemx -mno-mcount-ra-address
17762 @opindex mmcount-ra-address
17763 @opindex mno-mcount-ra-address
17764 Emit (do not emit) code that allows @code{_mcount} to modify the
17765 calling function's return address. When enabled, this option extends
17766 the usual @code{_mcount} interface with a new @var{ra-address}
17767 parameter, which has type @code{intptr_t *} and is passed in register
17768 @code{$12}. @code{_mcount} can then modify the return address by
17769 doing both of the following:
17772 Returning the new address in register @code{$31}.
17774 Storing the new address in @code{*@var{ra-address}},
17775 if @var{ra-address} is nonnull.
17778 The default is @option{-mno-mcount-ra-address}.
17783 @subsection MMIX Options
17784 @cindex MMIX Options
17786 These options are defined for the MMIX:
17790 @itemx -mno-libfuncs
17792 @opindex mno-libfuncs
17793 Specify that intrinsic library functions are being compiled, passing all
17794 values in registers, no matter the size.
17797 @itemx -mno-epsilon
17799 @opindex mno-epsilon
17800 Generate floating-point comparison instructions that compare with respect
17801 to the @code{rE} epsilon register.
17803 @item -mabi=mmixware
17805 @opindex mabi=mmixware
17807 Generate code that passes function parameters and return values that (in
17808 the called function) are seen as registers @code{$0} and up, as opposed to
17809 the GNU ABI which uses global registers @code{$231} and up.
17811 @item -mzero-extend
17812 @itemx -mno-zero-extend
17813 @opindex mzero-extend
17814 @opindex mno-zero-extend
17815 When reading data from memory in sizes shorter than 64 bits, use (do not
17816 use) zero-extending load instructions by default, rather than
17817 sign-extending ones.
17820 @itemx -mno-knuthdiv
17822 @opindex mno-knuthdiv
17823 Make the result of a division yielding a remainder have the same sign as
17824 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
17825 remainder follows the sign of the dividend. Both methods are
17826 arithmetically valid, the latter being almost exclusively used.
17828 @item -mtoplevel-symbols
17829 @itemx -mno-toplevel-symbols
17830 @opindex mtoplevel-symbols
17831 @opindex mno-toplevel-symbols
17832 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
17833 code can be used with the @code{PREFIX} assembly directive.
17837 Generate an executable in the ELF format, rather than the default
17838 @samp{mmo} format used by the @command{mmix} simulator.
17840 @item -mbranch-predict
17841 @itemx -mno-branch-predict
17842 @opindex mbranch-predict
17843 @opindex mno-branch-predict
17844 Use (do not use) the probable-branch instructions, when static branch
17845 prediction indicates a probable branch.
17847 @item -mbase-addresses
17848 @itemx -mno-base-addresses
17849 @opindex mbase-addresses
17850 @opindex mno-base-addresses
17851 Generate (do not generate) code that uses @emph{base addresses}. Using a
17852 base address automatically generates a request (handled by the assembler
17853 and the linker) for a constant to be set up in a global register. The
17854 register is used for one or more base address requests within the range 0
17855 to 255 from the value held in the register. The generally leads to short
17856 and fast code, but the number of different data items that can be
17857 addressed is limited. This means that a program that uses lots of static
17858 data may require @option{-mno-base-addresses}.
17860 @item -msingle-exit
17861 @itemx -mno-single-exit
17862 @opindex msingle-exit
17863 @opindex mno-single-exit
17864 Force (do not force) generated code to have a single exit point in each
17868 @node MN10300 Options
17869 @subsection MN10300 Options
17870 @cindex MN10300 options
17872 These @option{-m} options are defined for Matsushita MN10300 architectures:
17877 Generate code to avoid bugs in the multiply instructions for the MN10300
17878 processors. This is the default.
17880 @item -mno-mult-bug
17881 @opindex mno-mult-bug
17882 Do not generate code to avoid bugs in the multiply instructions for the
17883 MN10300 processors.
17887 Generate code using features specific to the AM33 processor.
17891 Do not generate code using features specific to the AM33 processor. This
17896 Generate code using features specific to the AM33/2.0 processor.
17900 Generate code using features specific to the AM34 processor.
17902 @item -mtune=@var{cpu-type}
17904 Use the timing characteristics of the indicated CPU type when
17905 scheduling instructions. This does not change the targeted processor
17906 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
17907 @samp{am33-2} or @samp{am34}.
17909 @item -mreturn-pointer-on-d0
17910 @opindex mreturn-pointer-on-d0
17911 When generating a function that returns a pointer, return the pointer
17912 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
17913 only in @code{a0}, and attempts to call such functions without a prototype
17914 result in errors. Note that this option is on by default; use
17915 @option{-mno-return-pointer-on-d0} to disable it.
17919 Do not link in the C run-time initialization object file.
17923 Indicate to the linker that it should perform a relaxation optimization pass
17924 to shorten branches, calls and absolute memory addresses. This option only
17925 has an effect when used on the command line for the final link step.
17927 This option makes symbolic debugging impossible.
17931 Allow the compiler to generate @emph{Long Instruction Word}
17932 instructions if the target is the @samp{AM33} or later. This is the
17933 default. This option defines the preprocessor macro @code{__LIW__}.
17937 Do not allow the compiler to generate @emph{Long Instruction Word}
17938 instructions. This option defines the preprocessor macro
17943 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
17944 instructions if the target is the @samp{AM33} or later. This is the
17945 default. This option defines the preprocessor macro @code{__SETLB__}.
17949 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
17950 instructions. This option defines the preprocessor macro
17951 @code{__NO_SETLB__}.
17955 @node Moxie Options
17956 @subsection Moxie Options
17957 @cindex Moxie Options
17963 Generate big-endian code. This is the default for @samp{moxie-*-*}
17968 Generate little-endian code.
17972 Generate mul.x and umul.x instructions. This is the default for
17973 @samp{moxiebox-*-*} configurations.
17977 Do not link in the C run-time initialization object file.
17981 @node MSP430 Options
17982 @subsection MSP430 Options
17983 @cindex MSP430 Options
17985 These options are defined for the MSP430:
17991 Force assembly output to always use hex constants. Normally such
17992 constants are signed decimals, but this option is available for
17993 testsuite and/or aesthetic purposes.
17997 Select the MCU to target. This is used to create a C preprocessor
17998 symbol based upon the MCU name, converted to upper case and pre- and
17999 post-fixed with @samp{__}. This in turn is used by the
18000 @file{msp430.h} header file to select an MCU-specific supplementary
18003 The option also sets the ISA to use. If the MCU name is one that is
18004 known to only support the 430 ISA then that is selected, otherwise the
18005 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
18006 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
18007 name selects the 430X ISA.
18009 In addition an MCU-specific linker script is added to the linker
18010 command line. The script's name is the name of the MCU with
18011 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
18012 command line defines the C preprocessor symbol @code{__XXX__} and
18013 cause the linker to search for a script called @file{xxx.ld}.
18015 This option is also passed on to the assembler.
18019 Specifies the ISA to use. Accepted values are @samp{msp430},
18020 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
18021 @option{-mmcu=} option should be used to select the ISA.
18025 Link to the simulator runtime libraries and linker script. Overrides
18026 any scripts that would be selected by the @option{-mmcu=} option.
18030 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
18034 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
18038 This option is passed to the assembler and linker, and allows the
18039 linker to perform certain optimizations that cannot be done until
18044 Describes the type of hardware multiply supported by the target.
18045 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
18046 for the original 16-bit-only multiply supported by early MCUs.
18047 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
18048 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
18049 A value of @samp{auto} can also be given. This tells GCC to deduce
18050 the hardware multiply support based upon the MCU name provided by the
18051 @option{-mmcu} option. If no @option{-mmcu} option is specified then
18052 @samp{32bit} hardware multiply support is assumed. @samp{auto} is the
18055 Hardware multiplies are normally performed by calling a library
18056 routine. This saves space in the generated code. When compiling at
18057 @option{-O3} or higher however the hardware multiplier is invoked
18058 inline. This makes for bigger, but faster code.
18060 The hardware multiply routines disable interrupts whilst running and
18061 restore the previous interrupt state when they finish. This makes
18062 them safe to use inside interrupt handlers as well as in normal code.
18066 Enable the use of a minimum runtime environment - no static
18067 initializers or constructors. This is intended for memory-constrained
18068 devices. The compiler includes special symbols in some objects
18069 that tell the linker and runtime which code fragments are required.
18073 @node NDS32 Options
18074 @subsection NDS32 Options
18075 @cindex NDS32 Options
18077 These options are defined for NDS32 implementations:
18082 @opindex mbig-endian
18083 Generate code in big-endian mode.
18085 @item -mlittle-endian
18086 @opindex mlittle-endian
18087 Generate code in little-endian mode.
18089 @item -mreduced-regs
18090 @opindex mreduced-regs
18091 Use reduced-set registers for register allocation.
18094 @opindex mfull-regs
18095 Use full-set registers for register allocation.
18099 Generate conditional move instructions.
18103 Do not generate conditional move instructions.
18107 Generate performance extension instructions.
18109 @item -mno-perf-ext
18110 @opindex mno-perf-ext
18111 Do not generate performance extension instructions.
18115 Generate v3 push25/pop25 instructions.
18118 @opindex mno-v3push
18119 Do not generate v3 push25/pop25 instructions.
18123 Generate 16-bit instructions.
18126 @opindex mno-16-bit
18127 Do not generate 16-bit instructions.
18129 @item -misr-vector-size=@var{num}
18130 @opindex misr-vector-size
18131 Specify the size of each interrupt vector, which must be 4 or 16.
18133 @item -mcache-block-size=@var{num}
18134 @opindex mcache-block-size
18135 Specify the size of each cache block,
18136 which must be a power of 2 between 4 and 512.
18138 @item -march=@var{arch}
18140 Specify the name of the target architecture.
18142 @item -mcmodel=@var{code-model}
18144 Set the code model to one of
18147 All the data and read-only data segments must be within 512KB addressing space.
18148 The text segment must be within 16MB addressing space.
18149 @item @samp{medium}
18150 The data segment must be within 512KB while the read-only data segment can be
18151 within 4GB addressing space. The text segment should be still within 16MB
18154 All the text and data segments can be within 4GB addressing space.
18158 @opindex mctor-dtor
18159 Enable constructor/destructor feature.
18163 Guide linker to relax instructions.
18167 @node Nios II Options
18168 @subsection Nios II Options
18169 @cindex Nios II options
18170 @cindex Altera Nios II options
18172 These are the options defined for the Altera Nios II processor.
18178 @cindex smaller data references
18179 Put global and static objects less than or equal to @var{num} bytes
18180 into the small data or BSS sections instead of the normal data or BSS
18181 sections. The default value of @var{num} is 8.
18183 @item -mgpopt=@var{option}
18188 Generate (do not generate) GP-relative accesses. The following
18189 @var{option} names are recognized:
18194 Do not generate GP-relative accesses.
18197 Generate GP-relative accesses for small data objects that are not
18198 external or weak. Also use GP-relative addressing for objects that
18199 have been explicitly placed in a small data section via a @code{section}
18203 As for @samp{local}, but also generate GP-relative accesses for
18204 small data objects that are external or weak. If you use this option,
18205 you must ensure that all parts of your program (including libraries) are
18206 compiled with the same @option{-G} setting.
18209 Generate GP-relative accesses for all data objects in the program. If you
18210 use this option, the entire data and BSS segments
18211 of your program must fit in 64K of memory and you must use an appropriate
18212 linker script to allocate them within the addressible range of the
18216 Generate GP-relative addresses for function pointers as well as data
18217 pointers. If you use this option, the entire text, data, and BSS segments
18218 of your program must fit in 64K of memory and you must use an appropriate
18219 linker script to allocate them within the addressible range of the
18224 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
18225 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
18227 The default is @option{-mgpopt} except when @option{-fpic} or
18228 @option{-fPIC} is specified to generate position-independent code.
18229 Note that the Nios II ABI does not permit GP-relative accesses from
18232 You may need to specify @option{-mno-gpopt} explicitly when building
18233 programs that include large amounts of small data, including large
18234 GOT data sections. In this case, the 16-bit offset for GP-relative
18235 addressing may not be large enough to allow access to the entire
18236 small data section.
18242 Generate little-endian (default) or big-endian (experimental) code,
18245 @item -mbypass-cache
18246 @itemx -mno-bypass-cache
18247 @opindex mno-bypass-cache
18248 @opindex mbypass-cache
18249 Force all load and store instructions to always bypass cache by
18250 using I/O variants of the instructions. The default is not to
18253 @item -mno-cache-volatile
18254 @itemx -mcache-volatile
18255 @opindex mcache-volatile
18256 @opindex mno-cache-volatile
18257 Volatile memory access bypass the cache using the I/O variants of
18258 the load and store instructions. The default is not to bypass the cache.
18260 @item -mno-fast-sw-div
18261 @itemx -mfast-sw-div
18262 @opindex mno-fast-sw-div
18263 @opindex mfast-sw-div
18264 Do not use table-based fast divide for small numbers. The default
18265 is to use the fast divide at @option{-O3} and above.
18269 @itemx -mno-hw-mulx
18273 @opindex mno-hw-mul
18275 @opindex mno-hw-mulx
18277 @opindex mno-hw-div
18279 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
18280 instructions by the compiler. The default is to emit @code{mul}
18281 and not emit @code{div} and @code{mulx}.
18283 @item -mcustom-@var{insn}=@var{N}
18284 @itemx -mno-custom-@var{insn}
18285 @opindex mcustom-@var{insn}
18286 @opindex mno-custom-@var{insn}
18287 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
18288 custom instruction with encoding @var{N} when generating code that uses
18289 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
18290 instruction 253 for single-precision floating-point add operations instead
18291 of the default behavior of using a library call.
18293 The following values of @var{insn} are supported. Except as otherwise
18294 noted, floating-point operations are expected to be implemented with
18295 normal IEEE 754 semantics and correspond directly to the C operators or the
18296 equivalent GCC built-in functions (@pxref{Other Builtins}).
18298 Single-precision floating point:
18301 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
18302 Binary arithmetic operations.
18308 Unary absolute value.
18310 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
18311 Comparison operations.
18313 @item @samp{fmins}, @samp{fmaxs}
18314 Floating-point minimum and maximum. These instructions are only
18315 generated if @option{-ffinite-math-only} is specified.
18317 @item @samp{fsqrts}
18318 Unary square root operation.
18320 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
18321 Floating-point trigonometric and exponential functions. These instructions
18322 are only generated if @option{-funsafe-math-optimizations} is also specified.
18326 Double-precision floating point:
18329 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
18330 Binary arithmetic operations.
18336 Unary absolute value.
18338 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
18339 Comparison operations.
18341 @item @samp{fmind}, @samp{fmaxd}
18342 Double-precision minimum and maximum. These instructions are only
18343 generated if @option{-ffinite-math-only} is specified.
18345 @item @samp{fsqrtd}
18346 Unary square root operation.
18348 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
18349 Double-precision trigonometric and exponential functions. These instructions
18350 are only generated if @option{-funsafe-math-optimizations} is also specified.
18356 @item @samp{fextsd}
18357 Conversion from single precision to double precision.
18359 @item @samp{ftruncds}
18360 Conversion from double precision to single precision.
18362 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
18363 Conversion from floating point to signed or unsigned integer types, with
18364 truncation towards zero.
18367 Conversion from single-precision floating point to signed integer,
18368 rounding to the nearest integer and ties away from zero.
18369 This corresponds to the @code{__builtin_lroundf} function when
18370 @option{-fno-math-errno} is used.
18372 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
18373 Conversion from signed or unsigned integer types to floating-point types.
18377 In addition, all of the following transfer instructions for internal
18378 registers X and Y must be provided to use any of the double-precision
18379 floating-point instructions. Custom instructions taking two
18380 double-precision source operands expect the first operand in the
18381 64-bit register X. The other operand (or only operand of a unary
18382 operation) is given to the custom arithmetic instruction with the
18383 least significant half in source register @var{src1} and the most
18384 significant half in @var{src2}. A custom instruction that returns a
18385 double-precision result returns the most significant 32 bits in the
18386 destination register and the other half in 32-bit register Y.
18387 GCC automatically generates the necessary code sequences to write
18388 register X and/or read register Y when double-precision floating-point
18389 instructions are used.
18394 Write @var{src1} into the least significant half of X and @var{src2} into
18395 the most significant half of X.
18398 Write @var{src1} into Y.
18400 @item @samp{frdxhi}, @samp{frdxlo}
18401 Read the most or least (respectively) significant half of X and store it in
18405 Read the value of Y and store it into @var{dest}.
18408 Note that you can gain more local control over generation of Nios II custom
18409 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
18410 and @code{target("no-custom-@var{insn}")} function attributes
18411 (@pxref{Function Attributes})
18412 or pragmas (@pxref{Function Specific Option Pragmas}).
18414 @item -mcustom-fpu-cfg=@var{name}
18415 @opindex mcustom-fpu-cfg
18417 This option enables a predefined, named set of custom instruction encodings
18418 (see @option{-mcustom-@var{insn}} above).
18419 Currently, the following sets are defined:
18421 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
18422 @gccoptlist{-mcustom-fmuls=252 @gol
18423 -mcustom-fadds=253 @gol
18424 -mcustom-fsubs=254 @gol
18425 -fsingle-precision-constant}
18427 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
18428 @gccoptlist{-mcustom-fmuls=252 @gol
18429 -mcustom-fadds=253 @gol
18430 -mcustom-fsubs=254 @gol
18431 -mcustom-fdivs=255 @gol
18432 -fsingle-precision-constant}
18434 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
18435 @gccoptlist{-mcustom-floatus=243 @gol
18436 -mcustom-fixsi=244 @gol
18437 -mcustom-floatis=245 @gol
18438 -mcustom-fcmpgts=246 @gol
18439 -mcustom-fcmples=249 @gol
18440 -mcustom-fcmpeqs=250 @gol
18441 -mcustom-fcmpnes=251 @gol
18442 -mcustom-fmuls=252 @gol
18443 -mcustom-fadds=253 @gol
18444 -mcustom-fsubs=254 @gol
18445 -mcustom-fdivs=255 @gol
18446 -fsingle-precision-constant}
18448 Custom instruction assignments given by individual
18449 @option{-mcustom-@var{insn}=} options override those given by
18450 @option{-mcustom-fpu-cfg=}, regardless of the
18451 order of the options on the command line.
18453 Note that you can gain more local control over selection of a FPU
18454 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
18455 function attribute (@pxref{Function Attributes})
18456 or pragma (@pxref{Function Specific Option Pragmas}).
18460 These additional @samp{-m} options are available for the Altera Nios II
18461 ELF (bare-metal) target:
18467 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
18468 startup and termination code, and is typically used in conjunction with
18469 @option{-msys-crt0=} to specify the location of the alternate startup code
18470 provided by the HAL BSP.
18474 Link with a limited version of the C library, @option{-lsmallc}, rather than
18477 @item -msys-crt0=@var{startfile}
18479 @var{startfile} is the file name of the startfile (crt0) to use
18480 when linking. This option is only useful in conjunction with @option{-mhal}.
18482 @item -msys-lib=@var{systemlib}
18484 @var{systemlib} is the library name of the library that provides
18485 low-level system calls required by the C library,
18486 e.g. @code{read} and @code{write}.
18487 This option is typically used to link with a library provided by a HAL BSP.
18491 @node Nvidia PTX Options
18492 @subsection Nvidia PTX Options
18493 @cindex Nvidia PTX options
18494 @cindex nvptx options
18496 These options are defined for Nvidia PTX:
18504 Generate code for 32-bit or 64-bit ABI.
18507 @opindex mmainkernel
18508 Link in code for a __main kernel. This is for stand-alone instead of
18509 offloading execution.
18513 @node PDP-11 Options
18514 @subsection PDP-11 Options
18515 @cindex PDP-11 Options
18517 These options are defined for the PDP-11:
18522 Use hardware FPP floating point. This is the default. (FIS floating
18523 point on the PDP-11/40 is not supported.)
18526 @opindex msoft-float
18527 Do not use hardware floating point.
18531 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
18535 Return floating-point results in memory. This is the default.
18539 Generate code for a PDP-11/40.
18543 Generate code for a PDP-11/45. This is the default.
18547 Generate code for a PDP-11/10.
18549 @item -mbcopy-builtin
18550 @opindex mbcopy-builtin
18551 Use inline @code{movmemhi} patterns for copying memory. This is the
18556 Do not use inline @code{movmemhi} patterns for copying memory.
18562 Use 16-bit @code{int}. This is the default.
18568 Use 32-bit @code{int}.
18571 @itemx -mno-float32
18573 @opindex mno-float32
18574 Use 64-bit @code{float}. This is the default.
18577 @itemx -mno-float64
18579 @opindex mno-float64
18580 Use 32-bit @code{float}.
18584 Use @code{abshi2} pattern. This is the default.
18588 Do not use @code{abshi2} pattern.
18590 @item -mbranch-expensive
18591 @opindex mbranch-expensive
18592 Pretend that branches are expensive. This is for experimenting with
18593 code generation only.
18595 @item -mbranch-cheap
18596 @opindex mbranch-cheap
18597 Do not pretend that branches are expensive. This is the default.
18601 Use Unix assembler syntax. This is the default when configured for
18602 @samp{pdp11-*-bsd}.
18606 Use DEC assembler syntax. This is the default when configured for any
18607 PDP-11 target other than @samp{pdp11-*-bsd}.
18610 @node picoChip Options
18611 @subsection picoChip Options
18612 @cindex picoChip options
18614 These @samp{-m} options are defined for picoChip implementations:
18618 @item -mae=@var{ae_type}
18620 Set the instruction set, register set, and instruction scheduling
18621 parameters for array element type @var{ae_type}. Supported values
18622 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
18624 @option{-mae=ANY} selects a completely generic AE type. Code
18625 generated with this option runs on any of the other AE types. The
18626 code is not as efficient as it would be if compiled for a specific
18627 AE type, and some types of operation (e.g., multiplication) do not
18628 work properly on all types of AE.
18630 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
18631 for compiled code, and is the default.
18633 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
18634 option may suffer from poor performance of byte (char) manipulation,
18635 since the DSP AE does not provide hardware support for byte load/stores.
18637 @item -msymbol-as-address
18638 Enable the compiler to directly use a symbol name as an address in a
18639 load/store instruction, without first loading it into a
18640 register. Typically, the use of this option generates larger
18641 programs, which run faster than when the option isn't used. However, the
18642 results vary from program to program, so it is left as a user option,
18643 rather than being permanently enabled.
18645 @item -mno-inefficient-warnings
18646 Disables warnings about the generation of inefficient code. These
18647 warnings can be generated, for example, when compiling code that
18648 performs byte-level memory operations on the MAC AE type. The MAC AE has
18649 no hardware support for byte-level memory operations, so all byte
18650 load/stores must be synthesized from word load/store operations. This is
18651 inefficient and a warning is generated to indicate
18652 that you should rewrite the code to avoid byte operations, or to target
18653 an AE type that has the necessary hardware support. This option disables
18658 @node PowerPC Options
18659 @subsection PowerPC Options
18660 @cindex PowerPC options
18662 These are listed under @xref{RS/6000 and PowerPC Options}.
18665 @subsection RL78 Options
18666 @cindex RL78 Options
18672 Links in additional target libraries to support operation within a
18679 Specifies the type of hardware multiplication support to be used. The
18680 default is @samp{none}, which uses software multiplication functions.
18681 The @samp{g13} option is for the hardware multiply/divide peripheral
18682 only on the RL78/G13 targets. The @samp{rl78} option is for the
18683 standard hardware multiplication defined in the RL78 software manual.
18685 @item -m64bit-doubles
18686 @itemx -m32bit-doubles
18687 @opindex m64bit-doubles
18688 @opindex m32bit-doubles
18689 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
18690 or 32 bits (@option{-m32bit-doubles}) in size. The default is
18691 @option{-m32bit-doubles}.
18695 @node RS/6000 and PowerPC Options
18696 @subsection IBM RS/6000 and PowerPC Options
18697 @cindex RS/6000 and PowerPC Options
18698 @cindex IBM RS/6000 and PowerPC Options
18700 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
18702 @item -mpowerpc-gpopt
18703 @itemx -mno-powerpc-gpopt
18704 @itemx -mpowerpc-gfxopt
18705 @itemx -mno-powerpc-gfxopt
18708 @itemx -mno-powerpc64
18712 @itemx -mno-popcntb
18714 @itemx -mno-popcntd
18723 @itemx -mno-hard-dfp
18724 @opindex mpowerpc-gpopt
18725 @opindex mno-powerpc-gpopt
18726 @opindex mpowerpc-gfxopt
18727 @opindex mno-powerpc-gfxopt
18728 @opindex mpowerpc64
18729 @opindex mno-powerpc64
18733 @opindex mno-popcntb
18735 @opindex mno-popcntd
18741 @opindex mno-mfpgpr
18743 @opindex mno-hard-dfp
18744 You use these options to specify which instructions are available on the
18745 processor you are using. The default value of these options is
18746 determined when configuring GCC@. Specifying the
18747 @option{-mcpu=@var{cpu_type}} overrides the specification of these
18748 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
18749 rather than the options listed above.
18751 Specifying @option{-mpowerpc-gpopt} allows
18752 GCC to use the optional PowerPC architecture instructions in the
18753 General Purpose group, including floating-point square root. Specifying
18754 @option{-mpowerpc-gfxopt} allows GCC to
18755 use the optional PowerPC architecture instructions in the Graphics
18756 group, including floating-point select.
18758 The @option{-mmfcrf} option allows GCC to generate the move from
18759 condition register field instruction implemented on the POWER4
18760 processor and other processors that support the PowerPC V2.01
18762 The @option{-mpopcntb} option allows GCC to generate the popcount and
18763 double-precision FP reciprocal estimate instruction implemented on the
18764 POWER5 processor and other processors that support the PowerPC V2.02
18766 The @option{-mpopcntd} option allows GCC to generate the popcount
18767 instruction implemented on the POWER7 processor and other processors
18768 that support the PowerPC V2.06 architecture.
18769 The @option{-mfprnd} option allows GCC to generate the FP round to
18770 integer instructions implemented on the POWER5+ processor and other
18771 processors that support the PowerPC V2.03 architecture.
18772 The @option{-mcmpb} option allows GCC to generate the compare bytes
18773 instruction implemented on the POWER6 processor and other processors
18774 that support the PowerPC V2.05 architecture.
18775 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
18776 general-purpose register instructions implemented on the POWER6X
18777 processor and other processors that support the extended PowerPC V2.05
18779 The @option{-mhard-dfp} option allows GCC to generate the decimal
18780 floating-point instructions implemented on some POWER processors.
18782 The @option{-mpowerpc64} option allows GCC to generate the additional
18783 64-bit instructions that are found in the full PowerPC64 architecture
18784 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
18785 @option{-mno-powerpc64}.
18787 @item -mcpu=@var{cpu_type}
18789 Set architecture type, register usage, and
18790 instruction scheduling parameters for machine type @var{cpu_type}.
18791 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
18792 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
18793 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
18794 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
18795 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
18796 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
18797 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
18798 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
18799 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
18800 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8}, @samp{powerpc},
18801 @samp{powerpc64}, @samp{powerpc64le}, and @samp{rs64}.
18803 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
18804 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
18805 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
18806 architecture machine types, with an appropriate, generic processor
18807 model assumed for scheduling purposes.
18809 The other options specify a specific processor. Code generated under
18810 those options runs best on that processor, and may not run at all on
18813 The @option{-mcpu} options automatically enable or disable the
18816 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
18817 -mpopcntb -mpopcntd -mpowerpc64 @gol
18818 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
18819 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
18820 -mcrypto -mdirect-move -mpower8-fusion -mpower8-vector @gol
18821 -mquad-memory -mquad-memory-atomic}
18823 The particular options set for any particular CPU varies between
18824 compiler versions, depending on what setting seems to produce optimal
18825 code for that CPU; it doesn't necessarily reflect the actual hardware's
18826 capabilities. If you wish to set an individual option to a particular
18827 value, you may specify it after the @option{-mcpu} option, like
18828 @option{-mcpu=970 -mno-altivec}.
18830 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
18831 not enabled or disabled by the @option{-mcpu} option at present because
18832 AIX does not have full support for these options. You may still
18833 enable or disable them individually if you're sure it'll work in your
18836 @item -mtune=@var{cpu_type}
18838 Set the instruction scheduling parameters for machine type
18839 @var{cpu_type}, but do not set the architecture type or register usage,
18840 as @option{-mcpu=@var{cpu_type}} does. The same
18841 values for @var{cpu_type} are used for @option{-mtune} as for
18842 @option{-mcpu}. If both are specified, the code generated uses the
18843 architecture and registers set by @option{-mcpu}, but the
18844 scheduling parameters set by @option{-mtune}.
18846 @item -mcmodel=small
18847 @opindex mcmodel=small
18848 Generate PowerPC64 code for the small model: The TOC is limited to
18851 @item -mcmodel=medium
18852 @opindex mcmodel=medium
18853 Generate PowerPC64 code for the medium model: The TOC and other static
18854 data may be up to a total of 4G in size.
18856 @item -mcmodel=large
18857 @opindex mcmodel=large
18858 Generate PowerPC64 code for the large model: The TOC may be up to 4G
18859 in size. Other data and code is only limited by the 64-bit address
18863 @itemx -mno-altivec
18865 @opindex mno-altivec
18866 Generate code that uses (does not use) AltiVec instructions, and also
18867 enable the use of built-in functions that allow more direct access to
18868 the AltiVec instruction set. You may also need to set
18869 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
18872 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
18873 @option{-maltivec=be}, the element order for Altivec intrinsics such
18874 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
18875 match array element order corresponding to the endianness of the
18876 target. That is, element zero identifies the leftmost element in a
18877 vector register when targeting a big-endian platform, and identifies
18878 the rightmost element in a vector register when targeting a
18879 little-endian platform.
18882 @opindex maltivec=be
18883 Generate Altivec instructions using big-endian element order,
18884 regardless of whether the target is big- or little-endian. This is
18885 the default when targeting a big-endian platform.
18887 The element order is used to interpret element numbers in Altivec
18888 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
18889 @code{vec_insert}. By default, these match array element order
18890 corresponding to the endianness for the target.
18893 @opindex maltivec=le
18894 Generate Altivec instructions using little-endian element order,
18895 regardless of whether the target is big- or little-endian. This is
18896 the default when targeting a little-endian platform. This option is
18897 currently ignored when targeting a big-endian platform.
18899 The element order is used to interpret element numbers in Altivec
18900 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
18901 @code{vec_insert}. By default, these match array element order
18902 corresponding to the endianness for the target.
18907 @opindex mno-vrsave
18908 Generate VRSAVE instructions when generating AltiVec code.
18910 @item -mgen-cell-microcode
18911 @opindex mgen-cell-microcode
18912 Generate Cell microcode instructions.
18914 @item -mwarn-cell-microcode
18915 @opindex mwarn-cell-microcode
18916 Warn when a Cell microcode instruction is emitted. An example
18917 of a Cell microcode instruction is a variable shift.
18920 @opindex msecure-plt
18921 Generate code that allows @command{ld} and @command{ld.so}
18922 to build executables and shared
18923 libraries with non-executable @code{.plt} and @code{.got} sections.
18925 32-bit SYSV ABI option.
18929 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
18931 requires @code{.plt} and @code{.got}
18932 sections that are both writable and executable.
18933 This is a PowerPC 32-bit SYSV ABI option.
18939 This switch enables or disables the generation of ISEL instructions.
18941 @item -misel=@var{yes/no}
18942 This switch has been deprecated. Use @option{-misel} and
18943 @option{-mno-isel} instead.
18949 This switch enables or disables the generation of SPE simd
18955 @opindex mno-paired
18956 This switch enables or disables the generation of PAIRED simd
18959 @item -mspe=@var{yes/no}
18960 This option has been deprecated. Use @option{-mspe} and
18961 @option{-mno-spe} instead.
18967 Generate code that uses (does not use) vector/scalar (VSX)
18968 instructions, and also enable the use of built-in functions that allow
18969 more direct access to the VSX instruction set.
18974 @opindex mno-crypto
18975 Enable the use (disable) of the built-in functions that allow direct
18976 access to the cryptographic instructions that were added in version
18977 2.07 of the PowerPC ISA.
18979 @item -mdirect-move
18980 @itemx -mno-direct-move
18981 @opindex mdirect-move
18982 @opindex mno-direct-move
18983 Generate code that uses (does not use) the instructions to move data
18984 between the general purpose registers and the vector/scalar (VSX)
18985 registers that were added in version 2.07 of the PowerPC ISA.
18987 @item -mpower8-fusion
18988 @itemx -mno-power8-fusion
18989 @opindex mpower8-fusion
18990 @opindex mno-power8-fusion
18991 Generate code that keeps (does not keeps) some integer operations
18992 adjacent so that the instructions can be fused together on power8 and
18995 @item -mpower8-vector
18996 @itemx -mno-power8-vector
18997 @opindex mpower8-vector
18998 @opindex mno-power8-vector
18999 Generate code that uses (does not use) the vector and scalar
19000 instructions that were added in version 2.07 of the PowerPC ISA. Also
19001 enable the use of built-in functions that allow more direct access to
19002 the vector instructions.
19004 @item -mquad-memory
19005 @itemx -mno-quad-memory
19006 @opindex mquad-memory
19007 @opindex mno-quad-memory
19008 Generate code that uses (does not use) the non-atomic quad word memory
19009 instructions. The @option{-mquad-memory} option requires use of
19012 @item -mquad-memory-atomic
19013 @itemx -mno-quad-memory-atomic
19014 @opindex mquad-memory-atomic
19015 @opindex mno-quad-memory-atomic
19016 Generate code that uses (does not use) the atomic quad word memory
19017 instructions. The @option{-mquad-memory-atomic} option requires use of
19020 @item -mupper-regs-df
19021 @itemx -mno-upper-regs-df
19022 @opindex mupper-regs-df
19023 @opindex mno-upper-regs-df
19024 Generate code that uses (does not use) the scalar double precision
19025 instructions that target all 64 registers in the vector/scalar
19026 floating point register set that were added in version 2.06 of the
19027 PowerPC ISA. The @option{-mupper-regs-df} turned on by default if you
19028 use either of the @option{-mcpu=power7}, @option{-mcpu=power8}, or
19029 @option{-mvsx} options.
19031 @item -mupper-regs-sf
19032 @itemx -mno-upper-regs-sf
19033 @opindex mupper-regs-sf
19034 @opindex mno-upper-regs-sf
19035 Generate code that uses (does not use) the scalar single precision
19036 instructions that target all 64 registers in the vector/scalar
19037 floating point register set that were added in version 2.07 of the
19038 PowerPC ISA. The @option{-mupper-regs-sf} turned on by default if you
19039 use either of the @option{-mcpu=power8}, or @option{-mpower8-vector}
19043 @itemx -mno-upper-regs
19044 @opindex mupper-regs
19045 @opindex mno-upper-regs
19046 Generate code that uses (does not use) the scalar
19047 instructions that target all 64 registers in the vector/scalar
19048 floating point register set, depending on the model of the machine.
19050 If the @option{-mno-upper-regs} option is used, it turns off both
19051 @option{-mupper-regs-sf} and @option{-mupper-regs-df} options.
19053 @item -mfloat-gprs=@var{yes/single/double/no}
19054 @itemx -mfloat-gprs
19055 @opindex mfloat-gprs
19056 This switch enables or disables the generation of floating-point
19057 operations on the general-purpose registers for architectures that
19060 The argument @samp{yes} or @samp{single} enables the use of
19061 single-precision floating-point operations.
19063 The argument @samp{double} enables the use of single and
19064 double-precision floating-point operations.
19066 The argument @samp{no} disables floating-point operations on the
19067 general-purpose registers.
19069 This option is currently only available on the MPC854x.
19075 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
19076 targets (including GNU/Linux). The 32-bit environment sets int, long
19077 and pointer to 32 bits and generates code that runs on any PowerPC
19078 variant. The 64-bit environment sets int to 32 bits and long and
19079 pointer to 64 bits, and generates code for PowerPC64, as for
19080 @option{-mpowerpc64}.
19083 @itemx -mno-fp-in-toc
19084 @itemx -mno-sum-in-toc
19085 @itemx -mminimal-toc
19087 @opindex mno-fp-in-toc
19088 @opindex mno-sum-in-toc
19089 @opindex mminimal-toc
19090 Modify generation of the TOC (Table Of Contents), which is created for
19091 every executable file. The @option{-mfull-toc} option is selected by
19092 default. In that case, GCC allocates at least one TOC entry for
19093 each unique non-automatic variable reference in your program. GCC
19094 also places floating-point constants in the TOC@. However, only
19095 16,384 entries are available in the TOC@.
19097 If you receive a linker error message that saying you have overflowed
19098 the available TOC space, you can reduce the amount of TOC space used
19099 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
19100 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
19101 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
19102 generate code to calculate the sum of an address and a constant at
19103 run time instead of putting that sum into the TOC@. You may specify one
19104 or both of these options. Each causes GCC to produce very slightly
19105 slower and larger code at the expense of conserving TOC space.
19107 If you still run out of space in the TOC even when you specify both of
19108 these options, specify @option{-mminimal-toc} instead. This option causes
19109 GCC to make only one TOC entry for every file. When you specify this
19110 option, GCC produces code that is slower and larger but which
19111 uses extremely little TOC space. You may wish to use this option
19112 only on files that contain less frequently-executed code.
19118 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
19119 @code{long} type, and the infrastructure needed to support them.
19120 Specifying @option{-maix64} implies @option{-mpowerpc64},
19121 while @option{-maix32} disables the 64-bit ABI and
19122 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
19125 @itemx -mno-xl-compat
19126 @opindex mxl-compat
19127 @opindex mno-xl-compat
19128 Produce code that conforms more closely to IBM XL compiler semantics
19129 when using AIX-compatible ABI@. Pass floating-point arguments to
19130 prototyped functions beyond the register save area (RSA) on the stack
19131 in addition to argument FPRs. Do not assume that most significant
19132 double in 128-bit long double value is properly rounded when comparing
19133 values and converting to double. Use XL symbol names for long double
19136 The AIX calling convention was extended but not initially documented to
19137 handle an obscure K&R C case of calling a function that takes the
19138 address of its arguments with fewer arguments than declared. IBM XL
19139 compilers access floating-point arguments that do not fit in the
19140 RSA from the stack when a subroutine is compiled without
19141 optimization. Because always storing floating-point arguments on the
19142 stack is inefficient and rarely needed, this option is not enabled by
19143 default and only is necessary when calling subroutines compiled by IBM
19144 XL compilers without optimization.
19148 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
19149 application written to use message passing with special startup code to
19150 enable the application to run. The system must have PE installed in the
19151 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
19152 must be overridden with the @option{-specs=} option to specify the
19153 appropriate directory location. The Parallel Environment does not
19154 support threads, so the @option{-mpe} option and the @option{-pthread}
19155 option are incompatible.
19157 @item -malign-natural
19158 @itemx -malign-power
19159 @opindex malign-natural
19160 @opindex malign-power
19161 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
19162 @option{-malign-natural} overrides the ABI-defined alignment of larger
19163 types, such as floating-point doubles, on their natural size-based boundary.
19164 The option @option{-malign-power} instructs GCC to follow the ABI-specified
19165 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
19167 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
19171 @itemx -mhard-float
19172 @opindex msoft-float
19173 @opindex mhard-float
19174 Generate code that does not use (uses) the floating-point register set.
19175 Software floating-point emulation is provided if you use the
19176 @option{-msoft-float} option, and pass the option to GCC when linking.
19178 @item -msingle-float
19179 @itemx -mdouble-float
19180 @opindex msingle-float
19181 @opindex mdouble-float
19182 Generate code for single- or double-precision floating-point operations.
19183 @option{-mdouble-float} implies @option{-msingle-float}.
19186 @opindex msimple-fpu
19187 Do not generate @code{sqrt} and @code{div} instructions for hardware
19188 floating-point unit.
19190 @item -mfpu=@var{name}
19192 Specify type of floating-point unit. Valid values for @var{name} are
19193 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
19194 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
19195 @samp{sp_full} (equivalent to @option{-msingle-float}),
19196 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
19199 @opindex mxilinx-fpu
19200 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
19203 @itemx -mno-multiple
19205 @opindex mno-multiple
19206 Generate code that uses (does not use) the load multiple word
19207 instructions and the store multiple word instructions. These
19208 instructions are generated by default on POWER systems, and not
19209 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
19210 PowerPC systems, since those instructions do not work when the
19211 processor is in little-endian mode. The exceptions are PPC740 and
19212 PPC750 which permit these instructions in little-endian mode.
19217 @opindex mno-string
19218 Generate code that uses (does not use) the load string instructions
19219 and the store string word instructions to save multiple registers and
19220 do small block moves. These instructions are generated by default on
19221 POWER systems, and not generated on PowerPC systems. Do not use
19222 @option{-mstring} on little-endian PowerPC systems, since those
19223 instructions do not work when the processor is in little-endian mode.
19224 The exceptions are PPC740 and PPC750 which permit these instructions
19225 in little-endian mode.
19230 @opindex mno-update
19231 Generate code that uses (does not use) the load or store instructions
19232 that update the base register to the address of the calculated memory
19233 location. These instructions are generated by default. If you use
19234 @option{-mno-update}, there is a small window between the time that the
19235 stack pointer is updated and the address of the previous frame is
19236 stored, which means code that walks the stack frame across interrupts or
19237 signals may get corrupted data.
19239 @item -mavoid-indexed-addresses
19240 @itemx -mno-avoid-indexed-addresses
19241 @opindex mavoid-indexed-addresses
19242 @opindex mno-avoid-indexed-addresses
19243 Generate code that tries to avoid (not avoid) the use of indexed load
19244 or store instructions. These instructions can incur a performance
19245 penalty on Power6 processors in certain situations, such as when
19246 stepping through large arrays that cross a 16M boundary. This option
19247 is enabled by default when targeting Power6 and disabled otherwise.
19250 @itemx -mno-fused-madd
19251 @opindex mfused-madd
19252 @opindex mno-fused-madd
19253 Generate code that uses (does not use) the floating-point multiply and
19254 accumulate instructions. These instructions are generated by default
19255 if hardware floating point is used. The machine-dependent
19256 @option{-mfused-madd} option is now mapped to the machine-independent
19257 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
19258 mapped to @option{-ffp-contract=off}.
19264 Generate code that uses (does not use) the half-word multiply and
19265 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
19266 These instructions are generated by default when targeting those
19273 Generate code that uses (does not use) the string-search @samp{dlmzb}
19274 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
19275 generated by default when targeting those processors.
19277 @item -mno-bit-align
19279 @opindex mno-bit-align
19280 @opindex mbit-align
19281 On System V.4 and embedded PowerPC systems do not (do) force structures
19282 and unions that contain bit-fields to be aligned to the base type of the
19285 For example, by default a structure containing nothing but 8
19286 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
19287 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
19288 the structure is aligned to a 1-byte boundary and is 1 byte in
19291 @item -mno-strict-align
19292 @itemx -mstrict-align
19293 @opindex mno-strict-align
19294 @opindex mstrict-align
19295 On System V.4 and embedded PowerPC systems do not (do) assume that
19296 unaligned memory references are handled by the system.
19298 @item -mrelocatable
19299 @itemx -mno-relocatable
19300 @opindex mrelocatable
19301 @opindex mno-relocatable
19302 Generate code that allows (does not allow) a static executable to be
19303 relocated to a different address at run time. A simple embedded
19304 PowerPC system loader should relocate the entire contents of
19305 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
19306 a table of 32-bit addresses generated by this option. For this to
19307 work, all objects linked together must be compiled with
19308 @option{-mrelocatable} or @option{-mrelocatable-lib}.
19309 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
19311 @item -mrelocatable-lib
19312 @itemx -mno-relocatable-lib
19313 @opindex mrelocatable-lib
19314 @opindex mno-relocatable-lib
19315 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
19316 @code{.fixup} section to allow static executables to be relocated at
19317 run time, but @option{-mrelocatable-lib} does not use the smaller stack
19318 alignment of @option{-mrelocatable}. Objects compiled with
19319 @option{-mrelocatable-lib} may be linked with objects compiled with
19320 any combination of the @option{-mrelocatable} options.
19326 On System V.4 and embedded PowerPC systems do not (do) assume that
19327 register 2 contains a pointer to a global area pointing to the addresses
19328 used in the program.
19331 @itemx -mlittle-endian
19333 @opindex mlittle-endian
19334 On System V.4 and embedded PowerPC systems compile code for the
19335 processor in little-endian mode. The @option{-mlittle-endian} option is
19336 the same as @option{-mlittle}.
19339 @itemx -mbig-endian
19341 @opindex mbig-endian
19342 On System V.4 and embedded PowerPC systems compile code for the
19343 processor in big-endian mode. The @option{-mbig-endian} option is
19344 the same as @option{-mbig}.
19346 @item -mdynamic-no-pic
19347 @opindex mdynamic-no-pic
19348 On Darwin and Mac OS X systems, compile code so that it is not
19349 relocatable, but that its external references are relocatable. The
19350 resulting code is suitable for applications, but not shared
19353 @item -msingle-pic-base
19354 @opindex msingle-pic-base
19355 Treat the register used for PIC addressing as read-only, rather than
19356 loading it in the prologue for each function. The runtime system is
19357 responsible for initializing this register with an appropriate value
19358 before execution begins.
19360 @item -mprioritize-restricted-insns=@var{priority}
19361 @opindex mprioritize-restricted-insns
19362 This option controls the priority that is assigned to
19363 dispatch-slot restricted instructions during the second scheduling
19364 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
19365 or @samp{2} to assign no, highest, or second-highest (respectively)
19366 priority to dispatch-slot restricted
19369 @item -msched-costly-dep=@var{dependence_type}
19370 @opindex msched-costly-dep
19371 This option controls which dependences are considered costly
19372 by the target during instruction scheduling. The argument
19373 @var{dependence_type} takes one of the following values:
19377 No dependence is costly.
19380 All dependences are costly.
19382 @item @samp{true_store_to_load}
19383 A true dependence from store to load is costly.
19385 @item @samp{store_to_load}
19386 Any dependence from store to load is costly.
19389 Any dependence for which the latency is greater than or equal to
19390 @var{number} is costly.
19393 @item -minsert-sched-nops=@var{scheme}
19394 @opindex minsert-sched-nops
19395 This option controls which NOP insertion scheme is used during
19396 the second scheduling pass. The argument @var{scheme} takes one of the
19404 Pad with NOPs any dispatch group that has vacant issue slots,
19405 according to the scheduler's grouping.
19407 @item @samp{regroup_exact}
19408 Insert NOPs to force costly dependent insns into
19409 separate groups. Insert exactly as many NOPs as needed to force an insn
19410 to a new group, according to the estimated processor grouping.
19413 Insert NOPs to force costly dependent insns into
19414 separate groups. Insert @var{number} NOPs to force an insn to a new group.
19418 @opindex mcall-sysv
19419 On System V.4 and embedded PowerPC systems compile code using calling
19420 conventions that adhere to the March 1995 draft of the System V
19421 Application Binary Interface, PowerPC processor supplement. This is the
19422 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
19424 @item -mcall-sysv-eabi
19426 @opindex mcall-sysv-eabi
19427 @opindex mcall-eabi
19428 Specify both @option{-mcall-sysv} and @option{-meabi} options.
19430 @item -mcall-sysv-noeabi
19431 @opindex mcall-sysv-noeabi
19432 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
19434 @item -mcall-aixdesc
19436 On System V.4 and embedded PowerPC systems compile code for the AIX
19440 @opindex mcall-linux
19441 On System V.4 and embedded PowerPC systems compile code for the
19442 Linux-based GNU system.
19444 @item -mcall-freebsd
19445 @opindex mcall-freebsd
19446 On System V.4 and embedded PowerPC systems compile code for the
19447 FreeBSD operating system.
19449 @item -mcall-netbsd
19450 @opindex mcall-netbsd
19451 On System V.4 and embedded PowerPC systems compile code for the
19452 NetBSD operating system.
19454 @item -mcall-openbsd
19455 @opindex mcall-netbsd
19456 On System V.4 and embedded PowerPC systems compile code for the
19457 OpenBSD operating system.
19459 @item -maix-struct-return
19460 @opindex maix-struct-return
19461 Return all structures in memory (as specified by the AIX ABI)@.
19463 @item -msvr4-struct-return
19464 @opindex msvr4-struct-return
19465 Return structures smaller than 8 bytes in registers (as specified by the
19468 @item -mabi=@var{abi-type}
19470 Extend the current ABI with a particular extension, or remove such extension.
19471 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
19472 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
19473 @samp{elfv1}, @samp{elfv2}@.
19477 Extend the current ABI with SPE ABI extensions. This does not change
19478 the default ABI, instead it adds the SPE ABI extensions to the current
19482 @opindex mabi=no-spe
19483 Disable Book-E SPE ABI extensions for the current ABI@.
19485 @item -mabi=ibmlongdouble
19486 @opindex mabi=ibmlongdouble
19487 Change the current ABI to use IBM extended-precision long double.
19488 This is a PowerPC 32-bit SYSV ABI option.
19490 @item -mabi=ieeelongdouble
19491 @opindex mabi=ieeelongdouble
19492 Change the current ABI to use IEEE extended-precision long double.
19493 This is a PowerPC 32-bit Linux ABI option.
19496 @opindex mabi=elfv1
19497 Change the current ABI to use the ELFv1 ABI.
19498 This is the default ABI for big-endian PowerPC 64-bit Linux.
19499 Overriding the default ABI requires special system support and is
19500 likely to fail in spectacular ways.
19503 @opindex mabi=elfv2
19504 Change the current ABI to use the ELFv2 ABI.
19505 This is the default ABI for little-endian PowerPC 64-bit Linux.
19506 Overriding the default ABI requires special system support and is
19507 likely to fail in spectacular ways.
19510 @itemx -mno-prototype
19511 @opindex mprototype
19512 @opindex mno-prototype
19513 On System V.4 and embedded PowerPC systems assume that all calls to
19514 variable argument functions are properly prototyped. Otherwise, the
19515 compiler must insert an instruction before every non-prototyped call to
19516 set or clear bit 6 of the condition code register (@code{CR}) to
19517 indicate whether floating-point values are passed in the floating-point
19518 registers in case the function takes variable arguments. With
19519 @option{-mprototype}, only calls to prototyped variable argument functions
19520 set or clear the bit.
19524 On embedded PowerPC systems, assume that the startup module is called
19525 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
19526 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
19531 On embedded PowerPC systems, assume that the startup module is called
19532 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
19537 On embedded PowerPC systems, assume that the startup module is called
19538 @file{crt0.o} and the standard C libraries are @file{libads.a} and
19541 @item -myellowknife
19542 @opindex myellowknife
19543 On embedded PowerPC systems, assume that the startup module is called
19544 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
19549 On System V.4 and embedded PowerPC systems, specify that you are
19550 compiling for a VxWorks system.
19554 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
19555 header to indicate that @samp{eabi} extended relocations are used.
19561 On System V.4 and embedded PowerPC systems do (do not) adhere to the
19562 Embedded Applications Binary Interface (EABI), which is a set of
19563 modifications to the System V.4 specifications. Selecting @option{-meabi}
19564 means that the stack is aligned to an 8-byte boundary, a function
19565 @code{__eabi} is called from @code{main} to set up the EABI
19566 environment, and the @option{-msdata} option can use both @code{r2} and
19567 @code{r13} to point to two separate small data areas. Selecting
19568 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
19569 no EABI initialization function is called from @code{main}, and the
19570 @option{-msdata} option only uses @code{r13} to point to a single
19571 small data area. The @option{-meabi} option is on by default if you
19572 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
19575 @opindex msdata=eabi
19576 On System V.4 and embedded PowerPC systems, put small initialized
19577 @code{const} global and static data in the @code{.sdata2} section, which
19578 is pointed to by register @code{r2}. Put small initialized
19579 non-@code{const} global and static data in the @code{.sdata} section,
19580 which is pointed to by register @code{r13}. Put small uninitialized
19581 global and static data in the @code{.sbss} section, which is adjacent to
19582 the @code{.sdata} section. The @option{-msdata=eabi} option is
19583 incompatible with the @option{-mrelocatable} option. The
19584 @option{-msdata=eabi} option also sets the @option{-memb} option.
19587 @opindex msdata=sysv
19588 On System V.4 and embedded PowerPC systems, put small global and static
19589 data in the @code{.sdata} section, which is pointed to by register
19590 @code{r13}. Put small uninitialized global and static data in the
19591 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
19592 The @option{-msdata=sysv} option is incompatible with the
19593 @option{-mrelocatable} option.
19595 @item -msdata=default
19597 @opindex msdata=default
19599 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
19600 compile code the same as @option{-msdata=eabi}, otherwise compile code the
19601 same as @option{-msdata=sysv}.
19604 @opindex msdata=data
19605 On System V.4 and embedded PowerPC systems, put small global
19606 data in the @code{.sdata} section. Put small uninitialized global
19607 data in the @code{.sbss} section. Do not use register @code{r13}
19608 to address small data however. This is the default behavior unless
19609 other @option{-msdata} options are used.
19613 @opindex msdata=none
19615 On embedded PowerPC systems, put all initialized global and static data
19616 in the @code{.data} section, and all uninitialized data in the
19617 @code{.bss} section.
19619 @item -mblock-move-inline-limit=@var{num}
19620 @opindex mblock-move-inline-limit
19621 Inline all block moves (such as calls to @code{memcpy} or structure
19622 copies) less than or equal to @var{num} bytes. The minimum value for
19623 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
19624 targets. The default value is target-specific.
19628 @cindex smaller data references (PowerPC)
19629 @cindex .sdata/.sdata2 references (PowerPC)
19630 On embedded PowerPC systems, put global and static items less than or
19631 equal to @var{num} bytes into the small data or BSS sections instead of
19632 the normal data or BSS section. By default, @var{num} is 8. The
19633 @option{-G @var{num}} switch is also passed to the linker.
19634 All modules should be compiled with the same @option{-G @var{num}} value.
19637 @itemx -mno-regnames
19639 @opindex mno-regnames
19640 On System V.4 and embedded PowerPC systems do (do not) emit register
19641 names in the assembly language output using symbolic forms.
19644 @itemx -mno-longcall
19646 @opindex mno-longcall
19647 By default assume that all calls are far away so that a longer and more
19648 expensive calling sequence is required. This is required for calls
19649 farther than 32 megabytes (33,554,432 bytes) from the current location.
19650 A short call is generated if the compiler knows
19651 the call cannot be that far away. This setting can be overridden by
19652 the @code{shortcall} function attribute, or by @code{#pragma
19655 Some linkers are capable of detecting out-of-range calls and generating
19656 glue code on the fly. On these systems, long calls are unnecessary and
19657 generate slower code. As of this writing, the AIX linker can do this,
19658 as can the GNU linker for PowerPC/64. It is planned to add this feature
19659 to the GNU linker for 32-bit PowerPC systems as well.
19661 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
19662 callee, L42}, plus a @dfn{branch island} (glue code). The two target
19663 addresses represent the callee and the branch island. The
19664 Darwin/PPC linker prefers the first address and generates a @code{bl
19665 callee} if the PPC @code{bl} instruction reaches the callee directly;
19666 otherwise, the linker generates @code{bl L42} to call the branch
19667 island. The branch island is appended to the body of the
19668 calling function; it computes the full 32-bit address of the callee
19671 On Mach-O (Darwin) systems, this option directs the compiler emit to
19672 the glue for every direct call, and the Darwin linker decides whether
19673 to use or discard it.
19675 In the future, GCC may ignore all longcall specifications
19676 when the linker is known to generate glue.
19678 @item -mtls-markers
19679 @itemx -mno-tls-markers
19680 @opindex mtls-markers
19681 @opindex mno-tls-markers
19682 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
19683 specifying the function argument. The relocation allows the linker to
19684 reliably associate function call with argument setup instructions for
19685 TLS optimization, which in turn allows GCC to better schedule the
19690 Adds support for multithreading with the @dfn{pthreads} library.
19691 This option sets flags for both the preprocessor and linker.
19696 This option enables use of the reciprocal estimate and
19697 reciprocal square root estimate instructions with additional
19698 Newton-Raphson steps to increase precision instead of doing a divide or
19699 square root and divide for floating-point arguments. You should use
19700 the @option{-ffast-math} option when using @option{-mrecip} (or at
19701 least @option{-funsafe-math-optimizations},
19702 @option{-finite-math-only}, @option{-freciprocal-math} and
19703 @option{-fno-trapping-math}). Note that while the throughput of the
19704 sequence is generally higher than the throughput of the non-reciprocal
19705 instruction, the precision of the sequence can be decreased by up to 2
19706 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
19709 @item -mrecip=@var{opt}
19710 @opindex mrecip=opt
19711 This option controls which reciprocal estimate instructions
19712 may be used. @var{opt} is a comma-separated list of options, which may
19713 be preceded by a @code{!} to invert the option:
19718 Enable all estimate instructions.
19721 Enable the default instructions, equivalent to @option{-mrecip}.
19724 Disable all estimate instructions, equivalent to @option{-mno-recip}.
19727 Enable the reciprocal approximation instructions for both
19728 single and double precision.
19731 Enable the single-precision reciprocal approximation instructions.
19734 Enable the double-precision reciprocal approximation instructions.
19737 Enable the reciprocal square root approximation instructions for both
19738 single and double precision.
19741 Enable the single-precision reciprocal square root approximation instructions.
19744 Enable the double-precision reciprocal square root approximation instructions.
19748 So, for example, @option{-mrecip=all,!rsqrtd} enables
19749 all of the reciprocal estimate instructions, except for the
19750 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
19751 which handle the double-precision reciprocal square root calculations.
19753 @item -mrecip-precision
19754 @itemx -mno-recip-precision
19755 @opindex mrecip-precision
19756 Assume (do not assume) that the reciprocal estimate instructions
19757 provide higher-precision estimates than is mandated by the PowerPC
19758 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
19759 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
19760 The double-precision square root estimate instructions are not generated by
19761 default on low-precision machines, since they do not provide an
19762 estimate that converges after three steps.
19764 @item -mveclibabi=@var{type}
19765 @opindex mveclibabi
19766 Specifies the ABI type to use for vectorizing intrinsics using an
19767 external library. The only type supported at present is @samp{mass},
19768 which specifies to use IBM's Mathematical Acceleration Subsystem
19769 (MASS) libraries for vectorizing intrinsics using external libraries.
19770 GCC currently emits calls to @code{acosd2}, @code{acosf4},
19771 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
19772 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
19773 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
19774 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
19775 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
19776 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
19777 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
19778 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
19779 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
19780 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
19781 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
19782 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
19783 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
19784 for power7. Both @option{-ftree-vectorize} and
19785 @option{-funsafe-math-optimizations} must also be enabled. The MASS
19786 libraries must be specified at link time.
19791 Generate (do not generate) the @code{friz} instruction when the
19792 @option{-funsafe-math-optimizations} option is used to optimize
19793 rounding of floating-point values to 64-bit integer and back to floating
19794 point. The @code{friz} instruction does not return the same value if
19795 the floating-point number is too large to fit in an integer.
19797 @item -mpointers-to-nested-functions
19798 @itemx -mno-pointers-to-nested-functions
19799 @opindex mpointers-to-nested-functions
19800 Generate (do not generate) code to load up the static chain register
19801 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
19802 systems where a function pointer points to a 3-word descriptor giving
19803 the function address, TOC value to be loaded in register @code{r2}, and
19804 static chain value to be loaded in register @code{r11}. The
19805 @option{-mpointers-to-nested-functions} is on by default. You cannot
19806 call through pointers to nested functions or pointers
19807 to functions compiled in other languages that use the static chain if
19808 you use the @option{-mno-pointers-to-nested-functions}.
19810 @item -msave-toc-indirect
19811 @itemx -mno-save-toc-indirect
19812 @opindex msave-toc-indirect
19813 Generate (do not generate) code to save the TOC value in the reserved
19814 stack location in the function prologue if the function calls through
19815 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
19816 saved in the prologue, it is saved just before the call through the
19817 pointer. The @option{-mno-save-toc-indirect} option is the default.
19819 @item -mcompat-align-parm
19820 @itemx -mno-compat-align-parm
19821 @opindex mcompat-align-parm
19822 Generate (do not generate) code to pass structure parameters with a
19823 maximum alignment of 64 bits, for compatibility with older versions
19826 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
19827 structure parameter on a 128-bit boundary when that structure contained
19828 a member requiring 128-bit alignment. This is corrected in more
19829 recent versions of GCC. This option may be used to generate code
19830 that is compatible with functions compiled with older versions of
19833 The @option{-mno-compat-align-parm} option is the default.
19837 @subsection RX Options
19840 These command-line options are defined for RX targets:
19843 @item -m64bit-doubles
19844 @itemx -m32bit-doubles
19845 @opindex m64bit-doubles
19846 @opindex m32bit-doubles
19847 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
19848 or 32 bits (@option{-m32bit-doubles}) in size. The default is
19849 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
19850 works on 32-bit values, which is why the default is
19851 @option{-m32bit-doubles}.
19857 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
19858 floating-point hardware. The default is enabled for the RX600
19859 series and disabled for the RX200 series.
19861 Floating-point instructions are only generated for 32-bit floating-point
19862 values, however, so the FPU hardware is not used for doubles if the
19863 @option{-m64bit-doubles} option is used.
19865 @emph{Note} If the @option{-fpu} option is enabled then
19866 @option{-funsafe-math-optimizations} is also enabled automatically.
19867 This is because the RX FPU instructions are themselves unsafe.
19869 @item -mcpu=@var{name}
19871 Selects the type of RX CPU to be targeted. Currently three types are
19872 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
19873 the specific @samp{RX610} CPU. The default is @samp{RX600}.
19875 The only difference between @samp{RX600} and @samp{RX610} is that the
19876 @samp{RX610} does not support the @code{MVTIPL} instruction.
19878 The @samp{RX200} series does not have a hardware floating-point unit
19879 and so @option{-nofpu} is enabled by default when this type is
19882 @item -mbig-endian-data
19883 @itemx -mlittle-endian-data
19884 @opindex mbig-endian-data
19885 @opindex mlittle-endian-data
19886 Store data (but not code) in the big-endian format. The default is
19887 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
19890 @item -msmall-data-limit=@var{N}
19891 @opindex msmall-data-limit
19892 Specifies the maximum size in bytes of global and static variables
19893 which can be placed into the small data area. Using the small data
19894 area can lead to smaller and faster code, but the size of area is
19895 limited and it is up to the programmer to ensure that the area does
19896 not overflow. Also when the small data area is used one of the RX's
19897 registers (usually @code{r13}) is reserved for use pointing to this
19898 area, so it is no longer available for use by the compiler. This
19899 could result in slower and/or larger code if variables are pushed onto
19900 the stack instead of being held in this register.
19902 Note, common variables (variables that have not been initialized) and
19903 constants are not placed into the small data area as they are assigned
19904 to other sections in the output executable.
19906 The default value is zero, which disables this feature. Note, this
19907 feature is not enabled by default with higher optimization levels
19908 (@option{-O2} etc) because of the potentially detrimental effects of
19909 reserving a register. It is up to the programmer to experiment and
19910 discover whether this feature is of benefit to their program. See the
19911 description of the @option{-mpid} option for a description of how the
19912 actual register to hold the small data area pointer is chosen.
19918 Use the simulator runtime. The default is to use the libgloss
19919 board-specific runtime.
19921 @item -mas100-syntax
19922 @itemx -mno-as100-syntax
19923 @opindex mas100-syntax
19924 @opindex mno-as100-syntax
19925 When generating assembler output use a syntax that is compatible with
19926 Renesas's AS100 assembler. This syntax can also be handled by the GAS
19927 assembler, but it has some restrictions so it is not generated by default.
19929 @item -mmax-constant-size=@var{N}
19930 @opindex mmax-constant-size
19931 Specifies the maximum size, in bytes, of a constant that can be used as
19932 an operand in a RX instruction. Although the RX instruction set does
19933 allow constants of up to 4 bytes in length to be used in instructions,
19934 a longer value equates to a longer instruction. Thus in some
19935 circumstances it can be beneficial to restrict the size of constants
19936 that are used in instructions. Constants that are too big are instead
19937 placed into a constant pool and referenced via register indirection.
19939 The value @var{N} can be between 0 and 4. A value of 0 (the default)
19940 or 4 means that constants of any size are allowed.
19944 Enable linker relaxation. Linker relaxation is a process whereby the
19945 linker attempts to reduce the size of a program by finding shorter
19946 versions of various instructions. Disabled by default.
19948 @item -mint-register=@var{N}
19949 @opindex mint-register
19950 Specify the number of registers to reserve for fast interrupt handler
19951 functions. The value @var{N} can be between 0 and 4. A value of 1
19952 means that register @code{r13} is reserved for the exclusive use
19953 of fast interrupt handlers. A value of 2 reserves @code{r13} and
19954 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
19955 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
19956 A value of 0, the default, does not reserve any registers.
19958 @item -msave-acc-in-interrupts
19959 @opindex msave-acc-in-interrupts
19960 Specifies that interrupt handler functions should preserve the
19961 accumulator register. This is only necessary if normal code might use
19962 the accumulator register, for example because it performs 64-bit
19963 multiplications. The default is to ignore the accumulator as this
19964 makes the interrupt handlers faster.
19970 Enables the generation of position independent data. When enabled any
19971 access to constant data is done via an offset from a base address
19972 held in a register. This allows the location of constant data to be
19973 determined at run time without requiring the executable to be
19974 relocated, which is a benefit to embedded applications with tight
19975 memory constraints. Data that can be modified is not affected by this
19978 Note, using this feature reserves a register, usually @code{r13}, for
19979 the constant data base address. This can result in slower and/or
19980 larger code, especially in complicated functions.
19982 The actual register chosen to hold the constant data base address
19983 depends upon whether the @option{-msmall-data-limit} and/or the
19984 @option{-mint-register} command-line options are enabled. Starting
19985 with register @code{r13} and proceeding downwards, registers are
19986 allocated first to satisfy the requirements of @option{-mint-register},
19987 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
19988 is possible for the small data area register to be @code{r8} if both
19989 @option{-mint-register=4} and @option{-mpid} are specified on the
19992 By default this feature is not enabled. The default can be restored
19993 via the @option{-mno-pid} command-line option.
19995 @item -mno-warn-multiple-fast-interrupts
19996 @itemx -mwarn-multiple-fast-interrupts
19997 @opindex mno-warn-multiple-fast-interrupts
19998 @opindex mwarn-multiple-fast-interrupts
19999 Prevents GCC from issuing a warning message if it finds more than one
20000 fast interrupt handler when it is compiling a file. The default is to
20001 issue a warning for each extra fast interrupt handler found, as the RX
20002 only supports one such interrupt.
20006 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
20007 has special significance to the RX port when used with the
20008 @code{interrupt} function attribute. This attribute indicates a
20009 function intended to process fast interrupts. GCC ensures
20010 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
20011 and/or @code{r13} and only provided that the normal use of the
20012 corresponding registers have been restricted via the
20013 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
20016 @node S/390 and zSeries Options
20017 @subsection S/390 and zSeries Options
20018 @cindex S/390 and zSeries Options
20020 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
20024 @itemx -msoft-float
20025 @opindex mhard-float
20026 @opindex msoft-float
20027 Use (do not use) the hardware floating-point instructions and registers
20028 for floating-point operations. When @option{-msoft-float} is specified,
20029 functions in @file{libgcc.a} are used to perform floating-point
20030 operations. When @option{-mhard-float} is specified, the compiler
20031 generates IEEE floating-point instructions. This is the default.
20034 @itemx -mno-hard-dfp
20036 @opindex mno-hard-dfp
20037 Use (do not use) the hardware decimal-floating-point instructions for
20038 decimal-floating-point operations. When @option{-mno-hard-dfp} is
20039 specified, functions in @file{libgcc.a} are used to perform
20040 decimal-floating-point operations. When @option{-mhard-dfp} is
20041 specified, the compiler generates decimal-floating-point hardware
20042 instructions. This is the default for @option{-march=z9-ec} or higher.
20044 @item -mlong-double-64
20045 @itemx -mlong-double-128
20046 @opindex mlong-double-64
20047 @opindex mlong-double-128
20048 These switches control the size of @code{long double} type. A size
20049 of 64 bits makes the @code{long double} type equivalent to the @code{double}
20050 type. This is the default.
20053 @itemx -mno-backchain
20054 @opindex mbackchain
20055 @opindex mno-backchain
20056 Store (do not store) the address of the caller's frame as backchain pointer
20057 into the callee's stack frame.
20058 A backchain may be needed to allow debugging using tools that do not understand
20059 DWARF 2 call frame information.
20060 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
20061 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
20062 the backchain is placed into the topmost word of the 96/160 byte register
20065 In general, code compiled with @option{-mbackchain} is call-compatible with
20066 code compiled with @option{-mmo-backchain}; however, use of the backchain
20067 for debugging purposes usually requires that the whole binary is built with
20068 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
20069 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20070 to build a linux kernel use @option{-msoft-float}.
20072 The default is to not maintain the backchain.
20074 @item -mpacked-stack
20075 @itemx -mno-packed-stack
20076 @opindex mpacked-stack
20077 @opindex mno-packed-stack
20078 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
20079 specified, the compiler uses the all fields of the 96/160 byte register save
20080 area only for their default purpose; unused fields still take up stack space.
20081 When @option{-mpacked-stack} is specified, register save slots are densely
20082 packed at the top of the register save area; unused space is reused for other
20083 purposes, allowing for more efficient use of the available stack space.
20084 However, when @option{-mbackchain} is also in effect, the topmost word of
20085 the save area is always used to store the backchain, and the return address
20086 register is always saved two words below the backchain.
20088 As long as the stack frame backchain is not used, code generated with
20089 @option{-mpacked-stack} is call-compatible with code generated with
20090 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
20091 S/390 or zSeries generated code that uses the stack frame backchain at run
20092 time, not just for debugging purposes. Such code is not call-compatible
20093 with code compiled with @option{-mpacked-stack}. Also, note that the
20094 combination of @option{-mbackchain},
20095 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20096 to build a linux kernel use @option{-msoft-float}.
20098 The default is to not use the packed stack layout.
20101 @itemx -mno-small-exec
20102 @opindex msmall-exec
20103 @opindex mno-small-exec
20104 Generate (or do not generate) code using the @code{bras} instruction
20105 to do subroutine calls.
20106 This only works reliably if the total executable size does not
20107 exceed 64k. The default is to use the @code{basr} instruction instead,
20108 which does not have this limitation.
20114 When @option{-m31} is specified, generate code compliant to the
20115 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
20116 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
20117 particular to generate 64-bit instructions. For the @samp{s390}
20118 targets, the default is @option{-m31}, while the @samp{s390x}
20119 targets default to @option{-m64}.
20125 When @option{-mzarch} is specified, generate code using the
20126 instructions available on z/Architecture.
20127 When @option{-mesa} is specified, generate code using the
20128 instructions available on ESA/390. Note that @option{-mesa} is
20129 not possible with @option{-m64}.
20130 When generating code compliant to the GNU/Linux for S/390 ABI,
20131 the default is @option{-mesa}. When generating code compliant
20132 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
20138 Generate (or do not generate) code using the @code{mvcle} instruction
20139 to perform block moves. When @option{-mno-mvcle} is specified,
20140 use a @code{mvc} loop instead. This is the default unless optimizing for
20147 Print (or do not print) additional debug information when compiling.
20148 The default is to not print debug information.
20150 @item -march=@var{cpu-type}
20152 Generate code that runs on @var{cpu-type}, which is the name of a system
20153 representing a certain processor type. Possible values for
20154 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
20155 @samp{z9-109}, @samp{z9-ec}, @samp{z10}, @samp{z196}, and @samp{zEC12}.
20156 When generating code using the instructions available on z/Architecture,
20157 the default is @option{-march=z900}. Otherwise, the default is
20158 @option{-march=g5}.
20160 @item -mtune=@var{cpu-type}
20162 Tune to @var{cpu-type} everything applicable about the generated code,
20163 except for the ABI and the set of available instructions.
20164 The list of @var{cpu-type} values is the same as for @option{-march}.
20165 The default is the value used for @option{-march}.
20168 @itemx -mno-tpf-trace
20169 @opindex mtpf-trace
20170 @opindex mno-tpf-trace
20171 Generate code that adds (does not add) in TPF OS specific branches to trace
20172 routines in the operating system. This option is off by default, even
20173 when compiling for the TPF OS@.
20176 @itemx -mno-fused-madd
20177 @opindex mfused-madd
20178 @opindex mno-fused-madd
20179 Generate code that uses (does not use) the floating-point multiply and
20180 accumulate instructions. These instructions are generated by default if
20181 hardware floating point is used.
20183 @item -mwarn-framesize=@var{framesize}
20184 @opindex mwarn-framesize
20185 Emit a warning if the current function exceeds the given frame size. Because
20186 this is a compile-time check it doesn't need to be a real problem when the program
20187 runs. It is intended to identify functions that most probably cause
20188 a stack overflow. It is useful to be used in an environment with limited stack
20189 size e.g.@: the linux kernel.
20191 @item -mwarn-dynamicstack
20192 @opindex mwarn-dynamicstack
20193 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
20194 arrays. This is generally a bad idea with a limited stack size.
20196 @item -mstack-guard=@var{stack-guard}
20197 @itemx -mstack-size=@var{stack-size}
20198 @opindex mstack-guard
20199 @opindex mstack-size
20200 If these options are provided the S/390 back end emits additional instructions in
20201 the function prologue that trigger a trap if the stack size is @var{stack-guard}
20202 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
20203 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
20204 the frame size of the compiled function is chosen.
20205 These options are intended to be used to help debugging stack overflow problems.
20206 The additionally emitted code causes only little overhead and hence can also be
20207 used in production-like systems without greater performance degradation. The given
20208 values have to be exact powers of 2 and @var{stack-size} has to be greater than
20209 @var{stack-guard} without exceeding 64k.
20210 In order to be efficient the extra code makes the assumption that the stack starts
20211 at an address aligned to the value given by @var{stack-size}.
20212 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
20214 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
20216 If the hotpatch option is enabled, a ``hot-patching'' function
20217 prologue is generated for all functions in the compilation unit.
20218 The funtion label is prepended with the given number of two-byte
20219 Nop instructions (@var{pre-halfwords}, maximum 1000000). After
20220 the label, 2 * @var{post-halfwords} bytes are appended, using the
20221 larges nop like instructions the architecture allows (maximum
20224 If both arguments are zero, hotpatching is disabled.
20226 This option can be overridden for individual functions with the
20227 @code{hotpatch} attribute.
20230 @node Score Options
20231 @subsection Score Options
20232 @cindex Score Options
20234 These options are defined for Score implementations:
20239 Compile code for big-endian mode. This is the default.
20243 Compile code for little-endian mode.
20247 Disable generation of @code{bcnz} instructions.
20251 Enable generation of unaligned load and store instructions.
20255 Enable the use of multiply-accumulate instructions. Disabled by default.
20259 Specify the SCORE5 as the target architecture.
20263 Specify the SCORE5U of the target architecture.
20267 Specify the SCORE7 as the target architecture. This is the default.
20271 Specify the SCORE7D as the target architecture.
20275 @subsection SH Options
20277 These @samp{-m} options are defined for the SH implementations:
20282 Generate code for the SH1.
20286 Generate code for the SH2.
20289 Generate code for the SH2e.
20293 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
20294 that the floating-point unit is not used.
20296 @item -m2a-single-only
20297 @opindex m2a-single-only
20298 Generate code for the SH2a-FPU, in such a way that no double-precision
20299 floating-point operations are used.
20302 @opindex m2a-single
20303 Generate code for the SH2a-FPU assuming the floating-point unit is in
20304 single-precision mode by default.
20308 Generate code for the SH2a-FPU assuming the floating-point unit is in
20309 double-precision mode by default.
20313 Generate code for the SH3.
20317 Generate code for the SH3e.
20321 Generate code for the SH4 without a floating-point unit.
20323 @item -m4-single-only
20324 @opindex m4-single-only
20325 Generate code for the SH4 with a floating-point unit that only
20326 supports single-precision arithmetic.
20330 Generate code for the SH4 assuming the floating-point unit is in
20331 single-precision mode by default.
20335 Generate code for the SH4.
20339 Generate code for SH4-100.
20341 @item -m4-100-nofpu
20342 @opindex m4-100-nofpu
20343 Generate code for SH4-100 in such a way that the
20344 floating-point unit is not used.
20346 @item -m4-100-single
20347 @opindex m4-100-single
20348 Generate code for SH4-100 assuming the floating-point unit is in
20349 single-precision mode by default.
20351 @item -m4-100-single-only
20352 @opindex m4-100-single-only
20353 Generate code for SH4-100 in such a way that no double-precision
20354 floating-point operations are used.
20358 Generate code for SH4-200.
20360 @item -m4-200-nofpu
20361 @opindex m4-200-nofpu
20362 Generate code for SH4-200 without in such a way that the
20363 floating-point unit is not used.
20365 @item -m4-200-single
20366 @opindex m4-200-single
20367 Generate code for SH4-200 assuming the floating-point unit is in
20368 single-precision mode by default.
20370 @item -m4-200-single-only
20371 @opindex m4-200-single-only
20372 Generate code for SH4-200 in such a way that no double-precision
20373 floating-point operations are used.
20377 Generate code for SH4-300.
20379 @item -m4-300-nofpu
20380 @opindex m4-300-nofpu
20381 Generate code for SH4-300 without in such a way that the
20382 floating-point unit is not used.
20384 @item -m4-300-single
20385 @opindex m4-300-single
20386 Generate code for SH4-300 in such a way that no double-precision
20387 floating-point operations are used.
20389 @item -m4-300-single-only
20390 @opindex m4-300-single-only
20391 Generate code for SH4-300 in such a way that no double-precision
20392 floating-point operations are used.
20396 Generate code for SH4-340 (no MMU, no FPU).
20400 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
20405 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
20406 floating-point unit is not used.
20408 @item -m4a-single-only
20409 @opindex m4a-single-only
20410 Generate code for the SH4a, in such a way that no double-precision
20411 floating-point operations are used.
20414 @opindex m4a-single
20415 Generate code for the SH4a assuming the floating-point unit is in
20416 single-precision mode by default.
20420 Generate code for the SH4a.
20424 Same as @option{-m4a-nofpu}, except that it implicitly passes
20425 @option{-dsp} to the assembler. GCC doesn't generate any DSP
20426 instructions at the moment.
20429 @opindex m5-32media
20430 Generate 32-bit code for SHmedia.
20432 @item -m5-32media-nofpu
20433 @opindex m5-32media-nofpu
20434 Generate 32-bit code for SHmedia in such a way that the
20435 floating-point unit is not used.
20438 @opindex m5-64media
20439 Generate 64-bit code for SHmedia.
20441 @item -m5-64media-nofpu
20442 @opindex m5-64media-nofpu
20443 Generate 64-bit code for SHmedia in such a way that the
20444 floating-point unit is not used.
20447 @opindex m5-compact
20448 Generate code for SHcompact.
20450 @item -m5-compact-nofpu
20451 @opindex m5-compact-nofpu
20452 Generate code for SHcompact in such a way that the
20453 floating-point unit is not used.
20457 Compile code for the processor in big-endian mode.
20461 Compile code for the processor in little-endian mode.
20465 Align doubles at 64-bit boundaries. Note that this changes the calling
20466 conventions, and thus some functions from the standard C library do
20467 not work unless you recompile it first with @option{-mdalign}.
20471 Shorten some address references at link time, when possible; uses the
20472 linker option @option{-relax}.
20476 Use 32-bit offsets in @code{switch} tables. The default is to use
20481 Enable the use of bit manipulation instructions on SH2A.
20485 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
20486 alignment constraints.
20490 Comply with the calling conventions defined by Renesas.
20493 @opindex mno-renesas
20494 Comply with the calling conventions defined for GCC before the Renesas
20495 conventions were available. This option is the default for all
20496 targets of the SH toolchain.
20499 @opindex mnomacsave
20500 Mark the @code{MAC} register as call-clobbered, even if
20501 @option{-mrenesas} is given.
20507 Control the IEEE compliance of floating-point comparisons, which affects the
20508 handling of cases where the result of a comparison is unordered. By default
20509 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
20510 enabled @option{-mno-ieee} is implicitly set, which results in faster
20511 floating-point greater-equal and less-equal comparisons. The implcit settings
20512 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
20514 @item -minline-ic_invalidate
20515 @opindex minline-ic_invalidate
20516 Inline code to invalidate instruction cache entries after setting up
20517 nested function trampolines.
20518 This option has no effect if @option{-musermode} is in effect and the selected
20519 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
20521 If the selected code generation option does not allow the use of the @code{icbi}
20522 instruction, and @option{-musermode} is not in effect, the inlined code
20523 manipulates the instruction cache address array directly with an associative
20524 write. This not only requires privileged mode at run time, but it also
20525 fails if the cache line had been mapped via the TLB and has become unmapped.
20529 Dump instruction size and location in the assembly code.
20532 @opindex mpadstruct
20533 This option is deprecated. It pads structures to multiple of 4 bytes,
20534 which is incompatible with the SH ABI@.
20536 @item -matomic-model=@var{model}
20537 @opindex matomic-model=@var{model}
20538 Sets the model of atomic operations and additional parameters as a comma
20539 separated list. For details on the atomic built-in functions see
20540 @ref{__atomic Builtins}. The following models and parameters are supported:
20545 Disable compiler generated atomic sequences and emit library calls for atomic
20546 operations. This is the default if the target is not @code{sh*-*-linux*}.
20549 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
20550 built-in functions. The generated atomic sequences require additional support
20551 from the interrupt/exception handling code of the system and are only suitable
20552 for SH3* and SH4* single-core systems. This option is enabled by default when
20553 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
20554 this option also partially utilizes the hardware atomic instructions
20555 @code{movli.l} and @code{movco.l} to create more efficient code, unless
20556 @samp{strict} is specified.
20559 Generate software atomic sequences that use a variable in the thread control
20560 block. This is a variation of the gUSA sequences which can also be used on
20561 SH1* and SH2* targets. The generated atomic sequences require additional
20562 support from the interrupt/exception handling code of the system and are only
20563 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
20564 parameter has to be specified as well.
20567 Generate software atomic sequences that temporarily disable interrupts by
20568 setting @code{SR.IMASK = 1111}. This model works only when the program runs
20569 in privileged mode and is only suitable for single-core systems. Additional
20570 support from the interrupt/exception handling code of the system is not
20571 required. This model is enabled by default when the target is
20572 @code{sh*-*-linux*} and SH1* or SH2*.
20575 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
20576 instructions only. This is only available on SH4A and is suitable for
20577 multi-core systems. Since the hardware instructions support only 32 bit atomic
20578 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
20579 Code compiled with this option is also compatible with other software
20580 atomic model interrupt/exception handling systems if executed on an SH4A
20581 system. Additional support from the interrupt/exception handling code of the
20582 system is not required for this model.
20585 This parameter specifies the offset in bytes of the variable in the thread
20586 control block structure that should be used by the generated atomic sequences
20587 when the @samp{soft-tcb} model has been selected. For other models this
20588 parameter is ignored. The specified value must be an integer multiple of four
20589 and in the range 0-1020.
20592 This parameter prevents mixed usage of multiple atomic models, even if they
20593 are compatible, and makes the compiler generate atomic sequences of the
20594 specified model only.
20600 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
20601 Notice that depending on the particular hardware and software configuration
20602 this can degrade overall performance due to the operand cache line flushes
20603 that are implied by the @code{tas.b} instruction. On multi-core SH4A
20604 processors the @code{tas.b} instruction must be used with caution since it
20605 can result in data corruption for certain cache configurations.
20608 @opindex mprefergot
20609 When generating position-independent code, emit function calls using
20610 the Global Offset Table instead of the Procedure Linkage Table.
20613 @itemx -mno-usermode
20615 @opindex mno-usermode
20616 Don't allow (allow) the compiler generating privileged mode code. Specifying
20617 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
20618 inlined code would not work in user mode. @option{-musermode} is the default
20619 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
20620 @option{-musermode} has no effect, since there is no user mode.
20622 @item -multcost=@var{number}
20623 @opindex multcost=@var{number}
20624 Set the cost to assume for a multiply insn.
20626 @item -mdiv=@var{strategy}
20627 @opindex mdiv=@var{strategy}
20628 Set the division strategy to be used for integer division operations.
20629 For SHmedia @var{strategy} can be one of:
20634 Performs the operation in floating point. This has a very high latency,
20635 but needs only a few instructions, so it might be a good choice if
20636 your code has enough easily-exploitable ILP to allow the compiler to
20637 schedule the floating-point instructions together with other instructions.
20638 Division by zero causes a floating-point exception.
20641 Uses integer operations to calculate the inverse of the divisor,
20642 and then multiplies the dividend with the inverse. This strategy allows
20643 CSE and hoisting of the inverse calculation. Division by zero calculates
20644 an unspecified result, but does not trap.
20647 A variant of @samp{inv} where, if no CSE or hoisting opportunities
20648 have been found, or if the entire operation has been hoisted to the same
20649 place, the last stages of the inverse calculation are intertwined with the
20650 final multiply to reduce the overall latency, at the expense of using a few
20651 more instructions, and thus offering fewer scheduling opportunities with
20655 Calls a library function that usually implements the @samp{inv:minlat}
20657 This gives high code density for @code{m5-*media-nofpu} compilations.
20660 Uses a different entry point of the same library function, where it
20661 assumes that a pointer to a lookup table has already been set up, which
20662 exposes the pointer load to CSE and code hoisting optimizations.
20667 Use the @samp{inv} algorithm for initial
20668 code generation, but if the code stays unoptimized, revert to the @samp{call},
20669 @samp{call2}, or @samp{fp} strategies, respectively. Note that the
20670 potentially-trapping side effect of division by zero is carried by a
20671 separate instruction, so it is possible that all the integer instructions
20672 are hoisted out, but the marker for the side effect stays where it is.
20673 A recombination to floating-point operations or a call is not possible
20678 Variants of the @samp{inv:minlat} strategy. In the case
20679 that the inverse calculation is not separated from the multiply, they speed
20680 up division where the dividend fits into 20 bits (plus sign where applicable)
20681 by inserting a test to skip a number of operations in this case; this test
20682 slows down the case of larger dividends. @samp{inv20u} assumes the case of a such
20683 a small dividend to be unlikely, and @samp{inv20l} assumes it to be likely.
20687 For targets other than SHmedia @var{strategy} can be one of:
20692 Calls a library function that uses the single-step division instruction
20693 @code{div1} to perform the operation. Division by zero calculates an
20694 unspecified result and does not trap. This is the default except for SH4,
20695 SH2A and SHcompact.
20698 Calls a library function that performs the operation in double precision
20699 floating point. Division by zero causes a floating-point exception. This is
20700 the default for SHcompact with FPU. Specifying this for targets that do not
20701 have a double precision FPU defaults to @code{call-div1}.
20704 Calls a library function that uses a lookup table for small divisors and
20705 the @code{div1} instruction with case distinction for larger divisors. Division
20706 by zero calculates an unspecified result and does not trap. This is the default
20707 for SH4. Specifying this for targets that do not have dynamic shift
20708 instructions defaults to @code{call-div1}.
20712 When a division strategy has not been specified the default strategy is
20713 selected based on the current target. For SH2A the default strategy is to
20714 use the @code{divs} and @code{divu} instructions instead of library function
20717 @item -maccumulate-outgoing-args
20718 @opindex maccumulate-outgoing-args
20719 Reserve space once for outgoing arguments in the function prologue rather
20720 than around each call. Generally beneficial for performance and size. Also
20721 needed for unwinding to avoid changing the stack frame around conditional code.
20723 @item -mdivsi3_libfunc=@var{name}
20724 @opindex mdivsi3_libfunc=@var{name}
20725 Set the name of the library function used for 32-bit signed division to
20727 This only affects the name used in the @samp{call} and @samp{inv:call}
20728 division strategies, and the compiler still expects the same
20729 sets of input/output/clobbered registers as if this option were not present.
20731 @item -mfixed-range=@var{register-range}
20732 @opindex mfixed-range
20733 Generate code treating the given register range as fixed registers.
20734 A fixed register is one that the register allocator can not use. This is
20735 useful when compiling kernel code. A register range is specified as
20736 two registers separated by a dash. Multiple register ranges can be
20737 specified separated by a comma.
20739 @item -mindexed-addressing
20740 @opindex mindexed-addressing
20741 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
20742 This is only safe if the hardware and/or OS implement 32-bit wrap-around
20743 semantics for the indexed addressing mode. The architecture allows the
20744 implementation of processors with 64-bit MMU, which the OS could use to
20745 get 32-bit addressing, but since no current hardware implementation supports
20746 this or any other way to make the indexed addressing mode safe to use in
20747 the 32-bit ABI, the default is @option{-mno-indexed-addressing}.
20749 @item -mgettrcost=@var{number}
20750 @opindex mgettrcost=@var{number}
20751 Set the cost assumed for the @code{gettr} instruction to @var{number}.
20752 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
20756 Assume @code{pt*} instructions won't trap. This generally generates
20757 better-scheduled code, but is unsafe on current hardware.
20758 The current architecture
20759 definition says that @code{ptabs} and @code{ptrel} trap when the target
20761 This has the unintentional effect of making it unsafe to schedule these
20762 instructions before a branch, or hoist them out of a loop. For example,
20763 @code{__do_global_ctors}, a part of @file{libgcc}
20764 that runs constructors at program
20765 startup, calls functions in a list which is delimited by @minus{}1. With the
20766 @option{-mpt-fixed} option, the @code{ptabs} is done before testing against @minus{}1.
20767 That means that all the constructors run a bit more quickly, but when
20768 the loop comes to the end of the list, the program crashes because @code{ptabs}
20769 loads @minus{}1 into a target register.
20771 Since this option is unsafe for any
20772 hardware implementing the current architecture specification, the default
20773 is @option{-mno-pt-fixed}. Unless specified explicitly with
20774 @option{-mgettrcost}, @option{-mno-pt-fixed} also implies @option{-mgettrcost=100};
20775 this deters register allocation from using target registers for storing
20778 @item -minvalid-symbols
20779 @opindex minvalid-symbols
20780 Assume symbols might be invalid. Ordinary function symbols generated by
20781 the compiler are always valid to load with
20782 @code{movi}/@code{shori}/@code{ptabs} or
20783 @code{movi}/@code{shori}/@code{ptrel},
20784 but with assembler and/or linker tricks it is possible
20785 to generate symbols that cause @code{ptabs} or @code{ptrel} to trap.
20786 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
20787 It prevents cross-basic-block CSE, hoisting and most scheduling
20788 of symbol loads. The default is @option{-mno-invalid-symbols}.
20790 @item -mbranch-cost=@var{num}
20791 @opindex mbranch-cost=@var{num}
20792 Assume @var{num} to be the cost for a branch instruction. Higher numbers
20793 make the compiler try to generate more branch-free code if possible.
20794 If not specified the value is selected depending on the processor type that
20795 is being compiled for.
20798 @itemx -mno-zdcbranch
20799 @opindex mzdcbranch
20800 @opindex mno-zdcbranch
20801 Assume (do not assume) that zero displacement conditional branch instructions
20802 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
20803 compiler prefers zero displacement branch code sequences. This is
20804 enabled by default when generating code for SH4 and SH4A. It can be explicitly
20805 disabled by specifying @option{-mno-zdcbranch}.
20807 @item -mcbranch-force-delay-slot
20808 @opindex mcbranch-force-delay-slot
20809 Force the usage of delay slots for conditional branches, which stuffs the delay
20810 slot with a @code{nop} if a suitable instruction can't be found. By default
20811 this option is disabled. It can be enabled to work around hardware bugs as
20812 found in the original SH7055.
20815 @itemx -mno-fused-madd
20816 @opindex mfused-madd
20817 @opindex mno-fused-madd
20818 Generate code that uses (does not use) the floating-point multiply and
20819 accumulate instructions. These instructions are generated by default
20820 if hardware floating point is used. The machine-dependent
20821 @option{-mfused-madd} option is now mapped to the machine-independent
20822 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
20823 mapped to @option{-ffp-contract=off}.
20829 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
20830 and cosine approximations. The option @option{-mfsca} must be used in
20831 combination with @option{-funsafe-math-optimizations}. It is enabled by default
20832 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
20833 approximations even if @option{-funsafe-math-optimizations} is in effect.
20839 Allow or disallow the compiler to emit the @code{fsrra} instruction for
20840 reciprocal square root approximations. The option @option{-mfsrra} must be used
20841 in combination with @option{-funsafe-math-optimizations} and
20842 @option{-ffinite-math-only}. It is enabled by default when generating code for
20843 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
20844 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
20847 @item -mpretend-cmove
20848 @opindex mpretend-cmove
20849 Prefer zero-displacement conditional branches for conditional move instruction
20850 patterns. This can result in faster code on the SH4 processor.
20854 @node Solaris 2 Options
20855 @subsection Solaris 2 Options
20856 @cindex Solaris 2 options
20858 These @samp{-m} options are supported on Solaris 2:
20861 @item -mclear-hwcap
20862 @opindex mclear-hwcap
20863 @option{-mclear-hwcap} tells the compiler to remove the hardware
20864 capabilities generated by the Solaris assembler. This is only necessary
20865 when object files use ISA extensions not supported by the current
20866 machine, but check at runtime whether or not to use them.
20868 @item -mimpure-text
20869 @opindex mimpure-text
20870 @option{-mimpure-text}, used in addition to @option{-shared}, tells
20871 the compiler to not pass @option{-z text} to the linker when linking a
20872 shared object. Using this option, you can link position-dependent
20873 code into a shared object.
20875 @option{-mimpure-text} suppresses the ``relocations remain against
20876 allocatable but non-writable sections'' linker error message.
20877 However, the necessary relocations trigger copy-on-write, and the
20878 shared object is not actually shared across processes. Instead of
20879 using @option{-mimpure-text}, you should compile all source code with
20880 @option{-fpic} or @option{-fPIC}.
20884 These switches are supported in addition to the above on Solaris 2:
20889 Add support for multithreading using the POSIX threads library. This
20890 option sets flags for both the preprocessor and linker. This option does
20891 not affect the thread safety of object code produced by the compiler or
20892 that of libraries supplied with it.
20896 This is a synonym for @option{-pthreads}.
20899 @node SPARC Options
20900 @subsection SPARC Options
20901 @cindex SPARC options
20903 These @samp{-m} options are supported on the SPARC:
20906 @item -mno-app-regs
20908 @opindex mno-app-regs
20910 Specify @option{-mapp-regs} to generate output using the global registers
20911 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
20912 global register 1, each global register 2 through 4 is then treated as an
20913 allocable register that is clobbered by function calls. This is the default.
20915 To be fully SVR4 ABI-compliant at the cost of some performance loss,
20916 specify @option{-mno-app-regs}. You should compile libraries and system
20917 software with this option.
20923 With @option{-mflat}, the compiler does not generate save/restore instructions
20924 and uses a ``flat'' or single register window model. This model is compatible
20925 with the regular register window model. The local registers and the input
20926 registers (0--5) are still treated as ``call-saved'' registers and are
20927 saved on the stack as needed.
20929 With @option{-mno-flat} (the default), the compiler generates save/restore
20930 instructions (except for leaf functions). This is the normal operating mode.
20933 @itemx -mhard-float
20935 @opindex mhard-float
20936 Generate output containing floating-point instructions. This is the
20940 @itemx -msoft-float
20942 @opindex msoft-float
20943 Generate output containing library calls for floating point.
20944 @strong{Warning:} the requisite libraries are not available for all SPARC
20945 targets. Normally the facilities of the machine's usual C compiler are
20946 used, but this cannot be done directly in cross-compilation. You must make
20947 your own arrangements to provide suitable library functions for
20948 cross-compilation. The embedded targets @samp{sparc-*-aout} and
20949 @samp{sparclite-*-*} do provide software floating-point support.
20951 @option{-msoft-float} changes the calling convention in the output file;
20952 therefore, it is only useful if you compile @emph{all} of a program with
20953 this option. In particular, you need to compile @file{libgcc.a}, the
20954 library that comes with GCC, with @option{-msoft-float} in order for
20957 @item -mhard-quad-float
20958 @opindex mhard-quad-float
20959 Generate output containing quad-word (long double) floating-point
20962 @item -msoft-quad-float
20963 @opindex msoft-quad-float
20964 Generate output containing library calls for quad-word (long double)
20965 floating-point instructions. The functions called are those specified
20966 in the SPARC ABI@. This is the default.
20968 As of this writing, there are no SPARC implementations that have hardware
20969 support for the quad-word floating-point instructions. They all invoke
20970 a trap handler for one of these instructions, and then the trap handler
20971 emulates the effect of the instruction. Because of the trap handler overhead,
20972 this is much slower than calling the ABI library routines. Thus the
20973 @option{-msoft-quad-float} option is the default.
20975 @item -mno-unaligned-doubles
20976 @itemx -munaligned-doubles
20977 @opindex mno-unaligned-doubles
20978 @opindex munaligned-doubles
20979 Assume that doubles have 8-byte alignment. This is the default.
20981 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
20982 alignment only if they are contained in another type, or if they have an
20983 absolute address. Otherwise, it assumes they have 4-byte alignment.
20984 Specifying this option avoids some rare compatibility problems with code
20985 generated by other compilers. It is not the default because it results
20986 in a performance loss, especially for floating-point code.
20989 @itemx -mno-user-mode
20990 @opindex muser-mode
20991 @opindex mno-user-mode
20992 Do not generate code that can only run in supervisor mode. This is relevant
20993 only for the @code{casa} instruction emitted for the LEON3 processor. The
20994 default is @option{-mno-user-mode}.
20996 @item -mno-faster-structs
20997 @itemx -mfaster-structs
20998 @opindex mno-faster-structs
20999 @opindex mfaster-structs
21000 With @option{-mfaster-structs}, the compiler assumes that structures
21001 should have 8-byte alignment. This enables the use of pairs of
21002 @code{ldd} and @code{std} instructions for copies in structure
21003 assignment, in place of twice as many @code{ld} and @code{st} pairs.
21004 However, the use of this changed alignment directly violates the SPARC
21005 ABI@. Thus, it's intended only for use on targets where the developer
21006 acknowledges that their resulting code is not directly in line with
21007 the rules of the ABI@.
21009 @item -mcpu=@var{cpu_type}
21011 Set the instruction set, register set, and instruction scheduling parameters
21012 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
21013 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
21014 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
21015 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
21016 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21017 @samp{niagara3} and @samp{niagara4}.
21019 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
21020 which selects the best architecture option for the host processor.
21021 @option{-mcpu=native} has no effect if GCC does not recognize
21024 Default instruction scheduling parameters are used for values that select
21025 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
21026 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
21028 Here is a list of each supported architecture and their supported
21036 supersparc, hypersparc, leon, leon3
21039 f930, f934, sparclite86x
21045 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
21048 By default (unless configured otherwise), GCC generates code for the V7
21049 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
21050 additionally optimizes it for the Cypress CY7C602 chip, as used in the
21051 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
21052 SPARCStation 1, 2, IPX etc.
21054 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
21055 architecture. The only difference from V7 code is that the compiler emits
21056 the integer multiply and integer divide instructions which exist in SPARC-V8
21057 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
21058 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
21061 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
21062 the SPARC architecture. This adds the integer multiply, integer divide step
21063 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
21064 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
21065 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
21066 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
21067 MB86934 chip, which is the more recent SPARClite with FPU@.
21069 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
21070 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
21071 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
21072 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
21073 optimizes it for the TEMIC SPARClet chip.
21075 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
21076 architecture. This adds 64-bit integer and floating-point move instructions,
21077 3 additional floating-point condition code registers and conditional move
21078 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
21079 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
21080 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
21081 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
21082 @option{-mcpu=niagara}, the compiler additionally optimizes it for
21083 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
21084 additionally optimizes it for Sun UltraSPARC T2 chips. With
21085 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
21086 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
21087 additionally optimizes it for Sun UltraSPARC T4 chips.
21089 @item -mtune=@var{cpu_type}
21091 Set the instruction scheduling parameters for machine type
21092 @var{cpu_type}, but do not set the instruction set or register set that the
21093 option @option{-mcpu=@var{cpu_type}} does.
21095 The same values for @option{-mcpu=@var{cpu_type}} can be used for
21096 @option{-mtune=@var{cpu_type}}, but the only useful values are those
21097 that select a particular CPU implementation. Those are @samp{cypress},
21098 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{leon3},
21099 @samp{leon3v7}, @samp{f930}, @samp{f934}, @samp{sparclite86x}, @samp{tsc701},
21100 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21101 @samp{niagara3} and @samp{niagara4}. With native Solaris and GNU/Linux
21102 toolchains, @samp{native} can also be used.
21107 @opindex mno-v8plus
21108 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
21109 difference from the V8 ABI is that the global and out registers are
21110 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
21111 mode for all SPARC-V9 processors.
21117 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
21118 Visual Instruction Set extensions. The default is @option{-mno-vis}.
21124 With @option{-mvis2}, GCC generates code that takes advantage of
21125 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
21126 default is @option{-mvis2} when targeting a cpu that supports such
21127 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
21128 also sets @option{-mvis}.
21134 With @option{-mvis3}, GCC generates code that takes advantage of
21135 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
21136 default is @option{-mvis3} when targeting a cpu that supports such
21137 instructions, such as niagara-3 and later. Setting @option{-mvis3}
21138 also sets @option{-mvis2} and @option{-mvis}.
21143 @opindex mno-cbcond
21144 With @option{-mcbcond}, GCC generates code that takes advantage of
21145 compare-and-branch instructions, as defined in the Sparc Architecture 2011.
21146 The default is @option{-mcbcond} when targeting a cpu that supports such
21147 instructions, such as niagara-4 and later.
21153 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
21154 population count instruction. The default is @option{-mpopc}
21155 when targeting a cpu that supports such instructions, such as Niagara-2 and
21162 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
21163 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
21164 when targeting a cpu that supports such instructions, such as Niagara-3 and
21168 @opindex mfix-at697f
21169 Enable the documented workaround for the single erratum of the Atmel AT697F
21170 processor (which corresponds to erratum #13 of the AT697E processor).
21173 @opindex mfix-ut699
21174 Enable the documented workarounds for the floating-point errata and the data
21175 cache nullify errata of the UT699 processor.
21178 These @samp{-m} options are supported in addition to the above
21179 on SPARC-V9 processors in 64-bit environments:
21186 Generate code for a 32-bit or 64-bit environment.
21187 The 32-bit environment sets int, long and pointer to 32 bits.
21188 The 64-bit environment sets int to 32 bits and long and pointer
21191 @item -mcmodel=@var{which}
21193 Set the code model to one of
21197 The Medium/Low code model: 64-bit addresses, programs
21198 must be linked in the low 32 bits of memory. Programs can be statically
21199 or dynamically linked.
21202 The Medium/Middle code model: 64-bit addresses, programs
21203 must be linked in the low 44 bits of memory, the text and data segments must
21204 be less than 2GB in size and the data segment must be located within 2GB of
21208 The Medium/Anywhere code model: 64-bit addresses, programs
21209 may be linked anywhere in memory, the text and data segments must be less
21210 than 2GB in size and the data segment must be located within 2GB of the
21214 The Medium/Anywhere code model for embedded systems:
21215 64-bit addresses, the text and data segments must be less than 2GB in
21216 size, both starting anywhere in memory (determined at link time). The
21217 global register %g4 points to the base of the data segment. Programs
21218 are statically linked and PIC is not supported.
21221 @item -mmemory-model=@var{mem-model}
21222 @opindex mmemory-model
21223 Set the memory model in force on the processor to one of
21227 The default memory model for the processor and operating system.
21230 Relaxed Memory Order
21233 Partial Store Order
21239 Sequential Consistency
21242 These memory models are formally defined in Appendix D of the Sparc V9
21243 architecture manual, as set in the processor's @code{PSTATE.MM} field.
21246 @itemx -mno-stack-bias
21247 @opindex mstack-bias
21248 @opindex mno-stack-bias
21249 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
21250 frame pointer if present, are offset by @minus{}2047 which must be added back
21251 when making stack frame references. This is the default in 64-bit mode.
21252 Otherwise, assume no such offset is present.
21256 @subsection SPU Options
21257 @cindex SPU options
21259 These @samp{-m} options are supported on the SPU:
21263 @itemx -merror-reloc
21264 @opindex mwarn-reloc
21265 @opindex merror-reloc
21267 The loader for SPU does not handle dynamic relocations. By default, GCC
21268 gives an error when it generates code that requires a dynamic
21269 relocation. @option{-mno-error-reloc} disables the error,
21270 @option{-mwarn-reloc} generates a warning instead.
21273 @itemx -munsafe-dma
21275 @opindex munsafe-dma
21277 Instructions that initiate or test completion of DMA must not be
21278 reordered with respect to loads and stores of the memory that is being
21280 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
21281 memory accesses, but that can lead to inefficient code in places where the
21282 memory is known to not change. Rather than mark the memory as volatile,
21283 you can use @option{-msafe-dma} to tell the compiler to treat
21284 the DMA instructions as potentially affecting all memory.
21286 @item -mbranch-hints
21287 @opindex mbranch-hints
21289 By default, GCC generates a branch hint instruction to avoid
21290 pipeline stalls for always-taken or probably-taken branches. A hint
21291 is not generated closer than 8 instructions away from its branch.
21292 There is little reason to disable them, except for debugging purposes,
21293 or to make an object a little bit smaller.
21297 @opindex msmall-mem
21298 @opindex mlarge-mem
21300 By default, GCC generates code assuming that addresses are never larger
21301 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
21302 a full 32-bit address.
21307 By default, GCC links against startup code that assumes the SPU-style
21308 main function interface (which has an unconventional parameter list).
21309 With @option{-mstdmain}, GCC links your program against startup
21310 code that assumes a C99-style interface to @code{main}, including a
21311 local copy of @code{argv} strings.
21313 @item -mfixed-range=@var{register-range}
21314 @opindex mfixed-range
21315 Generate code treating the given register range as fixed registers.
21316 A fixed register is one that the register allocator cannot use. This is
21317 useful when compiling kernel code. A register range is specified as
21318 two registers separated by a dash. Multiple register ranges can be
21319 specified separated by a comma.
21325 Compile code assuming that pointers to the PPU address space accessed
21326 via the @code{__ea} named address space qualifier are either 32 or 64
21327 bits wide. The default is 32 bits. As this is an ABI-changing option,
21328 all object code in an executable must be compiled with the same setting.
21330 @item -maddress-space-conversion
21331 @itemx -mno-address-space-conversion
21332 @opindex maddress-space-conversion
21333 @opindex mno-address-space-conversion
21334 Allow/disallow treating the @code{__ea} address space as superset
21335 of the generic address space. This enables explicit type casts
21336 between @code{__ea} and generic pointer as well as implicit
21337 conversions of generic pointers to @code{__ea} pointers. The
21338 default is to allow address space pointer conversions.
21340 @item -mcache-size=@var{cache-size}
21341 @opindex mcache-size
21342 This option controls the version of libgcc that the compiler links to an
21343 executable and selects a software-managed cache for accessing variables
21344 in the @code{__ea} address space with a particular cache size. Possible
21345 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
21346 and @samp{128}. The default cache size is 64KB.
21348 @item -matomic-updates
21349 @itemx -mno-atomic-updates
21350 @opindex matomic-updates
21351 @opindex mno-atomic-updates
21352 This option controls the version of libgcc that the compiler links to an
21353 executable and selects whether atomic updates to the software-managed
21354 cache of PPU-side variables are used. If you use atomic updates, changes
21355 to a PPU variable from SPU code using the @code{__ea} named address space
21356 qualifier do not interfere with changes to other PPU variables residing
21357 in the same cache line from PPU code. If you do not use atomic updates,
21358 such interference may occur; however, writing back cache lines is
21359 more efficient. The default behavior is to use atomic updates.
21362 @itemx -mdual-nops=@var{n}
21363 @opindex mdual-nops
21364 By default, GCC inserts nops to increase dual issue when it expects
21365 it to increase performance. @var{n} can be a value from 0 to 10. A
21366 smaller @var{n} inserts fewer nops. 10 is the default, 0 is the
21367 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
21369 @item -mhint-max-nops=@var{n}
21370 @opindex mhint-max-nops
21371 Maximum number of nops to insert for a branch hint. A branch hint must
21372 be at least 8 instructions away from the branch it is affecting. GCC
21373 inserts up to @var{n} nops to enforce this, otherwise it does not
21374 generate the branch hint.
21376 @item -mhint-max-distance=@var{n}
21377 @opindex mhint-max-distance
21378 The encoding of the branch hint instruction limits the hint to be within
21379 256 instructions of the branch it is affecting. By default, GCC makes
21380 sure it is within 125.
21383 @opindex msafe-hints
21384 Work around a hardware bug that causes the SPU to stall indefinitely.
21385 By default, GCC inserts the @code{hbrp} instruction to make sure
21386 this stall won't happen.
21390 @node System V Options
21391 @subsection Options for System V
21393 These additional options are available on System V Release 4 for
21394 compatibility with other compilers on those systems:
21399 Create a shared object.
21400 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
21404 Identify the versions of each tool used by the compiler, in a
21405 @code{.ident} assembler directive in the output.
21409 Refrain from adding @code{.ident} directives to the output file (this is
21412 @item -YP,@var{dirs}
21414 Search the directories @var{dirs}, and no others, for libraries
21415 specified with @option{-l}.
21417 @item -Ym,@var{dir}
21419 Look in the directory @var{dir} to find the M4 preprocessor.
21420 The assembler uses this option.
21421 @c This is supposed to go with a -Yd for predefined M4 macro files, but
21422 @c the generic assembler that comes with Solaris takes just -Ym.
21425 @node TILE-Gx Options
21426 @subsection TILE-Gx Options
21427 @cindex TILE-Gx options
21429 These @samp{-m} options are supported on the TILE-Gx:
21432 @item -mcmodel=small
21433 @opindex mcmodel=small
21434 Generate code for the small model. The distance for direct calls is
21435 limited to 500M in either direction. PC-relative addresses are 32
21436 bits. Absolute addresses support the full address range.
21438 @item -mcmodel=large
21439 @opindex mcmodel=large
21440 Generate code for the large model. There is no limitation on call
21441 distance, pc-relative addresses, or absolute addresses.
21443 @item -mcpu=@var{name}
21445 Selects the type of CPU to be targeted. Currently the only supported
21446 type is @samp{tilegx}.
21452 Generate code for a 32-bit or 64-bit environment. The 32-bit
21453 environment sets int, long, and pointer to 32 bits. The 64-bit
21454 environment sets int to 32 bits and long and pointer to 64 bits.
21457 @itemx -mlittle-endian
21458 @opindex mbig-endian
21459 @opindex mlittle-endian
21460 Generate code in big/little endian mode, respectively.
21463 @node TILEPro Options
21464 @subsection TILEPro Options
21465 @cindex TILEPro options
21467 These @samp{-m} options are supported on the TILEPro:
21470 @item -mcpu=@var{name}
21472 Selects the type of CPU to be targeted. Currently the only supported
21473 type is @samp{tilepro}.
21477 Generate code for a 32-bit environment, which sets int, long, and
21478 pointer to 32 bits. This is the only supported behavior so the flag
21479 is essentially ignored.
21483 @subsection V850 Options
21484 @cindex V850 Options
21486 These @samp{-m} options are defined for V850 implementations:
21490 @itemx -mno-long-calls
21491 @opindex mlong-calls
21492 @opindex mno-long-calls
21493 Treat all calls as being far away (near). If calls are assumed to be
21494 far away, the compiler always loads the function's address into a
21495 register, and calls indirect through the pointer.
21501 Do not optimize (do optimize) basic blocks that use the same index
21502 pointer 4 or more times to copy pointer into the @code{ep} register, and
21503 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
21504 option is on by default if you optimize.
21506 @item -mno-prolog-function
21507 @itemx -mprolog-function
21508 @opindex mno-prolog-function
21509 @opindex mprolog-function
21510 Do not use (do use) external functions to save and restore registers
21511 at the prologue and epilogue of a function. The external functions
21512 are slower, but use less code space if more than one function saves
21513 the same number of registers. The @option{-mprolog-function} option
21514 is on by default if you optimize.
21518 Try to make the code as small as possible. At present, this just turns
21519 on the @option{-mep} and @option{-mprolog-function} options.
21521 @item -mtda=@var{n}
21523 Put static or global variables whose size is @var{n} bytes or less into
21524 the tiny data area that register @code{ep} points to. The tiny data
21525 area can hold up to 256 bytes in total (128 bytes for byte references).
21527 @item -msda=@var{n}
21529 Put static or global variables whose size is @var{n} bytes or less into
21530 the small data area that register @code{gp} points to. The small data
21531 area can hold up to 64 kilobytes.
21533 @item -mzda=@var{n}
21535 Put static or global variables whose size is @var{n} bytes or less into
21536 the first 32 kilobytes of memory.
21540 Specify that the target processor is the V850.
21544 Specify that the target processor is the V850E3V5. The preprocessor
21545 constant @code{__v850e3v5__} is defined if this option is used.
21549 Specify that the target processor is the V850E3V5. This is an alias for
21550 the @option{-mv850e3v5} option.
21554 Specify that the target processor is the V850E2V3. The preprocessor
21555 constant @code{__v850e2v3__} is defined if this option is used.
21559 Specify that the target processor is the V850E2. The preprocessor
21560 constant @code{__v850e2__} is defined if this option is used.
21564 Specify that the target processor is the V850E1. The preprocessor
21565 constants @code{__v850e1__} and @code{__v850e__} are defined if
21566 this option is used.
21570 Specify that the target processor is the V850ES. This is an alias for
21571 the @option{-mv850e1} option.
21575 Specify that the target processor is the V850E@. The preprocessor
21576 constant @code{__v850e__} is defined if this option is used.
21578 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
21579 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
21580 are defined then a default target processor is chosen and the
21581 relevant @samp{__v850*__} preprocessor constant is defined.
21583 The preprocessor constants @code{__v850} and @code{__v851__} are always
21584 defined, regardless of which processor variant is the target.
21586 @item -mdisable-callt
21587 @itemx -mno-disable-callt
21588 @opindex mdisable-callt
21589 @opindex mno-disable-callt
21590 This option suppresses generation of the @code{CALLT} instruction for the
21591 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
21594 This option is enabled by default when the RH850 ABI is
21595 in use (see @option{-mrh850-abi}), and disabled by default when the
21596 GCC ABI is in use. If @code{CALLT} instructions are being generated
21597 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
21603 Pass on (or do not pass on) the @option{-mrelax} command line option
21607 @itemx -mno-long-jumps
21608 @opindex mlong-jumps
21609 @opindex mno-long-jumps
21610 Disable (or re-enable) the generation of PC-relative jump instructions.
21613 @itemx -mhard-float
21614 @opindex msoft-float
21615 @opindex mhard-float
21616 Disable (or re-enable) the generation of hardware floating point
21617 instructions. This option is only significant when the target
21618 architecture is @samp{V850E2V3} or higher. If hardware floating point
21619 instructions are being generated then the C preprocessor symbol
21620 @code{__FPU_OK__} is defined, otherwise the symbol
21621 @code{__NO_FPU__} is defined.
21625 Enables the use of the e3v5 LOOP instruction. The use of this
21626 instruction is not enabled by default when the e3v5 architecture is
21627 selected because its use is still experimental.
21631 @opindex mrh850-abi
21633 Enables support for the RH850 version of the V850 ABI. This is the
21634 default. With this version of the ABI the following rules apply:
21638 Integer sized structures and unions are returned via a memory pointer
21639 rather than a register.
21642 Large structures and unions (more than 8 bytes in size) are passed by
21646 Functions are aligned to 16-bit boundaries.
21649 The @option{-m8byte-align} command line option is supported.
21652 The @option{-mdisable-callt} command line option is enabled by
21653 default. The @option{-mno-disable-callt} command line option is not
21657 When this version of the ABI is enabled the C preprocessor symbol
21658 @code{__V850_RH850_ABI__} is defined.
21662 Enables support for the old GCC version of the V850 ABI. With this
21663 version of the ABI the following rules apply:
21667 Integer sized structures and unions are returned in register @code{r10}.
21670 Large structures and unions (more than 8 bytes in size) are passed by
21674 Functions are aligned to 32-bit boundaries, unless optimizing for
21678 The @option{-m8byte-align} command line option is not supported.
21681 The @option{-mdisable-callt} command line option is supported but not
21682 enabled by default.
21685 When this version of the ABI is enabled the C preprocessor symbol
21686 @code{__V850_GCC_ABI__} is defined.
21688 @item -m8byte-align
21689 @itemx -mno-8byte-align
21690 @opindex m8byte-align
21691 @opindex mno-8byte-align
21692 Enables support for @code{double} and @code{long long} types to be
21693 aligned on 8-byte boundaries. The default is to restrict the
21694 alignment of all objects to at most 4-bytes. When
21695 @option{-m8byte-align} is in effect the C preprocessor symbol
21696 @code{__V850_8BYTE_ALIGN__} is defined.
21699 @opindex mbig-switch
21700 Generate code suitable for big switch tables. Use this option only if
21701 the assembler/linker complain about out of range branches within a switch
21706 This option causes r2 and r5 to be used in the code generated by
21707 the compiler. This setting is the default.
21709 @item -mno-app-regs
21710 @opindex mno-app-regs
21711 This option causes r2 and r5 to be treated as fixed registers.
21716 @subsection VAX Options
21717 @cindex VAX options
21719 These @samp{-m} options are defined for the VAX:
21724 Do not output certain jump instructions (@code{aobleq} and so on)
21725 that the Unix assembler for the VAX cannot handle across long
21730 Do output those jump instructions, on the assumption that the
21731 GNU assembler is being used.
21735 Output code for G-format floating-point numbers instead of D-format.
21738 @node Visium Options
21739 @subsection Visium Options
21740 @cindex Visium options
21746 A program which performs file I/O and is destined to run on an MCM target
21747 should be linked with this option. It causes the libraries libc.a and
21748 libdebug.a to be linked. The program should be run on the target under
21749 the control of the GDB remote debugging stub.
21753 A program which performs file I/O and is destined to run on the simulator
21754 should be linked with option. This causes libraries libc.a and libsim.a to
21758 @itemx -mhard-float
21760 @opindex mhard-float
21761 Generate code containing floating-point instructions. This is the
21765 @itemx -msoft-float
21767 @opindex msoft-float
21768 Generate code containing library calls for floating-point.
21770 @option{-msoft-float} changes the calling convention in the output file;
21771 therefore, it is only useful if you compile @emph{all} of a program with
21772 this option. In particular, you need to compile @file{libgcc.a}, the
21773 library that comes with GCC, with @option{-msoft-float} in order for
21776 @item -mcpu=@var{cpu_type}
21778 Set the instruction set, register set, and instruction scheduling parameters
21779 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
21780 @samp{mcm}, @samp{gr5} and @samp{gr6}.
21782 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
21784 By default (unless configured otherwise), GCC generates code for the GR5
21785 variant of the Visium architecture.
21787 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
21788 architecture. The only difference from GR5 code is that the compiler will
21789 generate block move instructions.
21791 @item -mtune=@var{cpu_type}
21793 Set the instruction scheduling parameters for machine type @var{cpu_type},
21794 but do not set the instruction set or register set that the option
21795 @option{-mcpu=@var{cpu_type}} would.
21799 Generate code for the supervisor mode, where there are no restrictions on
21800 the access to general registers. This is the default.
21803 @opindex muser-mode
21804 Generate code for the user mode, where the access to some general registers
21805 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
21806 mode; on the GR6, only registers r29 to r31 are affected.
21810 @subsection VMS Options
21812 These @samp{-m} options are defined for the VMS implementations:
21815 @item -mvms-return-codes
21816 @opindex mvms-return-codes
21817 Return VMS condition codes from @code{main}. The default is to return POSIX-style
21818 condition (e.g.@ error) codes.
21820 @item -mdebug-main=@var{prefix}
21821 @opindex mdebug-main=@var{prefix}
21822 Flag the first routine whose name starts with @var{prefix} as the main
21823 routine for the debugger.
21827 Default to 64-bit memory allocation routines.
21829 @item -mpointer-size=@var{size}
21830 @opindex mpointer-size=@var{size}
21831 Set the default size of pointers. Possible options for @var{size} are
21832 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
21833 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
21834 The later option disables @code{pragma pointer_size}.
21837 @node VxWorks Options
21838 @subsection VxWorks Options
21839 @cindex VxWorks Options
21841 The options in this section are defined for all VxWorks targets.
21842 Options specific to the target hardware are listed with the other
21843 options for that target.
21848 GCC can generate code for both VxWorks kernels and real time processes
21849 (RTPs). This option switches from the former to the latter. It also
21850 defines the preprocessor macro @code{__RTP__}.
21853 @opindex non-static
21854 Link an RTP executable against shared libraries rather than static
21855 libraries. The options @option{-static} and @option{-shared} can
21856 also be used for RTPs (@pxref{Link Options}); @option{-static}
21863 These options are passed down to the linker. They are defined for
21864 compatibility with Diab.
21867 @opindex Xbind-lazy
21868 Enable lazy binding of function calls. This option is equivalent to
21869 @option{-Wl,-z,now} and is defined for compatibility with Diab.
21873 Disable lazy binding of function calls. This option is the default and
21874 is defined for compatibility with Diab.
21878 @subsection x86 Options
21879 @cindex x86 Options
21881 These @samp{-m} options are defined for the x86 family of computers.
21885 @item -march=@var{cpu-type}
21887 Generate instructions for the machine type @var{cpu-type}. In contrast to
21888 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
21889 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
21890 to generate code that may not run at all on processors other than the one
21891 indicated. Specifying @option{-march=@var{cpu-type}} implies
21892 @option{-mtune=@var{cpu-type}}.
21894 The choices for @var{cpu-type} are:
21898 This selects the CPU to generate code for at compilation time by determining
21899 the processor type of the compiling machine. Using @option{-march=native}
21900 enables all instruction subsets supported by the local machine (hence
21901 the result might not run on different machines). Using @option{-mtune=native}
21902 produces code optimized for the local machine under the constraints
21903 of the selected instruction set.
21906 Original Intel i386 CPU@.
21909 Intel i486 CPU@. (No scheduling is implemented for this chip.)
21913 Intel Pentium CPU with no MMX support.
21916 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
21919 Intel Pentium Pro CPU@.
21922 When used with @option{-march}, the Pentium Pro
21923 instruction set is used, so the code runs on all i686 family chips.
21924 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
21927 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
21932 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
21936 Intel Pentium M; low-power version of Intel Pentium III CPU
21937 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
21941 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
21944 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
21948 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
21949 SSE2 and SSE3 instruction set support.
21952 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
21953 instruction set support.
21956 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
21957 SSE4.1, SSE4.2 and POPCNT instruction set support.
21960 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
21961 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
21964 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
21965 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
21968 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
21969 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
21970 instruction set support.
21973 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
21974 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
21975 BMI, BMI2 and F16C instruction set support.
21978 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
21979 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
21980 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
21983 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
21984 instruction set support.
21987 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
21988 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
21991 AMD K6 CPU with MMX instruction set support.
21995 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
21998 @itemx athlon-tbird
21999 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
22005 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
22006 instruction set support.
22012 Processors based on the AMD K8 core with x86-64 instruction set support,
22013 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
22014 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
22015 instruction set extensions.)
22018 @itemx opteron-sse3
22019 @itemx athlon64-sse3
22020 Improved versions of AMD K8 cores with SSE3 instruction set support.
22024 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
22025 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
22026 instruction set extensions.)
22029 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
22030 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
22031 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
22033 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
22034 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
22035 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
22038 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
22039 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
22040 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
22041 64-bit instruction set extensions.
22043 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
22044 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
22045 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
22046 SSE4.2, ABM and 64-bit instruction set extensions.
22049 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
22050 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
22051 instruction set extensions.)
22054 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
22055 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
22056 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
22059 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
22063 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
22064 instruction set support.
22067 VIA C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
22068 implemented for this chip.)
22071 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
22073 implemented for this chip.)
22076 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
22079 @item -mtune=@var{cpu-type}
22081 Tune to @var{cpu-type} everything applicable about the generated code, except
22082 for the ABI and the set of available instructions.
22083 While picking a specific @var{cpu-type} schedules things appropriately
22084 for that particular chip, the compiler does not generate any code that
22085 cannot run on the default machine type unless you use a
22086 @option{-march=@var{cpu-type}} option.
22087 For example, if GCC is configured for i686-pc-linux-gnu
22088 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
22089 but still runs on i686 machines.
22091 The choices for @var{cpu-type} are the same as for @option{-march}.
22092 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
22096 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
22097 If you know the CPU on which your code will run, then you should use
22098 the corresponding @option{-mtune} or @option{-march} option instead of
22099 @option{-mtune=generic}. But, if you do not know exactly what CPU users
22100 of your application will have, then you should use this option.
22102 As new processors are deployed in the marketplace, the behavior of this
22103 option will change. Therefore, if you upgrade to a newer version of
22104 GCC, code generation controlled by this option will change to reflect
22106 that are most common at the time that version of GCC is released.
22108 There is no @option{-march=generic} option because @option{-march}
22109 indicates the instruction set the compiler can use, and there is no
22110 generic instruction set applicable to all processors. In contrast,
22111 @option{-mtune} indicates the processor (or, in this case, collection of
22112 processors) for which the code is optimized.
22115 Produce code optimized for the most current Intel processors, which are
22116 Haswell and Silvermont for this version of GCC. If you know the CPU
22117 on which your code will run, then you should use the corresponding
22118 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
22119 But, if you want your application performs better on both Haswell and
22120 Silvermont, then you should use this option.
22122 As new Intel processors are deployed in the marketplace, the behavior of
22123 this option will change. Therefore, if you upgrade to a newer version of
22124 GCC, code generation controlled by this option will change to reflect
22125 the most current Intel processors at the time that version of GCC is
22128 There is no @option{-march=intel} option because @option{-march} indicates
22129 the instruction set the compiler can use, and there is no common
22130 instruction set applicable to all processors. In contrast,
22131 @option{-mtune} indicates the processor (or, in this case, collection of
22132 processors) for which the code is optimized.
22135 @item -mcpu=@var{cpu-type}
22137 A deprecated synonym for @option{-mtune}.
22139 @item -mfpmath=@var{unit}
22141 Generate floating-point arithmetic for selected unit @var{unit}. The choices
22142 for @var{unit} are:
22146 Use the standard 387 floating-point coprocessor present on the majority of chips and
22147 emulated otherwise. Code compiled with this option runs almost everywhere.
22148 The temporary results are computed in 80-bit precision instead of the precision
22149 specified by the type, resulting in slightly different results compared to most
22150 of other chips. See @option{-ffloat-store} for more detailed description.
22152 This is the default choice for x86-32 targets.
22155 Use scalar floating-point instructions present in the SSE instruction set.
22156 This instruction set is supported by Pentium III and newer chips,
22157 and in the AMD line
22158 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
22159 instruction set supports only single-precision arithmetic, thus the double and
22160 extended-precision arithmetic are still done using 387. A later version, present
22161 only in Pentium 4 and AMD x86-64 chips, supports double-precision
22164 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
22165 or @option{-msse2} switches to enable SSE extensions and make this option
22166 effective. For the x86-64 compiler, these extensions are enabled by default.
22168 The resulting code should be considerably faster in the majority of cases and avoid
22169 the numerical instability problems of 387 code, but may break some existing
22170 code that expects temporaries to be 80 bits.
22172 This is the default choice for the x86-64 compiler.
22177 Attempt to utilize both instruction sets at once. This effectively doubles the
22178 amount of available registers, and on chips with separate execution units for
22179 387 and SSE the execution resources too. Use this option with care, as it is
22180 still experimental, because the GCC register allocator does not model separate
22181 functional units well, resulting in unstable performance.
22184 @item -masm=@var{dialect}
22185 @opindex masm=@var{dialect}
22186 Output assembly instructions using selected @var{dialect}. Also affects
22187 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
22188 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
22189 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
22190 not support @samp{intel}.
22193 @itemx -mno-ieee-fp
22195 @opindex mno-ieee-fp
22196 Control whether or not the compiler uses IEEE floating-point
22197 comparisons. These correctly handle the case where the result of a
22198 comparison is unordered.
22201 @opindex msoft-float
22202 Generate output containing library calls for floating point.
22204 @strong{Warning:} the requisite libraries are not part of GCC@.
22205 Normally the facilities of the machine's usual C compiler are used, but
22206 this can't be done directly in cross-compilation. You must make your
22207 own arrangements to provide suitable library functions for
22210 On machines where a function returns floating-point results in the 80387
22211 register stack, some floating-point opcodes may be emitted even if
22212 @option{-msoft-float} is used.
22214 @item -mno-fp-ret-in-387
22215 @opindex mno-fp-ret-in-387
22216 Do not use the FPU registers for return values of functions.
22218 The usual calling convention has functions return values of types
22219 @code{float} and @code{double} in an FPU register, even if there
22220 is no FPU@. The idea is that the operating system should emulate
22223 The option @option{-mno-fp-ret-in-387} causes such values to be returned
22224 in ordinary CPU registers instead.
22226 @item -mno-fancy-math-387
22227 @opindex mno-fancy-math-387
22228 Some 387 emulators do not support the @code{sin}, @code{cos} and
22229 @code{sqrt} instructions for the 387. Specify this option to avoid
22230 generating those instructions. This option is the default on FreeBSD,
22231 OpenBSD and NetBSD@. This option is overridden when @option{-march}
22232 indicates that the target CPU always has an FPU and so the
22233 instruction does not need emulation. These
22234 instructions are not generated unless you also use the
22235 @option{-funsafe-math-optimizations} switch.
22237 @item -malign-double
22238 @itemx -mno-align-double
22239 @opindex malign-double
22240 @opindex mno-align-double
22241 Control whether GCC aligns @code{double}, @code{long double}, and
22242 @code{long long} variables on a two-word boundary or a one-word
22243 boundary. Aligning @code{double} variables on a two-word boundary
22244 produces code that runs somewhat faster on a Pentium at the
22245 expense of more memory.
22247 On x86-64, @option{-malign-double} is enabled by default.
22249 @strong{Warning:} if you use the @option{-malign-double} switch,
22250 structures containing the above types are aligned differently than
22251 the published application binary interface specifications for the x86-32
22252 and are not binary compatible with structures in code compiled
22253 without that switch.
22255 @item -m96bit-long-double
22256 @itemx -m128bit-long-double
22257 @opindex m96bit-long-double
22258 @opindex m128bit-long-double
22259 These switches control the size of @code{long double} type. The x86-32
22260 application binary interface specifies the size to be 96 bits,
22261 so @option{-m96bit-long-double} is the default in 32-bit mode.
22263 Modern architectures (Pentium and newer) prefer @code{long double}
22264 to be aligned to an 8- or 16-byte boundary. In arrays or structures
22265 conforming to the ABI, this is not possible. So specifying
22266 @option{-m128bit-long-double} aligns @code{long double}
22267 to a 16-byte boundary by padding the @code{long double} with an additional
22270 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
22271 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
22273 Notice that neither of these options enable any extra precision over the x87
22274 standard of 80 bits for a @code{long double}.
22276 @strong{Warning:} if you override the default value for your target ABI, this
22277 changes the size of
22278 structures and arrays containing @code{long double} variables,
22279 as well as modifying the function calling convention for functions taking
22280 @code{long double}. Hence they are not binary-compatible
22281 with code compiled without that switch.
22283 @item -mlong-double-64
22284 @itemx -mlong-double-80
22285 @itemx -mlong-double-128
22286 @opindex mlong-double-64
22287 @opindex mlong-double-80
22288 @opindex mlong-double-128
22289 These switches control the size of @code{long double} type. A size
22290 of 64 bits makes the @code{long double} type equivalent to the @code{double}
22291 type. This is the default for 32-bit Bionic C library. A size
22292 of 128 bits makes the @code{long double} type equivalent to the
22293 @code{__float128} type. This is the default for 64-bit Bionic C library.
22295 @strong{Warning:} if you override the default value for your target ABI, this
22296 changes the size of
22297 structures and arrays containing @code{long double} variables,
22298 as well as modifying the function calling convention for functions taking
22299 @code{long double}. Hence they are not binary-compatible
22300 with code compiled without that switch.
22302 @item -malign-data=@var{type}
22303 @opindex malign-data
22304 Control how GCC aligns variables. Supported values for @var{type} are
22305 @samp{compat} uses increased alignment value compatible uses GCC 4.8
22306 and earlier, @samp{abi} uses alignment value as specified by the
22307 psABI, and @samp{cacheline} uses increased alignment value to match
22308 the cache line size. @samp{compat} is the default.
22310 @item -mlarge-data-threshold=@var{threshold}
22311 @opindex mlarge-data-threshold
22312 When @option{-mcmodel=medium} is specified, data objects larger than
22313 @var{threshold} are placed in the large data section. This value must be the
22314 same across all objects linked into the binary, and defaults to 65535.
22318 Use a different function-calling convention, in which functions that
22319 take a fixed number of arguments return with the @code{ret @var{num}}
22320 instruction, which pops their arguments while returning. This saves one
22321 instruction in the caller since there is no need to pop the arguments
22324 You can specify that an individual function is called with this calling
22325 sequence with the function attribute @code{stdcall}. You can also
22326 override the @option{-mrtd} option by using the function attribute
22327 @code{cdecl}. @xref{Function Attributes}.
22329 @strong{Warning:} this calling convention is incompatible with the one
22330 normally used on Unix, so you cannot use it if you need to call
22331 libraries compiled with the Unix compiler.
22333 Also, you must provide function prototypes for all functions that
22334 take variable numbers of arguments (including @code{printf});
22335 otherwise incorrect code is generated for calls to those
22338 In addition, seriously incorrect code results if you call a
22339 function with too many arguments. (Normally, extra arguments are
22340 harmlessly ignored.)
22342 @item -mregparm=@var{num}
22344 Control how many registers are used to pass integer arguments. By
22345 default, no registers are used to pass arguments, and at most 3
22346 registers can be used. You can control this behavior for a specific
22347 function by using the function attribute @code{regparm}.
22348 @xref{Function Attributes}.
22350 @strong{Warning:} if you use this switch, and
22351 @var{num} is nonzero, then you must build all modules with the same
22352 value, including any libraries. This includes the system libraries and
22356 @opindex msseregparm
22357 Use SSE register passing conventions for float and double arguments
22358 and return values. You can control this behavior for a specific
22359 function by using the function attribute @code{sseregparm}.
22360 @xref{Function Attributes}.
22362 @strong{Warning:} if you use this switch then you must build all
22363 modules with the same value, including any libraries. This includes
22364 the system libraries and startup modules.
22366 @item -mvect8-ret-in-mem
22367 @opindex mvect8-ret-in-mem
22368 Return 8-byte vectors in memory instead of MMX registers. This is the
22369 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
22370 Studio compilers until version 12. Later compiler versions (starting
22371 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
22372 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
22373 you need to remain compatible with existing code produced by those
22374 previous compiler versions or older versions of GCC@.
22383 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
22384 is specified, the significands of results of floating-point operations are
22385 rounded to 24 bits (single precision); @option{-mpc64} rounds the
22386 significands of results of floating-point operations to 53 bits (double
22387 precision) and @option{-mpc80} rounds the significands of results of
22388 floating-point operations to 64 bits (extended double precision), which is
22389 the default. When this option is used, floating-point operations in higher
22390 precisions are not available to the programmer without setting the FPU
22391 control word explicitly.
22393 Setting the rounding of floating-point operations to less than the default
22394 80 bits can speed some programs by 2% or more. Note that some mathematical
22395 libraries assume that extended-precision (80-bit) floating-point operations
22396 are enabled by default; routines in such libraries could suffer significant
22397 loss of accuracy, typically through so-called ``catastrophic cancellation'',
22398 when this option is used to set the precision to less than extended precision.
22400 @item -mstackrealign
22401 @opindex mstackrealign
22402 Realign the stack at entry. On the x86, the @option{-mstackrealign}
22403 option generates an alternate prologue and epilogue that realigns the
22404 run-time stack if necessary. This supports mixing legacy codes that keep
22405 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
22406 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
22407 applicable to individual functions.
22409 @item -mpreferred-stack-boundary=@var{num}
22410 @opindex mpreferred-stack-boundary
22411 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
22412 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
22413 the default is 4 (16 bytes or 128 bits).
22415 @strong{Warning:} When generating code for the x86-64 architecture with
22416 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
22417 used to keep the stack boundary aligned to 8 byte boundary. Since
22418 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
22419 intended to be used in controlled environment where stack space is
22420 important limitation. This option leads to wrong code when functions
22421 compiled with 16 byte stack alignment (such as functions from a standard
22422 library) are called with misaligned stack. In this case, SSE
22423 instructions may lead to misaligned memory access traps. In addition,
22424 variable arguments are handled incorrectly for 16 byte aligned
22425 objects (including x87 long double and __int128), leading to wrong
22426 results. You must build all modules with
22427 @option{-mpreferred-stack-boundary=3}, including any libraries. This
22428 includes the system libraries and startup modules.
22430 @item -mincoming-stack-boundary=@var{num}
22431 @opindex mincoming-stack-boundary
22432 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
22433 boundary. If @option{-mincoming-stack-boundary} is not specified,
22434 the one specified by @option{-mpreferred-stack-boundary} is used.
22436 On Pentium and Pentium Pro, @code{double} and @code{long double} values
22437 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
22438 suffer significant run time performance penalties. On Pentium III, the
22439 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
22440 properly if it is not 16-byte aligned.
22442 To ensure proper alignment of this values on the stack, the stack boundary
22443 must be as aligned as that required by any value stored on the stack.
22444 Further, every function must be generated such that it keeps the stack
22445 aligned. Thus calling a function compiled with a higher preferred
22446 stack boundary from a function compiled with a lower preferred stack
22447 boundary most likely misaligns the stack. It is recommended that
22448 libraries that use callbacks always use the default setting.
22450 This extra alignment does consume extra stack space, and generally
22451 increases code size. Code that is sensitive to stack space usage, such
22452 as embedded systems and operating system kernels, may want to reduce the
22453 preferred alignment to @option{-mpreferred-stack-boundary=2}.
22499 @opindex mclfushopt
22517 @itemx -mprefetchwt1
22518 @opindex mprefetchwt1
22566 These switches enable the use of instructions in the MMX, SSE,
22567 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
22568 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
22569 BMI, BMI2, FXSR, XSAVE, XSAVEOPT, LZCNT, RTM, MPX or 3DNow!@:
22570 extended instruction sets. Each has a corresponding @option{-mno-} option
22571 to disable use of these instructions.
22573 These extensions are also available as built-in functions: see
22574 @ref{x86 Built-in Functions}, for details of the functions enabled and
22575 disabled by these switches.
22577 To generate SSE/SSE2 instructions automatically from floating-point
22578 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
22580 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
22581 generates new AVX instructions or AVX equivalence for all SSEx instructions
22584 These options enable GCC to use these extended instructions in
22585 generated code, even without @option{-mfpmath=sse}. Applications that
22586 perform run-time CPU detection must compile separate files for each
22587 supported architecture, using the appropriate flags. In particular,
22588 the file containing the CPU detection code should be compiled without
22591 @item -mdump-tune-features
22592 @opindex mdump-tune-features
22593 This option instructs GCC to dump the names of the x86 performance
22594 tuning features and default settings. The names can be used in
22595 @option{-mtune-ctrl=@var{feature-list}}.
22597 @item -mtune-ctrl=@var{feature-list}
22598 @opindex mtune-ctrl=@var{feature-list}
22599 This option is used to do fine grain control of x86 code generation features.
22600 @var{feature-list} is a comma separated list of @var{feature} names. See also
22601 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
22602 on if it is not preceded with @samp{^}, otherwise, it is turned off.
22603 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
22604 developers. Using it may lead to code paths not covered by testing and can
22605 potentially result in compiler ICEs or runtime errors.
22608 @opindex mno-default
22609 This option instructs GCC to turn off all tunable features. See also
22610 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
22614 This option instructs GCC to emit a @code{cld} instruction in the prologue
22615 of functions that use string instructions. String instructions depend on
22616 the DF flag to select between autoincrement or autodecrement mode. While the
22617 ABI specifies the DF flag to be cleared on function entry, some operating
22618 systems violate this specification by not clearing the DF flag in their
22619 exception dispatchers. The exception handler can be invoked with the DF flag
22620 set, which leads to wrong direction mode when string instructions are used.
22621 This option can be enabled by default on 32-bit x86 targets by configuring
22622 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
22623 instructions can be suppressed with the @option{-mno-cld} compiler option
22627 @opindex mvzeroupper
22628 This option instructs GCC to emit a @code{vzeroupper} instruction
22629 before a transfer of control flow out of the function to minimize
22630 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
22633 @item -mprefer-avx128
22634 @opindex mprefer-avx128
22635 This option instructs GCC to use 128-bit AVX instructions instead of
22636 256-bit AVX instructions in the auto-vectorizer.
22640 This option enables GCC to generate @code{CMPXCHG16B} instructions.
22641 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
22642 (or oword) data types.
22643 This is useful for high-resolution counters that can be updated
22644 by multiple processors (or cores). This instruction is generated as part of
22645 atomic built-in functions: see @ref{__sync Builtins} or
22646 @ref{__atomic Builtins} for details.
22650 This option enables generation of @code{SAHF} instructions in 64-bit code.
22651 Early Intel Pentium 4 CPUs with Intel 64 support,
22652 prior to the introduction of Pentium 4 G1 step in December 2005,
22653 lacked the @code{LAHF} and @code{SAHF} instructions
22654 which are supported by AMD64.
22655 These are load and store instructions, respectively, for certain status flags.
22656 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
22657 @code{drem}, and @code{remainder} built-in functions;
22658 see @ref{Other Builtins} for details.
22662 This option enables use of the @code{movbe} instruction to implement
22663 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
22667 This option enables built-in functions @code{__builtin_ia32_crc32qi},
22668 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
22669 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
22673 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
22674 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
22675 with an additional Newton-Raphson step
22676 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
22677 (and their vectorized
22678 variants) for single-precision floating-point arguments. These instructions
22679 are generated only when @option{-funsafe-math-optimizations} is enabled
22680 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
22681 Note that while the throughput of the sequence is higher than the throughput
22682 of the non-reciprocal instruction, the precision of the sequence can be
22683 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
22685 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
22686 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
22687 combination), and doesn't need @option{-mrecip}.
22689 Also note that GCC emits the above sequence with additional Newton-Raphson step
22690 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
22691 already with @option{-ffast-math} (or the above option combination), and
22692 doesn't need @option{-mrecip}.
22694 @item -mrecip=@var{opt}
22695 @opindex mrecip=opt
22696 This option controls which reciprocal estimate instructions
22697 may be used. @var{opt} is a comma-separated list of options, which may
22698 be preceded by a @samp{!} to invert the option:
22702 Enable all estimate instructions.
22705 Enable the default instructions, equivalent to @option{-mrecip}.
22708 Disable all estimate instructions, equivalent to @option{-mno-recip}.
22711 Enable the approximation for scalar division.
22714 Enable the approximation for vectorized division.
22717 Enable the approximation for scalar square root.
22720 Enable the approximation for vectorized square root.
22723 So, for example, @option{-mrecip=all,!sqrt} enables
22724 all of the reciprocal approximations, except for square root.
22726 @item -mveclibabi=@var{type}
22727 @opindex mveclibabi
22728 Specifies the ABI type to use for vectorizing intrinsics using an
22729 external library. Supported values for @var{type} are @samp{svml}
22730 for the Intel short
22731 vector math library and @samp{acml} for the AMD math core library.
22732 To use this option, both @option{-ftree-vectorize} and
22733 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
22734 ABI-compatible library must be specified at link time.
22736 GCC currently emits calls to @code{vmldExp2},
22737 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
22738 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
22739 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
22740 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
22741 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
22742 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
22743 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
22744 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
22745 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
22746 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
22747 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
22748 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
22749 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
22750 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
22751 when @option{-mveclibabi=acml} is used.
22753 @item -mabi=@var{name}
22755 Generate code for the specified calling convention. Permissible values
22756 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
22757 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
22758 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
22759 You can control this behavior for specific functions by
22760 using the function attributes @code{ms_abi} and @code{sysv_abi}.
22761 @xref{Function Attributes}.
22763 @item -mtls-dialect=@var{type}
22764 @opindex mtls-dialect
22765 Generate code to access thread-local storage using the @samp{gnu} or
22766 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
22767 @samp{gnu2} is more efficient, but it may add compile- and run-time
22768 requirements that cannot be satisfied on all systems.
22771 @itemx -mno-push-args
22772 @opindex mpush-args
22773 @opindex mno-push-args
22774 Use PUSH operations to store outgoing parameters. This method is shorter
22775 and usually equally fast as method using SUB/MOV operations and is enabled
22776 by default. In some cases disabling it may improve performance because of
22777 improved scheduling and reduced dependencies.
22779 @item -maccumulate-outgoing-args
22780 @opindex maccumulate-outgoing-args
22781 If enabled, the maximum amount of space required for outgoing arguments is
22782 computed in the function prologue. This is faster on most modern CPUs
22783 because of reduced dependencies, improved scheduling and reduced stack usage
22784 when the preferred stack boundary is not equal to 2. The drawback is a notable
22785 increase in code size. This switch implies @option{-mno-push-args}.
22789 Support thread-safe exception handling on MinGW. Programs that rely
22790 on thread-safe exception handling must compile and link all code with the
22791 @option{-mthreads} option. When compiling, @option{-mthreads} defines
22792 @option{-D_MT}; when linking, it links in a special thread helper library
22793 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
22795 @item -mno-align-stringops
22796 @opindex mno-align-stringops
22797 Do not align the destination of inlined string operations. This switch reduces
22798 code size and improves performance in case the destination is already aligned,
22799 but GCC doesn't know about it.
22801 @item -minline-all-stringops
22802 @opindex minline-all-stringops
22803 By default GCC inlines string operations only when the destination is
22804 known to be aligned to least a 4-byte boundary.
22805 This enables more inlining and increases code
22806 size, but may improve performance of code that depends on fast
22807 @code{memcpy}, @code{strlen},
22808 and @code{memset} for short lengths.
22810 @item -minline-stringops-dynamically
22811 @opindex minline-stringops-dynamically
22812 For string operations of unknown size, use run-time checks with
22813 inline code for small blocks and a library call for large blocks.
22815 @item -mstringop-strategy=@var{alg}
22816 @opindex mstringop-strategy=@var{alg}
22817 Override the internal decision heuristic for the particular algorithm to use
22818 for inlining string operations. The allowed values for @var{alg} are:
22824 Expand using i386 @code{rep} prefix of the specified size.
22828 @itemx unrolled_loop
22829 Expand into an inline loop.
22832 Always use a library call.
22835 @item -mmemcpy-strategy=@var{strategy}
22836 @opindex mmemcpy-strategy=@var{strategy}
22837 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
22838 should be inlined and what inline algorithm to use when the expected size
22839 of the copy operation is known. @var{strategy}
22840 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
22841 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
22842 the max byte size with which inline algorithm @var{alg} is allowed. For the last
22843 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
22844 in the list must be specified in increasing order. The minimal byte size for
22845 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
22848 @item -mmemset-strategy=@var{strategy}
22849 @opindex mmemset-strategy=@var{strategy}
22850 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
22851 @code{__builtin_memset} expansion.
22853 @item -momit-leaf-frame-pointer
22854 @opindex momit-leaf-frame-pointer
22855 Don't keep the frame pointer in a register for leaf functions. This
22856 avoids the instructions to save, set up, and restore frame pointers and
22857 makes an extra register available in leaf functions. The option
22858 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
22859 which might make debugging harder.
22861 @item -mtls-direct-seg-refs
22862 @itemx -mno-tls-direct-seg-refs
22863 @opindex mtls-direct-seg-refs
22864 Controls whether TLS variables may be accessed with offsets from the
22865 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
22866 or whether the thread base pointer must be added. Whether or not this
22867 is valid depends on the operating system, and whether it maps the
22868 segment to cover the entire TLS area.
22870 For systems that use the GNU C Library, the default is on.
22873 @itemx -mno-sse2avx
22875 Specify that the assembler should encode SSE instructions with VEX
22876 prefix. The option @option{-mavx} turns this on by default.
22881 If profiling is active (@option{-pg}), put the profiling
22882 counter call before the prologue.
22883 Note: On x86 architectures the attribute @code{ms_hook_prologue}
22884 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
22886 @item -mrecord-mcount
22887 @itemx -mno-record-mcount
22888 @opindex mrecord-mcount
22889 If profiling is active (@option{-pg}), generate a __mcount_loc section
22890 that contains pointers to each profiling call. This is useful for
22891 automatically patching and out calls.
22894 @itemx -mno-nop-mcount
22895 @opindex mnop-mcount
22896 If profiling is active (@option{-pg}), generate the calls to
22897 the profiling functions as nops. This is useful when they
22898 should be patched in later dynamically. This is likely only
22899 useful together with @option{-mrecord-mcount}.
22901 @item -mskip-rax-setup
22902 @itemx -mno-skip-rax-setup
22903 @opindex mskip-rax-setup
22904 When generating code for the x86-64 architecture with SSE extensions
22905 disabled, @option{-skip-rax-setup} can be used to skip setting up RAX
22906 register when there are no variable arguments passed in vector registers.
22908 @strong{Warning:} Since RAX register is used to avoid unnecessarily
22909 saving vector registers on stack when passing variable arguments, the
22910 impacts of this option are callees may waste some stack space,
22911 misbehave or jump to a random location. GCC 4.4 or newer don't have
22912 those issues, regardless the RAX register value.
22915 @itemx -mno-8bit-idiv
22916 @opindex m8bit-idiv
22917 On some processors, like Intel Atom, 8-bit unsigned integer divide is
22918 much faster than 32-bit/64-bit integer divide. This option generates a
22919 run-time check. If both dividend and divisor are within range of 0
22920 to 255, 8-bit unsigned integer divide is used instead of
22921 32-bit/64-bit integer divide.
22923 @item -mavx256-split-unaligned-load
22924 @itemx -mavx256-split-unaligned-store
22925 @opindex mavx256-split-unaligned-load
22926 @opindex mavx256-split-unaligned-store
22927 Split 32-byte AVX unaligned load and store.
22929 @item -mstack-protector-guard=@var{guard}
22930 @opindex mstack-protector-guard=@var{guard}
22931 Generate stack protection code using canary at @var{guard}. Supported
22932 locations are @samp{global} for global canary or @samp{tls} for per-thread
22933 canary in the TLS block (the default). This option has effect only when
22934 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
22938 These @samp{-m} switches are supported in addition to the above
22939 on x86-64 processors in 64-bit environments.
22950 Generate code for a 16-bit, 32-bit or 64-bit environment.
22951 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
22953 generates code that runs on any i386 system.
22955 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
22956 types to 64 bits, and generates code for the x86-64 architecture.
22957 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
22958 and @option{-mdynamic-no-pic} options.
22960 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
22962 generates code for the x86-64 architecture.
22964 The @option{-m16} option is the same as @option{-m32}, except for that
22965 it outputs the @code{.code16gcc} assembly directive at the beginning of
22966 the assembly output so that the binary can run in 16-bit mode.
22968 @item -mno-red-zone
22969 @opindex mno-red-zone
22970 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
22971 by the x86-64 ABI; it is a 128-byte area beyond the location of the
22972 stack pointer that is not modified by signal or interrupt handlers
22973 and therefore can be used for temporary data without adjusting the stack
22974 pointer. The flag @option{-mno-red-zone} disables this red zone.
22976 @item -mcmodel=small
22977 @opindex mcmodel=small
22978 Generate code for the small code model: the program and its symbols must
22979 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
22980 Programs can be statically or dynamically linked. This is the default
22983 @item -mcmodel=kernel
22984 @opindex mcmodel=kernel
22985 Generate code for the kernel code model. The kernel runs in the
22986 negative 2 GB of the address space.
22987 This model has to be used for Linux kernel code.
22989 @item -mcmodel=medium
22990 @opindex mcmodel=medium
22991 Generate code for the medium model: the program is linked in the lower 2
22992 GB of the address space. Small symbols are also placed there. Symbols
22993 with sizes larger than @option{-mlarge-data-threshold} are put into
22994 large data or BSS sections and can be located above 2GB. Programs can
22995 be statically or dynamically linked.
22997 @item -mcmodel=large
22998 @opindex mcmodel=large
22999 Generate code for the large model. This model makes no assumptions
23000 about addresses and sizes of sections.
23002 @item -maddress-mode=long
23003 @opindex maddress-mode=long
23004 Generate code for long address mode. This is only supported for 64-bit
23005 and x32 environments. It is the default address mode for 64-bit
23008 @item -maddress-mode=short
23009 @opindex maddress-mode=short
23010 Generate code for short address mode. This is only supported for 32-bit
23011 and x32 environments. It is the default address mode for 32-bit and
23015 @node x86 Windows Options
23016 @subsection x86 Windows Options
23017 @cindex x86 Windows Options
23018 @cindex Windows Options for x86
23020 These additional options are available for Microsoft Windows targets:
23026 specifies that a console application is to be generated, by
23027 instructing the linker to set the PE header subsystem type
23028 required for console applications.
23029 This option is available for Cygwin and MinGW targets and is
23030 enabled by default on those targets.
23034 This option is available for Cygwin and MinGW targets. It
23035 specifies that a DLL---a dynamic link library---is to be
23036 generated, enabling the selection of the required runtime
23037 startup object and entry point.
23039 @item -mnop-fun-dllimport
23040 @opindex mnop-fun-dllimport
23041 This option is available for Cygwin and MinGW targets. It
23042 specifies that the @code{dllimport} attribute should be ignored.
23046 This option is available for MinGW targets. It specifies
23047 that MinGW-specific thread support is to be used.
23051 This option is available for MinGW-w64 targets. It causes
23052 the @code{UNICODE} preprocessor macro to be predefined, and
23053 chooses Unicode-capable runtime startup code.
23057 This option is available for Cygwin and MinGW targets. It
23058 specifies that the typical Microsoft Windows predefined macros are to
23059 be set in the pre-processor, but does not influence the choice
23060 of runtime library/startup code.
23064 This option is available for Cygwin and MinGW targets. It
23065 specifies that a GUI application is to be generated by
23066 instructing the linker to set the PE header subsystem type
23069 @item -fno-set-stack-executable
23070 @opindex fno-set-stack-executable
23071 This option is available for MinGW targets. It specifies that
23072 the executable flag for the stack used by nested functions isn't
23073 set. This is necessary for binaries running in kernel mode of
23074 Microsoft Windows, as there the User32 API, which is used to set executable
23075 privileges, isn't available.
23077 @item -fwritable-relocated-rdata
23078 @opindex fno-writable-relocated-rdata
23079 This option is available for MinGW and Cygwin targets. It specifies
23080 that relocated-data in read-only section is put into .data
23081 section. This is a necessary for older runtimes not supporting
23082 modification of .rdata sections for pseudo-relocation.
23084 @item -mpe-aligned-commons
23085 @opindex mpe-aligned-commons
23086 This option is available for Cygwin and MinGW targets. It
23087 specifies that the GNU extension to the PE file format that
23088 permits the correct alignment of COMMON variables should be
23089 used when generating code. It is enabled by default if
23090 GCC detects that the target assembler found during configuration
23091 supports the feature.
23094 See also under @ref{x86 Options} for standard options.
23096 @node Xstormy16 Options
23097 @subsection Xstormy16 Options
23098 @cindex Xstormy16 Options
23100 These options are defined for Xstormy16:
23105 Choose startup files and linker script suitable for the simulator.
23108 @node Xtensa Options
23109 @subsection Xtensa Options
23110 @cindex Xtensa Options
23112 These options are supported for Xtensa targets:
23116 @itemx -mno-const16
23118 @opindex mno-const16
23119 Enable or disable use of @code{CONST16} instructions for loading
23120 constant values. The @code{CONST16} instruction is currently not a
23121 standard option from Tensilica. When enabled, @code{CONST16}
23122 instructions are always used in place of the standard @code{L32R}
23123 instructions. The use of @code{CONST16} is enabled by default only if
23124 the @code{L32R} instruction is not available.
23127 @itemx -mno-fused-madd
23128 @opindex mfused-madd
23129 @opindex mno-fused-madd
23130 Enable or disable use of fused multiply/add and multiply/subtract
23131 instructions in the floating-point option. This has no effect if the
23132 floating-point option is not also enabled. Disabling fused multiply/add
23133 and multiply/subtract instructions forces the compiler to use separate
23134 instructions for the multiply and add/subtract operations. This may be
23135 desirable in some cases where strict IEEE 754-compliant results are
23136 required: the fused multiply add/subtract instructions do not round the
23137 intermediate result, thereby producing results with @emph{more} bits of
23138 precision than specified by the IEEE standard. Disabling fused multiply
23139 add/subtract instructions also ensures that the program output is not
23140 sensitive to the compiler's ability to combine multiply and add/subtract
23143 @item -mserialize-volatile
23144 @itemx -mno-serialize-volatile
23145 @opindex mserialize-volatile
23146 @opindex mno-serialize-volatile
23147 When this option is enabled, GCC inserts @code{MEMW} instructions before
23148 @code{volatile} memory references to guarantee sequential consistency.
23149 The default is @option{-mserialize-volatile}. Use
23150 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
23152 @item -mforce-no-pic
23153 @opindex mforce-no-pic
23154 For targets, like GNU/Linux, where all user-mode Xtensa code must be
23155 position-independent code (PIC), this option disables PIC for compiling
23158 @item -mtext-section-literals
23159 @itemx -mno-text-section-literals
23160 @opindex mtext-section-literals
23161 @opindex mno-text-section-literals
23162 These options control the treatment of literal pools. The default is
23163 @option{-mno-text-section-literals}, which places literals in a separate
23164 section in the output file. This allows the literal pool to be placed
23165 in a data RAM/ROM, and it also allows the linker to combine literal
23166 pools from separate object files to remove redundant literals and
23167 improve code size. With @option{-mtext-section-literals}, the literals
23168 are interspersed in the text section in order to keep them as close as
23169 possible to their references. This may be necessary for large assembly
23172 @item -mtarget-align
23173 @itemx -mno-target-align
23174 @opindex mtarget-align
23175 @opindex mno-target-align
23176 When this option is enabled, GCC instructs the assembler to
23177 automatically align instructions to reduce branch penalties at the
23178 expense of some code density. The assembler attempts to widen density
23179 instructions to align branch targets and the instructions following call
23180 instructions. If there are not enough preceding safe density
23181 instructions to align a target, no widening is performed. The
23182 default is @option{-mtarget-align}. These options do not affect the
23183 treatment of auto-aligned instructions like @code{LOOP}, which the
23184 assembler always aligns, either by widening density instructions or
23185 by inserting NOP instructions.
23188 @itemx -mno-longcalls
23189 @opindex mlongcalls
23190 @opindex mno-longcalls
23191 When this option is enabled, GCC instructs the assembler to translate
23192 direct calls to indirect calls unless it can determine that the target
23193 of a direct call is in the range allowed by the call instruction. This
23194 translation typically occurs for calls to functions in other source
23195 files. Specifically, the assembler translates a direct @code{CALL}
23196 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
23197 The default is @option{-mno-longcalls}. This option should be used in
23198 programs where the call target can potentially be out of range. This
23199 option is implemented in the assembler, not the compiler, so the
23200 assembly code generated by GCC still shows direct call
23201 instructions---look at the disassembled object code to see the actual
23202 instructions. Note that the assembler uses an indirect call for
23203 every cross-file call, not just those that really are out of range.
23206 @node zSeries Options
23207 @subsection zSeries Options
23208 @cindex zSeries options
23210 These are listed under @xref{S/390 and zSeries Options}.
23212 @node Code Gen Options
23213 @section Options for Code Generation Conventions
23214 @cindex code generation conventions
23215 @cindex options, code generation
23216 @cindex run-time options
23218 These machine-independent options control the interface conventions
23219 used in code generation.
23221 Most of them have both positive and negative forms; the negative form
23222 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
23223 one of the forms is listed---the one that is not the default. You
23224 can figure out the other form by either removing @samp{no-} or adding
23228 @item -fbounds-check
23229 @opindex fbounds-check
23230 For front ends that support it, generate additional code to check that
23231 indices used to access arrays are within the declared range. This is
23232 currently only supported by the Java and Fortran front ends, where
23233 this option defaults to true and false respectively.
23235 @item -fstack-reuse=@var{reuse-level}
23236 @opindex fstack_reuse
23237 This option controls stack space reuse for user declared local/auto variables
23238 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
23239 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
23240 local variables and temporaries, @samp{named_vars} enables the reuse only for
23241 user defined local variables with names, and @samp{none} disables stack reuse
23242 completely. The default value is @samp{all}. The option is needed when the
23243 program extends the lifetime of a scoped local variable or a compiler generated
23244 temporary beyond the end point defined by the language. When a lifetime of
23245 a variable ends, and if the variable lives in memory, the optimizing compiler
23246 has the freedom to reuse its stack space with other temporaries or scoped
23247 local variables whose live range does not overlap with it. Legacy code extending
23248 local lifetime is likely to break with the stack reuse optimization.
23267 if (*p == 10) // out of scope use of local1
23278 A(int k) : i(k), j(k) @{ @}
23285 void foo(const A& ar)
23292 foo(A(10)); // temp object's lifetime ends when foo returns
23298 ap->i+= 10; // ap references out of scope temp whose space
23299 // is reused with a. What is the value of ap->i?
23304 The lifetime of a compiler generated temporary is well defined by the C++
23305 standard. When a lifetime of a temporary ends, and if the temporary lives
23306 in memory, the optimizing compiler has the freedom to reuse its stack
23307 space with other temporaries or scoped local variables whose live range
23308 does not overlap with it. However some of the legacy code relies on
23309 the behavior of older compilers in which temporaries' stack space is
23310 not reused, the aggressive stack reuse can lead to runtime errors. This
23311 option is used to control the temporary stack reuse optimization.
23315 This option generates traps for signed overflow on addition, subtraction,
23316 multiplication operations.
23320 This option instructs the compiler to assume that signed arithmetic
23321 overflow of addition, subtraction and multiplication wraps around
23322 using twos-complement representation. This flag enables some optimizations
23323 and disables others. This option is enabled by default for the Java
23324 front end, as required by the Java language specification.
23327 @opindex fexceptions
23328 Enable exception handling. Generates extra code needed to propagate
23329 exceptions. For some targets, this implies GCC generates frame
23330 unwind information for all functions, which can produce significant data
23331 size overhead, although it does not affect execution. If you do not
23332 specify this option, GCC enables it by default for languages like
23333 C++ that normally require exception handling, and disables it for
23334 languages like C that do not normally require it. However, you may need
23335 to enable this option when compiling C code that needs to interoperate
23336 properly with exception handlers written in C++. You may also wish to
23337 disable this option if you are compiling older C++ programs that don't
23338 use exception handling.
23340 @item -fnon-call-exceptions
23341 @opindex fnon-call-exceptions
23342 Generate code that allows trapping instructions to throw exceptions.
23343 Note that this requires platform-specific runtime support that does
23344 not exist everywhere. Moreover, it only allows @emph{trapping}
23345 instructions to throw exceptions, i.e.@: memory references or floating-point
23346 instructions. It does not allow exceptions to be thrown from
23347 arbitrary signal handlers such as @code{SIGALRM}.
23349 @item -fdelete-dead-exceptions
23350 @opindex fdelete-dead-exceptions
23351 Consider that instructions that may throw exceptions but don't otherwise
23352 contribute to the execution of the program can be optimized away.
23353 This option is enabled by default for the Ada front end, as permitted by
23354 the Ada language specification.
23355 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
23357 @item -funwind-tables
23358 @opindex funwind-tables
23359 Similar to @option{-fexceptions}, except that it just generates any needed
23360 static data, but does not affect the generated code in any other way.
23361 You normally do not need to enable this option; instead, a language processor
23362 that needs this handling enables it on your behalf.
23364 @item -fasynchronous-unwind-tables
23365 @opindex fasynchronous-unwind-tables
23366 Generate unwind table in DWARF 2 format, if supported by target machine. The
23367 table is exact at each instruction boundary, so it can be used for stack
23368 unwinding from asynchronous events (such as debugger or garbage collector).
23370 @item -fno-gnu-unique
23371 @opindex fno-gnu-unique
23372 On systems with recent GNU assembler and C library, the C++ compiler
23373 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
23374 of template static data members and static local variables in inline
23375 functions are unique even in the presence of @code{RTLD_LOCAL}; this
23376 is necessary to avoid problems with a library used by two different
23377 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
23378 therefore disagreeing with the other one about the binding of the
23379 symbol. But this causes @code{dlclose} to be ignored for affected
23380 DSOs; if your program relies on reinitialization of a DSO via
23381 @code{dlclose} and @code{dlopen}, you can use
23382 @option{-fno-gnu-unique}.
23384 @item -fpcc-struct-return
23385 @opindex fpcc-struct-return
23386 Return ``short'' @code{struct} and @code{union} values in memory like
23387 longer ones, rather than in registers. This convention is less
23388 efficient, but it has the advantage of allowing intercallability between
23389 GCC-compiled files and files compiled with other compilers, particularly
23390 the Portable C Compiler (pcc).
23392 The precise convention for returning structures in memory depends
23393 on the target configuration macros.
23395 Short structures and unions are those whose size and alignment match
23396 that of some integer type.
23398 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
23399 switch is not binary compatible with code compiled with the
23400 @option{-freg-struct-return} switch.
23401 Use it to conform to a non-default application binary interface.
23403 @item -freg-struct-return
23404 @opindex freg-struct-return
23405 Return @code{struct} and @code{union} values in registers when possible.
23406 This is more efficient for small structures than
23407 @option{-fpcc-struct-return}.
23409 If you specify neither @option{-fpcc-struct-return} nor
23410 @option{-freg-struct-return}, GCC defaults to whichever convention is
23411 standard for the target. If there is no standard convention, GCC
23412 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
23413 the principal compiler. In those cases, we can choose the standard, and
23414 we chose the more efficient register return alternative.
23416 @strong{Warning:} code compiled with the @option{-freg-struct-return}
23417 switch is not binary compatible with code compiled with the
23418 @option{-fpcc-struct-return} switch.
23419 Use it to conform to a non-default application binary interface.
23421 @item -fshort-enums
23422 @opindex fshort-enums
23423 Allocate to an @code{enum} type only as many bytes as it needs for the
23424 declared range of possible values. Specifically, the @code{enum} type
23425 is equivalent to the smallest integer type that has enough room.
23427 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
23428 code that is not binary compatible with code generated without that switch.
23429 Use it to conform to a non-default application binary interface.
23431 @item -fshort-double
23432 @opindex fshort-double
23433 Use the same size for @code{double} as for @code{float}.
23435 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
23436 code that is not binary compatible with code generated without that switch.
23437 Use it to conform to a non-default application binary interface.
23439 @item -fshort-wchar
23440 @opindex fshort-wchar
23441 Override the underlying type for @code{wchar_t} to be @code{short
23442 unsigned int} instead of the default for the target. This option is
23443 useful for building programs to run under WINE@.
23445 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
23446 code that is not binary compatible with code generated without that switch.
23447 Use it to conform to a non-default application binary interface.
23450 @opindex fno-common
23451 In C code, controls the placement of uninitialized global variables.
23452 Unix C compilers have traditionally permitted multiple definitions of
23453 such variables in different compilation units by placing the variables
23455 This is the behavior specified by @option{-fcommon}, and is the default
23456 for GCC on most targets.
23457 On the other hand, this behavior is not required by ISO C, and on some
23458 targets may carry a speed or code size penalty on variable references.
23459 The @option{-fno-common} option specifies that the compiler should place
23460 uninitialized global variables in the data section of the object file,
23461 rather than generating them as common blocks.
23462 This has the effect that if the same variable is declared
23463 (without @code{extern}) in two different compilations,
23464 you get a multiple-definition error when you link them.
23465 In this case, you must compile with @option{-fcommon} instead.
23466 Compiling with @option{-fno-common} is useful on targets for which
23467 it provides better performance, or if you wish to verify that the
23468 program will work on other systems that always treat uninitialized
23469 variable declarations this way.
23473 Ignore the @code{#ident} directive.
23475 @item -finhibit-size-directive
23476 @opindex finhibit-size-directive
23477 Don't output a @code{.size} assembler directive, or anything else that
23478 would cause trouble if the function is split in the middle, and the
23479 two halves are placed at locations far apart in memory. This option is
23480 used when compiling @file{crtstuff.c}; you should not need to use it
23483 @item -fverbose-asm
23484 @opindex fverbose-asm
23485 Put extra commentary information in the generated assembly code to
23486 make it more readable. This option is generally only of use to those
23487 who actually need to read the generated assembly code (perhaps while
23488 debugging the compiler itself).
23490 @option{-fno-verbose-asm}, the default, causes the
23491 extra information to be omitted and is useful when comparing two assembler
23494 @item -frecord-gcc-switches
23495 @opindex frecord-gcc-switches
23496 This switch causes the command line used to invoke the
23497 compiler to be recorded into the object file that is being created.
23498 This switch is only implemented on some targets and the exact format
23499 of the recording is target and binary file format dependent, but it
23500 usually takes the form of a section containing ASCII text. This
23501 switch is related to the @option{-fverbose-asm} switch, but that
23502 switch only records information in the assembler output file as
23503 comments, so it never reaches the object file.
23504 See also @option{-grecord-gcc-switches} for another
23505 way of storing compiler options into the object file.
23509 @cindex global offset table
23511 Generate position-independent code (PIC) suitable for use in a shared
23512 library, if supported for the target machine. Such code accesses all
23513 constant addresses through a global offset table (GOT)@. The dynamic
23514 loader resolves the GOT entries when the program starts (the dynamic
23515 loader is not part of GCC; it is part of the operating system). If
23516 the GOT size for the linked executable exceeds a machine-specific
23517 maximum size, you get an error message from the linker indicating that
23518 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
23519 instead. (These maximums are 8k on the SPARC and 32k
23520 on the m68k and RS/6000. The x86 has no such limit.)
23522 Position-independent code requires special support, and therefore works
23523 only on certain machines. For the x86, GCC supports PIC for System V
23524 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
23525 position-independent.
23527 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
23532 If supported for the target machine, emit position-independent code,
23533 suitable for dynamic linking and avoiding any limit on the size of the
23534 global offset table. This option makes a difference on the m68k,
23535 PowerPC and SPARC@.
23537 Position-independent code requires special support, and therefore works
23538 only on certain machines.
23540 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
23547 These options are similar to @option{-fpic} and @option{-fPIC}, but
23548 generated position independent code can be only linked into executables.
23549 Usually these options are used when @option{-pie} GCC option is
23550 used during linking.
23552 @option{-fpie} and @option{-fPIE} both define the macros
23553 @code{__pie__} and @code{__PIE__}. The macros have the value 1
23554 for @option{-fpie} and 2 for @option{-fPIE}.
23556 @item -fno-jump-tables
23557 @opindex fno-jump-tables
23558 Do not use jump tables for switch statements even where it would be
23559 more efficient than other code generation strategies. This option is
23560 of use in conjunction with @option{-fpic} or @option{-fPIC} for
23561 building code that forms part of a dynamic linker and cannot
23562 reference the address of a jump table. On some targets, jump tables
23563 do not require a GOT and this option is not needed.
23565 @item -ffixed-@var{reg}
23567 Treat the register named @var{reg} as a fixed register; generated code
23568 should never refer to it (except perhaps as a stack pointer, frame
23569 pointer or in some other fixed role).
23571 @var{reg} must be the name of a register. The register names accepted
23572 are machine-specific and are defined in the @code{REGISTER_NAMES}
23573 macro in the machine description macro file.
23575 This flag does not have a negative form, because it specifies a
23578 @item -fcall-used-@var{reg}
23579 @opindex fcall-used
23580 Treat the register named @var{reg} as an allocable register that is
23581 clobbered by function calls. It may be allocated for temporaries or
23582 variables that do not live across a call. Functions compiled this way
23583 do not save and restore the register @var{reg}.
23585 It is an error to use this flag with the frame pointer or stack pointer.
23586 Use of this flag for other registers that have fixed pervasive roles in
23587 the machine's execution model produces disastrous results.
23589 This flag does not have a negative form, because it specifies a
23592 @item -fcall-saved-@var{reg}
23593 @opindex fcall-saved
23594 Treat the register named @var{reg} as an allocable register saved by
23595 functions. It may be allocated even for temporaries or variables that
23596 live across a call. Functions compiled this way save and restore
23597 the register @var{reg} if they use it.
23599 It is an error to use this flag with the frame pointer or stack pointer.
23600 Use of this flag for other registers that have fixed pervasive roles in
23601 the machine's execution model produces disastrous results.
23603 A different sort of disaster results from the use of this flag for
23604 a register in which function values may be returned.
23606 This flag does not have a negative form, because it specifies a
23609 @item -fpack-struct[=@var{n}]
23610 @opindex fpack-struct
23611 Without a value specified, pack all structure members together without
23612 holes. When a value is specified (which must be a small power of two), pack
23613 structure members according to this value, representing the maximum
23614 alignment (that is, objects with default alignment requirements larger than
23615 this are output potentially unaligned at the next fitting location.
23617 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
23618 code that is not binary compatible with code generated without that switch.
23619 Additionally, it makes the code suboptimal.
23620 Use it to conform to a non-default application binary interface.
23622 @item -finstrument-functions
23623 @opindex finstrument-functions
23624 Generate instrumentation calls for entry and exit to functions. Just
23625 after function entry and just before function exit, the following
23626 profiling functions are called with the address of the current
23627 function and its call site. (On some platforms,
23628 @code{__builtin_return_address} does not work beyond the current
23629 function, so the call site information may not be available to the
23630 profiling functions otherwise.)
23633 void __cyg_profile_func_enter (void *this_fn,
23635 void __cyg_profile_func_exit (void *this_fn,
23639 The first argument is the address of the start of the current function,
23640 which may be looked up exactly in the symbol table.
23642 This instrumentation is also done for functions expanded inline in other
23643 functions. The profiling calls indicate where, conceptually, the
23644 inline function is entered and exited. This means that addressable
23645 versions of such functions must be available. If all your uses of a
23646 function are expanded inline, this may mean an additional expansion of
23647 code size. If you use @code{extern inline} in your C code, an
23648 addressable version of such functions must be provided. (This is
23649 normally the case anyway, but if you get lucky and the optimizer always
23650 expands the functions inline, you might have gotten away without
23651 providing static copies.)
23653 A function may be given the attribute @code{no_instrument_function}, in
23654 which case this instrumentation is not done. This can be used, for
23655 example, for the profiling functions listed above, high-priority
23656 interrupt routines, and any functions from which the profiling functions
23657 cannot safely be called (perhaps signal handlers, if the profiling
23658 routines generate output or allocate memory).
23660 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
23661 @opindex finstrument-functions-exclude-file-list
23663 Set the list of functions that are excluded from instrumentation (see
23664 the description of @option{-finstrument-functions}). If the file that
23665 contains a function definition matches with one of @var{file}, then
23666 that function is not instrumented. The match is done on substrings:
23667 if the @var{file} parameter is a substring of the file name, it is
23668 considered to be a match.
23673 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
23677 excludes any inline function defined in files whose pathnames
23678 contain @file{/bits/stl} or @file{include/sys}.
23680 If, for some reason, you want to include letter @samp{,} in one of
23681 @var{sym}, write @samp{\,}. For example,
23682 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
23683 (note the single quote surrounding the option).
23685 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
23686 @opindex finstrument-functions-exclude-function-list
23688 This is similar to @option{-finstrument-functions-exclude-file-list},
23689 but this option sets the list of function names to be excluded from
23690 instrumentation. The function name to be matched is its user-visible
23691 name, such as @code{vector<int> blah(const vector<int> &)}, not the
23692 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
23693 match is done on substrings: if the @var{sym} parameter is a substring
23694 of the function name, it is considered to be a match. For C99 and C++
23695 extended identifiers, the function name must be given in UTF-8, not
23696 using universal character names.
23698 @item -fstack-check
23699 @opindex fstack-check
23700 Generate code to verify that you do not go beyond the boundary of the
23701 stack. You should specify this flag if you are running in an
23702 environment with multiple threads, but you only rarely need to specify it in
23703 a single-threaded environment since stack overflow is automatically
23704 detected on nearly all systems if there is only one stack.
23706 Note that this switch does not actually cause checking to be done; the
23707 operating system or the language runtime must do that. The switch causes
23708 generation of code to ensure that they see the stack being extended.
23710 You can additionally specify a string parameter: @samp{no} means no
23711 checking, @samp{generic} means force the use of old-style checking,
23712 @samp{specific} means use the best checking method and is equivalent
23713 to bare @option{-fstack-check}.
23715 Old-style checking is a generic mechanism that requires no specific
23716 target support in the compiler but comes with the following drawbacks:
23720 Modified allocation strategy for large objects: they are always
23721 allocated dynamically if their size exceeds a fixed threshold.
23724 Fixed limit on the size of the static frame of functions: when it is
23725 topped by a particular function, stack checking is not reliable and
23726 a warning is issued by the compiler.
23729 Inefficiency: because of both the modified allocation strategy and the
23730 generic implementation, code performance is hampered.
23733 Note that old-style stack checking is also the fallback method for
23734 @samp{specific} if no target support has been added in the compiler.
23736 @item -fstack-limit-register=@var{reg}
23737 @itemx -fstack-limit-symbol=@var{sym}
23738 @itemx -fno-stack-limit
23739 @opindex fstack-limit-register
23740 @opindex fstack-limit-symbol
23741 @opindex fno-stack-limit
23742 Generate code to ensure that the stack does not grow beyond a certain value,
23743 either the value of a register or the address of a symbol. If a larger
23744 stack is required, a signal is raised at run time. For most targets,
23745 the signal is raised before the stack overruns the boundary, so
23746 it is possible to catch the signal without taking special precautions.
23748 For instance, if the stack starts at absolute address @samp{0x80000000}
23749 and grows downwards, you can use the flags
23750 @option{-fstack-limit-symbol=__stack_limit} and
23751 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
23752 of 128KB@. Note that this may only work with the GNU linker.
23754 @item -fsplit-stack
23755 @opindex fsplit-stack
23756 Generate code to automatically split the stack before it overflows.
23757 The resulting program has a discontiguous stack which can only
23758 overflow if the program is unable to allocate any more memory. This
23759 is most useful when running threaded programs, as it is no longer
23760 necessary to calculate a good stack size to use for each thread. This
23761 is currently only implemented for the x86 targets running
23764 When code compiled with @option{-fsplit-stack} calls code compiled
23765 without @option{-fsplit-stack}, there may not be much stack space
23766 available for the latter code to run. If compiling all code,
23767 including library code, with @option{-fsplit-stack} is not an option,
23768 then the linker can fix up these calls so that the code compiled
23769 without @option{-fsplit-stack} always has a large stack. Support for
23770 this is implemented in the gold linker in GNU binutils release 2.21
23773 @item -fleading-underscore
23774 @opindex fleading-underscore
23775 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
23776 change the way C symbols are represented in the object file. One use
23777 is to help link with legacy assembly code.
23779 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
23780 generate code that is not binary compatible with code generated without that
23781 switch. Use it to conform to a non-default application binary interface.
23782 Not all targets provide complete support for this switch.
23784 @item -ftls-model=@var{model}
23785 @opindex ftls-model
23786 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
23787 The @var{model} argument should be one of @samp{global-dynamic},
23788 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
23789 Note that the choice is subject to optimization: the compiler may use
23790 a more efficient model for symbols not visible outside of the translation
23791 unit, or if @option{-fpic} is not given on the command line.
23793 The default without @option{-fpic} is @samp{initial-exec}; with
23794 @option{-fpic} the default is @samp{global-dynamic}.
23796 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
23797 @opindex fvisibility
23798 Set the default ELF image symbol visibility to the specified option---all
23799 symbols are marked with this unless overridden within the code.
23800 Using this feature can very substantially improve linking and
23801 load times of shared object libraries, produce more optimized
23802 code, provide near-perfect API export and prevent symbol clashes.
23803 It is @strong{strongly} recommended that you use this in any shared objects
23806 Despite the nomenclature, @samp{default} always means public; i.e.,
23807 available to be linked against from outside the shared object.
23808 @samp{protected} and @samp{internal} are pretty useless in real-world
23809 usage so the only other commonly used option is @samp{hidden}.
23810 The default if @option{-fvisibility} isn't specified is
23811 @samp{default}, i.e., make every symbol public.
23813 A good explanation of the benefits offered by ensuring ELF
23814 symbols have the correct visibility is given by ``How To Write
23815 Shared Libraries'' by Ulrich Drepper (which can be found at
23816 @w{@uref{http://www.akkadia.org/drepper/}})---however a superior
23817 solution made possible by this option to marking things hidden when
23818 the default is public is to make the default hidden and mark things
23819 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
23820 and @code{__attribute__ ((visibility("default")))} instead of
23821 @code{__declspec(dllexport)} you get almost identical semantics with
23822 identical syntax. This is a great boon to those working with
23823 cross-platform projects.
23825 For those adding visibility support to existing code, you may find
23826 @code{#pragma GCC visibility} of use. This works by you enclosing
23827 the declarations you wish to set visibility for with (for example)
23828 @code{#pragma GCC visibility push(hidden)} and
23829 @code{#pragma GCC visibility pop}.
23830 Bear in mind that symbol visibility should be viewed @strong{as
23831 part of the API interface contract} and thus all new code should
23832 always specify visibility when it is not the default; i.e., declarations
23833 only for use within the local DSO should @strong{always} be marked explicitly
23834 as hidden as so to avoid PLT indirection overheads---making this
23835 abundantly clear also aids readability and self-documentation of the code.
23836 Note that due to ISO C++ specification requirements, @code{operator new} and
23837 @code{operator delete} must always be of default visibility.
23839 Be aware that headers from outside your project, in particular system
23840 headers and headers from any other library you use, may not be
23841 expecting to be compiled with visibility other than the default. You
23842 may need to explicitly say @code{#pragma GCC visibility push(default)}
23843 before including any such headers.
23845 @code{extern} declarations are not affected by @option{-fvisibility}, so
23846 a lot of code can be recompiled with @option{-fvisibility=hidden} with
23847 no modifications. However, this means that calls to @code{extern}
23848 functions with no explicit visibility use the PLT, so it is more
23849 effective to use @code{__attribute ((visibility))} and/or
23850 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
23851 declarations should be treated as hidden.
23853 Note that @option{-fvisibility} does affect C++ vague linkage
23854 entities. This means that, for instance, an exception class that is
23855 be thrown between DSOs must be explicitly marked with default
23856 visibility so that the @samp{type_info} nodes are unified between
23859 An overview of these techniques, their benefits and how to use them
23860 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
23862 @item -fstrict-volatile-bitfields
23863 @opindex fstrict-volatile-bitfields
23864 This option should be used if accesses to volatile bit-fields (or other
23865 structure fields, although the compiler usually honors those types
23866 anyway) should use a single access of the width of the
23867 field's type, aligned to a natural alignment if possible. For
23868 example, targets with memory-mapped peripheral registers might require
23869 all such accesses to be 16 bits wide; with this flag you can
23870 declare all peripheral bit-fields as @code{unsigned short} (assuming short
23871 is 16 bits on these targets) to force GCC to use 16-bit accesses
23872 instead of, perhaps, a more efficient 32-bit access.
23874 If this option is disabled, the compiler uses the most efficient
23875 instruction. In the previous example, that might be a 32-bit load
23876 instruction, even though that accesses bytes that do not contain
23877 any portion of the bit-field, or memory-mapped registers unrelated to
23878 the one being updated.
23880 In some cases, such as when the @code{packed} attribute is applied to a
23881 structure field, it may not be possible to access the field with a single
23882 read or write that is correctly aligned for the target machine. In this
23883 case GCC falls back to generating multiple accesses rather than code that
23884 will fault or truncate the result at run time.
23886 Note: Due to restrictions of the C/C++11 memory model, write accesses are
23887 not allowed to touch non bit-field members. It is therefore recommended
23888 to define all bits of the field's type as bit-field members.
23890 The default value of this option is determined by the application binary
23891 interface for the target processor.
23893 @item -fsync-libcalls
23894 @opindex fsync-libcalls
23895 This option controls whether any out-of-line instance of the @code{__sync}
23896 family of functions may be used to implement the C++11 @code{__atomic}
23897 family of functions.
23899 The default value of this option is enabled, thus the only useful form
23900 of the option is @option{-fno-sync-libcalls}. This option is used in
23901 the implementation of the @file{libatomic} runtime library.
23907 @node Environment Variables
23908 @section Environment Variables Affecting GCC
23909 @cindex environment variables
23911 @c man begin ENVIRONMENT
23912 This section describes several environment variables that affect how GCC
23913 operates. Some of them work by specifying directories or prefixes to use
23914 when searching for various kinds of files. Some are used to specify other
23915 aspects of the compilation environment.
23917 Note that you can also specify places to search using options such as
23918 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
23919 take precedence over places specified using environment variables, which
23920 in turn take precedence over those specified by the configuration of GCC@.
23921 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
23922 GNU Compiler Collection (GCC) Internals}.
23927 @c @itemx LC_COLLATE
23929 @c @itemx LC_MONETARY
23930 @c @itemx LC_NUMERIC
23935 @c @findex LC_COLLATE
23936 @findex LC_MESSAGES
23937 @c @findex LC_MONETARY
23938 @c @findex LC_NUMERIC
23942 These environment variables control the way that GCC uses
23943 localization information which allows GCC to work with different
23944 national conventions. GCC inspects the locale categories
23945 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
23946 so. These locale categories can be set to any value supported by your
23947 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
23948 Kingdom encoded in UTF-8.
23950 The @env{LC_CTYPE} environment variable specifies character
23951 classification. GCC uses it to determine the character boundaries in
23952 a string; this is needed for some multibyte encodings that contain quote
23953 and escape characters that are otherwise interpreted as a string
23956 The @env{LC_MESSAGES} environment variable specifies the language to
23957 use in diagnostic messages.
23959 If the @env{LC_ALL} environment variable is set, it overrides the value
23960 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
23961 and @env{LC_MESSAGES} default to the value of the @env{LANG}
23962 environment variable. If none of these variables are set, GCC
23963 defaults to traditional C English behavior.
23967 If @env{TMPDIR} is set, it specifies the directory to use for temporary
23968 files. GCC uses temporary files to hold the output of one stage of
23969 compilation which is to be used as input to the next stage: for example,
23970 the output of the preprocessor, which is the input to the compiler
23973 @item GCC_COMPARE_DEBUG
23974 @findex GCC_COMPARE_DEBUG
23975 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
23976 @option{-fcompare-debug} to the compiler driver. See the documentation
23977 of this option for more details.
23979 @item GCC_EXEC_PREFIX
23980 @findex GCC_EXEC_PREFIX
23981 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
23982 names of the subprograms executed by the compiler. No slash is added
23983 when this prefix is combined with the name of a subprogram, but you can
23984 specify a prefix that ends with a slash if you wish.
23986 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
23987 an appropriate prefix to use based on the pathname it is invoked with.
23989 If GCC cannot find the subprogram using the specified prefix, it
23990 tries looking in the usual places for the subprogram.
23992 The default value of @env{GCC_EXEC_PREFIX} is
23993 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
23994 the installed compiler. In many cases @var{prefix} is the value
23995 of @code{prefix} when you ran the @file{configure} script.
23997 Other prefixes specified with @option{-B} take precedence over this prefix.
23999 This prefix is also used for finding files such as @file{crt0.o} that are
24002 In addition, the prefix is used in an unusual way in finding the
24003 directories to search for header files. For each of the standard
24004 directories whose name normally begins with @samp{/usr/local/lib/gcc}
24005 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
24006 replacing that beginning with the specified prefix to produce an
24007 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
24008 @file{foo/bar} just before it searches the standard directory
24009 @file{/usr/local/lib/bar}.
24010 If a standard directory begins with the configured
24011 @var{prefix} then the value of @var{prefix} is replaced by
24012 @env{GCC_EXEC_PREFIX} when looking for header files.
24014 @item COMPILER_PATH
24015 @findex COMPILER_PATH
24016 The value of @env{COMPILER_PATH} is a colon-separated list of
24017 directories, much like @env{PATH}. GCC tries the directories thus
24018 specified when searching for subprograms, if it can't find the
24019 subprograms using @env{GCC_EXEC_PREFIX}.
24022 @findex LIBRARY_PATH
24023 The value of @env{LIBRARY_PATH} is a colon-separated list of
24024 directories, much like @env{PATH}. When configured as a native compiler,
24025 GCC tries the directories thus specified when searching for special
24026 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
24027 using GCC also uses these directories when searching for ordinary
24028 libraries for the @option{-l} option (but directories specified with
24029 @option{-L} come first).
24033 @cindex locale definition
24034 This variable is used to pass locale information to the compiler. One way in
24035 which this information is used is to determine the character set to be used
24036 when character literals, string literals and comments are parsed in C and C++.
24037 When the compiler is configured to allow multibyte characters,
24038 the following values for @env{LANG} are recognized:
24042 Recognize JIS characters.
24044 Recognize SJIS characters.
24046 Recognize EUCJP characters.
24049 If @env{LANG} is not defined, or if it has some other value, then the
24050 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
24051 recognize and translate multibyte characters.
24055 Some additional environment variables affect the behavior of the
24058 @include cppenv.texi
24062 @node Precompiled Headers
24063 @section Using Precompiled Headers
24064 @cindex precompiled headers
24065 @cindex speed of compilation
24067 Often large projects have many header files that are included in every
24068 source file. The time the compiler takes to process these header files
24069 over and over again can account for nearly all of the time required to
24070 build the project. To make builds faster, GCC allows you to
24071 @dfn{precompile} a header file.
24073 To create a precompiled header file, simply compile it as you would any
24074 other file, if necessary using the @option{-x} option to make the driver
24075 treat it as a C or C++ header file. You may want to use a
24076 tool like @command{make} to keep the precompiled header up-to-date when
24077 the headers it contains change.
24079 A precompiled header file is searched for when @code{#include} is
24080 seen in the compilation. As it searches for the included file
24081 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
24082 compiler looks for a precompiled header in each directory just before it
24083 looks for the include file in that directory. The name searched for is
24084 the name specified in the @code{#include} with @samp{.gch} appended. If
24085 the precompiled header file can't be used, it is ignored.
24087 For instance, if you have @code{#include "all.h"}, and you have
24088 @file{all.h.gch} in the same directory as @file{all.h}, then the
24089 precompiled header file is used if possible, and the original
24090 header is used otherwise.
24092 Alternatively, you might decide to put the precompiled header file in a
24093 directory and use @option{-I} to ensure that directory is searched
24094 before (or instead of) the directory containing the original header.
24095 Then, if you want to check that the precompiled header file is always
24096 used, you can put a file of the same name as the original header in this
24097 directory containing an @code{#error} command.
24099 This also works with @option{-include}. So yet another way to use
24100 precompiled headers, good for projects not designed with precompiled
24101 header files in mind, is to simply take most of the header files used by
24102 a project, include them from another header file, precompile that header
24103 file, and @option{-include} the precompiled header. If the header files
24104 have guards against multiple inclusion, they are skipped because
24105 they've already been included (in the precompiled header).
24107 If you need to precompile the same header file for different
24108 languages, targets, or compiler options, you can instead make a
24109 @emph{directory} named like @file{all.h.gch}, and put each precompiled
24110 header in the directory, perhaps using @option{-o}. It doesn't matter
24111 what you call the files in the directory; every precompiled header in
24112 the directory is considered. The first precompiled header
24113 encountered in the directory that is valid for this compilation is
24114 used; they're searched in no particular order.
24116 There are many other possibilities, limited only by your imagination,
24117 good sense, and the constraints of your build system.
24119 A precompiled header file can be used only when these conditions apply:
24123 Only one precompiled header can be used in a particular compilation.
24126 A precompiled header can't be used once the first C token is seen. You
24127 can have preprocessor directives before a precompiled header; you cannot
24128 include a precompiled header from inside another header.
24131 The precompiled header file must be produced for the same language as
24132 the current compilation. You can't use a C precompiled header for a C++
24136 The precompiled header file must have been produced by the same compiler
24137 binary as the current compilation is using.
24140 Any macros defined before the precompiled header is included must
24141 either be defined in the same way as when the precompiled header was
24142 generated, or must not affect the precompiled header, which usually
24143 means that they don't appear in the precompiled header at all.
24145 The @option{-D} option is one way to define a macro before a
24146 precompiled header is included; using a @code{#define} can also do it.
24147 There are also some options that define macros implicitly, like
24148 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
24151 @item If debugging information is output when using the precompiled
24152 header, using @option{-g} or similar, the same kind of debugging information
24153 must have been output when building the precompiled header. However,
24154 a precompiled header built using @option{-g} can be used in a compilation
24155 when no debugging information is being output.
24157 @item The same @option{-m} options must generally be used when building
24158 and using the precompiled header. @xref{Submodel Options},
24159 for any cases where this rule is relaxed.
24161 @item Each of the following options must be the same when building and using
24162 the precompiled header:
24164 @gccoptlist{-fexceptions}
24167 Some other command-line options starting with @option{-f},
24168 @option{-p}, or @option{-O} must be defined in the same way as when
24169 the precompiled header was generated. At present, it's not clear
24170 which options are safe to change and which are not; the safest choice
24171 is to use exactly the same options when generating and using the
24172 precompiled header. The following are known to be safe:
24174 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
24175 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
24176 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
24181 For all of these except the last, the compiler automatically
24182 ignores the precompiled header if the conditions aren't met. If you
24183 find an option combination that doesn't work and doesn't cause the
24184 precompiled header to be ignored, please consider filing a bug report,
24187 If you do use differing options when generating and using the
24188 precompiled header, the actual behavior is a mixture of the
24189 behavior for the options. For instance, if you use @option{-g} to
24190 generate the precompiled header but not when using it, you may or may
24191 not get debugging information for routines in the precompiled header.